Leading Edge
In This Issue Making Neurons from Rett Patient Stem Cells PAGE 527
Using Rett syndrome (RTT) as a prototype autism spectrum disorder (ASD), Marchetto et al. generated induced pluripotent stem cells (iPSCs). Neurons derived from RTT-iPSCs had fewer synapses and altered connectivity when compared to control neurons. Strikingly, synaptic defects in RTT neurons could be rescued pharmacologically, suggesting that there is a developmental window in an autism syndrome when potential therapies could be effective. These RTT patient-derived neurons also provide a promising cellular tool for drug screening, biomarker identification, and personalized treatment.
Translation Gets Pol II’s Stamp of Approval PAGE 552
Rpb4/7, a heterodimer of RNA polymerase II (Pol II) subunits, regulates premRNA synthesis. It stays bound to the mRNA and accompanies it to the cytoplasm where, Harel-Sharvit et al. now report, the dimer interacts directly with eIF3 to regulate translation. Efficient translation depends on Rpb4/7’s prior association with RNA Pol II in the nucleus. Thus, Rpb4/7 integrates the various stages of gene expression into one functional system.
Strippers for Protein Polyglutamylation PAGE 564
Proteins can be polyglutamylated, but the enzymes that remove these groups are unknown. Rogowski et al. now identify several protein deglutamylases as members of the cytosolic carboxypeptidase (CCP) family and describe the enzymatic mechanism of protein deglutamylation. Mice lacking CCP1 are subject to hyperglutamlyation of tubulin, and subsequent neurodegeneration, suggesting regulatory roles for control of polyglutamylation.
Hrd-ing Misfolded Proteins out of the ER PAGE 579
Misfolded proteins in the endoplasmic reticulum (ER) need to be transported back into the cytosol for polyubiquitination and degradation. In this issue, Carvalho et al. show that the ubiquitin ligase Hrd1p is the crucial membrane component in this retrotranslocation. The authors propose that Hrd1p cooperates with the cytosolic Cdc48p ATPase complex to move misfolded proteins through the ER membrane.
Paused Polymerase Takes On DNA-Encoded Nucleosome Formation PAGE 540
Pausing of RNA polymerase II during early elongation was initially thought to be a rare mechanism for attenuating transcription; however, it’s now known to be widespread, raising questions about the role of paused polymerase. Here, Gilchrist et al. report that pausing stimulates transcription potential by preventing formation of repressive promoter chromatin. Their findings suggest that competition between paused polymerase and nucleosomes for promoter occupancy has evolved as a prevalent gene regulatory strategy.
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 487
T Cells Help B Cells Help Themselves PAGE 592
Selection for B cells with high-affinity B cell receptors occurs in the germinal centers (GCs). This process was thought to involve competition of B cells for antigen in the GC. Using a novel live-microscopy approach, Victora and colleagues visualize the dynamics of GC selection. Their findings suggest that access to help from GC-resident T cells, and not competition for antigen, is the limiting factor in GC selection.
UPBEAT in the Zone PAGE 606
In Arabidopsis, stem cell progeny at the root tip rapidly proliferate in an amplifying zone and then increase in size and differentiate in an adjacent elongation zone. In this issue, Tsukagoshi et al. identify a transcription factor, UPBEAT1 (UPB1), that regulates the balance between proliferation and differentiation at the interface of these two zones. UPB1 regulates levels of reactive oxygen species (ROS) in the transition zone by controlling the expression of a set of peroxidases. Comparison to ROS-regulated growth control in animals suggests that a similar mechanism is used in plants and animals.
A Gemisch of EpiSCs PAGE 617
Pluripotent stem cells contribute to teratomas when injected into an organism or to chimeras when injected into an early embryo. However, stem cells derived from mouse epiblast tissue (EpiSCs) show restricted pluripotency and rarely form chimeras. Han et al. now show that EpiSCs actually comprise two different subpopulations of cells in culture. The major subpopulation shares features with late mouse epiblast cells and cannot form chimeras, whereas cells in the minor population resemble early mouse epiblast cells and can readily form chimeras, thus explaining the restricted potency of EpiSCs.
Sounds Painful PAGE 628
Acute and chronic pain affects millions of people worldwide. In this issue, Neely et al. use genome-wide RNAi screening in Drosophila to identify genes involved in heat nociception. Mice mutant for one of these genes, a2d3 (straightjacket in flies), have impaired heat pain sensitivity, and in humans, a2d3 SNP variants associate with reduced sensitivity to acute noxious heat and chronic back pain. Surprisingly, functional brain imaging of a2d3 mutant mice indicated that thermal pain and tactile stimulation trigger strong sensory cross-activation of brain regions involved in vision, olfaction, and hearing.
Metabolites Move up the Regulatory Chain PAGE 639
Small metabolites represent the vast majority of molecules in a cell. Li et al. now provide insight into protein-metabolite interactions, suggesting that they have an extensive role in the regulation of protein activity. Among the many prospective interactions identified, the authors specifically characterize those involving ergosterol, the yeast analog of cholesterol, which they show binds to many proteins including kinases, thereby affecting their functions and levels.
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 489
Leading Edge
Select: Manipulating Cellular Machinery Cells can be manipulated for any number of purposes, owing to the emergence of synthetic biology and bioengineering tools. Recent biotechnological advances have generated improved therapeutics, sophisticated biosensors, and new energy sources. The findings presented in this Select highlight innovative ways to both dissect and reconstruct various cellular machineries and components.
An Untaxing Way to Produce Taxol Precursors Taxol (paclitaxel) is a potent anticancer therapeutic typically used to treat breast, lung, and ovarian cancers. Historically it has been produced by synthetic or semisynthetic routes, requiring as many as 51 synthesis steps or the use of a plant-derived intermediate, baccatin III, respectively. Ajikumar et al. (2010) introduce a new approach to produce Taxol: engineering the isoprenoid pathway in Escherichia coli. There are two pathways to consider for Taxol biosynthesis: the native upstream isoprenoid pathway produces metabolic precursors, and the downstream taxadiene pathway has been reconstructed successfully in E. coli. To date, most groups have assumed that these pathways are additive, without paying due attention to toxic side effects of intermediate metabolites or competing pathways and metabolites that could inhibit the drug production. With traditional combinatorial approaches unavailable for taxadiene, the team undertook a ‘‘multivariate modular pathway engineering’’ approach, splitting the pathway into two individually addressable modules. The upstream module focused on the four rate-limiting genes of the methylerythritol phosphate pathway. The downstream module comprised the Microbes can be instructed to build complex two-gene pathway, resulting in taxadiene. By tweaking various genetic compochemical networks. Image courtesy of G. Stephnents and expression levels in the modules, the authors were able to optimize anopoulos. taxadiene production to levels 15,000-fold greater than the control, yielding 1.02 g/l (to date, the highest reported titers from metabolically engineered microbes have been a modest 10 mg/l). These yields notwithstanding, the key challenge in the generation of Taxolproducing microbes is the oxidation of taxadiene to taxadien-5a-ol via CYP450, a plant enzyme that is difficult to express in E. coli. By generating a chimeric CYP450, the authors are able to produce the alcohol but discover drastic reductions in taxadiene production, with concomitant increases in a pathway inhibitor, indole. Improvements to the CYP450-based oxidation chemistry that converts taxadiene to taxadien-5a-ol will open the door to commercial-scale synthesis of Taxol and other high-value terpenoids for use as chemicals or fuels. P.K. Ajikumar et al. (2010). Science 343, 70-74.
This Cell Can Smell Whole-cell biosensors have attracted great interest in recent years, owing to their ability to detect environmental contaminants with high specificity and sensitivity. Misawa et al. (2010) now show that oocytes from Xenopus laevis can also be engineered for use as chemical sensors, detecting odorants under controlled conditions in a portable fluidic device. Because of their large size (1 mm in diameter), oocytes have been used as model systems in electrophysiology experiments; and, like all living systems, they can respond to certain chemicals by converting chemical signals into an amplified current signal, which can be detected using capillary electrodes. The authors use microinjections to introduce RNAs encoding the odorant or pheromone receptors from moths or flies that detect bombykol, bombykal, Z11-16:Ald, or 2-heptanone. The oocytes are trapped within a microfluidic chamber under optimized temperature and flow conditions. The modified oocytes can ‘‘smell’’ odorants down to a concentration of 10 nM, with variation in sensing capabilities most likely due to variability in receptor expression levels. The livecell sensor is also able to discriminate between bombykol and bombykal, which differ only by a terminal alcohol or aldehyde, respectively, thereby highlighting the sensor’s inherent specificity in detecting slight differences in chemical structures. The bio-hybrid sensor is ultimately interfaced with a robot to demonstrate that it retains the capacity for chemical sensing under shaking conditions to mimic ‘‘real world’’ implementation of the technology. N. Misawa et al. (2010). Proc. Natl. Acad. Sci. 107, 15340-15344.
A portable fluidic device that detects odorants can be attached to a human-like robot. Image courtesy of S. Takeuchi.
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 491
Fast Track to Biofuels with Cellodextrin Transport The plant cell wall is highly resistant to degradation, which poses a significant challenge to the bioethanol industry. Saccharomyces cerevisiae is currently the yeast-of-choice for fermenting sugars from cornstarch or sugarcane; but these yeasts are unable to naturally ferment cellodextrins derived from plant cell walls without the assistance of cellulase cocktails containing b-glucosidases, which release yeast-fermentable glucose. The degradation of cellodextrins exhibited by the cellulolytic fungus Neurospora crassa has inspired Galazaka et al. (2010) to create new yeast strains for enhanced growth and production of bioethanol. Two sugar transporters, CDT-1 and CDT-2, are important for plant-fungi symbiosis and plant degradation. Genes encoding intracellular b-glucosidases are also widespread in fungi, attesting to their importance for optimal growth of fungi on cellulose-derived sugars. Toward complete cellodextrin catabolism and production of lignocellulosic biofuels, the authors re-engineered an S. cerevisiae strain to contain the functional cellodextrin transport system from N. crassa (CDT-1, CDT-2, and a b-glucosidase, GH1-1). Simultaneous saccharification and fermentation (SSF) is typically limited by the need for complete extracellular conversion of cellodextrin to glucose, which is then transported into the cell. For SSF in yeast re-engineered to express cdt-1 and gh1-1, the authors bypassed this extracellular step to allow the cell to take up cellodextrin via CDT, with intracellular conversion to glucose via b-glucosidases. The result is a biofuel-producing S. cerevisiae strain that sidesteps the fermentation bottleneck for lignocellulosic biomass, with yields comparable to those reported from the bioethanol industry. J.M. Galazka et al. (2010). Science 330, 84-86.
Redecorating the Walls Antibiotic-resistant Staphylococcus aureus has emerged as a major public health threat, with the number of S. aureus-infection-related deaths on the rise. Nelson et al. (2010) present an innovative engineering approach to incorporate nonnative small molecules into the S. aureus cell wall—a technique that could have myriad implications for imaging bacteria, engineering organisms with novel functions, and, perhaps most importantly, discovering new therapeutics. To incorporate small molecules into the bacterial cell wall, small molecules conjugated to ‘‘cell wall sorting’’ peptides with the sequence LPETG were administered to S. aureus. These peptide sequences were recognized by the periplasmic enzyme sortase A, which covalently cleaves the peptide and attaches the new N terminus to lipid II in the cell wall. The authors confirm, using immunocryoelectron microscopy and mass spectrometry, respecThe S. aureus cell wall is visualized using a biosynthetically tively, that the peptides localize and covalently attach to the cell wall. Covalent attachment of incorporated small molecule small molecules then permits the incorporation of azido groups into the cell wall, which can fluorophore. Image courtesy undergo click reactions with alkyne groups in a manner that does not interfere with native of D. Spiegel. cellular biochemistry. Thus, the authors demonstrate the ability to functionalize the bacterial surface with an array of reactive molecules, which could enable further perturbation and studies of bacterial process in vitro and in vivo. J.W. Nelson et al. (2010). ACS Chemical Biology published online October 5, 2010. 10.1021/cb100195d.
High-Frequency Genes, High-Specificity Antibodies Antigen-specific monoclonal antibodies are typically identified by screening immortalized B cells or recombinant antibody libraries. Both in vivo and in vitro methods suffer from complex design and technical drawbacks that limit their use for rapid and reliable antibody selection. In a recent report, Reddy et al. (2010) present an antibody isolation method that uses highthroughput DNA sequencing to analyze the variable heavy (VH) and light (VL) gene repertoires of antibody-secreting bone marrow plasma cells from immunized mice. Bioinformatic analysis then ranks the frequency of V genes, which permits the identification of highly expressed VH and VL genes whose products are the most likely to be antigen specific. V genes appearing at roughly the same frequency are reasoned to be expressed as a pair in the same plasma cell; these pairs are then synthesized and expressed as antibody fragments in Escherichia coli or as full-length immunoglobulin in mammalian cells. An antibody specific for the complement protein C1s isolated using this approach demonstrated subnanomolar binding affinity and functionality in ELISA and immunoprecipitation assays. Because this methodology relies on analysis of the V gene repertoire from bone marrow plasma cells, which produce the most abundant type of circulating antibodies, the authors hypothesize that this screening-free, rapid isolation approach could be used to generate therapeutic antibodies. In addition, this approach can be extended to other B cell populations and thus allow for the generation of antibodies that are tailored to an individual patient’s immune system. S.T. Reddy et al. (2010). Nature Biotechnology 28, 965-969. Megan Frisk
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 493
Leading Edge
Previews Modeling Rett Syndrome with Stem Cells Ryan M. Walsh1,2,3 and Konrad Hochedlinger1,2,3,* 1Howard Hughes Medical Institute at Massachusetts General Hospital, Center for Regenerative Medicine and Cancer Center, Boston, MA 02114, USA 2Harvard Stem Cell Institute, 42 Church Street, Cambridge, MA 02138, USA 3Department of Stem Cell and Regenerative Biology, Harvard University and Harvard Medical School, 7 Divinity Avenue, Cambridge, MA 02138, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.037
The discovery that somatic cells can be reprogrammed into induced pluripotent stem cells (iPSCs) raised the exciting possibility of modeling diseases with patient-specific cells. Marchetto et al. (2010) now use iPSC technology to generate, characterize, and treat an in vitro model for the autism spectrum disorder Rett syndrome. Rett syndrome is a severe X-linked neurodevelopmental disorder that affects 1 in 10,000–20,000 girls worldwide, making it one of the most common forms of mental retardation in females (Percy and Lane, 2005). A seminal discovery by Huda Zoghbi’s lab in 1999 identified a causative link between mutations in the methyl-CpG binding protein, MeCP2, and Rett syndrome (Amir et al., 1999), thus enabling mechanistic studies and providing a target for potential treatments. Importantly, restoration of MeCP2 function in a mouse model of the disease reverses the neurological symptoms in adult mice (Guy et al., 2007), raising the possibility that this disorder may be treatable in humans. Despite this progress, it remains unclear how loss of MeCP2 function leads to neurological defects, and no effective pharmacological treatments have yet been developed. A main limitation for human studies and thus drug development has been the inaccessibility of live neurons from human patients. To circumvent this shortcoming, Marchetto and colleagues (Marchetto et al., 2010 [this issue of Cell]) use induced pluripotent stem cells (iPSCs) to establish a human cellular model for Rett syndrome that is amenable to mechanistic studies and drug screens. To generate a human model of Rett syndrome, Marchetto et al. (2010) isolate fibroblasts from four female Rett patients and five healthy control individuals and then reprogram these fibroblasts into iPSC lines (Figure 1). These cell lines express the expected pluripotency markers and give rise to cell types of all
germ layers in teratomas (solid tumors derived from pluripotent cells), thus qualifying these cell lines as bona fide pluripotent cells. Given that hallmarks of Rett syndrome include changes in neuronal density and in brain size, the authors first determine whether these phenotypes are likely due to the abnormal proliferation of neural progenitor cells derived from iPSCs. Notably, they observe no overt defects in the cell cycle of neural progenitor cells, consistent with the notion that Rett is a disease of mature neurons. In contrast to neural progenitor cells, mature neurons derived from Rett iPSCs do show defects in structure and function when compared to neurons obtained from control iPSCs or embryonic stem cells (ESCs). For example, the authors detect a significant reduction in the number of synapses in glutamatergic neurons derived from Rett iPSCs when compared to neurons derived from either control iPSCs or ESCs. This reduction is likely the direct consequence of losing MeCP2 function, given that reducing the expression of MeCP2 in control ESCs produces a similar defect whereas the reintroduction of wild-type MeCP2 into mutant cells rescues the phenotype. The authors further confirm MeCP2’s role in regulating synapse formation by showing that overexpression of MeCP2 in control iPSCs leads to an increase in glutamatergic synapse numbers. These results are in accordance with observations from mouse models in which the loss or overexpression of MeCP2 leads to a respective decrease or increase of glutamatergic synapses (Chao et al.,
2007). In further agreement with research performed with mouse cells (Chen et al., 2001; Guy et al., 2001) and on human autopsies, neurons derived from Rett iPSCs are smaller in size and have fewer dendritic spines when compared to control iPSC and ESC neurons. Importantly, neurons derived from Rett iPSCs are also functionally impaired when compared to neurons derived from control iPSCs. Specifically, neurons derived from Rett iPSCs exhibit a reduction in the transient rise of intracellular calcium levels typical of active synapses. They also show a decrease in the frequency and amplitude of spontaneous excitatory and inhibitory postsynaptic currents when compared to control cells. Together, these findings provide compelling evidence that molecular and functional defects found in Rett syndrome patients can be recapitulated in iPSCderived neurons. Given the reversibility of the Rett phenotype in mouse, Marchetto et al. (2010) then ask whether they could rescue their in vitro phenotype using candidate drugs. First, they examine the effects of insulin-like growth factor 1 (IGF1), which had previously been shown to partially rescue Rett symptoms in Mecp2-deficient mice (Tropea et al., 2009). Indeed, treating neurons derived from human Rett iPSCs with this growth factor increases the number of glutamatergic synapses. Because the majority of MeCP2 mutations create premature stop codons in the gene (i.e., nonsense mutations), the authors also test the effect of the drug
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 499
Figure 1. Using iPSCs to Model Rett Syndrome In Vitro Mutations in the methyl-CpG binding protein (MeCP2) gene cause the neurodevelopmental disorder Rett syndrome. Marchetto et al. (2010) isolate fibroblasts from Rett patients with MeCP2 mutations. They then reprogram these cells into induced pluripotent stem cells (iPSCs) by the exogenous expression of the transcription factors Oct4, Sox2, Klf4, and c-Myc. These Rett iPSCs can then differentiate into neurons in vitro, recapitulating several of the defects found in Rett syndrome patients and in animal models of the disease. These defects can be partially reversed by candidate drugs, suggesting that this disease model will facilitate large-scale drug screens and future mechanistic studies of Rett syndrome.
gentamicin, which facilitates ribosomal readthrough of stop codons. Following treatment with low doses of gentamicin, Rett neurons harboring nonsense mutations in the MeCP2 gene express elevated levels of MeCP2 protein and display a striking increase of glutamatergic synapses, reaching levels similar to those seen in control neurons. Whether IGF1 or gentamicin treatment leads to a functional recovery of neurons, however, remains unexplored in this study. In summary, these results show that previously identified drugs are effective on neurons derived from Rett iPSCs and thus validate the use of large-scale drug screens to identify new and more effective compounds that could ameliorate the symptoms of Rett. Given the recent recognition that mouse iPSCs can exhibit molecular and functional differences compared with mouse ESCs (Stadtfeld and Hochedlinger, 2010), it will be important to ensure
that any cell culture disease model is an accurate representation of its in vivo counterpart rather than an artifact of the reprogramming procedure. In that regard, data presented by Marchetto and colleagues for their Rett syndrome model agree well with what is known from mouse models of Rett syndrome and postmortem analyses in humans. However, the observation by the authors that differentiated neurons derived from Rett iPSCs exhibit a severe skewing of X chromosome inactivation—meaning that one copy of the X chromosome is more likely to undergo inactivation than the other— remains unexplained. This observation is of particular relevance because MeCP2 is an X-linked gene and patients’ cells in vivo are mosaic for the MeCP2 mutation due to random X chromosome inactivation. One possible explanation for the skewing of X inactivation in vitro is that the previously inactive X chromosome in Rett fibroblasts does not fully reactivate
500 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
in iPSCs and remains inactive in iPSCderived neurons. Indeed, a recent study on skewed X chromosome inactivation in human iPSCs by Kathrin Plath’s lab is consistent with this interpretation (Tchieu et al., 2010). If the skewing reflects incomplete reactivation, then the studies performed by Marchetto and colleagues likely relied exclusively on iPSC clones derived from Rett fibroblasts that had inactivated the wild-type MeCP2 allele. Thus, to harness the power of cell culture disease models, future work will need to further explore the differences between iPSCs and ESCs. The ability to produce disease-specific differentiated cells is one of the major promises of iPSC technology because it holds the potential for disease modeling, drug development, and ultimately cell therapy. The model of Rett syndrome presented by Marchetto and colleagues is not only an additional proof of principle that human iPSCs may be useful in drug development but also a promising opportunity to gain insights into the pathology of Rett syndrome in live human neurons. ACKNOWLEDGMENTS K.H. is an advisor for iPierian, Inc. REFERENCES Amir, R.E., Van den Veyver, I.B., Wan, M., Tran, C.Q., Francke, U., and Zoghbi, H.Y. (1999). Nat. Genet. 23, 185–188. Chen, R.Z., Akbarian, S., Tudor, M., and Jaenisch, R. (2001). Nat. Genet. 27, 327–331. Chao, H., Zoghbi, H.Y., and Rosenmund, C. (2007). Neuron 56, 58–65. Guy, J., Hendrich, B., Holmes, M., Martin, J.E., and Bird, A. (2001). Nat. Genet. 27, 322–326. Guy, J., Gan, J., Selfridge, J., Cobb, S., and Bird, A. (2007). Science 315, 1143–1147. Marchetto, M.C.N., Carromeu, C., Acab, A., Yu, D., Yeo, G., Yangling, M., Chen, G., Gage, F.H., and Muotri, A.R. (2010). Cell 143, this issue, 527–539. Percy, A.K., and Lane, J.B. (2005). J. Child Neurol. 20, 718–721. Stadtfeld, M., and Hochedlinger, K. (2010). Genes Dev. 24, 2239–2263. Tchieu, J., Kuoy, E., Chin, M.H., Trinh, H., Patterson, M., Sherman, S.P., Aimiuwu, O., Lindgren, A., Hakimian, S., Zack, J.A., et al. (2010). Cell Stem Cell 7, 329–342. Tropea, D., Giacometti, E., Wilson, N.R., Beard, C., McCurry, C., Fu, D.D., Flannery, R., Jaenisch, R., and Sur, M. (2009). Proc. Natl. Acad. Sci. USA 106, 2029–2034.
Leading Edge
Previews Translation by Remote Control Pascal Preker1,* and Torben Heick Jensen1,* 1Centre for mRNP Biogenesis and Metabolism, Department of Molecular Biology, C.F. Møllers Alle ´ 3, Building 1130, Aarhus University, 8000 Aarhus, Denmark *Correspondence:
[email protected] (P.P.),
[email protected] (T.H.J.) DOI 10.1016/j.cell.2010.10.039
Efficient and accurate gene expression requires the coordination of multiple steps along the pathway of mRNA and protein synthesis. Now, Harel-Sharvit et al. (2010) show that transcriptional imprinting of mRNAs with two subunits of RNA polymerase II, Rbp4p and Rpb7p, guides transcripts to the translation apparatus. A defining feature of eukaryotic cells is compartmentalization. Whereas transcription of DNA into mRNA takes place in the nucleus, translation of the mRNA is physically separate and occurs in the cytoplasm, where ultimately the transcript is also degraded. In prokaryotes, transcription and translation are coupled in the protoplasm so that the translation machinery can directly engage the nascent mRNA. In this issue of Cell, Harel-Sharvit et al. (2010) provide surprising evidence that transcription and translation can also be coupled in eukaryotes—at least in the unicellular yeast Saccharomyces cerevisiae. The authors refer to this process as the ‘‘remote controlling’’ of translation by the transcription apparatus. Key to this coupling is a heterodimer composed of the Rpb4p and Rpb7p proteins (for review see, Choder, 2004; Sampath and Sadhale, 2005), a fraction of which forms a stalklike and highly conserved protrusion of the decameric core of RNA polymerase II (RNAPII), the enzyme responsible for the production of mRNAs and many noncoding RNAs (Figure 1, inset). This places the two proteins at a strategic position near the exit channel of the nascent RNA (Brueckner et al., 2009), and the heterodimer has been found to associate both in vivo and in vitro with mRNA in a transcription-dependent manner (GolerBaron et al., 2008; Ujva´ri and Luse, 2006). Rpb4/7p is only loosely associated with the RNAPII core, and the degree of its association varies with physiological conditions in yeast (Choder, 2004; Sampath and Sadhale, 2005). In addition, the two proteins are in vast excess over other RNAPII subunits and continuously shuttle between the nucleus and the cytoplasm
(Figure 1). Indeed, prior work has revealed a cytoplasmic function for Rpb4/7p in mRNA decay (for example, Goler-Baron et al., 2008). Presumably, the proteins influence the reversible transition of mRNA between the active translation machinery in the form of polysomes and the so-called processing (P) bodies, which are believed to be storage sites for RNAs and their degradation factors. The present study now demonstrates that Rpb4/7p also functions in translation, thus making it a ‘‘coordinator’’ of all major stages of gene expression. Using a variety of methods, HarelSharvit et al. first show that Rpb4/7p interacts, independent of RNA, with two subunits (Nip1p and Hcr1p) of the hexameric translation initiation factor eIF3, which serves as a platform for the assembly of the translation-initiation complex (Sonenberg and Hinnebusch, 2009). Importantly, eIF3 does not interact with other RNAPII subunits, implying that it contacts the soluble pool of Rpb4/7p. Prompted by these observations, the authors go on to show that Rpb4/7p is required for efficient translation initiation by analyzing a deletion mutant of the nonessential RPB4 gene (rpb4D) and a conditional mutant allele of the essential RPB7 gene (rpb7-26). Importantly, the latter mutation does not markedly affect transcription or mRNA degradation rates (Goler-Baron et al., 2008), thus minimizing the possibility of indirect effects. They find that the RPB4 and RPB7 mutant strains are hypersensitive to translation inhibitors and observe genetic interactions with regulators of translation. In addition, the two mutants exhibit reduced protein synthesis accompanied by a reduction in polysome content, as
well as loss of MFA2 and HYP2 mRNA from polysomes. They also elicit decreased disassembly of P bodies and slower movement of MFA2 mRNA from P bodies into polysomes. This last finding is important because mRNAs can also leave P bodies and (re)associate with polysomes (Figure 1; Parker and Sheth, 2007). Thus, in addition to directly docking mRNAs to the translation apparatus through eIF3, Rpb4/7p might facilitate translation initiation indirectly by stimulating P body disassembly and/or the release of mRNA from P bodies, thereby increasing the pool of translatable mRNA. Perhaps the most significant result of this work is the finding that mutations in two RNAPII subunits (Rpb1p and Rpb6p) at the interface with Rpb7p, which have been previously shown to compromise the recruitment of the Rpb4/7p subcomplex to the catalytic core, phenocopy the defects displayed by the RPB4 and RPB7 mutant strains. Similar results are obtained using a mutation of Rpb4p that fails to enter the nucleus. Taken together, these findings suggest a model whereby nucleocytoplasmic shuttling and recruitment of Rpb4/7p to the nascent transcript by the core RNAPII is required for Rpb4/ 7p’s cytoplasmic functions (Figure 1). The physical separation of transcription and translation has allowed the spread of intervening sequences (introns) in the eukaryotic lineage, which in turn is thought to have fueled genomic evolution by aiding in the creation of new protein domains (Schmidt and Davies, 2007). It now appears that eukaryotes, despite this segregation of the two main activities of the central dogma in molecular biology, have maintained a mechanism that allows transcription and translation to
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 501
Finally, the extent to which communicate through direct coupling of transcription and physical interactions. Apart translation by Rpb4/7p is from simply making translaconserved in higher eukarytion more streamlined and otes remains to be seen. The efficient, what could be the high degree of conservation functional implication(s) of of the heterodimer and its such a coupling? Earlier findinteraction partners make it ings suggested that, in yeast plausible that whatever the only, 20% of RNAPII moleprecise mechanism, it may cules contain Rpb4/7p under have coevolved—all the way optimal growth conditions, to humans. whereas, during stationary phase, all subunits are stoichiometric (Choder, 2004). REFERENCES Even though genome-wide Brueckner, F., Armache, K.J., analysis of Rpb7p occupancy Cheung, A., Damsma, G.E., Kettenhas contradicted these reberger, H., Lehmann, E., Sydow, J., sults (Jasiak et al., 2008), it and Cramer, P. (2009). Acta Crystalwould appear that any differlogr. D Biol. Crystallogr. 65, 112– ence in the stoichiometry of 120. Rpb4/7p could have bearing Choder, M. (2004). Trends Bioon the translational efficiency chem. Sci. 29, 674–681. of certain mRNAs under Goler-Baron, V., Selitrennik, M., different environmental conBarkai, O., Haimovich, G., Lotan, ditions. This effect could be R., and Choder, M. (2008). Genes further fine-tuned by the Dev. 22, 2022–2027. Figure 1. Rpb4/Rpb7p in Transcription, Translation, and Degradaimpact of Rpb4/7p on degraHarel-Sharvit, L., Eldad, N., Haimotion dation (Figure 1; Goler-Baron vich, G., Barkai, O., Dueck, L., and The inset (bottom left) represents the relevant interactions of RNA polymerase et al., 2008). The identificaChoder, M. (2010). Cell 143, this II (RNAPII) subunits: Rpb7p forms a heterodimer with Rpb4p and interacts with tion of the subsets of genes issue, 552–563. the catalytic core through Rpb1p and Rpb6p. The remaining eight subunits (gray) are placed arbitrarily. In the nucleus, Rpb4/7p is loaded cotranscriptionassociated with Rpb4/7p Jasiak, A.J., Hartmann, H., Karakaally onto the nascent RNA. The mature mRNA carrying a 50 cap and a 30 poly(A) under various conditions and sili, E., Kalocsay, M., Flatley, A., tail is exported into the cytoplasm in complex with Rpb4/7p, which targets the Kremmer, E., Stra¨sser, K., Martin, an understanding of the mRNA to the general translation initiation factor eIF3. The latter provides a scafD.E., So¨ding, J., and Cramer, P. fold for the assembly of other initiation factors, which together recruit ribomechanistic details that regu(2008). J. Biol. Chem. 283, 26423– somes. Translation initiation is further facilitated by interaction between the late the RNAPII-Rpb4/7p ininitiation complex and the poly(A) tail (not shown). Rpb4/7p and mRNA can 26427. teraction would be required also associate with processing (P) bodies or can return to polysomes. MutaParker, R., and Sheth, U. (2007). to foster this attractive hypotions in the RPB4 and RPB7 genes both adversely affect the assembly of Mol. Cell 25, 635–646. mRNA into polysomes and its transition between P bodies and polysomes. thesis. Transcription-coupled In addition, the mRNA can also be destined for degradation at various stages. Sampath, V., and Sadhale, P. imprinting of mRNAs with Re-import into the nucleus completes the ‘‘shuttling’’ of Rpb4/7p. (2005). IUBMB Life 57, 93–102. the Rpb4/7p subcomplex of Schmidt, E.E., and Davies, C.J. RNAPII in the nucleus could (2007). Bioessays 29, 262–270. potentially also serve as Paralogs of Rpb4/7p can also be found in a form of quality control, possibly by Sonenberg, N., and Hinnebusch, A.G. (2009). Cell ensuring that only mRNAs that have the RNA polymerases I and III, raising the 136, 731–745. been properly terminated and loaded possibility that they might utilize analogous Ujva´ri, A., and Luse, D.S. (2006). Nat. Struct. Mol. mechanisms to control RNA metabolism. Biol. 13, 49–54. with Rpb4/7p are translated.
502 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
Leading Edge
Previews Shining a Light on Germinal Center B Cells Jason G. Cyster1,* 1Howard Hughes Medical Institute and Department of Microbiology and Immunology, University of California, San Francisco, CA 94143, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.036
The mechanisms of B cell selection in lymphoid tissues are poorly understood. In this issue, Victora et al. (2010) use imaging of photoactivatable green fluorescent protein to define the movements of B cells in germinal centers and provide evidence that antibody affinity maturation is driven by competition for T cell help. Since the discovery in the 1960s that antibodies undergo affinity maturation as the immune response progresses, immunologists have been fascinated by the underlying mechanism, which is still only partially understood. Victora et al. now reveal the multi-day dynamics of B cell migration at sites of antibody affinity maturation and use this insight to further define the selection mechanism. Their efforts also mark a major technical advance for the study of cell behavior in a physiological context. Germinal centers of lymph nodes are a site of antibody affinity maturation. They are seeded by small numbers of rapidly dividing antigen-responding B cells and form over a period of days into structures of 10,000 cells (MacLennan, 1994). Germinal centers contain a dark zone and a light zone (named based onhistological staining); the former contains mostly B cells, whereas the latter also contains a small number of helper T cells and follicular dendritic cells. An early model posited that the dark zone is the site of B cell proliferation and somatic hypermutation of the immunoglobulin V (variable) gene, and that the light zone is where the newly mutated B cell receptors (BCRs) compete for antigen binding. B cell clones that have a strong enough BCR signal would then receive T cell help and survive. It was further proposed that germinal center B cells move cyclically between the dark and light zones to undergo repeated rounds of mutation and selection, with successful selection eventually giving rise to plasma cells and memory B cells (MacLennan, 1994).
Prior imaging studies established that germinal center B cells migrate in a continuous manner over the processes of antigen-bearing follicular dendritic cells (Allen et al., 2007b; Hauser et al., 2007; Schwickert et al., 2007). Small numbers of B cells have been observed moving bidirectionally between the light and dark zones, though in these cases the 30–60 min imaging windows were insufficient to determine whether there is a directional bias (Allen et al., 2007b; Hauser et al., 2007; Schwickert et al., 2007). Imaging cell interaction dynamics further reveals that only a fraction of the germinal center T cells are in stable contact with B cells, despite most B and T cells being specific for the same antigen (Allen et al., 2007b). These observations led to a revised model of affinity maturation in which B cells are selected not only based on receipt of a BCR signal, but also on their ability to compete for T cell help (Allen et al., 2007a). In the current work, Victora et al. (2010) obtain gene expression data for cells from the light and dark zones. To do this they photoactivate GFP in specific regions of the germinal centers using two-photon microscopy and then immediately isolate the cells by flow cytometry (Figure 1A). Although cells from the two regions are similar in their overall pattern of gene expression, consistent with previous studies (reviewed in Allen et al., 2007a), dark zone cells exhibit higher expression of genes related to mitosis, whereas light zone cells have higher expression of activation markers, cell-surface molecules, and apoptosis regulators. The small magnitude of these gene expression differ-
ences might reflect heterogeneity within the cell populations. Indeed, previous findings indicate that a minority of light zone cells exhibit plasma cell properties and nuclear localization of the transcription factor NF-kB (Allen et al., 2007a; Basso et al., 2004). Victora et al. take advantage of differences in the expression of cell-surface markers (specifically CD83, CD86, and the chemokine receptor CXCR4) to convincingly identify dark and light zone cells by flow cytometry. The authors find that both zones have cells in S phase, in agreement with other studies (Allen et al., 2007a), whereas almost all G2/M phase cells are restricted to the dark zone. Although this is consistent with long-held views, other studies have identified occasional cells undergoing mitosis in the light zone (Allen et al., 2007a). It is notable that the immunization strategy employed by Victora et al. to induce germinal center formation avoids use of adjuvant. Perhaps the level of innate immune stimuli accounts for these potential differences in the behavior of light zone cells. A major achievement of the photoactivation analysis by Victora et al. is the tracking of cell population behavior over many hours. They show that dark zone cells move to the light zone at a rate that replaces much of this compartment in 6 hr; movement from the light zone to the dark zone also occurs but is less prominent. These compelling data provide quantitative insight into the net flux of cells between zones and suggest that movement from the light zone to the dark zone is possible only for a selected subset of cells.
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 503
Figure 1. Germinal Center Dynamics (A) Transgenic B cells expressing photoactivatible-green fluorescent protein (GFP) are transferred to wild-type hosts in which germinal centers are subsequently induced by immunization. GFP is selectively photoactivated in the dark zone (left) or light zone (right) by two-photon laser scanning microscopy. The entire lymph node is then converted to a cell suspension, and GFP-expressing cells are isolated by flow cytometry. Alternatively, the lymph node is examined at 1 hr intervals to record the movements of photoactivated B cells. (B) Increased peptide loading onto major histocompatibility complex II (MHC II) is sufficient to give B cells a competitive advantage. A mouse containing a mixture of wild-type and DEC205-deficient B cells is immunized with antigen. After the germinal center has formed, the mouse is treated with DEC205 antibody coupled to antigen. The diagram shows a wild-type and DEC205-deficient B cell binding the same amount of antigen in the light zone via their B cell receptors (BCRs), but the wild-type cell also acquires antigen via DEC205, thus taking up more antigen and generating a greater number of MHC II peptide complexes compared to the DEC205-deficient cell. As a result it receives a greater level of survival and proliferation signals from helper T cells.
To test the model that germinal center B cells compete for T cell help, the authors use an approach to modulate the amount of peptide bound-major histocompatibiltiy complex II (MHC II) on B cells. In previous work, Nussenzweig, Steinman, and coworkers demonstrated that DEC205 antibody-antigen conjugates efficiently target antigens to dendritic cells for presentation to T cells. Conveniently, Victora et al. find that germinal center B cells also express DEC205, and they show that the same antigen-loading ‘‘trick’’ works to increase peptide-MHC II abundance within 6 hr on wild-type germinal center B cells (Victora et al., 2010). When immunized mice harboring a mixture of wild-type and DEC205-deficient B cells are treated with DEC205 antibody coupled to the relevant T cell antigen, there is an early accumulation of DEC205-positive B cells in the light zone (at 12 hr) followed by a remarkable enrichment in the dark
zone by 36–48 hr. The authors note that the DEC205-negative majority is counterselected by focusing T cell help on a small subset of DEC205-positive cells, concluding that T cell help is limiting for clonal expansion (Figure 1B). Although the cell frequency data are clear, the conclusion that cells with high amounts of peptide-MHC II complexes displace cells with less will be made still stronger when effects on absolute cell number are determined. Strikingly, the cells loaded with antigen via DEC205 fail to undergo affinity maturation, suggesting that when cells are equivalent in their ability to receive T cell help, differential BCR engagement is not sufficient to mediate affinity-based selection (Victora et al., 2010). The new findings, combined with the large body of prior information, support a model in which light zone cells gather antigen via their BCRs as they move over the network of follicular dendritic
504 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
cells, with cells that have higher affinity acquiring greater amounts of antigen per unit time. Higher antigen uptake via the BCR then leads to greater presentation of peptide-MHC II complexes (Batista and Neuberger, 1998). T cells are thought to reorient toward cells with the highest peptide-MHC II levels (Depoil et al., 2005), forming stable conjugates with the cells that have captured and processed the most antigen, and delivering CD40 ligand and cytokine signals essential for selection (Figure 1B). By turning on a light in germinal center B cells, Victora and colleagues open a new chapter in the study of antibody affinity maturation. Of course, many questions remain. Do B cells with high affinity display greater amounts of peptideMHC II complexes than low-affinity cells? How do T cells behave as they encounter B cells displaying different amounts of peptide-MHC II complexes? And how do T cells assist in the decision of a B cell
to differentiate into an antibody-secreting cell or return to the dark zone? In addition to providing a platform for addressing these questions, the powerful application of photoactivatable-GFP in this study should spur broader use of this technology to relate cell location with gene expression and cell fate.
REFERENCES Allen, C.D., Okada, T., and Cyster, J.G. (2007a). Immunity 27, 190–202.
Hauser, A.E., Junt, T., Mempel, T.R., Sneddon, M.W., Kleinstein, S.H., Henrickson, S.E., von Andrian, U.H., Shlomchik, M.J., and Haberman, A.M. (2007). Immunity 26, 655–667.
Allen, C.D., Okada, T., Tang, H.L., and Cyster, J.G. (2007b). Science 315, 528–531.
MacLennan, I.C.M. (1994). Annu. Rev. Immunol. 12, 117–139.
Basso, K., Klein, U., Niu, H., Stolovitzky, G.A., Tu, Y., Califano, A., Cattoretti, G., and Dalla-Favera, R. (2004). Blood 104, 4088–4096.
Schwickert, T.A., Lindquist, R.L., Shakhar, G., Livshits, G., Skokos, D., Kosco-Vilbois, M.H., Dustin, M.L., and Nussenzweig, M.C. (2007). Nature 446, 83–87.
ACKNOWLEDGMENTS
Batista, F.D., and Neuberger, M.S. (1998). Immunity 8, 751–759.
The author thanks C. Allen and O. Bannard for helpful discussions.
Depoil, D., Zaru, R., Guiraud, M., Chauveau, A., Harriague, J., Bismuth, G., Utzny, C., Muller, S., and Valitutti, S. (2005). Immunity 22, 185–194.
Victora, G.D., Schwickert, T.A., Fooksman, D.R., Kamphorst, A.O., Meyer-Hermann, M., Dustin, M.L., and Nussenzweig, M.C. (2010). Cell 143, this issue, 592–605.
A Straightjacket for Pain? Simon Beggs1 and Michael W. Salter1,* 1Program in Neurosciences and Mental Health, Hospital for Sick Children, 555 University Avenue, Toronto, ON M5G 1X8, Canada *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.038
Perception of pain involves both the peripheral and central nervous systems. Starting with a wholegenome RNA interference screen in Drosophila, Neely et al. (2010) identify a mammalian gene that is required not only for efficient transfer of pain signals between brain centers, but also for the suppression of inappropriate signaling between other sensory systems. Even though humanity’s attempts to control pain can be traced back through the millennia, much of our current arsenal of therapies still has its roots in folk remedies. Originally derived from natural plant products such as willow bark and poppies, these medicines have been the mainstay of chronic pain control for many years. Unfortunately, even modern therapies for chronic pain are too often ineffective, leaving untold numbers of individuals suffering worldwide. As we move into the postgenomic era, hopes have been raised that the problem of pain will be solved, largely by getting to grips with the underlying genetic, molecular, and cellular mechanisms involved in nociception. It is also expected that interindividual variability in pain experience, a major issue confounding effective treatment, will be accounted for in some part by variations in a definable subset of genes. The report by Neely et al. (2010) in this issue of Cell represents a major advance
in the search for the genes and mechanisms of pain and variability in its perception. In an impressively multifaceted study that extends from fruit flies to mice and ultimately to humans, the authors have uncovered a new molecular player, a2d3, in pain. Along the way, they also made some highly unexpected observations indicating that a2d3 is not only involved in pain, but also plays a role in the poorly understood mechanisms that separate one form of sensory information from another. The fruit fly provides an ideal model for unraveling genetic aspects of human disease, with advantages including a short generation time and the capacity to analyze extremely large numbers of animals in a high-throughput manner. Genome-wide RNA interference (RNAi) screens are a powerful investigative tool given the fly’s relatively small (15,000 genes) but extremely well-annotated genome and sufficient homology with
the human genome to permit the identification of conserved pathways in flies and humans. In a Herculean undertaking, Neely et al. take advantage of the Vienna global Drosophila RNAi library (Dietzl et al., 2007) to individually knock down almost every gene in the fly genome, specifically in neural tissues. Each line was then painstakingly tested using a behavioral assay to measure sensitivity to noxious heat. In total, 580 genes are identified that potentially regulate thermal nociception and possibly basic neuronal function. One of the genes identified is straightjacket, which is one of several a2d genes that encode calcium channel subunits. The mammalian homolog of straightjacket is CACNAD2D3, which encodes the a2d3 subunit. Of interest, a close relative of this protein, a2d1, is the molecular target for the analgesics gabapentin and pregabalin (Field et al., 2006). To further investigate the function of a2d3, the authors
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 505
to differentiate into an antibody-secreting cell or return to the dark zone? In addition to providing a platform for addressing these questions, the powerful application of photoactivatable-GFP in this study should spur broader use of this technology to relate cell location with gene expression and cell fate.
REFERENCES Allen, C.D., Okada, T., and Cyster, J.G. (2007a). Immunity 27, 190–202.
