Leading Edge
In This Issue An EnERgy Boost for Cancer PAGE 711
Rapidly growing cancer cells increase their rate of aerobic glycolysis in a metabolic shift known as the Warburg effect. Their proliferation also demands high protein folding capacity in the endoplasmic reticulum (ER). Fang et al. identify an ER-localized enzyme, ENTPD5, that is responsible for both of these features of tumor cells. Inhibition of ENTPD5, which is commonly upregulated in human cancers, blocked tumor growth in mice. Thus, ENTPD5 inhibition could potentially become an anticancer therapy.
A Nudge and a Kick for Histone Replacement PAGE 725
Most promoters in eukaryotes are marked with nucleosomes carrying a special histone H2A.Z, which is important for gene regulation. SWR1 incorporates H2A.Z into nucleosomes in a histone replacement reaction. Luk et al. now report a mechanism that ensures that only nucleosomes containing the canonical histone H2A are targeted for replacement. SWR1’s ATPase activity is sequentially stimulated by H2A-containing nucleosomes and free H2A.Z-H2B dimers, leading to eviction of nucleosomal H2A-H2B and deposition of H2A.Z-H2B. These stepwise events ensure the specificity of the nucleosome replacement reaction.
Locking Chromosome Cohesion during Replication PAGE 737
In eukaryotic cells, sister chromatids remain physically connected from the time of their synthesis during DNA replication until their separation during mitosis. Sister chromatid cohesion depends on the stable association of cohesin with DNA. Nishiyama et al. now show that Sororin binds cohesin during replication and stabilizes the cohesinDNA complex by displacing the cohesin ‘‘unloading’’ protein Wapl. Distant orthologs of Sororin exist in many species, implying that this may be a widespread mechanism for the maintenance of sister chromatid cohesion.
G Protein Lockdown for Channels PAGE 750
G protein-coupled potassium channels need to be turned off quickly, on a timescale faster than that afforded by either ligand clearance or receptor endocytosis. Raveh et al. now show that the GPCR kinase, GRK2, achieves rapid desensitization of the GIRK potassium channel by sequestering the G protein subunits required for GIRK activity. This kinase-independent function of GRK2 thus allows rapid control of ligand-stimulated channel function.
Actin Cherry Picks Recycling Receptors PAGE 761
Signaling receptors recycle efficiently during endocytosis in a manner that differs from bulk membrane recycling. Puthenveedu et al. use live cell imaging to show that distinct endosomal subdomains mediate active recycling of signaling receptors. The actin cytoskeleton binds in a sequence-dependent manner to the receptors, further concentrating and stabilizing these domains for recycling. Cell 143, November 24, 2010 ª2010 Elsevier Inc. 653
Shapewear for the ER PAGE 774
The endoplasmic reticulum (ER) consists of the nuclear envelope and an extensive peripheral network of tubules and membrane sheets. Shibata et al. demonstrate that ER sheets are formed through stabilization of their highly curved edges by the reticulon/DP1/Yop1 p proteins. The membrane protein Climp63 further shapes the sheets, acting as a spacer to regulate their area and luminal width.
HIV Pushes the T Cell Self-Destruct Button PAGE 789
The depletion of CD4 T cells during HIV infection is a hallmark of AIDS. Doitsh et al. show that abortive infection of CD4 T cells elicits cell death. Incomplete reverse transcripts of the virus accumulate in these cells and activate suicidal innate antiviral and inflammatory responses. Thus, T cell death is not triggered by new virus production but, rather, by a suicide mechanism, which likely evolved to protect the host but in fact contributes to immunodeficiency.
Hungry but Still Hearing PAGE 802
Caloric restriction (CR) extends the life span of many species and slows the progression of age-related hearing loss (AHL). Here, Someya et al. report that mitochondrial Sirt3 mediates the prevention of AHL and reduces oxidative damage in calorie-restricted mice. In response to CR, Sirt3 deacetylates and activates isocitrate dehydrogenase 2, leading to an enhanced glutathione antioxidant defense system in mitochondria. These results suggest that Sirt3-dependent mitochondrial adaptations may be a central mechanism to delay aging in mammals.
Outfoxing Aging PAGE 813
Loss of muscle strength is one of the most obvious changes that we experience as we age, but how this connects with systemic aging is unclear. Demontis and Perrimon report that accumulation of protein aggregates in aging Drosophila muscle is reduced by FOXO/4E-BP signaling, delaying muscle senescence. This pathway in muscle prevents overall aging and protein aggregation in other tissues. These results provide a framework to understand the coordination of tissue and organismal aging.
Golgi Decides, Axon or Dendrite PAGE 826
Neuronal cells polarize to develop an axon at one pole and dendrites at the other. Matsuki et al. identify two signaling pathways that influence Golgi morphogenesis to regulate this polarization. The Stk25 kinase acts through the Golgi protein GM130 to promote a condensed Golgi morphology and axon development. The Reelin-Dab1 signaling pathway, previously known to regulate other aspects of nervous system development, antagonizes the Stk25 pathway to promote Golgi extension and dendrite development. Thus, Golgi distribution is a central factor in neuronal development.
Structural Fingerprints of the Human Genome PAGE 837
Genomic structural variation—insertions, duplications, and deletions—are important contributors to human disease and genetic diversity. The precise molecular characteristics of these variants have been difficult to ascertain by standard highthroughput genome sequencing. Kidd et al. now report a resource of fosmid clones obtained from the genomes of 17 individuals. The authors characterize the breakpoints of more than a thousand structural variants, allowing inference of the molecular pathways that generated them and offering an in-depth view of the characteristics of human genomic variation. Cell 143, November 24, 2010 ª2010 Elsevier Inc. 655
Leading Edge
Select: Cell Cycle The phases of the cell cycle must be exquisitely timed and tightly regulated in order to ensure proper chromosome replication and segregation and cell division. New findings described in this issue’s Select address key regulatory events in the cell cycle and reveal potential clinical outcomes of errors in these processes.
An Epigenetic License to Replicate Chromosome replication needs to occur once and only once during the cell cycle to produce daughter cells with accurate genetic content. Licensing of replication origins is one form of DNA synthesis regulation, in which origins are loaded with pre-replication complex (RC) proteins during the end of M phase and throughout G1. Without this licensing event, replication origins cannot be activated. New findings from Tardat et al. identify the methyltransferase PR-Set7—and the histone modification that it catalyzes, methylation of histone H4 lysine 20 (H4K20me1)—as a key regulator of the onset of licensing in mammalian cells. The authors show that PR-Set7 and H4K20me1 levels are cell cycle regulated—both are high during M and G1 phases, dropping in S when synthesis begins—and that proteasomal degradation of PR-Set7 is needed to prevent DNA re-replication. The authors also show that silencing PR-Set7 leads to Re-replicating G2 cells (cyclin B1, red; decreased chromatin loading of pre-RC proteins and reduced origin firing during EdU, green). Image courtesy of E. Julien. S phase, whereas targeting PR-Set7 to nonorigin sites on the chromatin is sufficient to induce H4K20me1 and the assembly of pre-RC proteins. Future studies are needed to investigate how H4K20me1 facilitates chromatin loading of pre-RC proteins. M. Tardat et al. (2010). Nat. Cell Biol. Published online October 17, 2010. 10.1038/ncb2113.
Getting a Toehold on Microtubules The ability of the kinetochore to maintain an attachment between chromosomes and microtubules is necessary for proper chromosomal segregation during anaphase. The Ndc80 complex is known to be a key regulatory site for microtubule attachment, but, given the highly dynamic process of microtubule assembly and disassembly occurring during segregation, it has been a challenge to identify how the Ndc80 complex physically holds on to such a rapidly changing structure. Alushin et al. address this using cryo-electron microscopy to better reveal the metazoan Ndc80 complex bound to microtubules. The authors find that the Ndc80 complex binds both a- and b-tubulin monomers and identify a ‘‘toe’’—a short section of the NDC80 protein that recognizes a site between two tubulin monomers, a hinge point for tubulin bending. The toe appears to prefer binding straight tubulin, suggesting that it could act as a sensor for tubulin conformation. At the same time, the N terminus of NDC80 allows high-affinity microtubule binding and appears to mediate self-assembly of Ndc80 complexes in a manner that is modulated via phosphorylation by Aurora B kinase. The authors propose a model in which phosphorylated Ndc80 complexes bind a microtubule and spindle forces then pull the bound complex out of the Aurora B kinase phosphorylation zone. The resulting dephosphorylation of NDC80 results in Two Ndc80 molecules (blue and yellow; high-affinity clusters forming in linear arrays along the microtubule. This N terminus, magenta) binding tubulin cluster arrangement is consistent with a biased diffusion model of kineto- (green; C terminus, red). Image courtesy of E. Nogales. chore attachment and movement. On a shrinking microtubule, the Ndc80microtubule interaction would be reduced due to conformational changes in tubulin at the disassembling or depolymerizing end, and the cluster would diffuse along the microtubules toward the pole, thereby moving the chromosome in that direction. G.M. Alushin et al. (2010). Nature 467, 805–810. Cell 143, November 24, 2010 ª2010 Elsevier Inc. 657
Mounting Tension in Lead-Up to Fateful Decision Asymmetric cell division, which generates daughter cells with different developmental fates, is often achieved through asymmetric positioning of the mitotic spindle. However, some dividing cells start out with a centered spindle that becomes displaced during anaphase. This progressive asymmetry has been postulated to arise from greater elongation of microtubules on one side of the spindle. New findings from Ou et al. suggest that nonmuscle myosin II might also play a role. The authors show that in the QR.a neuroblast of Caenorhabditis elegans, myosin II becomes asymmetrically distributed during anaphase, concentrating at the anterior side of the cleavage furrow. Consequently, the anterior membrane becomes less dynamic and shrinks inward, whereas the posterior membrane expands like a balloon, suggesting that cortical tension and contractile forces driven by myosin II could be a factor in developing asymmetry. The authors also used CALI (chromophore-assisted laser inactivation) to specifically inactivate myosin II at the anterior membrane and find that this increases the size of the anterior daughter cell and can alter its fate from apoptosis to differentiation into a neuron-like cell. Future work is needed to better understand the respective contributions of microtubule elongation, myosin polarization, and perhaps other unknown mechanisms to the regulation of asymmetric division and cell fate. G. Ou et al. (2010). Science. Published online September 30, 2010. 10.1126/science.1196112.
Spindle Position, a Neuronal Mover and Maker Human microcephaly is a neurodevelopmental disorder characterized by a small brain, fewer surface ridges, and reduced cortical neuron numbers. Two recent papers used linkage analysis and genome capture in affected families to identify WDR62 as a common cause of genetic microcephaly and characterized the WDR62 protein as a spindle pole protein expressed in mitotic neural precursors. After sequencing affected individuals to identify specific disease-causing mutations, Nicholas et al. expressed mutant WDR62 in HeLa cells and showed that the normal accumulation of the protein at the spindle poles during mitosis is disrupted. Given the phenotype of reduced neuron numbers and small brain seen in microcephaly, one possibility the authors suggest is that WDR62 could be involved in proper positioning of the mitotic spindle and cleavage furrow, such that mutant WDR62 results in insufficient symmetric divisions—needed to produce neural precursors—early in cortical development. In Photograph of human microceagreement, Yu et al. show that the brain of an affected individual has profound cortical phalic brain. Image courtesy of defects, with thin sparse cortical layers and aberrant repositioning of neurons to C. Walsh. subcortical regions, suggesting deficits in neurogenesis and migration. Further description of the specific role of WDR62 at the spindle will clarify how it is involved in cerebral development and aid in our understanding of the etiology of microcephaly. A.K. Nicholas et al. (2010). Nat. Genet. Published online October 3, 2010. 10.1038/ng.682. T.W. Yu et al. (2010). Nat. Genet. Published online October 3, 2010. 10.1038/ng.683. Rebecca Alvania
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 659
Leading Edge
Previews ER Sheets Get Roughed Up Charles Barlowe1,* 1Department of Biochemistry, Dartmouth Medical School, Hanover, NH 03755, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.11.011
The molecular machinery that shapes the endoplasmic reticulum’s (ER’s) membrane into ordered networks of ‘‘smooth’’ tubules and ‘‘rough’’ sheets is poorly defined. Shibata et al. (2010) now report that sheet-inducing proteins, such as Climp-63, are enriched in the ‘‘rough’’ ER by their association with membrane-bound ribosomes, whereas curvature-inducing proteins localize at highly bent edges of membrane sheets. The elaborate morphologies of the endoplasmic reticulum have fascinated cell biologists for years. Compartments of the endoplasmic reticulum (ER) membrane form the nuclear envelope and then extend throughout the cell periphery in an interconnected network of membrane tubules and flattened discs called cisternae. How do these ordered arrays of membranes form, and how are their structures connected to their cellular function? In this issue of Cell, Shibata and coworkers define a class of sheetinducing membrane proteins that are enriched in the ribosome-studded ‘‘rough’’ ER. These proteins cooperate with membrane curvature-stabilizing factors to govern the relative level of sheets and tubules of the ER, providing a molecular basis for the longstanding morphological descriptions of ‘‘rough’’ and ‘‘smooth’’ ER. ER morphologies vary greatly across different species and cell types. For example, highly active secretory cells, such as pancreatic exocrine cells and plasma B cells, are packed full of flattened cisternae of rough ER. Live cell imaging also reveals that ER membranes are highly dynamic networks, undergoing constant remodeling often in response to physiological conditions. Previous studies focusing on the smooth ER found that tubule formation depends on a class of integral membrane proteins belonging to the reticulon and DP1 families (Voeltz et al., 2006). Reticulon and DP1 proteins are highly enriched in tubular ER elements, and they contain transmembrane segments with a double hairpin structure that induces positive membrane curvature by inserting like
a wedge into ER membranes (Figure 1). Indeed, reconstitution of purified reticulon and DP1 proteins into synthetic liposomes (i.e., artificial vesicles with a lipid bilayer) was sufficient to generate membrane tubules with a high degree of curvature (Hu et al., 2008). Thus, intrinsic properties of the reticulon and DP1 proteins are sufficient to induce membrane tubulation. However, ER tubules also form branched, reticular morphologies. Generation of these net-like structures requires additional factors, specifically atlastin GTPases, which drive fusion of ER tubules into branched networks (Hu et al., 2009; Orso et al., 2009). Of interest, atlastin isoforms were detected in association with the reticulon proteins, suggesting that the formation of tubules and branching are coordinated processes (Hu et al., 2009). In contrast to our understanding of ER tubules, the molecular mechanisms underlying the formation of ER sheets have been elusive. Now, Shibata et al. (2010) uncover an unexpected connection between the sheet-inducing factor Climp-63 and the reticulon and DP1 proteins. Their discovery begins with a key observation regarding the translocon complex, a large multisubunit channel that transports, or ‘‘translocates,’’ nascent polypeptides across ER membrane into the interior of the ER. Shibata and colleagues observe that components of the translocon complex are not only highly enriched in ER sheets, but they also form a specialized subdomain within ER membranes. Moreover, when the authors treat cells with the antibiotic puromycin, which disassembles
groups of ribosomes bound to the ER membranes (i.e., polysomes), proteins of the translocon complex redistribute between ER sheets and tubules. This finding suggests that actively translating polysomes concentrate translocon complexes into sheet subdomains of the ER. To identify the structural components of these ER sheet domains, Shibata and colleagues then perform a proteomic analysis of rough ER membranes from pancreatic secretory cells. Indeed, the most abundant protein constituents in ER sheets are components of the translocon complex and Climp-63. Moreover, microarray experiments reveal that Climp-63 messenger RNA (mRNA) levels are among the most highly induced messages during proliferation of ER sheet structures during the differentiation of immature B cells into IgG secreting plasma cells. Climp-63 is an ER transmembrane protein that contains an extended coiled-coil domain in the interior of the ER (i.e., the ER lumen). Previous studies suggested that this coiled-coil domain contributes to ER morphology by forming a scaffold in the ER lumen (Klopfenstein et al., 2001). To test the functional role of Climp-63 in ER sheet formation, Shibata and colleagues then overexpress Climp-63 in cultured cells, which causes a dramatic proliferation of ER sheets. Moreover, the distance between the sheets is 50 nm, the standard distance between ER sheets in mammalian cells (Figure 1). In contrast, decreasing the expression of Climp-63 does not deplete cells of ER sheets, but instead, it causes a marked reduction in the distance between cisternal sheets. Further, these sheets are spread diffusely
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 665
Finally, sheets and tubules throughout the cytoplasm, are not the only morphologies a similar phenotype as the of ER membranes. For authors observe when they example, specialized structreat cells with puromycin. tural domains of the ER are Finally, Climp-63 and the involved in metabolism of reticulon protein Rtn4 have hydrophobic compounds, opposing effects on ER morformation of ER-mitochonphology. Increased expresdrial junctions, transport of sion of Rtn4 reduces the Ca2+, formation of lipid dropnumber of ER sheets, lets, and protein export from whereas co-overexpression ER subdomains called transiwith Climp-63 restores sheet tional ER sites. The molecular structures in these cells. machinery that generates Importantly, reticulon prothese ER structures awaits teins strikingly localize to the elucidation. Although the highly curved edges of ER components that sculpt ER sheets, and this occurs sheets and tubules might when reticulon genes are Figure 1. Molecular Model for the Generation of ER Membrane also contribute to the morexpressed at endogenous Sheets and Tubules phology of these other struclevels or when both Climp-63 Cross-section of an endoplasmic reticulum (ER) cisterna showing the curvatures, it seems likely that and reticulon genes were ture-inducing proteins reticulons and DP1 (purple) enriched in highly bent novel mechanisms will also overexpressed together. membrane tubules and edges of the sheet. In contrast, the sheet-inducing protein Climp-63 (blue) is excluded from tubules and, instead, partitions into be discovered. The authors then propose sheet domains with translocon complexes. Climp-63 could assemble into the most basic mechanism parallel coiled-coil arrangements to flatten membranes and to serve as luminal for sheet formation that is ER spacers that keep individual sheets a specific distance apart (50 nm in REFERENCES mammalian cells). also consistent with their findings. In this model, reticulons Hu, J., Shibata, Y., Voss, C., Sheand DP1 proteins partition mesh, T., Li, Z., Coughlin, M., into the edges of sheets, where they 1998), suggesting an additional level of Kozlov, M.M., Rapoport, T.A., and Prinz, W.A. induce a high degree of curvature at the ER organization that is connected to the (2008). Science 319, 1247–1250. edges of closely apposed membrane bila- cell’s overall structure. Hu, J., Shibata, Y., Zhu, P.-P., Voss, C., Rismanchi, Although reticulon and DP1 proteins N., Prinz, W.A., Rapoport, T.A., and Blackstone, C. yers (Figure 1). However, assembling the ordered array of rough ER membranes in partition into sheet edges in vivo and ex- (2009). Cell 138, 549–561. active secretory cells also depends on pressing Climp-63 drives ER sheet prolif- Klopfenstein, D.R., Kappeler, F., and Hauri, H.P. the coiled-coil domain of Climp-63, which eration, it is still unknown whether these (1998). EMBO J. 17, 6168–6177. serves as a spacer between the sheets factors are sufficient for sheet formation Klopfenstein, D.R., Klumperman, J., Lustig, A., in the ER lumen (Figure 1). Lastly, the or whether other factors contribute to Kammerer, R.A., Oorschot, V., and Hauri, H.P. authors propose that Climp-63, together this process. A minimally reconstituted (2001). J. Cell Biol. 153, 1287–1300. with translocon complexes, partition into liposome system successfully demon- Nikonov, A.V., Hauri, H.P., Lauring, B., and Kreisheet domains with membrane-bound strated that reticulon and DP1 proteins bich, G. (2007). J. Cell Sci. 120, 2248–2258. drive tubule formation in vitro (Hu et al., polysomes to generate the rough ER. Orso, G., Pendin, D., Liu, S., Tosetto, J., Moss, This model proposed by Shibata and 2008). This system should provide a T.J., Faust, J.E., Micaroni, M., Egorova, A., Marticolleagues is also supported by previous powerful tool for determining whether nuzzi, A., McNew, J.A., and Daga, A. (2009). Nature studies showing that the coiled-coil adding purified Climp-63 is sufficient 460, 978–983. domain of Climp-63 assembles into for sheet formation. Furthermore, varying Shibata, Y., Shemesh, T., Prinz, W.A., Palazzo, a-helical rods that are required to restrict the ratio of curvature- and sheet- inducing A.F., Kozlov, M.M., and Rapoport, T.A. (2010). the lateral mobility of Climp-63 (Klopfen- proteins in liposomes of defined lipid Cell 143, this issue, 774–788. stein et al., 2001; Nikonov et al., 2007) compositions could provide insights into Voeltz, G.K., Prinz, W.A., Shibata, Y., Rist, J.M., (Figure 1). Moreover, Climp-63 is known the role that specific lipids play in gener- and Rapoport, T.A. (2006). Cell 124, to bind microtubules (Klopfenstein et al., ating observed ER morphology. 573–586.
666 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
Leading Edge
Previews SIRT3 in Calorie Restriction: Can You Hear Me Now? Carlos Sebastian1 and Raul Mostoslavsky1,* 1The Massachusetts General Hospital Cancer Center, Harvard Medical School, Boston, MA 02114, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.11.009
Caloric restriction decreases oxidative damage and extends life span in many organisms. Someya et al. (2010) show that the sirtuin SIRT3 mediates the protective effects of caloric restriction on agerelated hearing loss by promoting the mitochondrial antioxidant system through the regulation of isocitrate dehydrogenase 2 (Idh2). Despite two decades of effort, caloric restriction remains the only treatment demonstrated to extend life span and to delay the progression of several diseases normally associated with aging, such as cancer, diabetes, and neurological disorders. Early experiments in yeast showed that the life span extension mediated by caloric restriction involves Sir2, the founding member of the sirtuin family of histone deacetylases. However, later experiments have questioned this association (Longo and Kennedy, 2006), and the role of mammalian sirtuins in life span extension by caloric restriction is still under study. In this context, although SIRT1 appears to be the major mammalian sirtuin involved in the metabolic effects of caloric restriction (Haigis and Guarente, 2006), the precise role of sirtuins in the longevity response remains unclear. In this issue of Cell, Someya et al. (2010) bring some light to the field by describing a new function for the mitochondrial SIRT3 protein in the prevention of hearing loss mediated by caloric restriction during aging. These tantalizing results suggest that SIRT3 might play an important role in slowing the aging process in mammals. Age-related hearing loss is a hallmark of mammalian aging and the most common sensory disorder in the elderly (Liu and Yan, 2007). It is characterized by a gradual loss of spiral ganglion neurons and sensory hair cells in the cochlea of the inner ear (Liu and Yan, 2007). Given that the affected cells are postmitotic and do not regenerate, their loss leads to an age-associated decline in hearing function. Several groups have studied hearing loss as an example of age-related degen-
eration in mouse models. Remarkably, early work demonstrated that caloric restriction slows age-related hearing loss in animal models (Sweet et al., 1988). Moreover, in their previous work, Prolla and colleagues demonstrated that caloric restriction induces expression of the SIRT3 gene in the cochlea (Someya et al., 2007). They now elegantly follow up on these results, proving a role for this sirtuin in the delay in hearing loss due to caloric restriction and elucidating the molecular mechanisms underlying this effect. Someya et al. use wild-type and SIRT3deficient mice fed a diet in which caloric intake is reduced to 75% and compare them to control mice fed with a regular diet. The authors first look at the hearing response of the animals and find that, as expected, aging leads to hearing loss in both wild-type and SIRT3-deficient mice. However, whereas caloric restriction delays the progression of hearing loss in wild-type mice, this effect is completely abolished in SIRT3-deficient animals. These results are consistent with the effects of caloric restriction on spiral ganglion neurons and hair cells in these mice. In wild-type animals, a calorie restricted diet reduces the age-related loss of neurons and hair cells, whereas this effect is abrogated in SIRT3-deficient mice. Together, these results clearly pinpoint SIRT3 as a critical molecular determinant regulating the response to caloric restriction in age-related hearing loss. The authors next study the metabolic effects induced by caloric restriction in SIRT3-deficient mice. With a normal diet, SIRT3-deficient animals appear phenotypically normal, in accordance with
previous studies (Schwer et al., 2009). However, whereas wild-type mice display lower levels of serum insulin and triglycerides when fed a calorie-restricted diet, SIRT3-deficient mice do not show this response. Based on these results, the authors argue that SIRT3 plays a role in metabolic adaptations to caloric restriction. It remains unclear, however, whether SIRT3 can also mediate the effects of calorie restriction in other tissues or whether it does so specifically in the context of hearing loss. The authors then investigate the molecular mechanisms involved in this process. Given that caloric restriction reduces ageassociated oxidative damage to macromolecules (Sohal and Weindruch, 1996), Someya et al. analyze levels of oxidative damage to DNA in several tissues. They find that a calorie-restricted diet reduces this type of damage in wild-type mice, but not in SIRT3-deficient animals. Importantly, this is the first evidence that a mammalian sirtuin regulates levels of oxidative stress in response to caloric restriction. But how does SIRT3 regulate oxidative damage during caloric restriction? Given that SIRT3 localizes to the mitochondria, the authors hypothesize that SIRT3 could regulate the antioxidant systems present in this organelle. Using a combination of cellular and biochemical experiments, they discover that SIRT3 regulates the mitochondrial glutathione antioxidant defense system. Glutathione is the main small molecule antioxidant in cells and is generated by glutathione reductase in a reaction dependent on NADPH. The authors show that SIRT3 modulates the
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 667
conversion of oxidized glutaus closer to a healthier life thione to reduced glutathione span. In the words of Francois in response to caloric restricJacob, ‘‘In a world of unlimited tion. They find that, under imagination, we are continuthese conditions, SIRT3 ally inventing a possible world binds and deacetylates the or a piece of a world, and then mitochondrial isocitrate decomparing it with the real hydrogenase 2 enzyme world.’’ In the context of sir(Idh2), the enzyme that genertuins, it seems we are starting ates NADPH, increasing the to put some of these pieces enzyme’s activity. In agreetogether. ment with these results, Idh2 deacetylation and activity, as ACKNOWLEDGMENTS well as NADPH levels, increase during caloric restricWe would like to thank all of the tion in all wild-type tissues members of the Mostoslavsky lab tested, whereas SIRT3 defifor helpful comments. C.S. is the recipient of a Beatriu de Pinos Postciency impairs this response. doctoral Fellowship (Generalitat de Finally, overexpressing SIRT3 Catalunya). R.M. is a Sidney Kimmel and Idh2 promotes cell Scholar, a Massachusetts Life viability upon oxidative damScience Center New Investigator age. Together, these data Scholar, and the recipient of an lead the authors to propose AFAR Award. Work in the Mostoslavsky lab is funded, in part, by a model in which caloric National Institutes of Health. restriction promotes SIRT3 expression, leading to the deacetylation and activation of REFERENCES Idh2, thus providing resisFigure 1. Caloric Restriction, SIRT3, and Age-Related Hearing Loss tance to oxidative stress and During aging (left), oxidative damage (ROS, reactive oxygen species) leads to Finkel, T., Deng, C.X., and Mostothe loss of spiral ganglion neurons and sensory hair cells in the ear, leading to inhibiting the age-related slavsky, R. (2009). Nature 460, age-related hearing loss. Caloric restriction (right) prevents the age-associ587–591. loss of spiral ganglion neuated loss of spiral ganglion neurons and sensory hair cells. Someya et al. rons and hair cells (Figure 1). (2010) show that caloric restriction leads to an increase in SIRT3 levels in Haigis, M.C., and Guarente, L.P. the mitochondria. By promoting the deacetylation of isocitrate dehydrogenase Although Someya et al. (2006). Genes Dev. 20, 2913–2921. 2 (Idh2), SIRT3 promotes the accumulation of NADPH, hence activating glutaprovide enough data to Hirschey, M.D., Shimazu, T., Goetzthione reductase (GSR), which generates reduced glutathione (GSH), a cellular prove that the effects of man, E., Jing, E., Schwer, B., antioxidant. caloric restriction on ageLombard, D.B., Grueter, C.A., Harris, C., Biddinger, S., Ilkayeva, O.R., related hearing loss are et al. (2010). Nature 464, 121–125. dependent on SIRT3, key Liu, X.Z., and Yan, D. (2007). J. Pathol. 211, questions remain. First, does SIRT3 effects of caloric restriction using SIRT3 mediate the effects of caloric restriction activators? If so, such reagents would 188–197. in other tissues? And if so, what are its have significant therapeutic potential. Longo, V.D., and Kennedy, B.K. (2006). Cell 126, substrates? Multiple mitochondrial Finally, because other sirtuins also have 257–268. proteins are deacetylated upon caloric prominent roles in metabolic regulation Schwer, B., Eckersdorff, M., Li, Y., Silva, J.C., Ferrestriction in a SIRT3-dependent manner (Finkel et al., 2009), can we extend min, D., Kurtev, M.V., Giallourakis, C., Comb, M.J., (Schwer et al., 2009). It is therefore some of these findings to other sirtuins? Alt, F.W., and Lombard, D.B. (2009). Aging Cell 8, 604–606. important to determine whether Idh2 is SIRT1, for example, has been linked to the main SIRT3 target in preventing the response of mammals to caloric Sohal, R.S., and Weindruch, R. (1996). Science 273, 59–63. oxidative stress or whether other SIRT3 restriction (Haigis and Guarente, 2006), Someya, S., Yamasoba, T., Weindruch, R., Prolla, substrates contribute as well. Second, and it is therefore possible that the T.A., and Tanokura, M. (2007). Neurobiol. Aging what is the relationship between the activity of this and other sirtuins may be 28, 1613–1622. effect of SIRT3 on Idh2 and the recently regulated in a coordinated fashion Someya, S., Yu, W., Hallows, W.C., Xu, J., Vann, described role for SIRT3 in fatty acid following nutrient starvation. J.M., Leeuwenburg, C., Tanokura, M., Denu, Regardless of the utopian dream of life J.M., and Prolla, T.A. (2010). Cell 143, this issue, oxidation during nutrient stress (Hirschey et al., 2010)? Are these functions coordi- span extension, answering some of these 802–812. nated? If they are not, how is specificity questions may provide a step forward for Sweet, R.J., Price, J.M., and Henry, K.R. (1988). achieved? Third, can we mimic the treating age-related pathologies, bringing Audiology 27, 305–312.
668 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
Leading Edge
Previews ATP Consumption Promotes Cancer Metabolism William J. Israelsen1 and Matthew G. Vander Heiden1,* 1Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology, Cambridge, MA 02139, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.11.010
Cancer cells metabolize glucose by aerobic glycolysis, a phenomenon known as the Warburg effect. Fang et al. (2010) show that the endoplasmic reticulum enzyme ENTPD5 promotes ATP consumption and favors aerobic glycolysis. The findings suggest that nutrient uptake in cancer cells is limited by ATP and satisfies energy requirements other than ATP production. Mounting evidence suggests that cancer cells engage in a unique metabolic program that allows for rapid cell proliferation. Nonproliferating cells can use glycolysis products to generate ATP for their energy needs. Such cells generally have low rates of glycolysis followed by oxidation of pyruvate in the mitochondria, leading to efficient generation of ATP. Dividing cells, in contrast, also use glycolysis intermediates for the synthesis of macromolecules and must therefore balance their ATP requirements and biosynthetic needs (Vander Heiden et al., 2009). Metabolism of glucose by aerobic glycolysis, a phenomenon known as the Warburg effect, may help dividing cells strike this balance. The phosphoinositide 3-kinase (PI3K) signaling pathway, which is activated in many cancers, regulates cell growth and survival. PI3K signaling has been implicated in the altered glucose metabolism of cancer cells, and the serine/threonine kinase AKT, a major PI3K effector, promotes glucose uptake and increases the activity of glycolytic enzymes (DeBerardinis et al., 2008). In this issue of Cell, Fang et al. (2010) report an important mechanism by which AKT signaling leads to increased aerobic glycolysis. They show that AKT activation promotes protein glycosylation in the endoplasmic reticulum, which elevates ATP consumption and derepresses a rate-limiting enzyme in glycolysis that is otherwise inhibited by an elevated ratio of ATP to AMP. This work suggests how proliferating cells may integrate growth signals with energy status to enable increased glucose uptake to support cell proliferation.
Activation of the PI3K pathway in cancer can occur via genetic alterations that allow growth factor-independent kinase activation or via the loss of PTEN, a lipid phosphatase that attenuates PI3K signaling. Fang et al. now find that cell extracts from PTEN-deficient cells have an enhanced ability to generate AMP from ATP. Subsequent purification and biochemical characterization of this activity led to the identification of ectonucleoside triphosphate diphosphohydrolase 5 (ENTPD5) as the enzyme associated with the ATP hydrolysis activity. PI3K signaling leads to upregulation of ENTPD5, a UDPase that promotes the N-glycosylation and folding of glycoproteins in the ER by hydrolyzing UDP to UMP (Trombetta and Helenius, 1999) (Figure 1). UDP hydrolysis in the ER is a reaction necessary to promote protein folding via the calnexin/calreticulin pathway. It is linked to ATP hydrolysis in the cytosol by a cycle of glucose and phosphate transfer reactions. As part of this cycle, the UDP-glucose/UMP antiporter exports UMP out of the ER in exchange for importing UDP-glucose into the ER (Hirschberg et al., 1998). The UGGT enzyme then uses UDP-glucose to transfer glucose to proteins in the ER (Vembar and Brodsky, 2008). This glucose addition to nascent glycoproteins is necessary for their calnexin/calreticulinmediated protein folding. Thus, disruption of ENTPD5 in PTEN-deficient cells results in decreased protein N-glycosylation and causes ER stress. Cell surface proteins, including many growth factor receptors, are N-glycosylated. Fang et al. show that disruption of
ENTPD5 leads to decreased levels of several growth factor receptors, including epidermal growth factor receptor (EGFR), insulin-like growth factor receptor b (IGFR-b), and Her2/ErbB2. Given that growth factor signaling plays an important role in increasing nutrient metabolism in rapidly proliferating cells (DeBerardinis et al., 2008), these new findings suggest that cellular ATP levels can influence the folding and expression of growth factor receptors, perhaps ensuring that cells do not attempt to grow when ATP is limiting. Furthermore, because glucose metabolism by the hexosamine biosynthesis pathway provides the carbon for these receptor glycosylation events, the availability of glucose may provide a means to couple nutrient levels with growth factor receptor expression. These feedbacks may exist to prevent a metabolic catastrophe caused by activation of the cell growth machinery when the supply of nutrients or ATP is limiting. How does ENTPD5 regulate ATP levels? Fang et al. find that reconstitution of the ATP consuming activity also requires the presence of UMP/CMP kinase-1 and adenylate kinase-1. UMP/ CMP kinase-1 catalyzes the rephosphorylation of the UMP generated by ENTPD5 into UDP (Figure 1), in the process converting ATP to ADP. Adenylate kinase-1 then converts ADP molecules into ATP and AMP, thus allowing the cycle to continue. Surprisingly, this cycle involving ENTPD5 is a major source of ATP consumption in PTEN-deficient cells. Furthermore, these reactions directly affect the cell’s glycolytic rate. Whereas increased ENTPD5 expression has no
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 669
Figure 1. ENTPD5 Promotes Glycolysis in Proliferating Cells Fang et al. (2010) show that the endoplasmic reticulum (ER) UDPase ectonucleoside triphosphate diphosphohydrolase 5 (ENTPD5) is expressed in response to phosphoinositide 3-kinase (PI3K) signaling. Activation of PI3K results in FOXO phosphorylation by AKT and loss of ENTPD5 transcriptional repression. This leads to increased ENTPD5 enzyme activity in the ER, promoting protein folding. ENTPD5 activity promotes the import of UDP-glucose into the ER, where UGGT transfers glucose to an unfolded N-glycoprotein and produces UDP. Properly folded N-glycoproteins, such as growth factor receptors, exit the cycle, whereas unfolded proteins undergo further folding attempts or are degraded. ENTPD5 activity enables this process by hydrolyzing UDP to provide the UMP necessary for exchange with UDP-glucose in the cytosol. The activities of UMP/CMP kinase-1 and adenylate kinase-1 couple the energetic requirements of this cycle to the net conversion of ATP to AMP. Thus, increased ENTPD5 activity leads to a decrease in the cellular ATP/AMP ratio. Because this ratio is the major determinant of glucose flux through the phosphofructokinase (PFK) step in glycolysis, a lowered ATP/AMP ratio increases glycolysis, elevates lactate production, and provides glycolytic intermediates for biomass production.
effect on cellular respiration, it increases lactate production, suggesting a link between ATP consumption and increased glycolytic flux. In a series of experiments to determine how ENTPD5 increases glucose entry into glycolysis, Fang et al. find that the ratio of fructose6-phosphate to fructose-1-6-bisphosphate increases in cells following ENTPD5 knockdown, consistent with inhibition of this step in glycolysis. Phosphofructokinase (PFK), the enzyme that catalyzes this reaction, is the major enzyme controlling glucose commitment to the glycolytic pathway (Dunaway, 1983). A high ATP/ AMP ratio in the cell inhibits both PFK activity and glucose metabolism by glycolysis. In fact, the authors conclude that increased ATP consumption by ENTPD5 increases glycolysis by lowering the ATP/AMP ratio and relieving allosteric inhibition of PFK.
ATP is likely not the growth-limiting resource for most cells (Vander Heiden et al., 2009). The concept that glucose utilization by tumor cells may be limited by ATP consumption to prevent feedback inhibition of PFK has been suggested previously (Scholnick et al., 1973). This study finally provides a mechanism by which cells can increase ATP consumption to drive glucose uptake. An additional mechanism has also recently been described in which glucose incorporation into biosynthetic pathways occurs without producing excess ATP (Vander Heiden et al., 2010). Together, these studies support the notion that altered metabolism in cancer is not adapted to support ATP production. Fang et al. show that ENTPD5 expression correlates with PI3K activation in human prostate cancer cell lines and tumor tissue samples. Not all cancer cells
670 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
are dependent on activated PI3K, suggesting that increased ENTPD5 activity may not be a universal mechanism for lowering ATP levels in tumors. However, other enzymes involved in regulating nucleotide pools in cells have also been linked to cancer (Hartsough and Steeg, 2000), and there are additional homologs of ENTPD5 whose functions are not well understood. These enzymes may be involved in analogous cycles of ATP consumption, leading to enhanced glucose metabolism in other genetic contexts. Fang et al. also show that decreased ENTPD5 expression inhibits tumor growth, possibly via pleiotropic effects involving induction of ER stress and altered glucose metabolism. Consideration of ENTPD5 as a potential therapeutic target in PI3K-driven cancer is interesting given that pharmacological inhibition of ENTPD5 is predicted to decrease tumor ATP consumption. Although counterintuitive, the resulting increase in ATP/AMP ratio might reduce the entry of glucose into glycolysis and starve the cells of precursors necessary for biosynthesis. Successful efforts to target cancer metabolism will likely arise from understanding the feedbacks and complex regulation that are required for anabolic metabolism. The study by Fang et al. provides new insight by demonstrating that ATP consumption serves to increase glucose flux to satisfy the energetic and biosynthetic demands of a rapidly proliferating cell. ACKNOWLEDGMENTS We thank Brooke Bevis for her help preparing the figure and editing the manuscript. M.G.V.H. is a consultant to Agios Pharmaceuticals regarding development of compounds targeting cancer metabolism and is a member of its Scientific Advisory Board. REFERENCES DeBerardinis, R.J., Lum, J.J., Hatzivassiliou, G., and Thompson, C.B. (2008). Cell Metab. 7, 11–20. Dunaway, G.A. (1983). Mol. Cell. Biochem. 52, 75–91. Fang, M., Shen, Z., Huang, S., Zhao, L., Chen, S., Mak, T.M., and Wang, X. (2010). Cell 143, this issue, 711–724. Hartsough, M.T., and Steeg, P.S. (2000). J. Bioenerg. Biomembr. 32, 301–308. Hirschberg, C.B., Robbins, P.W., and Abeijon, C. (1998). Annu. Rev. Biochem. 67, 49–69.
Scholnick, P., Lang, D., and Racker, E. (1973). J. Biol. Chem. 248, 5175.
Vander Heiden, M.G., Cantley, L.C., and Thompson, C.B. (2009). Science 324, 1029–1033.
Asara, J.M., and Cantley, L.C. (2010). Science 329, 1492–1499.
Trombetta, E.S., and Helenius, A. (1999). EMBO J. 18, 3282–3292.
Vander Heiden, M.G., Locasale, J.W., Swanson, K.D., Sharfi, H., Heffron, G.J., Amador-Noguez, D., Christofk, H.R., Wagner, G., Rabinowitz, J.D.,
Vembar, S.S., and Brodsky, J.L. (2008). Nat. Rev. Mol. Cell Biol. 9, 944–957.
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 671
Leading Edge
Essay
Glycomics Hits the Big Time Gerald W. Hart1,* and Ronald J. Copeland1 1Department of Biological Chemistry, School of Medicine, Johns Hopkins University, 725 North Wolfe Street, Baltimore, MD 21205-2185, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.11.008
Cells run on carbohydrates. Glycans, sequences of carbohydrates conjugated to proteins and lipids, are arguably the most abundant and structurally diverse class of molecules in nature. Recent advances in glycomics reveal the scope and scale of their functional roles and their impact on human disease. By analogy to the genome, transcriptome, or proteome, the ‘‘glycome’’ is the complete set of glycans and glycoconjugates that are made by a cell or organism under specific conditions. Therefore, ‘‘glycomics’’ refers to studies that attempt to define or quantify the glycome of a cell, tissue, or organism (Bertozzi and Sasisekharan, 2009). In eukaryotes, protein glycosylation generally involves the covalent attachment of glycans to serine, threonine, or asparagine residues. Glycoproteins occur in all cellular compartments. Glycans are also attached to lipids, often ceramide, which is comprised of sphingosine, a hydrocarbon amino alcohol and a fatty acid. Complex glycans are mainly attached to secreted or cell surface proteins, and they do not cycle on and off of the polypeptide. In contrast, the monosaccharide O-linked N-acetylglucosamine (O-GlcNAc) cycles rapidly on serine or threonine residues of many nuclear and cytoplasmic proteins. Identifying the number, structure, and function of glycans in cellular biology is a daunting task but one that has been made easier in recent years by advances in technology and by our growing appreciation of how integral glycans are to biology (Varki et al., 2009). The scope of the glycomics challenge is immense. The covalent addition of glycans to proteins and lipids represents not only the most abundant posttranslational modification (PTM), but also by far the most structurally diverse. Although it is commonly stated that more than 50% of all polypeptides are covalently modified by glycans (Apweiler et al., 1999), even this estimate is far too low because it fails to include that myriad nuclear and cytoplasmic proteins are modified by
O-GlcNAc (Hart et al., 2007). Even though the generic term ‘‘glycosylation’’ is often used to categorize and lump all glycan modifications of proteins into one bin, side by side with other posttranslational modifications such as phosphorylation, acetylation, ubiquitination, or methylation, such a view is not only inaccurate, but also is completely misleading. If one only considers the linkage of the first glycan to the polypeptide in both prokaryotic and eukaryotic organisms, there are at least 13 different monosaccharides and 8 different amino acids involved in glycoprotein linkages, with a total of at least 41 different chemical bonds known to be linking the glycan to the protein (Spiro, 2002). Importantly, each one of these unique glycan:protein linkages is surely as different in both structure and function as protein methylation is from acetylation. Of course, this modification is not only about a single linkage. When structural diversity of the additional oligosaccharide branches of glycans and the added diversity of complex terminal saccharides on glycans, such as fucose or sialic acids (about 50 different sialic acids are known [Schauer, 2009]), are taken into account, the molecular diversity and varied functions of protein-bound glycans rapidly increase exponentially. Just the ‘‘sialome’’ (Cohen and Varki, 2010) rivals or exceeds many other posttranslational modifications in abundance and structural/functional diversity. In addition, chemical modifications, such as phosphorylation, sulfation, and acetylation, increase the glycan structural/functional diversity even more. Thus, categorizing glycosylation as a single type of posttranslational modification is neither useful nor at all reflective of reality.
672 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
Dynamic Structural Complexity Underlies Glycan Functions Glycoconjugates provide dynamic structural diversity to proteins and lipids that is responsive to cellular phenotype, to metabolic state, and to the developmental stage of cells. Complex glycans play critical roles in intercellular and intracellular processes, which are fundamentally important to the development of multicellularity (Figure 1). Unlike nucleic acids and proteins, glycan structures are not hardwired into the genome, depending upon a template for their synthesis. Rather, the glycan structures that end up on a polypeptide or lipid result from the concerted actions of highly specific glycosyltransferases (Lairson et al., 2008), which in turn are dependent upon the concentrations and localization of highenergy nucleotide sugar donors, such as UDP-N-acetylglucosamine, the endpoint of the hexosamine biosynthetic pathway. Therefore, the glycoforms of a glycoprotein depend upon many factors directly tied to both gene expression and cellular metabolism. There are at-least 250 glycosyltransferases in the human genome, and it has been estimated that about 2% of the human genome encodes proteins involved in glycan biosynthesis, degradation, or transport (Schachter and Freeze, 2009). Biosynthesis of the nucleotide sugar donors is directly regulated by nucleic acid, glucose, and energy metabolism, and the compartmentalization of these nucleotide sugar donors is highly regulated by specific transporters. Protein glycosylation is therefore controlled by rates of polypeptide translation and protein folding, localization of and competition between glycosyltransferases,
Figure 1. Glycans Permeate Cellular Biology Complex glycans at the cell surface are targets of microbes and viruses, regulate cell adhesion and development, influence metastasis of cancer cells, and regulate myriad receptor:ligand interactions. Glycans within the secretory pathway regulate protein quality control, turnover, and trafficking of molecules to organelles. Nucleocytoplasmic O-linked N-acetylglucosamine (O-GlcNAc) has extensive crosstalk with phosphorylation to regulate signaling, cytoskeletal functions, and gene expression in response to nutrients and stress.
cellular concentration and localization of nucleotide sugars, the localization of glycosidases, and membrane trafficking. Thus, individual glycosylation sites on the same polypeptide can contain different glycan structures that reflect both the type and status of the cell in which they are synthesized. For example, the glycoforms of the membrane protein Thy-1 are very different in lymphocytes than they are in brain, despite having the same polypeptide sequence (Rudd and Dwek, 1997). Conversely, even small changes in polypeptide sequence or structure will alter the types of glycan structures attached to a polypeptide. For example, histocompatibility antigen polypeptides with more than 90% sequence homology contain different N-linked glycan profiles at individual sites, reflective of their allelic type, even when they are synthesized within the same cells (Swiedler et al., 1985). Thus, site-specific protein glycosylation is highly regulated by
gene expression of glycan-processing enzymes, by polypeptide structure at all levels, and by cellular metabolism. Technology of Glycomics A detailed understanding of cellular processes will require a detailed appreciation of the glycans modulating proteins and pathways. Although this ultimate goal of glycomics is laudable, we are a very long way from having the technology to completely characterize the glycome of even a simple cell or tissue. Not only is the glycome much more complex than the genome, transcriptome, or proteome, as noted above, it is also much more dynamic, varying considerably not only with cell type, but also with the developmental stage and metabolic state of a cell. Even very conservative estimates indicate that there are well over a million different glycan structures in a mammalian cell’s glycome. However, upon considering ‘‘functional glycomics,’’
it is estimated that the binding sites of glycan-binding proteins (GBPs), such as antibodies, lectins, receptors, toxins, microbial adhesions, or enzymes (Figure 1), can accommodate only up to two to six monosaccharides within a glycan structure (Cummings, 2009). Therefore, the number of specific glycan substructures that bind to biologically important GBPs in a cell may be fewer than 10,000, a number that is within the realm of current analytical and, if targeted, chemical or enzymatic synthetic capabilities. Until recently, the lack of tools and the inherent complexity of glycans have been major barriers preventing most biologists from embracing the importance of glycans in biology. Recent technological advances have significantly lowered these barriers. Indeed, the tools of glycomics and the subfields of glycoproteomics, glycolipidomics, and proteoglycomics have all progressed substantially in recent years (Krishnamoorthy and Mahal, 2009; Laremore et al., 2010). Major technological advances, many of which are shared with proteomics, have recently allowed semiquantitative profiling of glycans and glycoproteins (Krishnamoorthy and Mahal, 2009; Vanderschaeghe et al., 2010). Some of these advances are the result of the National Institute of General Medical Science’s (NIGMS) support of the Consortium for Functional Glycomics (CFG), which has served to focus and assist more than 500 researchers on issues related to glycomics (Paulson et al., 2006; Raman et al., 2006). Kobata and colleagues were among the first to profile N-glycans, well before the current concepts of glycomics were conceived. Despite the lack of many modern methods, their pioneering work was characterized by a high level of rigor in defining the arrays of N-glycan structures present in cells and tissues and on specific proteins (Endo, 2010). Currently, a wide variety of high-resolution and highly sensitive methods are available, including capillary electrophoresis (CE), high-performance liquid chromatography (HPLC), and lectin microarrays. Glycans are often profiled after their release from polypeptides, which results in the loss of any information about proteins and sites to which they were attached. Even though it is much more difficult, it is also much preferable to perform
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 673
glycopeptide profiling (glycoproteomics) to first identify attachment sites prior to detailed profiling or structural analysis of the glycans present on a polypeptide. The ultimate goal of glycoproteomics, which is to define all of the molecular species (glycoforms) of glycoproteins in a cell or tissue, has not yet been realized for any glycoprotein with more than one glycan attachment site. N-glycans are generally released from proteins by peptide-N-glycosidase F (PNGase F), which cleaves most, but not all, N-glycans. Unfortunately, no such broadly specific enzyme exists for O-glycans, which are generally released by chemical methods, such as alkaliinduced b elimination, or by hydrazinolysis. However, for relatively pure glycoproteins, so called ‘‘top-down’’ mass spectrometric methods, which do not involve prior release of the glycans, may eventually prove useful, as instrumentation and methods improve (Reid et al., 2002). Due to the small sample sizes involved, most CE or HPLC separation methods require chemical modification of released glycans with fluorescent compounds. CE and HPLC methods provide high-resolution separation of glycans, and when combined with laser-induced fluorescent detection (LIF), tagged glycans can be detected in the low femtomole range. High pH anion-exchange chromatography (HPAEC) with pulsed-amperometric detection separates glycans with high resolution and detects them with high sensitivity without chemical modification, but the high alkalinity employed can be problematic for some labile structures. Lectins, which are defined as carbohydrate-binding proteins that are neither antibodies nor enzymes, have a wide range of glycan binding specificities, suitable for partial characterization of a glycome. Lectin microarrays use methods and equipment similar to that employed for nucleic acid arrays. Given the large number of different lectins available, lectin microarrays can provide information about the glycome in a high-throughput fashion, which is particularly useful in profiling glycans produced by infectious organisms (Hsu et al., 2006). In the future, it is highly likely that glycomics will play a central role in combating infectious disease. However, many technical issues remain to be resolved, such as standardization required for clinical use, the
development of purified recombinant lectins, and better definition of the specificities of many lectins (Gupta et al., 2010). Both matrix-assisted laser desorption ionization (MALDI) and electrospray mass spectrometry have played a key role in glycan profiling and in glycoproteomics (An et al., 2009; North et al., 2010; Zaia, 2010). For biomarker discovery, affinity enrichment approaches, based upon chemical modification and solidphase extraction of N-linked glycoproteins, have proven useful in profiling N-linked glycoprotein sites from serumor even from paraffin-embedded tissues (Tian et al., 2009). Recently, using lectin binding combined with advanced mass spectrometric methods, thousands of N-glycan attachment sites have been mapped, a prerequisite for understanding their functions (Zielinska et al., 2010). Given the structural diversity of glycans, all of these glycomic approaches generate vast amounts of data. Glycan bioinformatics has made great strides within recent years with major efforts from several laboratories (Aoki-Kinoshita, 2008). At least four major publicly available carbohydrate databases (Glycosciences.de, KEGG GLYCAN, EurocarbDB, and CFG) are now maintained, and efforts to structure them in a uniform format have been in progress for quite some time. In addition, the Carbohydrate-Active EnZyme database (CAZy) has played a key role in providing a global understanding of carbohydrate active enzymes, documenting their evolutionary relationships, providing a framework for elucidating common mechanisms, and establishing the relationship between glycogenomics and glycomes expressed by cells (Cantarel et al., 2009). Moreover, recent advances in bioinformatic analysis tools for complex glycomic mass spectrometry data sets have allowed complex data to be presented in formats useful to nonexperts in all fields of biology (Ceroni et al., 2008; Goldberg et al., 2005). Perhaps one of the most important contributions to the field of functional glycomics has been the development of well-defined glycan microarrays, which currently display more than 500 different glycan structures (Smith et al., 2010). The NIGMS-supported Consortium for Functional Glycomics (CFG) has generated and made publicly available
674 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
custom-made DNA microarrays that represent glycosyltransferases and glycan-binding proteins. The CFG also has developed databases that present phenotypic and biochemical data on glycosyltransferase knockout mice. Even though knocking out a single glycosyltransferase gene often affects hundreds of glycoconjugates and myriad biological processes, these mutant mice have proven valuable in revealing the fundamental biological importance of glycans. The microarrays and the databases produced by the CFG member community at large are publically available on the CFG website (http://www.functionalglycomics. org) and have resulted in a profound increase in our understanding of the binding specificities of GBPs, including lectins key to inflammation and immunity, and on infectious microbes or viruses. However, a major barrier preventing glycan biology from being incorporated more into the mainstream is the continued failure by the community to adopt a universally standard glycan structural format and database that are easily accessed worldwide. Most importantly, glycan databases must eventually be incorporated into standard interactive databases that are supported by public agencies (such as NCBI or EMBL) before glycan biology can be fully integrated into the wider research community. From Glycomics to Biology Glycans are directly involved in almost every biological process and certainly play a major role in nearly every human disease (Figure 1). Genetic studies in tissue culture cells indicate that specific complex glycan structures are generally not essential to a cell growing in culture, indicating that most of the functions of complex glycans are at the multicellular level. In contrast, the cycling monosaccharide, O-GlcNAc, on nuclear and cytoplasmic proteins, is essential even at the single cell level in mammals (Hart et al., 2007). The critical roles of glycans in mammals are now well established not only by the dearth of mutations in glycan biosynthetic enzymes that survive development, but also by the severe phenotypes generated when such mutations are not lethal. These severe phenotypes are clearly illustrated by the congenital disorders of
landscape, the pharmaceutical industry and the US Food and Drug Administration are rapidly realizing the critical importance, in terms of both bioactivity and safety, of carefully defining the glycoforms of any therapeutics derived from glycoconjugates.
Figure 2. Glycomic Complexity Reflects Cellular Complexity Given that glycan structures are regulated by metabolism and glyco-enzyme expression and glycans modify both proteins and lipids, functional glycomics also requires the tools of genomics, proteomics, lipidomics, and metabolomics (modified after Packer et al., 2008).
glycosylation (CDGs) (Schachter and Freeze, 2009), which are associated with severe mental and developmental abnormalities. Also, the severe muscular dystrophy that results from defective O-glycosylation of a-dystroglycan (Yoshida-Moriguchi et al., 2010) further illustrates how a mutation in a glycan biosynthetic enzyme results in a devastating disease. The interplay between O-GlcNAcylation and phosphorylation on nuclear and cytoplasmic proteins plays a key role in the etiology of diabetes, neurodegenerative disease, and cancer (Hart et al., 2007; Zeidan and Hart, 2010). It has long been appreciated that alterations in cell surface glycans contribute to the metastatic and neoplastic properties of tumor cells (Taniguchi, 2008). The functions of many receptors are modulated by their glycans, such as modulation of Notch receptors by the action of specific glycosyltransferases (Moloney et al., 2000), which regulate Notch’s activation by its ligands, affecting many developmental events. Selectins, which specifically bind to a subset of fucosylated and sialylated glycans, play a critical role in leukocyte homing to sites of inflammation. Indeed, a selectin inhibitor is currently in phase two clinical trials for vaso-occlusive sickle cell disease (Chang et al., 2010). Siglecs, which are a family of cell surface sialic acid-binding lectins, play a fundamental role in regulating lymphocyte functions and activation. Recent studies on galectins, a family of b-galactoside-binding lectins, have shown that they play a critical role in the
organization of receptors on the cell surface and play important roles in immunity, infections, development, and inflammation (Lajoie et al., 2009). Proteoglycans and glycosaminoglycans play a key role in the regulation of growth factors, in microbial binding, in tissue morphogenesis, and in the etiology of cardiovascular disease. Proteoglycans are perhaps the most complicated and information-rich molecules in biology, and progress in proteoglycomics has begun to accelerate (Ly et al., 2010). Nearly all microbes and viruses that infect humans bind to cells by attaching to specific cell surface glycans. Glycomics and glycan arrays will have a substantial impact upon future research toward both diagnosing and preventing infectious disease. Some of the most important drugs on the market are already the result of glycomics. The anti-flu virus drugs Relenza and Tamiflu are structural analogs of sialic acids that inhibit the flu virus neuraminidase and the transmission of the virus. Natural heparin, a sulfated glycosaminoglycan, and chemically defined synthetic heparin oligosaccharides have long been widely used in the clinic as anticoagulants and for many other clinical uses. Hyaluronic acid, a nonsulfated glycosaminoglycan, is used in the treatment of arthritis. Many recombinant pharmaceuticals, including therapeutic monoclonal antibodies, are glycoproteins, and their specific glycoforms are key to their bioactivity and half lives in circulation and to their possible induction of deleterious immune responses when they do not contain the correct glycans. Given this
Glycoproteomics, Glycolipidomics, and Biomarkers Clinical cancer diagnostic markers are often glycoproteins, but most current diagnostic tests only measure the expression of the polypeptide. Clearly, given the long known alterations in glycans associated with cancer, it is highly likely that cancer markers that detect specific glycoforms of a protein will have much higher sensitivity and specificity for early detection of cancer (Packer et al., 2008; Taniguchi, 2008). Thus, the convergence of glycomics and glycoproteomics is key to the discovery of biomarkers for the early detection of cancer (Taylor et al., 2009). Recently, the Food and Drug Administration has approved fucosylated a-fetoprotein as a diagnostic marker of primary hepatocarcinoma. In addition, fucosylated haptoglobin may be a much better marker of pancreatic cancer than simply monitoring the expression of the haptoglobin polypeptide. Indeed, The National Cancer Institute has begun an initiative to discover, develop, and clinically validate glycan biomarkers for cancer (http:// glycomics.cancer.gov/). System biology analyses of the glycome to identify biomarkers of human disease will, by necessity, also employ many of the same methods used by genomics, proteomics, metabolomics, and lipidomics (Figure 2) (Packer et al., 2008). Due to the critical roles of glycans in cardiovascular disease and lung disease and in the functions of blood cells, the National Heart Lung and Blood Institute (NHLBI) has recognized an acute need to train more researchers in the area of glycosciences by creating a ‘‘Program of Excellence in Glycosciences,’’ which will not only support collaborative research, but will also provide hands-on laboratory training in the methods of glycosciences to fellows. Thus, though our knowledge about the biology of glycans and glycomics continues to lag behind more mainstream fields of genomics and proteomics, technological advances in glycomics in the
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 675
last 5 years have begun to accelerate the integration of glycobiology into the other major fields of biomedical research. A complete mechanistic understanding of the etiology of almost any disease will depend upon the elucidation of the functions of all posttranslational modifications but will especially depend upon our understanding the many roles of glycans, the most abundant and structurally diverse type of posttranslational modification.
Chang, J., Patton, J.T., Sarkar, A., Ernst, B., Magnani, J.L., and Frenette, P.S. (2010). Blood 116, 1779–1786.
Raman, R., Venkataraman, M., Ramakrishnan, S., Lang, W., Raguram, S., and Sasisekharan, R. (2006). Glycobiology 16, 82R–90R.
Cohen, M., and Varki, A. (2010). OMICS 14, 455–464.
Reid, G.E., Stephenson, J.L., Jr., and McLuckey, S.A. (2002). Anal. Chem. 74, 577–583.
Cummings, R.D. (2009). Mol. Biosyst. 5, 1087– 1104.
Rudd, P.M., and Dwek, R.A. (1997). Crit. Rev. Biochem. Mol. Biol. 32, 1–100.
Endo, T. (2010). J. Biochem. 147, 9–17.
Schachter, H., and Freeze, H.H. (2009). Biochim. Biophys. Acta 1792, 925–930.
Goldberg, D., Sutton-Smith, M., Paulson, J., and Dell, A. (2005). Proteomics 5, 865–875. Gupta, G., Surolia, A., and Sampathkumar, S.G. (2010). OMICS 14, 419–436. Hart, G.W., Housley, M.P., and Slawson, C. (2007). Nature 446, 1017–1022.
ACKNOWLEDGMENTS We thank Dr. Chad Slawson for helpful suggestions. Original research in the author’s laboratory was supported by NIH grants R01CA42486, R01 DK61671, and R24 DK084949.
Hsu, K.L., Pilobello, K.T., and Mahal, L.K. (2006). Nat. Chem. Biol. 2, 153–157. Krishnamoorthy, L., and Mahal, L.K. (2009). ACS Chem. Biol. 4, 715–732. Lairson, L.L., Henrissat, B., Davies, G.J., and Withers, S.G. (2008). Annu. Rev. Biochem. 77, 521–555.
REFERENCES
Lajoie, P., Goetz, J.G., Dennis, J.W., and Nabi, I.R. (2009). J. Cell Biol. 185, 381–385.
An, H.J., Froehlich, J.W., and Lebrilla, C.B. (2009). Curr. Opin. Chem. Biol. 13, 421–426.
Laremore, T.N., Leach, F.E., III, Solakyildirim, K., Amster, I.J., and Linhardt, R.J. (2010). Methods Enzymol. 478, 79–108.
Aoki-Kinoshita, K.F. (2008). PLoS Comput. Biol. 4, e1000075. Apweiler, R., Hermjakob, H., and Sharon, N. (1999). Biochim. Biophys. Acta 1473, 4–8. Bertozzi, C.R., and Sasisekharan, R. (2009). Glycomics. In Essentials of Glycobiology, Second Edition, A. Varki, R.D. Cummings, J.D. Esko, H.H. Freeze, P. Stanley, C.R. Bertozzi, G.W. Hart, and M.E. Etzler, eds. (Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press). Cantarel, B.L., Coutinho, P.M., Rancurel, C., Bernard, T., Lombard, V., and Henrissat, B. (2009). Nucleic Acids Res. 37(Database issue), D233–D238. Ceroni, A., Maass, K., Geyer, H., Geyer, R., Dell, A., and Haslam, S.M. (2008). J. Proteome Res. 7, 1650–1659.
Ly, M., Laremore, T.N., and Linhardt, R.J. (2010). OMICS 14, 389–399. Moloney, D.J., Panin, V.M., Johnston, S.H., Chen, J., Shao, L., Wilson, R., Wang, Y., Stanley, P., Irvine, K.D., Haltiwanger, R.S., and Vogt, T.F. (2000). Nature 406, 369–375. North, S.J., Jang-Lee, J., Harrison, R., Canis, K., Ismail, M.N., Trollope, A., Antonopoulos, A., Pang, P.C., Grassi, P., Al-Chalabi, S., et al. (2010). Methods Enzymol. 478, 27–77. Packer, N.H., von der Lieth, C.W., Aoki-Kinoshita, K.F., Lebrilla, C.B., Paulson, J.C., Raman, R., Rudd, P., Sasisekharan, R., Taniguchi, N., and York, W.S. (2008). Proteomics 8, 8–20. Paulson, J.C., Blixt, O., and Collins, B.E. (2006). Nat. Chem. Biol. 2, 238–248.
676 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
Schauer, R. (2009). Curr. Opin. Struct. Biol. 19, 507–514. Smith, D.F., Song, X., and Cummings, R.D. (2010). Methods Enzymol. 480, 417–444. Spiro, R.G. (2002). Glycobiology 12, 43R–56R. Swiedler, S.J., Freed, J.H., Tarentino, A.L., Plummer, T.H., Jr., and Hart, G.W. (1985). J. Biol. Chem. 260, 4046–4054. Taniguchi, N. (2008). Mol. Cell. Proteomics 7, 626–627. Taylor, A.D., Hancock, W.S., Hincapie, M., Taniguchi, N., and Hanash, S.M. (2009). Genome Med 1, 57. Tian, Y., Gurley, K., Meany, D.L., Kemp, C.J., and Zhang, H. (2009). J. Proteome Res. 8, 1657–1662. Vanderschaeghe, D., Festjens, N., Delanghe, J., and Callewaert, N. (2010). Biol. Chem. 391, 149–161. Varki, A., Cummings, R.D., Esko, J.D., Freeze, H.H., Stanley, P., Bertozzi, C.R., Hart, G.W., and Etzler, M.E. (2009). Essentials of Glycobiology, Second Edition (Cold Spring Harbor, NY: Cold Spring Harbor Laboratory Press). Yoshida-Moriguchi, T., Yu, L., Stalnaker, S.H., Davis, S., Kunz, S., Madson, M., Oldstone, M.B., Schachter, H., Wells, L., and Campbell, K.P. (2010). Science 327, 88–92. Zaia, J. (2010). OMICS 14, 401–418. Zeidan, Q., and Hart, G.W. (2010). J. Cell Sci. 123, 13–22. Zielinska, D.F., Gnad, F., Wisniewski, J.R., and Mann, M. (2010). Cell 141, 897–907.
Leading Edge
Essay
What Determines the Specificity and Outcomes of Ubiquitin Signaling? Fumiyo Ikeda,1 Nicola Crosetto,1 and Ivan Dikic1,* 1Frankfurt Institute for Molecular Life Sciences and Institute of Biochemistry II, Goethe University School of Medicine, Theodor-Stern-Kai 7, D-60590 Frankfurt (Main), Germany *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.026
Ubiquitin signals and ubiquitin-binding domains are implicated in almost every cellular process, but how is their functionality achieved in cells? We assess recent advances in monitoring the dynamics and specificity of ubiquitin networks in vivo and discuss challenges ahead. Introduction A small protein modifier, ubiquitin, is the building block of a repertoire of molecular signals spanning from single ubiquitin to ubiquitin chains of different linkage used for posttranslational modification of dozens of cellular proteins (Hershko and Ciechanover, 1998). The seven lysines (K) of ubiquitin (K6, K11, K27, K29, K33, K48, and K63) and the amino-terminal methionine (M1) are connected to the C-terminal glycine for chain assembly, generating polymers (Ikeda and Dikic, 2008; Iwai and Tokunaga, 2009). Ubiquitin signals are recognized and processed by specialized ubiquitin-binding domains (UBDs) that form transient, noncovalent interactions either with ubiquitin moieties or with the linkage region in their chains. So far, roughly 200 intracellular proteins have been recognized to contain one or more UBDs (Dikic et al., 2009). Ubiquitin-UBD interactions regulate almost every aspect of cellular physiology, including protein degradation, receptor trafficking, DNA repair, cell-cycle progression, gene transcription, autophagy, and apoptosis (recently reviewed in Deshaies and Joazeiro, 2009; Kirkin et al., 2009; Raiborg and Stenmark, 2009; Ulrich and Walden, 2010; Wickliffe et al., 2009; Winget and Mayor, 2010). Yet, very little is known about the nature of ubiquitin signals and the dynamics of their interpretation by UBDs in the highly crowded molecular environment of the cell. In particular, it remains unclear how a relatively limited pool of signals (ubiquitin chains and UBDs) with partially overlapping biochemical properties can orchestrate the localization and function
of thousands of proteins involved in very different cellular processes. Here we summarize the most recent advances in understanding specificity determinants in ubiquitin signaling and discuss future challenges in the development of sensitive and reliable methods for monitoring spatial and temporal patterns of ubiquitination in vivo. Diversity of Ubiquitin Signals Despite its relatively rigid globular structure, ubiquitin is one of the most versatile signaling molecules in the cell. Although the surface of ubiquitin is mostly composed of polar residues, it is through its few hydrophobic patches that it interacts with most UBDs (Dikic et al., 2009; Winget and Mayor, 2010). Moreover, the presence of seven lysine residues and the N-terminal methionine within ubiquitin that can be fused to the C-terminal diglycine motif of another ubiquitin allows the formation of polymeric chains endowed with flexibility, as in the case of K63-linked or M1-linked chains (often referred to as linear) (Ikeda and Dikic, 2008; Iwai and Tokunaga, 2009). K48linked and K11-linked chains adopt a more rigid conformation, in which ubiquitin monomers are tightly packed against each other. This creates unique modules composed of aligned ubiquitin moieties in which the hydrophobic patch containing isoleucine 44 is either embedded or facing out toward the surface (Pickart and Fushman, 2004; Bremm et al., 2010; Matsumoto et al., 2010). Conversely, K6-linked chains form an asymmetric compact conformation distinct from any other known type of ubiquitin chain
(Virdee et al., 2010). The possibility of heterotypic ubiquitin chains (that is, with mixed linkages) has been shown in vitro, but their presence and biological functions in vivo remain to be confirmed. Altogether, monoubiquitin and homotypic polyubiquitin chains comprise no more than ten signal types. However, the ability to synthesize homotypic chains of various lengths indicates that the repertoire of ubiquitin signals in the cell may be larger than expected. Signals Decoders: Ubiquitin-Binding Domains Ubiquitin signals are read and processed by UBDs that bind noncovalently to mono- or polyubiquitin chains. To date, five structural folds are known with more than 20 UBDs identified overall (Dikic et al., 2009). UBDs are commonly a-helical structures, zinc fingers, pleckstrin homology (PH) folds, or similar to the ubiquitin-conjugating (Ubc) domain present in E2 enzymes (Dikic et al., 2009). In the majority of cases, isolated UBDs preferentially bind to monoubiquitin via a conserved hydrophobic patch surrounding isoleucine 44. The measured affinity of isolated UBDs for monoubiquitin typically falls in the micromolar range (Dikic et al., 2009; Winget and Mayor, 2010). In certain cases, monoubiquitinUBD interactions have also been demonstrated in the context of endogenous fullsize proteins. For example, UBDs present in Y family polymerases performing DNA translesion synthesis bind the monoubiquitinated sliding clamp PCNA (Bienko et al., 2005), and monoubiquitinated transmembrane receptors are recognized
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 677
by endocytic sorting proteins containing diverse UBDs (Hicke and Dunn, 2003). The affinity of UBD-containing proteins for their monoubiquitinated targets in the cellular environment, however, may be different from that inferred from in vitro studies. In fact, the way ubiquitin signals are decoded in cells may be influenced by multiple factors, including the presence of tandem copies of one UBD in the same protein, oligomerization, and protein compartmentalization (reviewed in Dikic et al., 2009; Winget and Mayor, 2010). In addition to monoubiquitin, many UBDs display either relative or absolute selectivity for certain types of chains (Ikeda and Dikic, 2008; Dikic et al., 2009; Winget and Mayor, 2010). For instance, the Pru (Plextrin receptor for ubiquitin) domain in the proteasome receptor Rpn13 preferentially interacts with K48linked diubiquitin (Husnjak et al., 2008), and the NZF (Npl4 zinc finger) domain in TAK1-binding protein 2 (TAB2) binds specifically to K63-linked ubiquitin (Kulathu et al., 2009; Sato et al., 2009). In contrast, UBDs in NEMO and ABIN proteins (UBAN) bind linear diubiquitin chains with approximately 100-fold higher affinity than K63 or K48 chains, and binding to monoubiqutitin could not be detected (Rahighi et al., 2009; Lo et al., 2009). The selectivity of UBAN for linear chains has been explained by the observation that a NEMO dimer binds symmetrically to linear diubiquitin, involving direct interactions with residues exposed in the glycine-methionine linkages (Rahighi et al., 2009). In addition, the crystal structures of the NZF domain of TAB2 and TAB3 in complex with K63-linked diubiquitin have shown a two-sided ubiquitinbinding surface thanks to a flexible K-linkage positioned away from the interaction surface (Kulathu et al., 2009; Sato et al., 2009). Linkage selectivity can also result from multivalent interaction between tandem UBD arrays in a given protein and ubiquitin monomers or linkages in a polyubiquitin chain. Tandem ubiquitin-interacting motifs (UIMs) in the DNA double-strand break response protein Rap80 are positioned to cross two K63-linked monomers, whereas Ataxin-3 UIMs display K48 avidity (Sims and Cohen, 2009). The proteasome receptor S5a has two UIMs separated by linker
regions and shows a 10-fold higher affinity for diubiquitin over monoubiquitin (Zhang et al., 2009). These observations suggest that the function of tandem UBD arrays is to increase the affinity for a given ubiquitinated substrate rather than simultaneously binding multiple substrates. Specificity and Plasticity of Ubiquitin Signaling Historically, distinct ubiquitin signals have been linked to specific cellular functions. For example, K48-linked chains, also known as ‘‘classical’’ ubiquitin chains, were originally described as the signal that targets substrates for proteasomal degradation (Hershko and Ciechanover, 1998). Nonclassical linkage types, such as K63-, K11-, or M1-linked chains are instead associated with DNA repair regulation, cell-cycle progression, innate immunity, and inflammation (Ikeda and Dikic, 2008; Iwai and Tokunaga, 2009; Matsumoto et al., 2010; Wickliffe et al., 2009). Recent reports, however, have challenged the notion that distinct chain types exclusively regulate specific processes in the cell. For instance, nonclassical ubiquitin signals, such as K11 chains generated by the anaphasepromoting complex (APC/C), can also target selected substrates for proteasomal degradation (Jin et al., 2008). In yeast, cyclin B1 is modified by a mix of K48-, K63-, and K11-linked chains rather than by K48 chains alone (Kirkpatrick et al., 2006). This heterogeneous pool is sufficient to bind to proteasomal ubiquitin receptors and drive cyclin B1 degradation (Kirkpatrick et al., 2006). Furthermore, linear chains, initially discovered as activators of the NF-kB pathway (Tokunaga et al., 2009), can also trigger proteasomal degradation when fused to artificial substrates (Zhao and Ulrich, 2010). So, how is functional specificity of ubiquitin signaling achieved in vivo? Even though evidence indicates that specific chain types control distinct molecular processes, as clearly exemplified by NF-kB signaling, we speculate that additional signals (monoubiquitin and chains with different linkage and length) can control the same molecular process with different kinetics and spatial constraints. It has also been speculated that unanchored ubiquitin chains can regulate NF-kB activation (Xia et al., 2009).
678 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
However, the importance of this regulatory mechanism in vivo remains to be further investigated. Therefore, the decoding of ubiquitin signals might be performed in vivo by different UBDs (not necessarily endowed with absolute selectivity toward monoubiquitin or a particular chain type) embedded in key proteins controlling a particular process. Although this scenario could allow a certain degree of plasticity in ubiquitin signaling, specificity might be determined by the localization and assembly of UBD-containing proteins and enzymes catalyzing ubiquitination reactions. Catching Ubiquitin Signaling in the Act The huge discrepancy between our current understanding of the ubiquitin system from in vitro studies compared to in vivo models stems from the fact that ubiquitination and its recognition and cleavage occur in milliseconds (Pierce et al., 2009), therefore making it challenging to analyze these events in living systems. The first attempts to study ubiquitin signaling in vivo have used antibodies against monoubiquitin, polyubiquitin chains, or, more recently, selective linkages, including K11, K48, K63, and linear chains (Matsumoto et al., 2010; Newton et al., 2008; Wang et al., 2008; Tokunaga et al., 2009) (Figure 1A). Raising linkage-selective antibodies is not easy, despite being urgently needed to provide tools to discriminate between different chain types in the cell. These antibodies were produced either by synthesizing peptides resembling specific linkage bonds (Wang et al., 2008; Tokunaga et al., 2009) or by using the phage-display method (Matsumoto et al., 2010; Newton et al., 2008). Although chain-selective antibodies have been used to demonstrate specific chain formation in several biological settings (such as the NF-kB pathway and cell-cycle progression), their ability to monitor substrates with low abundance and the dynamics of chain (de)conjugation as well as their distribution in vivo are still very limited. Monoclonal antibodies recognizing diglycine-modified lysines have been used in combination with mass spectrometry in efforts to increase the sensitivity of immune-based techniques (Xu et al., 2010) (Figure 1B). These antibodies enrich
for the C-terminal di-glycine mulate in the cell to facilitate motif of ubiquitin attached to their detection, including the the acceptor lysine following use of inhibitors of the proproteolysis of ubiquitinated teasome and of deubiquitiproteins by trypsin (Fignating enzymes (DUBs). This ure 1B). This method revealed has often led to the conclumore than 200 ubiquitinated sion that high-mobility ubiquiproteins from human embrytin-positive smears observed onic kidney 293 cells, the on immunoblots represent majority of which were previthe natural modification of ously unknown targets (Xu substrates by very long ubiqet al., 2010). This strategy uitin chains. This, however, can be coupled to stable can be misleading because isotope labeling with amino the combination of different acids in cell culture (SILAC) ubiquitin signals (monoubito quantitatively explore proquitin or ubiquitin chains) on tein ubiquitination in diverse the same type of substrate biological settings. However, can also yield high-mobility it needs to be noted that smears (Haglund et al., this approach can neither 2003; Huang et al., 2006), detect short-lived proteins and inhibition of DUBs and nor distinguish lysine modifithe proteasome may cause cation by NEDD8. an overrepresentation of The AQUA (absolute quanlong ubiquitin chains that tification) method developed might not naturally occur in in the Gygi laboratory is the cell. another promising approach The question of chain to measure the dynamics of length is important given that ubiquitin signaling in cells chains with different topology (Kirkpatrick et al., 2005). and length may regulate difAQUA relies on the use of ferent cellular functions. For Figure 1. Antibodies for Ubiquitin Signals (A) Linkage-specific antibodies, such as a-lysine 11(K11)-, a-K48-, a-K63stable isotope-labeled interinstance, the length of K48linked ubiquitin chains and a-linear ubiquitin chains, can be applied for the nal standard peptides that linked tetraubiquitin chains detection of endogenous ubiquitination of a specific linkage type. mimic those produced during is optimized for interaction (B) After trypsin digestion of total cell extracts, immunoprecipitation of the tryptic digestion of ubiquitiwith proteasomal receptors samples by a specific antibody against glycine-glycine-lysine peptides (a-GGK Ab) can enrich fragments with ubiquitinated K residues from both nated proteins of interest. (Pickart and Fushman, 2004), substrates and ubiquitin chains. Analysis by mass spectrometry enables the In case of mono- or polyubias a ternary complex can identification of new target proteins as well as sites of ubiquitination. quitinated proteins, tryptic be formed between the ubiqdigestion produces a series uitin chains and proteasomal of unbranched and di-glycine- tandem mass spectrometers, makes receptors Rpn13 and S5a (Zhang et al., branched peptides. Initial analysis of AQUA the tool of choice for quantitatively 2009). Moreover, given that trimming such mixtures allows identification of measuring ubiquitin modifications directly of ubiquitinated substrates occurs from ubiquitination sites in the substrate and in cell lysates (Kirkpatrick et al., 2006). the distal end of the chains, it seems the type of ubiquitin chain linkage (such that the length of K48-linked chains as monoubiquitination or K63- or K48- What Is Known about Ubiquitin dictates the duration of proteasomal ubiquitin chains). Based on this informa- Chain Length In Vivo? degradation (Lee et al., 2010). tion, substrate-, site-, and linkage-specific The methods described above are pre- Monoubiquitination can also drive proreference peptides are synthesized and dicted to provide quantitative information teins to proteasomal degradation (Shabek used as quantitative internal standards, on the repertoire of ubiquitin signals and et al., 2009). These observations collecallowing for precise quantification of ubiquitinated proteins generated in dif- tively suggest that the ubiquitin chain monoubiquitin and polyubiquitin chains ferent biological settings. However, these length required for proteasomal degradaby targeted proteomics approaches such methods cannot monitor the length of tion is determined by the substrate’s as selective reaction monitoring. With ubiquitin chains in vivo. At present, all affinity for the proteasome and must be this methodology, the stoichiometry of our knowledge on their length in vivo just high enough to allow processivity of ubiquitin moieties on a protein of interest relies on nonquantitative analysis of the proteolytic process. This kind of can be determined (Figure 2A). Its immunoblots. Several procedures have adjustment is most likely controlled by simplicity and sensitivity, coupled with been designed to cause ubiquitin chains a proteasome-associated complex, the current widespread availability of and polyubiquitinated substrates to accu- which is equipped with both ubiquitin Cell 143, November 24, 2010 ª2010 Elsevier Inc. 679
ligase (HUL5) and deubiquitinating (UBP6) activities (Crosas et al., 2006). In the case of the NF-kB pathway, distinct activation steps involve K63, linear, and K48 chains (Bianchi and Meier, 2009), which are further edited (in length and topology) by ligases and DUBs (Wertz et al., 2004; Newton et al., 2008). An initial mechanism proposed for NFkB activation implicated long K63-linked chains in the recruitment of TAK1 and IKK kinases via their respective adaptor proteins TAB2 and NEMO (reviewed in Bianchi and Meier, 2009). This model has been challenged by the demonstration that cells expressing ubiquitin lacking K63 have intact NF-kB signaling via tumor necrosis factor-a receptors (Xu et al., 2009). Interestingly, based on available structures it appears that chain-selective UBDs in TAB2 and NEMO interact with K63-linked or linear diubiquitin chains, respectively (Kulathu et al., 2009; Rahighi et al., 2009; Sato et al., 2009). Given that no data are available on the precise length of ubiquitin chains in the NF-kB pathway, it is tempting to speculate that diubiquitin chains are the fundamental units recognized by selective UBDs. However, UBDs also show promiscuous binding with lower affinities to other types of chains. Examples include the NZF domain of TAB2, which also binds K48 chains in solution (Kulathu et al., 2009), and the UBAN domain in NEMO, which interacts with K63- and K48-linked chains longer than diubiquitin (Rahighi et al., 2009). We speculate that diubiquitin units in longer chains may amplify signals that can be recognized by nonselective
UBDs. In such a scenario, short ubiquitin chains added to substrates will be preferentially decoded by linkageselective UBDs, whereas long chains may be promiscuously read by different UBDs, possibly placing chain length next to chain linkage type in determining ubiquitinUBD selectivity.
Figure 2. Quantification and Detection of Ubiquitin Chains In Vivo (A) The workflow for the AQUA (absolute quantification) method of quantitative mass spectrometry is depicted. First, a representative tryptic peptide is selected based on initial proteomic sequencing experiments and then synthesized with a stable isotope at one residue for identification. The tryptic peptide sequence for lysine 48 (K48)-linked ubiquitin chains is indicated (upper panel). AQUA peptide standards are added to the sample (cell lysates or immunocomplexes) prior to trypsin digestion and targeted proteomic analysis is performed using selective reaction monitoring. The amount of total protein and the extent of ubiquitination at that particular site can be determined by comparing the precise amounts of the unmodified and ubiquitinated versions of the peptide (lower panel). (B) Schematic models of ubiquitin sensors are shown. By using different ubiquitin-binding domains (UBDs), the sensor can be applied for specific linkage type of ubiquitin chains (left), such as K48, K63, and linear chains. Tandem UBDs may be used to determine the chain length (right). One UBD recognizes 1 unit of diubiquitin. The tag chosen depends on the experimental purposes.
680 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
Development of Sensors Using Selective UBDs In order to measure the dynamics of ubiquitin chain formation/disassembly and their length in vivo, functional ubiquitin sensors are needed (Figure 2B). A recently engineered sensor (TUBE, tandem repeated ubiquitin entities) possesses four tandem UBA domains of either HR23 or ubiquitin 1 (Hjerpe et al., 2009). The ubiquitin-binding capacity of TUBE is markedly higher for ubiquitin tetramers in comparison to monoubiquitin. In addition, the affinity of the interaction of TUBE with either K63- or K48-tetraubiquitin chains is much greater than that of a single UBA domain (Hjerpe et al., 2009). An intriguing feature of TUBE is its ability to protect ubiquitin chains from cleavage by blocking accessibility to DUBs. The design principle of TUBE could be easily adapted to other UBDs: for example, a K63 chain-specific sensor could be created by fusing multiple NZF domains of TAB2 in tandem, a K48specific sensor by merging multiple Pru domains of Rpn13, and a linear-specific sensor by arraying several copies of the UBAN domain of NEMO or ABINs. These UBD-derived ubiquitin sensors could be used to protect and purify substrates decorated with endogenous ubiquitin chains. They could also
be used to determine the predominant linkage type within these chains by competition experiments and for measuring the length of ubiquitin polymers in their natural environment. A further critical challenge will be to evaluate chain-specific ubiquitin sensors using advanced (high-throughput) singlecell or -molecule microscopy. This might permit the qualitative and quantitative assessment of ubiquitin chain formation and the interplay between different chain types in vivo. Analyzing additional properties, such as the spatial and temporal regulation of conjugation and deconjugation of ubiquitin chains as well as their length in vivo, could enable a highresolution, systems-level analysis of the ‘‘ubiquitinome.’’ Perspective Even though we have attained a sophisticated mechanistic understanding of the ubiquitin system, it has been more difficult to analyze the orchestration of its functions in vivo. Within the cellular environment, ubiquitin signals must select the correct binding partner at the right place and time, ensuring accurate signaling. To understand the specificity and dynamics of the ubiquitin system in its biological context, it is critical that highly sensitive methods, such as mass spectrometry and advanced microscopy, are deployed to measure key parameters, such as the amount of different ubiquitin signals, the kinetics of UBD-ubiquitin recognition, and the type and length of ubiquitin chains attached onto substrates in vivo. By shedding light onto these properties, we will gain a deeper understanding of one of the most important and widely used regulatory systems of cell physiology. ACKNOWLEDGMENTS We are grateful to C. Behrends, A. Ciechanover, K. Rittinger, and S. van Wijk for comments and discussions. Research in the I.D. laboratory is supported by the Deutsche Forschungsgemeinschaft, the Cluster of Excellence ‘‘Macromolecular Complexes’’ of the Goethe University Frankfurt (EXC115), and the European Research Council under the European Union’s Seventh Framework
Programme (FP7/2007-2013)/ERC grant agreement n [250241-LineUb]. REFERENCES Bianchi, K., and Meier, P. (2009). Mol. Cell 36, 736–742. Bienko, M., Green, C.M., Crosetto, N., Rudolf, F., Zapart, G., Coull, B., Kannouche, P., Wider, G., Peter, M., Lehmann, A.R., et al. (2005). Science 310, 1821–1824. Bremm, A., Freund, S.M., and Komander, D. (2010). Nat. Struct. Mol. Biol. 17, 939–947. Crosas, B., Hanna, J., Kirkpatrick, D.S., Zhang, D.P., Tone, Y., Hathaway, N.A., Buecker, C., Leggett, D.S., Schmidt, M., King, R.W., et al. (2006). Cell 127, 1401–1413. Deshaies, R.J., and Joazeiro, C.A. (2009). Annu. Rev. Biochem. 78, 399–434. Dikic, I., Wakatsuki, S., and Walters, K.J. (2009). Nat. Rev. Mol. Cell Biol. 10, 659–671. Haglund, K., Sigismund, S., Polo, S., Szymkiewicz, I., Di Fiore, P.P., and Dikic, I. (2003). Nat. Cell Biol. 5, 461–466. Hershko, A., and Ciechanover, A. (1998). Annu. Rev. Biochem. 67, 425–479. Hicke, L., and Dunn, R. (2003). Annu. Rev. Cell Dev. Biol. 19, 141–172. Hjerpe, R., Aillet, F., Lopitz-Otsoa, F., Lang, V., England, P., and Rodriguez, M.S. (2009). EMBO Rep. 10, 1250–1258.
Matsumoto, M.L., Wickliffe, K.E., Dong, K.C., Yu, C., Bosanac, I., Bustos, D., Phu, L., Kirkpatrick, D.S., Hymowitz, S.G., Rape, M., et al. (2010). Mol. Cell 39, 477–484. Newton, K., Matsumoto, M.L., Wertz, I.E., Kirkpatrick, D.S., Lill, J.R., Tan, J., Dugger, D., Gordon, N., Sidhu, S.S., Fellouse, F.A., et al. (2008). Cell 134, 668–678. Pickart, C.M., and Fushman, D. (2004). Curr. Opin. Chem. Biol. 8, 610–616. Pierce, N.W., Kleiger, G., Shan, S.O., and Deshaies, R.J. (2009). Nature 462, 615–619. Rahighi, S., Ikeda, F., Kawasaki, M., Akutsu, M., Suzuki, N., Kato, R., Kensche, T., Uejima, T., Bloor, S., Komander, D., et al. (2009). Cell 136, 1098– 1109. Raiborg, C., and Stenmark, H. (2009). Nature 458, 445–452. Sato, Y., Yoshikawa, A., Yamashita, M., Yamagata, A., and Fukai, S. (2009). EMBO J. 28, 3903–3909. Shabek, N., Herman-Bachinsky, Y., and Ciechanover, A. (2009). Proc. Natl. Acad. Sci. USA 106, 11907–11912. Sims, J.J., and Cohen, R.E. (2009). Mol. Cell 33, 775–783. Tokunaga, F., Sakata, S., Saeki, Y., Satomi, Y., Kirisako, T., Kamei, K., Nakagawa, T., Kato, M., Murata, S., Yamaoka, S., et al. (2009). Nat. Cell Biol. 11, 123–132. Ulrich, H.D., and Walden, H. (2010). Nat. Rev. Mol. Cell Biol. 11, 479–489.
Huang, F., Kirkpatrick, D., Jiang, X., Gygi, S., and Sorkin, A. (2006). Mol. Cell 21, 737–748.
Virdee, S., Ye, Y., Nguyen, D.P., Komander, D., and Chin, J.W. (2010). Nat. Chem. Biol. 6, 750–757.
Husnjak, K., Elsasser, S., Zhang, N., Chen, X., Randles, L., Shi, Y., Hofmann, K., Walters, K.J., Finley, D., and Dikic, I. (2008). Nature 453, 481–488.
Wang, H., Matsuzawa, A., Brown, S.A., Zhou, J., Guy, C.S., Tseng, P.H., Forbes, K., Nicholson, T.P., Sheppard, P.W., Hacker, H., et al. (2008). Proc. Natl. Acad. Sci. USA 105, 20197–20202.
Ikeda, F., and Dikic, I. (2008). EMBO Rep. 9, 536–542. Iwai, K., and Tokunaga, F. (2009). EMBO Rep. 10, 706–713. Jin, L., Williamson, A., Banerjee, S., Philipp, I., and Rape, M. (2008). Cell 133, 653–665. Kirkin, V., McEwan, D.G., Novak, I., and Dikic, I. (2009). Mol. Cell 34, 259–269. Kirkpatrick, D.S., Weldon, S.F., Tsaprailis, G., Liebler, D.C., and Gandolfi, A.J. (2005). Proteomics 5, 2104–2111. Kirkpatrick, D.S., Hathaway, N.A., Hanna, J., Elsasser, S., Rush, J., Finley, D., King, R.W., and Gygi, S.P. (2006). Nat. Cell Biol. 8, 700–710. Kulathu, Y., Akutsu, M., Bremm, A., Hofmann, K., and Komander, D. (2009). Nat. Struct. Mol. Biol. 16, 1328–1330. Lee, M.J., Lee, B.H., Hanna, J., King, R.W., and Finley, D. (2010). Mol. Cell. Proteomics 2010, 8. Lo, Y.C., Lin, S.C., Rospigliosi, C.C., Conze, D.B., Wu, C.J., Ashwell, J.D., Eliezer, D., and Wu, H. (2009). Mol. Cell 33, 602–615.
Wertz, I.E., O’Rourke, K.M., Zhou, H., Eby, M., Aravind, L., Seshagiri, S., Wu, P., Wiesmann, C., Baker, R., Boone, D.L., et al. (2004). Nature 430, 694–699. Wickliffe, K., Williamson, A., Jin, L., and Rape, M. (2009). Chem. Rev. 109, 1537–1548. Winget, J.M., and Mayor, T. (2010). Mol. Cell 38, 627–635. Xia, Z.P., Sun, L., Chen, X., Pineda, G., Jiang, X., Adhikari, A., Zeng, W., and Chen, Z.J. (2009). Nature 461, 114–119. Xu, G., Paige, J.S., and Jaffrey, S.R. (2010). Nat. Biotechnol. 28, 868–873. Xu, M., Skaug, B., Zeng, W., and Chen, Z.J. (2009). Mol. Cell 36, 302–314. Zhang, N., Wang, Q., Ehlinger, A., Randles, L., Lary, J.W., Kang, Y., Haririnia, A., Storaska, A.J., Cole, J.L., Fushman, D., et al. (2009). Mol. Cell 35, 280–290. Zhao, S., and Ulrich, H.D. (2010). Proc. Natl. Acad. Sci. USA 107, 7704–7709.
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 681
Leading Edge
Minireview Ubiquitin: Same Molecule, Different Degradation Pathways Michael J. Clague1,* and Sylvie Urbe´1,* 1Cellular and Molecular Physiology, Institute of Translational Medicine, University of Liverpool, Crown Street, Liverpool L69 3BX, UK *Correspondence:
[email protected] (M.J.C.),
[email protected] (S.U.) DOI 10.1016/j.cell.2010.11.012
Ubiquitin is a common demoninator in the targeting of substrates to all three major protein degradation pathways in mammalian cells: the proteasome, the lysosome, and the autophagosome. The factors that direct a substrate toward a particular route of degradation likely include ubiquitin chain length and linkage type, which may favor interaction with particular receptors or confer differential susceptibility to deubiquitinase activities associated with each pathway. The dynamic state of bodily proteins was established by analyzing the fate of stable isotope-labeled amino acids that had been fed to mice. These classic experiments, conducted by Rudolf Schoenheimer in the late 1930s, presage modern stable isotope labeling techniques (such as SILAC), which allow determination of the turnover rate of hundreds to thousands of individual proteins in a single mass spectrometry experiment (Kristensen et al., 2008). After its discovery, the lysosomal compartment was considered the principal site of degradation of cellular proteins, through the action of resident acid-dependent proteases. However, this view perished with the demonstration that the half-lives of most cellular proteins are insensitive to alkalinization of the lysosomes. The subsequent discovery of the ubiquitin-proteasome degradation system as the major route to protein degradation generated a new orthodoxy. Central to this model is the idea that covalent modification of substrate proteins with a polypeptide ubiquitin tag targets them to the large (26S) proteolytic complex known as the proteasome. It came then as a surprise to discover that ubiquitin tagging also provides a signal to route endocytosed receptors to the lysosomal degradation pathway and more recently to mark organelles for disposal by the third major cellular degradative pathway of autophagocytosis. The role of ubiquitin in protein degradation is more ubiquitous than once thought (Figure 1). In this Minireview, we consider how a ubiquitin tag selects for specific degradation pathways and also highlight the interplay between these pathways that a shared dependence on ubiquitin engenders. General Considerations Substrate proteins are selected for modification of lysine residues by ubiquitin through interaction with an E3 ligase protein that recruits an E2-enzyme charged with ubiquitin. This can result in transfer of a single ubiquitin molecule (monoubiquitination) or coupling of further ubiquitin molecules, through integral lysine residues, to form a chain. The seven lysines of ubiquitin provide for the formation of different isopeptide chain linkages, which adopt different three-dimensional structures, and all of which are represented in eukaryotic cells (Xu et al., 2009). The specific combination of E2 and E3 enzymes recruited to 682 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
a substrate dictates the chain linkage type. The human genome encodes more than 20 different types of ubiquitin-binding domains, and proof of principle for linkage specificity of binding has been established (see Essay by F. Ikeda, N. Crosetto, and I. Dikic on page 677 of this issue). One means to achieve this is through the spatial arrangement of tandem ubiquitin-binding domains (UBDs) either encoded in a single protein or by combining domains within a multimolecular complex, such that simultaneous occupancy of two binding sites is restricted to particular chain configurations. Proteasomal Degradation Early work suggested that proteasomal targeting requires a lysine 48 (K48)-linked ubiquitin chain consisting of at least four conjoined ubiquitin molecules. This was based first upon the biochemical analysis of chains formed on a model substrate, b-galactosidase, in a reticulocyte lysate system and second upon studies showing that unique among lysine mutant versions of ubiquitin, K48R cannot serve as the sole source of ubiquitin in yeast (Finley, 2009; Xu et al., 2009). The affinity of unanchored K48 polyubiquitin chains for the proteasome increases more than 100-fold from di- to tetraubiquitin (170 nM) and less steeply thereafter (Thrower et al., 2000). A body of work now suggests that in fact the proteasome happily accepts other ubiquitin chain types. Indirect evidence for this comes from the observation that acute proteasome inhibition does not lead to the selective accumulation of K48 chains. Rather, all chain types with the exception of K63 are increased (Jacobson et al., 2009; Xu et al., 2009). During cell division, the human anaphase-promoting complex (APC/C) recruits two E2 ligases (UbcH10 and Ube2S), which combine to exclusively generate K11-linked chains on substrates. Loss of this unit leads to strong defects in mitotic progression due to failure of the necessary degradation processes (Song and Rape, 2010). In vitro studies have even shown that K63-modified dihydrofolate reductase provides an efficient proteasome substrate (Hofmann and Pickart, 1999). The proteasome is composed of a core (20S) particle containing multiple proteolytic sites and a 19S regulatory particle that
Figure 1. Ubiquitin Is a Common Denominator of Protein Degradation Pathways Specific ubiquitin receptors are associated with each degradation pathway. Autophagosomal and multivesicular body (MVB) pathways merge at the lysosome and share a dependence on v-ATPase activity (inhibited by bafilomycin). Both pathways also share sensitivity to inhibitors of phosphoinositide 3-kinase activity, such as wortmannin or 3-methyladenine, as the family member hVPS34 is required both for recruitment of ESCRT (endosomal sorting complex required for transport) components to MVBs and for expansion of the double-membrane preautophagosomal structure. Proteasomal inhibitors include lactacystin and epoxomicin.
governs access to the core. To enter the core, substrates must be amenable to unfolding by a hexamer of ATPases associated with the base of the regulatory particle. Other constituents of the regulatory particle are implicated in the recruitment of substrates (Finley, 2009). Rpn10 and Rpn13 interact with ubiquitinated substrates through UIM (ubiquitin-interacting motif) domains and a Pru (pleckstrin-like receptor for ubiquitin) domain, respectively. The UBL/UBA family of proteins are substoichiometric components of purified proteasomes that bind ubiquitin via their UBA (ubiquitin-associated) domain and the proteasome regulatory particle through its UBL (ubiquitin-like) domain. They are proposed to remotely scavenge ubiquitinated substrates and present them to the proteasome (Figure 2). Particular proteasome-associated ubiquitin receptors have been linked with the degradation of specific substrates (reviewed in Finley, 2009). The mammalian regulatory particle has three associated deubiquitinating enzymes (DUBs), POH1/PSMD14, USP14, and UCH37 (Rpn11 and Ubp6 in budding yeast), which have distinct specificities for different chain linkages (Finley, 2009). What is the function of these DUB activities? One important function is to salvage ubiquitin in order to maintain the cellular ubiquitin pool. The JAMM/MPN+ metalloprotease POH1 is thought to specifically disassemble K63-linked chains, as well as cleave the isopeptide bond that links the substrate and the proximal ubiquitin, allowing for en bloc removal of an ubiquitin chain. It also governs entry into the central proteolytic chamber, thereby coupling substrate degradation to recycling of ubiquitin (Yao and Cohen, 2002). Ubiquitin-aldehyde-sensitive cysteine
Figure 2. Ubiquitin Recognition by the Major Degradative Pathways Depiction of the ‘‘ubiquitin receptors’’ associated with each degradative pathway. The domain structures shown are for the human representatives of each protein family, except for yeast Ddi1, the human ortholog of which does not contain a UBA domain. CB: clathrin-binding motif; CC: coiled coil; ESCRT: endosomal sorting complex required for transport; GGA: golgi-associated, gamma adaptin ear containing, ARF-binding protein; GAE: gamma adaptin ear; GAT: GGA and TOM1; GLUE: GRAM-like ubiquitin-binding in Eap45; HRS: HGF receptor tyrosine kinase substrate; LIR: LC3-interacting region; PB1: Phox and Bem1; PRU: Pleckstrin-like receptor for ubiquitin; SH3: Src homology domain 3; STAM: signal transducing adaptor molecule; TOM1: target of myb1; TSG101: tumor susceptibility gene 101; UBA: ubiquitin-associated domain; UBL: ubiquitin-like domain; UEV: ubiquitin E2 variant domain; UIM: ubiquitin-interacting motif; VHS: Vps27, HRS, and STAM; VPS36: vacuolar protein sorting 36; vWFA: von Willebrand Factor type A; ZZ: zinc finger. Note the following gene names and commonly used alternative names also apply: p62; SQSTM1 (sequestosome), NDP52; CALCOCO2, UBQLN1; PLIC1; DSK2. Domain annotation based on PFAM and UNIPROT.
protease activities (that is, USP14 and UCH37) account for all activity directed toward K48-linked chains and also contribute to K63-linked chain disassembly (Jacobson et al., 2009). One attractive notion is that the integration of these DUB activities may provide for a proof-reading mechanism, facilitating release from the proteasome if commitment to degradation is not accomplished within a given time window. For example, preferential proteasomal DUB activity against K63-linked chains has been proposed to select against these substrates for degradation (Jacobson et al., 2009). Also in line with this principal, Cell 143, November 24, 2010 ª2010 Elsevier Inc. 683
a specific chemical inhibitor of USP14 has recently been shown to enhance the rate of protein degradation (Lee et al., 2010). In yeast, a ubiquitin ligase, Hul5 (mammalian ortholog is KIAA10/E3a), that is associated with proteasomes can oppose Ubp6 activity through chain elongation (E4) (Crosas et al., 2006). Thus a balance between proteasome-associated ubiquitin ligase and DUB activity may determine receptor fate. Endolysosomal Degradation The lysosomal degradation pathway is the principle means by which a cell turns over plasma membrane proteins, such as receptors or channels. Its defining characteristic is a requirement for organelle acidification, mediated by the v-ATPase. Endocytosed proteins are either recycled to the plasma membrane or captured into lumenal vesicles of the multivesicular body (MVB) as it matures from the sorting endosome, before fusing directly with lysosomes. Some receptors use ubiquitin as an internalization signal, but for other ubiquitinated receptors, such as epidermal growth factor receptor, this is secondary to, or redundant with, other adaptor-binding motifs. Ubiquitination directs internalized proteins toward lysosomal degradation by engagement with endosomal sorting complexes required for transport (ESCRTs) (reviewed in Clague and Urbe´, 2006). Monoubiquitination, in the form of an irreversible linear fusion appended to various receptors, is a sufficient signal for this sorting step. However, evidence suggests K63 as the primary ubiquitin chain type involved in endosomal sorting. Early studies in yeast cells, which suggested that appendage of K63-linked diubiquitin enhances vacuolar sorting, have been recently elaborated on with a detailed analysis of the downregulation of the Gap1 permease. These studies conclude that monoubiquitination is sufficient for initial internalization (at least so long as it is an irreversible linear fusion) but that efficient sorting at the endosome by the ESCRT machinery requires K63linked polyubiquitin (Lauwers et al., 2009). Concordantly, studies of the mammalian TrkA and MHC class I proteins reveal their utilization of K63-linked polyubiquitination for routing to the lysosome (Duncan et al., 2006; Geetha and Wooten, 2008). The first point of engagement of ubiquitinated cargo with the MVB sorting machinery is proposed to be the ESCRT-0 complex, comprising HRS and STAM, both of which possess UIM and VHS (Vps27, HRS, and STAM) domains, which can bind ubiquitin (Figure 2). Intact ESCRT-0 binds 50 times more tightly to K63linked tetraubiquitin than to monoubiquitin, but only 2-fold more tightly than to K48-tetraubiquitin (Ren and Hurley, 2010). ESCRT-0 is recruited to endosomes through binding to phosphatidylinositol 3-phosphate but also binds to clathrin and the downstream ESCRT-I complex. An alternative ESCRT-0 complex comprising TOM1, Tollip, and Endofin possesses all these salient features of the HRS-STAM complex. It is currently unclear whether these two complexes are redundant or used to receive different cargoes. In a further striking parallel to the proteasomal system, the ESCRT machinery has known associations with at least two DUB activities, AMSH and USP8 (UBPY), drawn from the JAMM/MPN+ and USP families, respectively. In yeast, the dominant endocytic E3 ligase activity Rsp5 can also associate with the STAM ortholog Hse1, providing a counterbalance to Ubp2 and Ubp7 (Ren et al., 2007), while a third ESCRT-associated DUB Doa4 is required for ubiquitin recycling of receptors 684 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
that are committed to degradation. Although deubiquitination is not an obligate step for MVB sorting, proof-reading and ubiquitin recycling roles akin to those suggested for proteasomal DUBs are consistent with available data (Clague and Urbe´, 2006). Autophagy The signature of autophagy is the capture of cytosol and organelles through envelopment within a double-membrane compartment derived from the preautophagosomal structure. In common with the MVB, the autophagosome can then directly fuse with late endosomes or lysosomes to form the autolysosome, wherein the double-membrane structure is digested. It is well suited for the digestion of cytosolic entities, which are incompatible with unfolding by the proteasome, such as organelles or protein aggregates. Identification of autophagy (Atg) genes and subsequent biochemical characterization revealed two essential posttranslational modification pathways, which resemble ubiquitination. In one case, Atg12 is stably conjugated to Atg5 in a constitutive fashion. In the second case, Atg8 is conjugated to the lipid phosphatidylethanolamine by transfer from an E2 enzyme following the onset of autophagy (for example, as induced by amino acid depravation). This is a prerequisite for the expansion of the preautophagosomal structure, perhaps by facilitating fusion between membranes. In mammalian cells, Atg8 is known as LC3 and its lipidated form as LC3-II. In fact, there are six Atg8 homologs in the human genome collectively known as the LC3/GABARAP family. Whereas autophagy is generally thought of as a nonselective degradation process, certain structures and organelles are selectively removed by this pathway. For example, mitochondria are lost during reticulocyte maturation and as a consequence of uncoupling (disconnecting the electron transport chain from ATP production) in cultured cells. Ribosomes, peroxisomes, and pathophysiological protein aggregates can also be degraded by autophagy. Recent studies have led to the proposal of a common principle involved in ‘‘selective autophagies’’ and once again ubiquitin plays a critical role (Kirkin et al., 2009). In general if the body to be cleared is ubiquitinated, then an adaptor molecule is required to couple this to the preautophagosomal membrane rich in Atg8/LC3. The prototypical adaptor of this class is p62/sequestosome 1, which contains both a ubiquitin-interacting domain (UBA) and a LIR motif (LC3-interacting region), a domain structure shared with Neighbor of BRCA1 gene 1 (NBR1) (Figure 2) (Pankiv et al., 2007). p62 has been previously implicated in the clearing of protein aggregates, which are known to be concentrated in ubiquitin. Recent data have indicated an essential role for ubiquitin (K63 and K27 polyubiquitin chain linkages have been implicated) in the selective autophagy of depolarized mitochondria, which become ubiquitinated following recruitment of the E3 ubiquitin ligase Parkin (Geisler et al., 2010). Using a lysine-less mutant of ubiquitin fused with red fluorescent protein, Kim et al. established that irreversible monoubiquitination is sufficient to concentrate a soluble protein within autophagosomal structures in a p62dependent manner (Kim et al., 2008). A selective pathway requiring the Ubp3:Bre5 DUB complex in Saccharomyces cerevisiae operates in the removal of mature ribosomes (Kraft and Peter, 2008). In cells deficient in Ubp3, ribosomal fractions are enriched with ubiquitin. Although an intimate
connection has been established, the exact role of ubiquitin in ribophagy is unclear. One model posits that ubiquitin may be protecting ribosomes from autophagy, which is then promoted by Ubp3 activity. Alternatively, a dynamic modification with ubiquitin may be required, perhaps as an engulfment signal similar to that of mitochondria. In support of this notion, a temperature-sensitive defect in the E3 ligase Rsp5 shows a synthetic ribophagy defect with loss of Ubp3 as compared with cells lacking Ubp3 alone (Kraft and Peter, 2008). If correct, then the principle of ensuring ubiquitin homeostasis through deubiquitination may be conserved by each of the selective degradation pathways we have discussed. The Interdependence of Degradation Pathways The relative contribution of degradation pathways may vary greatly between cell types. In most cases of cells cultured under stress-free conditions, proteasomal degradation predominates, but in muscle cells, lysosomal pathways (principally autophagy) can account for 40% of degradation of long-lived proteins. In atrophying muscle cells, both pathways are proposed to be co-ordinately upregulated under the transcriptional control of FOXO3 (Zhao et al., 2007). However, the proteasome is itself degraded by starvation-induced bulk autophagy (Kristensen et al., 2008). The reliance of three major cellular degradation pathways upon ubiquitination suggests that specific inhibition of any one pathway may perturb the ubiquitin economy of the cell and hence indirectly affect other degradation events (Figure 1). A clear example of this is the activated Met receptor, for which its lysosomal degradation is exquisitely sensitive to the depletion in free ubiquitin caused by proteasomal inhibition (Carter et al., 2004). Proteasome inhibition may also induce autophagy as a compensatory response. The autophagy adaptor protein p62 has also been implicated in proteasomal degradation, whereas the E3 ligase Parkin generates an autophagy tag on mitochondria but elsewhere can target proteins to the proteasome. VCP/p97 co-ordinates a number of ubiquitin-dependent processes that include the proteasome-dependent ERAD (endoplasmic reticulum-associated degradation) pathway but interestingly has recently been identified as a necessary factor for autophagosome maturation under basal conditions and following proteasome inhibition (Tresse et al., 2010). The MVB and autophagy pathways merge at the late endosome/lysosome and are both sensitive to proton pump and phosphoinositide 3-kinase inhibitors. Autophagosome formation is inherently sensitive to perturbations earlier in the endocytic pathway, which change the character of later endosomal compartments (such as the composition of SNARE proteins). Occasionally, teleological distinctions between these systems blur, such that some ubiquitinated cytosolic proteins may be degraded in the lysosome and cytoplasm-exposed domains of receptors may be nibbled by the proteasome. Mounting evidence suggests that there is a proteasome pool associated with endosomes that influences receptor sorting (Geetha and Wooten, 2008; Gorbea et al., 2010). Concluding Remarks Ubiquitin tagging is common to the three major cellular pathways for protein degradation. Herein lies a conundrum: how is a given
substrate targeted to a particular pathway? Variable parameters include location, chain length, and linkage type. A clear bias of the endosomal pathway toward K63-linked chains has emerged. This may simply reflect the subcellular localization of specific E3 ligases in combination with a high local concentration of ubiquitin-binding proteins, which couple to the ESCRT-machinery rather than the proteasome. New techniques allow for the determination of individual protein turnover on a global scale (Kristensen et al., 2008). This will enable the generation of a comprehensive annotation of turnover rates as a function of experimental perturbations or disease states, opening the door to a systems-level understanding of protein degradation. ACKNOWLEDGMENTS S.U. is a Cancer Research UK Senior Research Fellow. REFERENCES Carter, S., Urbe´, S., and Clague, M.J. (2004). J. Biol. Chem. 279, 52835–52839. Clague, M.J., and Urbe´, S. (2006). Trends Cell Biol. 16, 551–559. Crosas, B., Hanna, J., Kirkpatrick, D.S., Zhang, D.P., Tone, Y., Hathaway, N.A., Buecker, C., Leggett, D.S., Schmidt, M., King, R.W., et al. (2006). Cell 127, 1401–1413. Duncan, L.M., Piper, S., Dodd, R.B., Saville, M.K., Sanderson, C.M., Luzio, J.P., and Lehner, P.J. (2006). EMBO J. 25, 1635–1645. Finley, D. (2009). Annu. Rev. Biochem. 78, 477–513. Geetha, T., and Wooten, M.W. (2008). Traffic 9, 1146–1156. Geisler, S., Holmstro¨m, K.M., Skujat, D., Fiesel, F.C., Rothfuss, O.C., Kahle, P.J., and Springer, W. (2010). Nat. Cell Biol. 12, 119–131. Gorbea, C., Pratt, G., Ustrell, V., Bell, R., Sahasrabudhe, S., Hughes, R.E., and Rechsteiner, M. (2010). J. Biol. Chem. 285, 31616–31633. Hofmann, R.M., and Pickart, C.M. (1999). Cell 96, 645–653. Jacobson, A.D., Zhang, N.Y., Xu, P., Han, K.J., Noone, S., Peng, J., and Liu, C.W. (2009). J. Biol. Chem. 284, 35485–35494. Kim, P.K., Hailey, D.W., Mullen, R.T., and Lippincott-Schwartz, J. (2008). Proc. Natl. Acad. Sci. USA 105, 20567–20574. Kirkin, V., McEwan, D.G., Novak, I., and Dikic, I. (2009). Mol. Cell 34, 259–269. Kraft, C., and Peter, M. (2008). Autophagy 4, 838–840. Kristensen, A.R., Schandorff, S., Høyer-Hansen, M., Nielsen, M.O., Ja¨a¨ttela¨, M., Dengjel, J., and Andersen, J.S. (2008). Mol. Cell. Proteomics 7, 2419–2428. Lauwers, E., Jacob, C., and Andre´, B. (2009). J. Cell Biol. 185, 493–502. Lee, B.H., Lee, M.J., Park, S., Oh, D.C., Elsasser, S., Chen, P.C., Gartner, C., Dimova, N., Hanna, J., Gygi, S.P., et al. (2010). Nature 467, 179–184. Pankiv, S., Clausen, T.H., Lamark, T., Brech, A., Bruun, J.A., Outzen, H., Øvervatn, A., Bjørkøy, G., and Johansen, T. (2007). J. Biol. Chem. 282, 24131– 24145. Ren, J., Kee, Y., Huibregtse, J.M., and Piper, R.C. (2007). Mol. Biol. Cell 18, 324–335. Ren, X., and Hurley, J.H. (2010). EMBO J. 29, 1045–1054. Song, L., and Rape, M. (2010). Mol. Cell 38, 369–382. Thrower, J.S., Hoffman, L., Rechsteiner, M., and Pickart, C.M. (2000). EMBO J. 19, 94–102. Tresse, E., Salomons, F.A., Vesa, J., Bott, L.C., Kimonis, V., Yao, T.P., Dantuma, N.P., and Taylor, J.P. (2010). Autophagy 6, 217–227. Xu, P., Duong, D.M., Seyfried, N.T., Cheng, D., Xie, Y., Robert, J., Rush, J., Hochstrasser, M., Finley, D., and Peng, J. (2009). Cell 137, 133–145. Yao, T., and Cohen, R.E. (2002). Nature 419, 403–407. Zhao, J., Brault, J.J., Schild, A., Cao, P., Sandri, M., Schiaffino, S., Lecker, S.H., and Goldberg, A.L. (2007). Cell Metab. 6, 472–483.
Cell 143, November 24, 2010 ª2010 Elsevier Inc. 685
Leading Edge
Perspective Will the Ubiquitin System Furnish as Many Drug Targets as Protein Kinases? Philip Cohen1,2,* and Marianna Tcherpakov3 1MRC
Protein Phosphorylation Unit Institute for Cell Signalling Sir James Black Centre, Dow Street, Dundee DD1 5EH, Scotland, UK 3BCC Research, 40 Washington Street, Suite 110, Wellesley, MA 02481, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.11.016 2Scottish
Protein phosphorylation and protein ubiquitination regulate most aspects of cell life, and defects in these control mechanisms cause cancer and many other diseases. In the past decade, protein kinases have become one of the most important classes of drug targets for the pharmaceutical industry. In contrast, drug discovery programs that target components of the ubiquitin system have lagged behind. In this Perspective, we discuss the reasons for the delay in this pipeline, the drugs targeting the ubiquitin system that have been developed, and new approaches that may popularize this area of drug discovery in the future. Protein Phosphorylation Drug Discovery It can take years, even decades, before a field of research reaches the stage of maturity at which its discoveries can obviously be exploited for the improvement of health. An excellent example of this paradigm is the regulation of protein function by reversible phosphorylation. Phosphorylation was identified in the mid 1950s as a mechanism for controlling glycogenolysis. Twentyfive years later, it was still largely thought of simply as a control switch for metabolism. Indeed, researchers finally realized that protein phosphorylation regulates most aspects of cell life only after many advances made throughout the 1980s and early 1990s (Cohen, 2002a). Surprisingly, the idea that it would be possible to treat diseases with drugs targeting protein kinases was even slower to take root. Indeed, as late as 1998, the Head of Research and Development at one major pharmaceutical company (which no longer exists) told one of the authors that ‘‘there was absolutely no future in kinase drug discovery.’’ Later that same year, researchers revealed the remarkable clinical efficacy of a tyrosine kinase inhibitor, called Gleevec, for treating chronic myelogenous leukemia. Quite quickly, protein kinases then became one of the most popular classes of drug targets for the pharmaceutical industry, especially in the field of cancer treatment. Over the past decade, 16 drugs targeting one or more protein kinases have been approved for clinical use in cancer, 12 taken orally as pills and 4 that are injected. As of 2009, 153 other protein kinase inhibitors were undergoing clinical trials, and 23 of these drugs were in the most advanced stage of development, termed Phase III (Table 1) (Lawler, 2009). The current global market for kinase therapies is about US$15 billion per annum, and this value is forecasted to double by 2020. Research on protein kinases currently accounts for 30% of the drug discovery programs in the pharmaceutical industry and over 50% of cancer research and development. The kinase inhibitors 686 Cell 143, November 24, 2010 ª2010 Elsevier Inc.
undergoing Phase III clinical trials include Pfizer’s JAK3 inhibitor for rheumatoid arthritis (CP-690550) and Incyte Pharmaceutical’s JAK1/JAK2 inhibitor (INCB18424) for treating inflammatory diseases. If these drugs are approved, it will likely spark a new wave of interest in developing kinase inhibitors for the treatment of diseases other than cancer. Even by the late 1970s and early 1980s researchers had shown that oncogenes, such as Src (sarcoma), are protein kinases; phorbol esters, which promote tumors, are kinase activators; and, growth factor receptors, which have kinase domains, are overexpressed or mutated in human cancer (reviewed in Cohen, 2002b). So why did it take so long for most pharmaceutical companies to capitalize on the therapeutic potential of kinase inhibitors? In retrospect, one realizes that many researchers believed that kinase inhibitors were bad drug targets because they thought that it would be difficult to achieve the requisite specificity and potency. Most protein kinase inhibitors target the ATP-binding pockets of these enzymes, and the structural similarities of this site among many different kinases raised the suspicion that it would be impossible to develop drugs that inhibited only one type of protein kinase. Furthermore, the concentration of ATP in the cell is extremely high (i.e., millimolar), leading researchers to doubt whether compounds could be developed with the potency needed to compete successfully with intracellular ATP. These were, and indeed still are, challenging problems for many developing kinase inhibitors, but they have proven to be quite surmountable. Indeed, considerable potency and specificity have been achieved by developing compounds that target not only the ATP-binding site but also small hydrophobic pockets located proximal to the ATP-binding site. Moreover, researchers are identifying an increasing number of ‘‘allosteric’’ inhibitors that bind to other regions of a kinase. These compounds induce conformational changes in the kinase, which either suppress
Table 1. Phosphorylation, Ubiquitination, and Drug Discovery Phosphorylation
Ubiquitination
First publication 1955a >500 protein kinases
First publication 1978b 10 E1sf, 40 E2sf, >600 E3 ligasesf
c
140 protein phosphatases
c
90 deubiquitinasesc
d
Nobel Prize awarded 2004e
Nobel Prize awarded 1992
First drug approved in 2001 (Gleevec)
First drug approved in 2003 (Bortezomib)
16 drugs approved, over 150 undergoing clinical trials
One drug approved, 16 undergoing clinical trials
Current sales US$15 billion per year
Current sales US$1.4 billion per year
30% of pharmaceutical research and development
98% of cells (Figure S2B). Rapid turnover of endosomal actin was also independently confirmed by fluorescence recovery after photobleaching (FRAP) studies. When a single endosomal actin spot was bleached, the fluorescence recovered rapidly within 20 s (Figure 4C). As a control for more stable actin filaments, stress fibers showed little recovery of fluorescence after bleaching in this interval (Figure 4C). Exponential curve fits yielding a t1/2 of 8.26 s (99% CI = 7.65 to 8.97 s), consistent with rapid actin turnover (Figure 4D). In contrast, only part of the fluorescence (30%) was recovered in stress fibers in the same cells by 20 s, with curve fits yielding a t1/2 of 50.35 s (99% CI = 46.05 to 55.54 s). These results indicate that actin is dynamically assembled on the B2AR recycling tubules. Considering the rapid turnover of actin, we next explored the machinery responsible for localizing actin at the tubule. The Arp2/3 complex is a major nucleator of dynamic actin polymerization that has been implicated in polymerization-based endosome motility (Stamnes, 2002; Girao et al., 2008; Pollard, 2007). Arp3, an integral part of the Arp2/3 complex useful for visualizing this complex in intact cells (Merrifield et al., 2004), was specifically concentrated at the base of the B2AR tubules on the endosome (e.g., in Figure 4E and fluorescence trace in Figure 4F, Movie S7). Every B2AR tubule observed had a corresponding Arp3 spot at its base (n = 200). Surprisingly, however, we did not see N-WASP and WAVE-2, canonical members of the
two main families of Arp2/3 activators (Millard et al., 2004), on the endosome (Figure 4G). Similarly, we did not see endosomal recruitment of activated Cdc42, as assessed by a previously characterized GFP-fusion reporter consisting of the GTPase binding domain of N-WASP (Benink and Bement, 2005) (data not shown). All three proteins were readily detected at lamellipodia and filopodia as expected, indicating that the proteins were functional in these cells. While we cannot rule out a weak or transitory interaction of these activators with Arp2/3 at the endosome, the lack of enrichment prompted us to test for alternate Arp2/3 activators. Cortactin, an Arp- and actin- binding protein present on endosomes, has been proposed to be such an activator (Kaksonen et al., 2000; Millard et al., 2004; Daly, 2004). Cortactin-GFP was clearly concentrated at the base of the B2AR tubule on the endosome (Figure 4G), in a pattern identical to Arp2/3. When quantified (>200 endosomes each), every B2AR tubule was marked by cortactin, while none of the endosomes showed detectable N-WASP, WAVE-2, or Cdc42. Similarly, the WASH protein complex, which has been recently implicated in trafficking from the endosome (Derivery et al., 2009; Gomez and Billadeau, 2009; Duleh and Welch, 2010), was also clearly localized to B2AR tubules (Figure 4G). Together, these data suggest that an Arp2/3-, cortactin- and WASH-based machinery mediates dynamic actin assembly on the endosome. B2AR-Containing Tubules Are a Specialized Subset of Recycling Tubules on the Endosome Since the traditional view is that the endosomal tubules that mediate direct recycling to the plasma membrane are a uniform population, we next tested whether these tubules were the same as those that recycle bulk cargo. When B2AR recycling was visualized along with bulk recycling of TfR, endosomes containing both cargo typically extruded three to four tubules containing TfR. Strikingly, however, only one of these contained detectable amounts of B2AR (Example in Figure 5A, quantified in Figure 5B). This was consistent with fast 3D confocal live cell imaging of B2AR in endosomes, which showed that most endosomes extruded only one B2AR containing tubule, with a small fraction containing two. When quantified, only 24.4% of all TfR tubules contained detectable B2AR (n = 358 tubules). B2AR Tubules Are a Kinetically and Biochemically Distinct from Bulk Recycling Tubules When the lifetimes of tubules were quantified, the majority (>80%) of B2AR tubules lasted more than 30 s. In contrast, the majority of TfR tubules devoid of B2AR lasted less than 30 s (Figures 5B and 5C, Movie S8). Each endosome extruded several tubules containing TfR, only a subset (30%) of which were marked by actin, coronin, or cortactin (Figures 5D and 5E, arrows). Time-lapse movies indicated that the highly transient TfR-containing tubules were extruded from endosomal domains that were lacking cortactin (Figure 5E, arrows), while the relatively stable B2AR containing tubules were marked by cortactin (Figure 5E, arrowheads). Importantly, the relative stability of the subset of tubules was conferred by the actin cytoskeleton, as disruption of actin using latrunculin virtually abolished the stable fraction of TfR tubules (Figures 5B and 5C). Cell 143, 761–773, November 24, 2010 ª2010 Elsevier Inc. 765
Figure 4. Actin on B2AR Tubules Is Dynamic and Arp2/3-Nucleated (A) Cells expressing actin-GFP imaged live after treatment with 10 mM latrunculin for the indicated times, show rapid loss of endosomal actin. A time series of the boxed area, showing several endosomal actin loci, is shown at the lower panel. (B) The change in endosomal and cytoplasmic actin fluorescence over time after latrunculin normalized to initial endosomal actin fluorescence (n = 10). Onephase exponential curve fits (solid lines) show a t1/2 of 3.5 s for actin loss (R2 = 0.984, d.f = 23, Sy.x = 2.1 for endosomal actin, R2 = 0.960, d.f = 23, Sy.x = 1.9 for cytoplasmic). Endosomal and cytoplasmc actin fluorescence becomes statistically identical within 15 s after latrunculin. Error bars denote SEM. (C) Time series showing FRAP of representative examples of endosomal actin (top) and stress fibers (bottom). (D) Kinetics of FRAP of actin (mean ± s.e.m) quantified from 14 endosomes and 17 stress fibers. One-phase exponential curve fits (lines), show a t1/2 of 8.26 s for endosomal actin (R2 = 0.973, d.f = 34, Sy.x = 4.8) and 50.35 s for stress fibers (R2 = 0.801, d.f = 34, Sy.x = 3.9).