Hauser, A.E., Junt, T., Mempel, T.R., Sneddon, M.W., Kleinstein, S.H., Henrickson, S.E., von Andrian, U.H., Shlomchik, M.J., and Haberman, A.M. (2007). Immunity 26, 655–667.
Allen, C.D., Okada, T., Tang, H.L., and Cyster, J.G. (2007b). Science 315, 528–531.
MacLennan, I.C.M. (1994). Annu. Rev. Immunol. 12, 117–139.
Basso, K., Klein, U., Niu, H., Stolovitzky, G.A., Tu, Y., Califano, A., Cattoretti, G., and Dalla-Favera, R. (2004). Blood 104, 4088–4096.
Schwickert, T.A., Lindquist, R.L., Shakhar, G., Livshits, G., Skokos, D., Kosco-Vilbois, M.H., Dustin, M.L., and Nussenzweig, M.C. (2007). Nature 446, 83–87.
ACKNOWLEDGMENTS
Batista, F.D., and Neuberger, M.S. (1998). Immunity 8, 751–759.
The author thanks C. Allen and O. Bannard for helpful discussions.
Depoil, D., Zaru, R., Guiraud, M., Chauveau, A., Harriague, J., Bismuth, G., Utzny, C., Muller, S., and Valitutti, S. (2005). Immunity 22, 185–194.
Victora, G.D., Schwickert, T.A., Fooksman, D.R., Kamphorst, A.O., Meyer-Hermann, M., Dustin, M.L., and Nussenzweig, M.C. (2010). Cell 143, this issue, 592–605.
A Straightjacket for Pain? Simon Beggs1 and Michael W. Salter1,* 1Program in Neurosciences and Mental Health, Hospital for Sick Children, 555 University Avenue, Toronto, ON M5G 1X8, Canada *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.038
Perception of pain involves both the peripheral and central nervous systems. Starting with a wholegenome RNA interference screen in Drosophila, Neely et al. (2010) identify a mammalian gene that is required not only for efficient transfer of pain signals between brain centers, but also for the suppression of inappropriate signaling between other sensory systems. Even though humanity’s attempts to control pain can be traced back through the millennia, much of our current arsenal of therapies still has its roots in folk remedies. Originally derived from natural plant products such as willow bark and poppies, these medicines have been the mainstay of chronic pain control for many years. Unfortunately, even modern therapies for chronic pain are too often ineffective, leaving untold numbers of individuals suffering worldwide. As we move into the postgenomic era, hopes have been raised that the problem of pain will be solved, largely by getting to grips with the underlying genetic, molecular, and cellular mechanisms involved in nociception. It is also expected that interindividual variability in pain experience, a major issue confounding effective treatment, will be accounted for in some part by variations in a definable subset of genes. The report by Neely et al. (2010) in this issue of Cell represents a major advance
in the search for the genes and mechanisms of pain and variability in its perception. In an impressively multifaceted study that extends from fruit flies to mice and ultimately to humans, the authors have uncovered a new molecular player, a2d3, in pain. Along the way, they also made some highly unexpected observations indicating that a2d3 is not only involved in pain, but also plays a role in the poorly understood mechanisms that separate one form of sensory information from another. The fruit fly provides an ideal model for unraveling genetic aspects of human disease, with advantages including a short generation time and the capacity to analyze extremely large numbers of animals in a high-throughput manner. Genome-wide RNA interference (RNAi) screens are a powerful investigative tool given the fly’s relatively small (15,000 genes) but extremely well-annotated genome and sufficient homology with
the human genome to permit the identification of conserved pathways in flies and humans. In a Herculean undertaking, Neely et al. take advantage of the Vienna global Drosophila RNAi library (Dietzl et al., 2007) to individually knock down almost every gene in the fly genome, specifically in neural tissues. Each line was then painstakingly tested using a behavioral assay to measure sensitivity to noxious heat. In total, 580 genes are identified that potentially regulate thermal nociception and possibly basic neuronal function. One of the genes identified is straightjacket, which is one of several a2d genes that encode calcium channel subunits. The mammalian homolog of straightjacket is CACNAD2D3, which encodes the a2d3 subunit. Of interest, a close relative of this protein, a2d1, is the molecular target for the analgesics gabapentin and pregabalin (Field et al., 2006). To further investigate the function of a2d3, the authors
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 505
generated a knockout MRI signal in the mouse. Intriguingly, its primary cortical area (the expression in the mouse is barrel fields of the somatoentirely restricted to the brain, sensory cortex) remain being completely absent from unchanged in mutant the spinal cord and dorsal compared to wild-type mice, root ganglia. This suggests but cross-activation to other that, unlike the function of sensory brain regions is also Straightjacket in flies, observed. It therefore mammalian a2d3 likely mediappears that mice lacking ates the last leg of the nocia2d3 develop a form of ceptive relay, the transmissynesthesia, a condition sion of the signal from the whereby stimulation of one thalamus up to the highest sensory pathway elicits level, the cortex. perception in one or more To test the relevance of other sensory modalities. their findings in humans, Although transmission of Neely et al. identify and thermal nociception is diminscreen four single-nucleotide ished, other sensory systems Figure 1. Sounds, Sights, and Smells of Pain polymorphisms (SNPs) in the tested, such as touch, remain In the mouse brain, nociceptive signals are relayed from the thalamus to the human CACNA2D3 gene. unaffected. somatosensory and motor regions of the cortex. Visualized by functional One of these, rs6777055, is The study by Neely et al. magnetic resonance imaging (fMRI), mice lacking the gene encoding a2d3 display a significant reduction in activation of these regions, consistent with significantly associated with offers a compelling insight reduced perception of a painful thermal stimulus. Surprisingly, however, other reduced thermal sensitivity. into nociceptive transmission regions of the cortex, including those involved in olfaction, vision, and hearing, In these experiments, the in higher areas of the central are activated by the same stimulus. volunteers are given a series nervous system. In doing so, of short pulses of painful heat. it also raises many questions Although the temperature of the pulses mally activates a characteristic set of for future investigation. On a molecular remains the same, most people tend to brain structures associated with pain per- level, a2d3 is known to regulate calcium perceive the stimulus as becoming ception. Known as the pain matrix, this channel expression and its targeting to progressively more painful. In a small center includes the thalamus and somato- presynaptic sites (Dolphin, 2009). A number of subjects, however, this does sensory cortex. In the a2d3 mutant recent report suggests that a2d3 can be not occur, and it is these individuals that mouse, activation of the pain matrix is GPI anchored to the membrane and that are found to express the rs6777055 minor markedly diminished, with the fMRI signal this modification is necessary for a2d3seemingly unable to spread from the thal- mediated enhancement of calcium allele of the CACNA2D3 gene. Surprising similarities emerge between amus to higher cortical regions. There- currents (Davies et al., 2010; Bauer the a2d3 knockout mouse and humans fore, it appears that, in mammals, a2d3 et al., 2010). Evidence is also now with SNPs that interfere with a2d3 func- is necessary not for the detection of emerging that, in Drosophila, a2d3 is tion. For both, it appears that acute pain noxious heat but for the transmission of required for synaptic morphogenesis sensitivity is reduced (although spinal thermal nociceptive signals to higher (Kurshan et al., 2009). Of interest, howreflex-mediated events such as tail flick areas of the central nervous system, in ever, this function appears to be indepenremain normal, in line with the absence particular those associated with the pro- dent of its role in regulating voltageof a2d3 expression in peripheral and cessing of pain signals. gated calcium channels. In light of However, this is not the only finding of this multifunctionality, the question arises spinal sensory circuitry in the mouse). Furthermore, both thermal hypersensi- the fMRI study. When looking at total acti- as to what exactly is deficient in the tivity following inflammation in mice and vation in the brain, the authors find no a2d3-deficient mice. In particular, the perception of chronic back pain in difference between control and a2d3- synaptic consequences of eliminating humans appear to involve a2d3; patients deficient mice. This at first seems coun- a2d3 throughout mouse development expressing the rs6777055 SNP report terintuitive, given the large reduction in remain unknown. reduced pain 1 year after corrective fMRI signal from thalamocortical projecGiven the close relationship between tions in the mutant mouse. Upon closer a2d3 and a2d1, it would also be of great surgery compared to non-SNP patients. In addition to the remarkable insight investigation, it becomes evident that interest to test what happens to a2d3 into pain processing that these studies other entirely unrelated brain regions are expression after peripheral nerve injury, reveal, a fascinating and unexpected instead being activated, including visual, as a2d1 is upregulated in the somatosenphenomenon emerges from their studies auditory, and olfactory regions (Figure 1). sory system after nerve injury, concurrent using functional magnetic resonance Even more remarkably, when another with the onset of behavioral pain hyperimaging (fMRI) of the a2d3-deficient sensory modality, tactile stimulation of sensitivity. Furthermore, although the mice. Noxious thermal stimulation nor- the vibrissae, is tested, not only does the human studies presented indicate that 506 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
chronic pain states may indeed be mediated by the action of a2d3, it remains to be seen whether this can be modeled in mice. Finally, the cross-sensory activation in a2d3-deficient mice raises the question as to whether synesthesia in humans is in any way related to a2d3 polymorphisms. Despite considerable research and major mechanistic breakthroughs in recent years, pain remains the most common cause of disability and impaired quality of life and, as such, represents an enormous socioeconomic problem. The identification of a2d3 as a player in the mechanisms of pain and a molecule accounting for a portion of interindividual
pain variability opens up new possibilities for rationally designed and individualized pain therapies. The biggest challenge now is to find safe and effective ways to use this discovery to truly put a straightjacket on pain.
Dietzl, G., Chen, D., Schnorrer, F., Su, K.C., Barinova, Y., Fellner, M., Gasser, B., Kinsey, K., Oppel, S., Scheiblauer, S., et al. (2007). Nature 448, 151–156. Dolphin, A.C. (2009). Curr. Opin. Neurobiol. 19, 237–244.
REFERENCES
Field, M.J., Cox, P.J., Stott, E., Melrose, H., Offord, J., Su, T.Z., Bramwell, S., Corradini, L., England, S., Winks, J., et al. (2006). Proc. Natl. Acad. Sci. USA 103, 17537–17542.
Bauer, C.S., Tran-Van-Minh, A., Kadurin, I., and Dolphin, A.C. (2010). Curr. Opin. Neurobiol. 20, 563–571.
Kurshan, P.T., Oztan, A., and Schwarz, T.L. (2009). Nat. Neurosci. 12, 1415–1423.
Davies, A., Kadurin, I., Alvarez-Laviada, A., Douglas, L., Nieto-Rostro, M., Bauer, C.S., Pratt, W.S., and Dolphin, A.C. (2010). Proc. Natl. Acad. Sci. USA 107, 1654–1659.
Neely, G.G., Hess, A., Costigan, M., Keene, A.C., Goulas, S., Langeslag, M., Griffin, R.S., Belfer, I., Dai, F., Smith, S., et al. (2010). Cell 143, this issue, 628–638.
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 507
Leading Edge
Review Pluripotency and Cellular Reprogramming: Facts, Hypotheses, Unresolved Issues Jacob H. Hanna,1,3,* Krishanu Saha,1 and Rudolf Jaenisch1,2,* 1The
Whitehead Institute for Biomedical Research of Biology Massachusetts Institute of Technology, Cambridge, MA 02142, USA 3Present address: Department of Molecular Genetics, Weizmann Institute of Science, Rehovot 76100, Israel *Correspondence:
[email protected] (J.H.H.),
[email protected] (R.J.) DOI 10.1016/j.cell.2010.10.008 2Department
Direct reprogramming of somatic cells to induced pluripotent stem cells by ectopic expression of defined transcription factors has raised fundamental questions regarding the epigenetic stability of the differentiated cell state. In addition, evidence has accumulated that distinct states of pluripotency can interconvert through the modulation of both cell-intrinsic and exogenous factors. To fully realize the potential of in vitro reprogrammed cells, we need to understand the molecular and epigenetic determinants that convert one cell type into another. Here we review recent advances in this rapidly moving field and emphasize unresolved and controversial questions. Introduction Epigenetic changes, such as modifications to DNA and histones, alter gene expression patterns and regulate cell identity (Goldberg et al., 2007). Global epigenetic states must be tightly regulated during development to allow for the proper transitions between cellular states. However, cell fates during development are neither restrictive nor irreversible. The generation of animals by the nuclear transplantation of somatic nuclei into eggs (Gurdon, 1962) demonstrated that indeed the epigenome of differentiated cells can be reset to a pluripotent state. Derived from cells at various embryonic and postnatal stages, stem cells are characterized by self-renewal and the capacity for differentiation (Jaenisch and Young, 2008). Pluripotent cells have the ability to form all somatic lineages, and the first pluripotent cells were derived from a type of germline tumor, called teratocarcinoma. When explanted in tissue culture, the teratocarcinoma cells generated embryonal carcinoma cells, demonstrating that cancer cells can be reprogrammed to pluripotent cells (Hogan, 1976). The next breakthrough in the field came when researchers isolated embryonic stem cells (ESCs) from normal mouse embryos, creating a platform for the genetic engineering of animals (Evans and Kaufman, 1981). The generation of ESCs from human embryos came less than a decade later (Thomson et al., 1998), and this technology, combined with the direct reprogramming of somatic cells to pluripotent cells (Takahashi and Yamanaka, 2006), is now paving the way for ‘‘personalized’’ regenerative medicine (Hanna et al., 2007). This Review focuses on mechanisms that control the transition of cells between different states of pluripotency and differentiation. We will emphasize new concepts and unresolved questions in mammalian systems while concentrating on three aspects of epigenetic reprogramming: (1) the molecular definition of different pluripotent states and strategies to convert one cell state into another; (2) molecular concepts of somatic cell reprog508 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
ramming to pluripotency; and (3) direct transdifferentiation between somatic cell states. Distinct Pluripotent Cells Derived during Development Development proceeds from a state of totipotency, characteristic of the zygote and blastomeres during the early cleavage of the embryo, to cells that are restricted in their potential for development. It is from these later stages that pluripotent cells can be derived. At the 16-cell stage, the outer cells of the mouse embryo are allocated to two lineages: the trophoblast lineage, which will form part of the placenta; and the bipotential inner cell mass, which generates the epiblast and the hyphoblast. The epiblast and hyphoblast will form the embryo and the yolk sac, respectively. Cells of the epiblast lineage are termed pluripotent because they are the origin of all somatic cells and germline cells of the developing embryo. Primordial germ cells emerge at gastrulation and, in male embryos, give rise to spermatogonial stem cells. Pluripotent cells have been derived from all of these cell types by explanting the cells from embryos at different stages of development (Figure 1). As outlined below, the state of the donor cells, as well as the culture conditions, have a profound effect on the characteristics of the derived cells. We focus on pluripotent cells that have unrestricted developmental potential and thus can give rise to all cell types in the developing embryo or in the culture dish. A. Embryonic Stem Cells ESCs were the first pluripotent cells isolated from normal embryos. They were created by explanting the inner cell mass of the embryos from a strain of mice called ‘‘129’’ (Evans and Kaufman, 1981). Mouse ESCs recapitulate full developmental potential when injected into mouse blastocysts, contributing cells to the three germ layers and to the germline of chimeric animals. Consistent with their origin from the inner cell mass, ESCs express key pluripotency genes, such as Oct4, Sox2,
Figure 1. Developmental Origins of Pluripotent Stem Cells Different types of pluripotent cells can be derived by explanting cells at various stages of early embryonic development. Induced pluripotent stem cells (iPSCs) are derived by direct reprogramming of somatic cells in vitro.
and Nanog, and they exist in a pre-X-inactivation state with both X chromosomes active in female cells (Nichols and Smith, 2009). However, distinct biological and molecular characteristics distinguish ESCs from their in vivo counterparts of the inner cell mass. For example, cells of the inner cell mass are not self-renewing, and they are characterized by a genome that is globally hypomethylated (Santos et al., 2002). In contrast, ESCs have unlimited proliferation potential, and their genome is highly methylated (Meissner et al., 2008). Maintaining the pluripotent state of ESCs depends on key molecular signaling pathways (Table 1). Initially, researchers established ESCs in the presence of fetal bovine serum and other undefined factors secreted from irradiated mouse embryonic feeder cells. Studies identified the leukemia inhibitory factor (LIF) as an important mediator that supports maintenance of the pluripotency of mouse ESCs by signaling predominantly through the Stat3 pathway (Ying et al., 2008). Cultivating the cells in defined medium containing bone morphogenetic protein 4 (BMP4) allowed propagation of ESCs in the absence of feeders and serum. BMP4 acts in concert with LIF and supports stabilization of mouse ESCs by inducing inhibitors of differentiation (Id) genes. LIF and small-molecule inhibitors (termed ‘‘2i’’) of protein kinases ERK1/2 and GSK3b, which stimulates the WNT pathway, can replace the serum and thus allow the maintenance of ESCs in fully defined medium without embryonic feeder cells (Ying et al., 2008). A core set of transcription factors consisting of Oct4, Nanog, Sox2, and Tcf3 maintains the pluripotent state of ESCs (Boyer et al., 2005; Loh et al., 2006; Marson et al., 2008). Oct4 is expressed throughout early mammalian development and is essential for formation of the pluripotent inner cell mass and for maintaining ESCs (Nichols et al., 1998). Nanog is another important pluripotency regulator that is activated at the 8-cell stage. However, Nanog is also expressed later in a subset of inner cell mass cells and cooperates with other factors in X chromosome reactivation (Silva
et al., 2009). Oct4, Sox2, and Nanog induce and cross-regulate their own expression. Thus, Oct4-Sox2-Nanog constitutes a core transcriptional circuit wired in a feed-forward type of regulation (Figure 4D). These factors also coactivate redundant target genes and cooperate with secondary transcription factors that provide further stability to the ESC state (Chen et al., 2008). B. Epiblast Stem Cells Pluripotent cells can also be derived from the epiblast of the implanted embryo. The epiblast is a single layer of epithelial cells and originates from the inner cell mass after implantation of the embryo. Independent pluripotent lines, called ‘‘EpiSCs,’’ were established by explanting the epiblast from embryonic day 5.5–7.5 of postimplantation mouse embryos in growth conditions supplemented with FGF2 (basic fibroblast growth factor) and Activin (Brons et al., 2007; Tesar et al., 2007). EpiSCs express pluripotency markers and are pluripotent by a number of criteria, such as multilineage differentiation into embryoid bodies and teratomas. EpiSCs display flat colony morphology and grow poorly as single-cell clones following treatment with the protease trypsin (i.e., single-cell dissociation by trypsinization). Although these cells are termed pluripotent, they have more limited developmental potential than ESCs; they are highly inefficient in generating chimeras, have already undergone X chromosome inactivation, and demonstrate heterogeneous expression of early lineage-commitment markers. In contrast to ESCs, the growth of EpiSCs depends on signaling by Activin, FGF2, ERK1/2, and TGF-b; their growth is inhibited by BMP4 but is independent of LIF/Stat3 activity (Table 1). Molecularly, EpiSCs share a gene expression program reminiscent of the postimplantation epiblast, rather than that of the inner cell mass. EpiSCs exhibit reduced expression levels of the transcription factors Nanog, Rex1, and Klf, but differentiation markers, such as FGF5 and major histocompatibility complex (MHC) class I, are readily expressed in these cells (Tesar et al., 2007). Cell 143, November 12, 2010 ª2010 Elsevier Inc. 509
Table 1. Characteristics of Mouse Pluripotent States Characteristic
Naive Mouse Pluripotent State
Primed Mouse Pluripotent State
Alternative names
Preimplantation inner cell mass-like; mouse ESC-like
Post-implantation epiblast-like; mouse epiblast stem cell-like
Stem cell types
Embryonic stem cells (ESCs); embryonic germ cells (EGCs); spermatogonial germ stem cell (maGSC); induced pluripotent stem cell (iPSC)
Epiblast stem cell (EpiSCs)
Embryonic body formation
Yes
Yes
Teratoma
Yes
Yes
Blastocyst chimera
Yes
No
Susceptibility for primordial germ cell specification
Low
High
Single-cell clonogenicity
High
Low
Ability to grow independent of feeders and/or in defined medium
Yes
Yes
Doubling time in vitro
10–14 hr
14–16 hr
Morphology
Domed
Flattened
Positive regulators of the state
LIF/Stat3 BMP4 WNT IGF
TGF-b Activin FGF2 ERK1/2 WNT IGF
Negative regulators of the state
TGF-b Activin FGF2 ERK1/2
BMP4
Developmental Potential
Growth properties
Regulation by exogenous signaling pathways
Gene and marker expression signatures SSEA1, alkaline phosphatase
+
+
TRA1-60, TRA1-81, SSEA3/4
Oct4, Sox2
+
+
Nanog, Klf2, Klf4, Rex1, Stella
High ++
Low/absent +/
Lineage specification markers (e.g., FGF5, Blimp1, Cer1)
Absent
Positive with heterogeneous expression pattern
MHC class I
Nearly absent
Present
Oscillatory gene patterns
Nanog, Rex1
Blimp1
Oct4 enhancer activity
Distal element
Proximal element
XX status
XaXa
XaXi
Epigenetic state
C. Embryonic Germ Cells and Adult Germline Stem Cells Pluripotent cells have also been derived from cells of the germline lineage. When cultivated in adequate growth conditions, primordial germ cells isolated from embryonic day 8.5 embryos generated ES-like cells, termed embryonic germ cells (Surani, 1999). These cells were pluripotent, capable of generating teratomas and chimeras. Interestingly, in contrast to ESCs, when embryonic germ cells are fused with somatic cells, they can induce demethylation of the somatic imprinted genes, reflecting an enzymatic activity that resets the imprints during the development of primordial germ cells (Surani, 1999). Embryonic germ 510 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
cells have also been derived from human fetuses, but their characteristics are not as well defined (Shamblott et al., 1998). Spermatogonial stem cells from newborn and adult male gonads also generate ES-like cells, although at very low efficiency and after an extended time when these cells are explanted in vitro. Called male germ stem cells, these ES-like cells can be propagated in serum and LIF (Kanatsu-Shinohara et al., 2004). They carry a male-specific imprinting pattern, and they can induce teratomas and contribute to chimeras. Nevertheless, embryonic germ cells and male germ stem cells are pluripotent and share defining features with ESCs. Adult germline stem cells
have also been isolated from adult human testicular tissues (Conrad et al., 2008), but the identity of these cells has been questioned (Ko et al., 2010). Molecular Definition of Distinct Pluripotent States Much work has been devoted to establish molecular signatures that define pluripotency. As outlined above, ESCs are derived from the inner cell mass whereas EpiSCs are derived from the epiblast. Both display distinct biological characteristics that are reminiscent of their developmental origin, with some adaptations that occur upon explantation in vitro and during growth selection. It is becoming increasingly evident that different pluripotent cell types can be classified into two fundamentally distinct states of pluripotency: (1) the inner cell mass-like (ICMlike) pluripotent state, which is typical for ESCs derived from the inner mass cells, as well as embryonic germ cells and male germ stem cells derived from primordial germ cells or spermatogonial stem cells; and (2) the postimplantation epiblast-like state, characteristic of EpiSCs. The two pluripotent states, which are stabilized in vitro by different growth conditions, exhibit distinct molecular signatures and in vitro growth properties. In addition, the two states depend on signaling pathways that often antagonize each other (Table 1). BMP4 signaling stabilizes ESCs in conjunction with LIF, but it induces differentiation of EpiSCs; TGF-b and FGF2 support renewal of EpiSCs, but they induce differentiation of ESCs; and, EpiSCs require signaling of the ERK1/2 pathway whereas the self-renewal of mouse ESCs is enhanced by inhibition of ERK1/2 signaling. Nichols and Smith (2009) designated the ICM-like state of ESCs as the ‘‘naive’’ state and that of the epiblast-derived EpiSCs as the ‘‘primed’’ pluripotent state. This definition implies that the primed state is prone to differentiate whereas the naive ESCs correspond to a more immature state of pluripotency. This difference is, for example, reflected in the state of X chromosome inactivation. Naive ESCs are in the preinactivation state with both X chromosomes active (XaXa) in female cells; in contrast, primed EpiSCs have already undergone X chromosome inactivation. In addition, primed EpiSCs are poised to generate precursors of primordial germ cells in response to BMP4 signaling. This is consistent with the similarity of EpiSCs to the postimplantation epiblast state whereas BMP4 sustains the undifferentiated state of ESCs (Tesar et al., 2007). A recent report described an EpiSC-like cell type, termed FABSCs (i.e., FGF2, Activin, and BIO-derived stem cells) (Chou et al., 2008), which share expression markers with EpiSCs but are unable to differentiate. These results lead to the suggestion that FAB-SCs may represent a novel state of pluripotency. However, the nature and relevance of these cells is unknown, given that the FAB-SCs are not pluripotent and that they lack differentiation potential unless exposed to LIF/BMP4. Stability of and Transitions between Pluripotent States The molecular and biological definitions of naive and primed pluripotent cells raised the question of whether these epigenetic states are derived from independent coexisting progenitors or whether they reflect distinct pluripotent states characteristic of embryonic cells at successive developmental stages. This issue
was highlighted by the failure to derive naive ESCs from certain mouse strains, other rodents, and other species, including humans. Naive ESCs are readily isolated from only a limited number of ‘‘permissive’’ mouse strains, such as 129, C57BL/6, and BALB/ C. Explanted blastocysts from rats and ‘‘nonpermissive’’ mouse strains, such as nonobese diabetic (NOD) mice, yielded exclusively EpiSC-like pluripotent cells (Buehr et al., 2008; Hanna et al., 2009a). These results mistakenly suggested that the primed state is the only or ‘‘default’’ state of pluripotency in ‘‘nonpermissive’’ donor strains or species. The restriction of isolating naive cells from only particular strains or species is odd and difficult to explain, but most importantly, it has triggered researchers to determine how the two pluripotent states are related and whether it would be possible to switch one state into the other. Both in vitro and in vivo experiments have recently defined the relationships between the two distinct types of pluripotent states. Whereas naive pluripotent cells can differentiate into a primed EpiSC-like state in vitro by promoting the signaling of TGF-b, Activin, and FGF2, EpiSCs can epigenetically revert back to naive pluripotency by a variety of genetic manipulations and culture conditions (Figure 2) (Bao et al., 2009; Guo et al., 2009; Hanna et al., 2009a). Exposure to LIF/Stat3 signaling reverts EpiSCs from permissive strains to naive mouse ESClike cells. This conversion can be boosted by cultivating the cells in LIF and ‘‘2i’’ conditions (2i: GSK3b inhibitor and ERK1/2 inhibitor or Kenpaullone) or by the transient expression of pluripotency factors, including Klf4, Klf2, Nanog, or c-Myc, with different latencies and efficiencies (Figure 2A). These observations suggest that the naive and primed pluripotent states in the permissive genetic background are interconvertible and can be stabilized by appropriate culture conditions. In contrast to deriving cells from blastocysts of the permissive 129 strain, when NOD mouse blastocysts were explanted under standard ESC conditions in LIF and on feeders, they generated only EpiSC-like cells, not ESCs (Hanna et al., 2009a). This is consistent with the observation that LIF/Stat3 signaling alone is not sufficient to generate ESCs from NOD mouse blastocysts or to facilitate epigenetic reversion of NOD EpiSCs. However, cultivation of the NOD embryos in LIF and 2i conditions generated naive ESCs that were dependent on the continuous presence of additional exogenous factors that promote naive pluripotency (Figure 2B). Hence, in contrast to ESCs generated from the 129 strain, which can be maintained in LIF and serum, stabilization of pluripotent naive cells from the NOD strain requires additional factors in the medium. This property underscores the inherent ‘‘bistability’’ (Figure 5) of pluripotency that allows the naive and primed states to interconvert. Notably, in both mice and humans, naive and primed states have not been observed to coexist stably in the same culture conditions. Thus, the transitions between primed and naive states are distinct from cell-tocell differences that coexist during the culturing of mammalian ESCs (Figure 5), including oscillation in the expression of pluripotency markers Nanog and Stella (Hayashi et al., 2008). In summary, pluripotent cells isolated in different mammalian species can exist as distinct pluripotent states, known as the naive and the primed states, and specific extrinsic and intrinsic Cell 143, November 12, 2010 ª2010 Elsevier Inc. 511
Figure 3. Three Parameters that Impact Pluripotency Exogenous factors, genetic background, and epigenetics of the tissue origin of cells.
factors can induce transitions between the states (Figure 3). The genetic background of the permissive 129 mouse strain has the least requirements for naive pluripotency (LIF in Figure 2B) whereas the nonpermissive NOD background (or rat cells) requires additional factors that support the naive pluripotent state, such as the 2i conditions. The intrinsic genetic determinants of human cells are even less permissive because additional extrinsic factors are necessary for stabilizing the naive state (see next section). If the minimal growth requirements are not maintained, the cells lose the naive pluripotent state and may differentiate or convert into the primed state. In addition, the tissue origin of the cells may play a key role in establishing a particular state of pluripotency (Figure 3). For example, induced pluripotent stem cells (iPSCs) and ESCs derived from the NOD mouse strain require supplementation of 2i in addition to LIF whereas NOD germline stem cells derived from adult male gonads need only LIF for stabilization of the naive pluripotent state (Ohta et al., 2009). The molecular basis underlying this observation is still undefined. Human Embryonic Stem Cells Like mouse ESCs, human ESCs are isolated from explanted blastocysts before implantation (Thomson et al., 1998). Nevertheless, human ESCs share multiple defining features with mouse EpiSCs rather than mouse ESCs. These characteristics include flat morphology, dependence on FGF2/Activin signaling, propensity for X chromosome inactivation, and reduced tolerance to single-cell dissociation by trypsinization (Table 1). These molecular and biological similarities with mouse EpiSCs suggest Figure 2. Transitions between Naive and Primed Pluripotent States (A) Naive and primed cell types represent distinct gene expression states, corresponding to those observed in the preimplantation inner cell mass and postimplantation epiblast, respectively. To stabilize the primed state in vitro, supplementation with basic fibroblast growth factor (FGF2) and Activin support the core transcriptional circuitry governing this state. In contrast, the naive state has distinct active signaling pathways and thus requires different exogenous signals to induce and stabilize this state in vitro. Depending on the genetic background, combinations of these perturbations promote reversion to naive pluripotency and differentiation into the primed pluripotent state. (B) In this illustration, stabilization of the naive and primed states in vitro by leukemia inhibitory factor (LIF) and FGF2/Activin signaling, respectively, is depicted as creating a well in a landscape. These factors can promote or antagonize interconversion between the states. LIF signaling promotes transfer of primed cells to the naive state and continuously prevents differentiation
512 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
of the naive state. Shielding FGF2/Activin signaling can further enhance conversion into naive cells, as this signaling pathway is inhibitory to the naive state. In the nonpermissive NOD mouse strain, LIF signaling alone is not sufficient to maintain the naive state in vitro, as pluripotent cells derived from the inner cell mass or by in vitro reprogramming assume a primed state that can be stabilized by FGF2/Activin. However, the modulation of additional pathways, which are known to promote the naive state and prevent differentiation, allowed derivation of naive pluripotent cells from the NOD strain. This was achieved by altering the culture conditions, either with small molecules or by adding LIF, increasing Wnt signaling (small molecule ‘‘CH’’), and inhibiting ERK1/2 (small molecule ‘‘PD’’). The human genetic background is less permissive, as it required modulation of additional signaling pathways to induce the naive state.
that human ESCs correspond, at least partially, to the primed pluripotent state rather than to the naive state of mouse ESCs. For pluripotent cells from mice, isolation and culture conditions, as well as genetic differences of the blastocysts from ‘‘permissive’’ and ‘‘nonpermissive’’ mouse strains, profoundly affect the cell’s state of pluripotency (Figure 2). In fact, the interconversion between pluripotent states of mouse cells (Hanna et al., 2009a) raised the possibility that the appropriate culture conditions would allow isolation of naive human stem cells. Consistent with this possibility is that the isolation and culture conditions of human ESCs profoundly influence the state of X chromosome inactivation. Conventional human ESCs, isolated at atmospheric oxygen concentrations, have undergone X chromosome inactivation (XiXa), similar to mouse EpiSCs (Silva et al., 2008). However, human ESCs isolated and propagated under physiological oxygen conditions (5% O2), display a pre-X-inactivation status (XaXa) and, similar to mouse ESCs, initiate random inactivation upon differentiation (Lengner et al., 2010). This result argues that human blastocysts contain pre-X-inactivation cells and that oxidative stress upon culture of the embryos in atmospheric oxygen accelerates precocious X inactivation. Thus, suboptimal culture conditions interfere with the in vitro capturing of the more immature XaXa state of human inner cell mass cells. These results suggest that the proper conditions have not yet been devised for isolating human pluripotent cells with features similar to mouse naive ESCs or for converting human ESCs and iPSCs (which are similar to NOD EpiSCs) to a naive pluripotent state. Indeed, human ESCs and iPSCs were recently converted to a naive pluripotent state by propagating the cells in LIF and 2i conditions (i.e., the addition of PD/CH inhibitors) and simultaneously overexpressing Oct4/Klf4 or addition of Forskolin to the medium (Figure 2B) (Hanna et al., 2010). These naive human ESCs and iPSCs corresponded to naive mouse ESCs by several criteria. They reactivated the inactive X chromosome, resulting in a pre-X-inactivation status (XaXa in female cells). They exhibited a high efficiency of single-cell cloning and were dependent on LIF/Stat3 signaling instead of FGF/Activin signaling. They could routinely be passaged as single cells and showed a gene expression pattern that resembled that of naive mouse ESCs. Furthermore, the XaXa state of these naive ESCs was observed when the cells were cultured under 20% oxygen in contrast to the labile pre-X-inactivation state of primed ESCs isolated from human embryos under physiological oxygen conditions (Lengner et al., 2010). These results provide the first direct evidence for a validated naive state of pluripotency in humans that is highly similar to that of mouse ESCs but that conventional isolation conditions of ESCs failed to capture. Nevertheless, this naive state could be maintained only for limited passages (Hanna et al., 2010) even when cells were cultivated in the presence of Forskolin, which allows propagation of human embryonic germ cells (Shamblott et al., 1998). Thus, a crucial challenge will be to define growth conditions that allow robust, long-term maintenance of the ‘‘naı¨ve ground state’’ in genetically unmodified human cells. Conventional human ESCs are impractical for use in diseaserelated research because of the laborious culture conditions required for their maintenance, their low efficiencies of genetargeting by homologous recombination, and the dramatic
heterogeneity in differentiation potential among different human ESC lines (Osafune et al., 2008). Recently, novel gene targeting strategies using zinc-finger nucleases have overcome some of these limitations (Hockemeyer et al., 2009; Zou et al., 2009). Nevertheless, it will be interesting to determine whether genetic manipulation by homologous recombination is as efficient in the new naive human pluripotent cells as in mouse ESCs. A recent report (Buecker et al., 2010) described a cell state termed ‘‘hLR5’’ generated by the ectopic expression of the five different transcription factors Oct4, Sox2, c-Myc, Klf4, and Nanog in human fibroblasts. The cells were amenable to gene targeting of the HPRT (hypoxanthine-guanine phosphoribosyltransferase) locus by homologous recombination. However, the targeting efficiency at this particular locus was 2-fold lower than that reported previously for conventional human ESCs (Zwaka and Thomson, 2003). Although designated as ‘‘murineESC-like cells’’ and iPSCs, the hLR5 cells were not pluripotent because they failed to activate the endogenous pluripotency genes, lacked any informative expression of pluripotency markers, and were unable to differentiate (Buecker et al., 2010). Therefore, these cells likely represent transformed or partially reprogrammed cells (Mikkelsen et al., 2008). Direct Reprogramming of Somatic Cells to Pluripotency Epigenetic reprogramming of somatic cells to a pluripotent state has been achieved by nuclear transplantation, cell fusion, and direct reprogramming by expression of transcription factors. We focus here on unresolved or controversial issues of direct reprogramming because nuclear transfer and cell fusion approaches have been extensively reviewed elsewhere (Jaenisch and Young, 2008; Yamanaka and Blau, 2010). Takahashi and Yamanaka (2006) achieved a breakthrough in this field by demonstrating that the overexpression of four transcription factors, Oct4, Sox2, Klf4, and c-Myc, can convert somatic fibroblasts to pluripotent cells (iPSCs) that can contribute to the germline in chimeric mice (Okita et al., 2007; Wernig et al., 2007). These factors initiate poorly defined events that lead eventually to the reactivation of the endogenous pluripotency genes encoding Oct4, Nanog, and Sox2 and to the activation of the autoregulatory loop that maintains the pluripotent state independent of the transgenes (Figure 4D). Murine iPSCs share all defining features with naive mouse ESCs, including expression of pluripotency markers, reactivation of both X chromosomes, and the ability to generate chimeras and all-iPSC mice following tetraploid complementation (Boland et al., 2009; Kang et al., 2009; Zhao et al., 2009). Reprogramming can be induced not only by Oct4, Sox2, Klf4, and c-Myc but also by alternative combinations that employ Nanog, Lin28, ESRRB, NR5A2, and other genes that promote the establishment of the core transcriptional circuitry of stem cells (Ichida et al., 2009; Yu et al., 2007). The redundancy and cooperative action of reprogramming factors in establishing iPSCs likely results from the highly interconnected DNA-binding properties of the pluripotency factors that form the regulatory circuitry and that allow the transduced transcription factors to reestablish the autoregulatory and feed-forward loops by activating the endogenous pluripotency genes (Boyer et al., 2005). Cell 143, November 12, 2010 ª2010 Elsevier Inc. 513
Furthermore, given that the transcriptional circuit of pluripotency is positively or negatively regulated by cytokines or small molecules added to the medium that activate or suppress specific signaling cascades, it is not surprising that the same pathways influence the establishment and maintenance of both ESCs and iPSCs. For example, WNT signaling upregulates c-Myc expression and promotes naive pluripotency, but it also increases iPSC formation (Marson et al., 2008). Furthermore, the effect a small molecule has on iPSC formation depends on whether the small molecule affects a pathway that also stabilizes the given pluripotent state. For example, conventional human ESCs rely on TGF-b signaling, and as expected, addition of TGF-b inhibitors impedes human iPSC formation because these inhibitors are detrimental to the primed pluripotent state (Maherali and Hochedlinger, 2009). In contrast, inhibitors of this pathway enhance mouse iPSC formation because TGF-b destabilizes the naive pluripotent state of mouse ESCs. Reprogramming of somatic cells to pluripotency is accompanied by extensive remodeling of epigenetic marks, including DNA demethylation of key pluripotency genes such as Oct4 and Nanog. In somatic cells, the promoters of Oct4 and Nanog are highly methylated, reflecting their transcriptionally repressed state. The formation of iPSCs involves activation of these genes, and their demethylation is widely used to monitor successful reprogramming (Mikkelsen et al., 2008). In principle, demethylation can occur by a passive mechanism, such as the inhibition of DNA methyltransferase 1 (Dnmt1) during DNA replication, or by an active mechanism in which the methylated base is removed from nonreplicating DNA. Active demethylation is convincingly demonstrated only in cells that do not replicate their DNA. For example, global demethylation of the paternal genome occurs after fertilization and prior to the onset of DNA replication. In addition, it has been suggested that global demethylation of the genome in primordial germ cells may also occur by an active mechanism involving methyl-cytosine deamination by AID (activation-induced cytidine deaminase) (Popp et al., 2010). However, the expression levels of the enzymes in the AID pathway are low in primordial germ cells and the zygote, and a recent report suggests that enzymes of the base excision repair pathway rather than AID may catalyze this global demethylation in both types of cells (Hajkova et al., 2010). Because the pluripotent state is dominant over the somatic state, one strategy for reprogramming somatic cells is to fuse them with ESCs. It has been suggested that AID is involved in active demethylation of the somatic Oct4 gene at 48 to 72 hr after fusion but before the onset of DNA replication (Bhutani et al., 2009). Although the Oct4 promoter was found to be partially demethylated, the only evidence for reprogramming was a residual level of Oct4 expression (100-fold lower than the levels in ESCs). This is reminiscent of partially reprogrammed cells that express low levels of the endogenous Oct4 gene without being demethylated (Mikkelsen et al., 2008). Although this fusion approach may allow the dissection of the initial events taking place at the beginning of direct reprogramming, it is still unknown whether demethylation of the Oct4 gene detected in the early ES-somatic cell heterokaryons, combined with the very low levels of Oct4 expression, do actually reflect events relevant 514 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
for iPSC formation. A previous study, which fused B cells with ESCs, concluded that high Oct4 expression occurred only later after DNA replication and cell proliferation of the hybrid cells had occurred (Pereira et al., 2008). Cell proliferation and DNA replication precede iPSC formation. Thus, it is likely that the activation of the somatic pluripotency genes in somatic cells fused with ESCs occurs, at least in part, by a passive mechanism involving inhibition of Dnmt1. This is consistent with the observation that inhibition of Dnmt1 increases the efficiency of reprogramming (Mikkelsen et al., 2008). Clearly, much remains to be learned about the nature of the biochemical machinery involved in facilitating chromatin modification during iPSC formation (Singhal et al., 2010). Dynamics and Heterogeneity of Direct Reprogramming Even when different strategies are used to induce reprogramming, a consistent finding is that only a small fraction of donor cells will become iPSCs, with the first ones appearing no earlier than 5–10 days after expression of the reprogramming factors (Jaenisch and Young, 2008). Thus, the reprogramming efficiency appears to be quite low. Parameters thought to restrict the reprogramming efficiency include the possibility that only rare somatic stem cells may be susceptible to reprogramming or that activation of additional genes by insertional mutagenesis might be crucial. However, iPSCs from mouse (Hanna et al., 2008) and human (Loh et al., 2010; Seki et al., 2010; Staerk et al., 2010) were recently generated from lymphocytes with high efficiency, providing conclusive evidence that terminally differentiated cells can be reprogrammed to pluripotency. In contrast to previous claims (Eminli et al., 2009), control of in vitro cell plating efficiency, growth expansion, and gene delivery demonstrated that iPSCs need not preferentially arise from less differentiated somatic cells (Hanna et al., 2009a, 2009b) and that differentiation progression is not accompanied with an intrinsic decrease in reprogramming amenability. Furthermore, the generation of genetically unmodified iPSCs argues that insertional mutagenesis is not an essential step in the process (Okita et al., 2008; Stadtfeld et al., 2008). Determination of ‘‘reprogramming efficiency’’ and kinetics of in vitro reprogramming is typically based upon the appearance of iPSC colonies at a single and arbitrarily chosen time point (typically 3–6 weeks) after polyclonal somatic cell populations are transduced with reprogramming factors and plated. Efficiency is calculated by the fraction of reprogrammed cells (indicated by expression of pluripotency markers or reporter genes) divided by the total number of plated cells. Although such measurements can be informative, they provide limited mechanistic insights because it is difficult to quantify the extensive expansion and/or apoptosis of the original cells, which can be an immediate technical consequence of factor expression. The mouse embryo fibroblasts used in most studies represent a heterogeneous population of cells that are highly variable in their predisposition for immortalization, senescence, and tolerance to ectopic expression of exogenous factors (such as Klf4 and c-Myc), which affect the survival of single cells. Further, pluripotency markers, such as alkaline phosphatase and SSEA1 or SSEA4, are unspecific, and only a small fraction of cells with these markers will develop later
into genuine iPSCs (Jaenisch and Young, 2008). Moreover, the possibility that individual iPSC colonies may be sister clones from the same infected cell has also been ignored hitherto. Thus, although p53 inhibition was reported to variably increase reprogramming efficiency by 4- to 100-fold (Krizhanovsky and Lowe, 2009), this variation may be due to technical parameters of somatic survival, senescence, and apoptosis after expression of genes, such as c-Myc and Klf4. Because viral infections are usually used to induce reprogramming, heterogeneity in the expression of the reprogramming factors may be quite large, and thus, transgenic approaches were recently developed to overcome this experimental factor. Transgenic mice carrying a defined set of drug-inducible (doxycycline, DOX) proviruses, transposons, or a polycistronic construct encoding the reprogramming factors inserted into a single expression locus generated ‘‘secondary’’ somatic cells that could be reprogrammed by mere addition of DOX to the medium (Hanna et al., 2008; Woltjen et al., 2009). The use of these secondary cells avoided the need for a new virus infection to induce reprogramming and increased the reprogramming efficiency. However, still no more than 5%–10% of the cells eventually became reprogrammed, with a latency of 7–10 days before the first iPSCs appeared. This delay and the low efficiency are consistent with stochastic mechanisms involved in inducing reprogramming (Hanna et al., 2009b). Moreover, although the secondary cells after the initiation phases are genetically homogenous, cells at intermediate stages of reprogramming represent highly heterogeneous cell populations, and only a minority of these cells will ever become iPSCs. Thus, gene expression or epigenetic analyses of such heterogeneous cell populations may not be informative for characterizing those few cells that eventually form an iPSC. Long-term analyses of clonal cell populations derived from Pro/Pre-B cells were conducted to determine the dynamics of reprogramming and to assess the fraction of donor cells that are susceptible to reprogramming (Hanna et al., 2009b). In comparison to mouse embryonic fibroblasts, Pro/Pre-B cells have a higher cloning efficiency for single cells and better tolerance of reprogramming factors. Plus, these cells represent a well-defined lineage-committed population with the rearrangement of the IgH locus allowing for the unambiguous retrospective identification of the donor cell. In these experiments, reprogramming efficiency was defined as the potential of a donor cell to generate an iPSC daughter at some point. This experiment demonstrated that, with a balanced reprogramming factor stoichiometry, nearly every cell was able to generate iPSCs. These results argue that differentiation does not restrict the ability of somatic cells to be reprogrammed although the efficiency of a given daughter cell to become an iPSC is exceedingly small (Figure 4A). The kinetics of reprogramming displayed a broad distribution for the time before iPSCs appear, spanning 2–18 weeks, consistent with the notion that the process involves stochastic and rate-limiting epigenetic event(s). Additional inhibition of the p53/p21 pathway or ectopic expression of Lin28, while controlling for the same growth conditions, accelerated conversion into iPSCs and was directly proportional to the increase in cell division rate. Thus, p53 inhibition did not increase the fraction of cells that could be
reprogrammed but rather accelerated the formation of iPSCs in time, which appeared after a similar number of cell divisions. In contrast, ectopic expression of Nanog accelerated reprogramming in a manner that was independent of the rate of cell division (Figure 4B). These observations suggest that the cell cycle is a key parameter in iPSC generation and that reprogramming may be driven in a mode that is either dependent or predominately independent of cell division (Hanna et al., 2009b). Two possibilities could explain why increased cell proliferation accelerates the kinetics of iPSC formation. Accelerated cell division could amplify the number of target cells in which each daughter cell has an independent probability of becoming an iPSC, or DNA replication may be the prerequisite for permitting the epigenetic changes, such as DNA and histone modifications, to occur that allow the transitions to pluripotency. Several lines of evidence suggest that an important ratelimiting epigenetic event for reprogramming may be the reactivation of the key endogenous, autoregulatory circuitry that maintains the ESC state (Jaenisch and Young, 2008). The inability to generate iPSCs from somatic cells in which Nanog is disrupted (Silva et al., 2009) and the observation that ectopic expression of Nanog induces iPSCs in fewer cell divisions (Hanna et al., 2009b) suggest that activation of the endogenous Nanog gene is a required event in the establishment of pluripotency. In addition, a recent gene expression analysis of secondary populations of mouse embryonic fibroblasts indicated that irreversible commitment to reprogramming coincides with endogenous activation of Nanog (Samavarchi-Tehrani et al., 2010). Further, partially reprogrammed cell lines may represent stable intermediate stages in the reprogramming process that depend on the continuous expression of exogenous reprogramming factors (Mikkelsen et al., 2008; Sridharan et al., 2009). However, these cells can be induced to give rise to fully reprogrammed iPSCs upon additional manipulations that lead to the activation of the autoregulatory circuitry. Although these observations indicate that the reactivation of the endogenous autoregulatory circuitry is important and that the process is accompanied by many epigenetic changes, it is still unknown how many of these required epigenetic events are rate limiting for reprogramming. Modeling Direct Reprogramming to Pluripotency The transitions between lineage-committed and pluripotent cell states can be tracked quantitatively in terms of the fraction of cells in a given culture environment that change state over time. Mathematical modeling using experimental data of transition rates is consistent with the conclusion that a single epigenetic event is rate limiting. Both simulations and experiments indicated a single peak in the reprogramming latency distribution when reprogramming was sampled daily over 14 days in polyclonal populations of mouse embryonic fibroblasts (Smith et al., 2010) and every week over an extended 18 week culture in clonal pre-B cell and monocytes populations (Hanna et al., 2009b). In these simulations, when somatic cells express key reprogramming transcription factors, they can transition to the iPSC state with a particular probability (Figure 5). This probability Cell 143, November 12, 2010 ª2010 Elsevier Inc. 515
Figure 4. Trajectories of Epigenetic Reprogramming to Pluripotency (A) In direct reprogramming, the progression of clonal populations to a reprogrammed state first involves an initial technical phase (I), which depends on the survival of plated cells and their entry into the cell cycle. Once cell division occurs, the most critical phase begins. During this second phase (II), direct reprogramming involves a stochastic event because clonal populations do not give rise to iPSCs at the same time after phase I. This variation in latency is represented by the blue line. (B) Phase II can be accelerated by two mechanisms involving cell division (purple) or mechanisms independent of cell division (orange). (C) As with direct reprogramming, the progression of reprogramming by nuclear transfer and cell fusion involves two phases. However, compared to direct reprogramming, much less heterogeneity is observed with nuclear transfer and cell fusion. For one, partially reprogrammed lines are not observed with nuclear transfer or fusion, and the reprogramming is hypothesized to progress in a more deterministic manner. This suggests that the current protocols of direct reprogramming are not optimal and may be accelerated by the supplementing with more factors to eliminate the stochasticity and achieve the deterministic conversion observed in fusion and nuclear transfer. (D) One key rate-limiting step during direct reprogramming may be reactivation of the core pluripotency regulatory circuitry.