766 Cell 143, 761–773, November 24, 2010 ª2010 Elsevier Inc.
Together, these results suggest that sequence-dependent recycling of B2AR is mediated by specialized tubules that are kinetically and biochemically distinct from the bulk recycling tubules containing only TfR. A Kinetic Model for Sorting of B2AR into a Subset of Endosomal Tubules The relative stability of B2AR tubules suggested a simple model, based on kinetic sorting, for how sequence-dependent cargo was sorted into a specific subset of tubules and excluded from the transient TfR-containing bulk-recycling tubules. We hypothesized that B2AR diffuses more slowly on the endosomal membrane relative to bulk recycling cargo. The short lifetimes of the bulk-recycling tubules would then create a kinetic barrier for B2AR entry, while this barrier would be overcome in the subset of tubules stabilized by actin. To test the key prediction of this model, that B2AR diffuses more slowly than TfR on the endosomal membrane, we directly measured the diffusion rates of B2AR and TfR using FRAP. When B2AR or TfR was bleached on a small part of the endosomal membrane, B2AR fluorescence took significantly longer to recover than TfR (Figure 5F). When quantified, the rate of recovery of fluorescence of B2AR (t1/2 = 25.77 s, 99% CI 23.45 to 28.6 s) was 4 times slower than that of TfR (t1/2 = 6.21 s, 99% CI 5.49 to 7.17 s), indicating that B2AR diffuses significantly slower on the endosomal membrane than TfR (Figures 5F and 5G). Neither B2AR or TfR recovered within the time analyzed when the whole endosome was bleached (Figure 5H), confirming that the recovery of fluorescence was due to diffusion from the unbleached part of the endosome and not due to delivery of new receptors via trafficking. Further, B2AR on the plasma membrane diffused much faster than on the endosome (t1/2 = 6.45 s, 99% CI 5.62 to 7.66 s), comparable to TfR, suggesting that B2AR diffusion was slower specifically on the endosome (Figure 5H). We next tested whether the diffusion of B2AR into endosomal tubules was slower than that of TfR, by using the rate of increase of B2AR fluorescence as an index of receptor entry into tubules. B2AR fluorescence continuously increased throughout the duration of the tubule lifetimes (Figure S3A). Further, in a single tubule containing TfR and B2AR, TfR fluorescence reached its maximum at a markedly faster rate than that of B2AR (Figure S3B). Together, these results suggest that slow diffusion of B2AR on the endosome and stabilization of recycling tubules by actin can provide a kinetic basis for specific sorting of sequence-dependent cargo into subsets of endosomal tubules. Local Actin Assembly Is Required for B2AR Entry into the Subset of Tubules Because actin stabilizes the B2AR-containing subset of tubules, the model predicts that endosomal actin would be required for
sequence-dependent concentration of B2AR into these tubules. Consistent with this, B2AR was no longer concentrated in endosomal tubules when endosomal actin was acutely removed using latrunculin (e.g., in Figure 6A). When the pixel fluorescence along the limiting membrane of multiple endosomes was quantified, B2AR was distributed more uniformly along the endosomal membrane in the absence of actin (Figures 6B and 6C). We further confirmed this by comparing the variance in B2AR fluorescence along the endosomal perimeter, irrespective of their orientation. B2AR fluorescence was significantly more uniform in endosomes without actin (Figure 6D), indicating that actin was required for endosomes to concentrate B2AR in microdomains. Less than 20% of endosomes showed B2AR-containing tubules in the absence of endosomal actin, in contrast to control cells where over 75% of endosomes showed B2AR-containing tubules (Figure 6E). Further, cytochalasin D, a barbed-end capping drug that prevents further actin polymerization but does not actively cause depolymerization, also inhibited B2AR entry into tubules (Figure 6E) and B2AR surface recycling (Figure S4A). Neither TfR tubules on endosomes (Figure 6E) nor TfR recycling (Figure S4B) was inhibited by actin depolymerization, consistent with a role for actin specifically in sequencedependent recycling of B2AR (Cao et al., 1999). Further, depletion of cortactin using siRNA (Figure 6F) also inhibited B2AR entry into tubules (Figures 6G and 6H). This inhibition was specific to cortactin depletion, as it was rescued by exogenous expression of cortactin (Figure 6H). Together, these results indicate that a localized actin cytoskeleton concentrates sequencedependent recycling cargo into a specific subset of recycling tubules on the endosome. B2AR Sorting into the Recycling Subdomains Is Mediated by Its C-Terminal PDZ-Interacting Domain We next asked whether this actin-dependent concentration of receptors into endosomal tubules depended on the PDZ-interacting sequence present in the B2AR cytoplasmic tail that mediates sequence-dependent recycling (Cao et al., 1999; Gage et al., 2005). To test if the sequence was required, we used a mutant B2AR (B2AR-ala) in which the recycling sequence was specifically disrupted by the addition of a single alanine (Cao et al., 1999). Unlike B2AR, internalized B2AR-ala was not able to enter the tubular domains in the endosome (e.g., in Figure 6I, quantified in Figure 6J), or recycle to the cell surface (Figure S4). To test if this sequence was sufficient, we used a chimeric DOR construct with the B2AR-derived recycling sequence fused to its cytoplasmic tail, termed DOR-B2 (Gage et al., 2005), which recycles much more efficiently than DOR (Figure S4). In contrast to DOR, which showed little concentration in endosomal tubules, DOR-B2 entered tubules (Figures 6I and 6J) and recycled in an actin-dependent manner similar to B2AR (Figure S4D). Together, these results indicate that the
(E) Example endosomes in live cells coexpressing B2AR and Arp3-GFP showing Arp3 at the base of B2AR tubules (arrowhead in the inset). (F) Trace of linear pixel fluorescence of B2AR and Arp3 shows Arp3 specifically on the endosomal tubule. (G) Example endosomes from cells coexpressing B2AR and N-WASP-, WAVE2-, cortactin-, or WASH-GFP. N-WASP and WAVE2 were not detected on endosomes, while cortactin and WASH were concentrated at the B2AR tubules (arrowheads). Scale bars represent 1 mm. See also Figure S2 and Movie S7.
Cell 143, 761–773, November 24, 2010 ª2010 Elsevier Inc. 767
Figure 5. B2AR Is Enriched Specifically in a Subset of Endosomal Tubules that Are Stabilized by Actin (A) A representative example of an endosome with two tubules containing TfR, only one of which is enriched for B2AR. (B) The number of tubules with B2AR, TfR, and TfR in the presence of 10 mM latrunculin, per endosome per min, binned into lifetimes less than or more than 30 s, quantified across 28 endosomes and 281 tubules. (C) The percentages of B2AR, TfR, and TfR + latrunculin tubules with lifetimes less than or more than 30 s, normalized to total number of tubules in each case. (D) An example endosome containing TfR and coronin, showing that coronin is present on a subset of the TfR tubules. Arrowheads indicate a TfR tubule that is marked by coronin, and arrows show a TfR tubule that is not. (E) Time lapse series showing TfR-containing tubules extruding from endosomal domains without detectable cortactin. Arrowheads indicate a relatively stable TfR tubule that is marked by coronin, and arrows denote rapid transient TfR tubules without detectable cortactin. (F) Frames from a representative time lapse movie showing FRAP of B2AR (top row) or TfR (bottom row). The circles mark the bleached area of the endosome. TfR fluorescence recovers rapidly, while B2AR fluorescence recovers slowly. (G) Fluorescence recovery of B2AR (red circles) and TfR (green diamonds) on endosomes quantified from 11 experiments. Exponential fits (solid lines) show that B2AR fluorescence recovers with a t1/2 of 25.77 s (R2 = 0.83, d.f = 37, Sy.x = 6.3), while TfR fluorescence recovers with a t1/2 of 6.21 s (R2 = 0.91, d.f = 30, Sy.x = 7.1).
768 Cell 143, 761–773, November 24, 2010 ª2010 Elsevier Inc.
PDZ-interacting recycling sequence on B2AR was both required and sufficient to mediate concentration of receptors in the actinstabilized endosomal tubular domains. As PDZ-domain interactions have been established to indirectly link various integral membrane proteins to cortical actin (Fehon et al., 2010), we tested whether linking DOR to actin was sufficient to drive receptor entry into endosomal tubules. Remarkably, fusion of the actin-binding domain of the ERM protein ezrin (Turunen et al., 1994) to the C terminus of DOR was sufficient to localize the receptor (termed DOR-ABD) to endosomal tubules (Figure 6J). The surface recycling of B2AR, DOR-B2, and DOR-ABD were dependent on the presence of an intact actin cytoskeleton (Figure S4), consistent with previous publications (Cao et al., 1999; Gage et al., 2005; Lauffer et al., 2009). Further, transplantation of the actin-binding domain was also sufficient to specifically confer recycling to a version of B2AR lacking its native recycling signal (Figure S4F). These results indicate that the concentration of B2AR in the actin-stabilized recycling tubules is mediated by linking receptors to the local actin cytoskeleton through PDZ interactions. DISCUSSION Even though endocytic receptor sorting was first appreciated over two decades ago (e.g., Brown et al., 1983; Farquhar, 1983; Steinman et al., 1983), our understanding of the principles of this process has been limited. A major reason for this has been the lack of direct assays to visualize signaling receptor sorting in the endosome. Here we directly visualized, in living cells, endosomal sorting between two prototypic members of the largest known family of signaling receptors for which sequence-specific recycling is critical for physiological regulation of cell signaling (Pippig et al., 1995; Lefkowitz et al., 1998; Xiang and Kobilka, 2003). We resolve sorting at the level of single trafficking events on individual endosomes, and define a kinetic and affinity-based model for how sequence-dependent receptors are sorted away from bulk-recycling and degrading proteins. By analyzing individual sorting and recycling events on single endosomes, we demonstrate a remarkable diversity in recycling pathways emanating from the same organelle (Scita and Di Fiore, 2010). The traditional view has been that recycling to the plasma membrane is mediated by a uniform set of endosomal tubules from a single endosome. In contrast to this view, we demonstrate that the recycling pathway is highly specialized, and that specific cargo can segregate into specialized subsets of tubules that are biochemically, biophysically, and functionally distinct. Receptor recycling plays a critical role in controlling the rate of cellular re-sensitization to signals (Lefkowitz et al., 1998; Sorkin and von Zastrow, 2009), and recent data suggest that the sequence-dependent recycling of signaling receptors is selectively controlled by signaling pathways (Yudowski et al., 2009). The physical separation between bulk and sequencedependent recycling that we demonstrate here allows for such
selective control without affecting the recycling of constitutively cycling nutrient receptors. Further, such physical separation might also reflect the differences in molecular requirements that have been observed between bulk and sequence-dependent recycling (Hanyaloglu and von Zastrow, 2007). Endosome-associated actin likely plays a dual role in endosomal sorting, both of which contribute to sequence-dependent entry of cargo selectively into special domains. First, by stabilizing the specialized endosomal tubules relative to the much more dynamic tubules that mediate bulk recycling, the local actin cytoskeleton could allow sequence-dependent cargo to overcome a kinetic barrier that limits their entry into the bulk pathway. Supporting this, we show that most endosomal tubules are highly transient, lasting less than a few seconds (Figures 5B and 5C), which allows enough time for entry of the fast-diffusing bulk recycling cargo, but not the slow-diffusing sequencedependent cargo (Figures 5F and 5G), into these tubules. A subset of these tubules representing the sequence-dependent recycling pathway is stabilized by the presence of an actin cytoskeleton (Figures 5B and 5C). This stabilization allows time for B2AR to diffuse into these tubules (Figure S3), which eventually pinch off membranes that can directly fuse with the plasma membrane (Figure 2). Interestingly, inhibition of actin caused a decrease in the total number of tubules by approximately 25% (Figure 5B), suggesting that the actin cytoskeleton plays a role in maintaining the B2AR-containing subset of tubules, and not just in the sorting of B2AR into these tubules. Second, a local actin cytoskeleton could provide the machinery for active concentration of recycling proteins like the B2AR, which interact with actin-associated sorting proteins (ERM and ERM-binding proteins) through C-terminal sequences (Weinman et al., 2006; Wheeler et al., 2007; Lauffer et al., 2009; Fehon et al., 2010), in specialized recycling tubules. Consistent with this, the C-terminal sequence on B2AR was both required and sufficient for sorting to the endosome and for recycling, and a distinct actin-binding sequence was sufficient for both receptor entry into tubules and recycling (Figure 6 and Figure S4). PDZ-interacting sequences have been identified on several signaling receptors, including multiple GPCRs, with different specificities for distinct PDZ-domain proteins (Weinman et al., 2006). Further, actin-stabilized subsets of tubules were present even in the absence of B2AR in the endosome. We propose that, using a combination of kinetic and affinity-based sorting principles, discrete Actin-Stabilized SEquence-dependent Recycling Tubule (ASSERT) domains could thus mediate efficient sorting of sequence-dependent recycling cargo away from both degradation and bulk recycling pathways that diverge from the same endosomes. Our results, therefore, uncover an additional role for actin polymerization in endocytic sorting, separate from its role in endosome motility. It will be interesting to investigate the mechanism and signals that control the nucleation of such a spatially localized actin cytoskeleton on the endosome. The lack of obvious
(H) Fluorescence recovery of B2AR (blue triangles) and TfR (green diamonds) on endosomes when the whole endosome was bleached, or of B2AR on the cell surface (red circles) quantified from 12 experiments. B2AR fluorescence on the surface recovers with a t1/2 of 6.49 s (R2 = 0.94, d.f = 27, Sy.x = 8.1). Error bars denote SEM. Scale bars represent 1 mm. See also Figure S3 and Movie S8.
Cell 143, 761–773, November 24, 2010 ª2010 Elsevier Inc. 769
Figure 6. B2AR Enrichment in Tubules Depends on Endosomal Actin and a PDZ-Interacting Sequence on the B2AR Cytoplasmic Domain (A) Representative fields from B2AR-expressing cells exposed to isoproterenol showing B2AR endosomes before (top panel) or after (bottom panel) exposure to 10 mM latrunculin for 5 min. Tubular endosomal domains enriched in B2AR (arrowheads) are lost upon exposure to latrunculin. (B) Schematic of measurement of endosomal B2AR fluorescence profiles in the limiting membrane. The profile was measured in a clockwise manner starting from the area diametrically opposite the tubule (an angle of 0 ). (C) B2AR concentration along the endosomal membrane, calculated from fluorescence profiles of 20 endosomes, normalized to the average endosomal B2AR fluorescence. In the presence of latrunculin, B2AR enrichment in tubules is abolished, and B2AR fluorescence shows little variation along the endosomal membrane. (D) Variance in endosomal B2AR fluorescence values measured before and after latrunculin. B2AR distribution becomes more uniform after latrunculin. (E) The percentages of endosomes extruding B2AR-containing tubules, calculated before (n = 246) and after (n = 106) treatment with latrunculin, or before (n = 141) and after (n = 168) cytochalasin-D, show a significant reduction after treatment with either drug. As a control, the percentages of endosomes extruding TfR-containing tubules before (n = 317) and after (n = 286), respectively, are shown. (F) Cortactin immunoblot showing reduction in protein levels after siRNA. (G) Representative fields from B2AR-containing endosomes in cells treated with control and cortactin siRNA. Arrowheads denote endosomal tubules in the control siRNA-treated cells. (H) Percentages of endosomes extruding B2AR tubules calculated in control siRNA-treated cells (n = 210), cortactin siRNA-treated cells (n = 269), and cortactin siRNA-treated cells expressing an siRNA-resistant cortactin (n = 250). (I) Representative examples of endosomes from agonist-exposed cells expressing B2AR, B2AR-ala, DOR, or DOR-B2. Arrowheads denote receptor-containing tubules on B2AR and DOR-B2 endosomes. (J) The percentage of endosomes with tubular domains containing B2AR, B2AR-ala, DOR, DOR-B2, or DOR-ABD (n = 246, 302, 137, 200, and 245, respectively) were quantified. Scale bars represent 1 mm; and error bars represent SEM. See also Figure S4.
770 Cell 143, 761–773, November 24, 2010 ª2010 Elsevier Inc.
concentration of the canonical Arp2/3 activators, WASP and WAVE, suggests a novel mode of actin nucleation involving cortactin. Cortactin can act as a nucleation-promoting factor for Arp2/3, at least in vitro (Ammer and Weed, 2008), and can interact with dynamin (Schafer et al., 2002; McNiven et al., 2000), which makes it an attractive candidate for coordinating actin dynamics on membranes. Interestingly, inhibition of WASH, a recently described Arp regulator that is present on B2AR tubules, has been reported to result in an increase in endosomal tubules (Derivery et al., 2009). Although its role in sequence-dependent recycling remains to be tested, this suggests the presence of multiple actin-associated proteins with distinct functions on the endosome. The simple kinetic and affinity-based principle that we propose likely provides a physical basis for sequence-dependent sorting of internalized membrane proteins between essentially opposite fates in distinct endosomal domains. Proteins that bind sequence-dependent degrading receptors and are required for their degradation (Whistler et al., 2002; Marley and von Zastrow, 2010) might act as scaffolds and provide a similar kinetic barrier to prevent them from accessing the rapid bulk-recycling tubules. Entry of these receptors into the involution pathway might then be accelerated by their association with the well-characterized ESCRT-associated domains on the vacuolar portion of endosomes (Hurley, 2008; Saksena et al., 2007; Williams and Urbe´, 2007), complementary to the presently identified ASSERT domains on a subset of endosomal tubules. Such diversity at the level of individual trafficking events to the same destination from the same organelle raises the possibility that there exists yet further specialization among the pathways that mediate exit out of the endosome, including in the degradative pathway and the retromer-based pathway to the trans-Golgi network. Importantly, the physical separation in pathways that we report here potentially allows for cargo-mediated regulation as a mode for controlling receptor recycling to the plasma membrane. Such a mechanism can provide virtually an unlimited level of selectivity in the post-endocytic system using minimal core trafficking machineries, as has been observed for endocytosis at the cell surface (Puthenveedu and von Zastrow, 2006). As the principles of such sorting depend critically on kinetics, the high-resolution imaging used here to analyze domain kinetics and biochemistry, and to achieve single-event resolution in living cells, provides a powerful method to elucidate biologically important sorting processes in the future. EXPERIMENTAL PROCEDURES Constructs and Reagents Receptor constructs and stably transfected HEK293 cell lines are described previously (Gage et al., 2005; Lauffer et al., 2009) Transfections were performed using Effectene (QIAGEN) according to manufacturer’s instructions. For visualizing receptors, FLAG-tagged receptors were labeled with M1 antibodies (Sigma) conjugated with Alexa-555 (Invitrogen) as described (Gage et al., 2005), or fusion constructs were generated where receptors were tagged on the N-terminus with GFP. Latrunculin and Cytochalasin D (Sigma) were used at 10 mM final concentration. Live-Cell and Fluorescence Imaging Cells were imaged using a Nikon TE-2000E inverted microscope with a 1003 1.49 NA TIRF objective (Nikon) and a Yokagawa CSU22 confocal head (Sola-
mere), or an Andor Revolution XD Spinning disk system on a Nikon Ti microscope. A 488 nm Ar laser and a 568 nm Ar/Kr laser (Melles Griot), or 488 nm and 561 nm solid-state lasers (Coherent) were used as light sources. Cells were imaged in Opti-MEM (GIBCO) with 2% serum and 30 mM HEPES (pH 7.4), maintained at 37 C using a temperature-controlled incubation chamber. Time lapse images were acquired with a Cascade II EM-CCD camera (Photometrics) driven by MicroManager (www.micro-manager.org) or an Andor iXon+ EM-CCD camera using iQ (Andor). The same lasers were used as sources for bleaching in FRAP experiments. Structured illumination microscopy was performed as described earlier (Gustafsson et al., 2008).
Electron Microscopy EM studies were carried out using MDCK cells because they are amenable to a previously described pre-embedding processing that facilitates detection of cytoplasmic actin filaments (Ikonen et al., 1996; Parton et al., 1991), and because they contain morphologically similar endosomes to HEK293 cells. Cells were grown on polycarbonate filters (Transwell 3412; Costar, Cambridge, MA) for 4 days as described previously (Parton et al., 1991). To allow visualization of early endosomes and any associated filaments a pre-embedding approach was employed. Cells were incubated with HRP (Sigma type II, 10mg/ml) in the apical and basolateral medium for 10min at 37 C and then washed, perforated, and immunogold labeled with a rabbit anti- actin antibody, a gift of Professor Jan de Mey (Strasbourg), followed by 9nm protein A-gold. HRP visualization and epon embedding was as described previously (Parton et al., 1991; Ikonen et al., 1996).
Image and Data Analysis Acquired image sequences were saved as 16-bit tiff stacks, and quantified using ImageJ (http://rsb.info.nih.gov/ij/). For estimating receptor enrichment, a circular mask 5 px in diameter was used to manually select the membrane at the base of the tubule or membranes derived from endosomes. Fluorescence values measured were normalized to that of the endosomal membrane devoid of tubules. An area of the coverslip lacking cells was used to estimate background fluorescence. For estimating linear pixel values along the tubules, a line selection was drawn along the tubule and across the endosome, and the Plot Profile function used to measure pixel values. For obtaining the average value plot across multiple sorting events, the linear pixels were first normalized to the diameter of the endosome and then averaged. To generate pixel values along the endosomal limiting membranes, the Oval Profile plugin, with 60 segments, was used after manually selecting the endosomal membrane using an oval ROI. Lifetimes of tubules were calculated by manually tracking the extension and retraction of tubules over time-lapse series. Microsoft Excel was used for simple data analyses and graphing. Curve fits of data were performed using GraphPad Prism. All P-values are from two-tailed Mann-Whitney tests unless otherwise noted.
SUPPLEMENTAL INFORMATION Supplemental Information includes four figures and eight movies and can be found with this article online at doi:10.1016/j.cell.2010.10.003.
ACKNOWLEDGMENTS The majority of the imaging was performed at the Nikon Imaging Center at UCSF. We thank David Drubin, Matt Welch, John Sedat, Aylin Hanyaloglu, Aaron Marley, and James Hislop for essential reagents and valuable help. M.A.P. was supported by a K99/R00 grant DA024698, M.v.Z. by an R37 grant DA010711, and O.D.W. by an RO1 grant GM084040, all from the NIH. J.T. is an investigator of the Howard Hughes Medical Institute. Received: October 31, 2009 Revised: April 7, 2010 Accepted: September 27, 2010 Published: November 24, 2010
Cell 143, 761–773, November 24, 2010 ª2010 Elsevier Inc. 771
REFERENCES Ammer, A.G., and Weed, S.A. (2008). Cortactin branches out: roles in regulating protrusive actin dynamics. Cell Motil. Cytoskeleton 65, 687–707. Brown, M.S., Anderson, R.G., and Goldstein, J.L. (1983). Recycling receptors: the round-trip itinerary of migrant membrane proteins. Cell 32, 663–667. Benink, H.A., and Bement, W.M. (2005). Concentric zones of active RhoA and Cdc42 around single cell wounds. J. Cell Biol. 168, 429–439. Cao, T.T., Deacon, H.W., Reczek, D., Bretscher, A., and von Zastrow, M. (1999). A kinase-regulated PDZ-domain interaction controls endocytic sorting of the b2-adrenergic receptor. Nature 401, 286–290. Cullen, P.J. (2008). Endosomal sorting and signalling: an emerging role for sorting nexins. Nat. Rev. Mol. Cell Biol. 9, 574–582. Daly, R.J. (2004). Cortactin signalling and dynamic actin networks. Biochem. J. 382, 13–25. Derivery, E., Sousa, C., Gautier, J.J., Lombard, B., Loew, D., and Gautreau, A. (2009). The Arp2/3 activator WASH controls the fission of endosomes through a large multiprotein complex. Dev. Cell 17, 712–723. Duleh, S.N., and Welch, M.D. (2010). WASH and the Arp2/3 complex regulate endosome shape and trafficking. Cytoskeleton (Hoboken) 67, 193–206. Dunn, K.W., and Maxfield, F.R. (1992). Delivery of ligands from sorting endosomes to late endosomes occurs by maturation of sorting endosomes. J. Cell Biol. 117, 301–310. Farquhar, M.G. (1983). Intracellular membrane traffic: pathways, carriers, and sorting devices. Methods Enzymol. 98, 1–13. Fehon, R.G., McClatchey, A.I., and Bretscher, A. (2010). Organizing the cell cortex: the role of ERM proteins. Nat. Rev. Mol. Cell Biol. 11, 276–287. Gage, R.M., Matveeva, E.A., Whiteheart, S.W., and von Zastrow, M. (2005). Type I PDZ ligands are sufficient to promote rapid recycling of G Proteincoupled receptors independent of binding to N-ethylmaleimide-sensitive factor. J. Biol. Chem. 280, 3305–3313. Girao, H., Geli, M.I., and Idrissi, F.Z. (2008). Actin in the endocytic pathway: from yeast to mammals. FEBS Lett. 582, 2112–2119. Gomez, T.S., and Billadeau, D.D. (2009). A FAM21-containing WASH complex regulates retromer-dependent sorting. Dev. Cell 17, 699–711. Gustafsson, M.G., Shao, L., Carlton, P.M., Wang, C.J., Golubovskaya, I.N., Cande, W.Z., Agard, D.A., and Sedat, J.W. (2008). Three-dimensional resolution doubling in wide-field fluorescence microscopy by structured illumination. Biophys. J. 94, 4957–4970. Hanyaloglu, A.C., and von Zastrow, M. (2007). A novel sorting sequence in the beta2-adrenergic receptor switches recycling from default to the Hrs-dependent mechanism. J. Biol. Chem. 282, 3095–3104. Hanyaloglu, A.C., and von Zastrow, M. (2008). Regulation of GPCRs by endocytic membrane trafficking and its potential implications. Annu. Rev. Pharmacol. Toxicol. 48, 537–568. Hanyaloglu, A.C., McCullagh, E., and von Zastrow, M. (2005). Essential role of Hrs in a recycling mechanism mediating functional resensitization of cell signaling. EMBO J. 24, 2265–2283. Hurley, J.H. (2008). ESCRT complexes and the biogenesis of multivesicular bodies. Curr. Opin. Cell Biol. 20, 4–11. Ikonen, E., Parton, R.G., Lafont, F., and Simons, K. (1996). Analysis of the role of p200-containing vesicles in post-Golgi traffic. Mol. Biol. Cell 7, 961–974. Kaksonen, M., Peng, H.B., and Rauvala, H. (2000). Association of cortactin with dynamic actin in lamellipodia and on endosomal vesicles. J. Cell Sci. 113, 4421–4426.
Lefkowitz, R.J., Pitcher, J., Krueger, K., and Daaka, Y. (1998). Mechanisms of beta-adrenergic receptor desensitization and resensitization. Adv. Pharmacol. 42, 416–420. Marchese, A., Paing, M.M., Temple, B.R., and Trejo, J. (2008). G proteincoupled receptor sorting to endosomes and lysosomes. Annu. Rev. Pharmacol. Toxicol. 48, 601–629. Marley, A., and von Zastrow, M. (2010). Dysbindin promotes the post-endocytic sorting of G protein-coupled receptors to lysosomes. PLoS ONE 5, e9325. Maxfield, F.R., and McGraw, T.E. (2004). Endocytic recycling. Nat. Rev. Mol. Cell Biol. 5, 121–132. Mayor, S., Presley, J.F., and Maxfield, F.R. (1993). Sorting of membrane components from endosomes and subsequent recycling to the cell surface occurs by a bulk flow process. J. Cell Biol. 121, 1257–1269. McNiven, M.A., Kim, L., Krueger, E.W., Orth, J.D., Cao, H., and Wong, T.W. (2000). Regulated interactions between dynamin and the actin-binding protein cortactin modulate cell shape. J. Cell Biol. 151, 187–198. Merrifield, C.J., Qualmann, B., Kessels, M.M., and Almers, W. (2004). Neural Wiskott Aldrich Syndrome Protein (N-WASP) and the Arp2/3 complex are recruited to sites of clathrin-mediated endocytosis in cultured fibroblasts. Eur. J. Cell Biol. 83, 13–18. Miesenbo¨ck, G., De Angelis, D.A., and Rothman, J.E. (1998). Visualizing secretion and synaptic transmission with pH-sensitive green fluorescent proteins. Nature 394, 192–195. Millard, T.H., Sharp, S.J., and Machesky, L.M. (2004). Signalling to actin assembly via the WASP (Wiskott-Aldrich syndrome protein)-family proteins and the Arp2/3 complex. Biochem. J. 380, 1–17. Parton, R.G., Dotti, C.G., Bacallao, R., Kurtz, I., Simons, K., and Prydz, K. (1991). pH-induced microtubule-dependent redistribution of late endosomes in neuronal and epithelial cells. J. Cell Biol. 113, 261–274. Perrais, D., and Merrifield, C.J. (2005). Dynamics of endocytic vesicle creation. Dev. Cell 9, 581–592. Piper, R.C., and Katzmann, D.J. (2007). Biogenesis and function of multivesicular bodies. Annu. Rev. Cell Dev. Biol. 23, 519–547. Pippig, S., Andexinger, S., and Lohse, M.J. (1995). Sequestration and recycling of beta 2-adrenergic receptors permit receptor resensitization. Mol. Pharmacol. 47, 666–676. Pollard, T.D. (2007). Regulation of actin filament assembly by Arp2/3 complex and formins. Annu. Rev. Biophys. Biomol. Struct. 36, 451–477. Puthenveedu, M.A., and von Zastrow, M. (2006). Cargo regulates clathrincoated pit dynamics. Cell 127, 113–124. Saksena, S., Sun, J., Chu, T., and Emr, S.D. (2007). ESCRTing proteins in the endocytic pathway. Trends Biochem. Sci. 32, 561–573. Schafer, D.A., Weed, S.A., Binns, D., Karginov, A.V., Parsons, J.T., and Cooper, J.A. (2002). Dynamin2 and cortactin regulate actin assembly and filament organization. Curr. Biol. 12, 1852–1857. Scita, G., and Di Fiore, P.P. (2010). The endocytic matrix. Nature 463, 464–473. Shinozaki-Narikawa, N., Kodama, T., and Shibasaki, Y. (2006). Cooperation of phosphoinositides and BAR domain proteins in endosomal tubulation. Traffic 7, 1539–1550. Sorkin, A., and von Zastrow, M. (2009). Endocytosis and signalling: intertwining molecular networks. Nat. Rev. Mol. Cell Biol. 10, 609–622. Stamnes, M. (2002). Regulating the actin cytoskeleton during vesicular transport. Curr. Opin. Cell Biol. 14, 428–433.
Kaksonen, M., Toret, C.P., and Drubin, D.G. (2005). A modular design for the clathrin- and actin-mediated endocytosis machinery. Cell 123, 305–320.
Steinman, R.M., Mellman, I.S., Muller, W.A., and Cohn, Z.A. (1983). Endocytosis and the recycling of plasma membrane. J. Cell Biol. 96, 1–27.
Lauffer, B.E., Chen, S., Melero, C., Kortemme, T., von Zastrow, M., and Vargas, G.A. (2009). Engineered protein connectivity to actin mimics PDZ-dependent recycling of G protein-coupled receptors but not its regulation by Hrs. J. Biol. Chem. 284, 2448–2458.
Traer, C.J., Rutherford, A.C., Palmer, K.J., Wassmer, T., Oakley, J., Attar, N., Carlton, J.G., Kremerskothen, J., Stephens, D.J., and Cullen, P.J. (2007). SNX4 coordinates endosomal sorting of TfnR with dynein-mediated transport into the endocytic recycling compartment. Nat. Cell Biol. 9, 1370–1380.
772 Cell 143, 761–773, November 24, 2010 ª2010 Elsevier Inc.
Turunen, O., Wahlstro¨m, T., and Vaheri, A. (1994). Ezrin has a COOH-terminal actin-binding site that is conserved in the ezrin protein family. J. Cell Biol. 126, 1445–1453. Uetrecht, A.C., and Bear, J.E. (2006). Coronins: the return of the crown. Trends Cell Biol. 16, 421–426. Weinman, E.J., Hall, R.A., Friedman, P.A., Liu-Chen, L.Y., and Shenolikar, S. (2006). The association of NHERF adaptor proteins with g protein-coupled receptors and receptor tyrosine kinases. Annu. Rev. Physiol. 68, 491–505. Wheeler, D., Sneddon, W.B., Wang, B., Friedman, P.A., and Romero, G. (2007). NHERF-1 and the cytoskeleton regulate the traffic and membrane dynamics of G protein-coupled receptors. J. Biol. Chem. 282, 25076–25087. Whistler, J.L., Enquist, J., Marley, A., Fong, J., Gladher, F., Tsuruda, P., Murray, S.R., and Von Zastrow, M. (2002). Modulation of postendocytic sorting of G protein-coupled receptors. Science 297, 615–620.
Williams, R.L., and Urbe´, S. (2007). The emerging shape of the ESCRT machinery. Nat. Rev. Mol. Cell Biol. 8, 355–368. Xiang, Y., and Kobilka, B.K. (2003). Myocyte adrenoceptor signaling pathways. Science 300, 1530–1532. Yarar, D., Waterman-Storer, C.M., and Schmid, S.L. (2005). A dynamic actin cytoskeleton functions at multiple stages of clathrin-mediated endocytosis. Mol. Biol. Cell 16, 964–975. Yudowski, G.A., Puthenveedu, M.A., and von Zastrow, M. (2006). Distinct modes of regulated receptor insertion to the somatodendritic plasma membrane. Nat. Neurosci. 9, 622–627. Yudowski, G.A., Puthenveedu, M.A., Henry, A.G., and von Zastrow, M. (2009). Cargo-mediated regulation of a rapid Rab4-dependent recycling pathway. Mol. Biol. Cell 20, 2774–2784. Zerial, M., and McBride, H. (2001). Rab proteins as membrane organizers. Nat. Rev. Mol. Cell Biol. 2, 107–117.
Cell 143, 761–773, November 24, 2010 ª2010 Elsevier Inc. 773
Mechanisms Determining the Morphology of the Peripheral ER Yoko Shibata,1,2 Tom Shemesh,3 William A. Prinz,4 Alexander F. Palazzo,1,5 Michael M. Kozlov,3,* and Tom A. Rapoport1,2,* 1Howard
Hughes Medical Institute of Cell Biology Harvard Medical School, 240 Longwood Avenue, Boston, MA 02115, USA 3Department of Physiology and Pharmacology, Sackler Faculty of Medicine, Tel Aviv University, Ramat Aviv, 69978 Tel Aviv, Israel 4Laboratory of Cell Biochemistry and Biology, National Institute of Diabetes and Digestive and Kidney Disorders, National Institute of Health, Bethesda, MD 02892, USA 5Department of Biochemistry, University of Toronto, 1 King’s College Circle, Toronto, ON M5S 1A8, Canada *Correspondence:
[email protected] (M.M.K.),
[email protected] (T.A.R.) DOI 10.1016/j.cell.2010.11.007 2Department
SUMMARY
The endoplasmic reticulum (ER) consists of the nuclear envelope and a peripheral network of tubules and membrane sheets. The tubules are shaped by the curvature-stabilizing proteins reticulons and DP1/Yop1p, but how the sheets are formed is unclear. Here, we identify several sheet-enriched membrane proteins in the mammalian ER, including proteins that translocate and modify newly synthesized polypeptides, as well as coiled-coil membrane proteins that are highly upregulated in cells with proliferated ER sheets, all of which are localized by membrane-bound polysomes. These results indicate that sheets and tubules correspond to rough and smooth ER, respectively. One of the coiled-coil proteins, Climp63, serves as a ‘‘luminal ER spacer’’ and forms sheets when overexpressed. More universally, however, sheet formation appears to involve the reticulons and DP1/Yop1p, which localize to sheet edges and whose abundance determines the ratio of sheets to tubules. These proteins may generate sheets by stabilizing the high curvature of edges. INTRODUCTION How the characteristic shape of a membrane-bound organelle is generated is a fundamental question in cell biology. We have started to address this question for the endoplasmic reticulum (ER), an organelle that has a particularly intriguing morphology. It is a continuous membrane system that is comprised of the nuclear envelope as well as of a peripheral network of tubules and sheets (Baumann and Walz, 2001; Shibata et al., 2009; Voeltz et al., 2002). Both the tubules and sheets are dynamic, i.e., they are continuously forming and collapsing. Previous work has identified proteins that are responsible for shaping 774 Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc.
the tubular ER network (Hu et al., 2008, 2009; Shibata et al., 2008; Voeltz et al., 2006), but essentially nothing is known about how ER sheets are generated. In addition, it is unknown whether proteins specifically segregate into ER sheets and whether there is a functional significance to the existence of different ER morphologies. ER tubules are characterized by high membrane curvature in cross-section and are shaped by two families of curvature-stabilizing proteins, the reticulons and DP1/Yop1p (Voeltz et al., 2006). Members of both families are ubiquitously expressed in all eukaryotic cells. These proteins localize to the ER tubules, and their depletion leads to the loss of tubules. Conversely, the overexpression of certain isoforms results in long, unbranched tubules. Purified members of the two families deform reconstituted proteoliposomes into tubules (Hu et al., 2008). Together, these results indicate that the reticulons and DP1/Yop1p are both necessary and sufficient for ER tubule formation. These two protein families do not share sequence homology, but both have a conserved domain containing two long hydrophobic segments that sit in the membrane as hairpins (Voeltz et al., 2006). These hairpins may stabilize the high curvature of tubules in cross-section by forming a wedge in the lipid bilayer. In addition, oligomerization of these proteins may generate arc-like scaffolds around the tubules (Shibata et al., 2008). The peripheral ER sheets vary in size but always consist of two closely apposed membranes whose distance is approximately the same as the diameter of the tubules (30 nm in yeast [Bernales et al., 2006] and 50 nm in mammals). Consequently, the edges of sheets have a similarly high curvature as the crosssection of tubules. In ‘‘professional’’ secretory cells, such as plasma B cells or pancreatic cells, the ER sheets extend throughout the entire cell and are studded with membranebound ribosomes. They are stacked tightly with regular distances between the membranes on both the cytoplasmic and luminal sides (Fawcett, 1981). By contrast, cells that do not secrete many proteins contain mostly tubular ER. These observations have led to the idea that ER sheets correspond to rough ER (Shibata et al., 2006), the region of the ER that contains membrane-bound ribosomes, i.e., ribosomes associated with
the translocons, the sites of translocation and modification of newly synthesized secretory and membrane proteins. On the other hand, ER tubules would correspond to smooth ER (Shibata et al., 2006), the ER region devoid of ribosomes, which may be specialized in lipid metabolism or Ca2+ signaling. While these ideas are attractive, the tubular ER clearly contains membranebound ribosomes, and a segregation of rough ER proteins into sheets has not yet been demonstrated. Several mechanisms of ER sheet formation have been considered. One possibility is that integral membrane proteins would form bridges across the luminal space of the ER (Senda and Yoshinaga-Hirabayashi, 1998; Shibata et al., 2009). A second possibility is that proteins form flat cytoplasmic or luminal scaffolds, as suggested for the formation of flat Golgi cisternae (Short et al., 2005). It has also been proposed that the membrane association of ribosomes could directly be responsible for the generation of ER sheets (Puhka et al., 2007). Finally, given that the reticulons and DP1/Yop1p generate high curvature membranes, one might imagine that they generate sheets by stabilizing the sheet edges, bringing the apposing membranes in close proximity (Shibata et al., 2009). Here, we show that rough ER proteins partition into ER sheets. This includes both proteins involved in translocation and modification of newly synthesized polypeptides, as well as coiled-coil membrane proteins that are highly upregulated in cells containing proliferated ER sheets. Membrane-bound polysomes are required for the segregation of these rough ER proteins into sheets, and one of the coiled-coil proteins, Climp63, serves as a luminal ER spacer. However, neither the polysomes nor the coiled-coil proteins are essential for sheet formation per se. Instead, a major mechanism of sheet formation appears to involve the reticulons and DP1/Yop1p proteins, which can stabilize the high membrane curvature at sheet edges. Our results suggest that, in many cells, their abundance is the major determinant of ER morphology. RESULTS Segregation of Proteins into ER Sheets The different morphologies of the ER imply that, despite the continuity of the membrane system, some proteins are likely enriched in certain domains. So far, the only proteins known with a specific localization are the tubule-preferring reticulons, DP1/Yop1p, and atlastins/Sey1p (Hu et al., 2009; Shibata et al., 2008; Voeltz et al., 2006). These proteins localize to tubules even when highly overexpressed. By contrast, other overexpressed ER proteins distribute indiscriminately throughout the entire ER, making it impossible to draw conclusions about their endogenous localizations. We therefore first tested whether several endogenous ER proteins segregate into different ER domains using immunofluorescence and confocal microscopy in BSC1 cells. As expected, the luminal ER protein calreticulin, which is involved in the folding of glycoproteins, was found in peripheral ER sheets, which are mostly located close to the nucleus, as well as in the tubular ER network and the nuclear envelope (Figure 1A). Calreticulin almost perfectly colocalized with GFP-tagged Sec61b, stably overexpressed in the same cell. Endogenous Sec61b is part of the Sec61 complex, the
component forming the protein-conducting channel in the ER, but due to its tagging with GFP and overexpression, GFP-Sec61b is not associated with the translocon and distributes throughout the ER (Shibata et al., 2008). Antibodies recognizing the luminal chaperones BiP and Grp94 (anti-KDEL) also stained the entire ER (Figure 1C, middle). The integral membrane proteins calnexin and Bap31 showed a similar ubiquitous localization as overexpressed GFP-Sec61b (Figure 1B and Figure S1 available online). These results suggest that many luminal and membrane ER proteins do not localize to a specific ER domain, consistent with the continuity of the membrane system. Next, we tested the endogenous localization of components of the translocon. In contrast to overexpressed GFP-Sec61b, endogenous Sec61b was found concentrated in ER sheets when compared to the localization of the luminal ER proteins BiP and GRP94 (Figure 1C), although some weak staining of the tubular network and nonspecific staining of the cytoplasm were also seen. Because endogenous Sec61b is contained in the Sec61 complex, these data suggest that translocons are enriched in ER sheets. This is supported by the localization of endogenous TRAPa, a component tightly associated with the ribosome-bound Sec61 complex (Me´ne´tret et al., 2008); TRAPa was strongly enriched in the peripheral ER sheets (Figure 1D). Finally, Dad1, a component of the translocon-associated oligosaccharyl transferase complex that glycosylates nascent secretory and membrane proteins, also showed a similar localization; GFP-tagged Dad1 that was stably expressed in Dad1deficient cells at a level just sufficient to sustain viability (Nikonov et al., 2002) showed a clear preference for ER sheets, in contrast to calreticulin in the same cell (Figure 1E). Together, these data indicate that translocon components are enriched in ER sheets. To identify additional sheet-segregating proteins that could potentially be required for sheet formation, we reasoned that such proteins would be abundant in highly secretory cells that contain proliferated ER sheets. We therefore identified by mass spectrometry the most abundant, integral ER membrane proteins in dog pancreatic rough microsomes. The 25 most abundant proteins include translocon components, such as subunits of the oligosaccharyl transferase complex, signal peptidase, SRP receptor, components of the TRAP complex, and the Sec61 complex (Table S1). Of interest, the list also includes p180 and Climp63. Kinectin, which is sequence related to p180, is somewhat less abundant. All of these proteins have a single transmembrane segment and an extended coiled-coil domain, which is located on the luminal side of the ER membrane in the case of Climp63 and on the cytoplasmic side in the case of p180 and kinectin (Figure S2A). The molecular function of these coiled-coil proteins is not well understood. Climp63 has been implicated in the interaction of ER membranes with microtubules (Klopfenstein et al., 1998). P180 was originally proposed to be a ribosome receptor (Savitz and Meyer, 1990); it also interacts with microtubules (Ogawa-Goto et al., 2007) and is now thought to play a role in the differentiation of certain monocytic cells (Benyamini et al., 2009). Kinectin was initially identified as a receptor for the molecular motor kinesin (Toyoshima et al., 1992). Another way to identify potential sheet-segregating proteins is to analyze components that are upregulated during the differentiation of immature B cells to IgG-secreting plasma cells, which Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc. 775
Figure 1. Localization of Proteins to Different ER Domains (A) The localization of endogenous luminal ER protein calreticulin is compared with that of the stably overexpressed membrane protein GFP-Sec61b using confocal microscopy in BSC1 cells. Calreticulin was detected with specific antibodies by indirect immunofluorescence (left) and Sec61b by GFP fluorescence (middle). The right panel shows a merged image. Scale bar, 10 mm. (B) As in (A) but comparing the localization of the ER membrane protein calnexin with that of GFP-Sec61b. (C) The localization of endogenous Sec61b is compared to that of the endogenous ER luminal proteins BiP and GRP94 (anti-KDEL), using indirect immunofluorescence with specific antibodies and confocal microscopy. (D) As in (A) but comparing the localization of the translocon membrane protein TRAPa with that of GFP-Sec61b. Also note that TRAPa is noticeably depleted from the nuclear envelope. (E) The localization of stably expressed GFP-Dad1 in a BHK cell line lacking endogenous Dad1 is compared with that of endogenous calreticulin. All insets show a magnified view of the boxed areas highlighting the junctions between ER sheets and tubules. See also Figure S1.
involves massive ER sheet proliferation. To identify mRNAs whose abundance is greatly increased, we sorted through published microarray data (Luckey et al., 2006). The list of the 776 Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc.
25 most upregulated mRNAs coding for ER membrane proteins (Table S2) includes components of the translocon, of the unfolded protein response, and of the ER protein degradation
machinery. It also includes Climp63 and p180 (their mRNAs are upregulated by a factor of 19–26; kinectin mRNA was not analyzed). Together with the mass spectrometry data, these results raise the possibility that the coiled-coil membrane proteins Climp63, p180, and kinectin localize to ER sheets. Because these proteins have no known function in protein translocation or modification, they are also candidates for being involved in sheet formation. Next, we tested whether the coiled-coil proteins are enriched in ER sheets, using immunofluorescence and confocal microscopy. At endogenous levels, all three proteins indeed segregated to ER sheets, whereas in the same cells, calreticulin distributed throughout the entire ER (Figures 2A–2C). P180-GFP overexpressed at moderate levels was also enriched in ER sheets (Figure S2B). Thus, in addition to the translocon proteins, at least three other abundant integral membrane proteins are enriched in ER sheets. All three proteins were noticeably depleted from the nuclear envelope (Figure 2 and Figure S2), as reported previously for Climp63 (Klopfenstein et al., 1998). A Role for Polysomes in Protein Enrichment in ER Sheets Because translocon-associated proteins were found enriched in ER sheets and are also generally associated with ribosomes, we tested whether the sheet-preferring proteins are localized by their association with membrane-bound translating ribosomes. We treated tissue culture cells with puromycin, a drug that releases nascent polypeptide chains from ribosomes and disassembles polysomes; the localization of endogenous sheet-preferring proteins was subsequently analyzed by immunofluorescence. TRAPa moved into the tubular network (Figure 3A). Quantification shows that, in untreated cells, TRAPa is enriched in sheets, as compared to the general ER marker GFP-Sec61b, but 15 min after puromycin addition, TRAPa was almost equally abundant in sheets and tubules (Figure 3E). The disassembly of the polysomes did not abolish the ER sheets, which in fact occupied a larger surface in many cells (Figure S3A). To rule out the possibility that inhibition of translation causes the redistribution of TRAPa, we performed control experiments with cycloheximide, a drug that inhibits the elongation of polypeptide chains but leaves polysomes intact. Cycloheximide inhibited protein synthesis as effectively as puromycin (Figure S4), but TRAPa stayed in ER sheets (Figures 3B and 3E). All of the other tested ER sheet-preferring proteins behaved in the same way as TRAPa (Figures 3C–3E). On the other hand, the localization of calnexin and Bap31, membrane proteins that did not segregate into ER sheets, remained unchanged after treatment with either puromycin or cycloheximide, as was also the case for the luminal protein calreticulin (Figure 3E and Figures S3B and S3C). Pactamycin, an inhibitor of translation initiation, which allows ribosomes to run off the mRNAs, had a similar effect as puromycin on sheet-segregating proteins, i.e., they were no longer concentrated in sheets (Figure S3D). Again, the sheets did not disappear but often occupied a larger area of the cell (Figure S3E). These results indicate that polysomes concentrate sheet domains and localize certain membrane proteins to ER sheets, likely because these proteins have a direct or indirect affinity for membrane-bound polysomes.
Climp63 Serves as a ‘‘Luminal ER Spacer’’ To test for a possible role of the coiled-coil membrane proteins in ER morphology, we performed RNAi experiments. The depletion of Climp63, p180, and kinectin (Figure S5A) either individually or together did not abolish the existence of ER sheets (Figure 2D versus 2E). Nevertheless, these proteins have an effect on ER morphology, as the sheets in depleted cells spread throughout the cytoplasm (Figures S5B and S5C), similarly to what is observed when cells are treated with puromycin or pactamycin (Figure S3). It thus seems that the coiled-coil membrane proteins are not required for sheet formation per se may but function in segregating sheet domains close to the cell nucleus. Thin-section electron microscopy of COS7 cells confirmed that peripheral ER sheets persist after puromycin treatment or depletion of Climp63, p180, and kinectin (compare Figures 4B and 4C with 4A). No bulging of the two membrane sheets was observed, but of interest, the luminal width was significantly reduced in triple knockdown cells (from 45–50 nm to 25–30 nm; Figure 4E). A similar effect was seen when Climp63 alone was depleted (Figures 4D and 4E), whereas single or double knockdown of p180 and kinectin had no obvious phenotype (Figure 4E and data not shown). These results indicate that Climp63 serves to maintain the normal luminal width of peripheral ER sheets, likely by forming bridges through their luminal coiled-coil domains (Klopfenstein et al., 2001). Consistent with a luminal spacer function, organisms that lack Climp63, including Drosophila S2 cells (Figure 4E), silkworm (Senda and Yoshinaga-Hirabayashi, 1998), and S. cerevisiae (Bernales et al., 2006), all appear to have narrower ER sheets than mammals. It should be noted that the distance between the inner and outer nuclear membranes was unaffected by Climp63 depletion and was the same in mammalian and insect cells (Figure 4E), consistent with the absence of this protein from the nuclear envelope. Linking the Formation of ER Sheets and Tubules The overexpression of Climp63 led to a dramatic proliferation of ER sheets; we observed a good correlation between the expression level of a FLAG-tagged version of Climp63 in COS7 cells and the generation of sheets, an effect that is most strikingly seen in three-dimensional (3D) reconstructions of the ER (Figures 5A and 5B; quantification in 5C). In thin-section electron microscopy, prominent membrane structures were seen that consisted of anastomosing sheets containing membrane-bound ribosomes (Figure 5D). The sheets had a constant luminal width of 50 nm, and at the highest expression levels, the luminal protein calreticulin was displaced from areas of Climp63 localization (Figure S6), consistent with Climp63 filling the luminal space. We also observed organized smooth ER (OSER) structures in which the membranes were tightly stacked and the internal membranes were devoid of ribosomes (Figure S7). Although these structures are likely caused by oligomerization of the cytoplasmic GFP tag (Snapp et al., 2003), they differ from normal OSERs by having a constant luminal spacing of 50 nm. Given that ER sheet proliferation was also observed when the curvature-stabilizing reticulons are depleted in mammalian cells by RNAi (Anderson and Hetzer, 2008) or when the reticulons and Yop1p are lacking in S. cerevisiae (Voeltz et al., 2006), we tested whether Climp63 and the reticulons have opposing effects on ER Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc. 777
Figure 2. Membrane Proteins Enriched in ER Sheets (A) The endogenous localization of the membrane protein Climp63 is compared with that of the luminal ER protein calreticulin in COS7 cells, using indirect immunofluorescence with specific antibodies. The far-right panel shows a merged image. Junctions between peripheral ER sheets and tubules are highlighted in the magnified view of the boxed area (inset). Scale bar, 10 mm. (B) As in (A) but comparing the localization of kinectin (KTN) and calreticulin. (C) As in (A) but comparing the localization of p180 and calreticulin. (D) Climp63, p180, and kinectin were depleted in COS7 cells by RNAi (C/P/K siRNA), and Climp63, TRAPa, and calreticulin were visualized using indirect immunofluorescence with specific antibodies. Scale bar, 10 mm. (E) As in (D) but with cells transfected with control siRNA oligonucleotides. See also Figure S2 and Figure S5.
sheet formation. Indeed, when the reticulon Rtn4b was overexpressed in COS7 cells, peripheral ER sheets became diminished with increasing expression levels (Figures 5E and 5F; quantifica778 Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc.
tion in Figures 5G and 5H). Concomitant with the decrease in sheet structures, the normal tubular network was gradually replaced with long, unbranched tubules (quantification in Figure 5I).
When Climp63 and Rtn4a were both highly overexpressed, the normal ER morphology was almost restored (Figure 5J). Taken together, these results indicate that Climp63 and the curvature-promoting proteins undergo a ‘‘tug-of-war’’ that determines the amount of membrane partitioning into these domains. Curvature-Stabilizing Proteins Localize to Sheet Edges Because the reticulons and DP1/Yop1p localize to tubules, one might expect that they are also found at sheets edges because these have a similarly high membrane curvature as tubules in cross-section. Indeed, in many cells, the endogenous reticulons localized to the edges of sheets, as demonstrated by immunofluorescence using antibodies recognizing both Rtn4a and 4b (Figure 6A). Similar observations were made in plant cells (Sparkes et al., 2010). In Climp63-overexpressing cells with proliferated sheets, Rtn4a/b lined the edges of essentially all sheets in an even more striking manner (Figure 6B). To test whether the curvature-stabilizing proteins generally localize to sheet edges, we tested the localization of a reticulon in S. cerevisiae. We expressed Rtn1p-GFP from the chromosome together with ssRFP-HDEL, a general, luminal ER marker. Indeed, peripheral ER sheets were generally lined by Rtn1p-GFP (Figure 6C). The edge localization of Rtn1p-GFP was even more obvious in cells where ER sheet proliferation was induced by deletion of the genes encoding the tubule-shaping protein Yop1p and the GTPase Sey1p (Figure 6D) (Hu et al., 2009). Similar results were obtained when ER sheets were induced by deletion of OPI1 (Figure 6E) (Schuck et al., 2009). Thus, as in mammalian cells, the reticulons localize to the edges of peripheral ER sheets. These results indicate that the reticulons stabilize the high curvature of both tubules in cross-section and of sheet edges. A Role for Curvature-Stabilizing Proteins in Sheet Formation Given the localization of the curvature-generating proteins to sheet edges, we considered the possibility that they can generate sheets by bringing the apposing membranes into close proximity. In this model, the ratio of sheets and tubules would be determined by the relative amounts of lipids and curvaturestabilizing proteins. Indeed, the sheet proliferation seen upon OPI1 deletion in S. cerevisiae (Figure 6E) is likely caused by an increase in phospholipid synthesis; Opi1p normally inhibits the transcription factors Ino2p and Ino4p, which control many phospholipid synthesis enzymes (Ambroziak and Henry, 1994; Carman and Henry, 2007). To test whether expression of a curvature-stabilizing protein would convert the sheets into tubules, we used opi1D cells that express Rtn1p-GFP from the chromosome as well as the luminal ER marker ssRFP-HDEL. The overexpression of untagged Rtn1p from a CEN plasmid led to a partial conversion of sheets into tubules (Figure 6F; quantification in Figure 6H). When untagged Rtn1p was expressed at a still higher level from a 2 m plasmid, the sheet-to-tubule ratio converted back to about the level seen in wild-type cells (Figures 6G and 6H). These data support the idea that the abundance of the reticulons determines the relative amounts of sheets and tubules in the cell.
A Model for the Generation of ER Sheets and Tubules To test whether the curvature-stabilizing proteins alone could explain the relative amounts of sheets and tubules in a cell, we developed a simple theoretical model. We assume that the reticulons and DP1/Yop1p localize exclusively to tubules and sheet edges, generating and stabilizing these high curvature membranes by forming oligomeric scaffolds that are shaped as rigid arcs. Based on previous estimates, the energetically optimal distance between the arcs is assumed to be 40 nm (Hu et al., 2008). The edge membrane can be seen as a half-cylinder, whose axis bends in the sheet plane forming the sheet circumference. The protein-driven formation of a sheet edge enables the two membranes of a sheet to adopt planar shapes (Figure 7A). A tubule forms by self-folding of a part of the edge into a complete cylinder and therefore represents an edge extension (Figure 7A). We assume negligible bulging between the arc-like scaffolds, as supported by previous results (Hu et al., 2008), and a diameter of 30 nm for both sheet edges and tubules (Figure 4) (Bernales et al., 2006). Our model calculates for a given membrane surface area the total length of the tubules and the shape and dimensions of the sheets in dependence of the number of curvature-stabilizing proteins, Nc. We characterize the edge length by a parameter G = Le/ Le0, wherein Le0 is the circumference of a flat circular disc with the same overall membrane area (i.e., G = 1 for a flat disc). G is proportional to Nc (Supplemental Information). In our calculations, we assume that Nc is at least large enough to generate a circular sheet (G R 1). For each G value, we computed the overall membrane shape by minimizing the energy of the edge bending in the sheet plane (see Experimental Procedures and Supplemental Information). The top view of the shapes is presented in Figure 7B. Starting from the circular disc configuration at G = 1 (Figure 7B, blue line), the sheet shape elongates with increasing G (and Nc) (light blue line) and then acquires a flattened dumbbell appearance with a narrowing neck (aqua and yellow lines) and, finally, at G2, splits into two droplet-like sheets with a short tubule between them (orange line). Further increase of G results in tubule elongation and a decrease in the sizes of the two sheets (Figure 7B, dark red line). Eventually, the whole membrane converts into a tubule (not shown in Figure 7B). Thus, the curvature-producing proteins alone can generate both sheets, and tubules and their abundance determines the relative amounts of the two membrane domains. Next, we extended the model to include the effect of proteins enriched in sheets. We assume that polysome-bound Climp63, p180, and kinectin, as well as translocons, can diffuse throughout the sheets but cannot move into high curvature membrane areas, i.e., sheet edges and tubules. The number of all of these ‘‘sheet proteins’’ together is denoted by Ns. The sheet proteins may be considered as generating an ‘‘osmotic pressure’’ on the sheet edges, a force that resists the shrinkage of a sheet domain. The magnitude of this effect is determined by the interplay between the effective ‘‘osmotic pressure’’ produced by the sheet proteins and the effective stretching elasticity of the edge, the latter being determined by the curvature-stabilizing proteins (see Experimental Procedures and Supplemental Information). Our model does not take into Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc. 779
Figure 3. Polysome-Dependent Membrane Protein Enrichment in ER Sheets (A) The localization of the translocon component TRAPa is compared with that of stably expressed GFP-Sec61b after 15 min of treatment with puromycin (PURO). The far-right panel shows a merged image. Junctions between peripheral ER sheets and tubules are highlighted in the magnified view of the boxed area (inset). Scale bar, 10 mm. (B) As in (A) but after 15 min of treatment with cycloheximide (CHX). (C) As in (A) but comparing the localization of Climp63 with calreticulin after puromycin treatment. (D) As in (C) but after cycloheximide treatment. (E) Quantification of sheet enrichment of different ER proteins in untreated cells (blue bars) and in cells treated with puromycin (PURO; green) or cycloheximide (CHX; red). The ratio of the average fluorescence intensity in sheets versus tubules was determined for calnexin (CNX), Bap31, calreticulin (CRT), TRAPa, and kinectin and was divided by the sheet-to-tubule fluorescence ratio for stably expressed GFP-Sec61b, a protein that shows no preference for either ER domain.
780 Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc.
Figure 4. Climp63 Affects the Luminal Width of Peripheral ER Sheets (A) Rough ER sheets in a COS7 cell visualized by thin-section electron microscopy. Scale bar, 0.5 mm. (B) As in (A) but after treatment with puromycin (PURO) for 15 min. (C) As in (A) but after RNAi-depletion of Climp63, p180, and kinectin (C/P/K siRNA). (D) As in (A) but after RNAi-depletion of Climp63. (E) Quantification of the luminal width of peripheral ER sheets and the nuclear envelope (NE) in differently treated COS7 cells. For comparison, Drosophila S2R+ cells were also analyzed. Shown are the means and standard errors of n cells analyzed for each sample. Kinectin, KTN.
account that Climp63 affects sheet formation by serving as a luminal spacer, and it does not make any assumptions about the specific roles of p180 and kinectin. We computed the G values and membrane configurations for different values of Nc and Ns (Figure 7C). The colored lines on the bottom plane of the diagram represent the relationship between Nc and Ns for a given shape of the system, with the colors corresponding to the shapes as in Figure 7B. Figure 7C demonstrates that an increase of Nc at a given Ns results in larger G (blue to red transition) and thus in more tubules, whereas an increase of Ns at a given Nc results in lower G, i.e., more sheets. This is further illustrated in Figure 7D, in which the increase of
Ns at a constant Nc converts two small sheet areas connected by a narrow tubule into a larger sheet area. Thus, the model recapitulates the experimental observation of a tug-of-war between sheet-promoting Climp63 and curvature-stabilizing proteins. DISCUSSION Our results indicate that several mechanisms shape peripheral ER sheets. The most basic and universal mechanism appears to involve the previously identified curvature-stabilizing proteins, the reticulons and DP1/Yop1p. These proteins would stabilize not only the high curvature of narrow tubules, but also the
A similar analysis was done for GFP-Dad1 and Climp63 but with calreticulin as reference. Shown are the means and standard errors of data obtained from 7 to 30 cells for each condition. See also Figure S3 and Figure S4.
Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc. 781
Figure 5. Climp63 and Reticulon Overexpression Change the Abundance of Sheets and Tubules (A) FLAG-Climp63 overexpressed at relatively high levels in a COS7 cell was visualized by indirect immunofluorescence using FLAG antibodies. A 3D image was generated from a complete series of z sections (step size 0.25 mm) taken with a confocal microscope. Scale bar, 10 mm. (B) As in (A) but in a cell expressing FLAG-Climp63 at the highest observed levels. (C) Quantification of the effect of Climp63 overexpression on ER sheet abundance. Shown are the percentages of cells with normal reticular ER (blue bars), of cells with both large sheets and reticular ER (red), and of cells with large ER sheets lacking reticular ER (green) at different expression levels of FLAG-Climp63. The cells were divided into five groups according to their expression levels, as determined by overall average fluorescence intensity. (D) Thin-section electron micrograph of a COS7 cell overexpressing GFP-Climp63. The inset shows an enlargement of the boxed region. Scale bar, 0.5 mm. (E) HA-Rtn4b (red) was expressed in COS7 cells at relatively low levels and was localized with HA antibodies by indirect immunofluorescence and confocal microscopy. Endogenous Climp63 (green) was localized in the same cells with specific antibodies. (F) As in (G) but with the highest observed expression level of HA-Rtn4b. Note that Climp63 appears in bright punctae and in the nuclear envelope.
782 Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc.
curvature of sheet edges, a mechanism that is sufficient to keep the two membranes of a sheet closely apposed. The reticulons and DP1/Yop1p probably stabilize high curvature by two mechanisms, ‘‘hydrophobic insertion/wedging’’ and ‘‘scaffolding’’ (Shibata et al., 2009). The conserved segments of these proteins may form a wedge in the lipid bilayer that occupies more space in the cytoplasmic leaflet than in the luminal leaflet. Oligomerization of these proteins may generate scaffolds around curved membranes, which may take the shape of open arcs, given that they can localize to sheet edges. Our theoretical model demonstrates that the reticulons and DP1/Yop1p alone can generate both tubules and sheets, with their abundance determining the ratio of these domains. Consistent with the proposed dual role of the reticulons and DP1/Yop1p in tubule and sheet formation, they localize to both tubules and sheet edges, their depletion leads to increased sheet areas, and their overexpression converts sheets into tubules. In S. cerevisiae, the amount of membrane surface and the abundance of the reticulons and Yop1p appear to be the decisive factors determining the ratio of peripheral ER sheets and tubules. Generating more lipid increases the sheet area, whereas increasing the abundance of the curvature-stabilizing proteins increases the number of tubules. The observation of sheets in cells lacking the reticulons and Yop1p may be explained by the presence of other low-abundance curvature-promoting proteins or by the association of the cortical ER with the plasma membrane. Although we cannot exclude the existence of sheetpromoting proteins in yeast, the current data are consistent with a model in which curvature-stabilizing proteins are the major determinant of peripheral ER morphology. Our data suggest that, in mammalian cells, there are several additional factors that determine the morphology of peripheral ER sheets. This includes the coiled-coil membrane protein Climp63, which serves as a luminal spacer. After its depletion, the luminal width of the sheets decreases from 50 to 30 nm, a spacing that is also seen in organisms that lack the protein. Climp63 is highly upregulated in mammalian cells with proliferated ER sheets, and it induces sheets at the expense of tubules when overexpressed in tissue culture cells. Thus, at high concentrations, Climp63 appears to generate sheets all by itself, and the lack of extensive sheet edges may make the contribution of the curvature-stabilizing proteins less important. However, with luminal spacers alone, one would expect bulging of the sheet edges, in contrast to our observations (Figure 4), indicating that the curvature-stabilizing proteins may have a role even in cells with proliferated ER sheets. Climp63’s function may be to
optimize the size of the luminal space of peripheral ER sheets, such that sufficient luminal chaperones can be accommodated and the sheets are packed into a minimal space. Our analysis also identified two other coiled-coil membrane proteins, p180 and kinectin, with a potential role in shaping ER sheets. These proteins are enriched in sheets and abundant in cells with proliferated ER sheets. Overexpression of p180 has been reported to induce sheets in S. cerevisiae and in a monocytic cell line (Becker et al., 1999; Benyamini et al., 2009), although in our own experiments and those of others, the effects were smaller (Ueno et al., 2010 and data not shown). The depletion of p180 and kinectin had no effect on ER sheet morphology. Although the precise role of these proteins remains to be established, all coiled-coil membrane proteins could stabilize sheets simply by being excluded from high-curvature regions, as shown by our theoretical considerations. They may be considered as generating an ‘‘osmotic pressure,’’ a force that counteracts the shrinkage of sheet domains. Consistent with experimental observations for Climp63, the coiled-coil proteins are predicted to be in a tug-of-war with the reticulons and DP1/Yop1p, with the former shifting the balance toward sheets and the latter toward tubules. In this model, it does not actually matter how proteins are excluded from tubules and sheet edges. Given that all identified sheet-promoting proteins contain extended coiled-coil domains, they all have the propensity to oligomerize, which may contribute to their exclusion from high-curvature regions. The coiled-coil membrane proteins are not essential for sheet formation per se, as is obvious from our observation that their depletion by RNAi does not abolish ER sheets. This suggests that, like in yeast, the reticulons and DP1/Yop1p may provide the basic mechanism by which both sheets and tubules are generated. Consistent with this hypothesis, Climp63, p180, and kinectin are not known in lower organisms, in contrast to the reticulons and DP1/Yop1p, which are present in all eukaryotes. All sheet-enriched proteins tested, including translocon components and the coiled-coil membrane proteins, appear to be concentrated by membrane-bound polysomes; upon polysome disassembly, all of these proteins distribute equally between sheets and tubules throughout the cell. Thus, these proteins must have a direct or indirect affinity for membranebound polysomes. Indeed, several of the tested sheet-preferring proteins are known to be associated with membranebound translating ribosomes, including components of the Sec61 complex, the TRAP complex, the oligosaccharyl
(G) Quantification of the peripheral ER sheet area relative to the total ER area for different expression levels of HA-Rtn4b. The areas of ER sheets and tubules were determined from the fluorescence of Climp63 and Rtn4b, respectively, after subtraction of background. The cells were divided into five groups according to their expression levels of HA-Rtn4b, as determined by overall average fluorescence intensity, and the mean and standard error were calculated for each group. (H) Quantification of the effect of Rtn4b overexpression on ER sheet morphology, as determined by Climp63 staining. Shown are the percentages of cells with normal ER sheets (blue bars), of cells with disc-like ER sheets (red), and of cells with punctae (green) at different expression levels of Rtn4b. The cells were divided into five groups according to their expression levels. (I) Quantification of the effect of Rtn4b overexpression on ER tubule morphology, as determined by HA-Rtn4b staining. Shown are the percentages of cells with normal reticular ER (blue bars), of cells with an abnormally dense ER network (red), and of cells with unbranched, long tubules (green) at different expression levels of Rtn4b. The cells were divided into five groups according to their expression levels. (J) Myc-Rtn4a and FLAG-Climp63 were both highly expressed in COS7 cells. The far-right panel shows a merged image. Note that the ER morphology is almost normal. See also Figure S6 and Figure S7.
Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc. 783
Figure 6. The Reticulons Localize to the Edges of ER Sheets (A) The localization of endogenous Rtn4a and 4b is compared with that of Climp63 using indirect immunofluorescence with specific antibodies in COS7 cells. The insets show enlargements of the boxed region. Arrows point to reticulons lining the sheets. The far-right panel shows merged images. Scale bar, 10 mm. (B) As in (A) but with cells overexpressing FLAG-Climp63. (C) Rtn1p-GFP (green) and ssRFP-HDEL (red) were coexpressed in wild-type S. cerevisiae cells, and the cortical ER was visualized by fluorescence microscopy. Scale bar, 5 mm. (D) As in (C) except that the cells had proliferated ER sheets caused by deletion of SEY1 and YOP1 (sey1Dyop1D). (E) As in (C) except the cells had proliferated ER sheets caused by deletion of OPI1 (opi1D). The cells also contained an empty vector as a control for panels (F) and (G). (F) As in (E) except that untagged Rtn1p was expressed under the endogenous promoter from a CEN plasmid. (G) As in (E) except that untagged Rtn1p was expressed under the endogenous promoter from a 2 m plasmid. (H) Quantification of the experiments in (C) and (E–G). The relative area of ER sheets was determined from the area of ssRFP-HDEL fluorescence that did not colocalize with Rtn1p-GFP fluorescence and was divided by the total area of ssRFP-HDEL fluorescence. 14 to 38 cells were analyzed per condition, and the means and standard errors were calculated.
784 Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc.
Figure 7. Modeling of the Effect of Curvature-Stabilizing and Sheet-Promoting Proteins on ER Morphology (A) The reticulons and DP1/Yop1p (yellow arcs) are assumed to localize exclusively to tubules and sheet edges, generating and stabilizing these high-curvature membranes. Stabilization of sheet edges enables the upper and lower membranes of the sheet to adopt planar shapes. (B) Top view of membrane shapes computed by the theoretical model for increasing G values. The computation was performed for a total membrane area corresponding to 1 mm radius of the initial disc-like shape, a 15 nm cross-section radius of the tubules and edges, and a 40 nm optimal distance between the arc-like proteins at the edge (see Supplemental Information). Change of G from 1 to 2.1(blue to red) corresponds to increasing the number of curvature-stabilizing proteins Nc from 140 to 290. (C) G values and membrane shapes were calculated for different numbers of curvature-stabilizing and sheet-promoting proteins, Nc and Ns. The colors correspond to the membrane shapes shown in Figure 7B. The colored lines on the bottom plane of the diagram represent the relationship between Nc and Ns for a given shape of the system. (D) G values and membrane shapes were computed for different Ns values at Nc = 290. The shapes refer to Ns = 0, 500, and 1000.
transferase complex, and p180 (Go¨rlich and Rapoport, 1993). These proteins stay bound to ribosomes upon detergent solubilization of rough ER membranes, but they can be released from the ribosomes by puromycin/high salt treatment. Climp63 and kinectin are not bound to detergent-solubilized translocons (data not shown), so how they are recruited remains to be clarified.
Our results indicate that ER sheets correspond to rough ER and tubules to smooth ER. We propose that the assembly of translating membrane-bound ribosomes into polysomes concentrates the associated membrane-proteins, including Climp63, p180, and kinectin. Their concentration might facilitate their higher-order oligomerization, which may be required for their exclusion from high-curvature areas and thus for their Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc. 785
sheet-promoting function. Once sheets are formed, the membrane binding of polysomes would be facilitated. Polysomes often form spirals that could have an inherent preference for associating with ER sheets (Christensen and Bourne, 1999); whereas individual ribosomes or small polysomes can bind to narrow tubules, it is unlikely that each ribosome of a large polysome could be efficiently arranged on a narrow tubule. The binding of large polysomes could therefore be restricted to membrane sheets. The assembly of membrane-bound polysomes would concentrate more coiled-coil membrane proteins, and these in turn would generate more sheet area by the ‘‘osmotic effect,’’ allowing more polysomes to bind, and so on, a mechanism that would ultimately lead to a segregated rough ER domain. This model is consistent with the observation that the disassembly of polysomes or the depletion of Climp63 increases the mobility of translocons in the plane of the membrane (Nikonov et al., 2007; Nikonov et al., 2002). It also agrees with our results showing that the disassembly of polysomes leads to the spreading of ER sheets similar to that seen upon depletion of the sheet-promoting proteins. Our model explains the classic observation that, in many cells, membranebound ribosomes are not randomly distributed throughout the ER but, rather, concentrated in a separate membrane domain, the rough ER. An active sorting of proteins into the rough ER is consistent with previous cell fractionation experiments, which demonstrated that general ER proteins indiscriminately distribute throughout the ER, whereas translocon-associated proteins are enriched in the rough ER (Hinman and Phillips, 1970; Kreibich et al., 1978; Vogel et al., 1990). The nuclear envelope is a prominent ER domain whose structure is determined independently of the peripheral ER. Although the reticulons have been implicated in the assembly of the nuclear envelope and in the insertion of nuclear pores (Anderson and Hetzer, 2008; Dawson et al., 2009), they are nearly absent from the nuclear envelope, and their depletion or overexpression has no significant effect on this domain’s morphology. Similarly, DP1/Yop1p or the coiled-coil membrane proteins Climp63, p180, and kinectin are also nearly absent from the nuclear envelope and have no obvious effect on its structure. Of interest, TRAPa was also depleted from the nuclear envelope, raising the possibility that translocons are preferentially located in peripheral ER sheets. Thus, distinct mechanisms may determine the formation and function of the sheet-like domains of the nuclear envelope and peripheral ER. In summary, our results lead to a simple model, according to which the basic morphological elements of the peripheral ER, the tubules and sheets, are generated by the curvaturestabilizing proteins. Superimposed on this mechanism, membrane-bound polysomes and associated coiled-coil membrane proteins may cooperate to form segregated rough ER sheets in mammalian cells, domains that are functionally specialized in protein translocation. Other factors probably contribute to the morphology of the peripheral ER. Microtubules keep the mammalian ER under tension and stabilize membrane tubules, but they could also potentially form an additional scaffold that stabilizes sheets, as suggested by the fact that both Climp63 and p180 are microtubule-binding proteins (Klopfenstein et al., 1998; Ogawa-Goto et al., 2007). It will be interesting to elucidate 786 Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc.
how these factors collaborate with the identified membraneshaping principles. EXPERIMENTAL PROCEDURES Mammalian Tissue Culture and Transfections BSC1 cells stably expressing GFP-Sec61b and COS7 cells were grown in DMEM containing 10% fetal bovine serum at 37 C and 5% CO2 and were passaged every 2–3 days. GFP-Dad1 BHK cells (M3/18; Nikonov et al., 2002) were maintained in 10% CO2 at 39.5 C to degrade endogenous Dad1. For translation inhibition experiments, cells were treated with 200 mM cycloheximide, 200 mM puromycin, or 100 nM pactamycin in complete media for 15 min. To deplete Climp63, kinectin, and p180, COS7 cells were plated onto acid-washed coverslips at 20% confluency and were transfected with 120 nM total siRNA using Oligofectamine (Invitrogen). After1.5 days, cells were retransfected with the same amount of siRNA oligonucleotides and then processed for immunofluorescence 1.5 days afterward. Experiments with control siRNA oligonucleotides (QIAGEN) were done in parallel using the same conditions. Transient DNA transfections were performed using Lipofectamine 2000 (Invitrogen). See Supplemental Information for a list of DNA and siRNA constructs. Indirect Immunofluorescence and Confocal Microscopy Indirect immunofluorescence with mammalian cells was done as described (Shibata et al., 2008). Cells grown on acid-washed coverslips were fixed with 4% paraformaldehyde (EMS), permeabilized with 0.1% Triton X-100 (Pierce), and immunostained with various primary antibodies and then washed in PBS and probed with various fluorophore-conjugated secondary antibodies. See Supplemental Information for a list of antibodies used. Images were captured using a Yokogawa spinning-disk confocal on a Nikon TE2000U inverted microscope with a 603 or 1003 Plan Apo NA 1.4 objective lens, a Hamamatsu ORCA ER-cooled CCD camera, and MetaMorph software. All analyses/quantifications were done on raw 16 bit images using MetaMorph. For presentation, brightness levels were adjusted across the entire image and were changed from 16 to 8 bits using Adobe Photoshop. Quantification was performed as described in Supplemental Information. Thin-Section Electron Microscopy Thin-section EM experiments were performed as described previously (Shibata et al., 2008) except that cells were fixed directly in culture plates. Quantification was performed as described in Supplemental Information. Microscopy and Image Quantification of S. cerevisiae Cells Yeast strains and constructs used are described in Supplemental Information. Yeast cells were imaged live in complete medium at room temperature using an Olympus BX61 microscope, UPlanApo 100 3 /1.35 lens, Qimaging Retiga EX camera, and IVision version 4.0.5 software. To calculate relative peripheral sheet amounts, cortical ER images of cells expressing ssRFP-HDEL and Rtn1-GFP were taken. Images were thresholded above background, and the percentage of sheet area was calculated for each cell as the percentage of area of ssRFP-HDEL that did not overlap with Rtn1-GFP using Metamorph software. Means and standard errors were calculated using Microsoft Excel. For presentation, brightness levels were adjusted across the entire image, changed from 16 to 8 bits, and cropped using Adobe Photoshop. Identification of Abundant Coiled-Coil Membrane Proteins Mass spectrometry of dog pancreatic microsomal proteins and identification of mRNAs coding for ER membrane proteins that are upregulated during B cell differentiation (Luckey et al., 2006) were performed as described in the Supplemental Information. Modeling of Sheet versus Tubule Generation To compute the membrane configurations (the length of the tubule as well as the areas and shapes of the sheets) in dependence of the numbers of the curvature-producing, Nc, and the sheet-promoting proteins, Ns, we minimize the system energy, Ftot, for the given total membrane area, Atot. The total energy Ftot consists of three contributions: the effective stretching energy of
the edge, Fs; the energy of the effective osmotic pressure of the sheetpromoting proteins, Fp; and the energy of edge bending in the sheet plane, Fb. h i2 The energy Fs is given by Fs = 12kB TNc ðLe NNccðll00 + la ÞÞ , wherein, Le is the total length of the edge including the tubules; l0 is the energetically preferred distance between the arc-like proteins measured along the edge; la is the width of one protein arc; and kB Tz4$1021 Joule is the product of the Boltzmann constant and the absolute temperature. According to this expression, the length of the edge in a stress-free state is Le = Nc ðl0 + la Þ, and the effective rigidity of the edge stretching-compression with respect to Le is kstr = kB T$Nc . Based on previous estimates, we take l0 = 40nm and la = 4nm (Hu et al., 2008). The osmotic pressure energy Fp is given by Fp = kB T$Ns $lnðNs b=Aflat Þ, wherein Aflat is the flat area available to the sheet proteins and b is the area of one sheet protein. The area Aflat is related to the total area and length of the edge by Aflat = 12ðAtot a$Le Þ, wherein a is the membrane area absorbed by a unit length of the edge, which can be estimated as a = p$Re z50nm (Re z15nm is the radius of the edge cross-section), and Atot is the total membrane area. The energy Fb is given by Fb = 12B#c2e dLe , wherein ce is the in-plane curvature of the edge, and B is the modulus of the edge in-plane bending, which can be estimated using the membrane bending modulus kz20kB T (Helfrich, 1973) as BzpRe kz900kB T$nm. The integration is performed over the whole edge length, including the tubules. Estimates supported by numerical computations show that the total length of the edge Le and the corresponding value of the parameter G are determined by the energies Fs and Fp and are largely independent of Fb. At the same time, the system configuration resulting from minimization of Fb depends of the parameter G. Therefore, we determine the system configuration in two steps. First, we minimize the sum of Fc + Fs with respect to Le for every set of numbers Nc and Ns and determine the corresponding function G (Nc, Ns). Second, for every value of G (Nc, Ns), we minimize Fb with respect to the system shape and find the equilibrium configuration. The Supplemental Information gives a more detailed discussion of the model. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven figures, and two tables and can be found with this article online at doi:10.1016/j.cell.2010.11.007. ACKNOWLEDGMENTS We thank C. Denison, J. Minsteris, and S. Gygi for mass spectrometry analysis; J. Baughman for microarray analysis; A. Condon and A. Boye-Doe for technical assistance; J. Iwasa for help with illustrations; G. Kreibich, K. Ogawa-Goto, L. Lu, and R. Yan for materials; the Nikon Imaging Center and the Electron Microscopy facility at HMS for microscopy assistance; and R. Klemm and A. Osborne for critical reading of the manuscript. Y.S. was supported by the American Heart Association and is a Howard Hughes Medical Institute postdoctoral fellow. W.A.P. is supported by the Intramural Research Program of the National Institute of Diabetes and Digestive and Kidney Diseases. T.A.R. is a Howard Hughes Medical Institute Investigator. M.M.K. is supported by the Israel Science Foundation (ISF) and the Marie Curie network ‘‘Virus Entry.’’ Received: May 19, 2010 Revised: September 3, 2010 Accepted: October 26, 2010 Published: November 24, 2010
Anderson, D.J., and Hetzer, M.W. (2008). Reshaping of the endoplasmic reticulum limits the rate for nuclear envelope formation. J. Cell Biol. 182, 911–924. Baumann, O., and Walz, B. (2001). Endoplasmic reticulum of animal cells and its organization into structural and functional domains. Int. Rev. Cytol. 205, 149–214. Becker, F., Block-Alper, L., Nakamura, G., Harada, J., Wittrup, K.D., and Meyer, D.I. (1999). Expression of the 180-kD ribosome receptor induces membrane proliferation and increased secretory activity in yeast. J. Cell Biol. 146, 273–284. Benyamini, P., Webster, P., and Meyer, D.I. (2009). Knockdown of p180 eliminates the terminal differentiation of a secretory cell line. Mol. Biol. Cell 20, 732–744. Bernales, S., McDonald, K.L., and Walter, P. (2006). Autophagy counterbalances endoplasmic reticulum expansion during the unfolded protein response. PLoS Biol. 4, e423. Carman, G.M., and Henry, S.A. (2007). Phosphatidic acid plays a central role in the transcriptional regulation of glycerophospholipid synthesis in Saccharomyces cerevisiae. J. Biol. Chem. 282, 37293–37297. Christensen, A.K., and Bourne, C.M. (1999). Shape of large bound polysomes in cultured fibroblasts and thyroid epithelial cells. Anat. Rec. 255, 116–129. Dawson, T.R., Lazarus, M.D., Hetzer, M.W., and Wente, S.R. (2009). ER membrane-bending proteins are necessary for de novo nuclear pore formation. J. Cell Biol. 184, 659–675. Fawcett, D.W. (1981). The Cell (Philadelphia: W.B. Saunders). Go¨rlich, D., and Rapoport, T.A. (1993). Protein translocation into proteoliposomes reconstituted from purified components of the endoplasmic reticulum membrane. Cell 75, 615–630. Helfrich, W. (1973). Elastic properties of lipid bilayers: theory and possible experiments. Z. Naturforsch. C 28, 693–703. Hinman, N.D., and Phillips, A.H. (1970). Similarity and limited multiplicity of membrane proteins from rough and smooth endoplasmic reticulum. Science 170, 1222–1223. Hu, J., Shibata, Y., Voss, C., Shemesh, T., Li, Z., Coughlin, M., Kozlov, M.M., Rapoport, T.A., and Prinz, W.A. (2008). Membrane proteins of the endoplasmic reticulum induce high-curvature tubules. Science 319, 1247–1250. Hu, J., Shibata, Y., Zhu, P.P., Voss, C., Rismanchi, N., Prinz, W.A., Rapoport, T.A., and Blackstone, C. (2009). A class of dynamin-like GTPases involved in the generation of the tubular ER network. Cell 138, 549–561. Klopfenstein, D.R., Kappeler, F., and Hauri, H.P. (1998). A novel direct interaction of endoplasmic reticulum with microtubules. EMBO J. 17, 6168–6177. Klopfenstein, D.R., Klumperman, J., Lustig, A., Kammerer, R.A., Oorschot, V., and Hauri, H.P. (2001). Subdomain-specific localization of CLIMP-63 (p63) in the endoplasmic reticulum is mediated by its luminal a-helical segment. J. Cell Biol. 153, 1287–1300. Kreibich, G., Ulrich, B.L., and Sabatini, D.D. (1978). Proteins of rough microsomal membranes related to ribosome binding. I. Identification of ribophorins I and II, membrane proteins characteristics of rough microsomes. J. Cell Biol. 77, 464–487. Luckey, C.J., Bhattacharya, D., Goldrath, A.W., Weissman, I.L., Benoist, C., and Mathis, D. (2006). Memory T and memory B cells share a transcriptional program of self-renewal with long-term hematopoietic stem cells. Proc. Natl. Acad. Sci. USA 103, 3304–3309. Me´ne´tret, J.F., Hegde, R.S., Aguiar, M., Gygi, S.P., Park, E., Rapoport, T.A., and Akey, C.W. (2008). Single copies of Sec61 and TRAP associate with a nontranslating mammalian ribosome. Structure 16, 1126–1137.
REFERENCES
Nikonov, A.V., Hauri, H.P., Lauring, B., and Kreibich, G. (2007). Climp-63mediated binding of microtubules to the ER affects the lateral mobility of translocon complexes. J. Cell Sci. 120, 2248–2258.
Ambroziak, J., and Henry, S.A. (1994). INO2 and INO4 gene products, positive regulators of phospholipid biosynthesis in Saccharomyces cerevisiae, form a complex that binds to the INO1 promoter. J. Biol. Chem. 269, 15344–15349.
Nikonov, A.V., Snapp, E., Lippincott-Schwartz, J., and Kreibich, G. (2002). Active translocon complexes labeled with GFP-Dad1 diffuse slowly as large polysome arrays in the endoplasmic reticulum. J. Cell Biol. 158, 497–506.
Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc. 787
Ogawa-Goto, K., Tanaka, K., Ueno, T., Tanaka, K., Kurata, T., Sata, T., and Irie, S. (2007). p180 is involved in the interaction between the endoplasmic reticulum and microtubules through a novel microtubule-binding and bundling domain. Mol. Biol. Cell 18, 3741–3751. Puhka, M., Vihinen, H., Joensuu, M., and Jokitalo, E. (2007). Endoplasmic reticulum remains continuous and undergoes sheet-to-tubule transformation during cell division in mammalian cells. J. Cell Biol. 179, 895–909. Savitz, A.J., and Meyer, D.I. (1990). Identification of a ribosome receptor in the rough endoplasmic reticulum. Nature 346, 540–544. Schuck, S., Prinz, W.A., Thorn, K.S., Voss, C., and Walter, P. (2009). Membrane expansion alleviates endoplasmic reticulum stress independently of the unfolded protein response. J. Cell Biol. 187, 525–536.
Short, B., Haas, A., and Barr, F.A. (2005). Golgins and GTPases, giving identity and structure to the Golgi apparatus. Biochim. Biophys. Acta 1744, 383–395. Snapp, E.L., Hegde, R.S., Francolini, M., Lombardo, F., Colombo, S., Pedrazzini, E., Borgese, N., and Lippincott-Schwartz, J. (2003). Formation of stacked ER cisternae by low affinity protein interactions. J. Cell Biol. 163, 257–269. Sparkes, I., Tolley, N., Aller, I., Svozil, J., Osterrieder, A., Botchway, S., Mueller, C., Frigerio, L., and Hawes, C. (2010). Five Arabidopsis reticulon isoforms share endoplasmic reticulum location, topology, and membrane-shaping properties. Plant Cell 22, 1333–1343. Toyoshima, I., Yu, H., Steuer, E.R., and Sheetz, M.P. (1992). Kinectin, a major kinesin-binding protein on ER. J. Cell Biol. 118, 1121–1131.
Senda, T., and Yoshinaga-Hirabayashi, T. (1998). Intermembrane bridges within membrane organelles revealed by quick-freeze deep-etch electron microscopy. Anat. Rec. 251, 339–345.
Ueno, T., Tanaka, K., Kaneko, K., Taga, Y., Sata, T., Irie, S., Hattori, S., and Ogawa-Goto, K. (2010). Enhancement of procollagen biosynthesis by p180 through augmented ribosome association on the endoplasmic reticulum in response to stimulated secretion. J. Biol. Chem. 285, 29941–29950.
Shibata, Y., Hu, J., Kozlov, M.M., and Rapoport, T.A. (2009). Mechanisms shaping the membranes of cellular organelles. Annu. Rev. Cell Dev. Biol. 25, 329–354.
Voeltz, G.K., Prinz, W.A., Shibata, Y., Rist, J.M., and Rapoport, T.A. (2006). A class of membrane proteins shaping the tubular endoplasmic reticulum. Cell 124, 573–586.
Shibata, Y., Voeltz, G.K., and Rapoport, T.A. (2006). Rough sheets and smooth tubules. Cell 126, 435–439.
Voeltz, G.K., Rolls, M.M., and Rapoport, T.A. (2002). Structural organization of the endoplasmic reticulum. EMBO Rep. 3, 944–950.
Shibata, Y., Voss, C., Rist, J.M., Hu, J., Rapoport, T.A., Prinz, W.A., and Voeltz, G.K. (2008). The reticulon and DP1/Yop1p proteins form immobile oligomers in the tubular endoplasmic reticulum. J. Biol. Chem. 283, 18892–18904.
Vogel, F., Hartmann, E., Go¨rlich, D., and Rapoport, T.A. (1990). Segregation of the signal sequence receptor protein in the rough endoplasmic reticulum membrane. Eur. J. Cell Biol. 53, 197–202.
788 Cell 143, 774–788, November 24, 2010 ª2010 Elsevier Inc.
Abortive HIV Infection Mediates CD4 T Cell Depletion and Inflammation in Human Lymphoid Tissue Gilad Doitsh,1 Marielle Cavrois,1,5 Kara G. Lassen,1,5 Orlando Zepeda,1 Zhiyuan Yang,1 Mario L. Santiago,1,4 Andrew M. Hebbeler,1 and Warner C. Greene1,2,3,* 1Gladstone
Institute of Virology and Immunology, 1650 Owens Street, San Francisco, CA 94158, USA of Medicine 3Department of Microbiology and Immunology University of California, San Francisco, CA 94143, USA 4Present address: University of Colorado, Denver, Aurora, CO 80045, USA 5These authors contributed equally to this work *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.11.001 2Department
SUMMARY
The mechanism by which CD4 T cells are depleted in HIV-infected hosts remains poorly understood. In ex vivo cultures of human tonsil tissue, CD4 T cells undergo a pronounced cytopathic response following HIV infection. Strikingly, >95% of these dying cells are not productively infected but instead correspond to bystander cells. We now show that the death of these ‘‘bystander’’ cells involves abortive HIV infection. Inhibitors blocking HIV entry or early steps of reverse transcription prevent CD4 T cell death while inhibition of later events in the viral life cycle does not. We demonstrate that the nonpermissive state exhibited by the majority of resting CD4 tonsil T cells leads to accumulation of incomplete reverse transcripts. These cytoplasmic nucleic acids activate a host defense program that elicits a coordinated proapoptotic and proinflammatory response involving caspase-3 and caspase-1 activation. While this response likely evolved to protect the host, it centrally contributes to the immunopathogenic effects of HIV. INTRODUCTION Despite extensive efforts over the past quarter century, the precise mechanism by which HIV-1 causes progressive depletion of CD4 T cells remains debated. Both direct and indirect cytopathic effects have been proposed. When immortalized T cell lines are infected with laboratory-adapted HIV-1 strains, direct CD4 T cell killing predominates. Conversely, in more physiological systems, such as infection of lymphoid tissue with primary HIV-1 isolates, the majority of dying cells appear as uninfected ‘‘bystander’’ CD4 T cells (Finkel et al., 1995; Jekle et al., 2003).
Various mechanisms have been proposed to contribute to the death of these bystander CD4 T cells including the action of host-derived factors like tumor necrosis factor-a, Fas ligand and TRAIL (Gandhi et al., 1998; Herbeuval et al., 2005), and viral factors like HIV-1 Tat, Vpr, and Nef released from infected cells (Schindler et al., 2006; Westendorp et al., 1995). Considerable interest has also focused on the role of gp120 and gp41 Env protein in indirect cell death, although it is not clear whether death signaling involves gp120 binding to its chemokine receptor or gp41-mediated fusion. It is also unclear whether such killing is caused by HIV-1 virions or by infected cells expressing Env. Most studies have focused on death mechanisms acting prior to viral entry. Less is known about the fate of HIV-1-infected CD4 T cells that do not express viral genes, in particular naive CD4 T cells in tissue that are refractory to productive HIV infection (Glushakova et al., 1995; Kreisberg et al., 2006). In these cells, infection is aborted after viral entry, as reverse transcription is initiated but fails to reach completion (Kamata et al., 2009; Swiggard et al., 2004; Zack et al., 1990; Zhou et al., 2005). Human lymphoid aggregated cultures (HLACs) prepared from tonsillar tissue closely replicate the conditions encountered by HIV in vivo and thus form an attractive, biologically relevant system for studying HIV-1 infection (Eckstein et al., 2001). Lymphoid organs are the primary sites of HIV replication and contain more than 98% of the body’s CD4 T cells. Moreover, events critical to HIV disease progression occur in lymphoid tissues, where the network of cell-cell interactions mediating the immune response deteriorates and ultimately collapses. Primary cultures of peripheral blood cells do not fully mimic the cytokine milieu, the cellular composition of lymphoid tissue, nor the functional relationships that are undoubtedly important in HIV pathogenesis. Finally, HLACs can be infected with a low number of viral particles in the absence of artificial mitogens, allowing analysis of HIV cytopathicity in a natural and preserved environment. In this study, we used the HLAC system to explore the molecular basis for HIV-induced killing of CD4 T cells. Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc. 789
Figure 1. Massive Depletion of CD4 T Cells in HLACs Containing Small Number of Productively Infected Cells Kinetics of spreading viral infection versus depletion of CD4 T cells after infection of HLACs with a replication-competent HIV reporter virus encoding GFP. CD4 downregulation in GFP-positive cells likely represents the combined action of the HIV Nef, Vpu, and Env proteins expressed by this virus. Ratios of viable CD4 versus CD8 T cells in HIV-infected and uninfected cultures are also shown. Flow cytometry plots represent live-gated cells, based on the forward-scatter versus side-scatter profile of the complete culture. These data are the representative results of six independent experiments utilizing tonsil cells from six different donors.
RESULTS Selective Depletion of CD4 T Cells by X4-Tropic HIV-1 To explore depletion of CD4 T cells by HIV-1, HLACs made from freshly dissected human tonsillar tissues were infected with a GFP reporter virus (NLENG1), prepared from the X4-tropic NL4-3 strain of HIV-1. This reporter produces fully replicationcompetent viruses. An IRES inserted upstream of the Nef gene preserves Nef expression and supports LTR-driven GFP expression (Levy et al., 2004), allowing simultaneous quantification of the dynamics of HIV-1 infection and T cell depletion. NL4-3 790 Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc.
was selected because tonsillar tissue contains a high percentage of CD4 T cells expressing CXCR4 (90%–100%). Productively infected GFP-positive cells appeared in small numbers 3 days after infection, peaked on days 6–9, and decreased until day 12, when few CD4 T cells remained in the culture (Figure 1). Fluorescence-linked antigen quantification (FLAQ) assay of HIV-1 p24 (Hayden et al., 2003) confirmed the accumulation of viral particles in the medium between day 3 and days 8 to 9, when a plateau was reached (data not shown). Interestingly, when HIV-1 p24 levels plateaued no more than 1.5% of all cells (about 5% of CD4 T cells) were GFP-positive.
Figure 2. CD4 T Cell Depletion in HIV-1-Infected HLACs Predominantly Involves Nonproductively Infected Cells (A) Experimental strategy to assess indirect cell killing in HIV-1-infected human lymphoid cultures. Fresh human tonsil tissue from a single donor is processed into HLAC, and then separated into two fractions. One fraction is challenged with HIV-1 and cultured for 6 days, allowing viral spread. On day 5, the uninfected fraction is treated with AZT (5 mM) and labeled with CFSE (1 mM). On day 6, the infected and CFSE-labeled cultures are mixed and cocultured in the presence of AZT. Because of its site of action, AZT does not block viral output from the HIV-infected cells but prevents productive infection of CFSE-labeled cells. After 6 days of coculturing, the number of viable CFSE-positive cells is determined by flow cytometry. (B) Flow cytometry analysis of the mixed HLACs. Indirect killing is determined by gating on live CFSE-positive cells in the mixed cultures. Effector cells are either infected or uninfected cells. (C) Extensive depletion of nonproductively infected CD4 T cells by HIV-1. CFSE-labeled cells mixed with uninfected or infected cells were cultured in the presence of 5 mM AZT alone or together with 250 nM AMD3100. Data represent live CFSE-positive cells 6 days after coculture with infected or uninfected effector cells. The absence of productive infection in the CFSE-positive cells was confirmed by internal p24 staining and monitoring GFP expression following infection with the NLENG1 HIV-1 reporter virus (data not shown). (D) Preferential depletion of nonproductively infected CD4 T cells by HIV-1. The absolute numbers of viable CFSE-positive CD4 and CD8 T cells and B cells were determined. Percentages are normalized to the number of viable CFSE-positive cells cocultured with uninfected cells in the presence of AZT, as depicted by an asterisk. Error bars represent standard deviations of three samples from the same donor. This experiment is representative of more than 10 independent experiments with more than 10 donors of tonsillar tissues. See also Figure S1.
However, although the number of CD4 T cells was not markedly altered in infected cultures through six days, the culture was almost completely devoid of CD4 T cells by day 9. CD8 T cells were not depleted in infected cultures, and CD4 T cells were not depleted in uninfected cultures. These findings reveal marked and selective depletion of CD4 T cells in HLAC cultures. However, due to the nature of the assay, we could not definitively conclude whether the principal mechanism of depletion involved direct or indirect effects of HIV-1.
Extensive Depletion of Nonproductively Infected CD4 T Cells in HLACs To determine if indirect killing (formerly indicated as ‘‘bystander’’) of CD4 T cells accounted for most of the observed cellular depletion, we took advantage of a reported experimental strategy (Jekle et al., 2003) that unambiguously distinguishes between the death of productively and nonproductively infected cells (Figure 2A). After 6 days of coculture, survival analysis of CFSE-labeled cells by flow cytometry (Figure 2B) showed Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc. 791
Figure 3. HIV-1 Fusion Is Necessary to Induce Killing of Nonproductively Infected Cells (A and C) Concentrations of T20 that block viral infection. HLACs were infected with the indicated clones of HIV-1 in the presence of the indicated concentrations of T20 or no drugs. One hour before incubation with the virus, cells were pretreated with T20 or left untreated. At 12 hr, cells were washed extensively and cultured under the same conditions. On day 9, the viral concentration was determined using a p24gag FLAQ assay. The amount of p24gag accumulated in the absence of drugs by each viral clone (A) or by SKY (C) was defined as 100%. (B and D) Effect of T20 on indirect killing. CFSE-labeled cells were cocultured with cells infected with the indicated viral clones in the presence of 5 mM AZT and the indicated concentrations of T20. After 6 days, indirect killing in the mixed cultures was assessed. The number of viable CFSE-positive CD4 T cells cocultured with uninfected cells in the presence of AZT was defined as 100% (data not shown). To allow successful initial infection we pseudotyped the GIA-SKY mutants with the VSV-G envelope. NL4-3, WT lab-adapted virus; WEAU 16-8, primary virus; SIM, T20-resistant virus; GIA-SKY, T20-dependent virus; GIA and SKY, single-domain mutant viruses. Representative data from three independent experiments with different donors are shown. See also Figure S2.
extensive depletion of CD4 T cells in cultures mixed with HIVinfected cells but not in those mixed with uninfected cells (Figure 2C). The relative proportion of CD8 T cells was not altered. CD3+/CD8– T cells were similarly depleted, indicating that the loss was not an artifact of downregulated surface expression of CD4 following direct infection. Loss of CFSE-labeled CD4 T cells was prevented by AMD3100, which blocks the engagement of gp120 with CXCR4, but not by the reverse transcriptase inhibitor AZT. Thus, productive viral replication is not required for CD4 T cell death. To estimate the absolute numbers of all CFSE-labeled cell subsets, we added a standard number of fluorescent beads to the cell suspensions (Figure 2D). In contrast to the sharp decline in CD4 T cells, the absolute numbers of CD8 T and B cells were unaltered. Separating the HLAC into distinct cell types revealed that cell death occurred in purified populations of CD4 T cells suggesting that other cell types did not mediate the killing. (Figure S1 available online). In all instances, CD4-specific killing was prevented by AMD3100 but not AZT. Importantly, the extent of CD4 T cell depletion in the presence of AZT was similar to that observed when no antiviral drugs were added (Figure 2C and Figure 1, respectively). Together, these results suggest that indirect killing is the predominant mechanism for CD4 T cell depletion in HIV-infected HLACs. 792 Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc.
HIV gp41-Mediated Fusion Is Necessary for Depletion of Nonproductively Infected CD4 T Cells Studies with AMD3100 and AZT indicated that indirect CD4 T cell killing is mediated by events occurring between gp120-CXCR4 binding and reverse transcription. Engagement of the chemokine coreceptor induces conformational changes in gp41, resulting in insertion of viral fusion peptide on gp41 into the target T cell membrane. To determine if the gp120-CXCR4 interaction alone or later events involving viral fusion are required for indirect killing, we evaluated the effects of enfuvirtide (T20), a fusion inhibitor that blocks six-helix bundle formation by gp41, a prerequisite for virion fusion and core insertion. We first determined the optimal concentrations of T20 that block viral infection (Figure 3A). In NL4-3-infected cells, T20 began to inhibit infection at concentrations > 2 mg/ml; complete inhibition required 10 mg/ml. In cells infected with a primary viral isolate, WEAU 16-8 (Figure S2), infection was completely inhibited by 0.1 mg/ml of T20. T20 did not inhibit infection with a T20-resistant mutant, SIM (Rimsky et al., 1998), regardless of concentration. Next, we investigated the effect of T20 on indirect CD4 T cell killing (Figure 3B). In the absence of T20, high levels of indirect killing were observed. T20 concentrations that blocked infection also greatly inhibited indirect killing. T20 did not inhibit indirect killing in cultures containing SIM-infected cells. Thus, blocking gp41-mediated fusion prevents indirect killing.
We then examined a T20-dependent mutant, GIA-SKY (Baldwin et al., 2004), which fuses only when T20 is present, but cannot initiate a spreading infection in the absence of T20 (Figure 3C). Consistent with its T20 dependency, in the presence of 1 mg/ml T20, the GIA-SKY mutant readily replicated while growth was inhibited at higher or lower T20 concentrations. The single-domain mutants GIA and SKY exhibited a T20-resistance phenotype similar to that of SIM. GIA-SKY-infected cells did not induce indirect killing of CD4 T cells in the absence of T20 (Figure 3D). Indirect killing was observed in cultures treated with 1 mg/ml T20 but was inhibited at higher or lower concentrations. Since T20-dependent viruses were bound to CXCR4 before T20 was added, these findings argue that CXCR4 signaling is not sufficient to elicit indirect CD4 T cell killing. Indirect Killing Requires a Close Interaction between Uninfected and HIV-Infected Cells Next we examined whether indirect killing requires close contact with HIV-infected cells or instead can be fully supported by virions accumulating in the supernatants of the infected histocultures. We found that cell-free supernatants from HIV-infected histocultures were much less efficient at inducing indirect killing (Figure 4A). To exclude the possibility that the concentration of virions in the supernatants was too low, we repeated this experiment using a 20-fold concentrated virion supernatants (1 mg p24/ml) but failed to detect indirect CD4 T cell killing (Figure 4B). Together, these findings suggest that close cell-cell contact is likely required for indirect killing. To further explore the potential requirement of close cell-cell contact for indirect killing (Sherer et al., 2007; Sourisseau et al., 2007), we repeated these assays using cells that had been washed daily with fresh RPMI to prevent accumulation of HIV-1 virions and soluble factors. Such cell washing did not affect the ability of the resultant infected cells to mediate indirect CD4 T cell killing (Figure 4B), suggesting that virions released into the medium do not participate in indirect killing. We confirmed these findings using a transwell culture system. CSFE-labeled cells and HIV-infected cells were mixed or physically separated by a transwell insert with 1 mm pores, which allows free diffusion of virions but not cells. Indirect killing was substantial in the mixed cultures but not in the transwell cultures (Figure 4C). Together, these findings indicate that indirect killing requires close interaction between CFSE-labeled and HIV-1-infected cells, consistent with in vitro (Garg et al., 2007; Holm and Gabuzda, 2005) and in vivo studies showing that apoptotic nonproductively infected cells in human lymph nodes often cluster near productively infected cells (Finkel et al., 1995). Indirect Killing Requires Fusion of Virions from Nearby HIV-Producing Cells Indirect killing required gp41-mediated fusion and close interaction with HIV-infected cells, suggesting that cell death may be caused by the fusion of HIV-1 virions to CD4 T cells, syncytia formation, or hemifusion (mixing of lipids in the absence of fusion pore formation) mediated by Env present on HIV-infected cells interacting with neighboring CD4 T cells. HIV-1 virions (Holm et al., 2004; Jekle et al., 2003; Vlahakis et al., 2001), cell-mediated fusion (LaBonte et al., 2000; Margolis et al., 1995), and hemifusion (Garg et al., 2007) have been proposed to be involved
in indirect killing. Therefore, the requirement for cell-cell interaction in indirect killing may be mediated either by effective delivery of HIV-1 virions or by cell-associated Env. To discriminate between virion-mediated and cell-associated Env induction of indirect killing, we tested the effects of HIV protease inhibitors. These inhibitors act during the budding process, resulting in immature viral particles that cannot fuse with target cells (Wyma et al., 2004). We first assessed the effect of protease inhibitors on viral maturation. NL4-3 viruses carrying a b-lactamase-Vpr (BlaM-Vpr) reporter protein were produced in 293T cells in the presence or absence of the HIV protease inhibitor amprenavir. We also produced a mutant virus, TR712, encoding a form of gp41 lacking 144 of the 150 amino acids in the C-terminal cytoplasmic tail. This deletion largely relieves the impaired fusogenic properties of immature HIV-1 particles (Wyma et al., 2004). Protein analysis of viral lysates showed that the NL4-3 and TR712 virions appropriately cleaved gp160 to generate gp120 in the presence and absence of amprenavir. However, in the presence of amprenavir, an uncleaved form of p55 Gag polypeptide rather than the mature p24 CA protein accumulated in both NL4-3 and TR712 virions (Figure 4D). These results confirm that amprenavir treatment of virus producing cells results in the accumulation of immature particles containing normal levels of incorporated Env proteins. To test the ability of these viruses to fuse with target cells, we used an HIV virion-based fusion assay that measures b-lactamase (BlaM) activity delivered to target cells upon the fusion of virions containing BlaM fused to the Vpr protein (BlaM-Vpr) (Cavrois et al., 2002). Immunoblotting for BlaM confirmed that NL4-3 and TR712 virions incorporated Blam-Vpr in the presence or absence of amprenavir (Figure 4D). Next, SupT1 cells were infected with mature or amprenavirtreated immature NL4-3 or TR712 virions containing BlaM-Vpr. Immature NL4-3 viruses displayed a 90% decline in fusogenic properties (Figure 4E). In contrast, immature TR712 retained 40% fusion capacity, indicating that the impaired fusion is not a result of a defective BlaM enzyme. Thus, immature virions generated in the presence of amprenavir display greatly reduced ability to fuse with target cells. Importantly, protease inhibitors did not affect the function of Env proteins expressed on infected cells and did not block cell-cell fusion (Figure S3C). We next investigated the effect of protease inhibitors on indirect killing. Remarkably, three different protease inhibitors inhibited indirect killing as efficiently as AMD3100 (Figure 4F). These results indicated that HIV-1 virions, not HIV-infected cells, are responsible for indirect CD4 T cell killing. Additionally, recapitulating the efficient viral delivery of close cell-cell interactions by spinoculation of free virions resulted in extensive and selective indirect killing of CD4 T cells while sparing CD8 T cells and B cells (Figures S3A and S3B). Thus, although indirect killing in lymphoid cultures requires a close interaction between nonproductively and productively infected cells, this killing involves virions rather than cell-associated Env. Nonpermissive CD4 T Cells Die from Abortive Infection Based on these findings, we hypothesized that ‘‘indirect killing’’ involves an abortive form of infection, like that which occurs in nonpermissive resting CD4 T cells. These naive CD4 T cells Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc. 793
Figure 4. Killing of Nonproductively Infected CD4 T Cells Requires Fusion of Virions from Nearby HIV-1-Producing Cells (A) Supernatants from HIV-infected HLACs are less efficient at inducing indirect killing than mixing of HIV-infected and uninfected HLACs. (B) HIV-1 virions released into the medium do not participate in indirect killing. Replacing the mixed culture with fresh RPMI every 24 hr did not impair indirect killing. Challenging HLACs with supernatants containing 20-fold more histoculture-derived virions (1 mg p24/ml) than normally accumulated in mixed cultures containing infected cells (50 ng p24/ml) did not induce indirect killing. Percentages are normalized to the number of viable CFSE-positive cells depicted by an asterisk. (C) CFSE-labeled cells are not killed when HIV-infected HLAC is physically separated by a 1 mm –pore transwell system that allows free diffusion of HIV-1 particles. Values represent the levels of viable CFSE-positive cells after 6 days of culture in the presence of the indicated drugs. Red, HIV-infected cells; blue, uninfected cells; green, CFSE-labeled cells. (D) Mature and immature viruses carry equivalent amounts of envelope protein and Blam-Vpr, but differ in their content of capsid and Gag precursor. NL4-3 and TR712 viruses were generated in 293T cells with or without amprenavir, lysed and subjected to SDS-PAGE immunoblotting analysis for gp120, p55 Gag, p24 CA, Blam-Vpr, and free Blam. (E) Immature viruses have reduced capacity to enter cells. SupT1 cells were mock infected or infected with mature or immature NL4-3 or TR712 virions containing Blam-Vpr. After loading of cells with CCF2 dye, fusion was analyzed by flow cytometry. Percentages are the fraction of cells displaying increased cleaved CCF2 fluorescence, indicating virion fusion. (F) Protease inhibitors inhibit indirect killing. CFSE-labeled cells were cocultured with NL4-3-infected or uninfected cells in the presence of AZT (5 mM) alone or together with AMD3100 (250 nM). To the indicated cultures were added 5 mM of amprenavir, saquinavir, or indinavir. Percentages are normalized to the number of viable CFSE-positive cells depicted by an asterisk. Error bars represent the SD obtained with three independent samples from the same donor. See also Figure S3.
exhibit an early post-entry block to HIV-1 infection that can be relieved by activation with phytohemagglutinin (PHA) and interleukin-2 (IL-2) (Kreisberg et al., 2006; Santoni de Sio and Trono, 794 Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc.
2009; Unutmaz et al., 1999; Zack et al., 1990). To test this hypothesis, we compared the killing of activated and nonactivated CFSE-labeled cells in HLACs.
Figure 5. Death of Abortively Infected CD4 T Cells Is Due to Impaired Reverse Transcription (A) Status of mixed HLACs containing either resting or activated CFSE-labeled cells, 4 days after coculturing with effector cells. Activated CFSE-labeled cells were stimulated with PHA and IL-2 48 hr before mixing, but not during coculturing with effector cells. To avoid direct killing of activated CFSE-labeled cells in cultures with no drugs, cell killing was terminated and analyzed 4 days after coculturing. (B) AZT renders activated CFSE-labeled CD4 T cells sensitive to indirect killing. Resting or activated CFSE-labeled cells were cocultured with effector cells in the presence of no drugs, AZT (5 mM) alone, or AZT and AMD3100 (250 nM). Data are from two independent experiments performed with tonsil cells from two different donors. (C) AZT-induced killing is lost when AZT-resistant viruses are tested. Resting or activated CFSE-labeled cells were cocultured with cells infected with NL4-3 or HIV-1 clones #629 and #964 in the presence of no drugs, AZT (0.5 mM) alone, or AZT and AMD3100 (250 nM). AZT-sensitive and AZT-resistant sub-clones are depicted. Data are representative of three independent experiments with three different donors. (D) NNRTIs prevent killing of abortive infected CD4 T cells. Resting or activated CFSE-labeled cells were cocultured with infected or uninfected effector cells, in the presence of no drugs, AZT (5 mM), AMD3100 (250 nM), the NNRTIs efavirenz (100 nM), and nevirapine (1 mM), or the integration inhibitors raltegravir (30 mM) and 118-D-24 (60 mM). Killing of resting CFSE-labeled CD4 T cells was blocked with equal efficiency by NNRTIs and AMD3100 (columns 15, 16), but not by integration inhibitors (columns 17, 18). In combination, NNRTIs prevented cell death induced by AZT in activated CFSE-labeled cells (compare column 38 to 44 and 45). Data are representative of four independent experiments with four different donors. The absolute numbers of CFSE-labeled CD8 T cells and B cells was unaltered in these experiments (data not shown). Percentages are normalized to the number of viable CFSE-positive cells depicted by an asterisk. See also Figure S4.
CFSE-labeled cells were activated with PHA and IL-2 two days before mixing with effector cells, and contained a large percentage of dividing CD25 and CD69 positive cells. Nonactivated (resting) CFSE-labeled cells did not divide and typically
contained a small percentage of cells expressing CD25 and CD69 (Figure 5A). Either in the presence or absence of AZT, killing of resting CFSE-labeled CD4 T cells was robust (Figure 5B, columns 4 and 5, and 16 and 17). In sharp contrast, activated Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc. 795
Figure 6. Cytoplasmic HIV-1 DNA Triggers Proapoptotic and Proinflammatory Responses in Abortively Infected CD4 T Cells (A) Critical reactions in HIV-1 reverse transcription as detected by probes monitoring different regions within the Strong stop, Nef, and Env DNA fragments. RDDP, RNA-dependent DNA polymerase. Adapted from S.J. Flint et al., Principles of Virology, 2000 ASM Press, Washington DC, with permission. (B) NNRTIs prevent accumulation of DNA elongation products. The amount of viral DNA detected by a particular probe was calculated as a fold change relative to cells treated with no drugs (i.e., calibrator). A bactin probe was used as an internal reference. Mean cycle threshold (Ct) values of calibrator samples are depicted. CD4 T cells were infected with WT NL4-3 produced in 293T cells, or with a Dvif NL4-3 collected from supernatants of infected HLAC, as described in Figure S4C. Data are representative of two independent experiments performed with cells from two different donors. (C and D) Abortive HIV-1 infection generates a coordinated proapoptotic and proinflammatory response involving caspase-3 and caspase 1 activation. HLACs were spinoculated with no virus or with NL4-3 and AZT (5 mM), Efavirenz (100 nM), and T20 (10 mg/ml), as indicated (see Figures S3A and S3B). After 3 days, cells were assessed by flow cytometry for intracellular levels of proinflammatory cytokines, serine 37 phosporylated p53, and activated caspases as indicated. Ethidium monoazide was used to exclude dead and necrotic cells from the annexinV binding analysis. Data are representative of three independent experiments with three different donors. (E) Death of abortively infected CD4 T cells requires caspase activation. CSFE-labeled cells were cocultured with effector cells in the presence of 20 mM of Z-VADFMK, a general caspase inhibitor, or Z-FA-FMK, a negative control for caspase inhibitors. AZT (5 mM); AMD3100 (250 nM). Percentages are normalized to the number of viable CFSE-positive cells depicted by an asterisk. Error bars represent standard error of the mean of three experiments from three different HLAC donors. (F) Abortive HIV infection promotes the maturation and secretion of IL-1b in tonsillar CD4 T cells. Isolated tonsillar CD4 T cells were either untreated, or stimulated with PMA (phorbol-12-myristate-12-acetate, 0.5 mM) and the potassium ionophore nigericin (10 mM), or spinoculated with or without NL4-3 in the presence of
796 Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc.
CFSE-labeled CD4 T cells were not depleted in the absence of AZT, but were extensively depleted in cultures containing AZT (Figure 5B, columns 10 and 11 and 22 and 23). Addition of AMD3100 prevented the AZT-induced killing of activated CFSE-labeled cells, excluding nonspecific toxic effects of AZT in the activated cells (Figure 5B, columns 12 and 24). The ability of AZT to promote indirect killing of activated CD4 T cells suggested that cell death is triggered by impaired reverse transcription. To investigate this possibility, we repeated the experiment with two pairs of AZT-resistant HIV-1 clones, 629 and 964 (Larder et al., 1989). We first determined that concentrations of 0.5 mM AZT block viral replication in NL4-3-infected and AZT-sensitive clones and achieve half maximal inhibitory effect in AZT-resistant clones (Figures S4A and S4B). When resting CFSE-labeled cells were used, the extent of killing by the AZT-resistant HIV-1 viruses was similar to that obtained with NL4-3 with or without AZT (Figure 5C resting CFSE-positive cells), demonstrating a redundant function for endogenous termination of reverse transcription and AZT. Alternatively, when activated CFSE-labeled cells were tested, AZTresistant HIV-1 clones did not deplete CFSE-labeled CD4 T cells in the presence of AZT (Figure 5C, columns 29 and 35). Death of Abortively Infected CD4 T Cells Is Triggered by Premature Termination of Viral DNA Elongation We next asked what stage of reverse transcription triggers abortive infection cell death. AZT inhibits DNA elongation but not early DNA synthesis (Arts and Wainberg, 1994). We therefore examined whether blocking early DNA synthesis with nonnucleoside reverse transcriptase inhibitors (NNRTIs) would have the same effect as AZT. Impaired reverse transcription may also lead to abortive integration, causing chromosomal DNA breaks and a genotoxic response. To exclude this possibility, we used integrase inhibitors. To discriminate between the cytopathic response induced by endogenous termination of reverse transcription and the response induced by AZT, we separately assessed resting and activated CFSE-labeled cells. Remarkably, the NNRTIs, efavirenz and nevirapine, blocked indirect killing of resting CD4 T cells as efficiently as AMD3100 (Figure 5D, columns 15 and 16). These findings suggested that allosteric inhibition of reverse transcriptase induced by these NNRTI’s interrupts reverse transcription sufficiently early to abrogate the death response. In contrast, the integrase inhibitors raltegravir and 118-D-24 did not prevent abortive infection killing (Figure 5D, columns 17 and 18), suggesting that cell death involves signals generated prior to viral integration. NNRTIs also protected activated CFSE-labeled cells from death induced by AZT (Figure 5D, column 38 versus columns 44 and 45),
demonstrating that a certain degree of DNA synthesis is required to elicit the cytopathic response. This notion was further strengthened in findings obtained with vif-deficient (Dvif) HIV-1 particles where reverse transcription is inhibited during strong-stop DNA synthesis due to incorporated APOBEC3G (A3G) (Bishop et al., 2008; Li et al., 2007). Abortively infected CD4 T cells were not depleted by Dvif NL4-3-infected cells (Figures S4C and S4D), indicating that termination of reverse transcription before the completion of strong-stop DNA synthesis is not sufficient to generate a cytopathic response. Other HIV-1 mutants containing substitutions in RNase H and nucleocapsid that promote early defects in reverse transcription failed to elicit indirect CD4 T cell killing (Figures S4E and S4F). Together, these findings indicate that accumulation of reverse-transcribed DNA, rather than any inherent activity of the HIV-1 proteins, is the key factor that triggers the death response. Abortively Infected CD4 T Cells Commence but Do Not Complete Reverse Transcription We next examined the status of HIV-1 reverse transcription in tonsillar CD4 T cells after infection. Specifically, we investigated the effect on reverse transcription after treatment with NNRTIs, such as efavirenz and nevirapine, which prevent the death of abortively infected CD4 T cells, or with AZT or integrase inhibitor (raltegravir) that do not prevent CD4 T cell death. Taqman-based quantitative real-time PCR (QPCR) was used to quantify the synthesis of reverse transcription products in isolated CD4 T cells from HLAC 16 hr after infection with NL4-3. We designed specific QPCR primers and probes (Table S1) to monitor sequential steps in reverse transcription including generation of strong-stop DNA, first template exchange (Nef), and DNA strand elongation (Env) (Figure 6A). Reverse transcription products corresponding to strong-stop DNA were similar in untreated CD4 T cells or cells treated with AZT, NNRTIs, or raltegravir but were greatly reduced in cells treated with AMD3100 or in cultures infected with Dvif NL4-3 where arrest occurs prior to the completion of strong-stop DNA synthesis (Figure 6B columns 1–8). In contrast, the accumulation of later reverse transcription products detected by the Nef and Env probes were dramatically inhibited by the NNRTIs but not by raltegravir. Levels of Nef (Figure 6B, columns 10 and 11) and Env (columns 18 and 19) DNA products were similar in untreated cells and cells treated with AZT, indicating that reverse transcription in most tonsillar CD4 T cells naturally terminates during DNA chain elongation, coinciding with the block induced by AZT. The minor inhibition detected by AZT is likely due to a small number of permissive CD4 T cells in the culture. These results show that abortively infected CD4 T cells accumulate incomplete reverse
AZT (5 mM), AMD3100 (250 nM), and efavirenz (100 nM) as indicated. After 3 days, half of the cells were lysed and subjected to SDS-PAGE immunobloting analysis. On day 5, the supernatants from the rest of the cells were collected and subjected to SDS-PAGE immunobloting analysis. The IL-1b antibody detects the pro-IL-1b (37kD) and the mature cleaved form (17kD). Data are the representative results of five independent experiments using tonsillar CD4 T cells isolated from five different donors. (G) DNA reverse transcription intermediates induce an IFN-stimulatory antiviral innate immune response (ISD). ISRE-GFP reporters were transfected with 1 mg of HIV-1 reverse transcription intermediate products as indicated by numbers (detailed description in Figure S5E), empty DNA plasmid, or polyinosinic:polycytidylic acid [poly(I:C)], and were analyzed by flow cytometry after 48 hr. Data are representative of three independent experiments; error bars show the SD for three independent samples from the same experiment. See also Figure S5 and Figure S6.
Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc. 797
transcription products representative of DNA strand elongation. Blocking earlier steps of reverse transcription by NNRTIs or by genetic mutations like deletion of Vif or mutation of RNase H restricts accumulation of such products, and prevents abortive infection-induced cell death (Figure S6A). DNA Reverse Transcription Intermediates Elicit a Coordinated Proapoptotic and Proinflammatory Response in Abortively Infected CD4 T Cells We next evaluated whether HIV-mediated indirect killing of CD4 T cells is associated with deregulation of cytokine production or a DNA damage response. To facilitate a vigorous and synchronized killing effect, HLACs were spinoculated with NL4-3 virions in the presence of various antiviral drugs. Interestingly, based on immunostaining after cytokine capture, abortively infected CD4 T cells expressed IFN-b, and high levels of the proinflammatory interleukin 1b (IL–1b), but not tumor necrosis factor (TNFa) (Figure 6C). Phosphorylation of S37 p53 was not observed, suggesting that abortive HIV-1 infection does not induce a DNA damage cascade. Abortively infected CD4 T cells also displayed caspase-1 and caspase-3 activity along with appearance of annexin V (Figure 6D). T20 and efavirenz but not AZT prevented activation of these caspases, indicating that apoptosis was induced by abortive HIV-1 infection. Cell death was completely prevented by Z-VAD-FMK, a pan-caspase inhibitor, suggesting that caspase activation is required for the observed cytopathic response (Figure 6E). Such mode of cytokine production and caspase activation was not observed in CD8 T or B cells (Figures S5B and S5C). We next examined whether abortive HIV-1 infection signals for the maturation and secretion of IL-1b. In cells, IL–1b activity is rigorously controlled. Cells can be primed to express inactive pro-IL-1b by various proinflammatory signals. However, the release of bioactive IL-1b requires a second signal leading to activation of inflammasomes, cleavage of pro-IL-1b by caspase 1 and secretion of the bioactive 17 kDa form of IL-1b (Schroder and Tschopp, 2010). Interestingly, Western blot analysis revealed high amounts of intracellular pro-IL-1b in untreated CD4 T cells, suggesting that tonsillar CD4 T cells are primed to release proinflammatory mediators (Figure 6F). Stimulating the CD4 T cells with PMA and nigericin induced further accumulation of pro-IL-1b and promoted the maturation and release of the bioactive 17 kDa IL-1b into the supernatant. Remarkably, infection of CD4 T cells with NL4-3 in the presence of AZT similarly resulted in maturation and release of the bioactive 17 kDa IL-1b into the supernatant. This response was completely prevented by efavirenz and AMD3100, suggesting that abortive HIV-1 infection signals the maturation and release of bioactive IL-1b in these CD4 T cells. To identify the nature of the nucleic acid species that trigger these responses, we used a recently described H35 rat hepatocyte cell line containing an IFN-sensitive response element (ISRE) linked to GFP (Patel et al., 2009). H35 cells were first infected with pseudotyped VSV-G HIV-1 virions. These virions induced GFP expression and cell death in the presence or absence of AZT. Importantly, the expression of GFP and cell death response were blocked by efavirenz but not raltegravir (Figure S5D). Thus, the H35 system successfully reconstitutes 798 Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc.
the cytokine and cytopathic response observed in tonsillar CD4 T cells. We next synthesized the various HIV-1 reverse transcription intermediates and tested their ability to activate the ISRE-GFP reporter. Interestingly, none of the RNA-containing oligonucleotides stimulated the ISRE-GFP reporter expression above baseline. In sharp contrast, ssDNA and dsDNA oligonucleotides longer than 500 bases in length, which corresponded to reverse transcription intermediates produced during DNA elongation, evoked a potent ISRE-GFP activation (Figure 6G). Similarly, when cells were stimulated with poly(I:C), a synthetic double-stranded RNA known to activate IRF3 via the RIG-I pathway elicited a comparable ISRE-GFP response. Taken together, these findings indicate that reverse transcription intermediates generated during DNA chain elongation induce a coordinated proapoptotic and proinflamatory innate immune response involving caspase-3 and caspase-1 activation in abortively infected CD4 T cells. DISCUSSION The mechanism through which HIV-1 kills CD4 T cells, a hallmark of AIDS, has been a topic of vigorous research and one of the most pressing questions for the field over the last 28 years (Thomas, 2009). In this study, we investigated the mechanism of HIV-1-mediated killing in lymphoid tissues, which carry the highest viral burdens in infected patients. We used HLACs formed with fresh human tonsil cells and an experimental strategy that clearly distinguishes between direct and indirect mechanisms of CD4 T cell depletion. We now demonstrate that indirect cell killing involving abortive HIV infection of CD4 T cells accounts for the vast majority of cell death occurring in lymphoid tissues. No more than 5% of the CD4 T cells are productively infected, but virtually all the remaining CD4 T cells are abortively infected ultimately leading to caspase-mediated cell death. Equivalent findings were observed in HLACs formed with fresh human spleen (Figures S6B and S6C), indicating this mechanism of CD4 T cell depletion can be generalized to other lymphoid tissues. The massive depletion of nonproductively infected CD4 T cells is in contrast to their survival after infection of intact blocks of tonsillar tissue in human lymphoid histoculture (HLH) (Grivel et al., 2003). This result probably reflects differences between the HLH and the HLAC experimental systems. In HLH, the complex three-dimensional spatial cellular organization of lymphoid tissue is preserved, but cellular movement and interaction are restricted, both of which are required for indirect killing. In HLAC, the tissue is dispersed, and cells are free to interact, resulting in a rapid and robust viral spread. While the mechanism triggering indirect CD4 T cell death is certainly identical in both settings, HLH allows only a slow, nearly undetectable progression of indirect CD4 T cell death. In HLAC, this process is accelerated, allowing the outcome to be detected in a few days. Interestingly, indirect killing was also less efficient when peripheral blood cells were tested (data not shown). It is possible that cellular factors specifically produced in lymphoid organs are required to accelerate indirect killing of peripheral blood CD4 T cells. Several mechanisms have been proposed to explain indirect CD4 T cell killing during HIV infection. Our finding that CD4
Figure 7. Consequences of Inhibiting Early Steps of HIV-1 Infection on CD4 T Cell Death (A) The nonpermissive state of most CD4 T cells in lymphoid tissue leads to endogenous termination of reverse transcription during DNA chain elongation (i.e., ‘‘killing zone’’). As a result, DNA intermediates accumulate in the cytoplasm and elicit a multifaceted proapoptotic and proinflammatory innate immune defense programs, coordinated by IFN-stimulatory DNA (ISD) response, caspase-3, caspase-1, and IL-1b, to restrict viral spread. Different classes of antiretroviral drugs act at different stage of the HIV life cycle. NNRTIs like efavirenz and nevirapine inhibit early steps of DNA synthesis and therefore prevent such response and the consequence CD4 T cell death. AZT is less efficient at blocking DNA synthesis and therefore unable to abrogate this response. (B) In permissive CD4 T cells reverse transcription proceeds, allowing HIV-1 to bypass the ‘‘killing zone’’ and move on to productive (or latent) infection. Interrupting reverse transcription with AZT traps the virus in the ‘‘killing zone’’ and induces cell death. EFV, efavirenz; NVP, nevirapine; and RTGR, raltegravir. See also Figure S6.
T cell death is blocked by entry and fusion inhibitors but not by AZT, strongly suggested that such killing involves nonproductive infection of CD4 T cells. Therefore, we focused on events that occur after HIV-1 entry. Our investigations demonstrate that abortive viral DNA synthesis occurring in nonpermissive, quiescent CD4 tonsil T cells, plays a key role in the cell death response. Conversely, in the small subset of permissive target cells, reverse transcription is not interrupted, minimizing the accumulation and subsequent detection of such reverse transcription intermediates (Figure 7). Interrupted or slowed reverse transcription may create persistent exposure to cytoplasmic DNA products that elicit an antiviral innate immune response coordinated by activation of type I IFNs (Stetson and Medzhitov, 2006). Such activation, termed IFN-stimulatory DNA (ISD) response, may be analogous to the type I IFN response triggered by the RIG-I-like receptor (RLR) family of RNA helicases that mediate a cell-intrinsic antiviral
defense (Rehwinkel and Reis e Sousa, 2010). Our results suggest that abortive HIV-1 infection also stimulates activation of caspase-3, which is linked to apoptosis, and caspase-1, which promotes the processing and secretion of the proinflammatory cytokines like IL–1b. It is certainly possible that pyroptosis elicited in response to caspase-1 activation also contributes to the observed cytopathic response (Schroder and Tschopp, 2010). The release of inflammatory cytokines during CD4 T cell death could also contribute to the state of chronic inflammation that characterizes HIV infection. This inflammation may fuel further viral spread by recruiting uninfected lymphocytes to the inflamed zone. While this innate response was likely designed to protect the host, it is subverted in the case of HIV infection and importantly contributes to the immunopathogenic effects characteristic of HIV infection and AIDS. Such antiviral pathways comprise an unrecognized cellintrinsic retroviral detection system (Manel et al., 2010; Stetson Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc. 799
et al., 2008). Viral RNA in infected cells is recognized by members of the RIG-I-like family of receptors that detect specific RNA patterns like uncapped 50 triphosphate (Rehwinkel and Reis e Sousa, 2010). Although uncapped RNA intermediates are generated by the HIV-1 RNase H, they contain a 50 monophosphate and therefore may be not recognized by the RIG-I system (Figure 6G). In contrast to RNA receptors, intracellular sensing of viral DNA remains poorly understood. Consequently, it is unclear how HIV-1 DNA intermediates are detected in the cytoplasm of abortively infected CD4 T cells. AIM2 (absent in melanoma 2) was recently identified as a cytoplasmic dsDNA receptor that induces cell death in macrophages through activation of caspase-1 in imflammasomes (Hornung et al., 2009). Our preliminary investigations have not supported a role for AIM2 in cell death induced by abortive HIV infection (data not shown), suggesting the potential involvement of a different DNA-sensing mechanism. We also have not identified a role for TLR9 and MYD88 signaling in this form of cell death. Additional candidate sensors recognizing cytoplasmic HIV-1 DNA are now under study. In summary, both productive and nonproductive forms of HIV infection contribute to the pathogenic effects of this lentivirus. The relative importance of these different cell death pathways might well vary with the stage of HIV infection. For example, direct infection and death might predominate during acute infection where CCR5-expressing memory CD4 T cells in gutassociated lymphoid tissue are effectively depleted. Conversely, the CXCR4-dependent indirect killing we describe in tonsil tissue may reflect later stages of HIV-induced disease where a switch to CXCR4 coreceptor usage occurs in approximately 50% of infected subjects. The current study demonstrates how a cytopathic response involving abortive viral infection of resting nonpermissive CD4 T cells can lead not only to CD4 T cell depletion but also to the release of proinflammatory cytokines. The ensuing recruitment of new target cells to the site of inflammation may fuel a vicious cycle of continuing infection and CD4 T cell death centrally contributing to HIV pathogenesis.
QPCR reactions were performed in an ABI Prism 7900HT (Applied Biosystems).
EXPERIMENTAL PROCEDURES
REFERENCES
Culture and Infection of HLACs Human tonsil or splenic tissues were obtained from the National Disease Research Interchange and the Cooperative Human Tissue Network and processed as previously described (Jekle et al., 2003). For a detailed description see Extended Experimental Procedures.
Arts, E.J., and Wainberg, M.A. (1994). Preferential incorporation of nucleoside analogs after template switching during human immunodeficiency virus reverse transcription. Antimicrob. Agents Chemother. 38, 1008–1016.
FACS Analysis and Gating Strategy, Preparation of HIV-1 Virions, and Virion-Based Fusion Assay Data were collected on a FACS Calibur (BD Biosciences) and analyzed with Flowjo software (Treestar). HIV-1 viruses were generated by transfection of proviral DNA into 293T cells by the calcium phosphate method. Virion-based fusion assay was performed as previously described (Cavrois et al., 2002). Detailed protocols are provided in the supplemental experimental procedures. Spinoculation and Taqman-Based QPCR Analysis of HIV-1-Infected CD4 T Cells The spinoculation method is described in detail in Figures S3A and S3B. Isolation of HLAC CD4 T cells and QPCR protocol are described in detail in supplemental experimental procedures. Primers and probes sequences used to detect reverse transcription products are provided in Table S1.
800 Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc.
ISRE-GFP H35 Reporter Cells, Microscopy, and Generation of Synthetic HIV-1 Reverse Transcription Intermediates H35 rat hepatic cells containing an ISRE-GFP reporter were maintained as described (Patel et al., 2009). For microscopy imaging, ISRE-GFP reporter H35 cells were infected with a replication competent VSV-G pseudotyped NL4-3 and analyzed using an Axio observer Z1 microscope (Zeiss). Transfections and generation of synthetic HIV-1 reverse transcription intermediates are described in detail in Figure S5E and Extended Experimental Procedures. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, six figures, and one table and can be found with this article online at doi:10.1016/j.cell.2010.11.001. ACKNOWLEDGMENTS We thank David N. Levy for the NLENG1 plasmid; David Fenard for the NL4-3 variant plasmids SIM, GIA, GIA-SKY, and SKY; George M. Shaw for the WEAU 16-8 env clone; and Suraj J. Patel, Kevin R. King, and Martin L. Yarmush for the H35 ISRE-GFP reporter cell line. The following reagents were obtained through the AIDS Research and Reference Reagent Program, Division of AIDS, NIAID, NIH: AMD3100, T-20, Saquinavir, Amprenavir, Indinavir, Nevirapine, Efavirenz, and AZT-resistant HIV-1 clones #629 and #964. Special thanks to Dr. Eva Herker for assistance with fluorescence microscopy; to Dr. Stefanie Sowinski for help with assessing inflammatory responses in primary immune cells; and to Jason Neidleman for stimulating discussions and technical advice. We also thank Marty Bigos for assistance with the flow cytometry; Stephen Ordway and Gary Howard for editorial assistance; John C.W. Carroll and Alisha Wilson for graphics; and Robin Givens and Sue Cammack for administrative assistance. Funding for this project was provided by the Universitywide AIDS Research Program, F04-GIVI-210 (G.D.); the UCSF-GIVI Center for AIDS Research, NIH/NIAID P30 AI027763 (M.C.); the Francis Goelet Fellowship (K.G.L.); and the UCSF Medical Scientist Training Program, NIH/NIGMS T32 GM007618-32 (O.Z.). Received: November 5, 2009 Revised: May 7, 2010 Accepted: October 29, 2010 Published: November 24, 2010
Baldwin, C.E., Sanders, R.W., Deng, Y., Jurriaans, S., Lange, J.M., Lu, M., and Berkhout, B. (2004). Emergence of a drug-dependent human immunodeficiency virus type 1 variant during therapy with the T20 fusion inhibitor. J. Virol. 78, 12428–12437. Bishop, K.N., Verma, M., Kim, E.Y., Wolinsky, S.M., and Malim, M.H. (2008). APOBEC3G inhibits elongation of HIV-1 reverse transcripts. PLoS Pathog. 4, e1000231. Cavrois, M., De Noronha, C., and Greene, W.C. (2002). A sensitive and specific enzyme-based assay detecting HIV-1 virion fusion in primary T lymphocytes. Nat. Biotechnol. 20, 1151–1154. Eckstein, D.A., Penn, M.L., Korin, Y.D., Scripture-Adams, D.D., Zack, J.A., Kreisberg, J.F., Roederer, M., Sherman, M.P., Chin, P.S., and Goldsmith, M.A. (2001). HIV-1 actively replicates in naive CD4(+) T cells residing within human lymphoid tissues. Immunity 15, 671–682. Finkel, T.H., Tudor-Williams, G., Banda, N.K., Cotton, M.F., Curiel, T., Monks, C., Baba, T.W., Ruprecht, R.M., and Kupfer, A. (1995). Apoptosis occurs
predominantly in bystander cells and not in productively infected cells of HIVand SIV-infected lymph nodes. Nat. Med. 1, 129–134. Gandhi, R.T., Chen, B.K., Straus, S.E., Dale, J.K., Lenardo, M.J., and Baltimore, D. (1998). HIV-1 directly kills CD4+ T cells by a Fas-independent mechanism. J. Exp. Med. 187, 1113–1122. Garg, H., Joshi, A., Freed, E.O., and Blumenthal, R. (2007). Site-specific mutations in HIV-1 gp41 reveal a correlation between HIV-1-mediated bystander apoptosis and fusion/hemifusion. J. Biol. Chem. 282, 16899–16906. Glushakova, S., Baibakov, B., Margolis, L.B., and Zimmerberg, J. (1995). Infection of human tonsil histocultures: a model for HIV pathogenesis. Nat. Med. 1, 1320–1322. Grivel, J.C., Biancotto, A., Ito, Y., Lima, R.G., and Margolis, L.B. (2003). Bystander CD4+ T lymphocytes survive in HIV-infected human lymphoid tissue. AIDS Res. Hum. Retroviruses 19, 211–216. Hayden, M.S., Palacios, E.H., and Grant, R.M. (2003). Real-time quantitation of HIV-1 p24 and SIV p27 using fluorescence-linked antigen quantification assays. AIDS 17, 629–631. Herbeuval, J.P., Grivel, J.C., Boasso, A., Hardy, A.W., Chougnet, C., Dolan, M.J., Yagita, H., Lifson, J.D., and Shearer, G.M. (2005). CD4+ T-cell death induced by infectious and noninfectious HIV-1: role of type 1 interferon-dependent, TRAIL/DR5-mediated apoptosis. Blood 106, 3524–3531. Holm, G.H., and Gabuzda, D. (2005). Distinct mechanisms of CD4+ and CD8+ T-cell activation and bystander apoptosis induced by human immunodeficiency virus type 1 virions. J. Virol. 79, 6299–6311. Holm, G.H., Zhang, C., Gorry, P.R., Peden, K., Schols, D., De Clercq, E., and Gabuzda, D. (2004). Apoptosis of bystander T cells induced by human immunodeficiency virus type 1 with increased envelope/receptor affinity and coreceptor binding site exposure. J. Virol. 78, 4541–4551. Hornung, V., Ablasser, A., Charrel-Dennis, M., Bauernfeind, F., Horvath, G., Caffrey, D.R., Latz, E., and Fitzgerald, K.A. (2009). AIM2 recognizes cytosolic dsDNA and forms a caspase-1-activating inflammasome with ASC. Nature 458, 514–518, Epub 2009 Jan 2021. Jekle, A., Keppler, O.T., De Clercq, E., Schols, D., Weinstein, M., and Goldsmith, M.A. (2003). In vivo evolution of human immunodeficiency virus type 1 toward increased pathogenicity through CXCR4-mediated killing of uninfected CD4 T cells. J. Virol. 77, 5846–5854. Kamata, M., Nagaoka, Y., and Chen, I.S. (2009). Reassessing the role of APOBEC3G in human immunodeficiency virus type 1 infection of quiescent CD4+ T-cells. PLoS Pathog 5, e1000342. Kreisberg, J.F., Yonemoto, W., and Greene, W.C. (2006). Endogenous factors enhance HIV infection of tissue naive CD4 T cells by stimulating high molecular mass APOBEC3G complex formation. J. Exp. Med. 203, 865–870. LaBonte, J.A., Patel, T., Hofmann, W., and Sodroski, J. (2000). Importance of membrane fusion mediated by human immunodeficiency virus envelope glycoproteins for lysis of primary CD4-positive T cells. J. Virol. 74, 10690– 10698. Larder, B.A., Darby, G., and Richman, D.D. (1989). HIV with reduced sensitivity to zidovudine (AZT) isolated during prolonged therapy. Science 243, 1731– 1734.
Margolis, L.B., Glushakova, S., Baibakov, B., and Zimmerberg, J. (1995). Syncytium formation in cultured human lymphoid tissue: fusion of implanted HIV glycoprotein 120/41-expressing cells with native CD4+ cells. AIDS Res. Hum. Retroviruses 11, 697–704. Patel, S.J., King, K.R., Casali, M., and Yarmush, M.L. (2009). DNA-triggered innate immune responses are propagated by gap junction communication. Proc. Natl. Acad. Sci. USA 106, 12867–12872. Rehwinkel, J., and Reis e Sousa, C. (2010). RIGorous detection: exposing virus through RNA sensing. Science 327, 284–286. Rimsky, L.T., Shugars, D.C., and Matthews, T.J. (1998). Determinants of human immunodeficiency virus type 1 resistance to gp41-derived inhibitory peptides. J. Virol. 72, 986–993. Santoni de Sio, F.R., and Trono, D. (2009). APOBEC3G-depleted resting CD4+ T cells remain refractory to HIV1 infection. PLoS ONE 4, e6571. Schindler, M., Mu¨nch, J., Kutsch, O., Li, H., Santiago, M.L., Bibollet-Ruche, F., Mu¨ller-Trutwin, M.C., Novembre, F.J., Peeters, M., Courgnaud, V., et al. (2006). Nef-mediated suppression of T cell activation was lost in a lentiviral lineage that gave rise to HIV-1. Cell 125, 1055–1067. Schroder, K., and Tschopp, J. (2010). The inflammasomes. Cell 140, 821–832. Sherer, N.M., Lehmann, M.J., Jimenez-Soto, L.F., Horensavitz, C., Pypaert, M., and Mothes, W. (2007). Retroviruses can establish filopodial bridges for efficient cell-to-cell transmission. Nat. Cell Biol. 9, 310–315. Sourisseau, M., Sol-Foulon, N., Porrot, F., Blanchet, F., and Schwartz, O. (2007). Inefficient human immunodeficiency virus replication in mobile lymphocytes. J. Virol. 81, 1000–1012. Stetson, D.B., Ko, J.S., Heidmann, T., and Medzhitov, R. (2008). Trex1 prevents cell-intrinsic initiation of autoimmunity. Cell 134, 587–598. Stetson, D.B., and Medzhitov, R. (2006). Recognition of cytosolic DNA activates an IRF3-dependent innate immune response. Immunity 24, 93–103. Swiggard, W.J., O’Doherty, U., McGain, D., Jeyakumar, D., and Malim, M.H. (2004). Long HIV type 1 reverse transcripts can accumulate stably within resting CD4+ T cells while short ones are degraded. AIDS Res. Hum. Retroviruses 20, 285–295. Thomas, C. (2009). Roadblocks in HIV research: five questions. Nat. Med. 15, 855–859. Unutmaz, D., KewalRamani, V.N., Marmon, S., and Littman, D.R. (1999). Cytokine signals are sufficient for HIV-1 infection of resting human T lymphocytes. J. Exp. Med. 189, 1735–1746. Vlahakis, S.R., Algeciras-Schimnich, A., Bou, G., Heppelmann, C.J., VillasisKeever, A., Collman, R.C., and Paya, C.V. (2001). Chemokine-receptor activation by env determines the mechanism of death in HIV-infected and uninfected T lymphocytes. J. Clin. Invest. 107, 207–215. Westendorp, M.O., Frank, R., Ochsenbauer, C., Stricker, K., Dhein, J., Walczak, H., Debatin, K.M., and Krammer, P.H. (1995). Sensitization of T cells to CD95-mediated apoptosis by HIV-1 Tat and gp120. Nature 375, 497–500.
Levy, D.N., Aldrovandi, G.M., Kutsch, O., and Shaw, G.M. (2004). Dynamics of HIV-1 recombination in its natural target cells. Proc. Natl. Acad. Sci. USA 101, 4204–4209.
Wyma, D.J., Jiang, J., Shi, J., Zhou, J., Lineberger, J.E., Miller, M.D., and Aiken, C. (2004). Coupling of human immunodeficiency virus type 1 fusion to virion maturation: a novel role of the gp41 cytoplasmic tail. J. Virol. 78, 3429–3435.
Li, X.Y., Guo, F., Zhang, L., Kleiman, L., and Cen, S. (2007). APOBEC3G inhibits DNA strand transfer during HIV-1 reverse transcription. J. Biol. Chem. 282, 32065–32074.
Zack, J.A., Arrigo, S.J., Weitsman, S.R., Go, A.S., Haislip, A., and Chen, I.S. (1990). HIV-1 entry into quiescent primary lymphocytes: molecular analysis reveals a labile, latent viral structure. Cell 61, 213–222.
Manel, N., Hogstad, B., Wang, Y., Levy, D.E., Unutmaz, D., and Littman, D.R. (2010). A cryptic sensor for HIV-1 activates antiviral innate immunity in dendritic cells. Nature 467, 214–217.
Zhou, Y., Zhang, H., Siliciano, J.D., and Siliciano, R.F. (2005). Kinetics of human immunodeficiency virus type 1 decay following entry into resting CD4+ T cells. J. Virol. 79, 2199–2210.
Cell 143, 789–801, November 24, 2010 ª2010 Elsevier Inc. 801
Sirt3 Mediates Reduction of Oxidative Damage and Prevention of Age-Related Hearing Loss under Caloric Restriction Shinichi Someya,1,3,5 Wei Yu,2,5 William C. Hallows,2 Jinze Xu,4 James M. Vann,1 Christiaan Leeuwenburgh,4 Masaru Tanokura,3 John M. Denu,2,* and Tomas A. Prolla1,* 1Departments
of Genetics and Medical Genetics of Biomolecular Chemistry University of Wisconsin, Madison, WI 53706, USA 3Department of Applied Biological Chemistry, University of Tokyo, Yayoi, Tokyo 113-8657, Japan 4Department of Aging and Geriatrics and The Institute on Aging, University of Florida, Gainesville, FL 32611, USA 5These authors contributed equally to this work *Correspondence:
[email protected] (J.M.D.),
[email protected] (T.A.P.) DOI 10.1016/j.cell.2010.10.002 2Department
SUMMARY
Caloric restriction (CR) extends the life span and health span of a variety of species and slows the progression of age-related hearing loss (AHL), a common age-related disorder associated with oxidative stress. Here, we report that CR reduces oxidative DNA damage in multiple tissues and prevents AHL in wild-type mice but fails to modify these phenotypes in mice lacking the mitochondrial deacetylase Sirt3, a member of the sirtuin family. In response to CR, Sirt3 directly deacetylates and activates mitochondrial isocitrate dehydrogenase 2 (Idh2), leading to increased NADPH levels and an increased ratio of reduced-to-oxidized glutathione in mitochondria. In cultured cells, overexpression of Sirt3 and/or Idh2 increases NADPH levels and protects from oxidative stress-induced cell death. Therefore, our findings identify Sirt3 as an essential player in enhancing the mitochondrial glutathione antioxidant defense system during CR and suggest that Sirt3-dependent mitochondrial adaptations may be a central mechanism of aging retardation in mammals. INTRODUCTION It is well established that reducing food consumption by 25%– 60% without malnutrition consistently extends both the mean and maximum life span of rodents (Weindruch and Walford, 1988; Koubova and Guarente, 2003). Caloric restriction (CR) is also known to extend life span in yeast, worms, fruit flies, spiders, birds, and monkeys and delays the progression of a variety of age-associated diseases such as cancer, diabetes, cataract, and age-related hearing loss (AHL) in mammals (Wein802 Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc.
druch and Walford, 1988; Sohal and Weindruch, 1996; Someya et al., 2007; Colman et al., 2009). Furthermore, CR reduces neurodegeneration in animal models of Parkinson’s disease (Mattson, 2000) as well as Alzheimer’s disease (Zhu et al., 1999). The mitochondrial free radical theory of aging postulates that aging results from accumulated oxidative damage caused by reactive oxygen species (ROS), originating from the mitochondrial respiratory chain (Balaban et al., 2005). Consistent with this hypothesis, mitochondria are a major source of ROS and of ROS-induced oxidative damage, and mitochondrial function declines during aging (Wallace, 2005). A large body of evidence suggests that CR reduces the age-associated accumulation of oxidatively damaged proteins, lipids, and DNA through reduction of oxidative damage to these macromolecules and/or enhanced antioxidant defenses to oxidative stress (Weindruch and Walford, 1988; Sohal and Weindruch, 1996; Masoro, 2000). Yet, whether the anti-aging action of CR in mammals is a regulated process and requires specific regulatory proteins such as sirtuins still remains unclear. Sirtuins are NAD+-dependent protein deacetylases that regulate life span in lower organisms and have emerged as broad regulators of cellular fate and mammalian physiology (Donmez and Guarente, 2010; Finkel et al., 2009). A previous report has shown that life span extension by CR in yeast requires Sir2, a member of the sirtuin family (Lin et al., 2000), linking sirtuins and CR-mediated retardation of aging. In mammals, there are seven sirtuins that display diverse cellular localization (Donmez and Guarente, 2010; Finkel et al., 2009). Previous studies have focused on the role of Sirt1 as the major sirtuin mediating the metabolic effects of CR in mammals (Chen et al., 2005; Bordone et al., 2007; Chen et al., 2008). However, recent studies indicate that upregulation of Sirt1 in response to CR is not observed in all tissues examined (Cohen et al., 2004; Barger et al., 2008), and currently, no study has provided conclusive evidence that sirtuins play an essential role in CR-mediated aging retardation in mammals. Sirt3 is a member of the mammalian sirtuin family that is localized to mitochondria and
regulates levels of ATP and the activity of complex I of the electron transport chain (Ahn et al., 2008) and, as such, may play a role in the metabolic reprogramming mediated by CR. A recent study has shown that CR increases Sirt3 levels in liver mitochondria (Schwer et al., 2009). Fasting also increases Sirt3 protein expression in liver mitochondria, and mice lacking Sirt3 display the hallmarks of fatty acid oxidation disorders, indicating that Sirt3 modulates mitochondrial fatty acid oxidation in mammals (Hirschey et al., 2010). Furthermore, CR increases expression of Sirt3 in primary mouse cardiomyocytes, whereas overexpression of Sirt3 protects these cells from oxidative stress-induced cell death (Sundaresan et al., 2008), suggesting a potential role of Sirt3 in the aging retardation associated with CR in mammals. AHL is a universal feature of mammalian aging and is the most common sensory disorder in the elderly (Someya and Prolla, 2010; Liu and Yan, 2007). AHL is characterized by an age-associated decline of hearing function associated with loss of spiral ganglion neurons and sensory hair cells in the cochlea of the inner ear (Someya and Prolla, 2010; Liu and Yan, 2007). The progressive loss of neurons and hair cells in the inner ear leads to the onset of AHL because these postmitotic cells do not regenerate in mammals. The onset of AHL begins in the high-frequency region and spreads toward the low-frequency region during aging (Keithley et al., 2004; Hunter and Willott, 1987). This is accompanied by the loss of neurons and hair cells beginning in the basal region and spreading toward the apex of the cochlea of the inner ear with age. A previous study has shown that CR slows the progression of AHL in CBA/J mice (Sweet et al., 1988), whereas we have shown previously that CR prevents AHL in C57BL/6J mice, reduces cochlear degeneration, and induces Sirt3 in the cochlea (Someya et al., 2007). Both strains of mice have been extensively used as a model of AHL, although the age of onset of AHL varies from 12–15 months of age in C57BL/6J mice to 18–22 months of age in CBA/J mice (Zheng et al., 1999). Experimental evidence suggests that oxidative stress plays a major role in AHL (Jiang et al., 2007; Someya et al., 2009) and that CR protects cochlear cells through reduction of oxidative damage and/or by enhancing cellular antioxidant defenses to oxidative stress (Someya et al., 2007). Yet, the molecular mechanisms by which CR reduces oxidative cochlear cell damage remain unknown. In this report, we show that the mitochondrial deacetylase Sirt3 is required for the CR-mediated prevention of AHL in mice. We also show that Sirt3 is required for the reduction of oxidative damage in multiple tissues under CR conditions, as evidenced by DNA damage levels. At the mechanistic level, Sirt3 directly deacetylates isocitrate dehydrogenase 2 (Idh2), an enzyme that converts NADP+ to NADPH in mitochondria. In response to CR, Sirt3 stimulates Idh2 activity in mitochondria, leading to increased levels of NADPH and an increased ratio of reduced glutathione/oxidized glutathione, the major redox couple in the cell. In cultured cells, overexpression of Sirt3 and/or Idh2 increases NADPH levels and protects these cells from oxidative stress. The data presented here provide the first conclusive evidence that CR-mediated reduction of oxidative damage and prevention of a common age-related
phenotype (AHL) require a member of the sirtuin family in mammals. RESULTS Sirt3 Is Required for the CR-Mediated Prevention of Age-Related Cochlear Cell Death and Hearing Loss First, to investigate whether Sirt3 plays a role in the CR prevention of AHL, we conducted a 10 month CR dietary study using WT and Sirt3/ mice that have been backcrossed onto the C57BL/6J background. The C57BL/6J strain is considered an excellent model to study the anti-aging action of CR because this mouse strain is the most widely used mouse model for the study of aging and responds to CR with a robust extension of life span (Weindruch and Walford, 1988) and prevention of AHL (Someya et al., 2007). We reduced the calorie intake of WT and Sirt3/ mice to 75% (a 25% CR) of that fed to control diet (CD) mice in early adulthood (2 months of age), and this dietary regimen was maintained until 12 months of age. The auditory brainstem response (ABR), a common electrophysiological test of hearing function, was used to monitor the progression of AHL in these mice (Someya et al., 2009). We first confirmed that aging resulted in increased ABR hearing thresholds at the high (32 kHz), middle (16 kHz), and low (8 kHz) frequencies in 12-month-old WT mice (Figure 1A), indicating that these mice displayed hearing loss. As predicted, CR delayed the progression of AHL at all tested frequencies in WT mice (Figure 1A). Strikingly, CR did not delay the progression of AHL in Sirt3/ mice (Figure 1A), although CR had the same effect on body weight reduction in both WT and Sirt3/ mice (Figures S2A and S2B available online). Neural and hair cell degeneration are hallmarks of AHL (Keithley et al., 2004). In agreement with the hearing test results, basal regions of the cochleae from calorie-restricted WT mice displayed only minor loss of spiral ganglion neurons (Figures 1J and 1K; see also Figures 1B, 1C, 1F and 1G) and hair cells (Figure S1E; see also Figures S1A and S1C), whereas CR failed to protect these cells in Sirt3/ mice (Figures 1L and 1M; see also Figures 1D, 1E, 1H, and 1I; Figure S1F; see also Figures S1B and S1D). Collectively, these results demonstrate that Sirt3 plays an essential role in the CRmediated prevention of age-related cochlear cell death and hearing loss in mice. Next, to investigate whether Sirt3 plays a role in the metabolic effects induced by CR, we conducted a 3 month CR dietary study using WT and Sirt3/ mice starting at 2 months of age. Mice lacking the Sirt3 gene appeared phenotypically normal under basal and CR conditions: Sirt3/ mice were viable and fertile, and no significant changes were observed in body weight (Figures S2A and S2B), bone mineral density (Figure S2C), body fat (Figure S2D), tissue weight (Figure S2E), serum glucose levels (Figure S3A), glucose tolerance (Figure S3B), serum Igf-1 (Figure S3C), and cholesterol (Figure S3D) levels between control diet WT and Sirt3/ mice or calorierestricted WT and Sirt3/ mice at 5 months of age. However, though we found that WT mice displayed lower levels of serum insulin (Figure S3E) and triglycerides (Figure S3F) in response to CR, no significant changes were observed in these serum markers between control diet-fed and calorie-restricted Sirt3/ Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc. 803
A
2mo CD 12mo CD 12mo CR
100 Threshold (dB SPL)
Sirt3-/-
WT
2mo CD 12mo CD 12mo CR
100 *
75
25
* *
**
*
50 **
**
*
75
*
50
Figure 1. CR Prevents AHL and Protects Cochlear Neurons in WT Mice, but Not in Sirt3/ Mice
25
0
0 8
16
32
8
16
32
Frequency (kHz)
WT
Sirt3-/-
WT
Sirt3-/-
C
D
E
F
G
H
I
12 mo CD
2 mo CD
B
K
L
12 mo CR
J
mice, suggesting a possible role of Sirt3 in metabolic adaptations to CR. Sirt3 Is Required for the CR-Mediated Reduction of Oxidative Damage in Multiple Tissues How does Sirt3 reduce cochlear cell degeneration and slow the progression of AHL in response to CR? It is well established that CR reduces oxidative damage to DNA, proteins, and lipids in multiple tissues in mammals (Sohal and Weindruch, 1996; Masoro, 2000; Hamilton et al., 2001). Hence, we hypothesized that Sirt3 may play a role in the CR-mediated reduction of oxidative damage in the cochlea and other tissues. To test this hypothesis, we measured oxidative damage to DNA in the cochleae, brain (neocortex), and liver of control diet and calorie-restricted WT and Sirt3/ mice at 12 months of age. We found that CR reduced oxidative DNA damage in WT mice, as determined by measurements of 8-hydroxyguanosine and apurinic/aprimidinic (AP) sites, but failed to reduce oxidative DNA damage in tissues from Sirt3/ mice (Figures 2A and 2B). In agreement with the oxidative damage results, CR increased spiral ganglion neuron survival (Figure 2C), outer hair cell survival (Figure 2D), and inner hair cell survival (Figure 2E) in the basal regions of the cochleae of WT mice, whereas CR failed to protect these cells in Sirt3/ 804 Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc.
M
(A) ABR hearing thresholds were measured at 32, 16, and 8 kHz from control diet and/or calorierestricted WT (left) and Sirt3/ (right) mice at 2 and 12 months of age (n = 9–12). *Significantly different from 2-month-old WT or Sirt3/ mice (p < 0.05), **significantly different from 12-monthold WT mice (p < 0.05). CD, control diet; CR, calorie restricted diet. (B–M) Neurons in the basal cochlear regions from WT mice in control diet at 2 (B and C) and 12 (F and G) months of age and calorie-restricted diet at 12 months of age (J and K). Neurons from control diet Sirt3/ mice at 2 (D and E) and 12 (H and I) months of age and calorie-restricted Sirt3/ mice at 12 months of age (L and M) (n = 5). Arrows in the lower-magnification photos indicate neuron regions. Scale bars, 100 mm (B, F, J, D, H, and L) and 20 mm (C, G, J, E, I, and M). Data are means ± SEM. See also Figure S1, Figure S2, and Figure S3.
mice (Figures 2C–2E). Together, these results provide evidence that Sirt3 plays an essential role in the CR-mediated reduction of oxidative DNA damage in multiple tissues.
Sirt3 Enhances the Mitochondrial Glutathione Antioxidant Defense System in Response to CR A previous study has shown that overexpression of Sirt3 increased mRNA expression of the antioxidant genes manganese superoxide dismutase (MnSOD) and catalase (Cat) in primary cardiomyocytes and that Sirt3/ primary cardiomyocytes displayed higher levels of ROS compared to those of WT cells (Sundaresan et al., 2009), suggesting that Sirt3 may regulate the antioxidant systems. Glutathione acts as the major small molecule antioxidant in cells (Anderson, 1998; Halliwell and Gutteridge, 2007; Marı´ et al., 2009; Rebrin et al., 2003), and NADPHdependent glutathione reductase regenerates reduced glutathione (GSH) from oxidized glutathione (GSSG) (Anderson, 1998; Marı´ et al., 2009). In healthy mitochondria from young mice, glutathione is found mostly in the reduced form, GSH (Marı´ et al., 2009). During aging, oxidized glutathione accumulates, and hence an altered ratio of mitochondrial GSH to GSSG is thought to be a marker of both oxidative stress and aging (Rebrin et al., 2003; Schafer and Buettner, 2001; Marı´ et al., 2009). Thus, we hypothesized that Sirt3 may regulate the mitochondrial glutathione antioxidant system under CR conditions. To test this hypothesis, we measured the ratio of GSH:GSSG in the mitochondria of the inner ear, brain, and liver of control diet and calorie-restricted WT and Sirt3/ mice at 5 months of age. Mitochondrial GSSG levels decreased during CR in the inner ear from WT mice, but not from Sirt3/ mice (Figure 3B; see also Figure 3C). We also found that the ratios of
Neurons/mm2
Sirt3 Sirt3-/-/-
WT
WT WT
Basal Region 3000
Sirt3 Sirt3-/-/-
3000
*
*
8 0
3000
2000
2000
1000
1000
1000
0
0
0 WT WT
Basal Region
D
CD CR
100
Sirt3-/Sirt3-/-
Middle Region 100
75
75
50
50
50
25
25
25
0
0
0
*
WT WT
Sirt3-/Sirt3-/-
WT WT
Basal Region
E 100
CD CR
Sirt3-/Sirt3-/-
Middle Region 100 75
50
50
50
25
25
25
0
0 Sirt3-/-
Sirt3-/-
WT
Brain 120
*
80
80
40
40
0
B
Sirt3-/Sirt3-/-
CD CR
0.4 0.2
*
0.0
0 Sirt3-/Sirt3-/-
Brain 0.8
0.8
0.5
0.5
0.3
0.3
Sirt3-/Sirt3-/-
CD CR
0.0 WT WT
Sirt3-/Sirt3-/-
Brain
Liver 40
15
15
30
10
10
20
5
5
10
0
0 Sirt3-/-
Sirt3-/-
WT
20
WT
Sirt3-/-
Liver 1.0
Inner Ear 20
WT
1.0
0.0 WT WT
C
80 40
WT WT
Inner Ear 0.6
*
0 WT WT
Liver 120
0 WT
Sirt3-/-
WT
Sirt3-/-
Figure 3. Sirt3 Increases the Ratios Of GSH:GSSG in Mitochondria during CR (A–C) Ratios of GSH:GSSG (A), GSSG (B), and GSH (C) were measured in the inner ear, brain (neocortex), and liver from control diet and calorie-restricted WT and Sirt3/ mice at 5 months of age (n = 4–5). *Significantly different from 12- or 5-month-old WT mice (p < 0.05). Data are means ± SEM.
*
0 WT
*
CD CR
Sirt3-/-
Apical Region
75
WT
*
WT
100
75
*
Sirt3-/-
Apical Region
100
75
*
WT
120
-/-
Apical Region
2000
Sirt3-/Sirt3-/-
Sirt3
WT
4000 *
CD CR
16
Middle Region
CD CR 4000
4000
WT WT
OH Cells (%)
*
24
Inner Ear
GSH:GSSG (nmole/mg protein)
*
C
IH Cells (%)
CD CR
100 80 60 40 20 0
A
Liver
GSSG (nmole/mg protein)
CD CR
100 80 60 40 20 0
B
Brain
GSH (nmole/mg protein)
Cochlea
8-oxodGuo/106 dGuo
AP Sites/105 bp DNA
A
Sirt3-/-
Figure 2. CR Reduces Oxidative DNA Damage and Increases Cell Survival in the Cochleae from WT Mice, but Not from Sirt3/ Mice (A) Oxidative damage to DNA (apurinic/apyrimidinic sites) was measured in the cochlea and neocortex from control diet and calorie-restricted WT and Sirt3/ mice at 12 months of age (n = 4–5). AP sites, apurinic/apyrimidinic sites. *Significantly different from 12-month-old WT mice (p < 0.05). (B) Oxidative damage to DNA (8-oxodGuo) was measured in the liver from control diet and calorie-restricted WT and Sirt3/ mice at 12 months of age (n = 4–5). (C) Neuron survival (neuron density) of basal, middle, and apical cochlear regions was measured from control diet and calorie-restricted WT and Sirt3/ mice at 12 months of age (n = 4–5). (D) OH (outer hair) cell survival (%) of basal, middle, and apical cochlear regions was measured from control diet and calorie-restricted WT and Sirt3/ mice at 12 months of age (n = 4–5). (E) IH (inner hair) cell survival (%) of basal, middle, and apical cochlear regions was measured from control diet and calorie-restricted WT and Sirt3/ mice at 12 months of age (n = 4–5). Data are means ± SEM. See also Figures 1B–1M.
GSH:GSSG in mitochondria increased during CR in all of the tested WT tissues (Figure 3A); however, CR failed to increase the ratios of GSH:GSSG in Sirt3/ tissues (Figure 3A). These results are consistent with the histological, cochlear cell counting, and oxidative DNA damage results that demonstrated that
CR reduces oxidative damage in WT tissues, but not in the Sirt3/ tissues. Thus, during CR, Sirt3 promotes a more reductive environment in mitochondria of multiple tissues, thereby enhancing the glutathione antioxidant defense system. Sirt3 Stimulates Idh2 Activity and Increases NADPH Levels in Mitochondria in Response to CR Enzymes of mitochondrial antioxidant pathways require NADPH to perform their reductive functions. NADP+-dependent Idh2 from mitochondria converts NADP+ to NADPH, thereby promoting regeneration of GSH by supplying NADPH to glutathione reductase (Jo et al., 2001). A previous in vitro study suggested that Idh2 might be a target of Sirt3, as incubation of Sirt3 with isocitrate dehydrogenase led to an apparent increase in dehydrogenase activity (Schlicker et al., 2008). Thus, we hypothesized that, in response to CR, the mitochondrial deacetylase Sirt3 might directly deacetylate and activate Idh2, thereby regulating the levels of NADPH and, consequently, the glutathione antioxidant defense system. To provide initial support for the hypothesis that Sirt3 regulates Idh2 activity through deacetylation, we measured the acetylation levels of Idh2 in the liver mitochondria of WT and Sirt3/ mice fed control and CR diets. In WT tissues, acetylation of Idh2 was substantial in the control diet fed tissues, but CR induced an 8-fold decrease in acetylation (Figures 4A and 4B). Robust acetylation of Idh2 was observed in Sirt3/ mice from both Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc. 805
Sirt3-/-
WT
A
CD
CR
CD
CR WB: -IDH2
INPUT WB: -Sirt3 WB: -IDH2 IP:
-IDH2 WB: -AcK
C 200
CD CR
*
150 100 50 0
*
WT WT
IDH2 Activity (μM/s/μg protein)
D
Sirt3-/Sirt3-/-
Liver 0.4
2.0 1.0 0.0 CD CR CD CR
*
Brain
CD CR
0.8
CD CR
WT
Sirt3-/-
0.4 0.2
0.0
0.0 WT WT
Sirt3-/Sirt3-/-
Liver
Sirt3-/Sirt3-/-
Inner Ear CD CR
1.0
*
Brain 1.0
0.8
0.8
0.5
0.5
0.5
0.3
0.3
0.3
0.0
0.0 WT
*
0.6
0.2
*
0.0
1.0
*
3.0
0.4
*
0.2
E
WT
4.0
Inner Ear CD 0.6 CR
0.6
WT WT
NADPH/total NADP
Relative Sirt3 Protein Level
Relative IDH2 Acetylation Level (%)
B
Sirt3
-/-
0.8
*
*
0.0 WT
Sirt3
-/-
WT
Sirt3-/-
Figure 4. Sirt3 Increases Idh2 Activity and NADPH Levels in Mitochondria by Decreasing the Acetylation State of Idh2 during CR (A) (Top) Western blot analysis of Sirt3 and Idh2 levels in the liver from 5-monthold WT or Sirt3/ fed either control or calorie-restricted diet. (Bottom) Endogenous acetylated Idh2 was isolated by immunoprecipitation with anti-Idh2 antibody followed by western blotting with anti-acetyl-lysine antibody (n = 3). (B and C) Quantification of the amounts of total Idh2 acetylation (B) and Sirt3 protein (C) from (A). Western blot was normalized with Idh2 levels or Sirt3 levels quantified and analyzed by Image software (n = 3). (D) Idh2 activities were measured in the liver, inner ear (cochlea), and brain (neocortex) from control diet and calorie-restricted WT and Sirt3/ mice at 5 months of age (n = 3–5). (E) Ratios of NADPH:total NADP (NADP+ + NADPH) were measured in the liver, inner ear, and brain (neocortex) from control diet and caloric restricted WT and Sirt3/ mice at 5 months of age (n = 3–5). *Significantly different from control diet fed WT mice (p < 0.05). Data are means ± SEM.
control and CR diet-fed conditions, indicating that Sirt3 is required for the CR-induced deacetylation of Idh2 (Figures 4A and 4B). As predicted, CR induced Sirt3 protein levels that were approximately three times higher than those observed with control diet tissues in WT mice (Figure 4C). To establish whether Idh2 activity is stimulated by Sirt3 under CR conditions, we measured Idh2 activity in the mitochondria 806 Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc.
from the liver, inner ear, and brain of control diet and calorierestricted WT and Sirt3/ mice. We found that Idh2 activity significantly increased during CR in all of the WT tissues (Figure 4D); however, CR failed to increase Idh2 activity in the Sirt3/ tissues (Figure 4D). If CR can induce a Sirt3-dependent increase in Idh2 activity, we anticipated increased levels of NADPH, providing the primary source of reducing equivalents for the glutathione antioxidant system (Jo et al., 2001; Schafer and Buettner, 2001). To test this hypothesis, we measured NADPH levels in mitochondria of WT and Sirt3/ mice. We found that levels of NADPH increased during CR in all tissues tested from WT mice (Figure 4E); however, no significant changes in NADPH levels were observed between control diet and CR Sirt3/ tissues. Collectively, these results provide evidence that, during CR, Sirt3 induces the deacetylation and activation of Idh2, leading to increased levels of NADPH in mitochondria of multiple tissues. We note that we observed a reduction in Idh2 activity in liver from Sirt3/ mice fed the control diet and that this correlates with a slightly increased level of acetylated Idh2 as compared to WT mice (Figure 4B). However, we did not observe reduced Idh2 activity or reduced NADPH levels in the inner ear or brain of Sirt3/ mice. We postulate that, under basal conditions (control diet fed), additional factors regulate mitochondrial Idh2 activity and NADPH levels. To provide direct evidence that Sirt3 deacetylates Idh2, a number of biochemical experiments were performed. Although most enzyme:substrate reactions are necessarily transient interactions to promote rapid turnover, coimmunoprecipitation (co-IP) experiments can sometimes trap these interactions. Co-IP experiments were performed in human kidney cells (HEK293) cotransfected with Sirt3 and Idh2. We found that precipitated Idh2-FLAG was able to co-IP Sirt3-HA (Figure 5A), whereas precipitated Sirt3-FLAG was able to co-IP Idh2-MYC (Figure 5B), suggesting that a physical interaction can occur between Sirt3 and Idh2 in human cells. However, co-IP experiments do not prove a direct functional interaction. To provide support for a functional interaction between Sirt3 and acetylated Idh2, deacetylation assays were carried out in HEK293 cells (Figure 5C) and in vitro using purified components (Figure 5D). Utilizing HEK293 cells, Idh2 was cotransfected with or without Sirt3, isolated by immunoprecipitation with anti-MYC antibody followed by western blotting with anti-acetyl-lysine antibody. Coexpression with Sirt3 induced the deacetylation of Idh2 to background levels (Figure 5C). For the in vitro analysis, acetylated Idh2 was prepared (see Figure S4 and Experimental Procedures) and utilized as a substrate for purified recombinant Sirt3 or Sirt5. Acetylation status was assessed by western blotting with anti-acetyl-lysine antibody (Figure 5D), and the resulting change in Idh2 activity was measured separately (Figure 5E). We found that Sirt3, but not Sirt5, deacetylated IDH2 in an NAD+-dependent fashion (Figure 5E). The corresponding Idh2 activity measurements indicated that deacetylation by Sirt3, but not Sirt5, stimulated Idh2 activity by 100% (Figure 5E). Together, these data provide strong biochemical evidence that Sirt3 deacetylates and stimulates Idh2 activity and increases NADPH levels in mitochondria in response to CR.
A
Sirt3-HA
+
IDH2-FLAG
-
IP α-IgG
+
+
+
-
B
α-FLAG
+
+
Sirt3-FLAG
-
+
Sirt3-FLAG
-
+
+
+
-
Figure 5. Sirt3 Directly Deacetylates Idh2 and Stimulates Activity
α-FLAG
INPUT
WB: -HA
IDH2-MYC
+
IP α-IgG
INPUT
C
IDH2-MYC
WB: -MYC
WB: -FLAG
WB: -FLAG
WB: -HA
WB: -MYC
D
IDH2-FLAG
+
+
+
+
+ +
NAD+
-
+
-
+
-
SIRTUIN
-
-
+
Sirt3 Sirt3 Sirt5 Sirt5
WB: -MYC WB: -FLAG
IP: α-MYC WB: -AcK WB: -FLAG
WB: -AcK COOMASSIE BLUE
IDH2+Sirt5
IDH2+Sirt5+NAD
IDH2+Sirt3
IDH2+Sirt3+NAD
*
IDH2+NAD
250 200 150 100 50 0
IDH2
E
Relative IDH2 Activity (%)
INPUT
IP: α-FLAG
Overexpression of Sirt3 and/or Idh2 Increases NADPH Levels and Protects Cells from Oxidative Stress-Induced Cell Death Our physiological, histological, and biochemical results indicate that Sirt3 mediates reduction of oxidative damage by deacetylation and stimulating the activity of Idh2, which increases NADPH levels for antioxidant systems in mitochondria during CR. To provide support for this mechanism, we investigated whether Sirt3 and Idh2 are sufficient to alter the NADPH levels in cultured cells. HEK293 cells stably transfected with vector, Sirt3, Idh2, or Sirt3 with Idh2 were generated, and their NADPH levels were measured. NADPH levels were significantly increased when either Idh2 or Sirt3 or both proteins were stably overexpressed in HEK293 cells (Figures 6A and 6B). Importantly, overexpression of both Sirt3 and Idh2 yielded a greater increase in NADPH levels than either Sirt3 or Idh2 overexpressed alone (Figure 6A). Finally, to investigate whether overexpression of Sirt3, Idh2, or Sirt3 with Idh2 can protect cells from oxidative stress, the four HEK293 cell lines were treated with oxidants H2O2 (hydrogen peroxide) (Figure 6C) or menadione (Figure 6D), and cell viability was measured. Overexpression of Sirt3 or Idh2 was sufficient to protect cells from oxidative stress induced by both oxidants (Figures 6C and 6D). Again, overexpression of both Sirt3 and Idh2 led to higher cell viability than either Sirt3 or Idh2 overexpressed alone (Figures 6C and 6D). These results provide strong biochemical evidence that Sirt3 mediates reduction of oxidative
(A and B) Sirt3 interacts with Idh2. Idh2 or Sirt3 were immunoprecipitated from HEK293 cell lysates with IgG antibody or FLAG beads. Precipitated Idh2-FLAG was detected by anti-FLAG antibody, and co-IP Sirt3-HA was detected by anti-HA as indicated (A). Precipitated Sirt3-FLAG was detected by anti-FLAG antibody, and co-IP Idh2MYC was detected by anti-MYC as indicated (B) (n = 3). (C) Sirt3 deacetylates Idh2 in HEK293 cells. Idh2 was cotransfected with or without Sirt3, isolated by immunoprecipitation with anti-MYC antibody followed by western blotting with anti-acetyllysine antibody (n = 3). (D) Sirt3, but not Sirt5, deacetylates Idh2 in vitro. Acetylated Idh2 was prepared as outlined in the Experimental Procedures and was incubated with purified recombinant Sirt3 or Sirt5 with or without NAD+ at 37 C for 1 hr. Acetylation status was assessed by western blotting with antiacetyl-lysine antibody (n = 3). An anti-FLAG western shows that equivalent Idh2 protein levels were used, and Coomassie staining shows purified Sirt3 and Sirt5. (E) In vitro deacetylation of Idh2 by Sirt3, but not Sirt5, stimulates Idh2 activity. Acetylated Idh2 in buffer (Tris [pH 7.5], with or without 1 mM NAD, and 1 mM DTT) was incubated with purified 50 nM Sirt3 or Sirt5 (Hallows et al., 2006) at 37 C for 1 hr, followed by Idh2 activity assay (n = 3). *Significantly different from Idh2 alone (p < 0.05). Data are means ± SEM. See also Figure S4.
stress by stimulating Idh2 activity and increasing NADPH levels under stress conditions. DISCUSSION Sirt3 Reduces Oxidative Damage and Enhances the Glutathione Antioxidant Defense System under CR Conditions A widely accepted hypothesis of how aging leads to age-related hearing loss is through the accumulation of oxidative damage in the inner ear (Someya and Prolla, 2010; Liu and Yan, 2007). In support of this hypothesis, oxidative protein damage increases in the cochlea of CBA/J mice (Jiang et al., 2007), and oxidative DNA damage increases in the cochlea of C57BL/6J mice during aging (Someya et al., 2009). Age-related hair cell loss is also enhanced in mice lacking the antioxidant enzyme superoxide dismutase 1 (McFadden et al., 1999), whereas the same mutant animals show enhanced susceptibility to noise-induced hearing loss (Ohlemiller et al., 1999). We have shown recently that overexpression of mitochondrially targeted catalase delays the onset of AHL in C57BL/6J mice, reduces hair cell loss, and reduces oxidative DNA damage in the inner ear (Someya et al., 2009). Of interest, overexpression of catalase in the mitochondria leads to extension of life span in C57BL/6J mice, but overexpression of catalase in the peroxisome or nucleus does not (Schriner et al., 2005). Under normal conditions, catalase decomposes Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc. 807
B 180
-
-
+
+
pCDNA3-Sirt3-FLAG
-
+
-
+
* *
WB: α-FLAG
*
30 VEC
Sirt3
IDH2
Sirt3+IDH2
1 mM H2O2
VEC
0
-
+
+
+
+
90
*
60
*
** *
(A and B) (A) NADPH concentrations were significantly increased when either Idh2 or Sirt3 or both were stably overexpressed in HEK293 cells. Measurements with errors are shown for the four different stable cell populations from each type of transfection (vector alone, Sirt3, Idh2, and Sirt3 with Idh2) (n = 3). *Significantly different from vector alone (p < 0.05); **Significantly different from Idh2 or Sirt3 (p < 0.05). (B) Western blotting confirms Idh2 and Sirt3 stable expression. (C and D) Sirt3 and/or Idh2 overexpression is sufficient to protect HEK293 cells from the exogenous oxidants hydrogen peroxide (H2O2) (C) and menadione (D). The four different stable cells were transiently exposed to either 1 mM H2O2 or 25 mM menadione (n = 16). Data are means ± SEM.
30 0
25 μM Menadione -
Sirt3+IDH2
*
120
VEC
90
** *
IDH2
D
Sirt3
Sirt3
IDH2
0
Sirt3-FLAG
Sirt3+IDH2
60
60
IDH2-FLAG
VEC
120
Figure 6. Overexpression of Sirt3 and/or Idh2 Is Sufficient to Increase NADPH Levels and Protects HEK293 Cells from Oxidative Stress
INPUT
120
Cell Viability (%)
pBabe-IDH2-FLAG
Cell Viability (%)
C
** *
VEC
[NADPH] (nmol/mg protein)
A
+
+
+
+
hydrogen peroxide in the peroxisome, whereas in mitochondria, hydrogen peroxide is decomposed into water by glutathione peroxidase or peroxiredoxin (Finkel and Holbrook, 2000; Marı´ et al., 2009). Hence, these results suggest that mitochondrial ROS play a critical role in cochlear aging, AHL, and aging in general. We have demonstrated that Sirt3 mediates the CR reduction of oxidative DNA damage in multiple tissues and that these effects are likely to arise through an enhanced mitochondrial glutathione antioxidant defense system. As discussed earlier, the GSH:GSSG ratio is thought to be a marker of oxidative stress (Rebrin and Sohal, 2008). Experimental evidence indicates that aging results in a decrease in the ratio of GSH:GSSG in the mitochondria of brain, liver, kidney, eye, heart, and testis from aged C57BL/6J mice due to elevated levels of GSSG, whereas CR decreases the ratio of GSH:GSSG in the mitochondria of these tissues by lowering GSSG levels (Rebrin et al., 2003, 2007). Our findings demonstrate that CR increases these ratios of GSH:GSSG in the mitochondria of brain, liver, and inner ear from WT mice but fails to increase the ratios in the same tissues from Sirt3/ mice. Consistent with these results, CR reduced oxidative DNA damage in tissues from WT mice but failed to reduce such damage in tissues from Sirt3/ mice. CR also increased spiral ganglion neuron and hair cell survival in the WT cochlea, but not in Sirt3/ mice. Tissues that are composed of postmitotic cells such as the brain and the inner ear are particularly vulnerable to oxidative damage because of their high energy requirements and inability to undergo regeneration. Therefore, we speculate that the Sirt3-mediated modulation of 808 Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc.
the glutathione antioxidant defense system may play a central role in reduction of oxidative stress in multiple tissues under CR conditions, leading to aging retardation. We also note that other mitochondrial effects of Sirt3, such as regulation of fatty acid oxidation (Hirschey et al., 2010) and modulation of complex I activity (Ahn, et al., 2008), are likely to contribute to the metabolic adaptations in response to CR. Idh2 Regulates the Redox State of Mitochondria under CR Conditions A large body of evidence indicates that the antioxidant defense systems do not keep pace with the age-related increase in ROS production, and thus the balance between antioxidant defenses and ROS production shifts progressively toward a more pro-oxidant state during aging (Sohal and Weindruch, 1996; Rebrin and Sohal, 2008). This balance is determined in part by the ratios of interconvertible forms of redox couples, such as GSH/GSSG, NADPH/NADP+, NADH/NAD+, thioredoxinred/thioredoxinoxid, and glutaredoxinred/glutaredoxinoxid. The GSH/GSSH couple is thought to be the primary cellular determinant of the cellular redox state because its abundance is three to four orders of magnitude higher than the other redox couples (Rebrin and Sohal, 2008). NADPH is the reducing equivalent required for the regeneration of GSH and the GSH-mediated antioxidant defense system, which includes glutathione peroxidases, glutathione transferases, and glutathione reductase, playing a critical role in oxidative stress resistance (Halliwell and Gutteridge, 2007). GSH is synthesized in the cytosol and transported into the mitochondria through protein channels in the outer mitochondrial membrane (Halliwell and Gutteridge, 2007; Anderson, 1998). Although GSH can cross the outer mitochondrial membrane through these channels, GSSG cannot be exported into the cytosol (Olafsdottir and Reed, 1988). Thus, GSSG is reduced to GSH by mitochondrial NADPH-dependent
glutathione reductase, preventing accumulation of GSSG in the mitochondrial matrix (Schafer and Buettner, 2001; Marı´ et al., 2009). We have demonstrated that Sirt3 directly deacetylates and activates Idh2 under CR conditions. In response to CR, deacetylated Idh2 displays increased catalytic activity, which is correlated with increased NADPH levels in the mitochondria of multiple tissues from WT mice, but not from Sirt3/ mice. Hence, we speculate that Idh2 may be a major player in regulating the redox state of mitochondria under CR conditions given its role in mitochondrial NADPH production. A previous study has shown that Idh2 is induced in response to ROS in mouse fibroblasts, whereas decreased levels of Idh2 lead to higher ROS and accumulation of oxidative damage to DNA and lipids (Jo et al., 2001). Our in vitro findings demonstrate that overexpression of Sirt3 and/or Idh2 increases NADPH levels and protects cells from oxidative stress-induced cell death. Thus, these observations underlie a critical role for Idh2 in the generation of NADPH in mitochondria under conditions of CR, providing reducing capacity for the glutathione antioxidant system and increasing oxidative stress resistance. A Role for Sirt3 in CR-Mediated Prevention of AHL The mouse is considered a good model for the study of human AHL because the mouse cochlea is anatomically similar to that of humans (Steel et al., 1996; Steel and Bock, 1983). Most inbred mouse strains display some degree of AHL, and the age of onset of AHL is known to vary from 3 months in DBA/2J mice to more than 20 months in CBA/CaJ mice (Zheng et al., 1999). The C57BL/6J mouse strain, which is the most widely used mouse model for the study of aging, displays the classic pattern of AHL by 12–15 months of age (Hunter and Willott, 1987; Keithley et al., 2004). We have previously shown that AHL in C57BL/6J mice occurs through Bak-mediated apoptosis and that it can be prevented by the intake of small molecule antioxidants (Someya et al., 2009). We note that C57BL/6J and many other mouse strains carry a specific mutation (Cdh23753A) in the Cdh23 gene, which encodes a component of the hair cell tip link, and this mutation is known to promote early onset of AHL in these animals (Noben-Trauth et al., 2003). Of interest, the Cdh23753A allele may increase the susceptibility to oxidative stress in hair cells because a Sod1 mutation greatly enhances AHL in mice carrying Cdh23753A, but not in mice wild-type for Cdh23 (Johnson, et al., 2010). However, oxidative damage increases with age in the cochlea of both C57BL/6J mice and the CBA/J mouse strain that does not carry the Cdh23753A allele, indicating that oxidative stress plays a role in AHL independent of Cdh23 (Someya et al., 2009; Jiang et al., 2007; Zheng et al., 1999). In both strains, the loss of hair cells and spiral ganglion neurons begins in the base of the cochlea and spreads toward the apex with age (Keithley et al., 2004; Hunter and Willott, 1987). Importantly, CR slows the progression of AHL in both C57BL/6J and CBA/J strains (Someya et al., 2007; Sweet et al., 1988). Therefore, the protective effects of Sirt3 in AHL are likely to be of general relevance to AHL. It is thought that some of the effects of CR in aging retardation require significant reduction of body weight through reducing food consumption. In agreement with this hypothesis, obesity promotes a variety of age-related diseases, such as cardiovas-
Figure 7. A Model for the CR-Mediated Prevention of AHL in Mammals In response to CR, SIRT3 activates IDH2, thereby increasing NADPH levels in mitochondria. This in turn leads to an increased ratio of GSH:GSSG and decreased levels of ROS, thereby resulting in protection from oxidative stress and prevention of AHL in mammals.
cular disease, diabetes, high blood pressure, hypertension, and certain cancers (Paeratakul et al., 2002; Poirier et al., 2006). Obesity is also associated with an increased risk of mortality (Poirier et al., 2006; Lee et al., 1993). Of interest, CR failed to reduce oxidative damage in multiple tissues and slow the progression of AHL in CR Sirt3/ mice, despite the fact that these mice were lean (Figures S2A and S2B). Thus, these results suggest that weight loss may not be sufficient for the anti-aging action of CR. Instead, we postulate that critical metabolic effectors such as Sirt3 mediate the positive effects of CR. Conclusions In summary, we propose that, in response to CR, Sirt3 activates Idh2, thereby increasing NADPH levels in mitochondria. This in turn leads to increased ratios of GSH:GSSG in mitochondria and decreased levels of ROS, resulting in protection of inner ear cells and prevention of AHL in mammals (Figure 7). Because we observed similar effects of CR in the mitochondrial GSH/GSSG ratios in multiple tissues, we postulate that this may be a major mechanism of aging retardation by CR. We also postulate that pharmaceutical interventions that induce Sirt3 activity in multiple tissues will mimic CR by increasing oxidative stress resistance and preventing the mitochondrial decay associated with aging. EXPERIMENTAL PROCEDURES Animals Male and female Sirt3+/ mice were purchased from the Mutant Mouse Resource Centers (MMRRC) at the University of North Carolina-Chapel Hill
Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc. 809
(Chapel Hill, NC). In brief, these mice were created by generating embryonic stem (ES) cells (Omni bank number OST341297) bearing a retroviral promoter trap that functionally inactivates one allele of the Sirt3 gene (MGI, 2010). Male and female C57BL/6J mice were purchased from Jackson Laboratory (Bar Harbor, ME). Sirt3+/ mice have been backcrossed for four generations onto the C57BL/6J background. All animal studies were conducted at the AAALAC-approved Animal Facility in the Genetics and Biotechnology Center of the University of Wisconsin-Madison. Experiments were performed in accordance with protocols approved by the University of Wisconsin-Madison Institutional Animal Care and Use Committee (Madison, WI).