reflects the randomness in cell population size, arising from stochasticity in cell division times, fluctuations in the number of cells from apoptosis, and potential loss of iPSCs during passaging and cell culture. Such randomness gives rise to variability among genetically identical cells in populations undergoing reprogramming and is typically ignored in more conventional quantitative analyses that depend on a population-averaged doubling time (e.g., Figure 4B). Simulations incorporating a single rate-limiting stochastic event were able to recapitulate 516 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
the experimentally observed kinetics of iPSC generation. Furthermore, simulations with multiple slow epigenetic events did not fit the experimental data better when the increase in model complexity was taken into account. Although this modeling suggests a single rate-limiting event, these results do not exclude the existence of other initial events that are not rate limiting but that occur with high probability and thus would not register in current models. Detailed tracking of cell division in simulations indicates that the rate of cell division is a key parameter controlling the kinetics of reprogramming. Simulations of several different reprogramming conditions showed that ectopic expression of Nanog could accelerate reprogramming largely independent of changing the cell division rate or cell population size (Hanna et al., 2009b). Nanog is a component of the core autoregulatory loop controlling pluripotency (Figure 4D), but it is not in the original reprogramming cocktail. Thus, ectopic expression of Nanog could provide a key missing core component required for the ratelimiting event to occur. Although ectopic Nanog expression accelerates reprogramming, the process remains stochastic and inefficient even when Nanog, Oct4, and Sox2 proteins are supplied ectopically. Current modeling approaches combine
Figure 5. Developing Probabilistic Descriptions of Cell State (A) First, a set of characteristics must be identified as being informative for the transition between two cell states of interest. These characteristics usually consist of levels of gene expression or levels of epigenetic methylation or acetylation marks on DNA or chromatin. For simplicity, we show here a state space generated by the level of N different genes, g1 to gN, where each arrow represents an axis corresponding to the expression level of that particular gene transcript (N104 for mammalian cells). A cell at any time t exists in a point in this space, and its state can change with time as a result of noise, reprogramming, or differentiation. One trajectory is plotted, ^ which consists of gene expresand the vector S, sion levels changing with time, fully describes the cell state transition during this time. (B) N can be reduced to a more manageable 2–3 dimensions through statistical techniques, such as principal component analysis (PCA). Shown here is a two-dimensional representation of the state space in (A), where each axis (ga and gb from PCA) is a linear combination of particular genes. Stable cells in vitro exist at particular points on this graph. By mapping quantitatively the gene expression levels of several single stable cells in the same space, regions with high densities of spots define observable cell types in vitro (as type ‘‘I’’ and ‘‘J’’). (C) When a large sample of single cells is mapped in (B), the probability of occupying each point in this space can be calculated and plotted in a continuous fashion. (D) The continuous probabilistic description of cell state in (C) can be simplified into a discrete representation, with a small number of discrete states that transition at particular rates, k. These rates, k, represent an average of all possible trajectories from region i to region j. (E) Alternatively, a continuous description of the ^ from (C) at each point. This landscape, probabilities of staying at a particular point in state space can be represented as a landscape, by calculating ln[P(S)] ^ represents ‘‘energy barriers’’ between transitions involving any two states and thus may provide a more thorough description of transitions than the descripV(S), tion in (D) .
and average several sources of biochemical noise (Figure 5) over the entire cell population and over the course of a week. Therefore, more detailed characterization through frequent single-cell tracking of gene expression, signaling activation, and the epigenetic state will probably provide more insights into the key parameters and required changes in gene expression controlling the kinetics of reprogramming. In contrast to current protocols for direct reprogramming, nuclear transfer appears to reprogram the somatic nucleus in a single event, as suggested by the activation of Oct4 in the four-cell stage cloned embryo (Boiani et al., 2002). Moreover, only fully reprogrammed ES-like cells have been derived by cell fusion or following nuclear transfer, with no evidence for partially reprogrammed cells found in the iPSC approach (Hasegawa et al., 2010). This might suggest that reprogramming by nuclear transfer or cell fusion, in contrast to factor-induced reprogramming, may follow a more synchronized trajectory during reprogramming and progresses in a deterministic pattern (Figure 4C). It is important to devise approaches that could achieve synchronized and deterministic reprogramming for direct in vitro reprogramming as well. For the interested reader,
we summarize in the next section approaches that are being used to model cell state transitions. Emerging Quantitative Models of Cell State Transitions Various approaches are being used to model transitions between different cell states. Depending on the differentiation pathway or the transitions between two states of interest, one may select a particular set of characteristics to describe quantitatively the cell state (Huang, 2009). Both gene expression and epigenetic changes on the timescale of days to weeks distinguish stable, functional cell types described by developmental biologists. Thus, the cell state can consequently be parameter^ which can be ized as a vector of molecular characteristics, S, ^ = [g1; g2; either a set of gene expression levels (Figure 5, S g3;.gN]) or a set of epigenetic marks, such as DNA methylation and histone acetylation (not shown in Figure 5). Although such molecular characteristics clearly can correlate with one another, they are not necessarily correlated in the same fashion (Lu et al., 2009). During differentiation or noisy gene expression in a partic^ is ular culture condition, the cell state can vary in time, t (i.e., S ^ a function of time S(t)). Cell 143, November 12, 2010 ª2010 Elsevier Inc. 517
The species or genetic background defines the architecture of the state space, meaning that particular gene-gene relationships, interaction modalities, and integrating transfer functions are ‘‘hard-wired’’ by the genome. Such state space provides a means to organize quantitatively and visualize different states of a cell with a fixed genome. Often cells cluster in particular regions of the state space, and developmental biologists typically describe these regions as stable cell types that express particular markers. There also may be regions where no cells are found, corresponding to cell states that are somehow not stable for the given genome of the cell in the extracellular environment considered. Using standard statistical techniques, dimensionality reduction can transform this space into lower dimensions and greatly simplify the system (Figure 5B). For example, using principal components analysis, cells of the mouse embryo from the 8-cell stage to the blastocyst were mapped to two dimensions, where each dimension was a linear combination of genes g1 to gn (Tang et al., 2010). We can assign probabilities to each point in state space if we assume that we have reasonably sampled the entire space considered (Figure 5C). This continuous probability space can also be simplified further into a discrete number of observed states with particular kinetics for their transitions, allowing easier analysis with fewer variables (Figure 5D). This probabilistic framework allows us to predict transition rates between two states when particular parameters are perturbed. Parameters incorporated in this framework include the genetic background, species differences, expression of ectopic transcription factors, culture conditions, and biochemical noise. Sources of biochemical noise among genetically identical cells include transcriptional noise in factor expression, biochemical noise in signaling processes, and biochemical noise in epigenetic modifications (Raj and van Oudenaarden, 2008). For example, noise in signaling pathways could arise as a consequence of the inherent stochastic nature of molecular binding events of ligands to receptors, receptors to secondary messengers, or secondary messengers to transcription factors. In addition, variable cell-cell contact in juxtacrine signaling can also create noise in signaling pathways. Finally, biochemical noise can occur when epigenetic modifiers bind to particular genomic loci. Whereas a probabilistic framework can comprehensively characterize transitions among cell states in many contexts, generating a ‘‘landscape’’ has also been a popular way to summarize all possible transitions that a single cell can make in a particular culture condition. Such summaries enable predictions on transition rates when several parameters are perturbed over a continuous state space. For example, quantitative models for directed evolution (Bloom et al., 2005) have already described noisy searches in configured landscape and in sequence space instead of state space. Similar models are helping to guide experimentation by attempting to describe cellular differentiation states in terms of gene expression patterns. To generate a landscape for a transition process driven by noise, one can estimate the rate of transitions from state i to j as proportional to exp(height of energy barrierij/noise), using Kramers’ escape-rate theory. To convert the landscape to a more intuitive plot in which ‘‘low energy (i.e., stability) = high 518 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
^ for each probability,’’ an inverse function of the probability P(S) ^ which then cell state can be used to plot an elevation, V(S), generates a quasi-potential energy landscape (Figure 5E). The height of the barrierij between states i and j is then No*ln(ki/j), where ki/j is the measured transition rate from state i to j, and No is a constant. Each well or local minimum in such a landscape is a ‘‘probable’’ or stable state (i.e., a stable attractor) that is analogous to a ‘‘low-energy state.’’ In contrast, hills are unstable states that are less likely (‘‘improbable’’) to be occupied and correspond to a ‘‘high-energy state.’’ Note that the elevation ^ on the z axis in Figure 5E does not constitute a true ‘‘potenV(S) tial energy’’ in the classical sense as proposed for systems like protein folding because the system equations for the regulatory networks are not integrable and constitute a non-equilibrium system (Huang, 2009). ‘‘Bistability’’ describes a landscape in which several stable states may coexist. For example, several blue wells contain stable coexisting states in Figure 5E, and these states may interconvert with one another at a particular rate (the corresponding discrete description is shown in Figure 5D). In contrast, ‘‘metastability’’ describes observed states that exist transiently and are highly sensitive to particular culture conditions or genetic determinants. These states do not sit at the bottom of stable wells in the landscape; for example, in Figure 5E, metastable states do not exist in blue wells but rather in the more yellow or green areas of the landscape. In practice, a metastable state can be defined as an observable state lasting at least an order of magnitude larger than its doubling time (e.g., >102 hr for mammalian cells). Mathematically, however, a metastable state in state space is time-invariant ^ parameters, with long lifetimes lasting of their state-describing S many times longer (e.g., 100-fold longer) than the shortest lived state. From a thermodynamics perspective, all cell states are metastable because cells do not operate at thermodynamic equilibrium. However in the biological context, metastability emphasizes the transient lifetime of cell states in contrast to stable self-renewing stem cells and terminally differentiated cells. In Figure 2, this framework is applied to describe the multiple pluripotency states observed in vitro. Ultimately, single-cell experiments coupled with numerical simulations could refine such landscapes to guide future experimentation aimed at dissecting the mechanisms of cell transitions. iPSCs versus ESCs: Are They Equivalent? A complex and unresolved question in the field is whether iPSCs are equivalent to ESCs. This is an important issue as genetic or epigenetic abnormalities may influence iPSCs during differentiation and/or transplantation, generating cells with molecular profiles and biological characteristics that are different from ESC-derived cells. The criteria used to compare iPSCs and ESCs include biological assays that test for developmental potency and molecular assays that compare gene expression and epigenetic characteristics. Assays for developmental potency are considered to be crucial for concluding that an iPSC is pluripotent. In mice, chimera formation and germline contribution are routinely used to assess the developmental potential of iPSCs. The production of ‘‘all-iPSC mice’’ by tetraploid complementation recently demonstrated that iPSCs
can indeed have the same developmental potential as ESCs (Zhao et al., 2009). In the human system, however, researchers are restricted to using less stringent functional assays, such as in vitro differentiation and teratoma formation. Although teratoma formation is considered to be a prerequisite for designating human cells as pluripotent, it should be acknowledged that this is a qualitative test that does not allow for easy quantification of differentiation. Compared to functional tests for pluripotency, molecular analyses allow for more quantitative comparisons of iPSCs and ESCs. Numerous studies indicate that, at least for some clones, iPSCs are similar if not indistinguishable from ESCs derived from embryo or nuclear transfer experiments. These include profiling of global gene expression, modifications of histone tails, the state of X chromosome inactivation, and profiles of DNA methylation (Mikkelsen et al., 2008). However, some studies using global expression analyses concluded that iPSCs are a unique subtype of pluripotent cells that retain a consistent gene expression signature distinguishable from ESCs, even after extended passaging (Chin et al., 2009). Yet, the reanalysis of a large collection of gene expression and histone modification data lead to the conclusion that small variations between human iPSCs and ESCs in chromatin structure and global gene expression may constitute experimental ‘‘noise’’ and do not reflect a consistent signature that distinguishes iPSCs from ESCs (Guenther et al., 2010; Newman and Cooper, 2010). Two recent studies concluded that the only distinguishable difference between ESCs and the vast majority of iPSCs was the abnormal reduction in the expression of the maternally imprinted Dlk1-Dio3 locus and that this expression difference was the underlying cause for the inability to generate all-iPS mice by tetraploid complementation (4n). In contrast, 4n-competent iPSC lines showed normal allelic imprinting at this locus (Liu et al., 2010; Stadtfeld et al., 2010). Abnormal expression of this cluster was not observed in human iPSCs (Stadtfeld et al., 2010). Although these studies represent interesting correlations, the conclusions need to be reconciled with the observation that mice with bi-allelic deletion in components of the Dlk1-Dio3 locus (e.g., Gtl2) are viable (Takahashi et al., 2009). Moreover, iPSC clones derived from the same transgenic donor mouse system (Stadtfeld et al., 2010) were later reported to display additional global perturbations in transcriptional patterns, depending on the cell of origin (Polo et al., 2010). It remains to be clarified whether such specific gene expression signatures observed in early passage iPSC lines (Polo et al., 2010) represent ‘‘epigenetic memory’’ or simply result from residual transgene induction levels that are specific to the cell of origin and that induce perturbations in gene expression (Soldner et al., 2009), which may subside with silencing of the transgenes upon extended cell passaging. Another study compared the patterns of global DNA methylation and in vitro differentiation of early passage iPSCs derived from B lymphocytes or fibroblasts with those of ESCs. This study concluded that reprogramming with transcription factors can leave an epigenetic memory mark in iPSCs reminiscent of the donor cell type (Kim et al., 2010). In contrast, such patterns were not seen in ESCs derived after nuclear transfer, suggesting that nuclear transfer might reset the epigenetic characteristics of
somatic cells more effectively than reprogramming in vitro with transcription factors. However, another explanation for these results is that the nuclear transfer-derived ESCs used in this study (Kim et al., 2010), but not iPSCs, were obtained in the presence of ERK inhibitors, which can facilitate complete reprogramming (Ying et al., 2008; Silva et al., 2009). Further, the fibroblast-derived iPSC lines used in this study showed only partial demethylation of the endogenous Nanog promoter, consistent with incomplete reprogramming (Kim et al., 2010). Finally, iPSCs derived from patients with fragile X syndrome exhibited a phenotype that was not recapitulated in ESCs carrying the same mutation (Urbach et al., 2010). Fragile X syndrome is a common form of inherited mental retardation caused by an expansion of CGG-triplet repeats in the 50 untranslated region of the FMR1 gene, which leads to its transcriptional silencing. Interestingly, in fragile X-ESCs derived from blastocysts, the full expansion of the CGG-triplet repeat did not inactivate the FMR1 gene and silencing occurred only after differentiation. However, upon in vitro reprogramming of fragile X fibroblasts, the FMR1 locus remained inactive and was not reset to the transcriptional active state, demonstrating that in vitro reprogramming does not always faithfully reset the epigenetic state of the somatic cell to that of ESCs. It would be interesting to investigate whether this methylation pattern is lost with extended passaging of the iPSC lines. Many of the studies summarized above suggested that somatic cells can be reprogrammed to a pluripotent state, which is molecularly and biologically indistinguishable from that of ESCs and compatible with the generation of all-iPSC mice by tetraploid complementation. However, in some circumstances subtle differences, which are inconsistent and often transient, can also be observed. Evaluating the frequency and origin of altered expression patterns in iPSCs is of biological and clinical importance. Unfortunately, a number of methodological limitations, which are known to affect the state of pluripotency, complicate a meaningful comparison of iPSCs and ESCs, and they must be simultaneously controlled for. As summarized in Table 2, such parameters include the presence and incomplete silencing of transgenes, different combinations of reprogramming factors used to induce iPSCs, natural heterogeneity that exists between different pluripotent ESC lines, incomplete reprogramming in early passage cell lines, and the genetic background of the cells. Also, even low levels of basal vector expression have been shown to significantly affect the global gene expression pattern of undifferentiated cells (Soldner et al., 2009). This is important because for some iPSCs derived from patients, a disease-specific in vitro phenotype has been reported (Carvajal-Vergara et al., 2010; Ebert et al., 2009). These iPSCs were generated with constitutively expressed lentivirusor Moloney virus-based vectors. Therefore, variable and undefined levels of basal vector expression may have influenced the phenotype. An unresolved question is whether the subtle epigenetic attributes resulting from incomplete reprogramming of the somatic donor nucleus have a meaningful and functional significance for the developmental potential of iPSCs. Commonly referred to as transient epigenetic memory, these technical attributes of incomplete reprogramming can be erased following additional Cell 143, November 12, 2010 ª2010 Elsevier Inc. 519
Table 2. Parameters that May Affect Gene Expression and Biological Characteristics of Pluripotent Stem Cells Parameter
Factors Influencing the Properties of Induced Pluripotent Cell States
Considerations for Experimental Design
Transgene-containing iPSCs
The presence and incomplete silencing of reprogramming transgenes commonly used in generating iPSCs can perturb the identity and functionality of the induced cells.
Generation of vector-free reprogrammed cells.
Genetic background
The gene expression pattern and in vivo developmental competency can vary between different mouse strains (e.g., 129 strain versus the nonobese diabetic strain).
Comparison of ESCs and iPSCs from an identical genetic background.
Incomplete reprogramming
Direct reprogramming involves several cell divisions.
Analysis of fully reprogrammed lines that have achieved ample cell divisions after transduction of the reprogramming factors.
In vitro molecular heterogeneity among ESCs
Culture adaptation of cell lines may be the source for heterogeneity; ESC clones generated from identical genetic backgrounds can display interclonal variability.
Comparison of several independent ESC and iPSC lines from genetically identical backgrounds grown in the same growth conditions.
Reprogramming factor combinations
iPSCs can be generated by transduction of different combinations of reprogramming factors or small molecules, and this may affect the epigenetic characteristics of the iPSCs.
Inclusion of cell lines derived through different combinations of reprogramming factor or in ‘‘2i’’ conditions.
A number of constraints may affect the epigenetic state and biology of iPSCs; controlling for these parameters may facilitate the comparison between iPSCs and ESCs and result in more reliable characterization of the pluripotent state.
cell division or by supplementing other exogenous factors during reprogramming (Silva et al., 2009; Polo et al., 2010; Kim et al., 2010). Further, aberrant and variable imprinting are evident in cloned mice and in ESC lines derived from embryos (Humpherys et al., 2001), indicating that even substantial deregulation of genes still allows development to birth and beyond. Thus, differentiation to functional cells may be rather tolerant of epigenetic aberrations of the genome and subtle abnormalities in gene expression, and clearly, it will be important to establish criteria and minimal requirements that define the safety of iPSCs for clinical applications and disease research. Transdifferentiation of Somatic Cells Reprogramming of somatic cells into iPSCs requires the resetting of the epigenetic state from a somatic to a pluripotent embryonic state, which can be achieved even without using drug resistance to select for the activation of marked pluripotency genes. Thus, the pluripotent epigenetic state may represent a default cellular state easily captured in tissue culture. One explanation for this observation is that iPSCs may have a growth advantage over the somatic cells. However, another possibility is that the gene expression circuitry of pluripotency is a ‘‘ground’’ or default state that is the most stable state after erasing the somatic cell identity and thus reflects a state with a ‘‘minimal’’ epigenetic dominance of somatic programs. Such hypotheses raise the question of whether cells can be induced to ‘‘transdifferentiate’’ directly into another state of differentiation. Here transdifferentiation is defined as the direct conversion of one somatic cell type into another type without first reprogramming into pluripotent cells and then differentiating into 520 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
functional somatic cells (Graf and Enver, 2009). As outlined in Table 3, it is useful to define the extent of the transdifferentiation but also to distinguish between transdifferentiation within a lineage with those between lineages of different germ layer origins. One of the first examples of transdifferentiation within the same germ layer was the conversion of fibroblasts into muscle cells by overexpressing MyoD (Weintraub et al., 1989). However, activation of the endogenous MyoD gene was not detected in the muscle cells, suggesting that conversion was incomplete and that the maintenance of the myogenic phenotype depended on the transgene. Recently, cardiac fibroblasts were elegantly converted into cardiomyocyte-like cells by the ectopic expression of Gata4, Mef2C, and Tbx5 (Ieda et al., 2010). The reprogrammed cells did not depend on expression of the transgenes but exhibited transcriptional and functional differences from neonatal cardiomyocytes, particularly when the cells were derived from dermal fibroblasts. Transdifferentiation has also been achieved within the hematopoietic lineage. Ectopic expression of the transcription factor C/EBPa converted lymphocytes to macrophage-like cells (Xie et al., 2004). However, the induced macrophage-like cells continued to express markers specific to Pro-B cells and failed to activate several macrophage-specific markers. Similarly, in vitro transdifferentiation between different germ layers was achieved by the ectopic expression of MITF (microphthalmia-associated transcription factor), which converted fibroblasts into melanocyte-like cells that synthesized melanin through the activation of direct downstream targets of MITF, such as tyrosinase (Tachibana et al., 1996). However, the
Table 3. Transdifferentiation between Somatic Cell States Somatic Cell Conversion (with Germ Layer)
Exogenous Reprogramming Factors
Experimental Setting
Complete Molecular Epigenetic Reprogramming
The Dependency of the New State on the Transgene
Ref
Fibroblasts (mesoderm) Converted to myocyte-like cells (mesoderm)
MyoD + 5-AzaC
–in vitro –intralineage conversion
No: Failed to reactivate endogenous MyoD and other myogenic markers
Dependent
[1]
B cells, T cells, and fibroblasts (mesoderm) Converted to macrophage-like cells (mesoderm)
C/EBPa ±PU.1
–in vitro –intralineage conversion
No: Failed to reactivate several macrophage-expressed genes; failed to suppress some donor cell somatic markers
Independent
[2]
Cardiac fibroblasts (mesoderm) Converted to induced cardiac myocte-like cells (mesoderm)
Gata4, Mef2c, and Tbx5
–in vitro –intralineage conversion
No: Gene expression signature distinguishable from neonatal cardiomyocytes; Induced cardiac myocte-like cells derived from dermal fibroblast have significantly limited functionality
Independent
[3]
Fibroblasts (mesoderm) Converted to Melanocyte-like cells (ectoderm)
MITF
–in vitro –cross-lineage conversion
No: Failed to reactivate several melanocyte-expressed genes; failed to suppress donor cell markers
Not determined
[4]
Fibroblasts (mesoderm) Converted to induced Neuron-like (iN) cells (ectoderm)
Ascl1, Brn2, and Mytl1
–in vitro –cross-lineage conversion
Not determined
Not determined
[5]
B cells (mesoderm) Converted to common lymphoid progenitors, macrophages, and T cells (mesoderm)
Pax5 deletion
–in vivo (and in vitro) –intralineage conversion
Not determined; Functional in vivo hematopoietic reconstitution
Independent
[6]
Pancreas exocrine cells (endoderm) Converted to endocrine-like cells (endoderm)
Ngn3, Pdx1, and MafA
–in vivo –intralineage conversion
Not determined; Functional insulin-producing cells capable of improving hyperglycemia
Independent
[7]
Granulosa and Theca cells (mesoderm) Converted to Sertoli and Leydig cells (mesoderm)
Foxl2 deletion
–in vivo –intralineage conversion
Not determined; Functional male hormonal production
Independent
[8]
References: [1] Weintraub et al., 1989; [2] Xie et al., 2004; [3] Ieda et al., 2010; [4] Tachibana et al., 1996; [5] Vierbuchen et al., 2010; [6] Cobaleda et al., 2007; [7] Zhou et al., 2008; [8] Uhlenhaut et al., 2009.
resultant cells retained expression of fibroblast markers and had reduced expression of most melanocyte markers, suggesting that the partial transdifferentiation predominantly resulted from activation of direct targets of MITF. In a more recent study (Vierbuchen et al., 2010), particular neural transcription factors converted mouse fibroblasts into neuron-like cells. The induced neural cells shared essential features with functional neurons, including morphological characteristics, expression of cortical markers, the generation
of action potentials, and the formation of synapses. However, it has not been resolved whether the induced neural cells have silenced fibroblast-specific genes and maintain their newly acquired state independent of the expression of the transgenes. Overall, several of these directed transdifferentiation experiments in vitro provide strong evidence that the induced cells exhibit aberrant gene expression patterns and incomplete reprogramming into the new lineage. Although such in vitro derived cells may prove valuable in the future, further Cell 143, November 12, 2010 ª2010 Elsevier Inc. 521
investigation is needed to determine whether a somatic gene expression program can be extensively and completely reinstated on another somatic program from a different lineage without first going through the ground state of pluripotency. The deletion of the Pax5 transcription factor induced dedifferentiation of B cells into common lymphoid progenitor cells. Indeed, these cells were able to reconstitute the entire lineage of T lymphocytes in mice (Cobaleda et al., 2007). Finally, inducible deletion of Foxl2 in adult ovarian follicles led to the upregulation of testis-specific genes, resulting in reprogramming of granulosa and theca cells into functional Sertoli-like and Leydig-like cells, respectively, which were capable of expressing normal testosterone levels (Uhlenhaut et al., 2009). Researchers have demonstrated that the expression of key transcription factors can convert one cell type into a developmentally related cell type inside an animal even in the absence of cell proliferation. This was most clearly demonstrated by the conversion of exocrine into endocrine pancreatic cells (Zhou et al., 2008). Transdifferentiation in the absence of DNA replication may be consistent with the notion that changes in the expression of transcription factors predominantly drives the conversion between cells states that are within the same lineage and closely related developmentally. Thus, these types of conversions may require only a limited amount of resetting of DNA or chromatin modifications. This appears to be different from direct reprogramming of somatic cells to a pluripotent state, which entails extensive epigenetic resetting and requires multiple rounds of DNA replication. Finally, it should be mentioned that multiple studies have claimed that somatic cells, such as bone marrow cells, can be transdifferentiated into cells of other lineages merely by culturing the cells under specific conditions. As discussed elsewhere, the evidence for such claims has so far been unconvincing (Wagers and Weissman, 2004). Concluding Remarks Although rapid progress in our understanding of pluripotency and stem cells has been made, a number of important questions and technical hurdles remain. These include the stabilization of the various pluripotent states in cell cultures derived from different species. It is particularly important to define culture conditions that robustly and stably maintain the naive pluripotent state and then use this technology to derive and fully characterize naive iPSCs and ESCs from human embryos. Current direct reprogramming approaches are inefficient and involve stochastic changes occurring in highly heterogeneous cell populations. Because only a small fraction of cells will ever form an iPSC, the information gained from molecular analysis of the heterogeneous intermediate cell populations is limited. Such analyses cannot distinguish between the rate-limiting and non-rate-limiting epigenetic changes in those cells. To understand the complex epigenetic remodeling that precedes iPSC formation, we need to establish new experimental and theoretical approaches that allow for molecular analyses at the singlecell level. It is likely that the generation of patient-specific iPSCs will have a significant impact on the study of human diseases and on regenerative medicine. However, a number of technical 522 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
issues need to be resolved before the technology can be used in a clinical setting (Saha and Jaenisch, 2009). These include the establishment of efficient reprogramming strategies that do not result in genetically modified cells. Although the approaches currently available for generating genetically unmodified iPSC are inefficient, we expect that these technical hurdles will be resolved soon. In addition, one of the key challenges for translating these new technologies to the clinic is devising robust protocols for differentiating ESCs or iPSCs to self-renewing adult stem cells and lineage-committed cells. Armed with such protocols, researchers can then begin to define experimental conditions that allow the development and detection of relevant in vitro phenotypes for a given human disease, putting ‘‘personalized’’ regenerative medicine just over our horizon. ACKNOWLEDGMENTS We thank Hillel and Liliana Bachrach, Susan Whitehead, and Landon Clay for their generous gifts supporting research conducted in our laboratory. We thank R. Young and members of the Jaenisch lab for discussions. J.H.H. is supported by a Genzyme Fellowship and a Helen Hay Whitney Foundation Fellowship; K.S. by a Society in Science: Branco-Weiss fellowship. R.J. is a cofounder of Fate Therapeutics and an adviser to Stemgent. We apologize to authors whose work has not been covered or directly cited due to space limitations. REFERENCES Bao, S., Tang, F., Li, X., Hayashi, K., Gillich, A., Lao, K., and Surani, M.A. (2009). Epigenetic reversion of post-implantation epiblast to pluripotent embryonic stem cells. Nature 461, 1292–1295. Bhutani, N., Brady, J.J., Damian, M., Sacco, A., Corbel, S.Y., and Blau, H.M. (2009). Reprogramming towards pluripotency requires AID-dependent DNA demethylation. Nature 463, 1042–1047. Bloom, J.D., Meyer, M.M., Meinhold, P., Otey, C.R., MacMillan, D., and Arnold, F.H. (2005). Evolving strategies for enzyme engineering. Curr. Opin. Struct. Biol. 15, 447–452. Boiani, M., Eckardt, S., Scholer, H.R., and McLaughlin, K.J. (2002). Oct4 distribution and level in mouse clones: consequences for pluripotency. Genes Dev. 16, 1209–1219. Boland, M.J., Hazen, J.L., Nazor, K.L., Rodriguez, A.R., Gifford, W., Martin, G., Kupriyanov, S., and Baldwin, K.K. (2009). Adult mice generated from induced pluripotent stem cells. Nature 461, 91–94. Boyer, L.A., Lee, T.I., Cole, M.F., Johnstone, S.E., Levine, S.S., Zucker, J.P., Guenther, M.G., Kumar, R.M., Murray, H.L., Jenner, R.G., et al. (2005). Core transcriptional regulatory circuitry in human embryonic stem cells. Cell 122, 947–956. Brons, I.G., Smithers, L.E., Trotter, M.W., Rugg-Gunn, P., Sun, B., Chuva de Sousa Lopes, S.M., Howlett, S.K., Clarkson, A., Ahrlund-Richter, L., Pedersen, R.A., et al. (2007). Derivation of pluripotent epiblast stem cells from mammalian embryos. Nature 448, 191–195. Buecker, C., Chen, H.H., Polo, J.M., Daheron, L., Bu, L., Barakat, T.S., Okwieka, P., Porter, A., Gribnau, J., Hochedlinger, K., et al. (2010). A murine ESC-like state facilitates transgenesis and homologous recombination in human pluripotent stem cells. Cell Stem Cell 6, 535–546. Buehr, M., Meek, S., Blair, K., Yang, J., Ure, J., Silva, J., McLay, R., Hall, J., Ying, Q.L., and Smith, A. (2008). Capture of authentic embryonic stem cells from rat blastocysts. Cell 135, 1287–1298. Carvajal-Vergara, X., Sevilla, A., D’Souza, S.L., Ang, Y.S., Schaniel, C., Lee, D.F., Yang, L., Kaplan, A.D., Adler, E.D., Rozov, R., et al. (2010). Patientspecific induced pluripotent stem-cell-derived models of LEOPARD syndrome. Nature 465, 808–812.
Chen, X., Xu, H., Yuan, P., Fang, F., Huss, M., Vega, V.B., Wong, E., Orlov, Y.L., Zhang, W., Jiang, J., et al. (2008). Integration of external signaling pathways with the core transcriptional network in embryonic stem cells. Cell 133, 1106–1117. Chin, M.H., Mason, M.J., Xie, W., Volinia, S., Singer, M., Peterson, C., Ambartsumyan, G., Aimiuwu, O., Richter, L., Zhang, J., et al. (2009). Induced pluripotent stem cells and embryonic stem cells are distinguished by gene expression signatures. Cell Stem Cell 5, 111–123. Chou, Y.F., Chen, H.H., Eijpe, M., Yabuuchi, A., Chenoweth, J.G., Tesar, P., Lu, J., McKay, R.D., and Geijsen, N. (2008). The growth factor environment defines distinct pluripotent ground states in novel blastocyst-derived stem cells. Cell 135, 449–461. Cobaleda, C., Jochum, W., and Busslinger, M. (2007). Conversion of mature B cells into T cells by dedifferentiation to uncommitted progenitors. Nature 449, 473–477. Conrad, S., Renninger, M., Hennenlotter, J., Wiesner, T., Just, L., Bonin, M., Aicher, W., Buhring, H.J., Mattheus, U., Mack, A., et al. (2008). Generation of pluripotent stem cells from adult human testis. Nature 456, 344–349. Ebert, A.D., Yu, J., Rose, F.F., Jr., Mattis, V.B., Lorson, C.L., Thomson, J.A., and Svendsen, C.N. (2009). Induced pluripotent stem cells from a spinal muscular atrophy patient. Nature 457, 277–280. Eminli, S., Foudi, A., Stadtfeld, M., Maherali, N., Ahfeldt, T., Mostoslavsky, G., Hock, H., and Hochedlinger, K. (2009). Differentiation stage determines potential of hematopoietic cells for reprogramming into induced pluripotent stem cells. Nat. Genet. 41, 968–976. Evans, M.J., and Kaufman, M.H. (1981). Establishment in culture of pluripotential cells from mouse embryos. Nature 292, 154–156. Goldberg, A.D., Allis, C.D., and Bernstein, E. (2007). Epigenetics: A landscape takes shape. Cell 128, 635–638. Graf, T., and Enver, T. (2009). Forcing cells to change lineages. Nature 462, 587–594. Guenther, M.G., Frampton, G.M., Soldner, F., Hockemeyer, D., Mitalipova, M., Jaenisch, R., and Young, R.A. (2010). Chromatin structure and gene expression programs of human embryonic and induced pluripotent stem cells. Cell Stem Cell 7, 249–257. Guo, G., Yang, J., Nichols, J., Hall, J.S., Eyres, I., Mansfield, W., and Smith, A. (2009). Klf4 reverts developmentally programmed restriction of ground state pluripotency. Development 136, 1063–1069. Gurdon, J.B. (1962). The transplantation of nuclei between two species of Xenopus. Dev. Biol. 5, 68–83. Hajkova, P., Jeffries, S.J., Lee, C., Miller, N., Jackson, S.P., and Surani, M.A. (2010). Genome-wide reprogramming in the mouse germ line entails the base excision repair pathway. Science 329, 78–82. Hanna, J., Wernig, M., Markoulaki, S., Sun, C.W., Meissner, A., Cassady, J.P., Beard, C., Brambrink, T., Wu, L.C., Townes, T.M., et al. (2007). Treatment of sickle cell anemia mouse model with iPS cells generated from autologous skin. Science 318, 1920–1923. Hanna, J., Markoulaki, S., Schorderet, P., Carey, B.W., Beard, C., Wernig, M., Creyghton, M.P., Steine, E.J., Cassady, J.P., Foreman, R., et al. (2008). Direct reprogramming of terminally differentiated mature B lymphocytes to pluripotency. Cell 133, 250–264. Hanna, J., Markoulaki, S., Mitalipova, M., Cheng, A.W., Cassady, J.P., Staerk, J., Carey, B.W., Lengner, C.J., Foreman, R., Love, J., et al. (2009a). Metastable pluripotent states in NOD-mouse-derived ESCs. Cell Stem Cell 4, 513–524. Hanna, J., Saha, K., Pando, B., van Zon, J., Lengner, C., Creyghton, M., van Oudenaarden, A., and Jaenisch, R. (2009b). Direct cell reprogramming is a stochastic process amenable to acceleration. Nature 462, 595–601. Hanna, J., Cheng, A.W., Saha, K., Kim, J., Lengner, C.J., Soldner, F., Cassady, J.P., Muffat, J., Carey, B.W., and Jaenisch, R. (2010). Human embryonic stem cells with biological and epigenetic characteristics similar to those of mouse ESCs. Proc. Natl. Acad. Sci. USA 107, 9222–9227. Hasegawa, K., Zhang, P., Wei, Z., Pomeroy, J.E., Lu, W., and Pera, M.F. (2010). Comparison of reprogramming efficiency between transduction of
reprogramming factors, cell-cell fusion, and cytoplast fusion. Stem Cells 28, 1338–1348. Hayashi, K., Lopes, S.M., Tang, F., and Surani, M.A. (2008). Dynamic equilibrium and heterogeneity of mouse pluripotent stem cells with distinct functional and epigenetic states. Cell Stem Cell 3, 391–401. Hockemeyer, D., Soldner, F., Beard, C., Gao, Q., Mitalipova, M., Dekelver, R.C., Katibah, G.E., Amora, R., Boydston, E.A., Zeitler, B., et al. (2009). Efficient targeting of expressed and silent genes in human ESCs and iPSCs using zinc-finger nucleases. Nat. Biotechnol. 27, 851–857. Hogan, B.L. (1976). Changes in the behaviour of teratocarcinoma cells cultivated in vitro. Nature 263, 136–137. Huang, S. (2009). Reprogramming cell fates: reconciling rarity with robustness. Bioessays 31, 546–560. Humpherys, D., Eggan, K., Akutsu, H., Hochedlinger, K., Rideout, W.M., 3rd, Biniszkiewicz, D., Yanagimachi, R., and Jaenisch, R. (2001). Epigenetic instability in ES cells and cloned mice. Science 293, 95–97. Ichida, J., Blanchard, J., Lam, K., Son, E., Chung, J., Egli, D., Loh, K., Carter, A., Di Giorgio, F., Koszka, K., et al. (2009). A small-molecule inhibitor of Tgfbeta signaling replaces Sox2 in reprogramming by inducing Nanog. Cell Stem Cell 5, 491–503. Ieda, M., Fu, J.D., Delgado-Olguin, P., Vedantham, V., Hayashi, Y., Bruneau, B.G., and Srivastava, D. (2010). Direct reprogramming of fibroblasts into functional cardiomyocytes by defined factors. Cell 142, 375–386. Jaenisch, R., and Young, R. (2008). Stem cells, the molecular circuitry of pluripotency and nuclear reprogramming. Cell 132, 567–582. Kanatsu-Shinohara, M., Inoue, K., Lee, J., Yoshimoto, M., Ogonuki, N., Miki, H., Baba, S., Kato, T., Kazuki, Y., Toyokuni, S., et al. (2004). Generation of pluripotent stem cells from neonatal mouse testis. Cell 119, 1001–1012. Kang, L., Wang, J., Zhang, Y., Kou, Z., and Gao, S. (2009). iPS cells can support full-term development of tetraploid blastocyst-complemented embryos. Cell Stem Cell 5, 135–138. Kim, K., Doi, A., Wen, B., Ng, K., Zhao, R., Cahan, P., Kim, J., Aryee, M., Ji, H., Ehrlich, L., et al. (2010). Epigenetic memory in induced pluripotent stem cells. Nature 467, 285–290. Ko, K., Arauzo-Bravo, M.J., Tapia, N., Kim, J., Lin, Q., Bernemann, C., Han, D.W., Gentile, L., Reinhardt, P., Greber, B., et al. (2010). Human adult germline stem cells in question. Nature 465, E1. Krizhanovsky, V., and Lowe, S.W. (2009). Stem cells: The promises and perils of p53. Nature 460, 1085–1086. Lengner, C.J., Gimelbrant, A.A., Erwin, J.A., Cheng, A.W., Guenther, M.G., Welstead, G.G., Alagappan, R., Frampton, G.M., Xu, P., Muffat, J., et al. (2010). Derivation of pre-X inactivation human embryonic stem cells under physiological oxygen concentrations. Cell 141, 872–883. Liu, L., Luo, G.Z., Yang, W., Zhao, X., Zheng, Q., Lv, Z., Li, W., Wu, H.J., Wang, L., Wang, X.J., et al. (2010). Activation of the imprinted Dlk1-Dio3 region correlates with pluripotency levels of mouse stem cells. J. Biol. Chem. 285, 19483– 19490. Loh, Y.H., Wu, Q., Chew, J.L., Vega, V.B., Zhang, W., Chen, X., Bourque, G., George, J., Leong, B., Liu, J., et al. (2006). The Oct4 and Nanog transcription network regulates pluripotency in mouse embryonic stem cells. Nat. Genet. 38, 431–440. Loh, Y.H., Hartung, O., Li, H., Guo, C., Sahalie, J.M., Manos, P.D., Urbach, A., Heffner, G.C., Grskovic, M., Vigneault, F., et al. (2010). Reprogramming of T cells from human peripheral blood. Cell Stem Cell 7, 15–19. Lu, R., Markowetz, F., Unwin, R.D., Leek, J.T., Airoldi, E.M., Macarthur, B.D., Lachmann, A., Rozov, R., Ma’ayan, A., Boyer, L.A., et al. (2009). Systems-level dynamic analyses of fate change in murine embryonic stem cells. Nature 462, 358–362. Maherali, N., and Hochedlinger, K. (2009). Tgfbeta signal inhibition cooperates in the induction of iPSCs and replaces Sox2 and cMyc. Curr. Biol. 19, 1718– 1723.
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 523
Marson, A., Foreman, R., Chevalier, B., Bilodeau, S., Kahn, M., Young, R.A., and Jaenisch, R. (2008). Wnt signaling promotes reprogramming of somatic cells to pluripotency. Cell Stem Cell 3, 132–135.
Silva, J., Nichols, J., Theunissen, T.W., Guo, G., van Oosten, A.L., Barrandon, O., Wray, J., Yamanaka, S., Chambers, I., and Smith, A. (2009). Nanog is the gateway to the pluripotent ground state. Cell 138, 722–737.
Meissner, A., Mikkelsen, T.S., Gu, H., Wernig, M., Hanna, J., Sivachenko, A., Zhang, X., Bernstein, B.E., Nusbaum, C., Jaffe, D.B., et al. (2008). Genomescale DNA methylation maps of pluripotent and differentiated cells. Nature 454, 766–770.
Silva, S.S., Rowntree, R.K., Mekhoubad, S., and Lee, J.T. (2008). X-chromosome inactivation and epigenetic fluidity in human embryonic stem cells. Proc. Natl. Acad. Sci. USA 105, 4820–4825.
Mikkelsen, T.S., Hanna, J., Zhang, X., Ku, M., Wernig, M., Schorderet, P., Bernstein, B.E., Jaenisch, R., Lander, E.S., and Meissner, A. (2008). Dissecting direct reprogramming through integrative genomic analysis. Nature 454, 49–55. Newman, A.M., and Cooper, J.B. (2010). Lab-specific gene expression signatures in pluripotent stem cells. Cell Stem Cell 7, 258–262. Nichols, J., and Smith, A. (2009). Naive and primed pluripotent states. Cell Stem Cell 4, 487–492. Nichols, J., Zevnik, B., Anastassiadis, K., Niwa, H., Klewe-Nebenius, D., Chambers, I., Scholer, H., and Smith, A. (1998). Formation of pluripotent stem cells in the mammalian embryo depends on the POU transcription factor Oct4. Cell 95, 379–391. Ohta, H., Ohinata, Y., Ikawa, M., Morioka, Y., Sakaide, Y., Saitou, M., Kanagawa, O., and Wakayama, T. (2009). Male germline and embryonic stem cell lines from NOD mice: efficient derivation of GS cells from a nonpermissive strain for ES cell derivation. Biol. Reprod. 81, 1147–1153. Okita, K., Ichisaka, T., and Yamanaka, S. (2007). Generation of germlinecompetent induced pluripotent stem cells. Nature 448, 313–317. Okita, K., Nakagawa, M., Hyenjong, H., Ichisaka, T., and Yamanaka, S. (2008). Generation of mouse induced pluripotent stem cells without viral vectors. Science 322, 949–953. Osafune, K., Caron, L., Borowiak, M., Martinez, R.J., Fitz-Gerald, C.S., Sato, Y., Cowan, C.A., Chien, K.R., and Melton, D.A. (2008). Marked differences in differentiation propensity among human embryonic stem cell lines. Nat. Biotechnol. 26, 313–315. Pereira, C.F., Terranova, R., Ryan, N.K., Santos, J., Morris, K.J., Cui, W., Merkenschlager, M., and Fisher, A.G. (2008). Heterokaryon-based reprogramming of human B lymphocytes for pluripotency requires Oct4 but not Sox2. PLoS Genet. 4, e1000170. Polo, J.M., Liu, S., Figueroa, M.E., Kulalert, W., Eminli, S., Tan, K.Y., Apostolou, E., Stadtfeld, M., Li, Y., Shioda, T., et al. (2010). Cell type of origin influences the molecular and functional properties of mouse induced pluripotent stem cells. Nat. Biotechnol. 28, 848–855. Popp, C., Dean, W., Feng, S., Cokus, S.J., Andrews, S., Pellegrini, M., Jacobsen, S.E., and Reik, W. (2010). Genome-wide erasure of DNA methylation in mouse primordial germ cells is affected by AID deficiency. Nature 463, 1101–1105. Raj, A., and van Oudenaarden, A. (2008). Nature, nurture, or chance: stochastic gene expression and its consequences. Cell 135, 216–226. Saha, K., and Jaenisch, R. (2009). Technical challenges in using human induced pluripotent stem cells to model disease. Cell Stem Cell 5, 584–595. Samavarchi-Tehrani, P., Golipour, A., David, L., Sung, H.-k., Beyer, T.A., Datti, A., Woltjen, K., Nagy, A., and Wrana, J.L. (2010). Functional genomics reveals a BMP-driven mesenchymal-to-epithelial transition in the initiation of somatic cell reprogramming. Cell Stem Cell 7, 64–77. Santos, F., Hendrich, B., Reik, W., and Dean, W. (2002). Dynamic reprogramming of DNA methylation in the early mouse embryo. Dev. Biol. 241, 172–182. Seki, T., Yuasa, S., Oda, M., Egashira, T., Yae, K., Kusumoto, D., Nakata, H., Tohyama, S., Hashimoto, H., Kodaira, M., et al. (2010). Generation of induced pluripotent stem cells from human terminally differentiated circulating T cells. Cell Stem Cell 7, 11–14. Shamblott, M.J., Axelman, J., Wang, S., Bugg, E.M., Littlefield, J.W., Donovan, P.J., Blumenthal, P.D., Huggins, G.R., and Gearhart, J.D. (1998). Derivation of pluripotent stem cells from cultured human primordial germ cells. Proc. Natl. Acad. Sci. USA 95, 13726–13731.
524 Cell 143, November 12, 2010 ª2010 Elsevier Inc.
Singhal, N., Graumann, J., Wu, G., Arauzo-Bravo, M.J., Han, D.W., Greber, B., Gentile, L., Mann, M., and Scholer, H.R. (2010). Chromatin-remodeling components of the BAF complex facilitate reprogramming. Cell 141, 943–955. Smith, Z.D., Nachman, I., Regev, A., and Meissner, A. (2010). Dynamic singlecell imaging of direct reprogramming reveals an early specifying event. Nat. Biotechnol. 28, 521–526. Soldner, F., Hockemeyer, D., Beard, C., Gao, Q., Bell, G.W., Cook, E.G., Hargus, G., Blak, A., Cooper, O., Mitalipova, M., et al. (2009). Parkinson’s disease patient-derived induced pluripotent stem cells free of viral reprogramming factors. Cell 136, 964–977. Sridharan, R., Tchieu, J., Mason, M.J., Yachechko, R., Kuoy, E., Horvath, S., Zhou, Q., and Plath, K. (2009). Role of the murine reprogramming factors in the induction of pluripotency. Cell 136, 364–377. Stadtfeld, M., Nagaya, M., Utikal, J., Weir, G., and Hochedlinger, K. (2008). Induced pluripotent stem cells generated without viral integration. Science 322, 945–949. Stadtfeld, M., Apostolou, E., Akutsu, H., Fukuda, A., Follett, P., Natesan, S., Kono, T., Shioda, T., and Hochedlinger, K. (2010). Aberrant silencing of imprinted genes on chromosome 12qF1 in mouse induced pluripotent stem cells. Nature 465, 175–181. Staerk, J., Dawlaty, M.M., Gao, Q., Maetzel, D., Hanna, J., Sommer, C.A., Mostoslavsky, G., and Jaenisch, R. (2010). Reprogramming of human peripheral blood cells to induced pluripotent stem cells. Cell Stem Cell 7, 20–24. Surani, M.A. (1999). Reprogramming a somatic nucleus by trans-modification activity in germ cells. Semin. Cell Dev. Biol. 10, 273–277. Tachibana, M., Takeda, K., Nobukuni, Y., Urabe, K., Long, J., Meyers, K., Aaronson, S., and Miki, T. (1996). Ectopic expression of MITF, a gene for Waardenburg syndrome type 2, converts fibroblasts to cells with melanocyte characteristics. Nat. Genet. 14, 50–54. Takahashi, K., and Yamanaka, S. (2006). Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell 126, 663–676. Takahashi, N., Okamoto, A., Kobayashi, R., Shirai, M., Obata, Y., Ogawa, H., Sotomaru, Y., and Kono, T. (2009). Deletion of Gtl2, imprinted non-coding RNA, with its differentially methylated region induces lethal parent-origindependent defects in mice. Hum. Mol. Genet. 18, 1879–1888. Tang, F., Barbacioru, C., Bao, S., Lee, C., Nordman, E., Wang, X., Lao, K., and Surani, M.A. (2010). Tracing the derivation of embryonic stem cells from the inner cell mass by single-cell RNA-Seq analysis. Cell Stem Cell 6, 468–478. Tesar, P.J., Chenoweth, J.G., Brook, F.A., Davies, T.J., Evans, E.P., Mack, D.L., Gardner, R.L., and McKay, R.D. (2007). New cell lines from mouse epiblast share defining features with human embryonic stem cells. Nature 448, 196–199. Thomson, J.A., Itskovitz-Eldor, J., Shapiro, S.S., Waknitz, M.A., Swiergiel, J.J., Marshall, V.S., and Jones, J.M. (1998). Embryonic stem cell lines derived from human blastocysts. Science 282, 1145–1147. Uhlenhaut, N.H., Jakob, S., Anlag, K., Eisenberger, T., Sekido, R., Kress, J., Treier, A.C., Klugmann, C., Klasen, C., Holter, N.I., et al. (2009). Somatic sex reprogramming of adult ovaries to testes by FOXL2 ablation. Cell 139, 1130– 1142. Urbach, A., Bar-Nur, O., Daley, G.Q., and Benvenisty, N. (2010). Differential modeling of fragile X syndrome by human embryonic stem cells and induced pluripotent stem cells. Cell Stem Cell 6, 407–411. Vierbuchen, T., Ostermeier, A., Pang, Z.P., Kokubu, Y., Su¨dhof, T.C., and Wernig, M. (2010). Direct conversion of fibroblasts to functional neurons by defined factors. Nature 463, 1035–1041.
Wagers, A.J., and Weissman, I.L. (2004). Plasticity of adult stem cells. Cell 116, 639–648. Weintraub, H., Tapscott, S.J., Davis, R.L., Thayer, M.J., Adam, M.A., Lassar, A.B., and Miller, A.D. (1989). Activation of muscle-specific genes in pigment, nerve, fat, liver, and fibroblast cell lines by forced expression of MyoD. Proc. Natl. Acad. Sci. USA 86, 5434–5438.
Ying, Q.L., Wray, J., Nichols, J., Batlle-Morera, L., Doble, B., Woodgett, J., Cohen, P., and Smith, A. (2008). The ground state of embryonic stem cell self-renewal. Nature 453, 519–523. Yu, J., Vodyanik, M.A., Smuga-Otto, K., Antosiewicz-Bourget, J., Frane, J.L., Tian, S., Nie, J., Jonsdottir, G.A., Ruotti, V., Stewart, R., et al. (2007). Induced pluripotent stem cell lines derived from human somatic cells. Science 318, 1917–1920.
Wernig, M., Meissner, A., Foreman, R., Brambrink, T., Ku, M., Hochedlinger, K., Bernstein, B.E., and Jaenisch, R. (2007). In vitro reprogramming of fibroblasts into a pluripotent ES-cell-like state. Nature 448, 318–324.
Zhao, X.Y., Li, W., Lv, Z., Liu, L., Tong, M., Hai, T., Hao, J., Guo, C.L., Ma, Q.W., Wang, L., et al. (2009). iPS cells produce viable mice through tetraploid complementation. Nature 461, 86–90.
Woltjen, K., Michael, I.P., Mohseni, P., Desai, R., Mileikovsky, M., Hamalainen, R., Cowling, R., Wang, W., Liu, P., Gertsenstein, M., et al. (2009). piggyBac transposition reprograms fibroblasts to induced pluripotent stem cells. Nature 458, 766–770.
Zhou, Q., Brown, J., Kanarek, A., Rajagopal, J., and Melton, D.A. (2008). In vivo reprogramming of adult pancreatic exocrine cells to beta-cells. Nature 455, 627–632.
Xie, H., Ye, M., Feng, R., and Graf, T. (2004). Stepwise reprogramming of B cells into macrophages. Cell 117, 663–676. Yamanaka, S., and Blau, H.M. (2010). Nuclear reprogramming to a pluripotent state by three approaches. Nature 465, 704–712.
Zou, J., Maeder, M.L., Mali, P., Pruett-Miller, S.M., Thibodeau-Beganny, S., Chou, B.K., Chen, G., Ye, Z., Park, I.H., Daley, G.Q., et al. (2009). Gene targeting of a disease-related gene in human induced pluripotent stem and embryonic stem cells. Cell Stem Cell 5, 97–110. Zwaka, T.P., and Thomson, J.A. (2003). Homologous recombination in human embryonic stem cells. Nat. Biotechnol. 21, 319–321.
Cell 143, November 12, 2010 ª2010 Elsevier Inc. 525
A Model for Neural Development and Treatment of Rett Syndrome Using Human Induced Pluripotent Stem Cells Maria C.N. Marchetto,1,5 Cassiano Carromeu,2,5 Allan Acab,2 Diana Yu,1 Gene W. Yeo,3 Yangling Mu,1 Gong Chen,4 Fred H. Gage,1 and Alysson R. Muotri2,* 1The
Salk Institute for Biological Studies, 10010 North Torrey Pines Road, La Jolla, CA 92037, USA
2University of California San Diego, School of Medicine, Department of Pediatrics, Rady Children’s Hospital San Diego, Department of Cellular
and Molecular Medicine, Stem Cell Program, 9500 Gilman Drive, La Jolla, CA 92093, USA 3University of California San Diego, School of Medicine, Department of Cellular & Molecular Medicine, Stem Cell Program, 9500 Gilman Drive, La Jolla, CA 92093, USA 4Pennylvania State University, Department of Biology. 201 Life Science Building, University Park, PA 6802, USA 5These authors contributed equally to the work *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.016
SUMMARY
Autism spectrum disorders (ASD) are complex neurodevelopmental diseases in which different combinations of genetic mutations may contribute to the phenotype. Using Rett syndrome (RTT) as an ASD genetic model, we developed a culture system using induced pluripotent stem cells (iPSCs) from RTT patients’ fibroblasts. RTT patients’ iPSCs are able to undergo X-inactivation and generate functional neurons. Neurons derived from RTT-iPSCs had fewer synapses, reduced spine density, smaller soma size, altered calcium signaling and electrophysiological defects when compared to controls. Our data uncovered early alterations in developing human RTT neurons. Finally, we used RTT neurons to test the effects of drugs in rescuing synaptic defects. Our data provide evidence of an unexplored developmental window, before disease onset, in RTT syndrome where potential therapies could be successfully employed. Our model recapitulates early stages of a human neurodevelopmental disease and represents a promising cellular tool for drug screening, diagnosis and personalized treatment. INTRODUCTION Autism spectrum disorders (ASD) are complex neurodevelopmental diseases affecting 1 in 150 children in the United States (Autism and Developmental Disabilities Monitoring Network Surveillance Year 2000 Prinicipal Investigators; Centers for Disease Control and Prevention, 2007). Such diseases are mainly characterized by impaired social interaction and repetitive behavior. Family history and twin studies suggest that, in
some cases, these disorders share genetic roots, but the degree to which environmental and genetic patterns account for individual differences within ASD is currently unknown (Piven et al., 1997; Ronald et al., 2006). A different combination of genetic mutations is likely to play a role in each individual. Nevertheless, the study of mutations in specific genes can help to identify molecular mechanisms responsible for subtle alterations in the nervous system, perhaps pointing to common mechanisms for ASD. Rett syndrome (RTT) is a progressive neurological disorder caused by mutations in the X-linked gene encoding MeCP2 protein (Amir et al., 1999). RTT patients have a large spectrum of autistic characteristics and are considered part of the ASD population (Hammer et al., 2002; Samaco et al., 2005, 2004; Zappella et al., 2003). These individuals undergo apparently normal development until 6–18 months of age, followed by impaired motor function, stagnation and then regression of developmental skills, hypotonia, seizures and autistic behavior (Amir et al., 1999). MeCP2 may be involved in the epigenetic regulation of target genes, by binding to methylated CpG dinucleotides within promoters, and may function as a transcriptional repressor, although this view has been challenged recently (Chahrour et al., 2008; Yasui et al., 2007). Pluripotent human embryonic stem cells (hESCs) have been successfully generated from early stage human embryos and can differentiate into various cell types (Thomson et al., 1998). However, to develop cellular models of human disease, it is necessary to generate cell lines with genomes predisposed to diseases. Recently, reprogramming of somatic cells to a pluripotent state by overexpression of specific genes (induced pluripotent stem cells, iPSCs) has been accomplished (Takahashi and Yamanaka, 2006; Yu et al., 2007). Isogenic pluripotent cells are attractive not only for their potential therapeutic use with lower risk of immune rejection but also for understanding complex diseases (Marchetto et al., 2010; Muotri, 2008). Although iPSCs have been generated for several neurological diseases (Dimos et al., 2008; Ebert et al., 2009; Hotta et al., 2009; Lee et al., 2009; Park et al., 2008; Soldner et al., 2009), the demonstration Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc. 527
of disease-specific pathogenesis and phenotypic rescue in relevant cell types is a current challenge in the field (Marchetto et al., 2010). We have developed a human model of RTT by generating iPSCs from fibroblasts of RTT patients carrying different MeCP2 mutations and unaffected individuals. We show that RTT-iPSCs retained the capacity to generate proliferating neural progenitor cells (NPCs) and functional neurons that underwent X-inactivation. We observed a reduced number of dendritic spines and synapses in iPSC-derived neurons. Moreover, we detected an altered frequency of intracellular calcium spikes and electrophysiological defects in RTT-derived neuronal networks, revealing potential new biomarkers for RTT pathology. Gain and loss of function experiments in iPSC-derived neurons confirmed that some of the alterations observed were related to MeCP2 expression levels. Finally, we used the iPSC system to test candidate drugs to rescue synaptic deficiency in RTT neurons. Together, our results suggest that RTT and other complex CNS diseases can be modeled using the iPSC technology to investigate the cellular and molecular mechanisms underlying their abnormalities.
RESULTS Generation of iPSCs from RTT Patients and Normal Individuals Nonaffected control fibroblasts and cells carrying four distinct MeCP2 mutations (Figure 1A and Table S1 available online) isolated from clinically affected female patients with RTT symptoms were infected with retroviral reprogramming vectors (Sox2, Oct4, c-Myc and Klf4), as described elsewhere (Takahashi et al., 2007). After 2 to 3 weeks, compact iPSC colonies emerged from a background of fibroblasts (Figures 1B and 1C). Colonies were manually picked and transferred to matrigel (Figures 1D and 1E). We obtained at least 10 clones from each control (WT)-iPSC and RTT-iPSC that continuously expressed pluripotent markers such as Nanog, Lin28, Tra-1-81 and Sox2 (Figures 1F and 1G and Figures S1A–S1C). All iPSC clones used in this study maintained a normal karyotype (Figure 1H). Teratomas containing derivatives from all 3 embryonic germ layers confirmed that the iPSCs were able to differentiate in vivo (Figure 1I). PCR fingerprinting confirmed their derivation from respective fibroblasts (data not shown). Next, we asked if the global molecular signatures of RTT-iPSC clones carrying the two distinct MeCP2 mutations (1155del32 and Q244X) and WT-iPSC clones (from AG09319) resembled those of available hESC lines (HUES6). Gene expression profiles measured using human genome Affymetrix Gene Chip arrays were grouped by hierarchical clustering, and correlation coefficients were computed for all pair-wise comparisons (GEO accession number GSE21037). We observed that the WT-iPSC and RTT-iPSC clones were almost indistinguishable. The results clearly revealed that the iPSC and hESC lines were more similar to each other than to the respective original fibroblasts (Figure S1D). These findings, combined with manual inspection of the gene expression of known pluripotent- and fibroblast-related genes (Figures S1E and S1F), indicated that the reprogramming was 528 Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc.
successful. In Table S2 we present a summary of all iPSC subjects and clones utilized for each experiment. Neural Induction of iPSCs Our protocol for neuronal differentiation is outlined in Figure 2A. We initiated neural differentiation by plating embryoid bodies (EBs). After a week, EB-derived rosettes became apparent (Figure 2B). Rosettes were then manually collected, dissociated and re-plated. The NPCs derived from rosettes formed a homogeneous population after a couple of passages. NPCs were positive for early neural precursor markers, such as Nestin, Sox1-2 and Musashi1 (Figure 2C). To obtain mature neurons, EBs in the presence of retinoic acid (RA) were dissociated and re-plated (Figure 2B). At this stage, cells were positive for Tuj1 (b-III-Tubulin) and Map2 (Microtubule-associated protein 2) (Figure 2D). Moreover, we detected expression of GABA (g-amino butyric acid) and VGLUT1 (vesicular glutamate transpoter-1). We also observed synapsin puncta outlining Map2-positive neurites (Figure 2D). We did not detect a significant alteration in RTT neuronal survival when compared to controls, as measured by Map2 staining (Figure 2E and Figure S2A). In addition, infection with a lentivirus expressing the DsRed gene under the control of Synapsin promoter (Syn::DsRed) did not reveal any difference in neuronal survival between RTT and controls (Figure 2E and Figure S2B). Interestingly, the number of GABA-positive neurons was also not affected between RTT and controls (Figure 2F and Figure S2C). X-Inactivation during Neuronal Differentiation of RTT-iPSCs In female hESCs, both chromosomes should be active, but one X chromosome becomes silenced upon differentiation (Dhara and Benvenisty, 2004). Similar to ESCs, female mouse iPSCs have shown reactivation of a somatically silenced X chromosome and have undergone random X-inactivation upon differentiation (Maherali et al., 2007). Because MeCP2 is an X-linked gene, we examined the ability of our RTT-iPSCs clones to reset the X chromosome (i.e., to erase X-inactivation) and whether X-inactivation would take place again after neuronal differentiation (Figure 3A). We stained RTT-iPSCs clones and their respective fibroblasts with an antibody against trimethylated histone 3 Lysine 27 (me3H3K27), an epigenetic silencing marker present on the inactive X chromosome in interphase nuclei (Silva et al., 2003). Some, but not all, undifferentiated RTT-iPSCs clones displayed diffuse immunoreactivity throughout the nucleus, similar to some hESCs, showing that the memory of the previous inactivation state had been erased (Figure 3B). For further analysis, we only selected clones that displayed a diffuse me3H3K27 pattern to differentiate into neurons. Upon neuronal differentiation, intense nuclear foci staining, a prominent diagnostic of the inactive X, was found in 80% of neurons labeled by the infection of a lentivirus carrying the neuron-specific Synapsin promoter driving the EGFP reporter (Syn::EGFP). Nuclear foci were also present in RTT fibroblasts before reprogramming (Figure 3B). We quantified the percentage of cells displaying either a diffuse or intense X-inactivation (nuclear foci) (Figure 3C). Our data suggest that the majority of cells in selected clones from both hESCs (99%) and iPSCs (95%) have a diffuse pattern. In contrast,
Figure 1. Generation of iPSCs (A) Schematic representation of the MeCP2 gene structure and mutations used in this study. UTR, untranslated region; MBD, methyl-CpG binding domain; NLS, nuclear localization signal; Poly-A, polyadenylation signal; TRD, transcriptional repression domain; WW, domain-containing WW; X, stop codon. Respective cell-line codes are shown close to their mutations. (B) Morphology of human fibroblasts before retroviral infection. (C) Aspect of iPSCs colonies 14 days after infection. (D and E) Representative images of established iPSC colonies. (F and G) Representative images of RTT-iPSCs showing expression of pluripotent markers. (H) No karyotypic abnormalities were observed. (I) Representative images of teratoma sections. The scale bar represents 100 mm. See also Figure S1.
differentiated populations of fibroblasts and iPSC-derived neurons have me3H3K27 nuclear foci staining, indicating X-inactivation. We also used fluorescent in situ hybridization (FISH) to visualize Xist RNA, a noncoding transcript involved in X chromosome silencing that physically wraps the inactive X (Lucchesi et al., 2005). Before reprogramming, the majority of fibroblasts exhibit a clear Xist cloud. The signal is lost after reprogramming, indicating that selected iPSC clones have two active X chromosomes in our culture conditions. A Xist cloud is also observed in iPSC-derived neurons (Figure 3D). Fluorescent in situ hybridization (FISH) analysis using a centromeric X chromosome probe in iPSC-derived NPCs and neurons showed the presence of two X
chromosomes (Figure 3E). As a consequence of both X-chromosomes’ activation after reprogramming, the MeCP2 protein can be detected in undifferentiated iPSCs from RTT patients (Figure 3F). However, after differentiation, RTT-iPSC-derived neurons recapitulated X-inactivation and the population became mosaic regarding MeCP2 expression. Immunostaining was performed on several RTT-iPSC clones, and a representative example of MeCP2 expression after differentiation is shown in Figure 3F. Clones obtained from RTT fibroblasts carrying the 1155del32 MeCP2 mutation do not produce a fulllength MeCP2 protein (Traynor et al., 2002). Next, we selected one WT-iPSC clone (WT-33 C1) and one RTT-iPSC clone (1155del32 C15) to determine whether the RTT-iPSC-derived neuronal population showed reduced MeCP2 protein levels. As expected, we observed a reduction in the full-length MeCP2 protein amounts in both fibroblasts and neurons derived from the RTT-iPSC clone (Figure 3G). We tested the original fibroblasts and iPSC-derived neurons from this patient for X-inactivation using standard methodology for the androgen receptor locus (Allen et al., 1992). RTT fibroblasts carrying the 1155del32 MeCP2 mutation had a 55:45 distribution, but RTT-derived neurons showed highly skewed X-inactivation, with a 96:4 distribution (Figure S3). The outcome of the X-inactivation process, measured by the androgen receptor locus, seems to be consistent within the same clone. An independent differentiation of the same clone (RTT-1155del32 C15) yielded a 98:2 distribution. Unfortunately, androgen receptor locus analysis was not conclusive for the MeCP2 mutation Q244X Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc. 529
Figure 2. Neural Differentiation of iPSCs (A) Schematic view of the neural differentiation protocol. (B) Representative images depicting morphological changes during neuronal differentiation. The scale bar represents 100 mm. (C) NPCs are positive for neural precursor markers: Sox1, Sox2, Musashi1, and Nestin. The scale bar represents 50 mm. (D–F) (D) Representative images of cells after neuronal differentiation. iPSC-derived neurons express mature neuronal markers: GABA, Map2 and Synapsin. The scale bar represents 20 mm. Similar numbers of Map2-positive and Syn::DsRed-positive (E) as well as GABA-positive (F) neurons from WT and RTT cultures. Data shown as mean ± SEM. See also Figure S2.
Our data show that X-inactivation was erased in selected reprogrammed RTT-iPSCs clones and subsequently restored during neuronal differentiation. Importantly, the recapitulation of X-inactivation produces mosaic neuronal cultures with different ratios of cells expressing normal MeCP2 levels, mimicking what is observed in RTT patients’ brains. Our data do not preclude that partial reprogramming from a single fibroblast or retention of the X-inactivation would lead to clones with highly skewed X-inactivation, where neurons would express only the normal or mutant form of MePC2. In fact, we do observe WT and RTT-iPSC clones retaining X-inactivation after reprogramming. The RTT-T158M C3-derived neurons showed 100:0 distribution. The expression of the mutant MeCP2 allele was confirmed by sequencing.
cells. However, a reduction of 50% in the amount of MeCP2 protein level (Figure S4E) is consistent with a random X-inactivation. We have not analyzed the distribution for RTT-R306C clones. 530 Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc.
Normal Cellular Proliferation from RTT-iPSC-Derived NPCs An increased incidence of large head size has been reported in autism (Piven et al., 1995). Other studies have suggested that the autistic brain is smaller at birth, followed by rapid head growth during early development and then a period of reduced brain growth (Courchesne et al., 2003). Head growth deceleration has also been reported for RTT patients (Hagberg et al., 2001). Since the cellular mechanism behind this phenomenon is unknown, we investigated whether a perturbed NPC replication cycle was affected in RTT. NPCs derived from RTT-iPSCs, WT-iPSCs and hESCs (Cyth25 and HUES6) were generated and kept under proliferating conditions in the presence of FGF2. NPCs were derived using the same protocol described above, had identical passage numbers and were analyzed for cell cycle by flow cytometry. Our results showed no significant differences in any cycle phase between HESC-, WT-iPSC- and RTT-iPSC-derived NPCs (Figure 4A), though we cannot exclude the possibility that
altered head growth in RTT patients is caused by eventual abnormal NPC proliferation in another developmental stage. We then investigated potential phenotypic changes in RTT neurons compared to controls. Reduced Glutamatergic Synapse Number and Morphological Alterations in RTT Neurons Strong evidence implicates synapse alteration in ASD, including RTT (Zoghbi, 2003). Loss of MeCP2 and doubling of MeCP2 dosage in mice have opposite effects on excitatory synapse numbers in individual neurons (Chao et al., 2007). These results suggest that MeCP2 may be a rate-limiting factor in regulating glutamatergic synapse formation and indicate that changes in excitatory synaptic strength may underlie global network alterations in RTT. Therefore, we determined whether excitatory synapse numbers were reduced in human RTT neurons. After 8 weeks of differentiation, glutamatergic neurons were identified using antibodies against VGLUT1 (Takamori et al., 2000), and dendrites were labeled with Map2 (Figure 4B). To confirm the specificity of glutamatergic neurons in our cultures, we showed that VGLUT1 puncta were mostly adjacent to the postsynaptic density-95 (Psd95) protein (Niethammer et al., 1996) (Figure S4A). We found a reduction in the density of VGLUT1 puncta from RTT-iPSCs clones carrying 3 different MeCP2 mutations compared to HUES6 and distinct WT-iPSCs-derived Map2positive neurons, suggesting a specific defect in glutamate transport in RTT cultures (Figure 4B and Figure S4B). Since neurons carrying different MeCP2 mutations showed reduced VGLUT1 puncta in our cultures, we tested whether loss of function of MeCP2 was directly related to the number of glutamatergic synapses in our neuronal cultures. We cloned an shRNA against MeCP2 in a lentiviral vector that is able to knockdown both isoforms of MeCP2 (Figure S4C). Neurons derived from WT-iPSCs expressing the shMeCP2 showed a similar reduction in VGLUT1 puncta when compared to control neurons expressing a scramble shRNA (shControl) (Figure 4C and Figure S4B). Overexpression of MeCP2 using a lentiviral vector (Figure S4C) increased the number of VGLUT1 puncta in WT and RTT neurons (Figure 4D and Figure S4B). Our data strongly suggest that MeCP2 is a rate-limiting factor in regulating glutamatergic synapse number in human neurons. We also investigated whether RTT neurons displayed any morphological alteration when compared to controls. To visualize neuronal anatomy, we infected the cultures with the Syn::EGFP lentivirus. Morphological analysis of RTT neurons revealed that the number of spines in RTT neurites was reduced when compared to WT neurons and after ectopic expression of shMeCP2 (Figure 4E). Consistent with this observation, the number of spines in dendrites of neurons from postmortem RTT patients’ brains was previously reported to be lower than that in normal individuals (Chapleau et al., 2009). Finally, we documented that the cell soma sizes from neurons derived from the RTT-iPSCs carrying different MeCP2 mutations were smaller when compared to controls (reduction of 14.31 ± 4.83%). Similarly, loss of function using the shMeCP2 knockdown strategy in WT neurons reduced soma size at levels comparable to RTT levels (reduction of 14.52 ± 4.31%) (Figure 4F and Figure S4D).
Rescuing a RTT Neuronal Phenotype Recent studies have shown that re-activation of MeCP2 expression knockout mice led to a prolonged life span and delayed onset or reversal of certain neurological symptoms (Giacometti et al., 2007; Guy et al., 2007). These reports suggest that some RTT phenotypes can be rescued in vivo. We used our model to analyze the effect of selected compounds that may revert the neuronal phenotype in culture as a validation for future highthroughput drug screening platforms. Administration of IGF1 was recently described to promote a partial reversal of the RTT-like symptoms in a mouse model (Tropea et al., 2009). We treated RTT-derived neurons carrying different MeCP2 mutations in culture with IGF1 and observed an increase in glutamatergic synapse number, suggesting that the drug treatment could correct the RTT neuronal phenotype (Figure 4B and Figure S4B). Around 60% of MeCP2 mutations in RTT are nonsense mutations (Laccone et al., 2001). Thus, we tested whether we could increase MeCP2 expression levels in affected neurons by suppressing the nonsense mutation (Q244X) with read-through of the premature stop codon using pharmacological treatments. High concentrations of aminoglycosides antibiotics, such as gentamicin, can bind to the 16S rRNA, impairing ribosomal proofreading (Kellermayer, 2006). As a consequence, a fulllength protein is produced by incorporating a random amino acid at the stop codon position. We treated RTT-Q244X clones 3- and 4-derived neurons with two different doses of gentamicin and found that MeCP2 protein levels and glutamatergic synapse numbers were increased after 1 week (Figure 4G and Figure S4E). Treatment with a higher gentamicin dose (400ug/ml) for the same period did not rescue RTT neurons and lowered the number of VGLUT1 puncta in control neurons (Figure 4G). The finding that RTT patient-derived neurons displayed changes in neuronal morphology and in number of synapses prompted us to explore putative circuit alterations in vitro. Altered Activity-Dependent Calcium Transients in RTT-iPSC-Derived Cells Early in neural development, spontaneous electrical activity leads to increases in intracellular calcium levels and activation of signaling pathways that are important in regulating several neuronal processes (Spitzer et al., 2004). Recently, a disturbance in calcium homeostasis during early postnatal development was reported in a MeCP2 knockout model (Mironov et al., 2009). Moreover, several studies showed that functional mutations in genes encoding voltage-gated calcium channels and in genes whose activity is modulated by calcium, such as MeCP2, could lead to ASD (Splawski et al., 2006; Zhou et al., 2006). Neuronal activity-induced calcium influx can trigger the calcium/calmodulin-dependent protein kinase (CamK). CamK activation has been reported to induce phosphorylation of MeCP2, which was further postulated to regulate neuronal spine maturation (Tao et al., 2009; Zhou et al., 2006). Although these studies raised an interesting link between neuronal activity and spine maturation, the extent of cellular alteration in human ASD neurons was never characterized. To test if RTT-iPSCs-derived neuronal networks are affected in our system, we preloaded the cells with the calcium indicator fluo-4AM and highlighted neurons using Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc. 531
Figure 3. RTT-iPSC Clones Undergo X-Inactivation during Differentiation (A) Schematic representation of X-inactivation dynamics during reprogramming and further neural differentiation. RTT fibroblasts are mosaic for the MeCP2 WT gene expression. During reprogramming, X-inactivation is erased and iPSCs express both MeCP2 alleles. Upon neuronal differentiation, X-inactivation is re-established and the resultant cells are mosaic for MeCP2 WT gene expression. (B) Immunofluorescence for me3H3K27 in fibroblasts, pluripotent cells (Nanog-positive) and after neuronal differentiation (Syn::EGFP-positive). Pluripotent cells (hESCs and iPSCs) show diffuse staining whereas differentiated cells (fibroblasts and neurons) exhibit prominent me3H3K27 nuclear foci (arrowheads). Cells were counterstained with Dapi. The scale bar represents 15 mm.
532 Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc.
the Syn::DsRed vector. Cultures with similar cell density and numbers of DsRed-positive neurons were used (Figure S2B). Spontaneous calcium transients were analyzed from WT and RTT neuronal networks in several independent experiments over time (Figure 5). In our analyses, we only considered calcium transients generated by synaptic activity. Neurons were selected after confirmation that calcium transients were blocked with TTX or with the glutamate receptor antagonists CNQX (AMPA) and APV (NMDA) treatments, indicating neuronal signaling dependence on local synaptic connections (Figures S5A, S5B, and S5D). Gabazine, an antagonist of GABAa receptors, increased the number of calcium transients in the networks, indicating the presence of glutamatergic and gabaergic synapses in our system (Figure S5C, D). A representative example of calcium tracing in control and RTT neurons is depicted in Figure 5A and shows a sharp increase in amplitude followed by a decrease over time. The frequency of calcium oscillations in RTT neurons and in WT neurons expressing shMeCP2 was abnormally decreased when compared to controls, suggesting a deficiency in the neuronal network connectivity and activity dynamics (Figures 5B and 5C and Figures S5E and S5F). The deficiency in connectivity was further corroborated by a decrease in the percentage of Syn::DsRed-positive neurons exhibiting calcium transients in the RTT cultures when compared to controls (Figure 5D and Figure S5F). Decreased Frequency of Spontaneous Postsynaptic Currents in RTT Neurons Next we determined the functional maturation of the iPSC-derived neurons using electrophysiological methods. Whole-cell recordings were performed from cells that had differentiated for at least 6 weeks in culture. Neurons were visualized by infection with the Syn::EGFP viral vector (Figure 6A). Both WT and RTT neurons showed similar transient sodium inward currents, sustained potassium outward currents in response to voltage step depolarizations, and action potentials evoked by somatic current injections (Figure 6B). Therefore, our data indicated that WT and RTT reprogramming did not affect the ability of WT-iPSC- and RTT-iPSC-derived neurons to mature and become electrophysiologically active. We also recorded spontaneous excitatory and inhibitory postsynaptic currents (sEPSCs and sIPSCs) as a way of measuring intercellular connectivity and network formation (Figures 6B and 6C). Cumulative probability plots of amplitudes and inter-event intervals of spontaneous postsynaptic currents revealed that RTT neurons have a significant decrease in frequency and amplitude when compared to WT neurons (Figures 6D and 6E). Together, our data suggest that the neuronal network is altered in RTT iPSCderived cultures.
DISCUSSION The lack of detectable symptoms in female RTT patients until 6–18 months of age and the apparent phenotypic reversibility of some RTT phenotypes in MeCP2 knockout animals indicate that MeCP2 is not essential for early wiring of the nervous system but instead may only be required at late stages. It is possible that RTT patients have aberrant excitatory synaptic strength at very early stages, when the disease phenotype is not yet clearly observed. In fact, increasing evidence from clinical studies and mouse models indicates the presence of alterations during the so-called presymptomatic developmental phase (Charman et al., 2002; De Filippis et al., 2009; Kerr et al., 1987; Picker et al., 2006; Santos et al., 2007). To study human RTT neurons in culture, we derived iPSCs from RTT fibroblasts. RTT iPSCs are pluripotent and able to recapitulate X-inactivation upon neuronal differentiation. Even though the ratio of neurons expressing mutant MeCP2 due to X-inactivation was variable, the phenotypes described here for all RTT-derived neurons are similar. One interpretation could be that astrocytes, or other nonneuronal cells, carrying MeCP2 mutations present in our cultures could also affect neurons expressing the normal MeCP2 protein. In fact, the non-cell-autonomous influence was recently described for RTT, indicating that glial cells carrying MeCP2 mutations can distress healthy neurons (Ballas et al., 2009; Kishi and Macklis, 2010; Maezawa et al., 2009). Using human neurons carrying MeCP2 mutations, we showed that RTT glutamatergic neurons have a reduced number of synapses and dendritic spines when compared to nonaffected controls. Moreover, electrophysiological recordings from RTT neurons showed a significant decrease in the frequency and amplitude of spontaneous synaptic currents compared to WT neurons. The reduced frequency in RTT neurons could reflect the presence of fewer release sites or a decreased release probability. The results of electrophysiology recordings are consistent with the decreased VGLUT1 puncta observed in Map2-positive dendrites from RTT neurons. Also consistent with these findings, the frequency of intracellular calcium transients was decreased in RTT neurons when compared to controls. Our data indicate a potential imbalance in the neuronal networks associated with RTT pathology. The observations described here provide valuable information for RTT and, potentially, ASD patients, since they suggest that presymptomatic defects may represent novel biomarkers to be exploited as diagnostic tools and that early intervention may be beneficial. Therapies aiming at earlier stages of development may attenuate the downstream consequences of MeCP2 mutations. Restoring protein levels may be challenging, since MeCP2 levels are tightly regulated and chronically overdosing neurons with the
(C) Quantification of cells with diffused or foci me3H3K27 nuclear staining. Data shown as mean ± SEM. (D) RNA FISH shows that Xist RNA domains are present in the original fibroblasts before reprogramming. iPSCs show no Xist expression. Neurons derived from normal and RTT iPSCs show clear Xist clouds, indicating transcriptional silencing of the X chromosome (arrows). The scale bar represents 5 mm. (E) Two DNA FISH signals are evident in the nuclei of iPSC-derived NPCs and neurons, revealing the presence of two X chromosomes. The scale bar represents 10 mm. (F) RTT-iPSCs (1155del32) expressed WT MeCP2 but derived neurons displayed mosaicism regarding WT (arrowhead) and mutant (arrow) MeCP2 forms. The scale bar represents 50 mm. (G) RTT-derived fibroblasts and neurons have reduced levels of WT MeCP2 protein by Western blot. See also Figure S3.
Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc. 533
Figure 4. Alterations in RTT Neurons (A) Proliferating RTT NPCs displayed no signal of aberrant cell cycle when compared to controls. (B) Representative images of neurons showing VGLUT1 puncta on Map2 neurites. Bar graphs show synaptic density in RTT and WT neurons. IGF1 treatment increased VGLUT1 puncta number in RTT-derived neurons. The scale bar represents 5 mm. (C) Reduction of MeCP2 expression decreased the number of glutamatergic synapses in WT neurons. (D) Overexpression of MeCP2 increased the number of glutamatergic synapses. (E) Representative images of neurites of different genetic backgrounds. Bar graph shows the spine density from independent experiments using different RTT backgrounds and controls and after expression of shMeCP2. The scale bar represents 5 mm. (F) Representative images of neuronal cell body size. Bar graph shows the percentage of soma size decrease in RTT compared to WT neurons. Neuronal morphology was visualized using the Syn::EGFP lentiviral vector. The scale bar represents 50 mm. (G) A lower dose of gentamicin was able to rescue glutamatergic synapses in RTT neurons. Numbers of neurons analyzed (n) are shown within the bars in graphs (E) and (G). For all clones and mutations used refer to Figure S4 and Table S2. Data shown as mean ± SEM.
534 Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc.
WT allele can be as harmful as a loss of expression (Collins et al., 2004; Ramocki et al., 2009; Van Esch et al., 2005). Thus, we tested pharmacological treatment as a way to recover the RTT neuronal phenotype. We investigated the use of IGF1 in human neuronal cultures. Although it likely acts in a nonspecific manner, IGF1 is considered to be a candidate for pharmacological treatment of RTT and potentially other CNS disorders in a future clinical trial (Tropea et al., 2009). While IGF1 treatment increased synapse number in some clones, it stimulated glutamatergic RTT neurons above normal levels. Our data indicate that the IGF1 dose and timing parameters need to be precisely tuned in future clinical trials to avoid side effects. In a different approach, we tested a read-through drug (gentamicin) to rescue neurons derived from iPSCs carrying a nonsense MeCP2 mutation. A lower dosage of gentamicin was enough to increase full-length MeCP2 levels in RTT neurons, rescuing glutamatergic synapses. New drugs with reduced toxicity and enhanced suppression of premature stop codon mutations might be good therapeutic candidates (Nudelman et al., 2009; Welch et al., 2007). Control of glutamatergic synapse number and the other neuronal phenotypes analyzed here may be caused by loss of MeCP2 function in the cell. Alternatively, significant experimental and genomic variability in our system could be directly responsible for the RTT differences displayed in our data. Our gain and loss of function data strongly suggest that MeCP2 is indeed the causative agent of the cellular phenotypes reported here that might be relevant to the clinical features of RTT. Our data indicate that iPSCs not only can recapitulate some aspects of a genetic disease but also can be used to better design and anticipate results from translational medicine. This cellular model has the potential to lead to the discovery of new compounds to treat RTT and other forms of ASD. Finally, other CNS diseases may be modeled in vitro using a similar approach. EXPERIMENTAL PROCEDURES
Figure 5. Altered Activity-Dependent Calcium Transients in RTTDerived Neurons (A) Representative examples of WT and RTT calcium signal traces. Red traces correspond to the calcium rise phase detected by the algorithm used (see Extended Experimental Procedures). (B) Fluorescence intensity changes reflecting intracellular calcium fluctuations in RTT and WT neurons in different Regions of Interest (ROI). (C) RTT neurons show a lower average of calcium spikes when compared to WT control neurons. (D) The percentage of Syn::DsRed-positive neurons signaling in the RTT neuronal network is significantly reduced when compared to controls. Data shown as mean ± SEM. See also Figure S5.