Idh2 Activity Activities of Idh2 were measured by the Kornberg method (Kornberg, 1955). In brief, 20 ml of the mitochondrial lysate sample was added in each well of a 96-well plate, and then 180 ml of a reaction mixture (33 mM KH2PO4dK2HPO4, 3.3 mM MgCl2, 167 mM NADP+, and 167 mM (+)-potassium Ds-threo-isocitrate monobasic) was added in each well. The absorbance was immediately read at 340 nm every 10 s for 1 min in a microplate reader (Bio-Rad, Hercules, CA). All samples were run in duplicate. The reaction rates were calculated, and the Idh2 activity in the sample was defined as the production of one mmole of NADPH per sec.
Dietary Study Details on the methods used to house and feed mice have been described previously (Pugh et al., 1999). Mice are housed individually. Control diet (CD) groups were fed 86.4 kcal/week of the precision pellet diet AIN-93M (BioServ, Frenchtown, NJ), and caloric-restricted (CR) groups were fed 64.8 kcal/week (a 25% CR) of the precision pellet diet AIN-93M 40%DR (BioServ, Frenchtown, NJ). The schedule of feeding for control diet was 7 g on Mondays and Wednesdays and 10 g on Fridays, whereas the schedule of feeding for calorierestricted diets was 5 g on Mondays and Wednesdays and 8 g on Fridays. This dietary regimen was maintained from 2 months of age until 5 months of age for a 3 month CR study and from 2 months of age until 12 months of age for a 10 month CR study.
In Vitro Deacetylation Assay Idh2-FLAG was transfected into HEK293 cells, which were then treated with 5 mM nicotinamide for 16 hr. Nicotinamide is a widely used sirtuin inhibitor. Nicotinamide treatment leads to increased acetylation of Idh2, with a corresponding decrease in enzymatic activity (Figure S4). Idh2 from cell lysates was immunoprecipitated with anti-FLAG beads at 4 C for 2 hr, and then Idh2-FLAG on beads was utilized in 200 ul deacetylation buffer (Tris [pH 7.5], with or without 1 mM NAD, and 1 mM DTT) and incubated with purified 50 nM Sirt3 or Sirt5 (Hallows et al., 2006) at 37 C for 1 hr. Aliquots were removed for Idh2 activity assay and western blotting with anti-FLAG antibody or anti-acetyl-lysine antibody.
ABR Hearing Test At 12 months of age, ABRs were measured with a tone burst stimulus at 8, 16, and 32 kHz using an ABR recording system (Intelligent Hearing System, Miami, FL) as previously described (Someya et al., 2009). Mice were anesthetized with a mixture of xylazine hydrochloride (10 mg/kg, i.m.) (Phoenix Urology of St. Joseph, St. Joseph, MO) and ketamine hydrochloride (40 mg/kg, i.m.) (Phoenix Urology of St. Joseph). Measurement of DNA Oxidation Levels At 12 months of age, cochlea and neocortex were collected, and DNA was extracted with ethanol precipitation. DNA concentrations for each sample were adjusted to 0.1 mg/ml, and numbers of apurinic/apyrimidinic (AP) sites were determined using the DNA Damage Quantification Kit (Dojindo, Rockville, MD) and performed according to the manufacturer’s instructions and as previously described (Kubo et al., 1992; Meira, et al., 2009; McNeill and Wilson, 2007). Liver was also collected from the same mice, and 8-hydroxyguanosine levels (8-oxo-7,8-20 -deoxyguanosine/106 deoxyguanosine) in the DNA were determined using a HPLC-ECD method as previously described (Hofer et al., 2006). Measurement of Total GSH and GSSG Just after mitochondrial lysate preparation, 100 ml of the lysate was mixed with 100 ml of 10% metaphosphoric acid, incubated for 30 min at 4 C, and centrifuged at 14,000 3 g for 10 min at 4 C. The supernatant was used for the measurements of mitochondrial glutathione contents. Total glutathione (GSH + GSSG) and GSSG levels were determined by the method of Rahman et al. (2006). All samples were run in duplicate. The rates of 2-nitro-5-thiobenzoic acid formation were calculated, and the total glutathione (tGSH) and GSSG concentrations in the samples were determined by using linear regression to calculate the values obtained from the standard curve. The GSH concentration was determined by subtracting the GSSG concentration from the tGSH concentration. Idh2 Acetylation Analysis Antibodies used for western blotting included anti-Idh2 antibody (Santa Cruz, Santa Cruz, CA), anti-Sirt3 antibody (gift of Dr. Eric Verdin, UCSF), protein A/G plus agarose (Santa Cruz, Santa Cruz, CA), and pan-acetylated lysine (generated following the procedure of Zhao, et al. [2010], GeneTel Laboratories LLC, Madison, WI). For immunoprecipitation, liver mitochondria lysates were incubated with anti-Idh2 antibody overnight at 4 C. Then protein A/G plus agarose were added and incubated for 3 hr. After resins were washed, samples were boiled with SDS loading buffer and subjected to western blotting (Smith et al., 2009).
810 Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc.
Statistical Analysis All Statistical analyses were carried out by one-way ANOVA with post-Tukey multiple comparison tests using the Prism 4.0 statistical analysis program (GraphPad, San Diego, CA). All tests were two-sided with statistical significance set at p < 0.05. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, four figures, and one table and can be found at doi:10.1016/j.cell.2010.10.002. ACKNOWLEDGMENTS We thank S. Kinoshita for histological processing. This research was supported by NIH grants AG021905 (T.A.P.) and GM065386 (J.M.D.), the National Projects on Protein Structural and Functional Analyses from the Ministry of Education, Culture, Sports, Science, and Technologies of Japan, and Marine Bio Foundation. Received: July 19, 2010 Revised: September 3, 2010 Accepted: September 30, 2010 Published online: November 18, 2010 REFERENCES Ahn, B.H., Kim, H.S., Song, S., Lee, I.H., Liu, J., Vassilopoulos, A., Deng, C.X., and Finkel, T. (2008). A role for the mitochondrial deacetylase Sirt3 in regulating energy homeostasis. Proc. Natl. Acad. Sci. USA 105, 14447–14452. Anderson, M.E. (1998). Glutathione: an overview of biosynthesis and modulation. Chem. Biol. Interact. 111-112, 1–14. Balaban, R.S., Nemoto, S., and Finkel, T. (2005). Mitochondria, oxidants, and aging. Cell 120, 483–495. Barger, J.L., Kayo, T., Vann, J.M., Arias, E.B., Wang, J., Hacker, T.A., Wang, Y., Raederstorff, D., Morrow, J.D., Leeuwenburgh, C., et al. (2008). A low dose of dietary resveratrol partially mimics caloric restriction and retards aging parameters in mice. PLoS ONE 3, e2264. Bordone, L., Cohen, D., Robinson, A., Motta, M.C., van Veen, E., Czopik, A., Steele, A.D., Crowe, H., Marmor, S., Luo, J., et al. (2007). SIRT1 transgenic mice show phenotypes resembling calorie restriction. Aging Cell 6, 759–767. Chen, D., Steele, A.D., Lindquist, S., and Guarente, L. (2005). Increase in activity during calorie restriction requires Sirt1. Science 310, 1641.
Chen, D., Bruno, J., Easlon, E., Lin, S.J., Cheng, H.L., Alt, F.W., and Guarente, L. (2008). Tissue-specific regulation of SIRT1 by calorie restriction. Genes Dev. 22, 1753–1757.
Marı´, M., Morales, A., Colell, A., Garcı´a-Ruiz, C., and Ferna´ndez-Checa, J.C. (2009). Mitochondrial glutathione, a key survival antioxidant. Antioxid. Redox Signal. 11, 2685–2700.
Cohen, H.Y., Miller, C., Bitterman, K.J., Wall, N.R., Hekking, B., Kessler, B., Howitz, K.T., Gorospe, M., de Cabo, R., and Sinclair, D.A. (2004). Calorie restriction promotes mammalian cell survival by inducing the SIRT1 deacetylase. Science 305, 390–392.
Masoro, E.J. (2000). Caloric restriction and aging: an update. Exp. Gerontol. 35, 299–305. Mattson, M.P. (2000). Apoptosis in neurodegenerative disorders. Nat. Rev. Mol. Cell Biol. 1, 120–129.
Colman, R.J., Anderson, R.M., Johnson, S.C., Kastman, E.K., Kosmatka, K.J., Beasley, T.M., Allison, D.B., Cruzen, C., Simmons, H.A., Kemnitz, J.W., and Weindruch, R. (2009). Caloric restriction delays disease onset and mortality in rhesus monkeys. Science 325, 201–204.
McFadden, S.L., Ding, D., Reaume, A.G., Flood, D.G., and Salvi, R.J. (1999). Age-related cochlear hair cell loss is enhanced in mice lacking copper/zinc superoxide dismutase. Neurobiol. Aging 20, 1–8.
Donmez, G., and Guarente, L. (2010). Aging and disease: connections to sirtuins. Aging Cell 9, 285–290.
McNeill, D.R., and Wilson, D.M.I.I.I., III. (2007). A dominant-negative form of the major human abasic endonuclease enhances cellular sensitivity to laboratory and clinical DNA-damaging agents. Mol. Cancer Res. 5, 61–70.
Finkel, T., and Holbrook, N.J. (2000). Oxidants, oxidative stress and the biology of ageing. Nature 408, 239–247. Finkel, T., Deng, C.X., and Mostoslavsky, R. (2009). Recent progress in the biology and physiology of sirtuins. Nature 460, 587–591. Halliwell, B., and Gutteridge, J.M.C. (2007). Free Radicals in Biology and Medicine (New York, NY: Oxford University Press). Hallows, W.C., Lee, S., and Denu, J.M. (2006). Sirtuins deacetylate and activate mammalian acetyl-CoA synthetases. Proc. Natl. Acad. Sci. USA 103, 10230–10235. Hamilton, M.L., Van Remmen, H., Drake, J.A., Yang, H., Guo, Z.M., Kewitt, K., Walter, C.A., and Richardson, A. (2001). Does oxidative damage to DNA increase with age? Proc. Natl. Acad. Sci. USA 98, 10469–10474. Hirschey, M.D., Shimazu, T., Goetzman, E., Jing, E., Schwer, B., Lombard, D.B., Grueter, C.A., Harris, C., Biddinger, S., Ilkayeva, O.R., et al. (2010). SIRT3 regulates mitochondrial fatty-acid oxidation by reversible enzyme deacetylation. Nature 464, 121–125. Hofer, T., Seo, A.Y., Prudencio, M., and Leeuwenburgh, C. (2006). A method to determine RNA and DNA oxidation simultaneously by HPLC-ECD: greater RNA than DNA oxidation in rat liver after doxorubicin administration. Biol. Chem. 387, 103–111. Hunter, K.P., and Willott, J.F. (1987). Aging and the auditory brainstem response in mice with severe or minimal presbycusis. Hear. Res. 30, 207–218. Jiang, H., Talaska, A.E., Schacht, J., and Sha, S.H. (2007). Oxidative imbalance in the aging inner ear. Neurobiol. Aging 28, 1605–1612. Jo, S.H., Son, M.K., Koh, H.J., Lee, S.M., Song, I.H., Kim, Y.O., Lee, Y.S., Jeong, K.S., Kim, W.B., Park, J.W., et al. (2001). Control of mitochondrial redox balance and cellular defense against oxidative damage by mitochondrial NADP+-dependent isocitrate dehydrogenase. J. Biol. Chem. 276, 16168– 16176. Johnson, K.R., Yu, H., Ding, D., Jiang, H., Gagnon, L.H., and Salvi, R.J. (2010). Separate and combined effects of Sod1 and Cdh23 mutations on age-related hearing loss and cochlear pathology in C57BL/6J mice. Hear. Res. 268, 85–92. Keithley, E.M., Canto, C., Zheng, Q.Y., Fischel-Ghodsian, N., and Johnson, K.R. (2004). Age-related hearing loss and the ahl locus in mice. Hear. Res. 188, 21–28. Kornberg, A. (1955). Isocitric dehydrogenase of yeast (TPN). Methods Enzymol. 1, 705–707.
Meira, L.B., Moroski-Erkul, C.A., Green, S.L., Calvo, J.A., Bronson, R.T., Shah, D., and Samson, L.D. (2009). Aag-initiated base excision repair drives alkylation-induced retinal degeneration in mice. Proc. Natl. Acad. Sci. USA 106, 888–893. MGI. Sirt3Gt(OST341297)Lex. (2010). Available at http://www.informatics.jax.org/ searches/accession_report.cgi?id=MGI:3529767. Noben-Trauth, K., Zheng, Q.Y., and Johnson, K.R. (2003). Association of cadherin 23 with polygenic inheritance and genetic modification of sensorineural hearing loss. Nat. Genet. 35, 21–23. Ohlemiller, K.K., McFadden, S.L., Ding, D.L., Flood, D.G., Reaume, A.G., Hoffman, E.K., Scott, R.W., Wright, J.S., Putcha, G.V., and Salvi, R.J. (1999). Targeted deletion of the cytosolic Cu/Zn-superoxide dismutase gene (Sod1) increases susceptibility to noise-induced hearing loss. Audiol. Neurootol. 4, 237–246. Olafsdottir, K., and Reed, D.J. (1988). Retention of oxidized glutathione by isolated rat liver mitochondria during hydroperoxide treatment. Biochim. Biophys. Acta 964, 377–382. Paeratakul, S., Lovejoy, J.C., Ryan, D.H., and Bray, G.A. (2002). The relation of gender, race and socioeconomic status to obesity and obesity comorbidities in a sample of US adults. Int. J. Obes. Relat. Metab. Disord. 26, 1205–1210. Poirier, P., Giles, T.D., Bray, G.A., Hong, Y., Stern, J.S., Pi-Sunyer, F.X., and Eckel, R.H. (2006). Obesity and cardiovascular disease: pathophysiology, evaluation, and effect of weight loss. Arterioscler. Thromb. Vasc. Biol. 26, 968–976. Pugh, T.D., Klopp, R.G., and Weindruch, R. (1999). Controlling caloric consumption: protocols for rodents and rhesus monkeys. Neurobiol. Aging 20, 157–165. Rahman, I., Kode, A., and Biswas, S.K. (2006). Assay for quantitative determination of glutathione and glutathione disulfide levels using enzymatic recycling method. Nat. Protoc. 1, 3159–3165. Rebrin, I., and Sohal, R.S. (2008). Pro-oxidant shift in glutathione redox state during aging. Adv. Drug Deliv. Rev. 60, 1545–1552. Rebrin, I., Kamzalov, S., and Sohal, R.S. (2003). Effects of age and caloric restriction on glutathione redox state in mice. Free Radic. Biol. Med. 35, 626–635.
Koubova, J., and Guarente, L. (2003). How does calorie restriction work? Genes Dev. 17, 313–321.
Rebrin, I., Forster, M.J., and Sohal, R.S. (2007). Effects of age and caloric intake on glutathione redox state in different brain regions of C57BL/6 and DBA/2 mice. Brain Res. 1127, 10–18.
Kubo, K., Ide, H., Wallace, S.S., and Kow, Y.W. (1992). A novel, sensitive, and specific assay for abasic sites, the most commonly produced DNA lesion. Biochemistry 31, 3703–3708.
Schafer, F.Q., and Buettner, G.R. (2001). Redox environment of the cell as viewed through the redox state of the glutathione disulfide/glutathione couple. Free Radic. Biol. Med. 30, 1191–1212.
Lee, I.M., Manson, J.E., Hennekens, C.H., and Paffenbarger, R.S., Jr. (1993). Body weight and mortality. A 27-year follow-up of middle-aged men. JAMA 270, 2823–2828.
Schlicker, C., Gertz, M., Papatheodorou, P., Kachholz, B., Becker, C.F., and Steegborn, C. (2008). Substrates and regulation mechanisms for the human mitochondrial sirtuins Sirt3 and Sirt5. J. Mol. Biol. 382, 790–801.
Lin, S.J., Defossez, P.A., and Guarente, L. (2000). Requirement of NAD and SIR2 for life-span extension by calorie restriction in Saccharomyces cerevisiae. Science 289, 2126–2128.
Schriner, S.E., Linford, N.J., Martin, G.M., Treuting, P., Ogburn, C.E., Emond, M., Coskun, P.E., Ladiges, W., Wolf, N., Van Remmen, H., et al. (2005). Extension of murine life span by overexpression of catalase targeted to mitochondria. Science 308, 1909–1911.
Liu, X.Z., and Yan, D. (2007). Ageing and hearing loss. J. Pathol. 211, 188–197.
Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc. 811
Schwer, B., Eckersdorff, M., Li, Y., Silva, J.C., Fermin, D., Kurtev, M.V., Giallourakis, C., Comb, M.J., Alt, F.W., and Lombard, D.B. (2009). Calorie restriction alters mitochondrial protein acetylation. Aging Cell 8, 604–606. Smith, B.C., Hallows, W.C., and Denu, J.M. (2009). A continuous microplate assay for sirtuins and nicotinamide-producing enzymes. Anal. Biochem. 394, 101–109. Sohal, R.S., and Weindruch, R. (1996). Oxidative stress, caloric restriction, and aging. Science 273, 59–63. Someya, S., Yamasoba, T., Weindruch, R., Prolla, T.A., and Tanokura, M. (2007). Caloric restriction suppresses apoptotic cell death in the mammalian cochlea and leads to prevention of presbycusis. Neurobiol. Aging 28, 1613– 1622. Someya, S., Xu, J., Kondo, K., Ding, D., Salvi, R.J., Yamasoba, T., Rabinovitch, P.S., Weindruch, R., Leeuwenburgh, C., Tanokura, M., and Prolla, T.A. (2009). Age-related hearing loss in C57BL/6J mice is mediated by Bak-dependent mitochondrial apoptosis. Proc. Natl. Acad. Sci. USA 106, 19432–19437.
Sundaresan, N.R., Samant, S.A., Pillai, V.B., Rajamohan, S.B., and Gupta, M.P. (2008). SIRT3 is a stress-responsive deacetylase in cardiomyocytes that protects cells from stress-mediated cell death by deacetylation of Ku70. Mol. Cell. Biol. 28, 6384–6401. Sundaresan, N.R., Gupta, M., Kim, G., Rajamohan, S.B., Isbatan, A., and Gupta, M.P. (2009). Sirt3 blocks the cardiac hypertrophic response by augmenting Foxo3a-dependent antioxidant defense mechanisms in mice. J. Clin. Invest. 119, 2758–2771. Sweet, R.J., Price, J.M., and Henry, K.R. (1988). Dietary restriction and presbyacusis: periods of restriction and auditory threshold losses in the CBA/J mouse. Audiology 27, 305–312. Wallace, D.C. (2005). A mitochondrial paradigm of metabolic and degenerative diseases, aging, and cancer: a dawn for evolutionary medicine. Annu. Rev. Genet. 39, 359–407. Weindruch, R., and Walford, R.L. (1988). The Retardation of Aging and Disease by Dietary Restriction (Springfield, IL: Charles C Thomas Publishing, LTD).
Someya, S., and Prolla, T.A. (2010). Mitochondrial oxidative damage and apoptosis in age-related hearing loss. Mech. Ageing Dev. 131, 480–486.
Zhao, S., Xu, W., Jiang, W., Yu, W., Lin, Y., Zhang, T., Yao, J., Zhou, L., Zeng, Y., Li, H., et al. (2010). Regulation of cellular metabolism by protein lysine acetylation. Science 327, 1000–1004.
Steel, K.P., and Bock, G.R. (1983). Hereditary inner-ear abnormalities in animals. Relationships with human abnormalities. Arch. Otolaryngol. 109, 22–29.
Zheng, Q.Y., Johnson, K.R., and Erway, L.C. (1999). Assessment of hearing in 80 inbred strains of mice by ABR threshold analyses. Hear. Res. 130, 94–107.
Steel, K.P., Moorjani, P., and Bock, G.R. (1996). Mixed conductive and sensorineural hearing loss in LP/J mice. Hear. Res. 28, 227–236.
812 Cell 143, 802–812, November 24, 2010 ª2010 Elsevier Inc.
Zhu, H., Guo, Q., and Mattson, M.P. (1999). Dietary restriction protects hippocampal neurons against the death-promoting action of a presenilin-1 mutation. Brain Res. 842, 224–229.
FOXO/4E-BP Signaling in Drosophila Muscles Regulates Organism-wide Proteostasis during Aging Fabio Demontis1,* and Norbert Perrimon1,2,* 1Department
of Genetics Hughes Medical Institute Harvard Medical School, 77 Avenue Louis Pasteur, Boston, MA 02115, USA *Correspondence:
[email protected] (F.D.),
[email protected] (N.P.) DOI 10.1016/j.cell.2010.10.007 2Howard
SUMMARY
The progressive loss of muscle strength during aging is a common degenerative event of unclear pathogenesis. Although muscle functional decline precedes age-related changes in other tissues, its contribution to systemic aging is unknown. Here, we show that muscle aging is characterized in Drosophila by the progressive accumulation of protein aggregates that associate with impaired muscle function. The transcription factor FOXO and its target 4E-BP remove damaged proteins at least in part via the autophagy/lysosome system, whereas foxo mutants have dysfunctional proteostasis. Both FOXO and 4E-BP delay muscle functional decay and extend life span. Moreover, FOXO/4E-BP signaling in muscles decreases feeding behavior and the release of insulin from producing cells, which in turn delays the agerelated accumulation of protein aggregates in other tissues. These findings reveal an organism-wide regulation of proteostasis in response to muscle aging and a key role of FOXO/4E-BP signaling in the coordination of organismal and tissue aging. INTRODUCTION Aging of multicellular organisms involves distinct pathogenic events that include higher mortality, the progressive loss of organ function, and susceptibility to degenerative diseases, some of which arise from protein misfolding and aggregation. Recent genetic studies in the mouse, the nematode Caenorhabditis elegans, and the fruitfly Drosophila melanogaster have expanded our understanding of the evolutionarily conserved signaling pathways regulating aging, with the identification of several mutants that have prolonged or shortened life spans (Kenyon, 2005). Manipulation of longevity-regulating pathways in certain tissues is sufficient to extend life expectancy, indicating that some tissues have a predominant role in life span extension (Libina et al., 2003; Wang et al., 2005; Wolkow et al., 2000). For example, foxo overexpression in the Drosophila fat
body extends life span, indicating a key role of this tissue in the regulation of longevity (Giannakou et al., 2004; Hwangbo et al., 2004). In addition, because most tissues undergo progressive deterioration during aging (Garigan et al., 2002), it is thought that organismal life span may be linked to tissue senescence. However, our understanding of the mechanisms regulating tissue aging and their interconnection to life span is limited. For example, analysis in Drosophila has revealed that the prevention of age-dependent changes in cardiac performance does not alter life span (Wessells et al., 2004), raising the possibility that functional decline in distinct tissues may have different outcomes on the systemic regulation of aging. The Insulin/IGF-1 signaling pathway has been implicated in the control of aging across evolution via its downstream signaling component FOXO (DAF-16 in C. elegans), a member of the fork-head box O transcription factor family (Salih and Brunet, 2008). FOXO regulates the expression of a series of target genes involved in metabolism, cell growth, cell proliferation, stress resistance, and differentiation via direct binding to target gene promoter regions (Salih and Brunet, 2008). Mutations in foxo/ daf-16 reduce life span and stress resistance in both C. elegans and flies, indicating a key role in organism aging (Junger et al., 2003; Salih and Brunet, 2008). In addition to regulating life span, FOXO has been reported to prevent the pathogenesis of some age-related diseases. For example, FOXO reduces the toxicity associated with aggregation-prone human mutant Alzheimer’s and Huntington’s disease proteins (proteotoxicity) in C. elegans and mice, suggesting that regulating protein homeostasis (proteostasis) during aging may have a direct effect on the pathogenesis of human neurodegenerative diseases (Cohen et al., 2006; Hsu et al., 2003; Morley et al., 2002). However, little is known on the protective mechanisms induced in response to FOXO signaling and whether they vary in different aging tissues and disease contexts. Among the plethora of age-related pathological conditions, the gradual decay in muscle strength is one of the first hallmarks of aging in many organisms, including Drosophila, C. elegans, mice and, importantly, humans (Augustin and Partridge, 2009; Herndon et al., 2002; Nair, 2005; Zheng et al., 2005). However, despite its medical relevance, the mechanisms underlying muscle aging are incompletely understood. Functional changes in skeletal muscles temporally precede the manifestation of Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc. 813
Figure 1. FOXO Signaling in Skeletal Muscles Preserves Proteostasis during Aging (A–D) Electron micrographs of immunogold-labeled Drosophila skeletal muscles of wild-type flies at one (A and B) and 5 weeks of age (C and D). Protein aggregates (PA) are detected in the cytoplasm in proximity to mitochondria (Mt) and myofibrils (Myof) in old (C and D) but not young flies (A and B). Numerous gold particles (indicative of anti-ubiquitin immunoreactivity) localize to filamentous structures at 5 weeks of age (C and D), while only a few are present in muscles from young flies. Scale bars are 1 mm (A and C) and 500 nm (B and D).
814 Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc.
aging in other tissues (Herndon et al., 2002), and reduced muscle strength is associated with an increased risk in developing Alzheimer’s and Parkinson’s diseases (Boyle et al., 2009; Chen et al., 2005). However, although aging-related changes in skeletal muscles have been proposed to affect physiological processes in distal organs (Nair, 2005), whether or not muscle senescence modulates the pathogenesis of degenerative events in other tissues is unknown. The fruit fly Drosophila is an excellent model to study muscle aging. The progressive decline in muscle strength and function observed in humans is recapitulated in this system (Rhodenizer et al., 2008), which is amenable to extensive genetic manipulation. By using this model organism, we have searched for the molecular mechanisms responsible for muscle aging and found that decreased protein quality control plays a role in the pathogenesis of age-related muscle weakness. Interestingly, increased activity of the transcription factor FOXO and its target Thor/4E-BP are sufficient to delay this process and preserve muscle function at least in part by promoting the basal activity of the autophagy/lysosome system, an intracellular protein degradation pathway that removes damaged protein aggregates (Rubinsztein, 2006). Moreover, we report that FOXO/4E-BP signaling in muscles extends life span and regulates proteostasis organism-wide by regulating feeding behavior, release of insulin from producing cells, and 4E-BP induction in nonmuscle tissues. Thus, we propose a model by which FOXO/4E-BP signaling in muscles preserves systemic proteostasis by mimicking some of the protective effects of decreased nutrient intake. RESULTS Loss of Proteostasis during Muscle Aging Is Prevented by FOXO To detect cellular processes that are responsible for decreased muscle strength in aging flies, we monitored cellular changes in indirect flight muscles of wild-type flies by immunogold-electron microscopy (IEM). In older flies, we detected filamentous cytoplasmic structures that were instead absent in muscles from young flies (Figures 1A–1D). Filamentous materials present in these structures stained with an anti-ubiquitin antibody (Figure 1D), a marker for proteins that are polyubiquitinated, suggesting that the cytoplasmic structures are aggregates of damaged proteins. Aggregates were variable in size and were detected in both resin-embedded sections (Figure 1) and cryosections (data not shown) of thoracic muscles of the old but not the young flies, in parallel with an increase in the overall
number of gold particles (Figure 1E). To test the hypothesis that muscle function during aging may decrease due to defects in protein homeostasis, we better characterized the age-related deposition of protein aggregates by immunofluorescence. In agreement with the IEM analysis (Figures 1A–1E), we observed that aging skeletal muscles progressively accumulate aggregates of polyubiquitinated proteins (ranging up to several mm) that colocalize with p62/Ref(2)P, an inclusion body component (Figures 1F and1I). The cumulative area of protein aggregates increases during aging (Figure 1L), suggesting that the progressive protein damage, together with a decrease in the turnover of muscle proteins, may result in the age-related decline of muscle strength. To better characterize how protein quality control is linked with aging in muscles, we analyzed the deposition of protein aggregates in syngenic flies with foxo overexpression. Foxo overexpression results in its activation (Giannakou et al., 2004; Hwangbo et al., 2004) and was achieved specifically in muscles via the UAS-Gal4 system using the Mhc-Gal4 driver (see Figure S1 available online). Increased FOXO activity in muscles did not affect developmental growth and differentiation (as estimated by body weight and sarcomere assembly) (Figure S2), and resulted in the delayed accumulation of aggregates containing polyubiquitinated proteins and Ref(2)P during aging (Figures 1G and 1J, compare with control muscles in Figures 1F and 1I). Next, we tested whether foxo null animals display accelerated muscle aging, and found an increased accumulation of protein aggregates (Figures 1H and1K), indicating that FOXO is both necessary and sufficient to modulate muscle proteostasis (Figure 1L). To further corroborate these findings, we overexpressed either the wild-type or the constitutive-active foxo transgenes using the Dmef2-Gal4 muscle driver in combination with the temperature-sensitive tubulin-Gal80ts transgene to achieve adult-onset foxo overexpression in muscles (Figure S3). Transgene overexpression significantly preserved muscle proteostasis in both cases, while the controls displayed an increased accumulation of protein aggregates (Figure S3). All together, these results indicate that protein homeostasis depends on FOXO activity during muscle aging. 4E-BP Controls Proteostasis in Response to Pten/FOXO Activity To dissect the stimuli that encroach on FOXO to control proteostasis, we tested whether Pten overexpression phenocopies FOXO activation. Consistent with its role in activating FOXO, we found that Pten decreased the accumulation of protein
(E) The number of gold particles, indicative of ubiquitin immunoreactivity, significantly increases in old age (standard error of the mean [SEM] is indicated with n; **p < 0.01). (F–L) Immunostaining of indirect flight muscles from flies with (UAS-foxo/+;Mhc-Gal4/+) or without (Mhc-Gal4/+) foxo overexpression at 1 week (F and G) and 5 weeks of age (I and J), and foxo homozygous null (MhcGal4, foxo21/25) flies (H and K). Polyubiquitin (red) and p62/Ref(2)P (green) immunoreactivities reveal an increased deposition of aggregates containing polyubiquitinated proteins during aging in muscles of control flies (F and I), and, to a lesser extent, in muscles overexpressing foxo (G and J). Conversely, muscles from foxo null animals display an accelerated deposition of protein aggregates (H and K) in comparison with controls (F and I). Note the significant increase in the cumulative area of protein aggregates (indicative of both aggregate size and number) in (K) versus (I), and in (I) versus (J), indicating that the control of protein homeostasis is linked to FOXO activity in muscles (quantification in [L]) (SEM is indicated with n; *p < 0.05, **p < 0.01). Representative polyubiquitin and Ref(2)P immunoreactivities are shown in insets. Phalloidin staining (blue) outlines F-actin, which is a component of muscle myofibrils. Scale bar is 20 mm (F–K). See also Figure S1, Figure S2, and Figure S3.
Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc. 815
Figure 2. 4E-BP Preserves Proteostasis in Response to Pten/FOXO Signaling (A–F) Immunostaining of muscles overexpressing Pten and constitutive active (CA) 4E-BP. In both cases, a decrease in the accumulation of polyubiquitin protein aggregates is observed at 5 weeks of age in comparison with age-matched controls, suggesting that these interventions can preserve proteostasis in aging muscles. Scale bar is 20 mm. Hsp70 overexpression has instead limited effects (Figure S4, Table S1, Table S2). (G) A reduction in the cumulative area of protein aggregates is observed upon increased activity of either Pten or 4E-BP in comparison with controls (SEM is indicated with n; **p < 0.01, ***p < 0.001). (H) Relative quantification of Thor/4E-BP mRNA levels from thoraces of syngenic flies at 1 and 5 weeks of age. A significant increase in 4E-BP expression is detected in response to fasting and Pten and FOXO activity (**p < 0.01, ***p < 0.001; SEM is indicated with n = 4).
aggregates during aging (Figures 2B and 2E; see controls in Figures 2A and 2D). Next, we examined the responses induced by Pten/FOXO signaling. First, we examined whether FOXO activity delays protein damage by inducing chaperones that are key for protein quality control (Tower, 2009). In response to FOXO activity in muscles, we detected an increase in the mRNA levels of Hsp70 and its cofactors involved in protein folding (Hip, Hop, Hsp40, and Hsp90) but not in protein degradation (Chip and Chap) (Figure S4 and Table S1). FOXO regulates directly the 816 Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc.
expression of Hsp70 and its cofactors, as estimated with Luciferase transcriptional reporters based on the proximal promoter region of target genes (Figure S4 and Table S2). On this basis, we tested whether Hsp70 overexpression preserves proteostasis during aging but found little changes in the age-related accumulation of protein aggregates (Figure S4). Thus, we conclude that additional FOXO-dependent responses are involved. Among the FOXO-target genes, Thor/4E-BP has a key role in delaying aging by regulating protein translation (Zid et al., 2009; Tain et al., 2009). However, the cellular mechanisms that
are regulated by 4E-BP are largely unknown. To test whether 4E-BP controls proteostasis during muscle aging, we overexpressed a constitutive active form of 4E-BP in muscles and observed limited accumulation of protein aggregates during aging (Figures 2C and 2F) compared with controls (Figures 2A and 2D). All together, increased activity of Pten or 4E-BP significantly decreases the cumulative area of protein aggregates (Figure 2G). In addition, a significant increase in 4E-BP mRNA levels is induced in muscles upon Pten, foxo overexpression, and fasting (Figure 2H). All together, these findings suggest that 4E-BP is key to control proteostasis in response to Pten/FOXO signaling. FOXO/4E-BP Signaling Regulates Proteostasis via the Autophagy/Lysosome System While FOXO/4E-BP signaling mounts a stress resistance response that may decrease the extent of protein damage due to various stressors (Salih and Brunet, 2008; Tain et al., 2009), we wondered whether it regulates the removal of damaged proteins via macroautophagy. In this process, entire regions of the cytoplasm are sequestered in a double membrane vesicle (autophagosome) that subsequently fuses with a lysosome, where the autophagic cargo is degraded (Rubinsztein, 2006). Although the primary role of autophagy is to mount an adaptive response to nutrient deprivation, its basal activity is required for normal protein turnover (Hara et al., 2006). In agreement with this notion, suppression of basal autophagy leads to the accumulation of polyubiquitin protein aggregates in a number of contexts (Korolchuk et al., 2009; Rubinsztein, 2006). To test whether autophagy is regulated in response to FOXO signaling in muscles, we used a GFP-tagged version of the autophagosome marker Atg5 (Rusten et al., 2004). While the number of Atg5-GFP punctae decreases during aging in control muscles (Figures 3A and 3B), it is in part maintained in response to foxo overexpression (Figures 3C and 3D, and quantification in Figure 3E). In addition, given the interconnection between the lysosome system and autophagy, we monitored a GFPtagged version of the lysosome marker Lamp1 (lysosome-associated membrane protein 1) and detected an overall increase in the number of GFP punctae in response to overexpression of the autophagy inducer kinase Atg1, foxo, and 4E-BP CA in muscles at both 1 and 5 weeks of age (Figures 3G–3I and 3K–3M in comparison with controls in Figures 3F and 3J and quantification in Figure 3N). Closer inspection revealed that the abundance of Lamp1-GFP vesicles inversely correlates with the progressive deposition of polyubiquitin protein aggregates, suggesting that FOXO/4E-BP signaling regulates proteostasis at least in part via the autophagy/lysosome system. To further test this hypothesis, we analyzed the age-related changes in autophagy gene expression, which have been previously used as a correlative measurement of autophagic activity (Gorski et al., 2003; Simonsen et al., 2008). Interestingly, the expression of several autophagy genes involved in autophagosome induction (Atg1), nucleation (Atg6), and elongation (Atg5, Atg7, and Atg8) progressively declines during aging in muscles (Figure 3O), suggesting that gene expression changes likely contribute to the accumulation of damaged proteins. Conversely, foxo overexpression increased
the basal expression of several Atg genes at both young and old age, suggesting that their increased expression contributes to the beneficial effects of FOXO on proteostasis. To test this hypothesis, we knocked down Atg7 levels in foxo-overexpressing flies and analyzed the deposition of polyuiquitinated protein aggregates. Interestingly, RNAi treatment brought about a 50% decrease in Atg7 mRNA levels and resulted in a partial increase in the buildup of insoluble ubiquitinated proteins at 8 weeks, compared with age-matched, mock-treated flies (white RNAi) and 1-week-old flies (Figure 3P). All together, these findings suggest that FOXO/4E-BP signaling prevents the buildup in protein damage, at least in part by promoting the basal activity of the autophagy/lysosome system. Prevention of Muscle Aging by FOXO and 4E-BP Extends Life Span To evaluate whether preserving proteostasis can prevent functional alterations in aging muscles, we assessed muscle strength with negative geotaxis and flight assays (see Experimental Procedures). As shown in Figures 4A and 4B, muscle functionality gradually decreases in aging flies, resulting in impaired climbing and flight ability. Notably, foxo (Figure 4A) and 4E-BP activity (Figure 4B) significantly preserve muscle strength during aging. Thus, FOXO and 4E-BP prevent both the cellular degenerative events and the functional decay of aging muscles. Epidemiological studies in humans have associated muscle senescence with increased mortality (Nair, 2005), implying that muscle aging may have organism-wide consequences beyond muscle function. To ask whether the prevention of muscle aging affects the organism life span, we manipulated the activity of components of the Akt pathway in muscles and scored for their effects on viability. As shown in Figures 4C and 4D, either Pten, foxo, or 4E-BP CA overexpression in muscles is sufficient to significantly extend longevity by increasing the median and maximum life span. 4E-BP increased life span also in foxo heterozygous null animals (Figure 4D), while Hsp70 overexpression on the other hand showed little effects (Figure S5). All together, these findings indicate that the extent of muscle aging is interconnected with the life span of the organism. FOXO/4E-BP Signaling in Muscles Influences Feeding Behavior and the Release of Insulin from Producing Cells Considering that both fasting and FOXO induce 4E-BP expression (Figure 2H), we wondered whether the systemic effect of FOXO signaling on life span extension can result, at least in part, from reduced food intake. To test this hypothesis, we examined whether feeding behavior would be decreased in adults with FOXO and 4E-BP activation in muscles. We first monitored the amount of liquid food ingested using the CAFE´ assay (capillary feeding) (Ja et al., 2007). Interestingly, feeding was decreased in response to FOXO/4E-BP signaling in muscles (Figure 5A). To substantiate this finding, we measured the ingestion of bluecolored food (Xu et al., 2008) and detected significant differences in food intake with this assay (Figure 5B), confirming Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc. 817
818 Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc.
Figure 4. FOXO/4E-BP Signaling Preserves Muscle Function and Extends Life Span (A) Muscle function gradually decreases during aging as indicated by an increase in the percentage of flies with climbing and flight defects. However, foxo preserves their function in comparison with controls (flight ability: n[flies] = 10 (week 1 and 5) and 30 (week 8) with n[batch] = 3 (week 1 and 5) and 2 (week 8); standard deviation (SD) is indicated and *p < 0.05. Climbing ability: (n[Mhc-Gal4/+] = 1264, n[Mhc-Gal4/UAS-foxo] = 966, with n indicating the number of flies at day 1; p < 0.001). (B) Similar to FOXO, 4E-BP activity also results in decreased age-related flight and climbing deficits in comparison with controls (flight ability: n[flies] R 10 (week 1 and 5) and 25 (week 8) with n[batch] R 3 (week 1 and 5) and 2 (week 8); SD is indicated and *p < 0.05. Climbing ability: (n[Mhc-Gal4/+] = 204, n[Mhc-Gal4/UAS-4EBP CA] = 403, p < 0.001). (C) Survival of flies during aging. Foxo overexpression in muscles significantly extends the median and maximum life span (median and maximum life span: Mhc-Gal4/+ = 61 and 82 days (n = 1264); UAS-foxo tr.#1/+;Mhc-Gal4/+ = 73 and 100 days (n = 1184); Mhc-Gal4/UAS-foxo tr.#2 = 76 and 94 days (n = 966); p < 0.001). (D) Life span of flies with increased Pten and 4E-BP activity in muscles is extended in comparison with matched controls (median and maximum life span of 4E-BP: Mhc-Gal4/+ = 63 and 78 days (n = 204); Mhc-Gal4/UAS-4E-BP CA = 71 and 84 days (n = 403); Pten: Mhc-Gal4/+ = 55 and 76 days (n = 162); Mhc-Gal4/UAS-Pten = 66 and 88 days (n = 130); p < 0.001). Similar increase in life span is brought about by 4E-BP CA overexpression in foxo21 heterozygous null flies. See also Figure S5 and Figure S7.
that feeding behavior is affected. Next, to assess whether decreased feeding behavior arises from developmental defects, we measured the body weight of adult flies, which is a sensitive indicator of developmental feeding (Demontis and Perrimon, 2009), but found no significant differences (Figure 5C). Thus,
the behavior of flies overexpressing foxo and 4E-BP CA in muscles most likely is not caused by developmental defects. To assess the metabolic status, we monitored the glucose concentration (glycemia) in the hemolymph. Similar to wildtype flies starved for 24 hr, we detected a significant decrease
Figure 3. FOXO and 4E-BP Regulate Proteostasis at Least in Part via the Autophagy/Lysosome System (A–E) Immunostaining of muscles expressing the marker of autophagosomes Atg5-GFP reveals a significant increase in their number (E) and maintenance at 1 and 5 weeks of age upon foxo overexpression (C and D) in comparison with controls (A and B). In (E), SEM is indicated with n; *p < 0.05 and **p < 0.01. (F–N) Immunostaining of muscles expressing the lysosomal marker Lamp1-GFP and overexpressing either Atg1, foxo, or 4E-BP CA. Note an increase in the number of lysosomes (N) at both 1 (G-I) and 5 weeks of age (K–M), which inversely correlates with polyubiquitin immunoreactivity in comparison with control muscles (F and J). Scale bar is 10 mm (A–D and F-–M). In (N), SEM is indicated with n; *p < 0.05 and ***p < 0.001. (O) Relative mRNA levels of autophagy genes from thoraces of 1- and 5-week-old flies decrease during normal muscle aging, while their expression increases and persists in response to FOXO. SEM is indicated with n = 4; *p < 0.05, **p < 0.01 and ***p < 0.001. (P) RNAi treatment against Atg7 results in a 50% knockdown of its mRNA levels in muscles and partially impairs FOXO-mediated proteostasis, as indicated by the increased detection of ubiquitin-conjugated proteins in Triton X-100 insoluble fractions at 8 weeks (old, red) in comparison with mock-treated (white RNAi) and young flies (1 week old, black). Normalized values based on a-tubulin levels are indicated.
Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc. 819
Figure 5. FOXO Signaling in Muscles Partially Mimics Systemic Metabolic Changes Associated with Fasting by Modulating Feeding Behavior (A–C) Flies in which FOXO/4E-BP activity has been altered specifically in muscles consume less food than matched controls. Food consumption was determined via capillary feeding CAFE´ assay over 2 hr periods (A), and by monitoring the ingestion of blue colored food in 24 hr (B). Error bars represent SEM with n[measurements] = 44, 46, 52, 37, 103, and 61 in (A) and n = 2 in (B), with *p < 0.05, **p < 0.01, ***p < 0.001. Decreased feeding does not result from developmental defects, as indicated by similar body weights of flies analyzed (C) (error bars represent SD with n R 3). (D) Relative glucose levels (glycemia) in the hemolymph of flies overexpressing either foxo or 4E-BP CA in muscles, and matched controls. Manipulation of FOXO/4E-BP signaling in muscles brings about a reduction of glycemia similar in part to that of wild-type flies starved for 24 hr, as estimated with the glucose hexokinase assay (SEM is indicated with n = 5, and **p < 0.01, ***p < 0.001). (E–H) Immunostaining of Dilp-producing median neurosecretory cells in the brain of starved wild-type flies, flies overexpressing foxo in muscles, and controls. Increase in the immunoreactivity of the insulin-like peptide Dilp2 (green) is detected in producing cells in response to either starvation (F) or foxo overexpression in muscles (H), in comparison respectively with fed wild-type flies (E) and controls with no foxo overexpression in muscles (G). Smaller changes in Dilp5 levels are observed. Phalloidin staining (blue) detects F-actin (scale bar is 20 mm; images in [E]–[H] have the same magnification). (I) Quantification of the intensity of staining indicates that differences in Dilp2 fluorescence are significant (SD is indicated with n[measurements] = 35, 69, 37, and 96 from n[brains] = 2, 4, 3, and 4; *p < 0.05). (J–L) Quantification and immunostaining of adipose tissue (peripheral fat body of the abdomen) from 2 week old flies. (J) Note a significant increase in nuclear b-galactosidase immunoreactivity (red) in the adipose tissue from flies with a nuclear 4E-BP-lacZ reporter and foxo overexpression in muscles (L) in comparison with controls (K). F-actin (green) and DAPI staining (indicative of nuclei, blue) are shown. Scale bar is 20 mm. In (J), SEM is indicated with n = 20 and ***p < 0.001.
of glycemia in flies with FOXO and 4E-BP activation in muscles (Figure 5D). All together, these findings suggest that FOXO and 4E-BP act as a metabolic brake in muscles that, by influencing 820 Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc.
feeding behavior, mimic at least in part the physiological changes that are associated with fasting. To gain mechanistic insights into the systemic regulation of aging by FOXO/4E-BP signaling in muscles, we next monitored the release of insulin-like peptides (Dilps) from the Dilpproducing median neurosecretory cells in the brain, which have been previously shown to mediate the response of life span to nutrition in Drosophila (Broughton et al., 2010). We detected a significant accumulation of the insulin-like peptide Dilp2 (and to a lesser extent, Dilp5) in starved wild-type flies in comparison with fed flies (Figures 5E and 5F). Increased immunoreactivity indicates decreased release of Dilps and has been previously shown to
occur in response to starvation (Geminard et al., 2009). Next, we tested whether similar changes would occur upon FOXO signaling in muscles and found a partial accumulation of Dilps (Figures 5G–5I). Assuming that decreased Dilps secretion may result in systemic FOXO activation, we monitored its activity using a nuclear 4E-BP-lacZ transcriptional reporter. By immunostaining adipose tissues with anti-b-galactosidase antibodies, we detected higher 4E-BP expression upon foxo activation in muscles in comparison with controls (Figures 5J–5L). Thus, FOXO signaling in muscles appears to systemically activate 4E-BP expression in other tissues by regulating food intake and insulin release.
activity in muscles also confers systemic protection from the age-related decline in proteostasis. To test whether this effect is muscle-specific, we overexpressed foxo in the adipose tissue (abdominal fat body) with the S106GS-Gal4 driver, and analyzed the deposition of polyubiquitinated proteins in Triton X-100 insoluble fractions from thoraces. Under these conditions, we seemingly detected no differences (Figure S6), suggesting that, although other tissues may be involved, muscles may play a key role in this regulation. Altogether, these observations suggest that FOXO and 4E-BP activity in muscles mitigates the loss of proteostasis nonautonomously by influencing feeding behavior, insulin release from producing cells, and 4E-BP activity in other tissues.
FOXO/4E-BP Signaling in Muscles Regulates Proteostasis in Other Aging Tissues Our demonstration that FOXO/4E-BP signaling in muscles extends life span in Drosophila and induces a systemic fastinglike response, along with the observation that muscles undergo age-related structural and functional changes precociously in comparison with other tissues (Herndon et al., 2002; Zheng et al., 2005), raises the possibility that muscle senescence may influence the progression of age-related degenerative events in the entire organism. To test this hypothesis, we examined whether, in addition to life span extension, FOXO signaling in muscles can affect protein homeostasis in other tissues. As in the case of muscles (Figure 1 and Figure 2), we found that Ref(2)P/polyubiquitin aggregates progressively accumulate in aging retinas (Figures 6A and 6D), brains (Figures 6B and 6E), and adipose tissue (Figures 6C and 6F) (peripheral fat body of the abdomen). However, foxo overexpression in muscle resulted in decreased accumulation of protein aggregates in other aging tissues (Figures 6D–6F; quantification in Figure 6G). Similar changes were observed in response to 4E-BP activity in muscles in comparison with syngenic controls (Figure 6H). Importantly, this regulation is muscle nonautonomous, as Mhc-Gal4 drives transgene expression only in muscles (and not in the retina, brain or adipose tissue) (Figure S1). To further test the finding that FOXO/4E-BP signaling in muscles delays the systemic impairment of proteostasis in other tissues (Figures 6A–6H), we analyzed by western blot the ubiquitin levels of Triton X-100 insoluble fractions, which included protein aggregates, from either thoraces (which mainly consist of foxo-overexpressing muscles) or heads and abdomens (which are enriched in nonmuscle tissues and muscles with little foxo overexpression) (Figure S1), at 1 and 8 weeks of age. In agreement with the increased deposition of protein aggregates observed during aging by immunofluorescence (Figure 1, Figure 2, and Figures 6A–6F), ubiquitin levels were dramatically increased in the Triton X-100 insoluble fractions from control thoraces, and head and abdominal extracts at 8 weeks of age, in comparison with 1 week of age (Figure 6I). However, ubiquitin levels were only partially increased in old foxo-overexpressing flies in both thoracic and head and abdominal extracts. No substantial differences were instead detected in the Triton X-100 soluble fractions (data not shown). Similar results were obtained by 4E-BP CA but not Hsp70 overexpression in muscles (Figure 6I; Figure S5), indicating that 4E-BP
DISCUSSION By using a number of behavioral, genetic, and molecular assays, we have described a mechanism in the pathogenesis of muscle aging that is based on the loss of protein homeostasis (proteostasis) and the resulting decrease in muscle strength (Figure 7). Increased activity of Pten and the transcription factor FOXO is sufficient to delay this process, while foxo null animals experience accelerated loss of proteostasis during muscle aging. Pten and FOXO induce multiple protective responses, including the expression of folding chaperones and the regulator of protein translation 4E-BP that has a pivotal role in preserving proteostasis. FOXO and 4E-BP preserve muscle function, at least in part by sustaining the basal activity of the autophagy/lysosome system, which removes aggregates of damaged proteins. However, additional mechanisms may be involved. For example, the proteasome system may degrade damaged proteins and thus avoid their accumulation in aggregates (Rubinsztein, 2006). Thus, perturbation in proteasome assembly and subunit composition may contribute to muscle aging in response to FOXO activity. In addition, whereas overexpression of a single chaperone had limited effects, interventions to effectively limit the extent of protein damage are likely to delay the decay in proteostasis by decreasing the workload for the proteasome and autophagy systems (Tower, 2009). By comparing the accumulation of polyubiquitinated proteins in aggregates of aging muscles, retinas, brains, and adipose tissue, we have found that reduced protein homeostasis is a general feature of tissue aging that is particularly prominent in muscles (Figure 1, Figure 6, and Figure S6). The observation that muscle aging is characterized by loss of proteostasis further suggests some similarity between muscle aging and neurodegenerative diseases, many of which are characterized by the accumulation of protein aggregates (Rubinsztein, 2006). Mechanical, thermal, and oxidative stressors occur during muscle contraction (Arndt et al., 2010), and therefore muscle proteins may be particularly susceptible to damage in comparison with other tissues. While our findings refer to the loss of proteostasis in the context of normal aging, it is likely that a better understanding of this process will help cure muscle pathologies associated with aging, as some of the underlying mechanisms of etiology may be shared. For example, most cases of inclusion body myositis (IBM) arise over the age of 50 years, defining aging as a major risk factor for the pathogenesis of this disease. Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc. 821
Figure 6. Systemic Proteostasis Is Remotely Controlled by FOXO/4E-BP Signaling in Muscles (A–F) Aggregates of polyubiquitinated proteins accumulate during aging in the retina (A and D), brain (B and E), and the adipose tissue (C and F) of control flies (Mhc-Gal4/+), but to a lesser extent in tissues from flies overexpressing foxo in muscles (UAS-foxo/+;Mhc-Gal4/+), as indicated by polyubiquitin (red) and p62/Ref (2)P (green) stainings. Phalloidin staining (blue) outlines F-actin. Note that Mhc-Gal4 does not drive transgene expression in these tissues (Figure S1). Scale bar is 10 mm. (G and H) The age-related increase in the cumulative area of protein aggregates is significantly less prominent in tissues from flies overexpressing foxo (G) or 4E-BP CA (H) in muscles in comparison with controls (SEM is indicated with n; *p < 0.05. **p < 0.01, and ***p < 0.001). (I) Ubiquitin levels (indicative of protein aggregates) are detected in Triton X-100 insoluble fractions from thoraces, and head and abdominal tissues from flies overexpressing foxo in muscles or control flies at 1 (young, black) and 8 (old, red) weeks of age. Ubiquitin levels are increased in old flies in comparison with young
822 Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc.
Figure 7. FOXO/4E-BP Signaling in Muscles Controls Proteostasis and Systemic Aging Muscle aging is characterized by protein damage and accumulation of cytoplasmic aggregates. Loss of protein homeostasis (proteostasis) associates with the progressive decrease in muscle strength and can affect the life span of the organism. Pten/FOXO signaling induces multiple targets including several folding chaperones and the regulator of protein translation 4E-BP. FOXO/4E-BP activity regulates muscle proteostasis at least in part via the autophagy/lysosome pathway of protein degradation, preserves muscle function, and extends life span. In addition, FOXO/ 4E-BP signaling in muscles decreases feeding behavior that, similar to fasting, results in reduced insulin release from producing cells. This in turn promotes FOXO and 4E-BP activity in other tissues, preserving proteostasis organism-wide and mitigating systemic aging.
Interestingly, muscle weakness in patients with IBM is characterized by the accumulation of protein aggregates (Needham and Mastaglia, 2008), which we have now described as occurring in the context of regular muscle aging in Drosophila. Thus, FOXO may interfere with the pathogenesis of muscle degenerative diseases in addition to muscle aging. Studies in animal disease models of IBM will be needed to test this hypothesis. There is an apparent contradiction between our findings and the data describing the FOXO-dependent induction of muscle atrophy in mice (Bodine et al., 2001; Sandri et al., 2004), a serious form of muscle degeneration that results in decreased muscle strength (Augustin and Partridge, 2009). The observation that different degrees of FOXO activation can promote stress resistance, or rather cell death (Salih and Brunet, 2008), could explain why FOXO activity can be protective or rather detrimental during muscle aging. In particular, while physiologic FOXO activation can preserve protein homeostasis and muscle function, its excessive activation may lead to decreased muscle function due to hyperactivation of the protein turnover pathways. Consistent with this view, the autophagy pathway has also been involved in both muscle atrophy (Mammucari et al., 2007; Zhao et al., 2007) and in the preservation of muscle sarcomere organization (Arndt et al., 2010; Masiero et al., 2009), highlighting the importance of fine-tuning the degree of activation of stress resistance pathways to maintain muscle homeostasis. In addition, the output of FOXO activity may radically differ in growing versus preexisting myofibers. In particular, our present study indicates that FOXO protects preexisting myofibers
against age-dependent changes in proteostasis while also blunting developmental muscle growth in flies (Demontis and Perrimon, 2009), as observed in mammals (Kamei et al., 2004). Thus, deleterious effects of FOXO activation as observed in mammalian muscles may result from the inhibition of the growth of novel myofibers in postnatal development and adulthood, a process which is thought to be limited to development in Drosophila (Grefte et al., 2007). An interesting observation of our study is that interventions that decrease muscle aging also extend the life span of the organism. In particular, our work raises the prospect that the extent of muscle aging may be a key determinant of systemic aging (Figure 7). Reduced muscle proteostasis may be detrimental per se for life expectancy, presumably due to the involvement of muscles in a number of key physiological functions. Consistent with this view, overexpression in muscles of aggregation-prone human Huntington’s disease proteins is sufficient to decrease life span (Figure S7). Moreover, FOXO signaling in muscles regulates proteostasis in other tissues, via the inhibition of feeding behavior and the decreased release of insulin from producing cells, which in turn promote 4E-BP activity systemically. Thus, we propose that FOXO/4E-BP signaling in muscles regulates life span and remotely controls aging events in other tissues by bringing about some of the protection associated with decreased food intake. In mammals, muscles produce a number of cytokines involved in the control of systemic metabolism (Nair, 2005; Pedersen and Febbraio, 2008). For example, interleukin-6 (IL-6) is produced by muscles and has been proposed to control glucose homeostasis and feeding behavior through peripheral and brain mechanisms (Febbraio and Pedersen, 2002; Plata-Salaman, 1998). Thus,
flies in extracts from both muscles (thoraces) and nonmuscle tissues (heads and abdomens). However, flies overexpressing foxo in muscles have reduced deposition of protein aggregates at 8 weeks of age in both muscles and nonmuscle tissues. Similar results are obtained in response to increased 4E-BP activity in muscles (I), but not Hsp70 (Figure S5). Quantification of ubiquitin-conjugated proteins normalized to a-tubulin or histone H3 levels is indicated. See also Figures S1, Figure S5, and Figure S6.
Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc. 823
a muscle-based network of systemic aging as observed in flies may occur in humans. This study supports the common belief that preserving muscle function is beneficial for overall aging (Boyle et al., 2009; Chen et al., 2005), and the notion that muscles are central tissues to coordinate organism-wide processes, including aging and metabolic homeostasis (Nair, 2005). Moreover, the observation that FOXO signaling in muscles influences aging events in other tissues suggests that the systemic regulation of aging relies on tissue-to-tissue communication (Russell and Kahn, 2007), which may provide the basis for interventions to extend healthy life span. EXPERIMENTAL PROCEDURES Drosophila Strains and Life Span Analysis Details on fly strains can be found in Extended Experimental Procedures. For longevity measurement, male flies were collected within 24 hr from eclosion and reared at standard density (20 flies per vial) on cornmeal/soy flour/yeast fly food at 25 C. Dead flies were counted every other day and food changed. For each genotype, at least two independent cohorts of flies, raised at different times from independent crosses, were analyzed. For starvation treatments, flies were kept in normal vials with 1.5% agar as a water source for the period of time indicated. For all experiments, Mhc-Gal4 females were mated with male transgenic and syngenic control flies, and the resulting male offspring analyzed in parallel by comparing transgene expressing flies with matched controls flies having the same genetic background. For transgene expression with the Gal4-UAS system, flies were reared at 25 C. Behavioral and Metabolic Assays Flight ability was scored according to Park et al. (2006), and negative geotaxis assays were performed as previously described (Rhodenizer et al., 2008). In brief, flies were gently tapped to the bottom of a plastic vial, and the number of flies that could climb to the top of the vial after 20 s was scored. Quantification of the glucose concentration in the hemolymph, and capillary (CAFE´) and blue-colored food feeding assays were done as previously described (Geminard et al., 2009; Xu et al., 2008) and are described in detail in Extended Experimental Procedures. Immunostaining, Confocal and Electron Microscopy, and Image Analysis For whole-mount immunostaining of the fly tissues, indirect flight muscles, and peripheral fat body of the abdomen, retinas, and brains were dissected from male flies and fixed for 30–40 min in PBS with 4% paraformaldehyde and 0.2% Triton X-100. After washing, samples were incubated overnight with appropriate primary and secondary antibodies. Image analysis was done with ImageJ and Photoshop. Immuno-gold electron microscopy was done similar to Nezis et al., (2008). See Extended Experimental Procedures for further information and a list of the antibodies used. Quantitative Real-Time RT-PCR qRT-PCR was done as previously described (Demontis and Perrimon, 2009). Total RNA was prepared from fly thoraces and qRT-PCR was performed with the QuantiTect SYBR Green PCR kit (QIAGEN). Alpha-Tubulin 84B was used as normalization reference. Relative quantification of mRNA levels was calculated using the comparative CT method. Statistical Analysis Statistical analysis was performed with Excel (Microsoft) and p values were calculated with Student’s t tests and log-rank tests. Western Blot and Biochemical Analysis of Detergent-Insoluble Fractions Western blot and biochemical analysis of detergent-insoluble fractions were done substantially as previously described (Nezis et al., 2008). In brief,
824 Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc.
dissected flies were homogenized in ice-cold PBS with 1% Triton X-100 and protease inhibitors, and the resulting unsoluble pellet resuspended in RIPA buffer with 5% SDS and 8M urea. See Extended Experimental Procedures for a complete protocol. SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven figures, and two tables and can be found with this article online at doi:10.1016/j.cell.2010.10.007. ACKNOWLEDGMENTS We are grateful to Andreas Brech, Didier Contamine, Ernst Hafen, Pierre Leopold, Susan Lindquist, Ioannis Nezis, Amita Sehgal, Marc Tatar, Robert Tjian, John Tower, the DRSC/TRiP, and members of the Perrimon lab for fly stocks, reagents, and advice. We thank Maria Ericsson for assistance with electron microscopy, Christians Villalta for embryo injection, and Chris Bakal, Rami Rahal, and Jonathan Zirin for critically reading the manuscript. This work was supported by the NIH (1P01CA120964-01A1) and a Pilot Project Grant from the Paul F. Glenn Labs for the Molecular Biology of Aging. F.D. is an Ellison Medical Foundation/AFAR postdoctoral fellow. N.P. is an investigator of the Howard Hughes Medical Institute. Received: February 3, 2010 Revised: June 24, 2010 Accepted: October 1, 2010 Published: November 24, 2010 REFERENCES Arndt, V., Dick, N., Tawo, R., Dreiseidler, M., Wenzel, D., Hesse, M., Furst, D.O., Saftig, P., Saint, R., Fleischmann, B.K., et al. (2010). Chaperone-assisted selective autophagy is essential for muscle maintenance. Curr. Biol. 20, 143–148. Augustin, H., and Partridge, L. (2009). Invertebrate models of age-related muscle degeneration. Biochim. Biophys. Acta 1790, 1084–1094. Bodine, S.C., Stitt, T.N., Gonzalez, M., Kline, W.O., Stover, G.L., Bauerlein, R., Zlotchenko, E., Scrimgeour, A., Lawrence, J.C., Glass, D.J., and Yancopoulos, G.D. (2001). Akt/mTOR pathway is a crucial regulator of skeletal muscle hypertrophy and can prevent muscle atrophy in vivo. Nat. Cell Biol. 3, 1014–1019. Boyle, P.A., Buchman, A.S., Wilson, R.S., Leurgans, S.E., and Bennett, D.A. (2009). Association of muscle strength with the risk of Alzheimer disease and the rate of cognitive decline in community-dwelling older persons. Arch. Neurol. 66, 1339–1344. Broughton, S.J., Slack, C., Alic, N., Metaxakis, A., Bass, T.M., Driege, Y., and Partridge, L. (2010). DILP-producing median neurosecretory cells in the Drosophila brain mediate the response of lifespan to nutrition. Aging Cell 9, 336–346. Chen, H., Zhang, S.M., Schwarzschild, M.A., Hernan, M.A., and Ascherio, A. (2005). Physical activity and the risk of Parkinson disease. Neurology 64, 664–669. Cohen, E., Bieschke, J., Perciavalle, R.M., Kelly, J.W., and Dillin, A. (2006). Opposing activities protect against age-onset proteotoxicity. Science 313, 1604–1610. Demontis, F., and Perrimon, N. (2009). Integration of Insulin receptor/Foxo signaling and dMyc activity during muscle growth regulates body size in Drosophila. Development 136, 983–993. Febbraio, M.A., and Pedersen, B.K. (2002). Muscle-derived interleukin-6: mechanisms for activation and possible biological roles. FASEB J. 16, 1335– 1347. Garigan, D., Hsu, A.L., Fraser, A.G., Kamath, R.S., Ahringer, J., and Kenyon, C. (2002). Genetic analysis of tissue aging in Caenorhabditis elegans: a role for heat-shock factor and bacterial proliferation. Genetics 161, 1101–1112.
Geminard, C., Rulifson, E.J., and Leopold, P. (2009). Remote control of insulin secretion by fat cells in Drosophila. Cell Metab. 10, 199–207. Giannakou, M.E., Goss, M., Junger, M.A., Hafen, E., Leevers, S.J., and Partridge, L. (2004). Long-lived Drosophila with overexpressed dFOXO in adult fat body. Science 305, 361. Gorski, S.M., Chittaranjan, S., Pleasance, E.D., Freeman, J.D., Anderson, C.L., Varhol, R.J., Coughlin, S.M., Zuyderduyn, S.D., Jones, S.J., and Marra, M.A. (2003). A SAGE approach to discovery of genes involved in autophagic cell death. Curr. Biol. 13, 358–363. Grefte, S., Kuijpers-Jagtman, A.M., Torensma, R., and Von den Hoff, J.W. (2007). Skeletal muscle development and regeneration. Stem Cells Dev. 16, 857–868. Hara, T., Nakamura, K., Matsui, M., Yamamoto, A., Nakahara, Y., SuzukiMigishima, R., Yokoyama, M., Mishima, K., Saito, I., Okano, H., and Mizushima, N. (2006). Suppression of basal autophagy in neural cells causes neurodegenerative disease in mice. Nature 441, 885–889. Herndon, L.A., Schmeissner, P.J., Dudaronek, J.M., Brown, P.A., Listner, K.M., Sakano, Y., Paupard, M.C., Hall, D.H., and Driscoll, M. (2002). Stochastic and genetic factors influence tissue-specific decline in ageing C. elegans. Nature 419, 808–814. Hsu, A.L., Murphy, C.T., and Kenyon, C. (2003). Regulation of aging and agerelated disease by DAF-16 and heat-shock factor. Science 300, 1142–1145. Hwangbo, D.S., Gershman, B., Tu, M.P., Palmer, M., and Tatar, M. (2004). Drosophila dFOXO controls lifespan and regulates insulin signalling in brain and fat body. Nature 429, 562–566. Ja, W.W., Carvalho, G.B., Mak, E.M., de la Rosa, N.N., Fang, A.Y., Liong, J.C., Brummel, T., and Benzer, S. (2007). Prandiology of Drosophila and the CAFE assay. Proc. Natl. Acad. Sci. USA 104, 8253–8256. Junger, M.A., Rintelen, F., Stocker, H., Wasserman, J.D., Vegh, M., Radimerski, T., Greenberg, M.E., and Hafen, E. (2003). The Drosophila forkhead transcription factor FOXO mediates the reduction in cell number associated with reduced insulin signaling. J. Biol. 2, 20. Kamei, Y., Miura, S., Suzuki, M., Kai, Y., Mizukami, J., Taniguchi, T., Mochida, K., Hata, T., Matsuda, J., Aburatani, H., et al. (2004). Skeletal muscle FOXO1 (FKHR) transgenic mice have less skeletal muscle mass, down-regulated Type I (slow twitch/red muscle) fiber genes, and impaired glycemic control. J. Biol. Chem. 279, 41114–41123. Kenyon, C. (2005). The plasticity of aging: insights from long-lived mutants. Cell 120, 449–460. Korolchuk, V.I., Mansilla, A., Menzies, F.M., and Rubinsztein, D.C. (2009). Autophagy inhibition compromises degradation of ubiquitin-proteasome pathway substrates. Mol. Cell 33, 517–527. Libina, N., Berman, J.R., and Kenyon, C. (2003). Tissue-specific activities of C. elegans DAF-16 in the regulation of lifespan. Cell 115, 489–502. Mammucari, C., Milan, G., Romanello, V., Masiero, E., Rudolf, R., Del Piccolo, P., Burden, S.J., Di Lisi, R., Sandri, C., Zhao, J., et al. (2007). FoxO3 controls autophagy in skeletal muscle in vivo. Cell Metab. 6, 458–471. Masiero, E., Agatea, L., Mammucari, C., Blaauw, B., Loro, E., Komatsu, M., Metzger, D., Reggiani, C., Schiaffino, S., and Sandri, M. (2009). Autophagy is required to maintain muscle mass. Cell Metab. 10, 507–515. Morley, J.F., Brignull, H.R., Weyers, J.J., and Morimoto, R.I. (2002). The threshold for polyglutamine-expansion protein aggregation and cellular toxicity is dynamic and influenced by aging in Caenorhabditis elegans. Proc. Natl. Acad. Sci. USA 99, 10417–10422. Nair, K.S. (2005). Aging muscle. Am. J. Clin. Nutr. 81, 953–963. Needham, M., and Mastaglia, F.L. (2008). Sporadic inclusion body myositis: a continuing puzzle. Neuromuscul. Disord. 18, 6–16.
Nezis, I.P., Simonsen, A., Sagona, A.P., Finley, K., Gaumer, S., Contamine, D., Rusten, T.E., Stenmark, H., and Brech, A. (2008). Ref(2)P, the Drosophila melanogaster homologue of mammalian p62, is required for the formation of protein aggregates in adult brain. J. Cell Biol. 180, 1065–1071. Park, J., Lee, S.B., Lee, S., Kim, Y., Song, S., Kim, S., Bae, E., Kim, J., Shong, M., Kim, J.M., and Chung, J. (2006). Mitochondrial dysfunction in Drosophila PINK1 mutants is complemented by parkin. Nature 441, 1157–1161. Pedersen, B.K., and Febbraio, M.A. (2008). Muscle as an endocrine organ: focus on muscle-derived interleukin-6. Physiol. Rev. 88, 1379–1406. Plata-Salaman, C.R. (1998). Cytokines and feeding. News Physiol. Sci. 13, 298–304. Rhodenizer, D., Martin, I., Bhandari, P., Pletcher, S.D., and Grotewiel, M. (2008). Genetic and environmental factors impact age-related impairment of negative geotaxis in Drosophila by altering age-dependent climbing speed. Exp. Gerontol. 43, 739–748. Rubinsztein, D.C. (2006). The roles of intracellular protein-degradation pathways in neurodegeneration. Nature 443, 780–786. Russell, S.J., and Kahn, C.R. (2007). Endocrine regulation of ageing. Nat. Rev. Mol. Cell Biol. 8, 681–691. Rusten, T.E., Lindmo, K., Juhasz, G., Sass, M., Seglen, P.O., Brech, A., and Stenmark, H. (2004). Programmed autophagy in the Drosophila fat body is induced by ecdysone through regulation of the PI3K pathway. Dev. Cell 7, 179–192. Salih, D.A., and Brunet, A. (2008). FoxO transcription factors in the maintenance of cellular homeostasis during aging. Curr. Opin. Cell Biol. 20, 126–136. Sandri, M., Sandri, C., Gilbert, A., Skurk, C., Calabria, E., Picard, A., Walsh, K., Schiaffino, S., Lecker, S.H., and Goldberg, A.L. (2004). Foxo transcription factors induce the atrophy-related ubiquitin ligase atrogin-1 and cause skeletal muscle atrophy. Cell 117, 399–412. Simonsen, A., Cumming, R.C., Brech, A., Isakson, P., Schubert, D.R., and Finley, K.D. (2008). Promoting basal levels of autophagy in the nervous system enhances longevity and oxidant resistance in adult Drosophila. Autophagy 4, 176–184. Tain, L.S., Mortiboys, H., Tao, R.N., Ziviani, E., Bandmann, O., and Whitworth, A.J. (2009). Rapamycin activation of 4E-BP prevents parkinsonian dopaminergic neuron loss. Nat. Neurosci. 12, 1129–1135. Tower, J. (2009). Hsps and aging. Trends Endocrinol. Metab. 20, 216–222. Wang, M.C., Bohmann, D., and Jasper, H. (2005). JNK extends life span and limits growth by antagonizing cellular and organism-wide responses to insulin signaling. Cell 121, 115–125. Wessells, R.J., Fitzgerald, E., Cypser, J.R., Tatar, M., and Bodmer, R. (2004). Insulin regulation of heart function in aging fruit flies. Nat. Genet. 36, 1275– 1281. Wolkow, C.A., Kimura, K.D., Lee, M.S., and Ruvkun, G. (2000). Regulation of C. elegans life-span by insulinlike signaling in the nervous system. Science 290, 147–150. Xu, K., Zheng, X., and Sehgal, A. (2008). Regulation of feeding and metabolism by neuronal and peripheral clocks in Drosophila. Cell Metab. 8, 289–300. Zhao, J., Brault, J.J., Schild, A., Cao, P., Sandri, M., Schiaffino, S., Lecker, S.H., and Goldberg, A.L. (2007). FoxO3 coordinately activates protein degradation by the autophagic/lysosomal and proteasomal pathways in atrophying muscle cells. Cell Metab. 6, 472–483. Zheng, J., Edelman, S.W., Tharmarajah, G., Walker, D.W., Pletcher, S.D., and Seroude, L. (2005). Differential patterns of apoptosis in response to aging in Drosophila. Proc. Natl. Acad. Sci. USA 102, 12083–12088. Zid, B.M., Rogers, A.N., Katewa, S.D., Vargas, M.A., Kolipinski, M.C., Lu, T.A., Benzer, S., and Kapahi, P. (2009). 4E-BP extends lifespan upon dietary restriction by enhancing mitochondrial activity in Drosophilia. Cell 139, 149–160.
Cell 143, 813–825, November 24, 2010 ª2010 Elsevier Inc. 825
Reelin and Stk25 Have Opposing Roles in Neuronal Polarization and Dendritic Golgi Deployment Tohru Matsuki,1 Russell T. Matthews,1 Jonathan A. Cooper,3 Marcel P. van der Brug,2,4 Mark R. Cookson,2 John A. Hardy,2,5 Eric C. Olson,1 and Brian W. Howell1,* 1Department
of Neuroscience and Physiology, SUNY Upstate Medical University, Syracuse, NY 13210, USA of Neurogenetics, National Institute on Aging, National Institutes of Health, Bethesda, MD 20892, USA 3Division of Basic Sciences, Fred Hutchinson Cancer Research Center, Seattle, WA 98109, USA 4Present address: Department of Neuroscience, The Scripps Research Institute, Jupiter, FL 33458, USA 5Present address: Department of Molecular Neuroscience and Reta Lila Weston Laboratories, University College, Queens Square House, London WC1 3BG, UK *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.029 2Laboratory
SUMMARY
The Reelin ligand regulates a Dab1-dependent signaling pathway required for brain lamination and normal dendritogenesis, but the specific mechanisms underlying these actions remain unclear. We find that Stk25, a modifier of Reelin-Dab1 signaling, regulates Golgi morphology and neuronal polarization as part of an LKB1-Stk25-Golgi matrix protein 130 (GM130) signaling pathway. Overexpression of Stk25 induces Golgi condensation and multiple axons, both of which are rescued by Reelin treatment. Reelin stimulation of cultured neurons induces the extension of the Golgi into dendrites, which is suppressed by Stk25 overexpression. In vivo, Reelin and Dab1 are required for the normal extension of the Golgi apparatus into the apical dendrites of hippocampal and neocortical pyramidal neurons. This demonstrates that the balance between ReelinDab1 signaling and LKB1-Stk25-GM130 regulates Golgi dispersion, axon specification, and dendrite growth and provides insights into the importance of the Golgi apparatus for cell polarization. INTRODUCTION The development of the exquisite morphology of neurons is a carefully orchestrated process that optimizes the ability of individual neurons to receive signals, integrate them, and transmit the output to target cells. Neuronal polarization, first observed as the rapid growth of a process that will ultimately become an axon, followed by the asymmetrical development of dendrites are key steps in morphological and functional maturation (Arimura and Kaibuchi, 2005). Interestingly, the Golgi apparatus has been implicated in these different aspects of neuronal polarity. In the nascent neuron, the position of the Golgi and 826 Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc.
the adjoined centrosome correlates with the site of axon emergence, which becomes the future basal side of a mature pyramidal neuron (de Anda et al., 2005, 2010; Zmuda and Rivas, 1998). Later, the Golgi apparatus is positioned on the apical side of pyramidal neurons, proximal to the major apical dendritic tree and opposite to the axon and minor basal dendrites (Horton et al., 2005). Dispersion of the Golgi apparatus away from the apical pole leads to a loss of dendrite asymmetry in these cells, with equal-sized apical and basal dendrites (Horton et al., 2005). Furthermore, specialized Golgi outposts, which populate dendrites, promote the elaboration of dendritic branches (Ye et al., 2007). However, it remains to be determined how Golgi positioning within neurons is regulated. Mutations in the genes encoding the Reelin-Dab1 signaling pathway lead to profound defects in neuronal positioning and dendritogenesis during brain development (Niu et al., 2004; Rice et al., 2001). The lamination of the cerebral cortex, hippocampus, and cerebellum is disorganized and appears approximately inverted compared to normal. Reelin is a secreted ligand that is produced in discreet layers in the developing brain (D’Arcangelo et al., 1995; Ogawa et al., 1995). Genetic and biochemical studies have shown that it regulates a signal transduction pathway requiring the ApoE receptors ApoER2 and VLDLR (D’Arcangelo et al., 1999; Hiesberger et al., 1999; Trommsdorff et al., 1999), the cytoplasmic adaptor protein Dab1 (Howell et al., 2000), and Src family kinases (Arnaud et al., 2003; Bock and Herz, 2003). Disparate functions have been proposed for Reelin-Dab1 signaling, though a clear biological response to clarify its role in brain development is lacking (Chai et al., 2009; Cooper, 2008; Fo¨rster et al., 2010; Sanada et al., 2004). The severity of dab1-dependent phenotypes depends on the genetic background (Brich et al., 2003). We have recently identified stk25 as a modifier of dab1 mutant phenotypes (unpublished data). Here we characterize the role of Stk25 (also YSK1, Sok1) in nervous system development. Previous work has implicated Stk25 in regulating Golgi morphology through the Golgi matrix protein GM130 (Preisinger et al., 2004), which we confirm here.
GM130 regulates the fusion of ER-to-Golgi vesicles with the Golgi cisternae and the fusion of Golgi cisternae into elongated ribbons (Barr and Short, 2003; Puthenveedu et al., 2006). Depletion or mitotic phosphorylation of GM130 leads to Golgi fragmentation and reduced efficiency of biosynthetic processing (Lowe et al., 1998; Marra et al., 2007; Puthenveedu et al., 2006). The protein kinase LKB1 and its associated factors STRAD and MO25 are known to be important for neuronal polarization, axon specification, and dendrite growth (Asada et al., 2007; Barnes et al., 2007; Shelly et al., 2007). In this study, we find that Stk25 is part of an LKB1 cell polarization pathway. Stk25, LKB1, and GM130 are shown to regulate Golgi morphology and axon initiation. In addition, we show that Stk25 and ReelinDab1 signaling have antagonistic effects on neuronal polarization and the morphology and subcellular distribution of the Golgi. As the position of the Golgi plays roles in cell polarization, process extension, and cell migration (Fidalgo et al., 2010; Horton et al., 2005; Yadav et al., 2009; Ye et al., 2007), this evidence is fundamental for understanding the molecular control of neuronal morphogenesis and provides new insights into the biological role of Reelin-Dab1 signaling. RESULTS Stk25 Regulates Neuronal Polarity Stk25 has previously been shown to regulate the polarized migration of epithelial cells. As other Ste20-like kinases have roles in neuronal polarization (Jacobs et al., 2007; Preisinger et al., 2004), we sought to assess a role for Stk25 in neuronal polarization by using hippocampal neuronal cultures (Dotti and Banker, 1987). These neurons have a stereotypic morphology and program of differentiation and respond to Reelin-Dab1 signaling (Matsuki et al., 2008). Soon after plating, they extend short uniform processes that have the potential to develop into either axons or dendrites (Arimura and Kaibuchi, 2007). By stage III, 48 to 72 hr later, one of the processes can be identified as an axon whereas the other processes differentiate into dendrites. We reduced Stk25 levels by infection with a lentivirus carrying GFP and Stk25 shRNA and identified axons 6 days later using SMI-312, a pan-axonal neurofilament marker. Depletion of Stk25 inhibited axon specification. At least 30% of the Stk25 shRNA lentivirus-infected, GFP-positive neurons lacked an axon (Figures 1B and 1F, lane 2), whereas axons were detected in all neurons infected with either empty vector (EV) or control shRNA vectors (Figures 1A and 1F, lanes 1 and 3 and insets). The longest process in Stk25 shRNA-expressing cells was also much shorter than the long axons of control cells (Figures 1A, 1B, and 1F, lane 2), consistent with a failure to induce an axon. To assess whether axon absence was specifically caused by reduced Stk25 expression, we tested for rescue by Stk25 overexpression. Both kinase-active and kinase-inactive versions of an shRNA-resistant Stk25 (Stk25*) were expressed as red fluorescent protein (RFP) fusion proteins in cultures that were also infected with the GFP-expressing, Stk25 shRNA virus (Figures S1A–S1D available online). Both kinase-active and kinase-inactive Stk25*-RFP rescued the axon-less phenotype caused by Stk25 knockdown (Figure 1F, lanes 7–9). This suggests that
the axon-less phenotype in Stk25 shRNA-expressing cells was the specific result of reducing Stk25 expression and that Stk25 kinase activity is not required for axon production. To investigate whether Stk25 affected axon initiation or maintenance, we examined stage III hippocampal neurons (Figures 1D and 1E). We found that 56% ± 5% of Stk25 knockdown neurons lacked an axon compared to only 7% ± 8% of control samples (Figure 1G). The longest neurite in Stk25 knockdown neurons was also significantly shorter than the incipient axon in control cultures. Moreover, overexpression of Stk25 induced multiple axons. Expression of either the wild-type or kinase-inactive Stk25*-RFP fusion proteins, or an Stk25-green fluorescent protein (GFP) fusion that has previously been shown to be biologically active (Preisinger et al., 2004), induced multiple SMIpositive axons in approximately 45%–50% of neurons as compared to 15% ± 3% in GFP-alone expressing controls (Figures 1C and 1F, lanes 5, 6, 8, and 9). Stk25 overexpression did not increase axon length (Figure 1F). Taken together, the results show that Stk25 regulates axon initiation but not axon growth in cultured neurons. Reelin-Dab1 Signaling Suppresses Multiple Axon Production Stk25 is expressed at relatively high levels in Reelin-Dab1 responsive cells in the developing cortical plate (Figure S1E) and in the adult hippocampus and cerebellar Purkinje cells (Figure S1F). Because we identified stk25 in a screen for modifiers of dab1 mutant phenotypes (unpublished data), we examined whether Reelin-Dab1 signaling might have an undiscovered role in axon initiation. Hippocampal neurons were cultured from dab1/ mutant embryos and infected with GFP-expressing lentiviruses to survey their morphology. Surprisingly, approximately 30% of the dab1/ mutant neurons produced multiple axons as compared to approximately 15% of the wild-type neurons (Figure 1H). To determine whether the multiple axon phenotype in dab1/ mutant neurons was sensitive to Stk25 expression level, we examined the effect of knocking down Stk25. Significantly fewer dab1/ mutant neurons infected with the Stk25 shRNA-expressing lentivirus produced multiple axons than the GFP-expressing control sample (Figure 1H). In addition, a significant number of the Stk25 shRNA-expressing neurons completely lacked axons. This shows that ReelinDab1 signaling regulates axon initiation and that the multiple axon phenotype in dab1/ mutant mice is dependent upon Stk25 expression. Congruent with this result, growth of neurons in the presence of Reelin suppressed the multiple axon phenotype caused by Stk25 overexpression (Figure 1I). This treatment did not, however, lead to the loss of axon production, which would be expected if Stk25 function was abolished. None of these treatments affected axon length. Therefore, Reelin-Dab1 signaling appears to counteract the effects of high Stk25 expression without completely blocking its function in axon induction. Stk25 Regulates Axon Formation and Dendrite Asymmetry In Vivo To investigate whether Stk25 regulates neuronal differentiation in vivo, we electroporated the Stk25 shRNA-expressing vector Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc. 827
Figure 1. Stk25 Expression Regulates Axon Differentiation in Culture (A) Primary hippocampal neurons (E17.5) infected with the GFP-expressing EV-control virus had typical pyramidal neuron morphologies, including a long SMI-positive axon (inset a) and shorter dendrites. (B) Neurons infected with the Stk25 shRNA virus had shorter processes and frequently lacked long (>250 mm) SMI-positive processes that met the criteria for axons (inset b). An SMI-positive process (arrowhead) from a noninfected neuron runs parallel to the GFP-positive process (arrow). (C) Cells overexpressing Stk25 wild-type (WT)GFP had multiple SMI-positive axons (insets c, c0 ). (D) At stage III (2DIV), EV-control infected neurons had one dominant SMI-positive axon. (E) In contrast, Stk25 shRNA-expressing neurons often lacked SMI-positive, axon-like processes. (F) The number of neurons with 0, 1, 2, or more axons and the length of the longest processes were determined for neurons infected with the indicated viruses. For rescue experiments, neurons were coinfected with the Stk25 shRNA (GFP-positive) and either RFP, Stk25* WT-RFP, or Stk25* K49R-RFP expressing viruses (lanes 7–9, Figure S1). (G) At stage III (2DIV), many Stk25 shRNAexpressing neurons lacked axons as compared to a small percentage of EV-control infected neurons. (H) The number of neurons with multiple axons was increased in dab1/ (lane 2) compared to wild-type neurons (lane 1, duplicated from F), and this was reduced by Stk25 shRNA expression (lanes 3). (I) Primary hippocampal neurons that were infected with either GFP- or Stk25 WT-GFP-expressing viruses were split into three groups and grown in either neurobasal (NB), control-conditioned (CCM), or Reelin-conditioned (RCM) media for 6 days. Statistical significance (*,**,***p < 0.0001, Student’s t test, compared between the sample pairs: (F) 1:2; 4:5,6,7; 7:8,9; (G) 1:2; (H) 1:2, 2:3; n > 60; (I) 5:6; n indicated in bars). Bars: (C) 50 mm; (a) 10 mm; (c0 ) 5 mm; and (E) 20 mm. See also Figure S1.
into the hippocampi of fetal mice. The brains of these mice were analyzed for GFP expression and neuronal polarization of Ctip2positive, pyramidal neurons in the CA1 region of the hippocampus at postnatal day 7 (P7). Stk25 shRNA did not interfere with the positioning of neurons, but their apical dendrites were 828 Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc.
significantly longer (Figures 2A, 2B, and 2E). In addition, approximately 40% of the strongly GFP-positive, Stk25 shRNAexpressing neurons lacked identifiable axon initial segments, detected using anti-phospho-IkBa antibodies, suggesting that axons were either absent or failed to mature normally (Figures 2D and 2F; Movie S1). By comparison, all of the GFP-positive, EV-control electroporated neurons examined had axon initial segments (Figures 2C and 2F; Movie S1). This suggests that Stk25 regulates axon specification and dendrite growth in hippocampal pyramidal neurons in vivo.
Figure 2. Stk25 Regulates Neuronal Polarity during Brain Development (A) EV-control vector (GFP-positive, green) electroporated at E16.5 in utero was expressed in Ctip2-positive (red), hippocampal-pyramidal neurons at P7. (B) Stk25 shRNA-expressing neurons (GFP-positive) were appropriately positioned in the CA1 layer, and their apical dendrites extended further than EV-control. (C) GFP-expressing, EV-control transfected CA1 neurons had the typical pyramidal shape and phospho-IkBa- (red), GFP-positive (green) axon initial segments (Sanchez-Ponce et al., 2008) (Movie S1). (D) In contrast, a high percentage of strongly GFP-positive, Stk25 shRNA-expressing neurons were often misshapen and lacked axon initial segments (Movie S1). (E) Quantification of apical dendrite length in EV-control and Stk25 shRNA hippocampi. (F) Quantification of the number of GFP-, Ctip2-positive pyramidal neurons that had axon initial segments (n indicated in bar.) (G) In EV-control neurons, the Golgi apparatus (trace of GRASP65 signal) is concentrated on the apical side of the neuron (Movie S2). (H) In Stk25 shRNA-expressing neurons, the Golgi apparatus is broadly distributed throughout the neuron (Movie S2). (I) Scheme used to determine Golgi distribution in (J). (J) The Golgi distribution in apical, lateral (combined), or basal quadrants was quantified. (K) The diameters of the largest apical and basal processes were determined (*p < 0.0005, Student’s t test, n R 12, neurons from three animals). Bars: (B) 200 mm; (D and H) 10 mm. Error bars indicate standard error of the mean (SEM) in all figures.
In addition to having longer apical dendrites, the basal dendrites of Stk25 shRNA-expressing neurons were also atypical. Normal pyramidal neurons have long, thick apical dendrites and much thinner and shorter basal dendrites (Horton et al., 2005; Figures 2G and 2K; Movie S2). The apical dendrites of Stk25 shRNA-expressing neurons had normal thickness, but the basal dendrites were thicker than normal (Figures 2H and 2K; Movie S2). We were not able to measure the length of the basal dendrites. Therefore, there is evidence for growth of both apical and basal dendrites, and this reduced the distinction between apical and basal dendrites in terms of thickness. This suggests that Stk25 is needed for normal axon production and dendrite asymmetry in vivo.
Stk25 Interacts with STRADa and Acts on the LKB1 Signaling Pathway The functions of Stk25 resemble those reported for LKB1STRAD signaling (Barnes et al., 2007; Kishi et al., 2005; Shelly et al., 2007). This pathway has a prominent role in cell polarity control across numerous cell types from Caenorhabditis elegans to man. LKB1 is partially regulated by binding STRAD, which both shuttles it from the nucleus to the cytoplasm and stabilizes it. We therefore investigated whether Stk25 associates with the LKB1-STRAD signaling complex. By immunoprecipitating tagged fusion proteins coexpressed in HEK293T cells, we found that both wild-type and kinase-inactive HA-Stk25 coimmunoprecipitated with myc-STRADa (Figure S2A). Identifying Stk25 Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc. 829
Figure 3. Stk25-RFP Overexpression Rescues the Neuronal Polarization Defect Caused by LKB1 but Not by GM130 Knockdown (A) Expression of LKB1 shRNA (GFP-positive, green) in hippocampal neurons led to an increase in the number of neurons that lack an axon at 6DIV in cells also expressing RFP (red). (a) Longest process lacks SMI immunoreactivity. (B) In contrast, overexpressing Stk25* WT-RFP in LKB1 knockdown neurons rescued axon production. (b) Long, axon-like process is SMI positive. (C) GM130 knockdown (GFP-positive) also caused a reduction in axon production in RFP-positive cells. (c) No SMI imunoreactivity was detected in processes of the GFP-, RFP-positive neuron. (D) Stk25* WT-RFP expression did not rescue axonogenesis in GM130 knockdown neurons. (d) Longest process is SMI negative. (E) Axon number and the length of the longest processes were quantified for the indicated treatment groups. (Lane 1 was duplicated from Figure 1F lane 1.) (*p < 0.005 compared to lane 1, **p = 0.01 compared to lane 2, Student’s t test.) Bars: (D) 50 mm; (d) 5 mm. See also Figure S2.
as a direct or indirect STRAD-binding protein suggests a potential role for Stk25 on the LKB1 pathway. To investigate whether Stk25 is important for LKB1 function, we took two approaches. We examined whether (1) Stk25 is required for LKB1-STRAD-regulated epithelial cell polarization and (2) Stk25 overexpression rescues the LKB1 knockdown phenotype in neurons. We first tested whether reduced Stk25 expression would inhibit the LKB1-STRAD-dependent polarization of W4 intestinal epithelial cells. These cells have been engineered to constitutively express LKB1 and express STRAD in response to doxycyline, which leads to their polarization (Baas et al., 2004). Most W4 cells infected with EV and control shRNA lentiviruses became polarized within 24 hr of doxycycline treatment (Figures S2C and S2E). In contrast, only 20% of cells infected by the humanized (h) Stk25 shRNA lentivirus were polarized by doxycycline treatment (Figures S2C and S2E). Furthermore, expression of either wild-type or kinase-inactive Stk25*-RFP rescued STRAD-induced polarization in Stk25 shRNA-expressing W4 epithelial cells (Figure S2F). Collectively, 830 Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc.
these experiments show that the Stk25 protein, not its kinase activity, is required for LKB1-STRAD-regulated epithelial cell polarization. We then confirmed that LKB1 knockdown leads to a loss of axon initiation in cultured hippocampal neurons (Figure 3A; Barnes et al., 2007; Shelly et al., 2007). We tested whether Stk25 can rescue or bypass the LKB1 requirement by overexpressing Stk25* wild-type (WT)-RFP in LKB1 shRNA-expressing neurons (Figure 3B). Ninety-two percent of LKB1 knockdown neurons that expressed Stk25* WT-RFP produced at least one axon compared to only 48% of RFP-, LKB1 shRNA-coexpressing neurons (Figure 3E). These results are consistent with a role of Stk25 on the LKB1 pathway to regulate axon induction. GM130 Interacts with Stk25 and Regulates Axon Induction The Golgi matrix protein GM130, which has critical roles in regulating Golgi dynamics, was identified in a yeast two-hybrid screen as an Stk25 binding partner (Preisinger et al., 2004). We confirmed this interaction by coimmunoprecipitating tagged
Figure 4. Golgi Apparatus Morphology Is Regulated by Stk25, LKB1, and GM130 Expression and Reelin Signaling (A) Stage III neurons that were infected with the EV-control virus had typical cis-Golgi ribbons (GRASP65, Movie S3). In contrast, the cis-Golgi in Stk25 shRNA-, LKB1 shRNA-, or GM130 shRNA-expressing neurons was fragmented (Movie S3). GFP signal was omitted for clarity. (B) Significantly more Stk25 knockdown neurons had fragmented Golgi complexes compared to the EV-control and the control shRNA (n, as indicated). LKB1 and GM130 knockdown also caused significant Golgi fragmentation as compared to EV-control infected neurons. Stk25*RFP expression rescued Golgi fragmentation in LKB1 shRNA but not GM130 shRNA-expressing neurons. (C) Neurons overexpressing either Stk25 WT-GFP or Stk25 K49R-GFP had condensed cis-Golgi (GRASP65 signal) compared to EV-controls when grown in either neurobasal or control-CM. Growth in Reelin-CM partially rescued the Golgi appearance in Stk25-overexpressing cells. GM130 and GRASP65 colocalized under all conditions (not shown). (D) Golgi volume (upper panel) and the length of the longest Golgi ribbon (lower panel) were determined (*p < 0.0001, Student’s t test, n indicated in bars). Bars: 5 mm. See also Figure S3.
fusions of GM130 and Stk25 (Figure S2B). Interestingly, kinaseinactive Stk25 consistently immunoprecipitated with GM130 more efficiently than wild-type, suggesting that Stk25-dependent phosphorylation may destabilize the complex. Stk25 colocalizes with GM130 at the Golgi apparatus of HeLa cells (Preisinger et al., 2004). To determine whether Stk25 localizes to the Golgi complex in neurons, we raised an antibody to a region of Stk25 that is divergent from the close relatives Mst3 and Mst4 (Extended Experimental Procedures). Endogenous Stk25 expression overlapped with the GM130-positive cis-Golgi in neurons at stage III, coincident with axon specification (Figure S2D). To asses whether GM130 plays a role in neuronal differentiation, we examined GM130 shRNA-expressing neurons for defects in polarity. Similar to Stk25 and LKB1 knockdown neurons, knockdown of GM130 reduced axon number at 6DIV (Figure 3C). GM130 knockdown also caused a significant reduction in axon initiation in stage III (2DIV) neurons (data not shown). Stk25*-RFP overexpression in GM130-deficient cells did not rescue axon number at 6DIV (Figure 3D), which suggests that GM130 is required for neuronal polarization downstream of Stk25.
Stk25, GM130, and LKB1 Regulate Golgi Distribution Previously it was shown that GM130 regulates Golgi morphology in HeLa cells (Puthenveedu et al., 2006). Given that Stk25, LKB1, and GM130 regulate axon initiation, and the position of the Golgi apparatus early in differentiation normally coincides with axonal localization (de Anda et al., 2005, 2010), we examined whether Stk25, LKB1, and GM130 regulate Golgi morphology (Figure 4). Individually knocking down Stk25, LKB1, and GM130 in stage III primary hippocampal neurons resulted in dispersion of Golgi elements in a high percentage of cells, in contrast to the typical elongated morphology observed in the EV-control neurons (Figures 4A and 4B; Movie S3). Interestingly, the Golgi fragmentation caused by LKB1 knockdown was rescued by Stk25*-RFP overexpression (Figure 4B), suggesting that Stk25 overexpression can compensate for reductions in LKB1 signaling. In contrast, Golgi fragmentation in GM130 shRNA-expressing cells was not rescued by Stk25 overexpression (Figure 4B). Overexpression of either Stk25 WT-GFP or Stk25 K49R-GFP led to the condensation of the Golgi into a smaller volume (Figure 4C, neurobasal). Therefore, increasing or decreasing Stk25 expression from endogenous levels has different consequences for Golgi morphology, in addition to having the opposite effects on axon production. These results suggest an LKB1-Stk25-GM130 pathway for Golgi regulation in cultured neurons. Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc. 831
Importantly, Stk25 knockdown in hippocampal pyramidal neurons also caused Golgi fragmentation in vivo, as determined by use of in utero electroporation. Normally, the Golgi is strictly localized to the apical side of the soma and forms outposts in the apical dendrite (Horton et al., 2005; Figures 2G and 2J; Movie S2). However, in Stk25 shRNA-expressing, Ctip2-positive neurons, the Golgi apparatus was often broadly distributed throughout the soma (Figures 2H and 2J; Movie S2). In summary, these results indicate that Stk25, LKB1, and GM130 are required for normal Golgi morphology in neurons at a time when axons are first appearing. Furthermore, the fragmented Golgi phenotype correlated with the loss of axon production in neurons, and both phenotypes were rescued by Stk25 overexpression in LKB1 knockdown cells. Reelin Signaling Regulates Golgi Morphology As Stk25 and Reelin have opposing effects on axon initiation (Figure 1H) and Stk25 affects Golgi morphology (Figures 4A and 4B), we investigated the role of Reelin in regulating Golgi morphology. First we examined the appearance of the Golgi apparatus in hippocampal and neocortical pyramidal neurons of reelin/ and dab1/ mutant mice. In the pyramidal layer of the wildtype CA1 zone and in developing neocortical layers, the Golgi apparati were linearly organized and extended tens of microns into the apical processes (Figure 5D; Figures S4D and S4G, insets). The Golgi of the reelin/ and dab1/ mutants often appear convoluted near the nucleus rather than extended into a dendrite (Figures 5E and 5F; Figures S4E and S4F, insets). The distance from the Ctip2-positive nucleus to the tip of the Golgi ribbon was significantly decreased in reelin/ and dab1/ mutants as compared to wild-type (Figure 5G and Figure S4G), indicating that the reelin and dab1 genes either directly or indirectly regulate Golgi extension into the apical process of pyramidal neurons. As reelin and dab1 also regulate the proper layering of hippocampal pyramidal neurons (Caviness and Sidman, 1973; Goffinet, 1984; Rice et al., 2001) (Figures 5B and 5C), the effects of reelin and dab1 on Golgi deployment may be indirect. Therefore, we tested whether Reelin-Dab1 signaling acutely induces changes in Golgi morphology or localization by treating hippocampal neuron cultures with Reelin for 30 min. Hippocampal pyramidal neurons were infected with a low titer GFP-expressing lentivirus to help visualize individual neurons. The Golgi was largely localized close to the nucleus in control-conditioned media (CM) and neurobasal-treated Ctip2-positive pyramidal neurons (Figures 6A and 6C). However, in approximately 80% ± 5% of Reelin-CM-treated neurons, the Golgi apparati extended into the largest dendritic process (Figures 6A and 6C). The distance between the nucleus and the most distal portion of the Golgi ribbon from randomly selected Ctip2-positive neurons was significantly larger in the Reelin-CM-treated samples compared to the control-CM- and neurobasal-treated samples (Figure 6B). The Golgi apparatus is therefore rapidly deployed into dendrites in response to Reelin stimulation. We next evaluated whether the Golgi response to Reelin was sensitive to elevated Stk25 expression levels. Hippocampal neurons were infected with Stk25 WT-GFP or Stk25 K49R-GFP 832 Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc.
Figure 5. The Golgi Apparatus Extends into an Apical Process in Neonatal Hippocampus in a reelin- and dab1-Dependent Manner (A) Ctip2-positive CA1 neurons are organized into a tight lamella in wild-type brain. (B) Homozygous disruption of reelin or (C) dab1 causes dispersion of these neurons. (D) Confocal imaging through the CA1 region of the wild-type hippocampus revealed that the Golgi apparatus (white or green, inset) extends radially into the presumptive apical dendrite of Ctip2-positive neurons (red, inset). (E) In equivalent reelin/ or (F) dab1/ mutant sections, the Golgi is more often convoluted proximal to the nucleus (inset). Insets were selected from regions where isolated cells could be distinguished. (G) The Golgi phenotype was quantified by measuring the distance from the nucleus to the furthest tip of the Golgi ribbon. (*p < 0.0001, Student’s t test, n indicated in bar from three animals per group.) Bar: 200 mm in (C), 20 mm in (F), and 2 mm in inset. See also Figure S4.
expressing viruses after 72 hr in culture and treated analogously to experiments described above. Expression of either Stk25 WT-GFP and Stk25 K49R-GFP reduced but did not eliminate the Golgi extension in response to Reelin (Figures 6B and 6C). Under these conditions, linear Golgi ribbons were observed extending into the dendrites, but on average this was approximately 50% the distance observed in the Reelin-treated, GFPexpressing cells (Figure 6B). Furthermore, Reelin signaling suppressed Golgi compaction induced by Stk25 overexpression (Figures 4C and 4D). In cultures that were grown in Reelin-CM for 2 days (Figure 4), we did not observe Golgi deployment into dendrites. This is not surprising as components of the ReelinDab1 pathway begin to be degraded within a few hours. In 60-day-old animals, Golgi extension into dendrites was also reduced (data not shown). Therefore, Golgi deployment appears
Figure 6. Reelin Stimulation Leads to Rapid Golgi Extension into Dendrites Primary hippocampal neurons were infected with GFP-expressing viruses after 3DIV and stimulated 3 days later. (A) The Golgi apparati in Reelin-CM-treated neurons extended tens of microns into dendrites, compared to little or no extension into dendrites of control-CM or neurobasal-treated neurons. (B) The distance between the nucleus and the tip of the Golgi was measured for GFP-, Ctip2-positive neurons. Expression of Stk25 WT-GFP and Stk25 K49R-GFP caused a significant reduction in Reelin-induced Golgi extension. (C) The Golgi of most GFP-, Ctip2-positive Reelin-CM-treated neurons extended at least 10 mm from the nucleus into or toward a dendrite. Significantly fewer Golgi were observed in the processes of control-treated samples or Reelin-CM-treated samples that also overexpressed Stk25. Yellow arrows indicate furthest tip of Golgi ribbon from nucleus. (*p < 0.0001, **p = 0.0002, ***p < 0.05, Student’s t test, between Reelin-CM- and control-treated samples and between GFP- and Stk25-expressing samples treated with Reelin-CM.) Bars: 10 mm.
to be a transient, developmental phenomenon. Thus, similar to the manifestation of the multiple axon phenotype caused by Stk25 overexpression or loss of dab1 gene function, the degree of Golgi extension seems to be determined by a competition between Reelin-Dab1 signaling and Stk25 levels. DISCUSSION In this study, we find that Reelin-Dab1 signaling acts in an opposing manner to LKB1, GM130, and Stk25 to regulate the polarization of axons, dendrites, and Golgi apparati of hippocampal neurons, as shown in Figure 7. Knocking down these three proteins led to Golgi fragmentation and inhibited axon initiation (Figure 1, Figure 3, and Figure 4). In contrast, Stk25 overexpression caused Golgi condensation and the formation of multiple axons (Figure 1 and Figure 4). It also rescued axon production and Golgi fragmentation caused by LKB1 knockdown but did not rescue either phenotype caused by reduced GM130 expression (Figure 3 and Figure 4), suggesting that Stk25 functions as an intermediary between LKB1 and GM130. Stk25 directly or indirectly binds to the LKB1-STRAD complex and GM130 and may play a scaffolding role to link LKB1 signaling to GM130 and Golgi regulation (Figure S2). Reelin-Dab1 signaling antagonizes the effects of Stk25 overexpression on Golgi morphology and neuronal polarization as well as inducing polarized deployment of the Golgi into the apical dendrite (Figure 1,
Figure 4, and Figure 6). Together this implicates the LKB1 pathway, GM130, Stk25, and Reelin-Dab1 signaling in Golgi regulation during neuronal polarization. Involvement of the Golgi Apparatus in Neuronal Polarization The Golgi apparatus and centrosomes reorient as neurons migrate into the cortical plate (de Anda et al., 2010; Nichols and Olson, 2010). At the time of axon initiation, the centrosome is near the basal pole (rear) of the cell. It then moves to the opposite pole (front) and is important for extending an apical process that is used for radial migration (de Anda et al., 2010). The apical process subsequently transforms into the apical dendritic tree, with the Golgi and centrosomes at its base (Barnes et al., 2008; Horton et al., 2005). The same events presumably occur during migration of hippocampal pyramidal neurons in vivo. When hippocampal neurons are cultured, the centrosome position determines which neurite becomes an axon (de Anda et al., 2005). Later, the apical localization of the Golgi apparatus promotes the asymmetric growth of the apical compared to the basal dendrites (Horton et al., 2005). Consistent with this, Stk25 knockdown led to Golgi disorganization, inhibited axon induction, and lessened the asymmetry between the long, thick apical dendrite and short, slender basal dendrites (Figures 2F, 2H, 2J, and 2K). The Golgi may influence axon initiation through nucleating microtubules, regulating secretory trafficking, or interacting Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc. 833
Figure 7. Model of Stk25 as a Scaffolding Protein Acting Competitively with Reelin-Dab1 Signaling LKB1 is known to act in complex with STRAD to regulate cellular polarity (Alessi et al., 2006). Reelin, the receptors ApoER2 and VLDLR, and Dab1 also form a signaling complex (Hiesberger et al., 1999; Trommsdorff et al., 1998). STK25 coimmunoprecipitates with STRAD and GM130 (Figure 2S). Overexpression of LKB1 and STRAD is known to induce the formation of multiple axons (Barnes et al., 2007; Shelly et al., 2007). Independent of its kinase activity, STK25 does so also and induces Golgi condensation (Figure 1F and Figure 4A). Knocking down LKB1, Stk25, or GM130 causes Golgi fragmentation/dispersion and lost axon production, the opposite to Golgi condensation and multiple axon formation (Figure 1, Figure 3, and Figure 4) (Barnes et al., 2007; Shelly et al., 2007). The overexpression phenotypes are suppressed by Reelin stimulation. Dab1/ neurons (Reelin signaling deficient) have multiple axons and shorter dendrites (Figure 1F) (Niu et al., 2004). Reelin stimulation induces Golgi deployment and dendrite growth, phenotypes suppressed by Stk25 expression/overexpression (Figure 2 and Figure 6).
with the centrosome (Efimov et al., 2007; Pfenninger, 2009; Rosso et al., 2004; Su¨tterlin and Colanzi, 2010). It seems less likely that the Golgi is required to supply materials to sustain axon growth, as none of our manipulations affected axon length, only axon number. Therefore, the Golgi probably has a signaling or microtubule nucleation role in axon specification. Indeed, microtubule stabilization has been shown to enhance axon formation (Witte et al., 2008), and inhibiting post-Golgi trafficking disrupts axo-dendritic polarization (Bisbal et al., 2008; Yin et al., 2008). In dendrites, however, the Golgi may have a role in supplying materials for dendrite growth, as we detected effects on dendrite thickness and length (Figures 2E and 2K). Deployment of the Golgi into the apical dendrite may initiate the formation of dendritic Golgi outposts, which have been shown to promote dendrite growth and branching (Horton et al., 2005; Ye et al., 2007). We found that Stk25 functions in Golgi morphology and axon specification as part of an LKB1 pathway (Figure 3 and Figure 4). LKB1, the mammalian Par-4 homolog, is an evolutionarily conserved cell polarity protein that is known to regulate axodendritic polarity in neurons (Barnes et al., 2008). LKB1 is activated upon binding STRAD and MO25 (Alessi et al., 2006). STRAD stabilized LKB1 in processes prior to axon production and in the nascent axon, suggesting a role in axon specification (Shelly et al., 2007). As a master kinase, LKB1 activates several downstream kinases that regulate various aspects of cell polarity. These include the Sad A and Sad B kinases, which are required for neuronal polarization (Barnes et al., 2007; Kishi et al., 2005). Mst4, another downstream kinase, is closely related to Stk25. Like Stk25, it binds to GM130 and is enriched in the Golgi apparatus (Preisinger et al., 2004). Both Mst4 and Stk25 are required downstream of LKB1-STRAD induction for polarized brush border formation in epithelial cells (ten Klooster 834 Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc.
et al., 2009; Figure S2). However, although Mst4 kinase activity is required during this process, the kinase activity of Stk25 is not needed to induce polarized brush border formation, regulate Golgi morphogenesis, or polarize hippocampal neurons (Figure 1F and Figures 4C and 4D). This suggests a kinase-independent scaffolding function for Stk25 (Figure 7), which is reminiscent of the pseudokinase STRAD (Lizcano et al., 2004). GM130 appears to be necessary for Stk25 effects on Golgi and neuronal polarization; however, it may not be sufficient. By linking LKB1 signaling to GM130, Stk25 may directly regulate GM130 or indirectly modulate the activity of other Golgi proteins. Reelin-Dab1 Signaling Regulates Neuronal Polarization and Golgi Deployment Our work also shows that Reelin-Dab1 signaling, acting in opposition to LKB1-Stk25-GM130, affects Golgi morphology and axon formation. The absence of Reelin or Dab1 inhibited Golgi deployment into the apical dendrite in vivo (Figure 5 and Figure S4), and long-term growth in Reelin opposed Golgi condensation induced by Stk25 overexpression in vitro (Figure 4). Similarly, Dab1 absence induced supernumerary axons in vitro (Figure 1H), the opposite effect to depleting Stk25. However, Reelin-Dab1 and LKB1-Stk25-GM130 do not fit into a simple epistatic relationship. For example, Stk25 depletion reduces axon number even when Dab1 is absent, suggesting that Stk25 does not require Dab1 to regulate axon number (Figure 1). This indicates that LKB1-Stk25-GM130 and Reelin-Dab1 act on the Golgi and axon initiation through different pathways, and the balance between the two pathways determines the outcome. In this respect, Golgi distribution is a quantitative trait, not all or none, and may be influenced by other factors. Indeed, extended Golgi were observed in a subset of neurons in reelin/ and dab1/ mutant brains (Figure 5 and Figure S4). One possibility is that Reelin-Dab1 and LKB1-Stk25-GM130 regulate different aspects of Golgi morphology through different mechanisms. For example, Reelin-Dab1 may regulate ER-Golgi vesicle movement, and LKB1-Stk25-GM130 may affect vesicle fusion. In sum, we have characterized Stk25, a modifier of the ReelinDab1 pathway, and shown that it acts on the LKB1-STRAD pathway to regulate Golgi morphology and neuronal polarization. Stk25 may play a scaffolding role to link LKB1-STRAD to Golgi regulation through binding GM130, as the kinase activity was shown to be dispensable for neuronal polarization and Golgi morphogenesis. We find that Reelin-Dab1 signaling regulates Golgi morphology and deployment into dendrites in a competitive manner with Stk25. Golgi position has been shown to enhance local secretory trafficking (Horton et al., 2005; Ye et al., 2007); thus, this competition may regulate membrane and protein cargo flow into proximal dendrites. Our findings provide new insights into the regulation of morphogenic changes in neurons that drive neuronal polarization and brain lamination. EXPERIMENTAL PROCEDURES Expression Vectors The lentiviral vectors used in this study were based on pLentiLox 3.7 (pLL3.7) vectors (Rubinson et al., 2003) with the following substitutions: (1) for shRNA experiments, instead of the CMV promoter, the CMV enhancer/chicken b-actin
promoter (Niwa et al., 1991) directs GFP expression; (2) for fusion protein experiments, instead of the U6 promoter the CMV enhancer/chicken b-actin promoter directs expression. The shRNA constructs include Stk25 shRNA AG GAGCTCCTGAAGCACAAAT and control shRNA AGTAGCTCCTAAAGCACA CAT. The lentivirus production was as previously described (Matsuki et al., 2008). The knockdown viruses were confirmed to reduce expression of either Stk25, LKB1, or GM130 (Figure S1 and Figure S3). The Stk25 K49R mutant has previously been reported to be kinase inactive, which we confirmed (Preisinger et al., 2004 and data not shown). Animals All animals were used in accordance with protocols approved by the Animal Care and Use Committees of SUNY Upstate Medical University, National Institutes of Neurological Disorders and Stroke, and the Fred Hutchinson Cancer Research Center, following NIH guidelines. Time pregnant mice (C57BL/6 for in vitro experiments and Swiss Webster for in utero electroporations) and rats (Sprague Dawley) were purchased from Charles River Laboratories and Taconic. The dab1/ (Howell et al., 1997) and reelin/ (Jackson Labs) mice were on the C57BL/6 strain. Immunocytochemistry Immunocytochemistry was done according to published methods (Matsuki et al., 2008) and is detailed in the Extended Experimental Procedures along with a list of the antibodies used. To measure Golgi volumes and length of the longest Golgi ribbon, we immunostained the neurons with anti-GRASP65, anti-GFP, and anti-Ctip2, which recognizes a CA1 and layer V pyramidal neuron-specific transcription factor. The area of the Golgi apparatus was calculated for each Z-plain (Image Examiner, Zeiss), multiplied by the thickness of the section, and summed to determine the volume. Cell Culture Hippocampal neuronal cultures were isolated from embryonic day (E) 17.5 mice or E18.5 rats and grown in neurobasal samples supplemented with 2% B27 (Invitrogen, Matsuki et al., 2008). For polarity studies, neurons (1 3 104 cells per cm2) were infected with the respective viruses on the day of culturing and replated 2 days later on poly-L-lysine coated coverslips placed over a monolayer of astrocytes. Axons were quantified at 2 days in vitro (DIV) or 6DIV as indicated, following standard criteria (Shelly et al., 2007). For Golgi deployment assays, rat cultured neurons (3 3 105 cells per cm2) were infected with low titer virus on day 3 and treated and fixed on day 6 in culture. Similar results were obtained with mouse neurons (data not shown). The controland Reelin-conditioned media were collected and concentrated as previously described (Matsuki et al., 2008). Analysis of In Utero Electroporated Brains To knock down Stk25 expression, DNA was injected into the lateral ventricle of E17.5 embryos of Swiss Webster mice in utero and electroporated (70 mV) as previously described (Olson et al., 2006) with the electrode paddles oriented to direct the DNA into the hippocampus. Perfused brains were processed for analysis on P7. Floating sections (70–100 mm) were immunostained with antibodies described in the figure legends. Confocal images were collected with overlapping optical sections through 30 mm, which were flattened for display. We assessed whether axon initial segments or Golgi elements belonged to a particular GFP-positive neuron (Figure 2), by examining movies of either 3D-rendered images or Z sections (Movie S1 and Movie S2). Golgi areas (Figures 2G and 2H) were produced by thresholding (Adobe Photoshop) flattened, 2D-negative images to match the GRASP65 signal channel in the original and discarding the signal extraneous to the GFP-positive cells (Movie S2). Process diameters were measured 12 mm from the nucleus (Figure 2K). These measurements were done using Image Examiner (Zeiss). Measurement of dendrite lengths was done using the softWoRx (AppliedPrecision).
ACKNOWLEDGMENTS We would like to thank Zainab Mansaray and Kristin Giamanco for experimental assistance, Michael Zuber for comments on the manuscript, Hans Clevers for cell lines, Louis Cantley and Jun-ichi Miyazaki for DNA vectors, Arvydas Matiukas and Melissa Pepling for assistance with confocal microscopy, and Bonnie Lee Howell for editing. This work was supported by funds from the NINDS intramural program and SUNY Upstate Medical University to B.W.H.; NIH grants NS066071 to E.C.O., NS069660 to R.T.M., and CA41072 to J.A.C.; and NIA intramural funds for M.R.C. Received: May 3, 2010 Revised: August 27, 2010 Accepted: October 20, 2010 Published: November 24, 2010 REFERENCES Alessi, D.R., Sakamoto, K., and Bayascas, J.R. (2006). LKB1-dependent signaling pathways. Annu. Rev. Biochem. 75, 137–163. Arimura, N., and Kaibuchi, K. (2005). Key regulators in neuronal polarity. Neuron 48, 881–884. Arimura, N., and Kaibuchi, K. (2007). Neuronal polarity: from extracellular signals to intracellular mechanisms. Nat. Rev. Neurosci. 8, 194–205. Arnaud, L., Ballif, B.A., Fo¨rster, E., and Cooper, J.A. (2003). Fyn tyrosine kinase is a critical regulator of disabled-1 during brain development. Curr. Biol. 13, 9–17. Asada, N., Sanada, K., and Fukada, Y. (2007). LKB1 regulates neuronal migration and neuronal differentiation in the developing neocortex through centrosomal positioning. J. Neurosci. 27, 11769–11775. Baas, A.F., Kuipers, J., van der Wel, N.N., Batlle, E., Koerten, H.K., Peters, P.J., and Clevers, H.C. (2004). Complete polarization of single intestinal epithelial cells upon activation of LKB1 by STRAD. Cell 116, 457–466. Barnes, A.P., Lilley, B.N., Pan, Y.A., Plummer, L.J., Powell, A.W., Raines, A.N., Sanes, J.R., and Polleux, F. (2007). LKB1 and SAD kinases define a pathway required for the polarization of cortical neurons. Cell 129, 549–563. Barnes, A.P., Solecki, D., and Polleux, F. (2008). New insights into the molecular mechanisms specifying neuronal polarity in vivo. Curr. Opin. Neurobiol. 18, 44–52. Barr, F.A., and Short, B. (2003). Golgins in the structure and dynamics of the Golgi apparatus. Curr. Opin. Cell Biol. 15, 405–413. Bisbal, M., Conde, C., Donoso, M., Bollati, F., Sesma, J., Quiroga, S., Dı´az An˜el, A., Malhotra, V., Marzolo, M.P., and Ca´ceres, A. (2008). Protein kinase d regulates trafficking of dendritic membrane proteins in developing neurons. J. Neurosci. 28, 9297–9308. Bock, H.H., and Herz, J. (2003). Reelin activates SRC family tyrosine kinases in neurons. Curr. Biol. 13, 18–26. Brich, J., Shie, F.S., Howell, B.W., Li, R., Tus, K., Wakeland, E.K., Jin, L.W., Mumby, M., Churchill, G., Herz, J., and Cooper, J.A. (2003). Genetic modulation of tau phosphorylation in the mouse. J. Neurosci. 23, 187–192. Caviness, V.S.J., Jr., and Sidman, R.L. (1973). Retrohippocampal, hippocampal and related structures of the forebrain in the reeler mutant mouse. J. Comp. Neurol. 147, 235–254. Chai, X., Fo¨rster, E., Zhao, S., Bock, H.H., and Frotscher, M. (2009). Reelin stabilizes the actin cytoskeleton of neuronal processes by inducing n-cofilin phosphorylation at serine3. J. Neurosci. 29, 288–299. Cooper, J.A. (2008). A mechanism for inside-out lamination in the neocortex. Trends Neurosci. 31, 113–119.
SUPPLEMENTAL INFORMATION
D’Arcangelo, G., Miao, G.G., Chen, S.C., Soares, H.D., Morgan, J.I., and Curran, T. (1995). A protein related to extracellular matrix proteins deleted in the mouse mutant reeler. Nature 374, 719–723.
Supplemental Information includes Extended Experimental Procedures, four figures, and three movies and can be found with this article online at doi: 10.1016/j.cell.2010.10.029.
D’Arcangelo, G., Homayouni, R., Keshvara, L., Rice, D.S., Sheldon, M., and Curran, T. (1999). Reelin is a ligand for lipoprotein receptors. Neuron 24, 471–479.
Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc. 835
de Anda, F.C., Pollarolo, G., Da Silva, J.S., Camoletto, P.G., Feiguin, F., and Dotti, C.G. (2005). Centrosome localization determines neuronal polarity. Nature 436, 704–708. de Anda, F.C., Meletis, K., Ge, X., Rei, D., and Tsai, L.H. (2010). Centrosome motility is essential for initial axon formation in the neocortex. J. Neurosci. 30, 10391–10406. Dotti, C.G., and Banker, G.A. (1987). Experimentally induced alteration in the polarity of developing neurons. Nature 330, 254–256. Efimov, A., Kharitonov, A., Efimova, N., Loncarek, J., Miller, P.M., Andreyeva, N., Gleeson, P., Galjart, N., Maia, A.R., McLeod, I.X., et al. (2007). Asymmetric CLASP-dependent nucleation of noncentrosomal microtubules at the transGolgi network. Dev. Cell 12, 917–930. Fidalgo, M., Fraile, M., Pires, A., Force, T., Pombo, C., and Zalvide, J. (2010). CCM3/PDCD10 stabilizes GCKIII proteins to promote Golgi assembly and cell orientation. J. Cell Sci. 123, 1274–1284. Fo¨rster, E., Bock, H.H., Herz, J., Chai, X., Frotscher, M., and Zhao, S. (2010). Emerging topics in Reelin function. Eur. J. Neurosci. 31, 1511–1518. Goffinet, A.M. (1984). Events governing organization of postmigratory neurons: studies on brain development in normal and reeler mice. Brain Res. 319, 261–296. Hiesberger, T., Trommsdorff, M., Howell, B.W., Goffinet, A., Mumby, M.C., Cooper, J.A., and Herz, J. (1999). Direct binding of Reelin to VLDL receptor and ApoE receptor 2 induces tyrosine phosphorylation of disabled-1 and modulates tau phosphorylation. Neuron 24, 481–489. Horton, A.C., Ra´cz, B., Monson, E.E., Lin, A.L., Weinberg, R.J., and Ehlers, M.D. (2005). Polarized secretory trafficking directs cargo for asymmetric dendrite growth and morphogenesis. Neuron 48, 757–771. Howell, B.W., Hawkes, R., Soriano, P., and Cooper, J.A. (1997). Neuronal position in the developing brain is regulated by mouse disabled-1. Nature 389, 733–737. Howell, B.W., Herrick, T.M., Hildebrand, J.D., Zhang, Y., and Cooper, J.A. (2000). Dab1 tyrosine phosphorylation sites relay positional signals during mouse brain development. Curr. Biol. 10, 877–885. Jacobs, T., Causeret, F., Nishimura, Y.V., Terao, M., Norman, A., Hoshino, M., , M. (2007). Localized activation of p21-activated kinase controls and Nikolic neuronal polarity and morphology. J. Neurosci. 27, 8604–8615. Kishi, M., Pan, Y.A., Crump, J.G., and Sanes, J.R. (2005). Mammalian SAD kinases are required for neuronal polarization. Science 307, 929–932. Lizcano, J.M., Go¨ransson, O., Toth, R., Deak, M., Morrice, N.A., Boudeau, J., Hawley, S.A., Udd, L., Ma¨kela¨, T.P., Hardie, D.G., and Alessi, D.R. (2004). LKB1 is a master kinase that activates 13 kinases of the AMPK subfamily, including MARK/PAR-1. EMBO J. 23, 833–843. Lowe, M., Rabouille, C., Nakamura, N., Watson, R., Jackman, M., Ja¨msa¨, E., Rahman, D., Pappin, D.J., and Warren, G. (1998). Cdc2 kinase directly phosphorylates the cis-Golgi matrix protein GM130 and is required for Golgi fragmentation in mitosis. Cell 94, 783–793. Marra, P., Salvatore, L., Mironov, A., Jr., Di Campli, A., Di Tullio, G., Trucco, A., Beznoussenko, G., Mironov, A., and De Matteis, M.A. (2007). The biogenesis of the Golgi ribbon: the roles of membrane input from the ER and of GM130. Mol. Biol. Cell 18, 1595–1608.
on Cajal-Retzius neurons is a crucial molecule for laminar organization of cortical neurons. Neuron 14, 899–912. Olson, E.C., Kim, S., and Walsh, C.A. (2006). Impaired neuronal positioning and dendritogenesis in the neocortex after cell-autonomous Dab1 suppression. J. Neurosci. 26, 1767–1775. Pfenninger, K.H. (2009). Plasma membrane expansion: a neuron’s Herculean task. Nat. Rev. Neurosci. 10, 251–261. Preisinger, C., Short, B., De Corte, V., Bruyneel, E., Haas, A., Kopajtich, R., Gettemans, J., and Barr, F.A. (2004). YSK1 is activated by the Golgi matrix protein GM130 and plays a role in cell migration through its substrate 14-33zeta. J. Cell Biol. 164, 1009–1020. Puthenveedu, M.A., Bachert, C., Puri, S., Lanni, F., and Linstedt, A.D. (2006). GM130 and GRASP65-dependent lateral cisternal fusion allows uniform Golgi-enzyme distribution. Nat. Cell Biol. 8, 238–248. Rice, D.S., Nusinowitz, S., Azimi, A.M., Martı´nez, A., Soriano, E., and Curran, T. (2001). The reelin pathway modulates the structure and function of retinal synaptic circuitry. Neuron 31, 929–941. Rosso, S., Bollati, F., Bisbal, M., Peretti, D., Sumi, T., Nakamura, T., Quiroga, S., Ferreira, A., and Ca´ceres, A. (2004). LIMK1 regulates Golgi dynamics, traffic of Golgi-derived vesicles, and process extension in primary cultured neurons. Mol. Biol. Cell 15, 3433–3449. Rubinson, D.A., Dillon, C.P., Kwiatkowski, A.V., Sievers, C., Yang, L., Kopinja, J., Rooney, D.L., Zhang, M., Ihrig, M.M., McManus, M.T., et al. (2003). A lentivirus-based system to functionally silence genes in primary mammalian cells, stem cells and transgenic mice by RNA interference. Nat. Genet. 33, 401–406. Sanada, K., Gupta, A., and Tsai, L.H. (2004). Disabled-1-regulated adhesion of migrating neurons to radial glial fiber contributes to neuronal positioning during early corticogenesis. Neuron 42, 197–211. Sanchez-Ponce, D., Tapia, M., Mun˜oz, A., and Garrido, J.J. (2008). New role of IKK alpha/beta phosphorylated I kappa B alpha in axon outgrowth and axon initial segment development. Mol. Cell. Neurosci. 37, 832–844. Shelly, M., Cancedda, L., Heilshorn, S., Sumbre, G., and Poo, M.M. (2007). LKB1/STRAD promotes axon initiation during neuronal polarization. Cell 129, 565–577. Su¨tterlin, C., and Colanzi, A. (2010). The Golgi and the centrosome: building a functional partnership. J. Cell Biol. 188, 621–628. ten Klooster, J.P., Jansen, M., Yuan, J., Oorschot, V., Begthel, H., Di Giacomo, V., Colland, F., de Koning, J., Maurice, M.M., Hornbeck, P., and Clevers, H. (2009). Mst4 and Ezrin induce brush borders downstream of the Lkb1/Strad/ Mo25 polarization complex. Dev. Cell 16, 551–562. Trommsdorff, M., Borg, J.P., Margolis, B., and Herz, J. (1998). Interaction of cytosolic adaptor proteins with neuronal apolipoprotein E receptors and the amyloid precursor protein. J. Biol. Chem. 273, 33556–33560. Trommsdorff, M., Gotthardt, M., Hiesberger, T., Shelton, J., Stockinger, W., Nimpf, J., Hammer, R.E., Richardson, J.A., and Herz, J. (1999). Reeler/ Disabled-like disruption of neuronal migration in knockout mice lacking the VLDL receptor and ApoE receptor 2. Cell 97, 689–701. Witte, H., Neukirchen, D., and Bradke, F. (2008). Microtubule stabilization specifies initial neuronal polarization. J. Cell Biol. 180, 619–632.
Matsuki, T., Pramatarova, A., and Howell, B.W. (2008). Reduction of Crk and CrkL expression blocks reelin-induced dendritogenesis. J. Cell Sci. 121, 1869–1875.
Yadav, S., Puri, S., and Linstedt, A.D. (2009). A primary role for Golgi positioning in directed secretion, cell polarity, and wound healing. Mol. Biol. Cell 20, 1728–1736.
Nichols, A.J., and Olson, E.C. (2010). Reelin promotes neuronal orientation and dendritogenesis during preplate splitting. Cereb. Cortex 20, 2213–2223.
Ye, B., Zhang, Y., Song, W., Younger, S.H., Jan, L.Y., and Jan, Y.N. (2007). Growing dendrites and axons differ in their reliance on the secretory pathway. Cell 130, 717–729.
Niu, S., Renfro, A., Quattrocchi, C.C., Sheldon, M., and D’Arcangelo, G. (2004). Reelin promotes hippocampal dendrite development through the VLDLR/ ApoER2-Dab1 pathway. Neuron 41, 71–84. Niwa, H., Yamamura, K., and Miyazaki, J. (1991). Efficient selection for highexpression transfectants with a novel eukaryotic vector. Gene 108, 193–199. Ogawa, M., Miyata, T., Nakajima, K., Yagyu, K., Seike, M., Ikenaka, K., Yamamoto, H., and Mikoshiba, K. (1995). The reeler gene-associated antigen
836 Cell 143, 826–836, November 24, 2010 ª2010 Elsevier Inc.
Yin, D.M., Huang, Y.H., Zhu, Y.B., and Wang, Y. (2008). Both the establishment and maintenance of neuronal polarity require the activity of protein kinase D in the Golgi apparatus. J. Neurosci. 28, 8832–8843. Zmuda, J.F., and Rivas, R.J. (1998). The Golgi apparatus and the centrosome are localized to the sites of newly emerging axons in cerebellar granule neurons in vitro. Cell Motil. Cytoskeleton 41, 18–38.
Resource
A Human Genome Structural Variation Sequencing Resource Reveals Insights into Mutational Mechanisms Jeffrey M. Kidd,1,4 Tina Graves,2 Tera L. Newman,1,5 Robert Fulton,2 Hillary S. Hayden,1 Maika Malig,1 Joelle Kallicki,2 Rajinder Kaul,1 Richard K. Wilson,2 and Evan E. Eichler1,3,* 1Department
of Genome Sciences, University of Washington School of Medicine, Seattle, WA 98195, USA University Genome Sequencing Center, School of Medicine, St Louis, MO 63108, USA 3Howard Hughes Medical Institute, Seattle, WA 98195, USA 4Present address: Department of Genetics, Stanford University, Stanford, CA 94305, USA 5Present address: iGenix, Seattle, WA 98110, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.027 2Washington
SUMMARY
Understanding the prevailing mutational mechanisms responsible for human genome structural variation requires uniformity in the discovery of allelic variants and precision in terms of breakpoint delineation. We develop a resource based on capillary end sequencing of 13.8 million fosmid clones from 17 human genomes and characterize the complete sequence of 1054 large structural variants corresponding to 589 deletions, 384 insertions, and 81 inversions. We analyze the 2081 breakpoint junctions and infer potential mechanism of origin. Three mechanisms account for the bulk of germline structural variation: microhomology-mediated processes involving short (2–20 bp) stretches of sequence (28%), nonallelic homologous recombination (22%), and L1 retrotransposition (19%). The high quality and long-range continuity of the sequence reveals more complex mutational mechanisms, including repeat-mediated inversions and gene conversion, that are most often missed by other methods, such as comparative genomic hybridization, single nucleotide polymorphism microarrays, and next-generation sequencing. INTRODUCTION Despite significant advances in the discovery and genotyping of human genome structural variation, only a small fraction of common structural variation has been resolved at the sequence level (Conrad et al., 2010b; Freeman et al., 2006; Itsara et al., 2009; Kidd et al., 2008; Lam et al., 2010; McCarroll et al., 2008b; Redon et al., 2006). The majority of human genome structural variation has been discovered with single nucleotide polymorphism (SNP) microarrays and array comparative genomic hybridization (arrayCGH), approaches that provide limited infor-
mation about the precise structure and location of identified variants. Because of their dependence on the reference genome, array-based approaches preferentially detect deletions over insertions and are unable to directly detect copy-number-neutral events such as inversions. Higher-density array platforms give a better estimation of variant sizes, but most breakpoints cannot be resolved at a scale finer than 50 bp regions (Conrad et al., 2010b), while targeted next-generation sequencing approaches have difficulty resolving breakpoints within homologous segments (Conrad et al., 2010a). These methodological biases threaten to skew our understanding of the underlying mechanisms responsible for the formation of structural variation and limit our ability to comprehensively discover and genotype this form of genetic variation. We resolve the breakpoints of 1054 structural variants based on capillary sequencing of clone inserts. The high-quality sequence of contiguous variant haplotypes allows alternative structures to be included in future human genome assemblies and provides the breakpoint resolution necessary to accurately genotype these variants in sequence data generated from next-generation sequencing platforms. The sequences and the associated clones also provide a resource for assessing future methods for structural variation discovery. RESULTS The Human Genome Structural Variation Clone Resource The high quality of the reference human genome is due, in large part, to the fact that it was assembled based on capillary sequencing of individual large insert clones whose complete sequence was resolved prior to final genome assembly. This strategy allowed complex duplicated and repetitive regions to be incorporated that were missed by other approaches (Istrail et al., 2004; She et al., 2004). Since genome structural variation is similarly biased to these regions, we proposed that developing clone libraries for a modest number of additional genomes would serve as a valuable resource for characterizing complex and difficult-to-assay regions of genome structural variation (Eichler Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc. 837
et al., 2007). The overall strategy involved the construction of individual genome libraries using a fosmid cloning vector (40 kb inserts) and capillary sequencing of the ends of the inserts to generate a high-quality end-sequence pair (ESP). Discrepancies in the length and orientation of these mapped ESPs with respect to the reference genome serve as signatures of copy-number variation and inversion, respectively. Since the underlying clones can be retrieved, the complete sequence context of the discovered structural variant can also be obtained. Previously, we discovered and cloned 1695 structural variants with fosmid libraries derived from nine individuals and presented sequence of 261 structural variants (Kidd et al., 2008; Tuzun et al., 2005). We expand this resource to include capillary end sequencing of 4.1 million additional fosmid clones from eight additional human genomes (Table S1, available online). The combined set includes 13.8 million clones derived from the genomes of six Yoruba Nigerians, five CEPH Europeans, three Japanese, two Han Chinese, and one individual of unknown ancestry. Structural Variant Alleles Using this resource, we searched for clusters of clones that suggest a structural difference when compared to the reference. We discovered a total of 2051 discordant regions (Table S1) having support from multiple clones for a structure different from the reference genome. The size distribution of the fosmid clone inserts limited us to the detection of structural variants greater than 5 kb in length. Inversions also tend to be biased to larger events because of the probability of capturing a breakpoint by a pair of end sequences. While there is no upper bound in the detection of deletions and inversions, the direct capturing of insertions larger than the insert size of the clone (40 kb) requires specialized approaches. For example, new tandem duplications may be identified with an everted clone mapping signature (Figure S1) (Cooper et al., 2008) and insertions of novel human sequence may be identified by read pairs for which only one end maps (Kidd et al., 2010). We targeted 1054 structural variants (Table S1) from nine human genomes and completely sequenced the inserts of 1167 fosmid clones (46.4 Mb of sequence). We identified 81 loci for which breakpoints could not be resolved because of difficulty in clone assembly and the limits of 40 kb fosmid inserts (see Supplemental Experimental Procedures). We defined breakpoints relative to the reference genome assembly following a two-stage procedure (Kidd et al., 2010) (Figure 1 and Table S2). We initially distinguished copy-number changes (n = 973 insertion and/or deletions) from balanced genome structural variants (81 inversions) (Figure 2). The analyzed variants altered 95 gene structures. We estimate that 1.04% (11/1054) of the sequenced alleles are already known risk factors for common and rare human diseases (Figure 3 and Table S3). Breakpoint Features Using the 40 kb of clone-based sequence, we examined the sequence features and inferred potential mechanism of origin for these variants (Table 1). We identified 30 variants associated with the expansion or contraction of a variable number of tandem repeats (VNTRs) (Buard et al., 2000; Jeffreys et al., 838 Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc.
1994; Richard et al., 2008). VNTR repeat units ranged from 17 bp to 6.5 kb with copy numbers ranging from 1 to 319 copies. We identified 198 events (20% of the total insertions and deletions) that we classified as being the result of L1 retrotransposition. Each of the 198 L1 elements associated with the retrotransposition events has a sequence identity of at least 97.5% when compared to the L1.3 reference sequence, and 152 are at least 6 kb in size, consistent with full-length elements that may be capable of subsequent retrotransposition (Beck et al., 2010). We find evidence for transduction of flanking sequence for 20% (40/198) of the sites, with the transduced segment size ranging from 45 to 968 nucleotides (median of 81.5) (Goodier et al., 2000; Moran et al., 1999; Pickeral et al., 2000). Using the transduced sequence as a marker, we identified the potential donor location for 30 of these retrotranspositions (20 insertions in the fosmid source sample and 10 insertions in the reference genome). We identified three positions that have each given rise to multiple LINE insertions (Figure 2B), suggesting the presence of L1 donor hotspots. We note that 11 of the 20 L1 insertions in the fosmid source (including the three recurrent L1 donors) correspond to elements that have been functionally determined to represent hot L1s, according to assays performed by Beck et al. (2010). We found two events consistent with the insertion of an intact HERV-K element: one insertion in the reference sequence (as indicated by clone AC209281) and an insertion contained in clone AC226770. Both events showed less than 1% divergence from the HERV-K sequence (Dewannieux et al., 2006) and were flanked by long terminal repeats (Tristem, 2000). Our discovery size thresholds (>5 kb) preclude the identification of smaller retrotransposition events arising from SVA or Alu repeats that are common when smaller structural variants are considered (Bennett et al., 2008; Korbel et al., 2007; Lam et al., 2010; Mills et al., 2006). We divided the remaining 824 structural variants into two broad categories. Class I consists of variants with no additional sequence at the breakpoint junction (Figures 4A–4D and Figure S2). Class II variants contain an additional sequence, found across the variant junction, that is not present at either of the other variant breakpoints (Figures 4E–4G). We also assessed the presence of extended sequence homology and the extent of matching sequence at the breakpoints. We note that microhomology is a qualitative term without clear delineation as 1 or 2 bp matches are expected to occur often by chance (Figure 4) and a range of homologous match lengths is observed (Conrad et al., 2010a; Lam et al., 2010). Similarly, there is ambiguity in assigning events to potential mechanisms based solely on the length of homologous segments. Consequently, we categorize events based on observed ranges of homology and consider assignment to specific mechanisms as speculative. Among the class I events, 49% (289/590) of copy-number variants contain 2–20 bp of matching sequence, indicating that microhomology-mediated mechanisms, such as microhomology-mediated end joining (MMEJ), contribute to a substantial fraction (30%) of human structural variation (Table 1) (Hastings et al., 2009; McVey and Lee, 2008; Payen et al., 2008; Roth and Wilson, 1986). Although there is large overlap in the variant size when broken down by extent of homologous sequence
A
C
10 kb deletion break 1
break 2
deletion junction
B
Align against common junction sequence
Combine pairwise alignments
Assess identity
|||||||| |||--||--|| |||||||| |||--||--|| | |----| || ||||||||
| |----| || ||||
||||
|||||||| |||--||--|| |||||||| ||| |||| |||--||--|| | || || | |----| || |||| | |----| || |||| *1*1111* **122**
|||| |||| **22
Figure 1. Sequence and Breakpoint Analyses Variant breakpoints were defined based on alignments of sequences from the sequenced insertion and deletion alleles. For example, (A) the sequence of fosmid clone AC207429 is compared with sequence from the corresponding region on chr2. A 10 kb deletion, relative to the reference sequence, is readily apparent (indicated by the red bracket). The position of segmental duplications, common repeats (LINEs are green, SINEs are purple, and LTR elements are orange), and RefSeq exons are shown. Sequence segments corresponding to three different breakpoint regions (red, green, and purple bars) are extracted for further analysis. (B) The sequence across the variant junction is aligned against each of the other two sequences and the resulting pairwise alignments are merged. The pattern of sequence identity is assessed to identify the positions where the junction sequence switches from being a better match to the first breakpoint to being a better match to sequence from the second breakpoint. The breakpoint coordinates correspond to the innermost positions that can be confidently assigned to be before and after the variant boundary. (C) The result of aligning the three segments depicted in (A). Alignment columns where the junction sequence matches the sequence from the first (leftmost) breakpoint are indicated by a 1 while alignment columns where the junction sequence matches the second (rightmost) breakpoint are indicated by a 2. Positions where all three sequences are the same are indicated by an asterisk (*). The red square highlights the position of the breakpoint coordinates (highlighted in red and green text). The two breakpoints are separated by seven nucleotides found at both breakpoints with perfect identity (blue text). Highlighted in gray is a 293 bp segment present at both breakpoints with a sequence identity of 91%. See also Tables S2 and S7.
(Figure 4C), we find that, as a class, the mean size of events associated with microhomology (2–20 bp of matching sequence, n = 289, mean size is 9.7 kb) is significantly smaller (p = 0.02926, two sample t test) than those showing a hallmark of nonallelic homologous recombination (NAHR) (R200 bp of matching sequence, n = 177, mean size is 21.0 kb). The analyzed inversions are overwhelming driven by large homologous segments with 69% (56/81) of all analyzed inversions containing stretches of matching sequence at least 200 bp in length. In contrast, only 30% (177/590) of the class I copy-number variants contain matching breakpoint sequences of at least that length. It is important to note, however, that our clone end-sequence mapping strategy is biased toward the detection of larger inversions when compared to copy-number variants. This is a direct consequence of the probability of capturing a breakpoint that diminishes when inversions become smaller than the clone insert
size. Overall, we find that younger Alu events and segmental duplications contribute most significantly to the process of NAHR (Table S4), as expected because of their higher levels of sequence identity. The strongest enrichment is found for paired Alu repeats at each breakpoint (5.2-fold enrichment). If each breakpoint is treated separately, rather than requiring that an element of the same subfamily be present at both breakpoints of a variant, then AluY also shows a substantial degree of enrichment (2.6-fold, Table S4). Since AluY is the most recently active Alu family, dispersed AluY elements are expected to have a higher degree of sequence identity than other Alu families (Batzer and Deininger, 2002; Cordaux and Batzer, 2009). Closer examination of the distribution of breakpoints within individual Alus reveals a nonuniform pattern of breakpoint density (Figure 3D). The highest density of breakpoints occurs near the position of a sequence motif (CCNCCNTNNCCNC) that has been Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc. 839
A
B
Figure 2. Sequenced Structural Variant Alleles (A) Size distribution for 1054 sequenced structural variants. Insertions, deletions, and inversions relative to the genome reference assembly are depicted separately. Note that the bins are not of equal sizes. The mean size of the sequenced variants is 14.9 kb for deletions, 6.1 kb for insertions, and 196 kb for inversions. Our variant selection methodology largely identifies deletions greater than 5 kb and insertions from 5 kb to 40 kb in size and is biased against inversions smaller than 40 kb. (B) The relationship between the donor site of transduced sequences and LINE insertion position are given for 30 events with a match to hg18 using BLAT. Relationships are shown for 20 LINE insertions in library source individuals relative to the reference (blue lines) and for 10 insertions in the genome reference (red lines). The blue circles represent three different loci associated with multiple distinct LINE insertions. See also Figure S1 and Table S1.
associated with meiotic recombination hotspots, is found in some Alu elements (Myers et al., 2008), and has also been observed for rearrangements between human and chimpanzee (Han et al., 2007; Sen et al., 2006). We find that 16% (153/973) of the insertion and deletion variants and 9% (7/81) of the inversions contain additional sequence at the variant breakpoints (class II events; Figure 4). Many of the additional insertion sequences are relatively short in length, consistent with nontemplate-directed repair associated with nonhomologous end joining (Figure 4B). For these shorter sequences, no inference could be made as to the source of the additions. However, 41% of all class II variants (66/160) contain additional sequence at the junction at least 20 bp in length. Of these longer fragments, 88% (58/66) map to another location within the human genome. Since we are limited in this study to directly capturing the breakpoints of insertions smaller than 40 kb, we repeated this comparison with only deletions relative to the assembly where we expect to have less of a bias in terms of variant size. We find that the additional junction sequences for 30 of 39 class II deletion events at least 20 bp long map elsewhere in the genome. Seventy-three percent (22/30) are found on the same chromosome as the variant. In fact, eight of the insertions map less than 1 kb away from the variant breakpoint (Figure 4G and Table S5) and all 22 are less than 250 kb from the breakpoint. This pattern suggests the action of a replication-associated process that involves template switching or strand invasion (Hastings et al., 2009; Lee et al., 2007; Smith et al., 2007). In contrast to the class I events, only 2% of the class II events (3/160) contained 840 Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc.
stretches of homologous sequence flanking the breakpoint insertion confirming they arose by mechanisms other than NAHR. Interestingly, if we examine the sequence context of these regions, we find that 20% (30/153) of class II events map within 5 kb of a segmental duplication. This represents a significant enrichment for proximity to duplicated sequence (p < 0.002 based on comparisons with randomly sampled sequences) indicating that regions flanking segmental duplications may be generally more unstable and susceptible to multiple mutational processes such as template switching during replication (Itsara et al., 2009; Lee et al., 2007; Payen et al., 2008). Gene Conversion and Structural Variation During our analysis of putative NAHR events, we identified 10 structural variants having a complex pattern of exchange inconsistent with a simple model of unequal crossover. The breakpoint region contains an interleaved pattern of alternating patches of sequences from flanking homologous segments (Figure 5). These patterns are reminiscent of multiple rounds of gene conversion, although each of these events was also associated with a copy-number variant event. Using paralogous sequence variants that distinguish the 50 and 30 homologous segments, we investigated the overall extent of this nonallelic exchange (referred to as the conversion tract length), and the number of switches before unambiguous homology to the 50 or 30 end was re-established. We determined that most (6/10) of the conversion tracts were relatively short (200–600 bp in length) with a relatively consistent number (4–6) and length (30–40 bp) of
A
C MST150
NEGR1
IRGM
chr1
chr5
AC210916 20
10
AC207974
40 kb
30
20
10
30
40 kb
Repeats SegDupMasker
TMEM50A
B
D
RHD
C1orf63
LCE3C
LCE3D
LCE3E
LCE3B
chr1
chr1
AC196522
AC196511 10
20
30
40 kb
LCE3A
0.0
10
20
30
40 kb
Repeats SegDupMasker
Figure 3. Examples of Sequenced Variants Examples of the complete sequence of structural variant alleles that have been associated with disease risk, including (A) a 45.5 kb deletion upstream of NEGR1, (B) a 72 kb deletion of RHD, (C) a 3.9 kb and a 20.1 kb deletion upstream of IRGM, and (D) a 32 kb deletion of LCE3C. See also Table S3.
switches before clear boundaries at the 50 and 30 could be re-established (Figure S3). Seven of these events have breakpoints that map within segmental duplications, and the remaining three have breakpoints that map within LINEs. Three of the variants contained at least ten switches. One variant (AC212911) showed the largest associated conversion tract with a remarkable 182 switches extending over 7.9 kb (Figure 5D). We sequenced the deletion allele with fosmids derived from three different individuals for one event (AC226182). Each of the three deletion haplotypes contained identical patterns of interleaved sequence, a finding that is consistent with the creation of the pattern at the time of variant formation, or shortly thereafter, rather than as a result of a continual conversion process between deletion and insertion alleles leading to a diverse set of related molecules over time (Figure S3). It is also possible that the conversion pattern arose before the formation of the structural variant and that the pattern we observe in sequenced variants is merely incidental or the result of a series of mismatch repair processes prior to variant formation. Nevertheless, the observed switch pattern is reminiscent of patterns of toggling previously observed at some LINE insertions (Gilbert et al., 2005, 2002; Symer et al., 2002) and suggests a mechanism of serial strand invasion/repair during the rearrangement process.
Comparison with Other Genome-wide Studies and Ascertainment Biases In this study we focused on systematically characterizing large structural variants at the single base-pair level. In order to identify events that may have been missed by the fosmid ESP approach, we compared our set of structural variants to other studies that have discovered and genotyped copy-number variants in the same DNA samples. We focused on five individuals analyzed by fosmid end sequencing (Kidd et al., 2008), Affymetrix 6.0 microarray (McCarroll et al., 2008b), and high-density oligonucleotide arrayCGH (Conrad et al., 2010b). A comparison of the three studies shows that 11%–65% of discovered variants are unique to a single study and corresponding experimental platform (Figure 6). The limited overlap should not be surprising since each approach preferentially identifies a subset of the total collection of genomic variation. For example, the fosmid ESP mapping approach can detect insertions of sequence not represented in the genome assembly (Kidd et al., 2008, 2010), as well as balanced events such as inversions (not depicted in Figure 6), whereas array approaches can more readily detect copynumber variation caused by large duplications. Differences in ascertainment extend to the resolution of breakpoint sequences. The sequenced variants described in this Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc. 841
Table 1. Summary of Events and Inferred Mechanisms Event Classification
Insertions and Deletions
Inversions
Potential Mechanisms
L1
198 (20.3%)
NA
Retrotransposition
HERV-K
2 (0.2%)
NA
Retrotransposition
Retroelements
VNTR
30 (3.1%)
Class I (no additional sequence at breakpoint)
590 (60.6%)
74 (91.3%)
0 or 1 matching nucleotides
82 (8.4%)
10 (12.3%)
NHEJ
2–20 matching nucleotides
289 (29.7%)
8 (9.9%)
NHEJ, MMEJ
21–100 matching nucleotides
28 (2.9%)
0
NAHR, other
101–199 matching nucleotides
14 (1.4%)
0
NAHR, other
R200 (NAHR)
177 (18.2%)
56 (69.1%)
NAHR
Class 2 (additional sequence at breakpoint)
Minisatellite, NAHR
153 (15.7%)
7 (8.6%)
1–10 additional nucleotides
76 (7.8%)
2 (2.5%)
NHEJ
>10 additional nucleotides
77 (7.9%)
5 (6.2%)
NHEJ, FoSTeS,template switching
973
81
Total
The number of events that fall into each breakpoint class is given. The following abbreviations are used: NHEJ, nonhomologous end joining; FoSTeS, fork stalling and template switching. See also Table S6.
manuscript include 237 of the regions targeted for array capture and 454 sequencing (Conrad et al., 2010a). Seventy of these targeted events were successfully resolved by breakpoint array-capture experiments (Table S6), with none of the events containing extended breakpoint homology successfully resolved by next-generation sequencing. We also reassessed regions discovered by other studies that were missed by the fosmid ESP approach. With the standard fosmid analysis criteria (two or more discordant clones with sufficient quality) (Tuzun et al., 2005), an overlapping deletion site is only identified for 53% (631/1193) of the corresponding deletion genotypes reported by Conrad et al. (2010b). The intersection rate increases to 75% (900/1193 sample-level genotypes) if individual deletion clones are considered with reduced quality thresholds. This suggests that much of the variation missed by the fosmid ESP approach is a result of random fluctuations in the level of clone coverage and the quality of individual sequencing reads (Cooper et al., 2008). Experimental approaches to discover structural variation can have reduced sensitivity in regions of segmental duplication because of difficulty in uniquely mapping reads or designing array probes (Cooper et al., 2008; Kidd et al., 2008; Tuzun et al., 2005). We compared the validated structural variants from Kidd et al. (2008) with those found by read-depth approaches (Alkan et al., 2009). Alkan et al. (2009) identified 113 genes that differ in copy number among three individuals. Only 38% of the genes greater than 5 kb (26/69) and identified as copy-number variable by read-depth intersect with a structural variant (reported in Kidd et al.[2008]). This result indicates that even the fosmid ESP approach has underascertained copy-number variation associated with the most variable duplicated sequences. We identified 81 loci during our sequence analysis with evidence for a nonreference structure for which we could not unambiguously define the variant breakpoint (see Supplemental Experimental Procedures). Of these 81 loci, 63 are associated 842 Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc.
with segmental duplications, including ten examples of tandem duplications. We note that 23 of these duplication-containing loci map near gaps in the National Center for Biotechnology Information (NCBI) build36 genome assembly or to sequences that have been assigned to a chromosome but not fully integrated into the genome reference sequence. Duplication-mediated copy-number variation remains underascertained in terms of sequence-level resolution of variant haplotypes and mutational mechanism analysis. If we adjust for these biases, we estimate that the fosmid ESP approach has minimally missed at least 106 structural variants associated with segmental duplications. DISCUSSION We describe a clone resource from 17 human DNA samples that provides 135-fold physical coverage of the human genome. The corresponding catalog and clones can be used to further characterize almost any segment of human euchromatin. We used this resource to assess breakpoint characteristics of 1054 events. The nature of our experimental design permitted us to discover more events mediated by larger segments of homology, providing a more complete assessment of human genetic variation. Of particular interest are complex events whose sequence features have been difficult to previously assess at a genome-wide level. The high quality and length of the sequenced fosmids combined with defined paralogous sequence events allowed us to quantify alternating sequence matches suggestive of interlocus gene conversion (Baye´s et al., 2003; Lagerstedt et al., 1997; Reiter et al., 1997; Visser et al., 2005). Using this resource, we obtained the complete structure of several alleles that have been associated with disease, including a deletion variant upstream of the NEGR1 gene associated with increased body mass index (Willer et al., 2009) (clone AC210916), two deletion polymorphisms upstream of the IRGM gene associated with Crohn’s disease (Barrett et al.,
A
B
C
D
Breakpoint Density
0.12 0.10 0.08 0.06 0.04 0.02
50
100
150
200
250
300
Position in Alu
G
E 5’bkpnt AC209239
CAAATGCAATGTTTATTAAGCAGGTACTTTGTGCTCAAGAGTATGATACAGAGCACTAT CAAATGCAATGTTTATTAAGCAGGTACTTTGTGCTCAAGAGTATGATACAGAGCACTAT
5’bkpnt AC209239
GCTGGG GCTGGGATTTGGCAGAGGGGGATTTGGCAGGGTCATAGGACAACAGCGGAGGGAAGGTC
AC209239 3’bkpnt
AGCTCAGGAGGCTTAGGCATGAGAATCACTTGAACCTGGTAGGCA CTCAGGAGGCTTAGGCATGAGAATCACTTGAACCTGGTAGGCA
F
Figure 4. Variant Breakpoint Analyses (A–D) Class I variants are defined as those without additional nucleotides at the breakpoint. (A) A histogram of the extent of matching breakpoint sequence (black) and extended breakpoint homology (gray) is shown for 590 class I copy-number events. The red line corresponds to the expected distribution of breakpoint match lengths found from 100 random permutations. Note that bin sizes are not equal. The increase in extended homology segments 250–299 bp in length corresponds to variants having Alus at their breakpoints. (B) As in (A) zoomed in to show variants having a matching sequence of 20 bp or less. (C) Box plot of variant size partitioned by length of extended breakpoint homology for 590 class I copy-number variants (red line: median; blue box: interquartile range; whiskers: within 1.53 interquartile range). (D) Breakpoint density map within a consensus Alu repeat sequence based on 269 copy-number variant events (blue box: RNA pol III promoter; black boxes: AT-rich segment between the two monomers that make up the Alu element and the poly A tail; purple box: position of motif (CCNCCNTNNCCNC) found in some Alus and associated with recombination hotspots [Myers et al., 2008]). (E–G) Class II variants contain additional sequence across the breakpoint junction. (E) A class II variant containing a 55 nucleotide-long stretch of additional sequence (in blue) that is not found at either breakpoint. (F) Histogram of the length of additional sequence found at variant breakpoints (black) and the length of detected extended homology between breakpoint sequences (gray) for 153 class II copy-number variants. (G) Genomic location for class II unmatched sequences (>20 bp) associated with deletions. The black lines connect the positions of a class II deletion variant (relative to the genome assembly) and the corresponding location where the additional sequence across the variant breakpoint can be found. The relationship for 31 deletion variants is depicted. One event involves a match to unlocalized sequence on chromosome 1 (chr1_rand). See also Figure S2 and Tables S4 and S5.
Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc. 843
A
Figure 5. Breakpoint Assessment Using Paralogous Sequence Variants
Insertion Allele Deletion Allele
B
AC216822 AC216064 AC206476 1
200
400
600
800
1000
C AC212994 AC225624 AC225305 AC203608 1
200
400
600
800
1000
1200
1400
1600
1800
D
AC212911 1
1000
E Accession AC225832 AC225305 AC216797 AC215992 AC211399 AC212994 AC226182 AC203608 AC225624 AC212911
2000
Number of switches 4 4 4 4 6 6 6 10 14 182
3000
4000
Conversion tract (bp) 2,632 632 250 116 211 205 122 1,249 454 7,899
5000
6000
7000
8000
Variant Size (kb) 27.6 6.7 8.7 16.2 10.1 3.9 108.7 20.6 5.9 30.7
2008; Bekpen et al., 2009; McCarroll et al., 2008a) (clone AC207974), and the deletion of the LCE3B and LCE3C genes. In total, we conservatively estimate that 1.04% (11/1,054) of the discovered variants are associated with disease. This yield of disease-causing alleles rivals that found by genome-wide association studies using SNPs, which have identified 779 genome-wide associations based on genotyping of at least 100,000 SNPs (http://www.genome.gov/multimedia/illustrations/ GWAS2010-3.pdf). Although the functional significance of many of the other structural variants remains to be determined, the clone resource and availability of the complete sequence of variant haplotypes will facilitate future disease association through the rapid design of assays to test for association with disease (Abe et al., 2009; An et al., 2009; Kidd et al., 2007) or direct comparison with short sequencing reads from next-generation sequence platforms (Kidd et al., 2010; Lam et al., 2010). We investigated this approach for 1024 non-VNTR sequenced structural variants (Table S7) and found that 71% (726/1024) of 844 Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc.
(A) Schematic comparison of the structures of the insertion and deletion haplotypes of a putative NAHR variant. The blue and red boxes represent homologous sequences present at the breakpoints, which mediate the rearrangement. The blue and red vertical lines identify paralogous sequence variants that distinguish the 50 and 30 copy of the matching sequence. Scanning along the deletion allele, which is missing the intervening sequence, one observes single nucleotides specific with the 50 breakpoint, followed by a stretch of sequence that matches both, then sequences that match the 30 breakpoint. (B) Representation for three variants showing a classic NAHR pattern. Each line represents the deletion allele corresponding to the indicated variant. We note a single unexpected paralogous sequence variant mismatch located 145 bp past the 30 breakpoint, which could correspond to a SNP, short gene conversion, or alignment artifact because of the placement of indels between 50 and 30 segments. (C) Representation of four variants having breakpoints that show a pattern of alternating sequences that match the 30 then 50 breakpoints. (D) An extreme pattern of alternating matches that contains 182 switches spanning over a 7.9 kb interval. (E) Rearrangements associated with gene conversion. See also Figure S3.
the variants are uniquely identifiable with a read length of 36 bp and uniqueness threshold permitting up to one substitution. This includes 32 inversions—balanced events that are invisible to arraybased genotyping approaches. As read lengths increase to 100 bp, we estimate that 88% (902/ 1024) of these variants could be genotyped. The construction of complete alternative haplotypes then facilitates the use of read-pair information to distinguish among distinct structural configurations (Antonacci et al., 2010). Although, short read technologies may miss some of the breakpoint sequences, there are many advantages to the application of short read technology to genome structural variation. This includes the detection of thousands more events per individual genome, especially variants below the detection threshold of the fosmid ESP approach. The dynamic range response and the sequence specificity of next-generation sequencing allow absolute copy number and the identity of duplicated genes to be accurately predicted. One of the strengths of this clone resource, however, is that it permits the iterative assessment of predicted variants. Clones may be retrieved corresponding to structural variants discovered by other methods applied to these 17 individuals, including newly developed approaches such as methods for identifying transposon insertions (Huang et al., 2010; Witherspoon et al., 2010). Sequencing would provide complete information regarding the structure of additional events, thereby providing a resource set of sequenced variant haplotypes. The availability of the underlying clones and potential location of the variant within a specific DNA sample provides an approach for more fully exploring the genetic architecture and mutational properties
Conrad et al. N=1,128
Kidd et al. N=1,206
634
283
790
278
128 132 130 5 5
84 76
25
McCarroll et al. N=236
Figure 6. Comparison of Events Detected from Three Studies Only variants estimated to be >5 kb are included. The Kidd et al. (2008) set includes sites of insertion or deletion in one of the five samples relative to the genome assembly; the Conrad et al. (2010b) set includes gains and losses in at least one of the five samples relative to a reference arrayCGH sample; and the McCarroll et al. (2008b) set includes CNVs that were successfully genotyped on the Affymetrix 6.0 platform and are variable among the five included samples. Prior to comparison, the variant sets within each study were merged into a single, nonredundant interval set, and any overlap among regions between studies was sufficient regardless of which sample a variant was detected in.
of these regions. Thus, we predict that such a resource will be a valuable complement for understanding the true complexity of human genetic variation as human genomes become routinely sequenced using short read sequencing technology. EXPERIMENTAL PROCEDURES Identifying and Sequencing Variant Clones Sites of structural variation, relative to the reference genome assembly, were identified through fosmid ESP mapping. Briefly, genomic DNA was obtained from transformed lymphoblastoid cell lines (available from the Coriell Cell Repository) and approximately 1 million 40 kb fragments from each individual were cloned into fosmid vectors. Paired end sequences were obtained from both ends of each fragment with standard capillary sequencing. The resulting ESPs were mapped onto the reference assembly to identify clusters of multiple clones from a single individual showing the same type of discordancy (Tuzun et al., 2005). We previously identified 1695 structural variants that have been experimentally validated (Kidd et al., 2008). In this manuscript, we focus on 1054 events for which complete, finished clone sequence is available. Highquality finished sequence was obtained for all fosmid inserts with capillarybased shotgun sequencing and assembly with the procedures established for sequencing clones as part of the Human Genome Project. Some sequenced clones contain gaps in simple sequence repeats that are not related to the detected structural variants. For one individual, NA18956, additional clones were selected with a relaxed threshold of two standard deviations larger or smaller than the observed mean insert. In some cases, multiple clones were sequenced for a single event, whereas in other loci a single clone sequence appeared to contain multiple distinct variants relative to the genome reference. Identifying Variant Breakpoints Sequences of individual fosmid inserts were initially compared to the NCBI build36 (UCSC hg18) genome reference assembly with the program miropeats
(Parsons, 1995) with a match threshold of s 400. Images summarizing these comparisons that included annotations of the repeat content, predicted and observed segmental duplications (with DupMasker [Jiang et al., 2008]), and RefSeq exons were prepared and examined to identify clones harboring a structural difference relative to the build36. Clones that mapped to unassigned or random parts of the reference genome or that do not contain an entire event (such as clones that contain one edge of a tandem duplication) were omitted from analysis. Approximate variant breakpoints were determined utilizing the context provided by long stretches of contiguous matching sequence. In many cases, the pattern of common repeats or segmental duplications was a useful aid in this assessment. For each variant, three sequences were extracted and aligned. In the case of a deletion, two sequences at the variant boundaries are extracted from the genome assembly and one sequence (termed the deletion junction sequence) is extracted from the clone. For insertions, the junction sequence is extracted from the genome assembly and two sequences corresponding to the variant boundaries in the fosmid clone are extracted. For inversions, a single breakpoint is directly captured in the sequenced clone. However, the position of the other breakpoint can be inferred based on a comparison with the genome assembly. Thus, for inversions, two sequences are extracted from the assembly at the edges of the inferred inversion and the third sequence is extracted from the clone. For inversion analysis, one of the chromosomederived segments is reverse-complimented prior to alignment. An alignment is then constructed from the extracted breakpoint segments (Kidd et al., 2010). First, an optimal global alignment is computed between the junction fragment and each of the other two fragments with the program needle with default parameters (Rice et al., 2000). These alignments are then merged to yield a single, three-sequence alignment. From this alignment, the innermost positions that can be confidently assigned to be before and after the structural variant are identified. The resulting positions are used to define membership as a class I or class II variant and correspond to the breakpoint match length depicted in Figure 4. Extended breakpoint homology was determined with both cross_match (http://www.phrap.org/, -minmatch 4 -maxmatch 4 -minscore 20 -masklevel 100 -raw -word_raw) without complexityadjusted scoring (Chiaromonte et al., 2002) and bl2seq (-W 7 -g F -F F -S 1 -e 20) to identify the longest extent and identity of additional matching sequence (termed extended breakpoint homology) that included the two breakpoints. For putative NAHR events, we additionally determined the longest stretch of 100% perfect identity as well as a parsimonious matching metric to account for mutations after the time of variant formation (Figure S2). VNTR and Retroelement Analysis Events associated with tandem repeats were characterized with the output from miropeats (Parsons, 1995), tandem repeats finder (Benson, 1999), DupMasker (Jiang et al., 2008), and RepeatMasker (Smit et al., 1996–2004). Potential L1 insertions were characterized with both the TSDfinder program (Szak et al., 2002) and the results of the breakpoint identification and characterization process. Genotyping Structural Variants with Diagnostic K-mers Diagnostic k-mers were identified for each variant (Table S7) by extracting overlapping k-mers of the indicated size across each sequenced breakpoint. K-mers were then searched against the build36 genome sequence and a set of sequenced fosmids with mrsFAST (http://mrfast.sourceforge.net/). To be considered diagnostic, a k-mer must be unique (within the given edit distance threshold) to the allele variant from which it was derived (Kidd et al., 2010). ACCESSION NUMBERS All sequence data have been deposited in GenBank under project ID 29893. SUPPLEMENTAL INFORMATION Supplemental Information includes Supplemental Experimental Procedures, three figures, and eight tables and can be found with this article online at doi:10.1016/j.cell.2010.10.027.
Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc. 845
ACKNOWLEDGMENTS We thank D. Smith and the staff at Agencourt Biosciences for library production, E. Kirkness and staff of the J. Craig Venter Institute for end-sequence data from the JVCI library, and L. Chen for computational assistance in the mapping of end-sequence data. We thank S. Girirajan, J. Moran, and C. Payen for thoughtful discussion; T. Brown for manuscript preparation assistance; and members of the University of Washington and Washington University Genome Centers for assistance with data generation. J.M.K. is supported by a National Science Foundation Graduate Research Fellowship. This work was supported by the National Institutes of Health Grant HG004120 to E.E.E., who is an investigator of the Howard Hughes Medical Institute. E.E.E is on the scientific advisory board for Pacific Biosciences. T.L.N. is an employee and founder of iGenix Inc. Received: July 6, 2010 Revised: September 15, 2010 Accepted: October 15, 2010 Published: November 24, 2010 REFERENCES Abe, H., Ochi, H., Maekawa, T., Hatakeyama, T., Tsuge, M., Kitamura, S., Kimura, T., Miki, D., Mitsui, F., Hiraga, N., et al. (2009). Effects of structural variations of APOBEC3A and APOBEC3B genes in chronic hepatitis B virus infection. Hepatol. Res. 39, 1159–1168. Alkan, C., Kidd, J.M., Marques-Bonet, T., Aksay, G., Antonacci, F., Hormozdiari, F., Kitzman, J.O., Baker, C., Malig, M., Mutlu, O., et al. (2009). Personalized copy number and segmental duplication maps using next-generation sequencing. Nat. Genet. 41, 1061–1067. An, P., Johnson, R., Phair, J., Kirk, G.D., Yu, X.F., Donfield, S., Buchbinder, S., Goedert, J.J., and Winkler, C.A. (2009). APOBEC3B deletion and risk of HIV-1 acquisition. J. Infect. Dis. 200, 1054–1058. Antonacci, F., Kidd, J.M., Marques-Bonet, T., Teague, B., Ventura, M., Girirajan, S., Alkan, C., Campbell, C.D., Vives, L., Malig, M., et al. (2010). A large and complex structural polymorphism at 16p12.1 underlies microdeletion disease risk. Nat. Genet. 42, 745–750. Barrett, J.C., Hansoul, S., Nicolae, D.L., Cho, J.H., Duerr, R.H., Rioux, J.D., Brant, S.R., Silverberg, M.S., Taylor, K.D., Barmada, M.M., et al; NIDDK IBD Genetics Consortium; Belgian-French IBD Consortium; Wellcome Trust Case Control Consortium. (2008). Genome-wide association defines more than 30 distinct susceptibility loci for Crohn’s disease. Nat. Genet. 40, 955–962. Batzer, M.A., and Deininger, P.L. (2002). Alu repeats and human genomic diversity. Nat. Rev. Genet. 3, 370–379. Baye´s, M., Magano, L.F., Rivera, N., Flores, R., and Pe´rez Jurado, L.A. (2003). Mutational mechanisms of Williams-Beuren syndrome deletions. Am. J. Hum. Genet. 73, 131–151. Beck, C.R., Collier, P., Macfarlane, C., Malig, M., Kidd, J.M., Eichler, E.E., Badge, R.M., and Moran, J.V. (2010). LINE-1 retrotransposition activity in human genomes. Cell 141, 1159–1170. Bekpen, C., Marques-Bonet, T., Alkan, C., Antonacci, F., Leogrande, M.B., Ventura, M., Kidd, J.M., Siswara, P., Howard, J.C., and Eichler, E.E. (2009). Death and resurrection of the human IRGM gene. PLoS Genet. 5, e1000403. Bennett, E.A., Keller, H., Mills, R.E., Schmidt, S., Moran, J.V., Weichenrieder, O., and Devine, S.E. (2008). Active Alu retrotransposons in the human genome. Genome Res. 18, 1875–1883. Benson, G. (1999). Tandem repeats finder: a program to analyze DNA sequences. Nucleic Acids Res. 27, 573–580. Buard, J., Shone, A.C., and Jeffreys, A.J. (2000). Meiotic recombination and flanking marker exchange at the highly unstable human minisatellite CEB1 (D2S90). Am. J. Hum. Genet. 67, 333–344. Chiaromonte, F., Yap, V.B., and Miller, W. (2002). Scoring pairwise genomic sequence alignments. Pacific Symposium on Biocomputing 7, 115–126.
846 Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc.
Conrad, D.F., Bird, C., Blackburne, B., Lindsay, S., Mamanova, L., Lee, C., Turner, D.J., and Hurles, M.E. (2010a). Mutation spectrum revealed by breakpoint sequencing of human germline CNVs. Nat. Genet. 42, 385–391. Conrad, D.F., Pinto, D., Redon, R., Feuk, L., Gokcumen, O., Zhang, Y., Aerts, J., Andrews, T.D., Barnes, C., Campbell, P., et al; Wellcome Trust Case Control Consortium. (2010b). Origins and functional impact of copy number variation in the human genome. Nature 464, 704–712. Cooper, G.M., Zerr, T., Kidd, J.M., Eichler, E.E., and Nickerson, D.A. (2008). Systematic assessment of copy number variant detection via genome-wide SNP genotyping. Nat. Genet. 40, 1199–1203. Cordaux, R., and Batzer, M.A. (2009). The impact of retrotransposons on human genome evolution. Nat. Rev. Genet. 10, 691–703. Dewannieux, M., Harper, F., Richaud, A., Letzelter, C., Ribet, D., Pierron, G., and Heidmann, T. (2006). Identification of an infectious progenitor for the multiple-copy HERV-K human endogenous retroelements. Genome Res. 16, 1548–1556. Eichler, E.E., Nickerson, D.A., Altshuler, D., Bowcock, A.M., Brooks, L.D., Carter, N.P., Church, D.M., Felsenfeld, A., Guyer, M., Lee, C., et al; Human Genome Structural Variation Working Group. (2007). Completing the map of human genetic variation. Nature 447, 161–165. Freeman, J.L., Perry, G.H., Feuk, L., Redon, R., McCarroll, S.A., Altshuler, D.M., Aburatani, H., Jones, K.W., Tyler-Smith, C., Hurles, M.E., et al. (2006). Copy number variation: new insights in genome diversity. Genome Res. 16, 949–961. Gilbert, N., Lutz-Prigge, S., and Moran, J.V. (2002). Genomic deletions created upon LINE-1 retrotransposition. Cell 110, 315–325. Gilbert, N., Lutz, S., Morrish, T.A., and Moran, J.V. (2005). Multiple fates of L1 retrotransposition intermediates in cultured human cells. Mol. Cell. Biol. 25, 7780–7795. Goodier, J.L., Ostertag, E.M., and Kazazian, H.H., Jr. (2000). Transduction of 30 -flanking sequences is common in L1 retrotransposition. Hum. Mol. Genet. 9, 653–657. Han, K., Lee, J., Meyer, T.J., Wang, J., Sen, S.K., Srikanta, D., Liang, P., and Batzer, M.A. (2007). Alu recombination-mediated structural deletions in the chimpanzee genome. PLoS Genet. 3, 1939–1949. Hastings, P.J., Ira, G., and Lupski, J.R. (2009). A microhomology-mediated break-induced replication model for the origin of human copy number variation. PLoS Genet. 5, e1000327. Huang, C.R., Schneider, A.M., Lu, Y., Niranjan, T., Shen, P., Robinson, M.A., Steranka, J.P., Valle, D., Civin, C.I., Wang, T., et al. (2010). Mobile interspersed repeats are major structural variants in the human genome. Cell 141, 1171– 1182. Istrail, S., Sutton, G.G., Florea, L., Halpern, A.L., Mobarry, C.M., Lippert, R., Walenz, B., Shatkay, H., Dew, I., Miller, J.R., et al. (2004). Whole-genome shotgun assembly and comparison of human genome assemblies. Proc. Natl. Acad. Sci. USA 101, 1916–1921. Itsara, A., Cooper, G.M., Baker, C., Girirajan, S., Li, J., Absher, D., Krauss, R.M., Myers, R.M., Ridker, P.M., Chasman, D.I., et al. (2009). Population analysis of large copy number variants and hotspots of human genetic disease. Am. J. Hum. Genet. 84, 148–161. Jeffreys, A.J., Tamaki, K., MacLeod, A., Monckton, D.G., Neil, D.L., and Armour, J.A.L. (1994). Complex gene conversion events in germline mutation at human minisatellites. Nat. Genet. 6, 136–145. Jiang, Z., Hubley, R., Smit, A., and Eichler, E.E. (2008). DupMasker: a tool for annotating primate segmental duplications. Genome Res. 18, 1362–1368. Kidd, J.M., Newman, T.L., Tuzun, E., Kaul, R., and Eichler, E.E. (2007). Population stratification of a common APOBEC gene deletion polymorphism. PLoS Genet. 3, e63. Kidd, J.M., Cooper, G.M., Donahue, W.F., Hayden, H.S., Sampas, N., Graves, T., Hansen, N., Teague, B., Alkan, C., Antonacci, F., et al. (2008). Mapping and sequencing of structural variation from eight human genomes. Nature 453, 56–64.
Kidd, J.M., Sampas, N., Antonacci, F., Graves, T., Fulton, R., Hayden, H.S., Alkan, C., Malig, M., Ventura, M., Giannuzzi, G., et al. (2010). Characterization of missing human genome sequences and copy-number polymorphic insertions. Nat. Methods 7, 365–371. Korbel, J.O., Urban, A.E., Affourtit, J.P., Godwin, B., Grubert, F., Simons, J.F., Kim, P.M., Palejev, D., Carriero, N.J., Du, L., et al. (2007). Paired-end mapping reveals extensive structural variation in the human genome. Science 318, 420–426. Lagerstedt, K., Karsten, S.L., Carlberg, B.M., Kleijer, W.J., To¨nnesen, T., Pettersson, U., and Bondeson, M.L. (1997). Double-strand breaks may initiate the inversion mutation causing the Hunter syndrome. Hum. Mol. Genet. 6, 627–633.
Reiter, L.T., Murakami, T., Koeuth, T., Gibbs, R.A., and Lupski, J.R. (1997). The human COX10 gene is disrupted during homologous recombination between the 24 kb proximal and distal CMT1A-REPs. Hum. Mol. Genet. 6, 1595–1603. Rice, P., Longden, I., and Bleasby, A. (2000). EMBOSS: the European Molecular Biology Open Software Suite. Trends Genet. 16, 276–277. Richard, G.F., Kerrest, A., and Dujon, B. (2008). Comparative genomics and molecular dynamics of DNA repeats in eukaryotes. Microbiol. Mol. Biol. Rev. 72, 686–727. Roth, D.B., and Wilson, J.H. (1986). Nonhomologous recombination in mammalian cells: role for short sequence homologies in the joining reaction. Mol. Cell. Biol. 6, 4295–4304.
Lam, H.Y., Mu, X.J., Stu¨tz, A.M., Tanzer, A., Cayting, P.D., Snyder, M., Kim, P.M., Korbel, J.O., and Gerstein, M.B. (2010). Nucleotide-resolution analysis of structural variants using BreakSeq and a breakpoint library. Nat. Biotechnol. 28, 47–55.
Sen, S.K., Han, K., Wang, J., Lee, J., Wang, H., Callinan, P.A., Dyer, M., Cordaux, R., Liang, P., and Batzer, M.A. (2006). Human genomic deletions mediated by recombination between Alu elements. Am. J. Hum. Genet. 79, 41–53.
Lee, J.A., Carvalho, C.M., and Lupski, J.R. (2007). A DNA replication mechanism for generating nonrecurrent rearrangements associated with genomic disorders. Cell 131, 1235–1247.
She, X., Jiang, Z., Clark, R.A., Liu, G., Cheng, Z., Tuzun, E., Church, D.M., Sutton, G., Halpern, A.L., and Eichler, E.E. (2004). Shotgun sequence assembly and recent segmental duplications within the human genome. Nature 431, 927–930.
McCarroll, S.A., Huett, A., Kuballa, P., Chilewski, S.D., Landry, A., Goyette, P., Zody, M.C., Hall, J.L., Brant, S.R., Cho, J.H., et al. (2008a). Deletion polymorphism upstream of IRGM associated with altered IRGM expression and Crohn’s disease. Nat. Genet. 40, 1107–1112. McCarroll, S.A., Kuruvilla, F.G., Korn, J.M., Cawley, S., Nemesh, J., Wysoker, A., Shapero, M.H., de Bakker, P.I., Maller, J.B., Kirby, A., et al. (2008b). Integrated detection and population-genetic analysis of SNPs and copy number variation. Nat. Genet. 40, 1166–1174. McVey, M., and Lee, S.E. (2008). MMEJ repair of double-strand breaks (director’s cut): deleted sequences and alternative endings. Trends Genet. 24, 529–538. Mills, R.E., Luttig, C.T., Larkins, C.E., Beauchamp, A., Tsui, C., Pittard, W.S., and Devine, S.E. (2006). An initial map of insertion and deletion (INDEL) variation in the human genome. Genome Res. 16, 1182–1190. Moran, J.V., DeBerardinis, R.J., and Kazazian, H.H., Jr. (1999). Exon shuffling by L1 retrotransposition. Science 283, 1530–1534. Myers, S., Freeman, C., Auton, A., Donnelly, P., and McVean, G. (2008). A common sequence motif associated with recombination hot spots and genome instability in humans. Nat. Genet. 40, 1124–1129. Parsons, J.D. (1995). Miropeats: graphical DNA sequence comparisons. Comput. Appl. Biosci. 11, 615–619. Payen, C., Koszul, R., Dujon, B., and Fischer, G. (2008). Segmental duplications arise from Pol32-dependent repair of broken forks through two alternative replication-based mechanisms. PLoS Genet. 4, e1000175.
Smit, A., Hubley, R., and Green, P. (1996–2004). RepeatMasker Open-3.0. http://www.repeatmasker.org/cgi-bin/WEBRepeatMasker. Smith, C.E., Llorente, B., and Symington, L.S. (2007). Template switching during break-induced replication. Nature 447, 102–105. Symer, D.E., Connelly, C., Szak, S.T., Caputo, E.M., Cost, G.J., Parmigiani, G., and Boeke, J.D. (2002). Human l1 retrotransposition is associated with genetic instability in vivo. Cell 110, 327–338. Szak, S.T., Pickeral, O.K., Makalowski, W., Boguski, M.S., Landsman, D., and Boeke, J.D. (2002). Molecular archeology of L1 insertions in the human genome. Genome Biol. 3, research0052. Tristem, M. (2000). Identification and characterization of novel human endogenous retrovirus families by phylogenetic screening of the human genome mapping project database. J. Virol. 74, 3715–3730. Tuzun, E., Sharp, A.J., Bailey, J.A., Kaul, R., Morrison, V.A., Pertz, L.M., Haugen, E., Hayden, H., Albertson, D., Pinkel, D., et al. (2005). Fine-scale structural variation of the human genome. Nat. Genet. 37, 727–732. Visser, R., Shimokawa, O., Harada, N., Kinoshita, A., Ohta, T., Niikawa, N., and Matsumoto, N. (2005). Identification of a 3.0-kb major recombination hotspot in patients with Sotos syndrome who carry a common 1.9-Mb microdeletion. Am. J. Hum. Genet. 76, 52–67.
Pickeral, O.K., Maka1owski, W., Boguski, M.S., and Boeke, J.D. (2000). Frequent human genomic DNA transduction driven by LINE-1 retrotransposition. Genome Res. 10, 411–415.
Willer, C.J., Speliotes, E.K., Loos, R.J., Li, S., Lindgren, C.M., Heid, I.M., Berndt, S.I., Elliott, A.L., Jackson, A.U., Lamina, C., et al; Wellcome Trust Case Control Consortium; Genetic Investigation of ANthropometric Traits Consortium. (2009). Six new loci associated with body mass index highlight a neuronal influence on body weight regulation. Nat. Genet. 41, 25–34.
Redon, R., Ishikawa, S., Fitch, K.R., Feuk, L., Perry, G.H., Andrews, T.D., Fiegler, H., Shapero, M.H., Carson, A.R., Chen, W., et al. (2006). Global variation in copy number in the human genome. Nature 444, 444–454.
Witherspoon, D.J., Xing, J., Zhang, Y., Watkins, W.S., Batzer, M.A., and Jorde, L.B. (2010). Mobile element scanning (ME-Scan) by targeted high-throughput sequencing. BMC Genomics 11, 410.
Cell 143, 837–847, November 24, 2010 ª2010 Elsevier Inc. 847
SnapShot: The SUMO System Sandrine Creton and Stefan Jentsch Max Planck Institute of Biochemistry, Martinsried 82152, Germany
848
Cell 143, November 24, 2010 ©2010 Elsevier Inc. DOI 10.1016/j.cell.2010.11.026
See online version for legend and references.
SnapShot: The SUMO System Sandrine Creton and Stefan Jentsch Max Planck Institute of Biochemistry, Martinsried 82152, Germany SUMO (small ubiquitin-related modifier) is a highly conserved eukaryotic member of the family of ubiquitin-related proteins. Despite limited sequence similarity, SUMO and ubiquitin have highly homologous structures (i.e., with a root-mean-square deviation [rmsd] difference of 2.1 Å). Like ubiquitin, SUMO is covalently attached to other proteins (SUMOylation) and thus functions as a posttranslational modifier. Although a major function of ubiquitination is to promote protein degradation, SUMOylation does not usually trigger proteolyis of the conjugated protein. Instead, a main function of SUMOylation is to foster—and occasionally disrupt—protein-protein interactions. Although less frequent than ubiquitination, SUMOylation regulates numerous processes and has many substrates in the cytosol and the nucleus. Lower eukaryotes possess only one SUMO form, but higher eukaryotes express several, perhaps functionally distinct, SUMO variants. In most organisms, SUMO is essential for viability. This SnapShot depicts the reactions involved in SUMO ligation, the effects of SUMOylation on target proteins, and cellular processes regulated by SUMOylation. SUMO Enzymology Similar to ubiquitination, SUMOylation requires ATP and a series of enzymes for conjugation (Table A). SUMO is initially produced as an inactive precursor, which is then processed at the C terminus by SUMO-specific proteases (yeast Ulp1, mammalian SENPs), giving rise to mature SUMO with a C-terminal Gly-Gly motif. Conjugation involves at least two enzymes, an ATP-dependent activating enzyme (E1, a heterodimer) and a conjugating enzyme (E2, Ubc9). SUMO is linked to these enzymes by the formation of a thioester bond between SUMO’s C terminus and cysteine residues in the active sites of these enzymes. For some substrates, SUMOylation further requires specific SUMO ligases (E3). SUMOylation targets lysine residues (by isopeptide formation), often, but not always, within a consensus site