Cell Culture and Retrovirus Infection Female RTT and control fibroblasts were generated from explants of dermal biopsies following informed consent under protocols approved by the University of California San Diego. The Syn::EGFP or DsRed reporter vector was obtained by cloning the Synapsin-1 promoter (a gift from Dr. G. Thiel, Hamburg, Germany) in a lentivirus backbone. The shRNA against a target sequence on the human MeCP2 gene was cloned in the LentiLox3.7 lentivirus vector. Retrovirus vectors containing the Oct4, c-Myc, Klf4 and Sox2 human cDNAs from Yamanaka’s group (Takahashi et al., 2007) were obtained from Addgene. Two days after infection, fibroblasts were plated on mitotically inactivated mouse embryonic fibroblasts (Chemicon) with hESC medium. After 2 weeks, iPSC colonies were directly transferred to feeder-free conditions on matrigelcoated dishes (BD) using mTeSR1 (StemCell Technologies, and passed manually. The detailed protocols to obtain NPCs and mature neurons are described in the supplemental material. For the rescue experiments, 10 hg/mL of IGF1 (Peprotech) or Gentamicin (Invitrogen; at 100 or 400 mg/mL) was added to neuronal cultures for 1 week. Protocols were previously approved by the University of California San Diego and Salk Institute Institutional Review Board and the Embryonic Stem Cell Research Oversight Committee. Immunocytochemistry and Neuronal Morphology Quantification Cells were briefly fixed in 4% paraformaldehyde and then permeabilized with 0.5% Triton X-100 in PBS. Cells were then blocked in PBS containing 0.5% Triton X-100 and 5% donkey serum for 1 hr before incubation with primary antibody overnight at 4 C. After three washes with PBS, cells were incubated
Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc. 535
Figure 6. Decreased Frequency of Spontaneous Postsynaptic Currents in RTT Neurons (A) Fluorescence micrographs of representative WT and RTT neurons. The scale bar represents 10 mm. (B) Electrophysiological properties of WT and RTT neurons. From top to bottom: Transient Na+ currents and sustained K+ currents in response to voltage step depolarizations (command voltage varied from 20 to +30 mV in 5 mV increments when cells were voltage-clamped at 70 mV, Bars = 400 pA and 50 ms). Action potentials evoked by somatic current injections (cells current-clamped at around 60 mV, injected currents from 10 to 40 pA, Bars = 20 mV and 100 ms), sEPSCs (Bars = right, 20 pA, 100 ms; left: 10 pA, 500 ms), and sIPSCs (Bars = right, 20 pA, 500 ms; left: 20 pA, 400 ms). (C) Sample 4 min recordings of spontaneous currents when the cells were voltage-clamped at 70 mV (Bars = 20 pA and 25 s). (D) Cumulative probability plot of amplitudes (left panel, 1 pA bins; p < 0.001) and inter-event intervals (right panel, 20 ms bins; p < 0.05) of sEPSCs from groups of WT (black) and RTT (red) cells, respectively. (E) Cumulative probability plot of amplitudes (left panel, 1 pA bins; p < 0.05) and inter-event intervals (right panel, 20 ms bins; p < 0.05) of sIPSCs from each group (WT, black; RTT, green).
with secondary antibodies (Jackson ImmunoResearch) for 1 hr at room temperature. Fluorescent signals were detected using a Zeiss inverted microscope and images were processed with Photoshop CS3 (Adobe Systems). Primary antibodies used in this study are described in the supplemental information. Cell soma size was measure in bright field using ImageJ software after identification of neurons using the Syn::EGFP. The morphologies of neuronal dendrites and spines were studied from an individual projection of z-stacks optical sections and scanned at 0.5-mm increments that correlated with the resolution valued at z-plane. Each optical section was the result of 3 scans at 500 lps followed by Kalman filtering. For synapse quantification, images were taken by a z-step of 1 mm using Biorad radiance 2100 confocal microscope. Synapse quantification was done blinded to genotype. Only VGLUT1 puncta along Map2-positive processes were counted. Statistical significances were tested using Two-way ANOVA test and Bonferroni post-test.
Cell Cycle Analysis One million NPCs were fixed in 70% EtOH for at least 2 hr at 4 C. After PBS washing, cells were stained with 1 ml of propidium iodide (PI) solution (50 mg/mL PI in 3.8 Mm sodium citrate) and treated with 20 mL/mL of RNaseA. Cells were analyzed by fluorescence-activated cell sorting (FACS) on a Becton
536 Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc.
Dickinson LSRI and cell cycle gating was examined using FLOWJO - Flow Cytometry Analysis Software. RNA Extraction and RT-PCR Total cellular RNA was extracted from 5x106 cells using the RNeasy Protect Mini kit (QIAGEN, Valencia, CA), according to the manufacturer’s instructions, and reverse transcribed using the SuperScript III First-Strand Synthesis System RT-PCR from Invitrogen. The cDNA was amplified by PCR using Accuprime Taq DNA polymerase system (Invitrogen). Primer sequences used are described in Supplemental information. Teratoma Formation in Nude Mice Around 1-3 x106 fibroblasts or iPSCs were injected subcutaneously into the dorsal flanks of nude mice (CByJ.Cg-Foxn1nu/J) anesthetized with isoflurane. Five to six weeks after injection, teratomas were dissected, fixed overnight in 10% buffered formalin phosphate and embedded in paraffin. Sections were stained with hematoxylin and eosin for further analysis. Control mice injected with RTT fibroblasts failed to form teratomas. Protocols were previously approved by the University of California San Diego Institutional Animal Care and Use Committee.
Karyotyping and DNA Fingerprinting Standard G-banding chromosome and DNA fingerprinting analysis was performed by Cell Line Genetics (Madison, WI). DNA and RNA FISH Xist RNA exon 6 probes (GenBank U80460: 75081-78658 – a gift from Dr. Jeannie T. Lee, Massachusetts General Hospital, Harvard Medical School) were transcribed by using T7 RNA polymerase (Roche) with AlexaFluor 488-5UTP. X chromosome probe and Xist slide hybridization were performed by Molecular Diagnostic Services, Inc. (San Diego, CA). Protein Isolation and Western Blot Analysis Cells were isolated, suspended in 13 RIPA lyses buffer (Upstate) supplemented with 1% protease inhibitor cocktail (Sigma), triturated and centrifuged at 10,000 3 g for 10 min at 4 C. Twenty micrograms of total protein was separated on 12% SDS-polyacrylamide gel, transferred to a nitrocellulose membrane and probed with a primary antibody against MeCP2 (1:5,000; Diagenode), followed by horseradish-peroxidase-conjugated secondary antibody (1:5,000; Promega), and then visualized using ECL chemiluminescence (Amersham). As a control, membranes were stripped and re-probed for b-actin (1:10,000; Ambion) or a–tubulin (1:5,000, Ambion). For semiquantitative analysis, MeCP2 signal intensity was analyzed and corrected with respect to b-actin. Microarray Analysis The Affymetrix Power Tools (APT) suite of programs and Affymetrix Human Gene 1.0 ST Arrays library files and annotation were obtained from http:// www.affymetrix.com/support and details of the analysis are available in Supplemental information. Calcium Imaging Neuronal networks derived from human iPSCs were previously infected with the lentiviral vector carrying the Syn:DsRed reporter construct. Cell cultures were washed twice with sterile Krebs HEPES Buffer (KHB) and incubated with 2–5 mM Fluo-4AM (Molecular Probes/Invitrogen, Carlsbad, CA) in KHB for 40 min at room temperature. Excess dye was removed by washing twice with KHB and an additional 20 min incubation was done to equilibrate intracellular dye concentration and allow de-esterification. Time-lapse image sequences (1003 magnification) of 5000 frames were acquired at 28 Hz with a region of 336 3 256 pixels, using a Hamamatsu ORCA-ER digital camera (Hamamatsu Photonics K.K., Japan) with a 488 nm (FITC) filter on an Olympus IX81 inverted fluorescence confocal microscope (Olympus Optical, Japan). Images were acquired with MetaMorph 7.7 (MDS Analytical Technologies, Sunnyvale, CA). Images were subsequently processed using ImageJ (http:// rsbweb.nih.gov/ij/) and custom written routines in Matlab 7.2 (Mathworks, Natick, MA). Detailed quantitative analysis of calcium transients is available in the Supplemental material. Electrophysiology Whole-cell patch clamp recordings were performed from cells co-cultured with astrocytes after 6 weeks of differentiation. The bath was constantly perfused with fresh HEPES-buffered saline (see supplemental methods for recipe). The recording micropipettes (tip resistance 3–6 MU) were filled with internal solution described in the Supplemental materials. Recordings were made using Axopatch 200B amplifier (Axon Instruments). Signals were filtered at 2 kHz and sampled at 5 kHz. The whole-cell capacitance was fully compensated. The series resistance was uncompensated but monitored during the experiment by the amplitude of the capacitive current in response to a 10 mV pulse. All recordings were performed at room temperature and chemicals were purchased from Sigma. Frequency and amplitude of spontaneous postsynaptic currents were measured with the Mini Analysis Program software (Synaptosoft, Leonia, NJ). Statistical comparisons of WT and RTT groups were made using the nonparametric Kolmogorov-Smirnov two-tailed test, with a significance criterion of p = 0.05. EPSCs were blocked by CNQX or DNQX (10–20 mM) and IPSPs were inhibited by bicuculine (20 mM).
SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, five figures, and two tables and can be found with this article online at doi:10. 1016/j.cell.2010.10.016. ACKNOWLEDGMENTS The work was supported by the Emerald Foundation and by the National Institutes of Health through the NIH Director’s New Innovator Award Program, 1-DP2-OD006495-01. F.H.G. is supported by California Institute for Regenerative Medicine RL1-00649-1 and RC1-00115-1, The Lookout Fund and the Mathers Foundation. C.C. is a fellow from the International Rett Syndrome Foundation. M.C.N.M. is a Christopher and Danna Reeve Foundation fellow. G.C. was supported by the Glenn Foundation. We would like to thank Monica Coenraads for critical discussion; the Greenwood Genetic Center clinical diagnostic laboratory for X-inactivation analysis; Dr. Jeannie T. Lee for the Xist probe; and M.L. Gage for editorial comments. Received: February 9, 2010 Revised: August 4, 2010 Accepted: October 8, 2010 Published: November 11, 2010 REFERENCES Allen, R.C., Zoghbi, H.Y., Moseley, A.B., Rosenblatt, H.M., and Belmont, J.W. (1992). Methylation of HpaII and HhaI sites near the polymorphic CAG repeat in the human androgen-receptor gene correlates with X chromosome inactivation. Am. J. Hum. Genet. 51, 1229–1239. Amir, R.E., Van den Veyver, I.B., Wan, M., Tran, C.Q., Francke, U., and Zoghbi, H.Y. (1999). Rett syndrome is caused by mutations in X-linked MECP2, encoding methyl-CpG-binding protein 2. Nat. Genet. 23, 185–188. Autism and Developmental Disabilities Monitoring Network Surveillance Year 2000 Prinicipal Investigators; Centers for Disease Control and Prevention. (2007). Prevalence of autism spectrum disorders–autism and developmental disabilities monitoring network, six sites, United States, 2000. MMWR Surveill Summ 56, 1–11. Ballas, N., Lioy, D.T., Grunseich, C., and Mandel, G. (2009). Non-cell autonomous influence of MeCP2-deficient glia on neuronal dendritic morphology. Nat. Neurosci. 12, 311–317. Chahrour, M., Jung, S.Y., Shaw, C., Zhou, X., Wong, S.T., Qin, J., and Zoghbi, H.Y. (2008). MeCP2, a key contributor to neurological disease, activates and represses transcription. Science 320, 1224–1229. Chao, H.T., Zoghbi, H.Y., and Rosenmund, C. (2007). MeCP2 controls excitatory synaptic strength by regulating glutamatergic synapse number. Neuron 56, 58–65. Chapleau, C.A., Calfa, G.D., Lane, M.C., Albertson, A.J., Larimore, J.L., Kudo, S., Armstrong, D.L., Percy, A.K., and Pozzo-Miller, L. (2009). Dendritic spine pathologies in hippocampal pyramidal neurons from Rett syndrome brain and after expression of Rett-associated MECP2 mutations. Neurobiol. Dis. 35, 219–233. Charman, T., Cass, H., Owen, L., Wigram, T., Slonims, V., Weeks, L., Wisbeach, A., and Reilly, S. (2002). Regression in individuals with Rett syndrome. Brain Dev. 24, 281–283. Collins, A.L., Levenson, J.M., Vilaythong, A.P., Richman, R., Armstrong, D.L., Noebels, J.L., David Sweatt, J., and Zoghbi, H.Y. (2004). Mild overexpression of MeCP2 causes a progressive neurological disorder in mice. Hum. Mol. Genet. 13, 2679–2689. Courchesne, E., Carper, R., and Akshoomoff, N. (2003). Evidence of brain overgrowth in the first year of life in autism. JAMA 290, 337–344. De Filippis, B., Ricceri, L., and Laviola, G. (2009). Early postnatal behavioral changes in the Mecp2-308 truncation mouse model of Rett syndrome. Genes Brain Behav. 9, 213–223.
Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc. 537
Dhara, S.K., and Benvenisty, N. (2004). Gene trap as a tool for genome annotation and analysis of X chromosome inactivation in human embryonic stem cells. Nucleic Acids Res. 32, 3995–4002.
Development of novel aminoglycoside (NB54) with reduced toxicity and enhanced suppression of disease-causing premature stop mutations. J. Med. Chem. 52, 2836–2845.
Dimos, J.T., Rodolfa, K.T., Niakan, K.K., Weisenthal, L.M., Mitsumoto, H., Chung, W., Croft, G.F., Saphier, G., Leibel, R., Goland, R., et al. (2008). Induced pluripotent stem cells generated from patients with ALS can be differentiated into motor neurons. Science 321, 1218–1221.
Park, I.H., Arora, N., Huo, H., Maherali, N., Ahfeldt, T., Shimamura, A., Lensch, M.W., Cowan, C., Hochedlinger, K., and Daley, G.Q. (2008). Disease-specific induced pluripotent stem cells. Cell 134, 877–886.
Ebert, A.D., Yu, J., Rose, F.F., Jr., Mattis, V.B., Lorson, C.L., Thomson, J.A., and Svendsen, C.N. (2009). Induced pluripotent stem cells from a spinal muscular atrophy patient. Nature 457, 277–280. Giacometti, E., Luikenhuis, S., Beard, C., and Jaenisch, R. (2007). Partial rescue of MeCP2 deficiency by postnatal activation of MeCP2. Proc. Natl. Acad. Sci. USA 104, 1931–1936. Guy, J., Gan, J., Selfridge, J., Cobb, S., and Bird, A. (2007). Reversal of neurological defects in a mouse model of Rett syndrome. Science 315, 1143–1147. Hagberg, G., Stenbom, Y., and Engerstro¨m, I.W. (2001). Head growth in Rett syndrome. Brain Dev. 23 (Suppl 1), S227–S229. Hammer, S., Dorrani, N., Dragich, J., Kudo, S., and Schanen, C. (2002). The phenotypic consequences of MECP2 mutations extend beyond Rett syndrome. Ment. Retard. Dev. Disabil. Res. Rev. 8, 94–98. Hotta, A., Cheung, A.Y., Farra, N., Vijayaragavan, K., Se´guin, C.A., Draper, J.S., Pasceri, P., Maksakova, I.A., Mager, D.L., Rossant, J., et al. (2009). Isolation of human iPS cells using EOS lentiviral vectors to select for pluripotency. Nat. Methods 6, 370–376. Kellermayer, R. (2006). Translational readthrough induction of pathogenic nonsense mutations. Eur. J. Med. Genet. 49, 445–450. Kerr, A.M., Montague, J., and Stephenson, J.B. (1987). The hands, and the mind, pre- and post-regression, in Rett syndrome. Brain Dev. 9, 487–490. Kishi, N., and Macklis, J.D. (2010). MeCP2 functions largely cell-autonomously, but also non-cell-autonomously, in neuronal maturation and dendritic arborization of cortical pyramidal neurons. Exp. Neurol. 222, 51–58. Laccone, F., Huppke, P., Hanefeld, F., and Meins, M. (2001). Mutation spectrum in patients with Rett syndrome in the German population: Evidence of hot spot regions. Hum. Mutat. 17, 183–190. Lee, G., Papapetrou, E.P., Kim, H., Chambers, S.M., Tomishima, M.J., Fasano, C.A., Ganat, Y.M., Menon, J., Shimizu, F., Viale, A., et al. (2009). Modelling pathogenesis and treatment of familial dysautonomia using patient-specific iPSCs. Nature 461, 402–406.
Picker, J.D., Yang, R., Ricceri, L., and Berger-Sweeney, J. (2006). An altered neonatal behavioral phenotype in Mecp2 mutant mice. Neuroreport 17, 541–544. Piven, J., Arndt, S., Bailey, J., Havercamp, S., Andreasen, N.C., and Palmer, P. (1995). An MRI study of brain size in autism. Am. J. Psychiatry 152, 1145–1149. Piven, J., Palmer, P., Jacobi, D., Childress, D., and Arndt, S. (1997). Broader autism phenotype: evidence from a family history study of multiple-incidence autism families. Am. J. Psychiatry 154, 185–190. Ramocki, M.B., Peters, S.U., Tavyev, Y.J., Zhang, F., Carvalho, C.M., Schaaf, C.P., Richman, R., Fang, P., Glaze, D.G., Lupski, J.R., and Zoghbi, H.Y. (2009). Autism and other neuropsychiatric symptoms are prevalent in individuals with MeCP2 duplication syndrome. Ann. Neurol. 66, 771–782. Ronald, A., Happe´, F., Bolton, P., Butcher, L.M., Price, T.S., Wheelwright, S., Baron-Cohen, S., and Plomin, R. (2006). Genetic heterogeneity between the three components of the autism spectrum: a twin study. J. Am. Acad. Child Adolesc. Psychiatry 45, 691–699. Samaco, R.C., Hogart, A., and LaSalle, J.M. (2005). Epigenetic overlap in autism-spectrum neurodevelopmental disorders: MECP2 deficiency causes reduced expression of UBE3A and GABRB3. Hum. Mol. Genet. 14, 483–492. Samaco, R.C., Nagarajan, R.P., Braunschweig, D., and LaSalle, J.M. (2004). Multiple pathways regulate MeCP2 expression in normal brain development and exhibit defects in autism-spectrum disorders. Hum. Mol. Genet. 13, 629–639. Santos, M., Silva-Fernandes, A., Oliveira, P., Sousa, N., and Maciel, P. (2007). Evidence for abnormal early development in a mouse model of Rett syndrome. Genes Brain Behav. 6, 277–286. Silva, J., Mak, W., Zvetkova, I., Appanah, R., Nesterova, T.B., Webster, Z., Peters, A.H., Jenuwein, T., Otte, A.P., and Brockdorff, N. (2003). Establishment of histone h3 methylation on the inactive X chromosome requires transient recruitment of Eed-Enx1 polycomb group complexes. Dev. Cell 4, 481–495.
Lucchesi, J.C., Kelly, W.G., and Panning, B. (2005). Chromatin remodeling in dosage compensation. Annu. Rev. Genet. 39, 615–651.
Soldner, F., Hockemeyer, D., Beard, C., Gao, Q., Bell, G.W., Cook, E.G., Hargus, G., Blak, A., Cooper, O., Mitalipova, M., et al. (2009). Parkinson’s disease patient-derived induced pluripotent stem cells free of viral reprogramming factors. Cell 136, 964–977.
Maezawa, I., Swanberg, S., Harvey, D., LaSalle, J.M., and Jin, L.W. (2009). Rett syndrome astrocytes are abnormal and spread MeCP2 deficiency through gap junctions. J. Neurosci. 29, 5051–5061.
Spitzer, N.C., Root, C.M., and Borodinsky, L.N. (2004). Orchestrating neuronal differentiation: patterns of Ca2+ spikes specify transmitter choice. Trends Neurosci. 27, 415–421.
Maherali, N., Sridharan, R., Xie, W., Utikal, J., Eminli, S., Arnold, K., Stadtfeld, M., Yachechko, R., Tchieu, J., Jaenisch, R., et al. (2007). Directly reprogrammed fibroblasts show global epigenetic remodeling and widespread tissue contribution. Cell Stem Cell 1, 55–70.
Splawski, I., Yoo, D.S., Stotz, S.C., Cherry, A., Clapham, D.E., and Keating, M.T. (2006). CACNA1H mutations in autism spectrum disorders. J. Biol. Chem. 281, 22085–22091.
Marchetto, M.C., Winner, B., and Gage, F.H. (2010). Pluripotent stem cells in neurodegenerative and neurodevelopmental diseases. Hum. Mol. Genet. 19 (R1), R71–R76. Mironov, S.L., Skorova, E., Hartelt, N., Mironova, L.A., Hasan, M.T., and Ku¨gler, S. (2009). Remodelling of the respiratory network in a mouse model of Rett syndrome depends on brain-derived neurotrophic factor regulated slow calcium buffering. J. Physiol. 587, 2473–2485. Muotri, A.R. (2008). Modeling epilepsy with pluripotent human cells. Epilepsy Behav. 14 (Suppl.), 81–85. Niethammer, M., Kim, E., and Sheng, M. (1996). Interaction between the C terminus of NMDA receptor subunits and multiple members of the PSD-95 family of membrane-associated guanylate kinases. J. Neurosci. 16, 2157–2163. Nudelman, I., Rebibo-Sabbah, A., Cherniavsky, M., Belakhov, V., Hainrichson, M., Chen, F., Schacht, J., Pilch, D.S., Ben-Yosef, T., and Baasov, T. (2009).
538 Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc.
Takahashi, K., Tanabe, K., Ohnuki, M., Narita, M., Ichisaka, T., Tomoda, K., and Yamanaka, S. (2007). Induction of pluripotent stem cells from adult human fibroblasts by defined factors. Cell 131, 861–872. Takahashi, K., and Yamanaka, S. (2006). Induction of pluripotent stem cells from mouse embryonic and adult fibroblast cultures by defined factors. Cell 126, 663–676. Takamori, S., Rhee, J.S., Rosenmund, C., and Jahn, R. (2000). Identification of a vesicular glutamate transporter that defines a glutamatergic phenotype in neurons. Nature 407, 189–194. Tao, J., Hu, K., Chang, Q., Wu, H., Sherman, N.E., Martinowich, K., Klose, R.J., Schanen, C., Jaenisch, R., Wang, W., and Sun, Y.E. (2009). Phosphorylation of MeCP2 at Serine 80 regulates its chromatin association and neurological function. Proc. Natl. Acad. Sci. USA 106, 4882–4887. Thomson, J.A., Itskovitz-Eldor, J., Shapiro, S.S., Waknitz, M.A., Swiergiel, J.J., Marshall, V.S., and Jones, J.M. (1998). Embryonic stem cell lines derived from human blastocysts. Science 282, 1145–1147.
Traynor, J., Agarwal, P., Lazzeroni, L., and Francke, U. (2002). Gene expression patterns vary in clonal cell cultures from Rett syndrome females with eight different MECP2 mutations. BMC Med. Genet. 3, 12. Tropea, D., Giacometti, E., Wilson, N.R., Beard, C., McCurry, C., Fu, D.D., Flannery, R., Jaenisch, R., and Sur, M. (2009). Partial reversal of Rett Syndrome-like symptoms in MeCP2 mutant mice. Proc. Natl. Acad. Sci. USA 106, 2029–2034. Van Esch, H., Bauters, M., Ignatius, J., Jansen, M., Raynaud, M., Hollanders, K., Lugtenberg, D., Bienvenu, T., Jensen, L.R., Gecz, J., et al. (2005). Duplication of the MECP2 region is a frequent cause of severe mental retardation and progressive neurological symptoms in males. Am. J. Hum. Genet. 77, 442–453. Welch, E.M., Barton, E.R., Zhuo, J., Tomizawa, Y., Friesen, W.J., Trifillis, P., Paushkin, S., Patel, M., Trotta, C.R., Hwang, S., et al. (2007). PTC124 targets genetic disorders caused by nonsense mutations. Nature 447, 87–91. Yasui, D.H., Peddada, S., Bieda, M.C., Vallero, R.O., Hogart, A., Nagarajan, R.P., Thatcher, K.N., Farnham, P.J., and Lasalle, J.M. (2007). Integrated epige-
nomic analyses of neuronal MeCP2 reveal a role for long-range interaction with active genes. Proc. Natl. Acad. Sci. USA 104, 19416–19421. Yu, J., Vodyanik, M.A., Smuga-Otto, K., Antosiewicz-Bourget, J., Frane, J.L., Tian, S., Nie, J., Jonsdottir, G.A., Ruotti, V., Stewart, R., et al. (2007). Induced pluripotent stem cell lines derived from human somatic cells. Science 318, 1917–1920. Zappella, M., Meloni, I., Longo, I., Canitano, R., Hayek, G., Rosaia, L., Mari, F., and Renieri, A. (2003). Study of MECP2 gene in Rett syndrome variants and autistic girls. Am. J. Med. Genet. B. Neuropsychiatr. Genet. 119, 102–107. Zhou, Z., Hong, E.J., Cohen, S., Zhao, W.N., Ho, H.Y., Schmidt, L., Chen, W.G., Lin, Y., Savner, E., Griffith, E.C., et al. (2006). Brain-specific phosphorylation of MeCP2 regulates activity-dependent Bdnf transcription, dendritic growth, and spine maturation. Neuron 52, 255–269. Zoghbi, H.Y. (2003). Postnatal neurodevelopmental disorders: meeting at the synapse? Science 302, 826–830.
Cell 143, 527–539, November 12, 2010 ª2010 Elsevier Inc. 539
Pausing of RNA Polymerase II Disrupts DNA-Specified Nucleosome Organization to Enable Precise Gene Regulation Daniel A. Gilchrist,1 Gilberto Dos Santos,1 David C. Fargo,2 Bin Xie,4 Yuan Gao,4,5 Leping Li,3 and Karen Adelman1,* 1Laboratory
of Molecular Carcinogenesis and Information Services 3Biostatistics Branch National Institute of Environmental Health Sciences, National Institutes of Health, Research Triangle Park, NC 27709, USA 4Division of Genomics, Epigenomics and Bioinformatics, Lieber Institute for Brain Development 5Neurogeneration and Stem Cell Biology Program, Institute of Cell Engineering Johns Hopkins University, Baltimore, MD 21205, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.004 2Library
SUMMARY
Metazoan transcription is controlled through either coordinated recruitment of transcription machinery to the gene promoter or regulated pausing of RNA polymerase II (Pol II) in early elongation. We report that a striking difference between genes that use these distinct regulatory strategies lies in the ‘‘default’’ chromatin architecture specified by their DNA sequences. Pol II pausing is prominent at highly regulated genes whose sequences inherently disfavor nucleosome formation within the gene but favor occlusion of the promoter by nucleosomes. In contrast, housekeeping genes that lack pronounced Pol II pausing show higher nucleosome occupancy downstream, but their promoters are deprived of nucleosomes regardless of polymerase binding. Our results indicate that a key role of paused Pol II is to compete with nucleosomes for occupancy of highly regulated promoters, thereby preventing the formation of repressive chromatin architecture to facilitate further or future gene activation. INTRODUCTION Eukaryotic gene expression begins with recruitment of the transcription machinery to a gene promoter and formation of a preinitiation complex composed of RNA polymerase II (Pol II) and general transcription factors (Roeder, 2005). This step is highly regulated and is enhanced by DNA sequence motifs within the promoter region, which are recognized by general transcription factors to stabilize transcription complex assembly (Juven-Gershon et al., 2008). Interestingly, these core promoter motifs are more prevalent at highly regulated genes than at constitutively active housekeeping genes, suggesting that these two classes of promoters might use different mechanisms to attract the transcription machinery (Basehoar et al., 2004; Hendrix et al., 2008). 540 Cell 143, 540–551, November 12, 2010 ª2010 Elsevier Inc.
Chromatin structure also impacts polymerase recruitment by modulating promoter accessibility, and activation of some genes requires disassembly of promoter nucleosomes by ATP-dependent chromatin-remodeling complexes (Cairns, 2009). In the yeast Saccharomyces cerevisiae, highly regulated promoters are particularly likely to be occluded by nucleosomes before activation, making these genes reliant on nucleosome remodeling for transcription (Tirosh and Barkai, 2008). However, global mapping of nucleosomes in yeast has revealed that most promoter regions display low nucleosome occupancy even when the gene is inactive (Yuan et al., 2005; Albert et al., 2007), suggesting that assembly of promoter nucleosomes is inherently disfavored. Indeed, yeast promoter DNA sequences often contain rigid poly (dA:dT) tracts that deter nucleosome assembly (Iyer and Struhl, 1995). Accordingly, intrinsic sequence preferences for nucleosome formation contribute significantly to accessibility of yeast promoters in vivo (Sekinger et al., 2005; Kaplan et al., 2009; Zhang et al., 2009). Human and Drosophila promoters are also generally nucleosome deprived in a manner that is not dependent on gene expression (Mito et al., 2005; Ozsolak et al., 2007; Mavrich et al., 2008; Schones et al., 2008). However, the mechanisms for this nucleosome depletion appear to be different than in yeast. Metazoan genes are much more G+C-rich than their yeast counterparts and, in contrast to yeast, are reported to intrinsically favor nucleosome formation around their promoters (Kaplan et al., 2009; Tillo et al., 2010). Thus, active mechanisms must contribute to the broad nucleosome depletion observed in metazoans, such as recruitment of chromatin-remodeling complexes or association of the transcription machinery (Kim et al., 2005; Ozsolak et al., 2007). Indeed, pausing of Pol II near promoters can affect both the positioning (Mavrich et al., 2008; Schones et al., 2008) and occupancy of nucleosomes (Gilchrist et al., 2008). Polymerase pausing was first described at the Drosophila heat shock genes, where Pol II synthesizes 25–50 nucleotides (nt) of RNA prior to heat shock and then halts to ‘‘wait’’ for an activating signal (Rougvie and Lis, 1988; Lis, 1998). Heat shock immediately triggers the release of paused polymerase into the gene,
allowing an extremely rapid and robust transcriptional response (Lis, 1998). Rapid activation of heat shock genes is also favored by the lack of nucleosomes within the initially transcribed region (Wu, 1980), which would otherwise present barriers to efficient elongation (Izban and Luse, 1992). Although promoter-proximal pausing was once considered a rare phenomenon, recent work has demonstrated that it is a common regulatory strategy in higher eukaryotes (Muse et al., 2007; Zeitlinger et al., 2007; Core et al., 2008; Nechaev et al., 2010; Rahl et al., 2010). However, despite the growing appreciation for the widespread nature of pausing, the functions of paused Pol II remain to be elucidated. We investigated the relationships among pausing, gene activity, and chromatin structure by performing high-resolution mapping of Pol II, pause-inducing factors, and nucleosomes across the Drosophila genome. Our data reveal that Pol II pausing occurs globally and plays a decisive role in determining promoter nucleosome occupancy. Moreover, we find that genes regulated by pausing rather than Pol II recruitment have distinct ‘‘default’’ chromatin architectures specified by their DNA sequences. Although recruitment-limited genes have intrinsically nucleosome-deprived promoters, genes with paused Pol II require polymerase occupancy to prevent promoter nucleosome assembly. These findings indicate that a gene’s intrinsic nucleosome occupancy in the naive, or default, state is instructive for gene regulation and suggest that the interplay between static information within promoter DNA sequences and the dynamics of polymerase pausing facilitates precise control of gene expression. RESULTS Pausing of Pol II Is Widespread and Occurs at Highly Active Genes Regulation of Pol II pausing involves the coordinated action of both negative and positive elongation factors (Marshall and Price, 1992). Shortly after transcription initiation, the pauseinducing factors negative elongation factor (NELF) and DRBsensitivity inducing factor (DSIF) associate with the polymerase and decrease elongation efficiency (Yamaguchi et al., 2002; Wu et al., 2003; Cheng and Price, 2007; Lee et al., 2008). To examine the prevalence of pausing during early elongation, we used genome-wide ChIP-chip on high-density tiling arrays to compare NELF and DSIF distribution in Drosophila S2 cells with that of Pol II (see Figure S1A and Table S1 available online). Heat maps representing fold enrichment over input DNA (Figure 1A) reveal a broad colocalization of NELF, total Pol II, and DSIF near promoters. In fact, the average promoter signals for these factors correspond extremely well (Figure 1B; Figure S1B), indicating that NELF and DSIF generally associate with Pol II in the promoter-proximal region. Additionally, in agreement with recent reports (Rahl et al., 2010), most genes show enrichment in Pol II signal near promoters relative to downstream regions, suggesting that recruited polymerases are generally released inefficiently into genes. Release of paused polymerase into productive elongation is triggered by the kinase activity of the positive transcription elongation factor b (P-TEFb) (Marshall and Price, 1995; Peterlin and
Price, 2006). P-TEFb phosphorylates the Serine-2 residues on the Pol II C-terminal domain, DSIF and NELF, leading to dissociation of NELF and recruitment of factors that facilitate transcription elongation and RNA processing. The tight correlation between NELF and Pol II signals near promoters suggests that each round of transcription involves NELF-mediated pausing, such that active genes should be enriched in NELF. To confirm this, we identified active genes by performing ChIP-chip with an antibody that recognizes the Serine-2 phosphorylated (Ser2-P) form of Pol II. All heat maps shown in Figure 1A have genes rank-ordered from highest Ser2-P Pol II enrichment within the gene to lowest, clustering active genes at the top. Expression analysis confirms that genes with elevated Ser2-P Pol II signal produced significant levels of mRNA (Figure 1A, mRNA). Notably, the most active promoters were highly enriched in NELF (e.g., Ef2b; Figure 1C; Figure S1C), suggesting that NELF is universally present during early elongation, even at the most highly expressed genes. To determine whether NELF-bound polymerases were engaged in transcription, we evaluated RNA production from each transcription start site (TSS). We found that >85% of Pol II-bound promoters generate significant short ( 0.5) by MannWhitney U test. Different movies were acquired using different settings and therefore should not be compared to each other.
598 Cell 143, 592–605, November 12, 2010 ª2010 Elsevier Inc.
U
h
h
nt
re
48
h
B1-8hi DEC205-/-
36
h
ed
Day 6/7: treatment αDEC-OVA or CS s.c. 3 μg/footpad
24
Day 0: boost NP-OVA s.c. 25 μg/footpad
32 16 8 4 2 1 0.5 0.25
at
2-4 wk Immunize OVA-alum i.p. 50 μg/mouse
DZ/LZ ratio
Day 8: flow cytometry histology
48h 36h 24h 12h
Host: C57BL/6
B1-8hi DEC205+/+
C
12
Day -1: cell transfer 5-10 x 106 total B cells (≈15% Igλ+) from B1-8hi PA-GFP+/+ (15%) + B1-8hi DEC205-/- CD45.1+ (85%)
A
3.98
at
h
h
ed
0
40 20 0
29.2
at re U
nt
DAPI
19.5
15.1
21
16.8
16.5
h
33.3
60
h
21.9
80
48
61.8
100
h
32.4
D
36
LZ 29.2
CD86
58.3
72.4
59.7
36.7
12
DZ 63.8
31
32.5
ed
18
% cells in GC
19.8
U
nt
re
DAPI
48
10.4
h
32.5
20
36
61.1
B1-8hi DEC205-/-
40
h
91
83.6
B1-8hi DEC205+/+
60
24
CXCR4
LZ 23
48h
h
57.8
29.3
CD86
36h
24
24h
12
DZ 68.7
12h
% DAPI high
Untreated
CXCR4
B1-8hi DEC205-/-
B1-8hi DEC205+/+
B
B1-8hi, DEC205-/B1-8hi, DEC205+/+ Endo. (DEC205+/+)
E
LZ
DZ
Untreated
12h
24h
36h
48h
PAGFP (B1-8hi DEC+/+) CD35 (FDC) IgD (Naive B cells)
Figure 6. T Cell Help Availability Limits Interzonal Migration (A) Diagrammatic representation of the experimental protocol. (B) LZ/DZ distribution and DNA content (DAPI) analysis of DEC205+/+ (top) and DEC205/ (bottom) of popliteal lymph node GC B cells at different times after injection of aDEC-OVA. (C) Quantification of (B). Each symbol represents one mouse. (D) Fraction of GC cells derived from transferred DEC205+/+ and DEC205/ B1-8hi and endogenous (DEC205+/+) cells in popliteal lymph nodes of mice treated with aDEC-OVA for different lengths of time. Bars = SEM. (E) Histology showing the distribution of DEC205+/+ cells in popliteal lymph node GCs at different times after injection of aDEC-OVA. DEC205+/+ cells, which express PA-GFP (used here as a genetic label), were stained with anti-GFP antibodies (green). LZ and DZ can be identified by presence of CD35hi follicular dendritic cells (FDCs, in red) and absence of IgD+ B cells (in blue), respectively. All data are representative of two experiments.
of BCR by using chimeric antibodies to DEC205 (aDEC-OVA). Targeting B cells in this manner results in efficient processing and presentation of OVA peptides (Figure S5E) (Kamphorst et al., 2010). To test the role of antigen presentation in interzonal migration, we produced mixed GCs containing DEC205+/+ and DEC205/ B1-8hi B cells by immunizing wild-type mice that had received
a mixture of the two cell types (Figure 6A). aDEC-OVA treatment initially resulted in a striking accumulation of DEC205+/+ cells in the LZ compartment (Figures 6B and 6C). However, by 24 hr after antigen delivery, DEC205+/+ cells began to reappear in the DZ, and by 48 hr, as much as 90% of this population was in the DZ (Figures 6B and 6C). Acquisition of DZ phenotype by DEC205+/+ cells was accompanied by a burst in proliferation Cell 143, 592–605, November 12, 2010 ª2010 Elsevier Inc. 599
Day -1: cell transfer 5-10 x 106 total B cells (≈15% Igλ+) from B1-8hi PA-GFP+/+ CD45.1+ (25%) + B1-8hi DEC205-/- CD45.1+ (75%)
C
1 0.5
PB S
S -O VA
-C
3.52
αD
10 2 10 1
DEC205+/+
EGFP
B220 104
H
Ratio NP3/NP23
1.0
3
10
14
28
Days after boost PBS: NP23 PBS: NP3
αDEC-CS
αDEC-OVA
80
112
***
0.8
**
0.6 0.4 0.2 0.0
αDEC-OVA: NP23 αDEC-OVA: NP3
10 3
B1-8hi DEC205+/+ B1-8 hi -/Endogenous DEC205 9.72 9.72 9 72 72
9
αDEC-CS
Endo (DEC205+/+)
86.6
αDEC-OVA
10 4
B1-8hiDEC205+/+
9.00
10 5
All
PA-GFP (inactive)
αDEC-CS CD138
41.5
No. CD138hi/pLN
Day 5: treatment αDEC-OVA or CS s.c. 3 μg/footpad
F 48.1
0.049
0.74
Day 8: flow cytometry
Day 0: boost NP-OVA s.c. 25 μg/footpad
Gated on live/single
αDEC-OVA
EC
αD
αD 2-4 wk Immunize OVA-alum i.p. 50 μg/mouse
0
αD
EC
PB S
-O VA
S -C
0
EC
EC
PB S
αD
Host: C57BL/6
102
10
Day -1: cell transfer 6 x 106 total B cells (≈15% Igλ+) from B1-8hi PA-GFP+/+ (50%) + B1-8hi DEC205-/- EGFP +/- (50%)
D
E
20
0.25
0
IgG1 titer
% DAPI high (DEC205-/-)
20
DZ/LZ ratio (DEC205-/-)
40
2
S
Endogenous (DEC205+/+)
VA
B1-8hiDEC205 +/+
60
**
30
-O
80
**
4
-C
B1-8hiDEC205 -/-
% of GC cells
100
EC
B
G
Day 5: treatment αDEC-OVA or CS s.c. 3 μg/footpad
Day 0: boost NP-OVA s.c. 25 μg/footpad
EC
2-4 wk Immunize OVA-alum i.p. 50 μg/mouse
Day 8: flow cytometry
αD
Host: C57BL/6
B1-8hiDEC205-/-
A
9
14
28
Days after boost αDEC-OVA
PBS
W33L+
W33L-
pχ2 50, ± SD); ** p < 0.001, as determined by a Student’s t test. See also Figure S1.
peroxidases. Peroxidases are known to regulate the levels of certain ROS, particularly hydrogen peroxide and superoxide. Staining for the presence of these ROS in the root and altering their concentrations using chemical reagents showed a clear correlation between growth rate, location of the TZ, and the relative distribution of different ROS species in the meristematic and elongation zones. Interestingly, differences in the localization of UPB1 in transcriptional and translational reporter lines suggest that this transcriptional regulator might also function as an intercellular signaling molecule. We show evidence that UPB1 provides a direct transcriptional link between ROS distribution and the proliferation status of the cells in the root tip. RESULTS UPBEAT1 Controls the Transition from Cellular Proliferation to Differentiation To identify transcription factors (TFs) that regulate the first stages of the transition from cellular proliferation to differentiation, we
analyzed the RootMap gene expression data (Brady et al., 2007a) and identified approximately 100 TFs with increased expression at the boundary between the meristematic and elongation zone, which marks the onset of differentiation. We screened T-DNA insertional mutant lines available for 96 of these genes, looking for alterations in primary root growth rates. One line (SALK_115536) developed a longer root than wild-type and contained an insertion in At2g47270 (Figures 1A and 1B). This gene showed a particularly prominent expression peak at the boundary between the meristematic and elongation zone (Figure S1A available online). By qRT-PCR, we confirmed that the T-DNA insertion caused a strong reduction in expression of At2g47270, which encodes an uncharacterized protein with a bHLH domain (Figures S1B and S1C). The bHLH domain occupies 70% of the 102–amino acid protein, which we named UPBEAT1 (UPB1) (Figure 1C). The Arabidopsis genome contains 147 genes that are predicted to encode proteins with a bHLH domain. UPB1 belongs to the bHLH-subfamily 14, but it is only distantly related to the other members of that subfamily (Toledo-Ortiz et al., 2003). Apart from the bHLH domain, no other functional domains were predicted in UPB1. To determine the effects of the insertional mutation (upb1-1) on root growth, we counted the number of cortex cells in a cell file extending from the quiescent center (QC) to the first elongated cell, as a measurement of meristem size (Dello Ioio et al., 2007). We found a significant increase in cortex cell number in the upb1-1 mutant, indicating enlargement of the meristem (Figures 1D and 1E). No difference was detected in Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc. 607
Figure 2. UPB1 Gene and Protein Expression Patterns in the Root (A–C) Expression of pUPB1::GFP (transcriptional fusion) in wild-type. (D–F) Localization of pUPB1::UPB1-GFP (translational fusion) in upb1-1 background. (G–I) Localization of pUPB1::UPB1-3YFP in upb1-1 background. The three zones with GFP signal are depicted: meristematic zone (A, D, and G), transition zone (B, E, and H), and the lateral root cap (C, F, and I). Note that the lateral root cap images are of the surface of the root, whereas the meristematic and transition zone images are of median longitudinal optical sections. Scale bars, 50 mm. See also Figure S2.
the radial pattern of root cell layers between upb1-1 and wildtype, suggesting that UPB1 functions primarily in regulating root growth as opposed to patterning. The only other available T-DNA insertion was located at the end of the 30 UTR of UPB1. We were unable to detect any phenotypic effects or significant UPB1 expression level changes in this line. To further explore the function of UPB1, we created ectopic expression lines with the constitutive 35S promoter and a fluorescent reporter (35S::UPB1-3YFP). Three independent lines showed reduced root length and a decrease in cortex cell number in the meristematic zone. (Figure 1). To determine whether cell size was also regulated by UPB1, we measured the length of the first cell in the maturation zone in upb1-1 and in the 35S::UPB1-3YFP transformants. In upb1-1, cells were longer than in wild-type, whereas in the 35S::UPB1-3YFP lines, cells were shorter than in wild-type (Figure S1D). Taken together, these results suggest that UPB1 acts as a regulator of root growth through modulation of the transition from cell proliferation to elongation as well as playing a role in controlling cell size. The UPB1 Protein Appears to Move from the Lateral Root Cap to the Elongation Zone A correlation exists between the position of the lateral root cap (LRC) furthest from the root tip and the TZ (Willemsen et al., 608 Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc.
2008), but the underlying mechanism is unknown. The tissuespecific expression data indicate that UPB1 mRNA is expressed in the vascular tissue as well as in the COBRA-like 9 (COBL9) expression domain, which includes root hair and LRC (Brady et al., 2007a, 2007b). To begin to determine whether UPB1 might play a role in the signaling process between the LRC and the TZ, we constructed transcriptional and translational fusions and transformed them into plants. For the transcriptional reporter, we used 3002 bp upstream of the putative start codon fused to GFP (pUPB1::GFP). Five independent lines exhibited strong fluorescence in cells of the LRC close to the TZ. Outside of the LRC, fluorescence was detected in the vascular tissue of the elongation and maturation zones (Figures 2A–2C). For the translational reporter, we used the same promoter upstream of the UPB1-coding region fused to GFP (pUPB1::UPB1-GFP). Interestingly, in plants containing this construct, we detected low GFP fluorescence in the LRC and the meristematic zone. Fluorescence was primarily localized to the nuclei of all cell types in the elongation zone (Figures 2D–2F). We also detected weak fluorescence in the maturation zone. Expression of the translational fusion in the upb1-1 line rescued the mutant phenotype (Figures S2A–S2C), indicating that the fusion protein functions in a manner similar to the native UPB1 protein. The difference between mRNA and protein localization suggests that the UPB1 protein might move from the LRC or vascular tissue to function in all cell files in the elongation zone. To test this, we used the same promoter but fused the UPB1-coding region to a triple yellow fluorescent protein (3YFP) tag, because this tag has a high molecular weight and has been used to reduce protein movement (Kurata et al., 2005). In pUPB1::UPB1-3YFP plants, YFP fluorescence was detected primarily in the nucleus of the LRC, in addition to all cells in the elongation zone (Figures 2G– 2I). We conclude from this localization pattern that movement was reduced in these lines but not eliminated. Consistent with this observation, the mutant phenotype was partially rescued (Figure S2D). Identification of Genes Regulated by UPB1 The spatial distribution of the UPB1 protein suggested that it might exert a different effect on gene expression in the meristematic and elongation zones. Therefore, we isolated the meristematic and elongation zones by microdissection and extracted RNA from each section independently. In the meristematic zone of upb1-1 mutants, expression of only 55 genes was significantly altered in comparison to wild-type (2-fold change, p < 0.05), whereas expression of 738 genes was affected in the elongation zone of the mutant roots (Figure 3A). This finding was consistent with the hypothesis that UPB1 is primarily active in the elongation zone. Further support came from microarray analysis of ectopic expression of UPB1 in the upb1-1 mutant. Stronger transcriptional effects were observed in the meristematic zone (1809 genes) than in the elongation zone (812 genes). If UPB1 normally acts in the elongation zone, then ectopic expression in the meristematic zone would be expected to have a greater effect there than in the elongation zone where it normally functions. Because of the minimal effects of the upb1-1 mutation on gene expression in the meristematic zone, we did not include these data in our subsequent analyses.
activity’’ (p < 1011). Another highly enriched category was ‘‘response to reactive oxygen species’’ (p < 104). We found these to be particularly interesting because recent reports indicated that peroxidases and ROS play an important role in controlling root growth (Dunand et al., 2007; Passardi et al., 2006). The ROS produced by peroxidases is essential in the peroxidase-mediated formation of lignin (Ros Barcelo, 1997), which is required for the formation of primary cell walls. We also found that ‘‘lignin biosynthetic process’’ (p < 103) and ‘‘lignan biosynthetic process’’ (p < 106) GO categories were overrepresented within derepressed genes in upb1-1. Further indication that UPB1 regulates cell wall remodeling is found in other enriched GO categories: ‘‘cellular glucan metabolic pathway’’ (p < 106), ‘‘cell wall’’ (p < 104), ‘‘phenylpropanoid biosynthesis’’ (p < 105), and ‘‘secondary cell wall biogenesis’’ (p < 103). These findings raise the possibility that UPB1 controls growth by regulating ROS activity, which in turn regulates cell wall remodeling. Among genes that were positively regulated by UPB1, the enriched GO categories were ‘‘tRNA processing’’ (p < 106), ‘‘ATPdependent helicase activity’’ (p < 106), ‘‘DNA-directed RNA polymerase activity’’ (p < 105), ‘‘maintenance of DNA methylation’’ (p < 104), and ‘‘regulation of cell cycle’’ (p < 103). These would also appear to be involved in growth and replication processes although we think it is likely that these are indirect responses. In particular, the genes in the cell cycle GO category contained both S-phase and M-phase activated genes, indicating that there is no specificity for a particular phase of the cell cycle. We note that we did not find enrichment for genes involved in plant hormone homeostasis or signaling. Figure 3. Expression and Function of UPB1 Response Genes (A) Transcriptome changes upon alteration of UPB1 expression levels. The heat map includes all genes that were differentially expressed in at least one of the microarray experiments. Green indicates an increase in expression, red indicates a decrease in expression; color intensity indicates the magnitude of the effect. (B and C) Enriched Gene Ontology (GO) categories within gene lists consisting of genes that are negatively (B) or positively (C) regulated by UPB1. See also Figure S3 and Table S1.
Meta-analysis approaches to integrate diverse and inherently noisy data have proven useful to identify transcription factor target genes (Busch et al., 2010; Levesque et al., 2006). We performed a meta-analysis of the three datasets in which significant changes occurred. Analogous to Busch et al. (2010), we obtained a cumulative Z score (cZS) for each gene and empirically determined a significance threshold for the cZS by random sampling (see Supplemental Information for details). Using the significance threshold of p < 0.01, we identified 2375 UPB1responsive genes that exhibited consistent expression changes upon modulation of UPB1 expression (Table S1). To investigate the biological function of these genes, we identified significantly enriched Gene Ontology (GO) categories (Figures 3B and 3C). GO enrichment analysis associates each gene of a list with distinct biological processes and then evaluates whether the list contains more genes than expected by chance for a certain biological process. The most significantly enriched GO category of genes negatively regulated by UPB1 was ‘‘peroxidase
Identification of UPB1 Direct Targets To identify UPB1 direct targets, we performed chromatin immunoprecipitation of UPB1 followed by hybridization to a custom oligonucleotide microarray (ChIP-chip). We reasoned that a subset of the genes found by microarray expression analysis should be overrepresented in the ChIP-chip binding data. On the basis of this assumption, we systematically explored the parameter space to optimize our detection settings for enriched regions in the ChIP-chip experiments (for details, see Extended Experimental Procedures). By using highly stringent conditions in which transcriptionally regulated genes were enriched, we identified 166 putative UPB1 direct target genes (Table S2). These included genes of various annotated functions, including quite a few transcription factors indicating that UPB1 might regulate a cascade of transcription factors. Of particular interest were three peroxidase genes (Figure 4A) that are expressed highly at the boundary of the meristematic and elongation zone in the RootMap datasets (Figure S3). All three (At4g11290; Per39, At4g16270; Per40, At5g17820; Per57) are up-regulated in the upb1-1 mutant and down-regulated in the UPB1 ectopic expressor (Figure 4B). Consistent with these results, the phenotype of an overexpressor of another peroxidase gene (Per34) had longer roots than wild-type, whereas the double knockdown line of Per33 and Per34 had shorter roots (Passardi et al., 2006). To investigate whether the UPB1 target peroxidases also play a role in root growth, we ectopically expressed Per57 by driving its cDNA with the 35S promoter. Three independent lines Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc. 609
Figure 4. UPB1 Binding of Upstream Regulatory Regions and Resulting Transcriptional Effects (A) Binding profile of UPB1 to upstream regulatory regions of At4g11290, At4g16270, and At5g17820. UPB1-bound chromosomal regions are shown by average Z scores of enrichments in the 4 ChIP-chip experiments (y axis). Genomic positions (x axis) are given relative to the annotated transcription start of the indicated primary RNA. Shaded areas indicate genomic regions that were detected as enriched. (B) Expression profiles of UPB1 direct targets, At4g11290, At4g16270, and At5g17820 from microarray data. (C) Root tip morphology of 6 dai Col-0, upb1-1, and 35S::Per57-GFP (line #1 to #3) plants. Blue arrowheads mark QC cells; white arrowheads indicate cortex transition zones. Scale bar, 50 mm. (D) Average number of cells in root meristems of Col-0, upb1-1 and 3 independent lines of 35S:: Per57-GFP (line #1 to #3) plants (n > 20, ± SD; **p < 0.001, Student’s t test; *p < 0.05). See also Table S2 and Figure S7.
displayed a significantly larger meristem than wild-type. The phenotype was not as pronounced as in upb1-1 mutants (Figures 4C and 4D), suggesting that the peroxidases directly controlled by UPB1 play additive roles in regulating root growth. Redox Processes Are Important for Regulating Root Growth Upon treatment with H2O2 for 24 hr, the root meristem in wildtype became significantly shorter (Figures 5A and 5C) and root length was reduced. In contrast, H2O2 treatment of the upb1-1 mutant and UPB1 ectopic expressor did not result in a significant change of meristem size (Figures 5A, 5F, and 5I) or growth. Scavenging H2O2 by treating with potassium iodide (KI) in wild-type resulted in a longer root, which contained a larger root meristem (Figures 5A and 5D). The same treatment had no effect on the meristem size in the upb1-1 mutant but significantly increased 610 Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc.
the meristem size of the UPB1 ectopic expressor (Figures 5A, 5G, and 5J). The insensitivity of the upb1-1 mutant to H2O2 scavenging is consistent with the up-regulation of peroxidase genes and a decrease in H2O2 levels (Figure S4). Moreover, subjecting the mutant to additional H2O2 has no effect on meristem size (Figure 5), consistent with the idea that the upregulated peroxidases are able to continually scavenge even an excess of H2O2. To directly investigate the role of peroxidases in root growth, we treated plants with salicylhydroxamic acid (SHAM), an inhibitor of peroxidase activity (Brouwer et al., 1986). Wild-type meristems treated with SHAM became significantly smaller than those of untreated roots. The same treatment led to a decreased size of the meristem in upb1-1 plants, to the extent that they were similar to untreated wildtype roots (Figure 5K). Conversely, the inhibitor had almost no effect on root meristem size in the UPB1 ectopic expressor (Figure 5K). Using a different peroxidase inhibitor, KCN (Bestwick et al., 1997), which inhibits a broader spectrum of peroxidases (Chen and Asada, 1989), even stronger effects could be observed. These included a decrease of meristem size in the UPB1 ectopic expressor (Figure 5K). These results indicated that H2O2 content and the regulation of H2O2 by UPB1-controlled peroxidase is important for root growth. To investigate the spatial aspect of this regulation, we determined the distribution of H2O2 by applying 30 -(p-hydroxyphenyl) fluorescein (HPF) to roots. HPF is known to stain
Figure 5. Effects of ROS Level Changes (A) Average number of cells in root meristems of Col-0, upb1-1, and 35S::UPB1-3YFP #2 plants (n > 30, ± SD; **p < 0.001, Student’s t test; *p < 0.05). (B–J) Root meristems of 6 dai plants upon various treatments for 24 hr. Scale bars, 50 mm. Panels show untreated Col-0 plant (B), 100 mM hydrogen peroxide (H2O2) (C), 1 mM potassium iodide (KI) treated Col-0 plants (D), untreated upb1-1 mutant (E), 100 mM H2O2 (F), 1 mM KI treated upb1-1 mutants (G), untreated 35S::UPB1-3YFP #2 (H), 100 mM H2O2 (I), and 1 mM KI treated 35S::UPB1-3YFP #2 plants (J); blue arrowheads indicate cells of the QC and white arrowheads indicate the cortex transition zone. (K) Average number of cells in root meristems of Col-0, upb1-1 and 35S::UPB1-3YFP plants after 24 hr treatment with 100 mM SHAM, 100 mM KCN, and 100 mM DPI (n > 30, ± SD; **p < 0.001, Student’s t test; *p < 0.05). See also Figure S4.
hydrogen peroxide (Dunand et al., 2007). The intensity of HPF fluorescence changed according to our expectations after different treatments that modified ROS levels in roots (Figures S4A–S4M). In wild-type, strong HPF fluorescence was detected in the columella, the LRC, in the vasculature, and the epidermis of the elongation zone. Fluorescence was substantially weaker in the meristematic zone than in the elongation zone (Figures 6B and 6D). In the upb1-1 mutant, less HPF fluorescence in the entire root was observed (Figures 6A and 6D). Conversely, in the UPB1 ectopic expressor, HPF fluorescence was increased (Figures 6C and 6D). We also used the highly specific H2O2 indi-
cator, BES-H2O2-Ac (Maeda, 2008; Maeda et al., 2004), with results similar to those of HPF staining (Figures 6E–6H). These two independent assays provide strong support for our hypothesis that increased peroxidase activity in upb1-1 results in lower H2O2 levels, whereas repression of peroxidases in the meristem in the ectopic expression lines results in an increase in H2O2 levels. Another aspect of ROS growth regulation involves superoxide (O2,), which is thought to be produced by the activity of NADPH oxidases and has been shown to affect root growth and root hair development (Foreman et al., 2003). To investigate the role of O2, in UPB1-regulated root growth, we treated plants with diphenylene iodonium (DPI), which primarily inhibits NADPH oxidase activities. In both wild-type and upb1-1, treatment with DPI resulted in a reduction in meristem size but there was no effect on the meristem of the UPB1 ectopic expressor (Figure 5K). Furthermore, to determine the distribution of O2, in the root, we stained roots with nitroblue tetrazolium (NBT), which is widely used as an indicator of O2, levels (Bielski et al., 1980). In wild-type, strong staining appeared in all cell types in the meristematic zone, whereas only the vascular tissue in the elongation zone showed staining, indicating that O2, preferentially accumulates in the root meristematic zone (Figure 6J). In upb1-1, staining of the meristematic and elongation zones was more intense compared to wild-type plants (Figures 6I and 6L). Interestingly, the O2, level in wild-type roots treated with KI appeared similar to that of upb1-1 roots (Figures S4N–S4Z). In contrast, in the UPB1 ectopic expressor, staining of the meristematic and elongation zones appeared weaker than in wild-type (Figures 6K and 6L). We also used dihydroethidium (DHE) as a second indicator for O2, (Owusu-Ansah et al., 2008). DHE fluorescence was similar to the NBT staining (Figures 6M–6P). These results are consistent with UPB1 functioning as a regulator of ROS production through repression of peroxidase gene expression. Finally, we performed simultaneous staining for H2O2 and O2, by using BES-H2O2-Ac and DHE on the same roots (Figures 6Q–6S). The simultaneous staining results suggest that there are opposing gradients of H2O2 and O2,. The crossing point of these gradients is altered in plants with modified UPB1 expression and coincides in each case with the onset of differentiation. This raises the possibility that the crossing point of the two gradients might determine the position of the TZ in the Arabidopsis root tip. To determine whether ROS also influences the process of cell expansion, we measured the size of the first mature cells in upb1-1, the UPB1 ectopic expressor and the Per57 overexpressor, as well as after different ROS treatments. We found a positive correlation between cell length and meristem size (Figure S5C). Finally, we examined UPB1 expression after treatment with H2O2 and KI. H2O2 treatment caused up-regulation of UPB1 expression, whereas KI treatment reduced UPB1 expression (Figure S5A). This indicates that H2O2 levels regulate UPB1 expression creating a feed back loop in ROS signaling. Taken together, our results are consistent with a model in which UPB1 acts to repress peroxidase expression in the elongation zone. In the upb1-1 mutant, this results in an increase in O2, and a decrease in H2O2 in this region of the root (Figure S5B). The balance of these ROS molecules appears to be Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc. 611
Figure 6. ROS Distribution and Proliferation/Differentiation (A–S) Six dai roots stained with 30 -(p-hydroxyphenyl) fluorescein (HPF) (A–C), BES-H2O2-Ac (E–G), nitroblue tetrazolium (NBT) (I–K), dihydroethidium (DHE) (M–O), and both DHE and BES-H2O2-Ac (Q–S). Panels show upb1-1 (A, E, I, M, and Q), Col-0 (B, F, J, N, and R), 35S::UPB1-3YFP #2 (C, G, K, O, and S). Scale bars, 50 mm. Quantification of HPF (D), BES-H2O2-Ac (H), and DHE fluorescent intensity (P) are shown (n = 20, ± SD). Quantification of NBT staining intensity is seen (L) (n = 20, ± SD; **p < 0.001, Student’s t test; *p < 0.05). (T) Model of UPB1-dependent regulation of meristem size. In Col-0 root tips (center panel), superoxide (O2,) accumulates in the meristem (blue area), whereas hydrogen peroxide (H2O2) accumulates in the elongation zone (green area). UPB1 represses expression of peroxidases (pink circles) in the elongation zone. In the upb1-1 mutant (left panel) peroxidases are derepressed and higher abundance of peroxidases leads to increased content of O2,. On the other hand, UPB1 ectopic expressor (right panel) represses peroxidases and leads to increased levels of H2O2. See also Figure S5.
important for making the transition from cell proliferation to differentiation (Figure 6T). UPB1 Does Not Appear to Act through Cytokinin and Auxin Signaling Our data strongly implicate ROS as being central to UPB1 function in regulating root growth. It is known that the ratio of cytokinin to auxin is important in controlling the balance of cell division and differentiation in the root, and that two key transcription factors, SHY2/IAA3 and ARR1, control this process (Dello Ioio et al., 2008). The arr1 mutant has a large meristem 612 Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc.
phenotype similar to that of upb1-1. A first indication that UPB1 is not involved in the same signaling pathway as ARR1 and SHY2 came from examination of our microarray datasets in which neither gene appears to be responsive to UPB1 (Table S1). We confirmed this by qRT-PCR (Figure S6). Exogenous cytokinin application decreased the meristem size of upb1-1 in a similar fashion to that of wild-type, whereas exogenous auxin increased the meristem cell number in upb1-1 and in the UPB1 ectopic expressor in a manner similar to wild-type (Figure 7). Furthermore, exogenous application of auxin and cytokinin had almost no effect on UPB1 expression in either
Figure 7. Effects of Cytokinin and Auxin on Developmental Zones of the Root (A) Average number of cells in root meristems of Col-0, upb1-1, and 35S::UPB1-3YFP #2 plants (n > 30, ± SD; **p < 0.001, Student’s t test; *p < 0.05). (B–J) Root meristems of 6 dai plants after hormonal treatment for 24 hr (scale bars, 50 mm): untreated Col-0 plant (B), 0.5 nM IAA (C), 5 mM transzeatin (Zt) treated Col-0 plants (D), untreated upb1-1 mutant (E), 0.5 nM IAA (F), 5 mM Zt treated upb1-1 mutants (G), untreated 35S::UPB1-3YFP #2 (H), 0.5 nM IAA (I), and 5 mM Zt treated 35S::UPB13YFP #2 plants (J). Blue arrowheads indicate cells of the QC and white arrowheads indicate the cortex transition zone. See also Figure S6.
the wild-type meristem or elongation zone (Figure S6). These results indicate that UPB1-mediated regulation of meristem size is likely to be independent of auxin and cytokinin signaling. DISCUSSION UPB1 Regulates ROS Signaling to Control the Transition from Proliferation to Differentiation Our microarray expression analysis coupled with UPB1 ChIPchip analysis indicated that UPB1 directly represses a set of peroxidases as cells begin to differentiate. The use of chemical inhibitors and the ectopic expression of one of the target peroxidase genes (Per57), as well as chemical indicators for ROS provided strong evidence that these peroxidases control ROS distribution, which in turn governs the transition from proliferation to differentiation. Further support for this hypothesis comes from treatment with peroxidase inhibitors, which cause a reduction in the size of the root meristem, indicating an earlier onset of differentiation (Figure 5K). These genetic and chemical studies reveal the importance of peroxidase activity in the root tip and are consistent with our interpretation of the upb1-1 phenotype as being directly related to UPB1 regulation of at least three Class III peroxidases. In Arabidopsis, there are 73 Class III peroxidase genes (Tognolli et al., 2002) of which 60 are represented on the ATH1 array, and 21 of these were affected by UPB1. All 21 genes are expressed primarily in the elongation zone (Figure S3). We obtained T-DNA insertion lines for many of the peroxidases that are UPB1 targets, but the single mutants did not show obvious phenotypes (data not shown), probably because of functional redundancy. Peroxidases have two opposite functions: one is the reduction of H2O2 by moving electrons to various donor molecules and the second is to catalyze the hydroxylic cycle, which results in the
formation of ROS, particularly O2,. In the upb1-1 mutant, the distribution of O2, and H2O2 was altered apparently as the result of derepression of a set of peroxidases. In living organisms, H2O2 is more stable than O2, (Pitzschke et al., 2006). O2, is transformed into H2O2 both spontaneously and through enzymatic activity of superoxide dismutase and other enzymes, such as apoplastic oxalate oxidase (Caliskan and Cuming, 1998), diamine oxidase (Federico and Angelini, 1986), or peroxidase (Elstner and Heupel, 1976). Our results indicate that, in wild-type, O2, accumulates primarily in the meristematic zone whereas H2O2 accumulates mainly in the elongation zone. Given the changes in ROS distribution in upb1-1 and 35S::UPB1-3YFP, it would appear that O2, and H2O2 distribution are important for localization of the transition zone. In the upb-1 mutant, the consumption of H2O2 by peroxidases in the elongation zone might drive production of O2, in the meristematic zone to maintain ROS homeostasis. In 35S::UPB1-3YFP, these peroxidases are repressed, which leads to accumulation of H2O2 in the meristematic zone. Consistent with this interpretation, SHAM treated roots have H2O2 accumulation in the meristematic zone (Figure 6, Figure S4, and Dunand et al., 2007) and at least five NADPH oxidase genes were up-regulated in the upb11 mutant (Table S1). We also determined the meristem size of 3 NADPH oxidase mutants (AtrbohC, AtrbohD, and AtrbohC). None of them showed a reduction in meristem size (data not shown). This finding suggests that there is functional redundancy among these genes for root meristem maintenance because inhibition of NADPH oxidase activity by DPI resulted in a reduction in meristem size (Figure 5K). These results lead to the following working model. Maintenance of cellular proliferation requires an accumulation of O2,, whereas cellular differentiation requires elevated H2O2 levels. These two different ROS environments coincide with the meristematic and the elongation zone, as visible in the double staining of H2O2 and O2,. Because of the gradient nature of the ROS species distribution, the cells that enter the transition zone can still proliferate. Once the ratio between O2, and H2O2 concentrations reaches a certain level, the cells stop Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc. 613
proliferating and begin to elongate (Figure 6T). According to this working model, the ROS balance in the transition zone plays a critical role and UPB1 is one of the key regulators in maintaining this balance. Because H2O2 itself affects UPB1 expression, this regulatory system contains a feedback loop that might play a role in ROS homeostasis. Because high levels of ROS can damage cells, a feedback loop could allow the plant to maintain proper ROS levels. It would also constitute a system for adjusting ROS levels to maintain the proper balance between proliferation and differentiation. To further investigate this hypothesis, we mined the oxidative stress data available in the public microarray databases. The AtGenExpress dataset contains experiments in which methyl viologen, a compound that causes continuous formation of O2, (Asada, 2006) was used. UPB1 was up-regulated after 24 hr of treatment, and the peroxidases that are direct targets of UPB1 showed decreased expression. Some peroxidases that are not targets of UPB1 showed increased expression, indicating that there might be compensatory expression responses to maintain ROS homeostasis (Figure S5D). Alternatively, the up-regulated peroxidases might be expressed in other organs or tissues. Other TFs have been implicated in ROS signaling in the Arabidopsis shoot. In the ascorbate peroxidase 1 (apx1) mutant, which is a cytosolic H2O2-scavenging enzyme in the Arabidopsis leaf, heat shock factor 21 (HSF21) was identified as a key transcriptional regulator for ROS response upon light stress (Davletova et al., 2005a). A dominant negative construct for HSF21 impaired ZAT12 expression, which was known as a transcriptional regulator of oxidative stress responses (Davletova et al., 2005b). However, neither HSF21 nor ZAT12 expression was affected in our microarray data. Furthermore, the meristem size of the apx1 mutant was the same as wild-type (data not shown). This may indicate that different ROS sensing and signaling systems exist in different tissues or organs. Redox homeostasis is also important for ROS signal transduction. Thioredoxins (TRX) are known as the key regulators for cell redox homeostasis (Meyer et al., 2005). We found a TRX reductase (NTRA; At2g17420) as one of the UPB1 direct target genes (Table S2). It was reported that ntra or ntrb single mutants do not show any phenotype (Reichheld et al., 2007) because of strong functional redundancy. However, the ntra/ntrb double mutant had a small meristem phenotype (Bashandy et al., 2010). NTR is important for reducing oxidized thioredoxin and thioredoxin plays an important role in providing reducing power to the peroxidases (Nordberg and Arner, 2001). These results also indicate that redox homeostasis in the root meristem plays an important role in the transition from cell proliferation to cell differentiation. ROS Distribution Is Important for the Transition from Proliferation to Differentiation in Plants and Animals ROS has also been shown to play an important role in maintaining the balance between cell proliferation and differentiation in animals. A redox-dependent signaling pathway controls the induction of cell division through the regulation of cyclinD1 expression (Burch and Heintz, 2005). Distribution of specific ROS appears to act as an important signal at the transcriptional and posttranscriptional levels during cell-cycle progression in animal cells (Menon and Goswami, 2007). For example, in 614 Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc.
Drosophila, changing ROS levels can switch the status of hematopoietic cells from proliferation to differentiation (Owusu-Ansah and Banerjee, 2009). Moreover, it has been shown that manganese superoxide dismutase (MnSOD) activity regulates cellcycle progression through modulation of ROS levels, which control expression of both the cyclinB1 and cyclinD1 genes in mouse cells (Sarsour et al., 2008). The authors proposed that O2, regulates the proliferative cycle whereas H2O2 induces quiescence (Sarsour et al., 2008). This would be analogous to our model, in which O2, accumulates in the meristematic zone and is necessary for proliferation, whereas H2O2 accumulates in the elongation zone when cells arrest division and begin differentiation. Intriguingly, we found that cell-cycle-related genes that are up-regulated by UPB1 including cyclinB and cyclinD genes do not appear to be direct targets according to our ChIP-chip data. Thus, it seems likely that UPB1 regulates cell-cycle progression indirectly by controlling ROS homeostasis. Peroxidases are known to modify cell walls, mainly through lignin modification. In fact, class III peroxidases tend to localize in the extracellular space known as the apoplast (Passardi et al., 2006), where they can directly modify cell wall structure. Interestingly, we detected the UPB1 direct target, Per57-GFP fusion protein in the cytoplasm as well as in the apoplast (Figure 4C). According to our microarray data, in addition to peroxidases, UPB1 also regulates the ‘‘lignin synthesis’’ GO category. This suggests that UPB1 may act both directly and indirectly to modify cell walls. Thus, regulation of ROS status could act on the cell cycle to stop proliferation and, at the same time, act to modify cell walls to initiate cell expansion. The joint analysis of ROS-dependent changes to meristem cell number and cell length indicated that the ROS effects on cell division and cell length are correlated (Figure S5C). This supports a dual role for ROS, even though these functions would be compartmentalized as the cell wall modification would occur in the apoplast and the cell-cycle modulation would occur inside the cell. UPB1 Regulation of Root Growth Is Independent of the Auxin/Cytokinin Signaling Pathway It has been reported that auxin and cytokinin play an important role in controlling the balance between cell division and differentiation in the root meristem through two transcription factors, ARR1 and SHY2 (Dello Ioio et al., 2008). UPB1 gene expression was not affected by either auxin or cytokinin, and in the upb1-1 mutant, SHY2 and ARR1 expression levels and response to either auxin or cytokinin were similar to wild-type indicating that UPB1 acts through a pathway independent of this hormonal signaling pathway. It is surprising that two pathways can exert a powerful control on the balance between cell division and differentiation independently of each other. One explanation might be that the root has to integrate different types of information. Hormones usually serve as long distance signals, whereas the ROS pathway may play a key role in response to local cues and homeostasis. Interestingly, a role for ROS in local, rapid developmental decisions has emerged in animals. In zebrafish, a local gradient of ROS is used to rapidly trigger and execute a developmental program to recruit leukocytes to wounded tissue sites (Niethammer et al., 2009).
Although our results strongly suggest that UPB1 is involved in a pathway distinct from hormonal signaling for controlling the transition from proliferation to differentiation in the root, it remains possible that these two signaling systems converge at some level.
contained H2O2 (100 mM), KI (1 mM), SHAM (100 mM), KCN (100 mM), DPI (100 mM), IAA (0.5 nM), and trans Zeatin (Zt) (5 mM), respectively. Nitroblue tetrazolium (NBT) and dihydroethidium (DHE) were used for superoxide (O2,) staining and 30 -(p-hydroxyphenyl) fluorescein (HPF) and BESH2O2-AC (WAKO, Japan) were used for hydrogen peroxide (H2O2) staining.
UPB1 May Act Non-Cell-Autonomously There is an intriguing correlation between the height of the LRC and the point of transition from proliferation to differentiation. Comparison of transcriptional and translational fusions of UPB1 suggests that UPB1 protein may be synthesized in the LRC and then move to the elongation zone where it becomes nuclear localized. Alternatively, UPB1 protein may be made at low levels in the elongation zone and have a long half-life, allowing it to accumulate in these cells. Consistent with the first hypothesis, UPB1 is only 102 amino acids in length, suggesting that it could move passively through plasmodesmata. Additional support for this hypothesis came from analysis of a 3YFP-–tagged UPB1 protein, which is localized to the LRC in addition to all cells in the elongation zone. This altered localization pattern is presumably due to the larger size of the 3YFP tagged version, which may prevent passive diffusion from the LRC. The 35S::UPB1-3YFP in upb1-1 affected gene expression in the opposite manner to that of the upb1-1 mutant. This finding indicates that the 3YFP tag does not disrupt UPB1 protein function as a transcriptional regulator. Taken together, these data suggest that at least some UPB1 is synthesized in the LRC and then moves to act in all cells of the elongation zone. In this way, UPB1 may act as a signal from the LRC indicating the proper location of the transition zone.
ACCESSION NUMBERS
EXPERIMENTAL PROCEDURES See Supplemental Information for details. Plant Material and Treatment Arabidopsis thaliana Columbia-0 (Col-0) was used as wild-type unless otherwise noted. The T-DNA insertion lines for upb1-1 (SALK_115536) and upb12 (SALK_133978) were confirmed using PCR with the primers listed in Table S3. For characterization of phenotypes, seeds were sown and allowed to germinate on vertically positioned media plates for 5 days, and then seedlings were transferred onto the treatment media. Microarray Experiments Total RNA was isolated from approximately 60 meristems and elongation zones of Col-0, upb1-1, and 35S::UPB1-3YFP #2 plants. Two biological replicates were performed for each experiment. Fragmented cRNA probes were prepared using the two-cycle amplification protocol recommended by Affymetrix. Samples were submitted to Expression Analysis Inc. (Durham, NC, USA) for hybridization to Arabidopsis whole genome ATH1 Affymetrix GeneChip. ChIP-chip Experiment pUPB1::UPB1-GFP lines were germinated and grown on the MS media for 6 days. Whole root tissue was fixed and chromatin immunoprecipitation (ChIP) was performed as described in Leibfried et al. (2005) with some modification, including the chromatin shearing by using a Bioruptor UCD-200 (Diagenode). ROS-Related and Plant Hormone Treatments Chemical treatments were performed by the transferring the 5-day seedling from MS media to the MS media containing those chemicals for 24 hr. Media
All microarray data, including the tilling array design, were submitted to the NCBI GEO database under accession numbers GSE21876 and GSE21741.
SUPPLEMENTAL INFORMATION Supplemental Information includes Extended Experimental Procedures, seven figures, and three tables and can be found with this article online at doi:10.1016/j.cell.2010.10.020. ACKNOWLEDGMENTS This work was supported by a grant from the NSF Arabidopsis 2010 program. H.T. was supported by a JSPS Postdoctoral Fellowship for Research Abroad. We thank H. Cui for advice on ChIP-chip; R. Mittler for the apx1 mutant; M.A. Torres for AtrbohC, D, and F mutants; K.D. Birnbaum, Y. Helariutta, Y. Tada, X. Dong, and B. Scheres for comments on the manuscript; and members of the Benfey laboratory for discussions and advice. P.N.B. is a founder and CEO of GrassRoots Biotechnology. Received: January 19, 2010 Revised: July 28, 2010 Accepted: October 13, 2010 Published: November 11, 2010 REFERENCES Aida, M., Beis, D., Heidstra, R., Willemsen, V., Blilou, I., Galinha, C., Nussaume, L., Noh, Y.S., Amasino, R., and Scheres, B. (2004). The PLETHORA genes mediate patterning of the Arabidopsis root stem cell niche. Cell 119, 109–120. Asada, K. (2006). Production and scavenging of reactive oxygen species in chloroplasts and their functions. Plant Physiol. 141, 391–396. Bashandy, T., Guilleminot, J., Vernoux, T., Caparros-Ruiz, D., Ljung, K., Meyer, Y., and Reichheld, P.J. (2010). Interplay between the NADP-linked thioredoxin and glutathione systems in Arabidopsis auxin signaling. Plant Cell 22, 376–391. Beemster, G.T., and Baskin, T.I. (1998). Analysis of cell division and elongation underlying the developmental acceleration of root growth in Arabidopsis thaliana. Plant Physiol. 116, 1515–1526. Bestwick, C.S., Brown, I.R., Bennett, M.H., and Mansfield, J.W. (1997). Localization of hydrogen peroxide accumulation during the hypersensitive reaction of lettuce cells to Pseudomonas syringae pv phaseolicola. Plant Cell 9, 209–221. Bielski, B.H.J., Shine, G.G., and Bajuk, S. (1980). Reduction of nitroblue tetrazolium by CO2- and O2- radicals. J. Phys. Chem. 84, 830–833. Brady, S.M., Orlando, D.A., Lee, J.Y., Wang, J.Y., Koch, J., Dinneny, J.R., Mace, D., Ohler, U., and Benfey, P.N. (2007a). A high-resolution root spatiotemporal map reveals dominant expression patterns. Science 318, 801–806. Brady, S.M., Song, S., Dhugga, K.S., Rafalski, J.A., and Benfey, P.N. (2007b). Combining expression and comparative evolutionary analysis. The COBRA gene family. Plant Physiol. 143, 172–187. Burch, P.M., and Heintz, N.H. (2005). Redox regulation of cell-cycle re-entry: cyclin D1 as a primary target for the mitogenic effects of reactive oxygen and nitrogen species. Antioxid. Redox Signal. 7, 741–751.
Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc. 615
Busch, W., Miotk, A., Ariel, F.D., Zhao, Z., Forner, J., Daum, G., Suzaki, T., Schuster, C., Schultheiss, S.J., Leibfried, A., et al. (2010). Transcriptional control of a plant stem cell niche. Dev. Cell 18, 849–861. Caliskan, M., and Cuming, A.C. (1998). Spatial specificity of H2O2-generating oxalate oxidase gene expression during wheat embryo germination. Plant J. 15, 165–171. Chen, G.-X., and Asada, K. (1989). Ascorbate peroxidase in tea leaves: occurrence of two isozymes and the differences in their enzymatic and molecular properties. Plant Cell Physiol. 30, 987–998. Davletova, S., Rizhsky, L., Liang, H., Shengqiang, Z., Oliver, D.J., Coutu, J., Shulaev, V., Schlauch, K., and Mittler, R. (2005a). Cytosolic ascorbate peroxidase 1 is a central component of the reactive oxygen gene network of Arabidopsis. Plant Cell 17, 268–281. Davletova, S., Schlauch, K., Coutu, J., and Mittler, R. (2005b). The zinc-finger protein Zat12 plays a central role in reactive oxygen and abiotic stress signaling in Arabidopsis. Plant Physiol. 139, 847–856. Dello Ioio, R., Linhares, F.S., Scacchi, E., Casamitjana-Martinez, E., Heidstra, R., Costantino, P., and Sabatini, S. (2007). Cytokinins determine Arabidopsis root-meristem size by controlling cell differentiation. Curr. Biol. 17, 678–682. Dello Ioio, R., Nakamura, K., Moubayidin, L., Perilli, S., Taniguchi, M., Morita, M.T., Aoyama, T., Costantino, P., and Sabatini, S. (2008). A genetic framework for the control of cell division and differentiation in the root meristem. Science 322, 1380–1384. Dunand, C., Crevecoeur, M., and Penel, C. (2007). Distribution of superoxide and hydrogen peroxide in Arabidopsis root and their influence on root development: possible interaction with peroxidases. New Phytol. 174, 332–341. Elstner, E.F., and Heupel, A. (1976). Inhibition of nitrite formation from hydroxylammoniumchloride: a simple assay for superoxide dismutase. Anal. Biochem. 70, 616–620. Federico, R., and Angelini, R. (1986). Occurrence of diamine oxidase in the apoplast of pea epicotyls. Planta 167, 300–302. Foreman, J., Demidchik, V., Bothwell, J.H., Mylona, P., Miedema, H., Torres, M.A., Linstead, P., Costa, S., Brownlee, C., Jones, J.D., et al. (2003). Reactive oxygen species produced by NADPH oxidase regulate plant cell growth. Nature 422, 442–446. Kurata, T., Ishida, T., Kawabata-Awai, C., Noguchi, M., Hattori, S., Sano, R., Nagasaka, R., Tominaga, R., Koshino-Kimura, Y., Kato, T., et al. (2005). Cellto-cell movement of the CAPRICE protein in Arabidopsis root epidermal cell differentiation. Development 132, 5387–5398. Leibfried, A., To, J.P.C., Busch, W., Stehling, S., Kehle, A., Demar, M., Kieber, J.J., and Lohmann, J.U. (2005). WUSCHEL controls meristem function by direct regulation of cytokinin-inducible response regulators. Nature 438, 1172–1175. Levesque, M.P., Vernoux, T., Busch, W., Cui, H., Wang, J.Y., Blilou, I., Hassan, H., Nakajima, K., Matsumoto, N., Lohmann, J.U., et al. (2006). Whole-genome analysis of the SHORT-ROOT developmental pathway in Arabidopsis. PLoS Biol. 4, e143.
Menon, S.G., and Goswami, P.C. (2007). A redox cycle within the cell cycle: ring in the old with the new. Oncogene 26, 1101–1109. Meyer, Y., Reichheld, P.J., and Vignols, F. (2005). Thioredoxins in Arabidopsis and other plants. Photosynth. Res. 86, 419–433. Niethammer, P., Grabher, C., Look, A.T., and Mitchison, T.J. (2009). A tissuescale gradient of hydrogen peroxide mediates rapid wound detection in zebrafish. Nature 459, 996–999. Nordberg, J., and Arner, E.S. (2001). Reactive oxygen species, antioxidants and the mammalian thioredoxin system. Free Radic. Biol. Med. 31, 1287– 1312. Owusu-Ansah, E., and Banerjee, U. (2009). Reactive oxygen species prime Drosophila haematopoietic progenitors for differentiation. Nature 461, 486–487. Owusu-Ansah, E., Yavari, A., and Banerjee, U. (2008). A protocol for in vivo detection of reactive oxygen species. Nature Protocols, http:// www.natureprotocols.com/2008/02/27/a_protocol_for_in_vivo_detecti.php. 10.1038/nprot.2008.23. Passardi, F., Tognolli, M., De Meyer, M., Penel, C., and Dunand, C. (2006). Two cell wall associated peroxidases from Arabidopsis influence root elongation. Planta 223, 965–974. Pitzschke, A., Forzani, C., and Hirt, H. (2006). Reactive oxygen species signaling in plants. Antioxid. Redox Signal. 8, 1757–1764. Reichheld, J.P., Khafif, M., Riondet, C., Droux, M., Bonnard, G., and Meyer, Y. (2007). Inactivation of thioredoxin reductases reveals a complex interplay between thioredoxin and glutathione pathways in Arabidopsis development. Plant Cell 19, 1851–1865. Ros Barcelo, A. (1997). Lignification in plant cell walls. Int. Rev. Cytol. 176, 87–132. Sarsour, E.H., Venkataraman, S., Kalen, A.L., Oberley, L.W., and Goswami, P.C. (2008). Manganese superoxide dismutase activity regulates transitions between quiescent and proliferative growth. Aging Cell 7, 405–417. Brouwer, K.S., van Valen, T., Day, D.A., and Lambers, H. (1986). Hydroxamatestimulated O2 uptake in roots of Pisum sativum and Zea mays, mediated by a peroxidase. Plant Physiol. 82, 236–240. Tognolli, M., Penel, C., Greppin, H., and Simon, P. (2002). Analysis and expression of the class III peroxidase large gene family in Arabidopsis thaliana. Gene 288, 129–138. Toledo-Ortiz, G., Huq, E., and Quail, P.H. (2003). The Arabidopsis basic/helixloop-helix transcription factor family. Plant Cell 15, 1749–1770. Ubeda-Tomas, S., Federici, F., Casimiro, I., Beemster, G.T., Bhalerao, R., Swarup, R., Doerner, P., Haseloff, J., and Bennett, M.J. (2009). Gibberellin signaling in the endodermis controls Arabidopsis root meristem size. Curr. Biol. 19, 1194–1199.
Maeda, H. (2008). Which are you watching, an individual reactive oxygen species or total oxidative stress? Ann. N. Y. Acad. Sci. 1130, 149–156.
Vernoux, T., Wilson, R.C., Seeley, K.A., Reichheld, J.P., Muroy, S., Brown, S., Maughan, S.C., Cobbett, C.S., Van Montagu, M., Inze, D., et al. (2000). The ROOT MERISTEMLESS1/CADMIUM SENSITIVE2 gene defines a glutathione-dependent pathway involved in initiation and maintenance of cell division during postembryonic root development. Plant Cell 12, 97–110.
Maeda, H., Fukuyasu, Y., Yoshida, S., Fukuda, M., Saeki, K., Matsuno, H., Yamauchi, Y., Yoshida, K., Hirata, K., and Miyamoto, K. (2004). Fluorescent probes for hydrogen peroxide based on a non-oxidative mechanism. Angew. Chem. Int. Ed. Engl. 43, 2389–2391.
Willemsen, V., Bauch, M., Bennett, T., Campilho, A., Wolkenfelt, H., Xu, J., Haseloff, J., and Scheres, B. (2008). The NAC domain transcription factors FEZ and SOMBRERO control the orientation of cell division plane in Arabidopsis root stem cells. Dev. Cell 15, 913–922.
616 Cell 143, 606–616, November 12, 2010 ª2010 Elsevier Inc.
Epiblast Stem Cell Subpopulations Represent Mouse Embryos of Distinct Pregastrulation Stages Dong Wook Han,1 Natalia Tapia,1 Jin Young Joo,1 Boris Greber,1 Marcos J. Arau´zo-Bravo,1 Christof Bernemann,1 Kinarm Ko,2,3 Guangming Wu,1 Martin Stehling,1 Jeong Tae Do,4 and Hans R. Scho¨ler1,5,* 1Department of Cell and Developmental Biology, Max Planck Institute for Molecular Biomedicine, Ro ¨ ntgenstrasse 20, 48149 Mu¨nster, Germany 2Center for Stem Cell Research, Institute of Biomedical Sciences and Technology 3Department of Neuroscience, School of Medicine Konkuk University, Hwayang-dong, Gwangjin-gu, Seoul 143-701, Republic of Korea 4Laboratory of Stem Cell and Developmental Biology, CHA Stem Cell Institute, CHA University, 605-21 Yoeksam 1-dong, Gangnam-gu, Seoul 135-081, Republic of Korea 5University of Mu ¨ nster, Medical Faculty, Domagkstrasse 3, 48149 Mu¨nster, Germany *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.10.015
SUMMARY
Embryonic stem cells (ESCs) comprise at least two populations of cells with divergent states of pluripotency. Here, we show that epiblast stem cells (EpiSCs) also comprise two distinct cell populations that can be distinguished by the expression of a specific Oct4-GFP marker. These two subpopulations, Oct4-GFP positive and negative EpiSCs, are capable of converting into each other in vitro. Oct4GFP positive and negative EpiSCs are distinct from ESCs with respect to global gene expression pattern, epigenetic profile, and Oct4 enhancer utilization. Oct4-GFP negative cells share features with cells of the late mouse epiblast and cannot form chimeras. However, Oct4-GFP positive EpiSCs, which only represent a minor EpiSC fraction, resemble cells of the early epiblast and can readily contribute to chimeras. Our findings suggest that the rare ability of EpiSCs to contribute to chimeras is due to the presence of the minor EpiSC fraction representing the early epiblast. INTRODUCTION Four different types of pluripotent stem cells have been established in culture to date: EC (embryonal carcinoma) cells from teratocarcinomas, ES (embryonic stem) cells from the inner cell mass (ICM), EG (embryonic germ) cells from PGCs (primordial germ cells), and EpiSCs (epiblast stem cells) from developing epiblasts (Lovell-Badge, 2007). ES, EC, and EG cells are capable of efficiently forming both teratomas and chimeras. However, although EpiSCs readily form teratomas, they rarely form chimeras (Brons et al., 2007; Tesar et al., 2007). This feature
suggests that EpiSCs may exhibit restricted pluripotency relative to other pluripotent stem cells despite similarities in gene expression and epigenetic profile (Chou et al., 2008). To our knowledge, EpiSCs represent only one cell type exhibiting this restricted potency. This limited potency of EpiSCs may be inherent to the tissue of origin or newly arises from the different culture environment of EpiSCs compared with ES cells (ESCs). This latter possibility is corroborated by evidence indicating that mouse ESCs require LIF and BMP4 to maintain pluripotency (Ying et al., 2003), whereas EpiSCs require bFGF and Activin A (Brons et al., 2007; Tesar et al., 2007). The low potency of EpiSCs can also be explained by the inability of these cells to grow and survive as single cells when injected into a blastocyst, preventing them from incorporating into the ICM. Another possible explanation concerns the developmental stage gap of the cells involved—donor EpiSCs are of embryonic day 5.5 (E5.5) to E7.5, whereas recipient blastocysts are of E3.5. However, these possible explanations do not fully explain the distinct pluripotency of EpiSCs. A final explanation may be that EpiSCs are actually heterogeneous in nature and thus comprise at least two distinct subpopulations, one with higher developing potency and the other with lower potency, possibly accounting for the apparent discrepancy in the potency of EpiSCs observed with different in vivo development assays. ESCs have long been considered to represent a homogeneous population of cells; however, recent studies based on Stella (Dppa3) and Rex1 (Zfp42) expression have demonstrated that ESCs comprise a heterogeneous cell population (Hayashi et al., 2008; Toyooka et al., 2008). Different cell subpopulations represent different in vivo developmental stages, with a dynamic interchange between cells of an ICM-like state and those of an epiblast- or primitive ectoderm-like state. Similarly, ESC expression levels of stage-specific embryonic antigen 1 (SSEA1) and platelet endothelial cell adhesion molecule 1 (Pecam1) have been shown to correlate with ESC pluripotency (Furusawa et al., 2004). Strikingly, intermittent Nanog expression has Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc. 617
Figure 1. Characterization EpiSCs
of
GOF18
(A) Morphology and Oct4-GFP expression in the established GOF18 EpiSC line at early and late passages. (B and C) Scatter plots of global gene expression microarrays comparing GOF18 EpiSCs with T9 EpiSCs (B) and ESCs (C). The black lines delineate the boundaries of 2-fold difference in gene expression levels. Genes highly expressed in ordinate samples compared with abscissas samples are shown as green circles; those less expressed are shown as red. Positions of pluripotent cell (Pou5f1/Oct4, Sox2, Nanog, Klf2, Klf4, Klf5, Sall4, Dnmt3l, Esrrb, Fbxo15, and Zfp42/Rex1), germ cell (Stella [Dppa3], Dppa4, Dppa5, Iftm3 [Fragilis]), and epiblast (Fgf5, T) markers are indicated with orange dots. The color bar to the right indicates the scattering density; the higher the scattering density, the darker the blue color. Gene expression levels are depicted on log2 scale. (D) Protein levels of pluripotency markers in EpiSCs. Levels of Oct4, Nanog, and Sox2 were compared in ESCs and EpiSCs by western blot; 3T3 cells were used as a negative control. (E) Immunofluorescence analysis for Oct4, Nanog, and Sox2 in ESCs and EpiSCs. DAPI was used for nuclear staining.
been associated with distinct functional ESC phenotypes (Chambers et al., 2007). Thus, undifferentiated ESC cultures comprise cells of different pluripotency levels corresponding to different developmental stages in the embryo. To date, EpiSCs have been considered to represent a homogeneous pluripotent cell population due to their high methylation compared with ESCs. This epigenetic profile may result in reduced EpiSC plasticity, thus preventing cellular heterogeneity, as observed in ESCs (Hayashi et al., 2008; Toyooka et al., 2008). Primitive ectoderm (PrE) cells of developing epiblasts have been generally described to be incapable of contributing to chimera formation, although they still retain differentiation potential and can give rise to cells of all three germ layers (Gardner et al., 1985; Lawson et al., 1991; Tesar et al., 2007). However, Gardner et al. demonstrated that PrE cells from an early stage, such as E5, still have the capability for chimeric contribution, but they lose it within the next 24 hr (Gardner et al., 1985). Therefore, if we consider the developing potency of early-stage epiblasts and the heterogeneity of ESCs, we can surmise that EpiSCs may comprise a heterogeneous population, like ESCs, and that a minor or rare subpopulation of EpiSCs exists that contributes to chimera formation, with the major EpiSC subpopulation having no such developing potency, like late-stage epiblasts. In this study, we explored the distinct pluripotential state of EpiSCs by investigating the homogeneity of EpiSCs in culture. Based on the distinct expression of an Oct4-GFP transgene, we found that EpiSCs are not homogeneous but, rather, 618 Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc.
comprise two distinct cell subpopulations: an Oct4-GFP positive population and an Oct4-GFP negative population. These two subpopulations displayed clear differences in molecular characteristics and developmental potential, including capability for chimera formation. We show that these two subpopulations do not represent spontaneously converted ES-like cells but, rather, resemble cells of early- and late-stage postimplantation embryos. Therefore, our findings suggest that the observed discrepancy in a pluripotential capacity of EpiSCs (teratoma versus chimera formation) is due to the presence of two functionally distinct subpopulations of EpiSCs in culture. RESULTS Dynamic Expression of an Oct4-GFP Transgene in Established GOF18 EpiSCs Epiblast stem cells were derived from GOF18 (genomic Oct4 fragment, 18 kb) mice containing a GFP transgene under the control of the entire regulatory region of the Oct4 gene (Yeom et al., 1996). A characteristic feature of mouse EpiSCs is the preferential use of the proximal enhancer (PE) over the distal enhancer (DE) for Oct4 gene transcription (Tesar et al., 2007). Thus, both established EpiSCs from GOF18 mice and postimplantation epiblasts should theoretically express Oct4 and be Oct4-GFP positive. We first examined the expression of Oct4GFP in both established EpiSCs (Figure 1A) and postimplantation epiblasts (Figure 5E). Whereas the in vivo epiblast was
Figure 2. Dynamic Expression of Oct4-GFP in GOF18 EpiSCs (A) Percentage of Oct4-GFP positive cells in EpiSCs measured by FACS sorting. (B) Oct4-GFP positive and Oct4-GFP negative EpiSCs were FACS sorted and cultured separately. An Oct4-GFP negative colony from Oct4GFP positive cells and an Oct4-GFP positive colony from Oct4-GFP negative cells observed under fluorescence microscope after 1 week in culture are shown. (C) Percentage of Oct4-GFP positive cells from both Oct4-GFP positive and Oct4-GFP negative cells was measured by FACS 7 days after the initial sorting. (D) Summary of the clonal assay. FACS-sorted Oct4-GFP positive and negative single cells were plated on 96-well plates, and Oct4-GFP expression was monitored under fluorescence microscope during 1 week. The numbers of clonal cell lines containing only GFP positive cells, only GFP negative cells, or a mixture of GFP positive and negative cells are indicated.
Oct4-GFP positive, only a very low percentage of established GOF18 EpiSCs retained Oct4-GFP expression after a few passages (Figure 1A and Figure 2A; see also Figure 5E). GOF18 EpiSCs exhibited a global gene expression profile very similar to that of an independently derived EpiSC line but distinct from that of ESCs, confirming that GOF18 cells are bona fide EpiSCs (Tesar et al., 2007) (Figures 1B and 1C). Oct4 protein levels were similar in ESCs and EpiSCs, but Sox2 and Nanog protein levels were lower in EpiSCs (Figures 1D and 1E), consistent with published data (Silva et al., 2009). Therefore, the expression of the Oct4-GFP transgene did not appear to correspond to that of the endogenous Oct4 gene in GOF18 EpiSCs, prompting us to speculate that this may be due to an adaptation of the in vivo epiblast to the in vitro culture conditions. Similarly, the ICM undergoes adaptive modifications during the establishment of ESCs, as previously described (Buehr and Smith, 2003; Rossant, 2001). Thus, inactivation of the Oct4-GFP transgene might actually result from epigenetic modifications occurring in the in vitro culture—a change that may affect the GFP transgene, but not the endogenous Oct4 locus at the gene and protein levels. However, 0.5% of GOF18 EpiSCs remained Oct4-GFP positive after serial passages (Figure 2A). Next, we sought to investigate the characteristics of this Oct4GFP positive population of EpiSCs. To this end, we sorted Oct4GFP positive and negative cells by fluorescence-activated cell sorting (FACS) and cultured them separately. Of interest, 48 hr after FACS sorting, Oct4-GFP positive cells started to lose GFP expression. On the other hand, some Oct4-GFP negative cells started to gain GFP expression (Figure 2B). After 1 week in culture, the percentage of Oct4-GFP positive cells in each culture was similar to that of the parental nonsorted EpiSCs. Oct4-GFP negative EpiSCs gave rise to a small proportion of Oct4-GFP positive EpiSCs (0.2% Oct4-GFP positive), whereas the majority of Oct4-GFP positive EpiSCs converted to Oct4GFP negative EpiSCs (0.5% Oct4-GFP positive) (Figure 2C).
To better characterize this conversion, we performed a clonal assay (Figure 2D). We then FACS sorted single Oct4-GFP positive and negative cells in 96-well plates (10 3 96-well plates with Oct4-GFP negative EpiSCs and 5 3 96-well plates with Oct4-GFP positive EpiSCs). Fifty-eight of 960 Oct4-GFP negative EpiSCs and 30 of 480 Oct4-GFP positive EpiSCs had survived after 1 day in culture. After 1 week in culture, five Oct4-GFP negative clonal lines showed partial reactivation of the Oct4-GFP transgene. Similarly, 12 Oct4-GFP positive single cell lines lost Oct4-GFP expression—either partially or entirely (Figure 2D). These results indicated that both Oct4-GFP positive and negative EpiSCs had the ability to interconvert, resulting in a state of dynamic equilibrium between Oct4-GFP positive and negative EpiSCs in culture. Oct4-GFP Positive and Negative EpiSCs Have Distinct Gene Expression Profiles Next, we assessed whether the Oct4-GFP positive and negative subpopulations differ at the molecular level. The global gene expression profile of both FACS-sorted subpopulations was determined by microarray analysis. The global gene expression profile of Oct4-GFP negative EpiSCs, which represented 99% of the entire GOF18 EpiSC culture, matched that of the parental nonsorted GOF18 EpiSCs (Figure 3A). Moreover, the global expression profile of the Oct4-GFP positive fraction was very similar to that of the negative fraction, but both differed markedly from that of an ESC control transcriptome (Figures 3A and 3B). In other words, both EpiSC subpopulations indeed constitute EpiSCs and not, for example, spontaneously converted ES-like cells. ESC marker genes, such as Klf2, Klf4, Dnmt3l, Esrrb, Stella (Dppa3), Fbxo15, Rex1 (Zfp42), and Dppa5, were upregulated in the Oct4-GFP positive subpopulation compared with the Oct4GFP negative fraction (Figure 3A and Figure S1 available online). Similarly, EpiSC markers, such as Fgf5 and Brachyury (T), were downregulated in GFP positive subpopulation compared with negative EpiSCs. Overall, these findings indicate that Oct4-GFP Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc. 619
Figure 3. Oct4-GFP Positive and Negative EpiSCs Have Distinct Gene Expression and Epigenetic Profiles (A) Scatter plots of global gene expression microarray comparing Oct4-GFP negative and positive EpiSCs with nonsorted EpiSCs. (B) Scatter plots of global gene expression microarrays comparing Oct4-GFP negative and positive EpiSCs with ESCs. The black lines delineate the boundaries of 2-fold difference in gene expression levels. Genes highly expressed in ordinate samples compared with abscissas samples are shown as green circles; those less expressed are shown as red. Positions of pluripotent cell (Pou5f1/Oct4, Sox2, Nanog, Klf2, Klf4, Klf5, Sall4, Dnmt3l, Esrrb, Fbxo15, and Zfp42/Rex1), germ cell (Stella [Dppa3], Dppa4, Dppa5, Iftm3 [Fragilis]), and epiblast (Fgf5, T) markers are indicated with orange dots. The color bar to the right indicates the scattering density; the higher the scattering density, the darker the blue color. Gene expression levels are depicted on log2 scale. (C) Heat map of global gene expression patterns in ESCs, Oct4-GFP positive and negative EpiSCs, T9 EpiSCs, and GOF18 EpiSCs. The color bar at top codifies the gene expression in log2 scale. Red and blue colors indicate high and low gene expression, respectively. (D) Hierarchical clustering shows that Oct4-GFP positive EpiSCs clustered close to the parental EpiSCs or T9 EpiSCs, but not to ESCs. (E) DNA methylation status of the promoter regions of both the endogenous Oct4 gene and the Oct4-GFP transgene was determined by bisulfite sequencing PCR. (F) Dppa5, Stella (Dppa3), Vasa, and Fragilis (Iftm3) promoter regions were analyzed by bisulfite sequencing PCR in ESCs, MEFs, and PGCs, as well as in Oct4GFP positive and negative EpiSCs. See also Figure S1.
positive cells were still EpiSCs by identity, but they differed from the more abundant EpiSC subpopulation by the expression of several prominent marker genes. 620 Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc.
Recent reports have described that EpiSCs are capable of converting to an ES-like state under specific culture conditions (Bao et al., 2009; Greber et al., 2010; Guo et al., 2009; Hanna
Figure 4. Oct4-GFP Positive and Negative EpiSCs Show Distinct ESC Conversion Efficiency (A) Conversion of both the Oct4-GFP positive and negative EpiSCs to an ES-like cell state using a conversion condition (2i+LIF). Morphology and alkaline phosphatase (AP) activity of EpiSCs after 4 days of treatment are shown. (B) Relative conversion efficiency of Oct4-GFP positive and negative EpiSCs was determined by counting the number of AP-positive colonies. Data represent mean ± SEM of triplicates; n = 3. (C) Dppa5, Stella, Vasa, and Fragilis promoter regions in the converted ES-like cells were analyzed by bisulfite sequencing PCR. See also Figure S2.
et al., 2009). We therefore sought to further rule out the possibility that Oct4-GFP positive EpiSCs represented EpiSCs that had converted to an ES-like state. As revealed by hierarchical clustering, Oct4-GFP positive as well as Oct4-GFP negative EpiSCs clearly differed from ESCs and clustered with an independent EpiSC line (Figures 3C and 3D). Thus, the Oct4-GFP positive EpiSC subpopulation did not represent EpiSC cells that had spontaneously converted to an ES-like state. Epigenetic Differences between Oct4-GFP Positive and Negative EpiSCs We then asked whether at least some of the differences in gene expression between Oct4-GFP positive and negative EpiSCs (Figure 3A) could have resulted from differential epigenetic modifications. To this end, we determined the DNA methylation status of the endogenous and transgenic Oct4 loci in both Oct4-GFP positive and negative EpiSCs. Consistent with the gene expression profiling data (Figure 3A), the promoter regions of endogenous Oct4 were completely unmethylated in both subpopulations (Figure 3E). However, the Oct4-GFP transgene promoter was hypomethylated in Oct4-GFP positive EpiSCs but highly methylated in Oct4-GFP negative EpiSCs (Figure 3E). These data indicate that the mechanism regulating the dynamic Oct4GFP fluctuations in EpiSC culture is related to the differential DNA methylation status of the Oct4-GFP promoter in Oct4GFP positive and negative EpiSCs. Next, we assessed whether Oct4-GFP positive cells may represent early PGCs. To this end, we examined the promoter regions of the germ cell markers Stella (Dppa3), Vasa, and Fragilis (Iftm3) in ESCs, Oct4-GFP positive as well as negative EpiSCs, PGCs, and mouse embryonic fibroblasts (MEFs). As expected, all promoter regions examined were fully methylated in MEFs
but completely unmethylated in PGCs. All of these genes exhibited an imprinted methylation pattern in ESCs, which is due to the heterogeneous expression of many germ cell markers (Carter et al., 2008). In contrast, germ cell marker genes were fully methylated in Oct4-GFP negative as well as the positive EpiSCs, like in MEFs, but not in ESCs or PGCs (Figure 3F). Finally, parental nonsorted EpiSCs exhibited the same methylation pattern as Oct4-GFP negative EpiSCs (data not shown). These data further support the notion that Oct4-GFP positive cells have an overall EpiSC identity that is neither ESlike nor PGC-like. In contrast, the methylation pattern of the Dppa5 promoter was similar in Oct4-GFP positive EpiSCs and ESCs (Figure 3F). This is consistent with the gene expression profiling data, revealing increased Dppa5 expression in Oct4-GFP positive EpiSCs compared with Oct4-GFP negative EpiSCs (Figure 3A). These data suggest that Oct4-GFP positive cells differ from Oct4-GFP negative cells in the expression of some prominent ESC marker genes, as reflected by differential epigenetic regulation. These data further suggest that the dynamic interchange between the Oct4-GFP positive and negative subpopulations of EpiSCs appears to be regulated by a DNA methylation-dependent mechanism. Oct4-GFP Positive and Negative EpiSCs Show Distinct ESC Conversion Rates Next, we asked whether these two EpiSC populations differ in function. EpiSCs can be converted to an ES-like state by switching to stringent ESC culture conditions, the so-called 2i+LIF medium (2i condition) (Greber et al., 2010; Guo et al., 2009; Hanna et al., 2009). As Oct4-GFP positive cells exhibited upregulation of some prominent ESC markers, we speculated that Oct4-GFP positive EpiSCs can be converted more easily than Oct4-GFP negative EpiSCs. Under the 2i condition, the conversion efficiency of Oct4-GFP positive EpiSCs was much higher compared with that of Oct4-GFP negative EpiSCs, as determined by the number of colonies staining positive for alkaline phosphatase (Figures 4A and 4B). Remarkably, conversion of Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc. 621
Oct4-GFP positive EpiSCs was noted only in the presence of 2i+LIF, and not in EpiSC medium. Consistent with a previous study (Bao et al., 2009), we also obtained converted ES-like cells in typical ESC medium (LIF alone). However, we were unable to maintain these converted cells with LIF alone (data not shown). Thus, although Oct4-GFP positive cells do not represent ESlike cells, they can be converted at substantially higher efficiency than Oct4-GFP negative EpiSCs. Finally, to definitively rule out the possibility that Oct4-GFP positive EpiSCs are actually spontaneously converted ES-like cells, we assessed the promoter methylation patterns of Stella, Fragilis, Vasa, and Dppa5 in the ES-like cells that had been converted from Oct4-GFP positive EpiSCs. The Stella promoter is known to be differentially methylated in EpiSCs and ESCs (Bao et al., 2009; Hayashi et al., 2008). As shown in Figure 4C, the converted ES-like cells exhibited an imprinted DNA methylation pattern of the Stella promoter, like ESCs, but a clearly different pattern compared with Oct4-GFP positive EpiSCs, from which they had been derived (Figure 3F). Similarly, the converted ESlike cells exhibited hypomethylation of the Rex1 promoter region, but both Oct4-GFP positive and negative EpiSCs showed high Rex1 promoter methylation (Figure S2). Moreover, the converted ES-like cells also showed reduced DNA promoter methylation of the germ cell markers Fragilis, Vasa, and Dppa5, like ESCs, whereas Oct4-GFP positive EpiSCs exhibited hypermethylation of these germ cell marker genes (Figure 4C and Figure 3F). Therefore, the distinct DNA methylation pattern in Oct4-GFP positive EpiSCs and converted ES-like cells allows us to definitively rule out that the Oct4-GFP positive EpiSC subpopulation represents EpiSCs that had spontaneously converted to an ES-like state in EpiSC culture. Oct4-GFP Positive and Negative EpiSCs Represent Pregastrulation Epiblasts of Different Developmental Stages Oct4-GFP positive and Oct4-GFP negative EpiSCs exhibit differential expression of some ESC marker genes (Figure 3A), distinguishable Dppa5 methylation pattern (Figure 3F), and distinct ESC conversion efficiency (Figures 4A and 4B). Of interest, Dppa5, which was less methylated in Oct4-GFP positive than in Oct4-GFP negative EpiSCs (Figure 3F), is expressed in E5.5 epiblasts, but not in E6.5 epiblasts (Western et al., 2005). On the other hand, the typical EpiSC marker T, which was highly expressed in Oct4-GFP negative compared with Oct4-GFP positive EpiSCs, is more abundant in E6.5 epiblasts than in E5.5 epiblasts (Rivera-Pe´rez and Magnuson, 2005). These observations prompted us to investigate whether the cells with different Oct4-GFP transgene expression actually correspond to in vivo epiblasts of different developmental stages. For this purpose, we compared the gene expression profiles of E5.5 and E6.5 in vivo epiblasts isolated from GOF18 mice. As epiblasts from such an early stage contain a very limited number of cells, the procedure of whole-genome profiling currently presents a number of challenges. Also, as Oct4-GFP positive EpiSCs are very similar to the parental EpiSCs based on microarray gene analysis (Figure 3C), we sought to compare the specific genes differentially expressed in Oct4-GFP positive and negative EpiSCs. To this end, Oct4-GFP positive cells from both E5.5 622 Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc.
and E6.5 epiblasts were sorted and analyzed by real-time PCR. Genes that were differentially expressed in Oct4-GFP positive and negative EpiSCs (Figure 3A), such as Rex1 (Zfp42), Esrrb, Fbxo15, Klf2, Klf4, Klf5, Stella, Dppa4, Dppa5, Dnmt3l, Piwil2, Fragilis, and T, were selected for analysis. Consistent with the microarray data, the majority of ESC markers were highly expressed in both Oct4-GFP positive EpiSCs and E5.5 epiblasts compared with Oct4-GFP negative EpiSCs and E6.5 epiblasts, respectively (Figures 5A and 5B). On the other hand, Fragilis and T were highly expressed in Oct4-GFP negative EpiSCs and E6.5 epiblasts compared with Oct4-GFP positive EpiSCs and E5.5 epiblasts, respectively (Figures 5A and 5B). Thus, the expression patterns of Oct4-GFP positive and negative EpiSCs correlate to those of E5.5 and E6.5 epiblasts, respectively. Next, we used computational analysis to determine the association between the EpiSC subpopulations and the in vivo epiblasts of different stages. The heat map calculated with the real-time RQ values again shows a very similar association (Figure 5C). Finally, the hierarchical clustering generated with the real-time PCR data clearly shows that Oct4-GFP positive EpiSCs are similar to E5.5 epiblasts, whereas Oct4-GFP negative EpiSCs rather represent E6.5 epiblasts (Figure S3). Oct4 transcription is regulated through the DE in the ICM and ESCs but through the PE in the epiblast and EpiSCs (Tesar et al., 2007; Yeom et al., 1996). Using a luciferase assay, we compared Oct4 enhancer utilization/activity in both Oct4-GFP positive and negative EpiSCs. As a control, we used ES-like cells that had been converted from Oct4-GFP positive EpiSCs. Oct4-GFP negative EpiSCs specifically used the PE, and the converted ES-like cells showed strong DE activity, as expected (Figure 5D). Although converted ES-like cells are maintained in the presence of 2i+LIF medium, spontaneous differentiation of the converted cells into EpiSCs could explain the residual PE activity observed. Accordingly, ESCs, which are derived directly from the ICM, showed exactly the same enhancer utilization as converted ES-like cells (data not shown). Of interest, Oct4-GFP positive EpiSCs preferentially utilize the PE, but they also exhibit some DE activity (Figure 5D). This unexpected DE activity in Oct4-GFP positive EpiSCs, although lower than PE activity, encouraged us to analyze the enhancer activity in early-stage in vivo epiblasts. We isolated in vivo epiblasts (E5.5, E6.6, and E7.5) from both GOF18 (containing DE and PE) and GOF18DPE mice. As mentioned above, in vivo epiblasts and their in vitro derivatives, EpiSCs, have been described to preferentially use the PE for Oct4 transcription (Guo et al., 2009; Hanna et al., 2009; Tesar et al., 2007; Yeom et al., 1996). The in vivo epiblasts isolated from GOF18 mice on E5.5, E6.6, and E7.5 were GFP positive, as expected (Figure 5E). However, the in vivo GOF18DPE epiblasts (E5.5 and E6.5), which were supposed to be GFP negative due to the absence of PE, were also GFP positive, but GFP expression was clearly weaker than that in GOF18 mice—i.e., the GFP signals were barely detectable without overexposure, consistent with a previous report (Yoshimizu et al., 1999). GFP intensity had gradually decreased in E6.5, with GFP negative expression evident at E7.5, except for PGCs (Figure 5E). This observation strongly suggests that the early-stage epiblast still retains residual DE activity, together with PE activity. Thus, the enhancer switch, an event widely
Figure 5. Oct4-GFP Positive and Negative EpiSCs Represent In Vivo Epiblasts of Different Developmental Stages (A and B) Comparison of gene expression patterns in in vivo samples (E5.5 versus E6.5 epiblasts) (A) and in vitro samples (Oct4-GFP positive versus negative EpiSCs) (B). Expression levels are normalized to those of E6.5 epiblast (A) and Oct4-GFP negative EpiSCs (B), respectively. (C) Heat map of the RT-PCR values. RT-PCR RQ ratios were used to compare in vivo with in vitro samples. The in vivo ratios were calculated by dividing the RT-PCR RQ value of the E6.5 epiblast by that of the E5.5 epiblast sample. The in vitro ratios were calculated by dividing the RT-PCR RQ value of the GFP negative EpiSC sample by that of the GFP positive EpiSC sample. The color bar at top codifies the ratio values; red and blue colors indicate high and low ratios, respectively. (D) Evaluation of the Oct4 enhancer activity in Oct4-GFP positive and negative EpiSCs. Relative luciferase activity was normalized to the activity of the empty vector. Data represent mean ± SEM of triplicates; n = 3. (E) Dynamics of the Oct4-GFP reporter expression in in vivo epiblasts. In vivo epiblasts were isolated from both GOF18 and GOF18DPE mice at E5.5, 6.5, and 7.5. Oct4-GFP expression was measured under fluorescence microscope. To show specific Oct4-GFP expression from GOF18DPE mice, all images were taken with the same overexposure, but the images from GOF18 mice were taken with normal exposure. The extra panel under the E7.5 GOF18DPE epiblast shows residual Oct4-GFP expression in PGCs. See also Figure S3.
accepted to occur around the time of implantation (E4.5), appears to occur later than expected. Therefore, the DE activity observed in Oct4-GFP positive EpiSCs also correlates with Oct4 enhancer utilization in early-stage epiblasts (Figures 5D and 5E). The enhancer utilization assay, combined with the real-time PCR data, suggests that Oct4-GFP positive EpiSCs correspond to an epiblast of an early developmental stage, whereas Oct4GFP negative EpiSCs are rather similar to an epiblast of a later stage. Oct4-GFP Positive EpiSCs Can Contribute to Chimeras Gardner et al. reported that the PrE from E5 embryos is capable of contributing to chimera formation by blastocyst injection, unlike that from E6 and E7 embryos (Gardner et al., 1985). To determine whether there is a functional difference between Oct4-GFP positive and negative EpiSCs, just like between early-
and late-stage in vivo epiblasts, we assessed whether these two cell populations can contribute to chimeras. Consistent with previous studies (Brons et al., 2007; Guo et al., 2009; Tesar et al., 2007), upon injection of Oct4-GFP positive EpiSCs into blastocysts, only 10% of the embryos showed ICM integration (Figure 6A). Surprisingly, following transfer of these embryos, germline contribution as well as coat-color chimerism was noted (Figures 6B and 6C and Table S1). However, we were unable to observe any evidence of germline transmission with Oct4-GFP positive EpiSCs (Table S1). To trace the fate of Oct4-GFP negative EpiSCs in reconstructed embryos, we first infected GOF18 EpiSCs with a Td-tomato lentivirus. We then established a new Td-tomato GOF18 EpiSC line (Epi-Red) (Figure 6D). Tomatopositive/Oct4-GFP negative cells were FACS sorted and injected into blastocysts, but neither ICM integration (Figure 6E and Figure S4 available online) nor chimera contribution was observed (Table S1). Thus, Oct4-GFP negative EpiSCs, which made up Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc. 623
Oct4-GFP negative EpiSCs are rather similar to those of a latestage epiblast (E6.5 or E7.5).
Figure 6. Oct4-GFP Positive EpiSCs Can Contribute to Chimera Formation (A) Integration of Oct4-GFP positive EpiSCs into the ICM following blastocyst injection. (B) Germline contribution of Oct4-GFP positive EpiSCs, as shown by the expression of Oct4-GFP in the gonad of a 13.5 days postcoitum (dpc) embryo. (C) Coat-color chimera from Oct4-GFP positive EpiSCs. (D) Morphology of Epi-Red cells and expression of Td-tomato lentivirus. (E) Epi-Red cells failed to integrate into the ICM following blastocyst injection. See also Figure S4.
99% of the EpiSCs in culture, had no developmental potential, unlike Oct4-GFP positive EpiSCs. Taken together, our data suggest that Oct4-GFP positive EpiSCs represent cells of an early-stage epiblast (E5.5), whereas
Dppa5 Overexpression Increases the Early-Stage Epiblast Fraction We have demonstrated that EpiSCs are heterogeneous and comprise two subpopulations—Oct4-GFP positive and negative cells, representing early- and late-stage epiblasts, respectively. Thus, we then tested whether overexpressing Dppa5, a gene that is specifically expressed in early-stage epiblasts and Oct4-GFP positive EpiSCs, but not in late-stage epiblasts and Oct4-GFP negative EpiSCs, can increase the Oct4-GFP positive fraction. Dppa5-overexpressing EpiSCs showed an 6-fold increase in Oct4-GFP positive cells compared with noninfected EpiSCs (Figure 7A). We then FACS sorted the GFP positive cells from Dppa5-overexpressing and control EpiSCs and cultured them for another 10 passages (Figure 7B). Oct4-GFP positive cells from both Dppa5-overexpressing and control EpiSCs immediately started to lose Oct4-GFP expression. After two passages, 10.5% of the Dppa5-overexpressing cells and 6.3% of the control EpiSCs remained GFP positive. The number of GFP positive cells from control EpiSCs continued to decrease until levels reached those of the parental nonsorted EpiSCs— 1.3% at passage 4. However, GFP positive cells were stably maintained in Dppa5-overexpressing EpiSCs at 4%–6%, even after several passages. To exclude the possibility that Dppa5 overexpression could revert EpiSCs to an ES-like state,
Figure 7. Dppa5 Overexpression Increases the Early-Stage Epiblast Fraction (A) The Oct4-GFP positive fraction was increased by overexpressing Dppa5. The numbers of Oct4GFP positive cells from both control and Dppa5overexpressing EpiSCs were compared by FACS. (B) Dynamics of Oct4-GFP expression in FACSsorted GFP positive cells from both control and Dppa5-overexpressing EpiSCs. FACS-sorted Oct4-GFP positive cells were cultured for another 10 passages, and the number of Oct4-GFP positive cells was measured after every two passages. (C) A dynamic interchange exists between ESCs positive and negative for Stella (Dppa3), Nanog, and Rex1 (Zfp42). When all three genes are expressed, ESCs represent the well-known ground state of pluripotency. However, when the expression of one of these genes is reduced, the pluripotent capability is also considerably reduced. EpiSCs are also heterogeneous, with two different populations distinguished by Oct4-GFP transgene expression. Oct4-GFP positive and negative EpiSCs exist in a state of dynamic equilibrium and correspond to in vivo epiblasts of early and late stages of development, respectively. The endogenous Oct4 locus is unmethylated and expressed in both populations, whereas the Oct4-GFP transgene is differentially methylated, allowing to distinguish the early- and late-stage epiblast subpopulations. See also Figure S2.
624 Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc.
Oct4-GFP cells were sorted from Dppa5-overexpressing EpiSCs, and the methylation status of the Rex1 promoter was assessed by bisulfite sequencing PCR. Rex1 promoter was highly methylated, like in Oct4-GFP positive and negative EpiSCs, indicating that the increased Oct4-GFP positive fraction still maintains EpiSC identity rather than acquires an ES-like state (Figure S2). Therefore, these data suggest that overexpression of Dppa5 increases the Oct4-GFP positive EpiSC fraction, which represents early-stage epiblasts. DISCUSSION The pluripotency of a cell can be assessed by determining the cell’s developmental potential both in vitro and in vivo. The in vivo differentiation of cells to form both teratomas and chimeras is a basic yet reliable tool for assessing a cell’s developing potency. Thus, cells that are pluripotent must meet these two basic criteria. EpiSCs, however, exhibit limited developing potency and thus have a unique pluripotential capacity compared with other pluripotent stem cell lines from different tissues of origin. The discrepancy in the potency of EpiSCs observed with different developmental assays, together with the heterogeneity of ESCs, prompted us to closely investigate the pluripotential state of EpiSCs. To this end, we assessed whether EpiSCs also exhibit heterogeneity, like ESCs. Our results show that EpiSCs are not as homogeneous as previously described (Hayashi et al., 2008; Toyooka et al., 2008). We have taken advantage of the differential expression of an Oct4-GFP transgene and have defined a novel type of EpiSCs. We found a dynamic interchange of two distinct cell subpopulations of EpiSCs that are in equilibrium in vitro: Oct4-GFP positive and Oct4-GFP negative EpiSCs (Figure 7C). We speculate that the distinct expression levels of the endogenous Oct4 and the Oct4-GFP reporter result from an adaptation of the embryonal epiblast to culture conditions during the establishment of the EpiSCs, consistent with adaptive modifications described for ESCs (Buehr and Smith, 2003; Rossant, 2001). Nevertheless, using a luciferase assay, we showed that the two subpopulations exhibit distinct Oct4 enhancer utilization. Therefore, although we discovered the existence of two types of EpiSCs—Oct4-GFP positive and negative EpiSC subpopulations—based on distinct transgene expression, this expression indeed represents distinct enhancer activity of the endogenous Oct4 locus (Figures 5D and 5E). Bao et al. recently showed the spontaneous conversion of EpiSCs to ES-like cells in ESC medium in the absence of exogenous transcription factors or small molecules (Bao et al., 2009). To investigate whether Oct4-GFP positive EpiSCs represent converted ES-like cells, we compared the distinctive features of EpiSCs and ESCs. First, the global gene expression profile of Oct4-GFP positive EpiSCs correlated with that of EpiSCs, but not with ESCs (Figures 3C and 3D). Second, as the promoter regions of Stella and Rex1 were shown to be hypomethylated in ESCs but hypermethylated in EpiSCs (Bao et al., 2009; Hayashi et al., 2008), we examined the methylation of these two genes as a hallmark for discriminating EpiSCs from ESCs or converted ES-like cells. Consistent with these previous studies, both the Stella and Rex1 promoters, as well as other germ cell marker
gene promoters, were hypomethylated in ESCs but highly methylated in Oct4-GFP positive EpiSCs. In contrast, ES-like cells that had been converted from Oct4-GFP positive EpiSCs showed exactly the same DNA methylation pattern as ESCs. Third, we investigated the signaling dependence of these cells by comparing their Oct4 enhancer utilization. We found that Oct4-GFP negative EpiSCs specifically use the PE, but converted ES-like cells and ESCs preferentially use the DE rather than the PE, as expected. Surprisingly, although Oct4-GFP positive EpiSCs mainly use the PE, a lower DE activity could also be detected, indicating that Oct4-GFP positive EpiSCs have a completely different pattern of Oct4 enhancer utilization than ESCs. Taken together, these results show that Oct4-GFP positive EpiSCs are not converted ES-like cells. We have also demonstrated that Oct4-GFP positive and negative EpiSCs correspond to developing epiblasts of two different stages. In terms of gene expression profile, Oct4-GFP positive EpiSCs correlate with E5.5 epiblasts and Oct4-GFP negative EpiSCs with E6.5 epiblasts. Moreover, Oct4 enhancer utilization of Oct4-GFP positive EpiSCs resembles that of early epiblasts (E5.5). Therefore, Oct4-GFP positive EpiSCs represent early-stage in vivo epiblast cells, whereas Oct4-GFP negative EpiSCs, the majority of EpiSCs, represent late-stage in vivo epiblast cells. Finally, we compared the developmental potential of Oct4GFP positive and negative EpiSCs. Although germline transmission could not be observed, Oct4-GFP positive EpiSCs could incorporate into the ICM and even contribute to the germline, unlike Oct4-GFP negative EpiSCs. PrE cells from E5 embryos retain the ability to give rise to adult chimeras, whereas PrE cells from E6 and E7 embryos have lost this ability (Gardner et al., 1985). Therefore, in terms of gene expression profile, epigenetic status, specific Oct4 enhancer utilization, and even functional developmental capacity, Oct4-GFP positive and negative EpiSCs represent in vivo epiblasts of early and late developmental stages, respectively. EpiSCs rarely exhibit chimeric contribution, with the chimera rate (0.5%) correlating with the percentage of Oct4-GFP positive cells in GOF18 EpiSCs (Brons et al., 2007). Thus, it is likely that the cells that contribute to chimera formation are the EpiSCs of an early in vivo epiblast— i.e., Oct4-GFP positive EpiSCs. Taken together, our findings demonstrate that EpiSC cultures are heterogeneous and comprise Oct4-GFP positive and negative EpiSC cell subpopulations that are in a state of dynamic equilibrium and that correspond to in vivo epiblasts of early and late developmental stages, respectively (Figure 7C). Moreover, our data also suggest that the introduction of an earlystage-specific transcription factor, such as Dppa5, could increase the EpiSC subpopulation that represents early-stage epiblasts (Figures 7A and 7B). Therefore, investigations with EpiSCs derived from GOF18 mice offer opportunities to better understand early embryonic development, differentiation processes, and germ cell development, as well as the developing potency of in vivo epiblasts of different developmental stages. As EpiSCs have many properties in common with human ESCs, it would be interesting to assess whether human ESCs also exhibit heterogeneity, similar to EpiSCs, which may affect the therapeutic potential of these cells. Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc. 625
EXPERIMENTAL PROCEDURES Cell Culture The derivation and characterization of GOF18 EpiSCs is described elsewhere (Greber et al., 2010). In brief, E5.5 embryos (129/Sv female 3 C56/Bl6 and DBA/2 background GOF18+/+ male) were collected and transferred into HBSS medium. For dissection, deciduas were removed with forceps, and the extraembryonic ectoderm was separated from the epiblast by using hand-pulled glass pipettes. After washing with PBS, the epiblast was cultured on MEFs in EpiSC medium: DMEM/F12 (GIBCO BRL) containing 20% knockout serum replacement (GIBCO BRL), 2 mM glutamine, 13 nonessential amino acids, and 5 ng/ml bFGF. After initial culture on MEFs for three to five passages, EpiSC colonies were picked and transferred onto dishes that had been precoated with FCS for 30 min. Feeder-free EpiSCs were cultured in MEF (CF1 mice) conditioned medium. For conditioning, irradiated MEFs were seeded at a density of 5 3 104 cells/cm2 and incubated in EpiSC medium for 24 hr. The conditioned medium was filtered, and bFGF (5 ng/ml) was added. For passaging feeder-free EpiSCs, colonies were incubated with collagenase IV (Invitrogen) for 5 min at 37 C and triturated by using a cell scraper. Cell clumps were replated on FCS-coated dishes, and the medium was changed every 24 hr. DNA Methylation Analysis DNA methylation status of EpiSCs was determined by the bisulfite sequencing method from our published protocol (Han et al., 2009). Detailed protocols, including PCR conditions and primer sequences, are described in our previous study (Han et al., 2008) and in the Extended Experimental Procedures. Whole-Genome Expression Analysis Global gene expression profiling was performed with the Illumina microarray. Detailed protocols are described in the Extended Experimental Procedures. Blastocyst Injection Blastocysts were collected from B6C3F1 mice. EpiSCs were recovered by treatment with collagenase IV or 0.25% trypsin EDTA, and 10–15 cells were loaded into an injection pipette and injected into B6C3F1 blastocysts by Piezo (Prime-tech). Each of the 15 injected blastocysts was transferred into the uterus of a pseudopregnant ICR mouse.
Brons, I.G., Smithers, L.E., Trotter, M.W., Rugg-Gunn, P., Sun, B., Chuva de Sousa Lopes, S.M., Howlett, S.K., Clarkson, A., Ahrlund-Richter, L., Pedersen, R.A., and Vallier, L. (2007). Derivation of pluripotent epiblast stem cells from mammalian embryos. Nature 448, 191–195. Buehr, M., and Smith, A. (2003). Genesis of embryonic stem cells. Philos. Trans. R. Soc. Lond. B Biol. Sci. 358, 1397–1402, discussion 1402. Carter, M.G., Stagg, C.A., Falco, G., Yoshikawa, T., Bassey, U.C., Aiba, K., Sharova, L.V., Shaik, N., and Ko, M.S. (2008). An in situ hybridization-based screen for heterogeneously expressed genes in mouse ES cells. Gene Expr. Patterns 8, 181–198. Chambers, I., Silva, J., Colby, D., Nichols, J., Nijmeijer, B., Robertson, M., Vrana, J., Jones, K., Grotewold, L., and Smith, A. (2007). Nanog safeguards pluripotency and mediates germline development. Nature 450, 1230–1234. Chou, Y.F., Chen, H.H., Eijpe, M., Yabuuchi, A., Chenoweth, J.G., Tesar, P., Lu, J., McKay, R.D., and Geijsen, N. (2008). The growth factor environment defines distinct pluripotent ground states in novel blastocyst-derived stem cells. Cell 135, 449–461. Furusawa, T., Ohkoshi, K., Honda, C., Takahashi, S., and Tokunaga, T. (2004). Embryonic stem cells expressing both platelet endothelial cell adhesion molecule-1 and stage-specific embryonic antigen-1 differentiate predominantly into epiblast cells in a chimeric embryo. Biol. Reprod. 70, 1452–1457. Gardner, R.L., Lyon, M.F., Evans, E.P., and Burtenshaw, M.D. (1985). Clonal analysis of X-chromosome inactivation and the origin of the germ line in the mouse embryo. J. Embryol. Exp. Morphol. 88, 349–363. Greber, B., Wu, G., Bernemann, C., Joo, J.Y., Han, D.W., Ko, K., Tapia, N., Sabour, D., Sterneckert, J., Tesar, P., and Scho¨ler, H.R. (2010). Conserved and divergent roles of FGF signaling in mouse epiblast stem cells and human embryonic stem cells. Cell Stem Cell 6, 215–226. Guo, G., Yang, J., Nichols, J., Hall, J.S., Eyres, I., Mansfield, W., and Smith, A. (2009). Klf4 reverts developmentally programmed restriction of ground state pluripotency. Development 136, 1063–1069. Han, D.W., Do, J.T., Gentile, L., Stehling, M., Lee, H.T., and Scho¨ler, H.R. (2008). Pluripotential reprogramming of the somatic genome in hybrid cells occurs with the first cell cycle. Stem Cells 26, 445–454. Han, D.W., Do, J.T., Arau´zo-Bravo, M.J., Lee, S.H., Meissner, A., Lee, H.T., Jaenisch, R., and Scho¨ler, H.R. (2009). Epigenetic hierarchy governing Nestin expression. Stem Cells 27, 1088–1097.
SUPPLEMENTAL INFORMATION
Hanna, J., Markoulaki, S., Mitalipova, M., Cheng, A.W., Cassady, J.P., Staerk, J., Carey, B.W., Lengner, C.J., Foreman, R., Love, J., et al. (2009). Metastable pluripotent states in NOD-mouse-derived ESCs. Cell Stem Cell 4, 513–524.
Supplemental Information includes Extended Experimental Procedures, four figures, and one table and can be found with this article online at doi: 10.1016/j.cell.2010.10.015.
Hayashi, K., Lopes, S.M., Tang, F., and Surani, M.A. (2008). Dynamic equilibrium and heterogeneity of mouse pluripotent stem cells with distinct functional and epigenetic states. Cell Stem Cell 3, 391–401.
ACKNOWLEDGMENTS
Lawson, K.A., Meneses, J.J., and Pedersen, R.A. (1991). Clonal analysis of epiblast fate during germ layer formation in the mouse embryo. Development 113, 891–911.
We are indebted to all members of the Scho¨ler laboratory for fruitful discussions on the results. We are especially grateful to Dr. Paul Tesar for providing the T9 EpiSC line. We are also grateful to Go¨ran Key, Inge Sobek-Klocke, Martina Sinn, and David Obridge for technical assistance and Areti Malapetsas for editing the manuscript. This work was supported by the Federal Ministry of Education and Research (BMBF) initiative ‘‘Cell-Based Regenerative Medicine’’ (Grant 01GN0539).
Lovell-Badge, R. (2007). Many ways to pluripotency. Nat. Biotechnol. 25, 1114–1116.
Received: April 9, 2010 Revised: August 17, 2010 Accepted: October 12, 2010 Published online: November 4, 2010
Silva, J., Nichols, J., Theunissen, T.W., Guo, G., van Oosten, A.L., Barrandon, O., Wray, J., Yamanaka, S., Chambers, I., and Smith, A. (2009). Nanog is the gateway to the pluripotent ground state. Cell 138, 722–737.
REFERENCES Bao, S., Tang, F., Li, X., Hayashi, K., Gillich, A., Lao, K., and Surani, M.A. (2009). Epigenetic reversion of post-implantation epiblast to pluripotent embryonic stem cells. Nature 461, 1292–1295.
626 Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc.
Rivera-Pe´rez, J.A., and Magnuson, T. (2005). Primitive streak formation in mice is preceded by localized activation of Brachyury and Wnt3. Dev. Biol. 288, 363–371. Rossant, J. (2001). Stem cells from the Mammalian blastocyst. Stem Cells 19, 477–482.
Tesar, P.J., Chenoweth, J.G., Brook, F.A., Davies, T.J., Evans, E.P., Mack, D.L., Gardner, R.L., and McKay, R.D. (2007). New cell lines from mouse epiblast share defining features with human embryonic stem cells. Nature 448, 196–199. Toyooka, Y., Shimosato, D., Murakami, K., Takahashi, K., and Niwa, H. (2008). Identification and characterization of subpopulations in undifferentiated ES cell culture. Development 135, 909–918.
Western, P., Maldonado-Saldivia, J., van den Bergen, J., Hajkova, P., Saitou, M., Barton, S., and Surani, M.A. (2005). Analysis of Esg1 expression in pluripotent cells and the germline reveals similarities with Oct4 and Sox2 and differences between human pluripotent cell lines. Stem Cells 23, 1436–1442. Yeom, Y.I., Fuhrmann, G., Ovitt, C.E., Brehm, A., Ohbo, K., Gross, M., Hu¨bner, K., and Scho¨ler, H.R. (1996). Germline regulatory element of Oct-4 specific for the totipotent cycle of embryonal cells. Development 122, 881–894.
Ying, Q.L., Nichols, J., Chambers, I., and Smith, A. (2003). BMP induction of Id proteins suppresses differentiation and sustains embryonic stem cell selfrenewal in collaboration with STAT3. Cell 115, 281–292. Yoshimizu, T., Sugiyama, N., De Felice, M., Yeom, Y.I., Ohbo, K., Masuko, K., Obinata, M., Abe, K., Scho¨ler, H.R., and Matsui, Y. (1999). Germline-specific expression of the Oct-4/green fluorescent protein (GFP) transgene in mice. Dev. Growth Differ. 41, 675–684.
Cell 143, 617–627, November 12, 2010 ª2010 Elsevier Inc. 627
A Genome-wide Drosophila Screen for Heat Nociception Identifies a2d3 as an Evolutionarily Conserved Pain Gene G. Gregory Neely,1,2,18 Andreas Hess,3,18 Michael Costigan,4 Alex C. Keene,5 Spyros Goulas,1 Michiel Langeslag,6 Robert S. Griffin,7 Inna Belfer,8 Feng Dai,8 Shad B. Smith,9 Luda Diatchenko,9 Vaijayanti Gupta,10 Cui-ping Xia,1 Sabina Amann,1 Silke Kreitz,3 Cornelia Heindl-Erdmann,3 Susanne Wolz,3 Cindy V. Ly,11 Suchir Arora,10 Rinku Sarangi,10 Debasis Dan,10 Maria Novatchkova,1 Mark Rosenzweig,12 Dustin G. Gibson,9 Darwin Truong,1 Daniel Schramek,1 Tamara Zoranovic,1 Shane J.F. Cronin,1 Belinda Angjeli,1 Kay Brune,3 Georg Dietzl,13 William Maixner,9 Arabella Meixner,1 Winston Thomas,14 J. Andrew Pospisilik,15 Mattias Alenius,16 Michaela Kress,6 Sai Subramaniam,10 Paul A. Garrity,12 Hugo J. Bellen,17 Clifford J. Woolf,4,* and Josef M. Penninger1,* 1Institute
of Molecular Biotechnology of the Austrian Academy of Sciences, Dr. Bohr Gasse 3-5, A-1030 Vienna, Austria Program, Garvan Institute of Medical Research, 384 Victoria Street, Darlinghurst, Sydney, New South Wales 2010, Australia 3Department of Experimental and Clinical Pharmacology and Toxicology, University of Erlangen-Nuremberg, Fahrstrasse 17, 91054 Erlangen, Germany 4Program in Neurobiology, Children’s Hospital Boston and Department of Neurobiology, Harvard Medical School, Boston, MA 02115, USA 5Biology Department, New York University, 100 Washington Square East, New York, NY 10003, USA 6Division of Physiology, Department of Physiology and Medical Physics, Innsbruck Medical University, Fritz-Pregl-Strasse 3, A-6020 Innsbruck, Austria 7Department of Anesthesia and Critical Care, Massachusetts General Hospital and Harvard Medical School, Boston, MA 02129, USA 8Molecular Epidemiology of Pain Program, Department of Anesthesiology, University of Pittsburgh, Pittsburgh, PA, USA 9Center for Neurosensory Disorders, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599, USA 10Strand Life Sciences Pvt. Ltd., 237 C V Raman Avenue, Rajmahal Vilas, Bangalore, India 11Department of Neuroscience, Baylor College of Medicine, Houston, TX 77030, USA 12National Center for Behavioral Genomics, Department of Biology, Brandeis University, Waltham, MA 02458, USA 13Howard Hughes Medical Institute, Stanford University, 318 Campus Drive, Stanford, CA 94305, USA 14Deltagen, Inc., 1900 South Norfolk Street, Suite 105, San Mateo, CA 94403, USA 15Max Planck Institute for Immunobiology, Stuebeweg 51, D-79108 Freiburg, Germany 16Department of Clinical and Experimental Medicine, Linkoping University, Linkoping, SE-581 85, Sweden 17HHMI, Department of Molecular and Human Genetics, Department of Neuroscience and Program in Developmental Biology, Baylor College of Medicine, Houston, TX 77030, USA 18These authors contributed equally to this work *Correspondence:
[email protected] (C.J.W.),
[email protected] (J.M.P.) DOI 10.1016/j.cell.2010.09.047 2Neuroscience
SUMMARY
Worldwide, acute, and chronic pain affects 20% of the adult population and represents an enormous financial and emotional burden. Using genome-wide neuronal-specific RNAi knockdown in Drosophila, we report a global screen for an innate behavior and identify hundreds of genes implicated in heat nociception, including the a2d family calcium channel subunit straightjacket (stj). Mice mutant for the stj ortholog CACNA2D3 (a2d3) also exhibit impaired behavioral heat pain sensitivity. In addition, in humans, a2d3 SNP variants associate with reduced sensitivity to acute noxious heat and chronic back pain. Functional imaging in a2d3 mutant mice revealed impaired transmission of thermal pain-evoked signals from the thalamus to higher-order pain centers. Intriguingly, in a2d3 mutant mice, thermal 628 Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc.
pain and tactile stimulation triggered strong crossactivation, or synesthesia, of brain regions involved in vision, olfaction, and hearing. INTRODUCTION Acute and chronic pain affects millions of people after injury or surgery and those suffering from diseases like arthritis, cancer, and diabetes. Nociception (the detection of noxious or damaging stimuli) serves a crucial biological purpose: it alerts living organisms to environmental dangers, inducing the sensation of pain, reflex withdrawal, and complex behavioral and emotional responses, which protect the organism from further damage (Cox et al., 2006). Noxious stimuli are detected by specialized high-threshold primary sensory neurons (nociceptors) (Lumpkin and Caterina, 2007), which transfer signals to the spinal cord and then transmit them to the brain for higherlevel processing that results in the conscious awareness of the sensation called pain (Tracey and Mantyh, 2007). The
A
Control
B
Avoidance
* *
% Prefe erence
100 80
100
*
80
60
60
40
40
20
20
0
Thermal challenge 46°C
painless
0
25 32 35 46
25 32 35 46
Temperature (ºC) p ( )
C
Pan-neuronal Driver (+ UAS-DCR2)
0.150
Control painless
Frequenc cy
0.125 0.100 0.075
confirm
UAS-IR (Transgene)
♂
♀
Z score ≥1.65 Cut-off
F1
0.050 0.025 0.000 100
% Avoidance 80
60
40
20
0
% Avoidance
Noxious heat (46°C)
D
E Frequency
0.5
Total: 16051 Hit 622 (3%) Hits: Genes: 580
0.3 0.05
0 00 0.00 100
80
60
40
20
0
% Avoidance
functional importance of pain perception is exemplified by individuals with defects in nociception; patients with congenital insensitivity to pain do not survive past their twenties (Basbaum et al., 2009). Drosophila (fruit flies) respond to noxious stimuli and have become a powerful model organism for studying the genetics of nociception (Manev and Dimitrijevic, 2004; Tracey et al., 2003; Xu et al., 2006). For instance, the TRP channel PAINLESS was identified as a heat-responsive channel mediating thermalbased nociception in fly larvae (Sokabe et al., 2008; Tracey et al., 2003). Using genome-wide neuronal-specific RNAi knockdown, we report a global screen for an innate behavior and identify hundreds of novel genes implicated in heat nociception in the fly, including the a2d family calcium channel subunit straightjacket. Conservation of the mammalian straightjacket ortholog, a2d3, in thermal nociception was confirmed in knockout mice that exhibit significantly impaired basal heat pain sensitivity and delayed thermal hyperalgesia after inflammation. In humans,
Figure 1. Thermal Nociception in Adult Drosophila (A) Schematic representation of the thermal nociception assay in adult Drosophila. (B) Avoidance of noxious temperature of 46 C, but not avoidance of ‘‘subnoxious’’ temperatures (25 C –35 C), is impaired in painless mutant (Painless(EP(2)2451) flies compared to the control strain Canton S (control). Data are presented as mean values ± SEM. Approximately 20 flies were tested per group in replicates of at least four cohorts. Significant differences (p < 0.001) were observed for temperature and strain responses. Further post hoc (Tukey’s) analysis showed a significant temperature avoidance response at 46 C for control (*p < 0.05), but not painless, flies when compared to responses at 25 C. (C) To set up the experimental screening system, w1118 (isogenic to the UAS-IR library) 3 elav-Gal4 flies (control, gray; n = 1706) and painless mutant flies (painless, blue; n = 1816) were tested for avoidance to noxious heat (46 C). Based on these data, a Z score R 1.65 was calculated as a specific cutoff to identify lines for further screening. ElavGal4 (also containing UAS-Dicer 2 [UAS-DRC2] for more efficient gene silencing) females were crossed to UAS-IR lines to knock down the target genes in all neurons. All lines that exhibited a thermal avoidance defect (Z score R 1.65) were rerested multiple times. (D) Results of the genome-wide screen. Approximately three percent (622 transformants) of total lines tested (16,051) exhibited a defect in thermal nociception, resulting in 580 candidate pain genes (622 transgenic lines). (E) Distribution of adult thermal nociception and developmental lethal hits for 16,051 Drosophila UAS-IR lines. 1427 elav-Gal4 3 UAS-IR lines were developmentally lethal (lethal). Among the 14624 viable lines, 562 lines exhibited defective thermal nociception (pain). An additional 60 lines that exhibited defective nociception as well as a semilethal phenotype were labeled as pain and lethal. See also Figure S1 and Table S1.
we found single-nucleotide polymorphisms (SNPs) in a2d3 that are associated with reduced acute heat pain sensitivity in healthy volunteers and chronic postsurgical back pain. RESULTS A Genome-wide Screen for Thermal Nociception To identify genes required for nociception, we developed a highthroughput behavioral assay to determine the response of adult Drosophila to noxious heat as a stimulus. When exposed to a surface at a constant temperature of 25 C, flies distribute evenly in the experimental chamber, but when given a choice between a noxious (46 C) and nonnoxious (31 C) surface, flies rapidly avoid the harmful temperature (Figures 1A and 1B). painless mutant flies respond normally to subnoxious temperatures (%39 C) but fail to avoid noxious heat (46 C) (Figure 1B). Thus, adult flies can rapidly avoid noxious heat, and this complex innate behavior is dependent on painless. Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc. 629
20 0 Co
nt st rol j-I st R 1 j-I R 2
Ca2+
CD8-GFP
GFP Nuclear G
E stj-Gal4>UAS-GFP
F
seco onds
**
40
** **
Anti-STJ antibody
G
stj
46 C 40 30 20 10 0 on tr st ol j-I R st 1 j-I R 2
β
**
60
1.2 1.0 0.8 06 0.6 0.4 0.2 0.0
(until 100% incapacitated)
α1
80
D
C
γ α1
100
C
nt st rol j-I st R 1 j-I R 2
α2
46 C
Relative ex xpression
Ca2+
B
Co
α2-δ
% Avoiidance
A
stj in situ stj dapi
leg
NC82 stj GFP
Using this assay, we performed a genome-wide behavioral screen using the Vienna global Drosophila RNAi library (Dietzl et al., 2007). The panneuronal-specific elav-Gal4 driver line was crossed to flies containing UAS-IR (IR, inverted repeat) transgenes covering the expressed genome (Figure 1C). Testing control flies (n = 1706) over many different days revealed that the vast majority avoided the noxious surface, with a mean avoidance response of 92% ± 6.4% SD, whereas painless mutants (n = 1816) exhibited a markedly reduced avoidance response (51% ± 9.97% SD). Based on these data, we set our primary cutoff at the 95 percentile of probability, corresponding to a Z score of > 1.65 (Figure 1C). At this threshold, we consistently observed impaired thermal nociception in painless mutant flies. To identify novel genes regulating pain, we tested 16051 elavGal4 > UAS-IR combinations targeting 11664 different Drosophila genes (82% of the Drosophila genome version 5.7) for effects on noxious temperature avoidance (Figure 1D and Figure S1B available online). Positive hits were retested, and 622 specific transgenic UAS-IR lines corresponding to 580 genes were identified as candidate thermal nociception genes (Figure 1E and Tables S1A–S1C). Approximately 9% of the neuronal elav-Gal4-driven UAS-IR lines were lethal, yielding no or few progeny (Figure 1E and Table S1D). Gene Ontology (GO) and GO gene set enrichment analysis of the total screen data (Figures S1C–S1D and Tables S1E and S1F) showed a significant enrichment of genes involved in ATP synthesis, 630 Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc.
Figure 2. Straightjacket Controls Thermal Nociception in Adult Drosophila (A) Diagram of the a2-d family encoding peripheral subunits of Ca2+ channels. (B) RNAi knockdown of stj impairs noxious thermal avoidance in adult Drosophila (percent avoidance of noxious temperature). stj-IR1 = inverted repeat 1, stj-IR2 = inverted repeat 2, both crossed to elav-Gal4;UAS-DCR2. (C) Q-PCR for stj knockdown efficiency in elav-Gal4 > UAS-stj-IR1/2 adult fly brains. (D) Kinetics of temperature-induced paralysis for control and elav-Gal4 > UAS-stj flies. (E) stj-Gal4 driving expression of lamin:GFP to label nuclei and cell surface CD8:GFP to visualize axonal projections in the brain of adult flies. The pars intercerebralis is marked with an arrow. (F) Colocalization of anti-STJ immunostaining and stj-Gal4 > UAS-lamin:GFP. The pars intercerebralis is marked with an arrow. (G) stj in situ hybridization in the leg of wild-type (w1118) flies. Of note, the sense control did not show any signal. DAPI counterstaining is shown as mark nuclei. All data are presented as mean ± SEM. *p < 0.05; **p < 0.01 (Student’s t test). See also Figure S2.
stj dapi
sensilla neurotransmission, and secretion. We further annotated 80 nociception hits with previously unknown functions (Table S1G). KEGG pathway analyses of the primary thermal nociception hits and their respective binding partners (Table S1H) revealed significant enrichment for oxidative phosphorylation, amino acid and fatty acid metabolism, ubiquitin-mediated proteolysis, and various signaling pathways such as Wnt, ErbB, hedgehog, JAK-Stat, Notch, mTOR, or TGFb (Table S1I). Thus, our thermal nociception screen and the subsequent bioinformatic analyses have revealed multiple genes and pathways that relate to the expression of an innate nociceptive behavior, many of which had no previous functional annotations. straightjacket Is a Thermal Nociception Gene in Drosophila One of the genes that we picked up in this screen was straightjacket (CG12295, stj), a member of the a2d family of genes that function as subunits of voltage-gated Ca2+ channels (Figure 2A) and control the function and development of synapses (Catterall, 2000; Dickman et al., 2008; Eroglu et al., 2009; Kurshan et al., 2009; Ly et al., 2008). The fly stj ortholog in mammals is a2d3, a close homolog of a2d1, which is the molecular target of gabapentin and pregabalin (Field et al., 2006), widely used analgesics for neuropathic pain in humans (Dworkin et al., 2007). We confirmed that stj is required for noxious heat avoidance in adult Drosophila using a second independent hairpin (Figure 2B). The two stj hairpins resulted in about 90% and 60% reduction of stj mRNA expression, respectively (Figure 2C). Importantly, when flies were exposed to 46 C without a choice to escape, stj knockdown did not alter the kinetics of temperature-induced
A
Futch
B
50% 40%
Response (%)
30%
Stj-Gal4>UAS-CD8:GFP
w1118 X elav
50%
30% 20%
10%
10%
50% 40% 30% 20%
0%
1 2 3 4 5 6 7 8 9 ≥10
stj2
40%
mean 5.23 sec (+/-0.4) P0.05
40%
mean 6.28sec (+/-0.67) P UAS-Lamin:GFP-positive cells of the pars intercerebralis (Figure 2F and data not shown). We also found GFP+ nuclei and projections in the ventral nerve cord (VNC) and ascending/descending axons from the VNC that innervate the central brain (Figure S2C). stj-specific in situ hybridization revealed stj transcripts in the sensory organ (sensilla) of the leg (Figure 2G), indicating expression in the peripheral and central nervous system of the fly. Further studies are required to fine map the site of stj action in the Drosophila pain circuit. We next assessed whether stj also controls thermal nociception in the larval heat pain paradigm (Tracey et al., 2003). In larvae, we found expression of stj-Gal4 > UAS-CD8:GFP in multidendritic sensory neurons (Figure 3A). Panneuronal knockdown of stj (UAS-stj-IR x elav-Gal4) abrogated the larval response to noxious heat to an extent even greater than painless (Figure 3B).
Thermal Analgesia in a2d3 Mutant Mice We next tested whether the fly stj data is predictive of altered nociceptive behavior in mammals. The closest stj ortholog in mammals is a2d3 (mouse a2d3 is 33% identical and 60% similar to the STJ protein, and the domain structures are conserved throughout evolution [Ly et al., 2008]). To examine the role of a2d3 in vivo, we studied a2d3 mutant mice generated by homologous recombination. Correct recombination and loss of protein expression were confirmed by Southern (Figure 4A) and western blotting (Figure 4B). a2d3 mutant mice are born at the expected Mendelian frequency and are fertile. Extensive characterization of these mice showed no obvious anatomical or histological abnormalities, including apparently normal brain morphology (Tables S2A and S2B). There were also no genotype-related or biologically significant differences noted between age- and gender-matched mutant and wild-type control mice for any of Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc. 631
A
+/+
Brain
B
+/-
23.1
+/+ +/- -/- +/+ +/- -/-
94 9.4
Targeted
α2δ3
Endogenous
α2δ1 2δ1 Actin
6.6
Hot plate 50
***
D
40
**
30
Inflammatory pain
α2δ3+/+ α2δ3-/-/
20
***
***
10 0
**
20
Latency (sec)
Latency (sec)
C
Figure 4. a2d3 Is Required for Thermal Pain Responses in Mice
DRG
/ α2δ3+/+ -/α2δ3
* *
15
#
#
10 5
50
52
54
Temperature (°C)
56
-2
1
3
5
Days after CFA
the parameters evaluated at necropsy or by serum chemistry and hematology (Tables S2C and S2E). Moreover, normal growth and body weights were recorded for mice at 49, 90, 180, and 300 days of age (Table S2F). Hence, by all anatomical and physiological parameters assessed, a2d3 mutant mice appear normal. Importantly, similar to Drosophila stj mutants, a2d3 mutant mice showed a defect in acute thermal nociception in the hot plate assay, with diminished responsiveness at 50, 52, 54, and 56 C (Figure 4C). In addition, a2d3 mutant mice exhibited delayed thermal sensitization in the Complete Freund’s Adjuvant (CFA) model of peripheral inflammatory pain (Figure 4D), indicating that a2d3 contributes to the acute phase of heat hyperalgesia. CFA induced inflammation, as determined by paw swelling, was comparable between a2d3 mutant and control mice (Figure S4A). By contrast, mechanical sensitivities (von Frey test) were unaffected in a2d3 mutant mice (Figure S4B). a2d3 mutant mice were also evaluated for other behavioral tasks (Crawley, 2008): open field test to assess locomotor activity, general exploratory behavior, intrasession habituation, and general anxiety; tail suspension to assess behavioral despair; and a rotarod test to assess basic motor skills and coordination. In these assays, no significant differences were observed between a2d3 mutant and control mice (Figures S4C–S4F and Table S2G). Thus, genetic deletion of a2d3 in mice results in substantially impaired acute heat pain responses and a delay in inflammatory heat hyperalgesia. a2d3 SNPs Associate with Human Pain Sensitivity Because knockdown of straightjacket in Drosophila and knockout of a2d3 in mice result in impaired sensitivity to thermal 632 Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc.
7
(A) Southern blotting of genomic DNA in a2d3 wild-type (+/+) and a2d3 heterozygous (+/) ES cells to confirm successful gene targeting. The endogenous wild-type and targeted alleles are indicated. A 50 probe was used on Nhe I digested genomic DNA. (B) a2d3 and a2d1 protein expression in brain and isolated DRG lysated from a2d3+/+ (+/+), a2d3+/ (+/), and a2d3/ (/) mice. Actin is shown as a loading control. (C) Using the hot plate assay, a2d3 mutant mice (n = 16) show a delayed acute thermal nociception response as compared to control a2d3+/+ mice (n = 12). Littermate mice were used as a control. Values represent the latency (seconds) to respond to the indicated temperatures. (D) CFA-induced inflammatory pain behavior. CFA (20 ml) was injected into the hindpaw of a2d3+/ (+/+, n = 10) and a2d3/ (/, n = 21) littermates, and mice were tested for thermal pain (54 C) using the hot plate assay on the indicated days. Days 1, 3, 5, and 7 indicate days after CFA injection. All data are presented as mean values ± SEM. *p < 0.05; **p < 0.01; and ***p < 0.001 comparing mutant versus control mice. #p < 0.05 comparing sensitization to baseline (day 2) of the same genotype (Student’s t test). See also Figure S4 and Table S2.
pain, we speculated that polymorphisms at the a2d3 (CACNA2D3) locus might be associated with heat pain variance in humans. To assay for potential association of a2d3 haplotypes relative to pain sensitivity, we screened four single-nucleotide polymorphisms (SNPs) within or close to the human CACNA2D3 gene (Figure 5A) in a cohort of 189 healthy volunteers subjected to a battery of experimental pain sensitivity tests (Diatchenko et al., 2005). Of these, the minor allele of the SNP rs6777055 was significantly associated with reduced thermal pain sensitivity, i.e., heat wind-up pain (Figure 5B, recessive model). Wind-up measures successive increases in perceived pain intensity to a repeated noxious heat stimulus (10 heat pulses of 1.5 s each at 50 C, each separated by 3 s). Thermosensitive neurons have been also implicated in chronic pain in humans (Premkumar, 2010). To explore this, we compared pain levels in 169 Caucasian adults who participated in a prospective observational study of surgical discectomy for persistent lumbar root pain, caused by an intervertebral disc herniation (Atlas et al., 2001) for an association with CACNA2D3 SNPs. The minor alleles of two CACNA2D3 SNPs (rs1851048 and rs6777055) were associated with less pain within the first year following surgery (Figure 5C, recessive model). Importantly, the rs6777055 SNP C/C was significantly associated with less pain in both healthy volunteer and chronic pain cohorts showing a recessive mode of inheritance. In both the experimental and lumbar pain groups, the minor allele frequency for rs6777055 was 0.2; that is, 4% of the human population is homozygous for this genetic variant. These data show that minor variants within the human CACNA2D3 gene are associated with less heat-induced pain in healthy volunteers and reduced chronic pain in lumbar back pain patients.
A 3p21.1
CACNA2D3 54.157
55.109
Acute (wind up)
C
rs6777055
PCA
-0.5
C/A + A/A C/C N=184
-1.0 -1.5
*
N=5 N 5
-2.0
rs6777055
Z sco ore
0.0
rs s6777055
Chronic
0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7
rs1851048
C/A + A/A C/C
Z sco ore
B
rs s9808962 rs s1918070
55.5
rs s1851048
54
N=153
*
N=8
0.2 0.1 -0.0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6
A/G + G/G A/A N=137
*
N=17
Figure 5. Polymorphisms in CACNA2D3 (a2d3) Associate with Decreased Acute and Chronic Pain in Humans (A) Schematic representation of the human CACNA2D3 gene locus on chromosome 3p21.1. The positions of the SNPs assayed are indicated. Blue boxes represent exons. The relative gene position is given in megabases (Mb). (B) Homozygous carriers of the rs6777055 minor allele (C/C) were significantly less sensitive to heat wind-up-induced sensitivity relative to the other genotypes (C/A or A/A). (C) Of 169 lumbar chronic root pain patients, 1 year postdiscectomy, those homozygous for the minor allele C/C at SNP rs6777055 and A/A at SNP rs1851048 were less sensitive than the other allele combinations. In each case, the homozygous minor allele is associated with significantly less pain. Note that genotyping was not always successful for every individual, hence the slightly different total numbers in the chronic pain group. All data are presented as mean values ± SEM. *p < 0.05 (Student’s t test).
a2d3 Controls Central Transmission of Pain Signals to the Sensory Cortex and Other Higher-Order Pain Centers Nociceptive processing involves the relay of sensory information from primary nociceptor neurons to second-order neurons in the dorsal horn of the spinal cord that then transfer nociceptive information to the brain stem, thalamus, and higher-order brain centers (Costigan et al., 2009; Lumpkin and Caterina, 2007; Tracey and Mantyh, 2007). Because our a2d3 knockout mice carry a LacZ reporter, we used b-Gal staining as a marker to assess a2d3 expression. In the brain, b-Gal labeled the thalamus, pyramidal cells of the ventro-posterior paraflocculus of the cerebellum, caudate, putamen, the dentate gyrus of the hippocampus, and the olfactory bulb and tubercle, as well as diffusely throughout the cortex (Figure 6A and data not shown). The LacZ expression profiles were confirmed by western blotting and quantitative PCR (data not shown) and matched in situ data from the Allen brain atlas (data not shown and Koester and Insel, 2007). We failed to detect LacZ expression in the spinal cord and DRG (data not shown). Absence of a2d3 expression in primary
sensory DRG neurons was confirmed by western blotting (Figure 4B). In line with these expression data, our behavioral experiments showed that loss of a2d3 does not affect the noxious heat-induced tail flick response (Figure S4G), a pain behavior mediated by a spinal reflex circuit (Pitcher et al., 1995). Finally, patch clamping showed that calcium current and kinetics were comparable among DRG neurons from control and a2d3 mutant mice (Figures S4H–S4K), indicating no requirement for a2d3 in calcium channel function in these cells. These data suggest that a2d3 is not required for thermal pain processing in nociceptors and the spinal cord, but a2d3 may regulate thermal pain processing in the brain. To address this, we employed noninvasive functional magnetic resonance imaging (fMRI) using the blood oxygenation level-dependent (BOLD) signal to generate activation maps of brain regions affected by noxious topical heat stimuli (Knabl et al., 2008; Ogawa et al., 1990; Thulborn et al., 1982). In wild-type mice (n = 20), noxious thermal stimuli activate brain structures known as the ‘‘pain matrix’’ (Melzack, 1999) such as the thalamus (Figure 6B), the S1 and S2 somatosensory cortex (Figure 6C), the cingulum, amygdala, hypothalamus, or the motor cortex (Figure S5). These areas are also involved in pain perception in human subjects (Tracey and Mantyh, 2007). In both wild-type (n = 20) and a2d3 mutant mice (n = 18), thermal pain induced activation of the thalamus (Figure 6B), the key subcortical pain relay center (Price, 2000). Intriguingly, loss of a2d3 expression interrupted the normal engagement of pain circuitry in the brain, resulting in markedly reduced BOLD peak amplitudes and activation volumes in higher-order pain centers such as the somatosensory S1 and S2 cortices, cingulate, or motor cortex after exposure to noxious temperatures (Figure 6C and Figure S5). Impaired activation of higher-order pain centers, i.e., sensory and motor cortices, was confirmed by calculation of Euclidian distances (data not shown). To additionally assess temporal information flow of the pain signal within different cerebral structures, we calculated a cross-correlation matrix of the response time profiles for each predefined region of the somatosensory pain matrix (Figure 6D). In control mice, the sensory input relays thermal-evoked neural signals to the thalamus, where it effectively spreads to other central brain centers such as the sensory and association cortex, limbic system, cerebellum, basal ganglia, and motor cortex. a2d3 mutant mice again exhibited normal activation of the thalamus but a reduced flow to nearly all of the higher-order pain centers, in particular the somatosensory cortex (SC) (Figure 6D). Moreover, whereas the pain signal spreads from the left (i.e., contralateral to the side of stimulation) in control mice, we found a considerable reduction in correlation coefficients of the pain signal from the left to the right brain in a2d3 mutant mice (Figure S6A). In addition, we observed increased negative BOLD signals in the S1 somatosensory, the motor, and the cingulate cortex on both hemispheres of a2d3 mutant mice (Figure S6B), suggesting that genetic inactivation of a2d3 not only results in impaired transmission of the signal to higher pain structures, but also in intracortical inhibition (Arthurs and Boniface, 2002; Shmuel et al., 2002). Thus, loss of a2d3 leads to impaired transmission of noxious heat-evoked signals from the thalamus to higher pain centers. Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc. 633
Figure 6. a2d3 Is Expressed in the Brain and Relays the Pain Signal to Higher-Order Brain Centers
Hippocampus
Cortex
A
Cerebellum
Olfactory Bulb
Olfactory Tubercle
α2δ3+/+ α2δ3-/-
1.50
1.75
Voxel
0.75
*
0.50
150 100
*
50
0.25 0.00 50° C
55° C
200 1.25
0.75 0.50
** **
50°C
45°C
55°C
**
100
**
0
0.00 45°C
150
50
α2δ3+/+
50°C
45°C
55°C
α2δ3-/-
50°C
55°C
SI
Th
SC
AC
LL
LS
HT
BG
CM
D
**
1.00
0.25
0 45° C
250
1.50 200
1.25
1.00
S1/S2 Somatosensory Cortex
250
ge [%] Bold Chang
1.75
Bold Change [%]
C
Thalamus
Voxel
B
Thalamus
(A) b-Gal staining of whole brain slices from a2d3+/ mice that carry the LacZ cassette. Different brain regions that are positive for LacZ expression are indicated. White lines indicate the brain slices displayed in Figure 7A. (B and C) Quantification of percent of BOLD change and mean activation volume (in voxels) for (B) the thalamus and (C) the S1 somatosensory cortex of a2d3+/+ and a2d3/ mice. Of note, it has been proposed that the S1 cortical region is involved in the localization of nociception (Treede et al., 1999). The different stimulation temperatures are indicated. Data are presented as mean ± SEM. *p < 0.05; **p < 0.01 (Student’s t test comparing the respective control and a2d3/ groups). (D) Cross-correlation matrix of time profiles. Whereas the pain signal spreads from the thalamus to other higher-order pain centers in a2d3+/+ mice (red areas), in a2d3/ mice, correlated activation can only be observed up to the level of the thalamus. Very weak activity is found in somatosensory cortex (SC) for a2d3/ mice (green stripes). Data from the structures of the left side of the brain are shown following challenge with noxious heat (55 C) at the right hindpaw. SI, sensory input; Th, thalamus; SC, somatosensory cortex; AC, association cortex; LL, link to limbic system; LS, limbic system; HT, hypothalamus; BG, basal ganglia; C, cerebellum; M, motor cortex; P, periaquaeductal gray. Correlation coefficients (cc) are given in the range from 0 (green) to +1 (red). n = 20 for a2d3+/+; n = 18 for a2d3/. See also Figures S5 and Figure S6.
SI
Th
SC
AC
LL
LS
HT
0
BG
cc
CM
SI
Th
SC
AC
1
Loss of a2d3 Results in Sensory Cross-Activation Although we found a marked impairment in the activation of known higher-order pain centers, one puzzling finding from our fMRI data was that we did not find statistically significant differences between control and a2d3 mutant mice in total activation volume and peak height when neuronal activity was surveyed in the entire brain (Figure S7A). Because we had initially only focused on the pain matrix, we speculated that, therefore, loss of a2d3 may result in hyperactivation of additional brain regions. Remarkably, noxious heat stimulation of all a2d3 mutant mice triggered significantly enhanced activation of the visual cortex and the auditory cortex, as well as olfactory brain regions (Figures 7A and 7B). Thus, in a2d3 mutant mice, noxious heat stimulation results in a significant sensory crossactivation of brain regions involved in vision, hearing, and olfaction. 634 Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc.
LL
To image basal neuronal activity integrated over a 24 hr time period, we LS HT BG CM employed manganese labeling (MEMRI) (Silva et al., 2004). We observed similar activity in all imaged brain regions among a group comparison of control and a2d3 mutant mice, indicating that the observed stimulus-induced sensory cross-activations are not due to altered basal neuronal activity. Moreover, diffusion tensor imaging (DTI) showed no overt structural changes with respect to fractional anisotropy between the thalamus and higher-order pain centers or the thalamus and the visual, auditory, and olfactory centers. Further, cross-correlation analysis of the time profiles of the structures of the pain matrix from resting state BOLD imaging showed no defects in spontaneous spreading from the thalamus to higher-order pain centers in a2d3 mutant mice (data not shown). Finally, network analysis of this resting state brain activity showed no overt changes in total functional connectivity within the pain matrix (1087 connections in wild-type versus 1040 connections in a2d3 mutant mice) and also apparently normal multisensory-thalamo-cortical network connectivity (1922 connections in wild-type versus 2099 connections in a2d3 mutant mice). Although we cannot exclude subtle
Heat
A M
M2 M1
Cg
Figure 7. a2d3 Mutant Mice Exhibit Sensory CrossActivation in Response to Thermal and Tactile Stimuli
Tactile
S1
Cg1 fmi Cg2 IL CPu DP
AID AIV DEn aca Pir
CPu
OT
OT
M
S1 S1BF
CA
S1
S1
RSG
S1 S2
S2
S2 AIP
CPu GP B
CM
VL
Pir
Amd
HT V2M V1 V2L S AuD Au1 CA1 TcA CA3
LEnt
VC AC
Op
Bold Change C [%]
2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2
VC AC
DpG PAG DpMe RMC SN IP
Sensory cross-activation (heat)
B
SC
*
*
*
OT
VC
AC
Sensory cross-activation (tactile)
C
Bold Change C [%]
Ect PRh
RSG
2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2
α2δ3+/+ α2δ3-/-
** **
(A) Second-order statistical parameter maps showing only the significant differences of heat (55 C) and tactile (vibrissae) stimulation-induced brain activation between a2d3+/+ and a2d3/ mutant mice. Activation was assessed by BOLD-fMRI. The three planes correspond to the white lines shown in Figure 6A. The green/blue scale indicates increased peak activation (55 C) in a2d3+/+ control mice compared to a2d3/ mutant mice. The yellow/red scale indicates increased activation in a2d3/ mutant mice compared to a2d3+/+ control mice. Images depict significant differences of second-order group statistics corrected for multiple comparisons over all mice tested (n = 20 for a2d3+/+ mice; n = 18 for a2d3/ mice). Arrows point to activated regions; note that, for heat stimulation, the S1/S2 somatosensory cortex, the cingulate (Cg) cortex and the motor (M) cortex show significantly higher activity in a2d3+/+ controls. In a2d3/ mice, heat stimulation leads to significantly higher activity in the auditory cortex (AC), the visual cortex (VC), and the olfactory tubercle (OT), as well as the amygdala (Amd) and the hypothalamus (HT). For tactile stimulation, only one small region in the S1 somatosensory cortex, ipsilateral to the side of stimulation (right), showed significantly higher activity in a2d3+/+ controls, whereas a2d3/ mice again exhibited increased activation of the VC, AC, and OT, in addition to the caudate putamen (Cpu), S1, and S2 regions of the somatosensory cortex and the superior colliculus (SC). (B and C) Percent of BOLD changes in the auditory cortex (AC), olfactory tubercle (OT), and visual cortex (VC) in control and a2d3/ mice following (B) heat (55 C) and (C) tactile vibrissal stimulation. Data are presented as mean values ± SEM. *p < 0.05; **p < 0.01 (Student’s t test). See also Figure S7.
** DISCUSSION
OT
VC
AC
developmental changes in defined neuronal populations of a2d3 mutant mice, these data suggest that the observed heat-induced sensory cross-activations (and defective transmission of the thermal pain signal from the thalamus to higher-order pain centers) are not due to altered basal neuronal connectivities. To assess whether loss of a2d3 also results in sensory crossactivation in other sensory modalities, we performed BOLD imaging in response to tactile vibrissal stimulation (Hess et al., 2000). Among control and a2d3 mutant mice, we observed apparently normal activation of the brain region that encompasses the barrel field (data not shown); the barrel field is the primary cortical somatosensory brain center for processing of vibrissal stimulation (Petersen, 2007). Remarkably, although we observed apparently normal activation of the barrel field, tactile stimulation again resulted in sensory cross-activation of visual, auditory, and olfactory brain centers in a2d3 knockout mice (Figures 7A and 7C). Thus, in a2d3 mutant mice, noxious heat stimulation, as well as tactile stimulation, trigger sensory cross-activation of brain regions involved in vision, hearing, and olfaction.
Our whole-genome, neuron-specific RNAi screen provides a global functional analysis of a complex, innate behavior. We have uncovered hundreds of candidate genes for thermal nociception, a large proportion of which had completely unknown functions until now. Because many of these genes are conserved across phyla, our data provide a starting point for large-scale human genomics efforts to finding novel pain genes and defining the molecular mechanisms of nociception. One of the screen hits was the calcium channel subunit straightjacket/a2d3. In both larva and adult Drosophila, we show that straightjacket is indeed required for heat nociception. Further work is required to define the spatial and temporal requirements for stj in thermal nociception. Importantly, similar to the fly, genetic deletion of a2d3 in mice also results in impaired acute heat pain responses. These results translate to humans because we found a2d3 polymorphisms that significantly associate with reduced heat pain sensitivity in healthy volunteers and lower levels of chronic pain in lumbar back pain patients. These data reinforce the extraordinary conservation of the neurobiological mechanisms of nociception, from its manifestation as avoidance of damage in primitive creatures like flies to the complex sensation of pain in humans. Our functional imaging data indicate that a2d3 appears to be specifically required for central transmission of the thermal Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc. 635
nociceptive signal from the thalamus to the sensory cortex and other higher-order pain centers. Until now, it has been assumed that the threshold for thermal pain sensitization is set exclusively in peripheral sensory neurons via thermosensitive TRP channels like TRPV1 and altered excitability of nociceptor terminals (Hucho and Levine, 2007). However, our results indicate that noxious heat-evoked behavior is reduced even when nociception appears to be intact up to the thalamus. Whether the loss of a2d3 results in defective signaling and/or subtle alterations in synaptic circuits that link the thalamus to higher-order pain centers remains to be determined. Intriguingly, in a2d3 mutant mice, thermal pain and tactile vibrissal stimulation triggered strong cross-activation of brain regions involved in vision, olfaction, and hearing. Sensory cross-activation, or synesthesia, is a neurological condition in which a stimulus in one sensory modality triggers perception of another sense (Hubbard and Ramachandran, 2005). In humans, synesthesia can be only objectively verified using functional brain imaging (Aleman et al., 2001). Multiple forms of synesthesia exist, including pain stimuli that trigger color (Dudycha and Martha, 1935). Synesthesia might affect up to 4% of the population, shows genetic linkage, and has been associated with intelligence and creativity (Hubbard and Ramachandran, 2005). In addition, thalamic lesions can also cause synesthesia (Beauchamp and Ro, 2008). Thus, a2d3 mutant mice might provide an animal model to enable the phenomenon of sensory cross-activation to be experimentally dissected. EXPERIMENTAL PROCEDURES Detailed experimental procedures are provided in the Supplemental Information. Fly Stocks All UAS-IR transgenic fly lines were obtained from the VDRC RNAi library (Dietzl et al., 2007) with the exception of the second stj hairpin, which was obtained from the Harvard trip stocks. elav and UAS-Dcr2 were gifts from B. Dickson (Dietzl et al., 2007). See Extended Experimental Procedures for a complete list of stocks used. Drosophila Behavioral Tests For avoidance of noxious heat, 20 4-day-old flies were placed in a sealed experimental chamber. All tests were performed in the dark. The bottom of the chamber was heated to 46 C, whereas the subnoxious zone was measured to be 31 C at the end of the 4 min experiment. Percent of avoidance was calculated by counting the number of flies that failed to avoid the noxious temperature compared to the total number of flies in the chamber. Larval pain assays were performed as described (Tracey et al., 2003). Mechanosensation (Kernan score) was performed as described (Kernan et al., 1994). Detection of stj Expression Brains and ventral nerve cords of adult flies and sensory nerves of larvae from stj-Gal4 > UAS-Lamin:GFP (to detect nuclei) and stj-Gal4 > UAS-CD8:GFP (to detect axonal projections) lines were imaged. stj-expressing neurons were further detected using antibody staining and in situ hybridization. Generation of a2d3 Knockout Mice For gene targeting of a2d3 in mice, a targeting vector was inserted into exon 15 of the murine a2d3 gene. Germline-transmitted F1 mice were backcrossed onto a C57BL/6 background. All behavioral and fMRI mouse studies were conducted in accordance with guidelines of the European Union Council (86/609/ EU) and following Austrian regulations for the use of laboratory animals.
636 Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc.
Mouse Behavioral Experiments For the hot plate assay, wild-type and a2d3 mutant littermate mice were tested for hot plate latency at 50 C –56 C. The mechanical pain test was performed by applying von Frey hairs to the dorsal surface of each hind paw until a hind limb withdrawal response was observed; the hair with the minimum bending force required to produce a response was recorded. Inflammatory thermal hyperalgesia was produced in the mouse right hind paw by intraplantar injection of CFA. Before and after CFA injection, nociceptive responses to heat were measured using the hot plate test (54 C). Paw swelling was measured to evaluate the inflammatory response elicited by CFA. Western Blotting The following primary antibodies were used: mouse anti-Actin, dilution 1/1000; goat anti- a2d1, dilution 1/500, and rabbit anti-a2d3, dilution 1/500. To generate the anti-a2d3 antibody, rabbits were immunized with the peptide VSERTIKETTGNIAC conjugated to KLH. Secondary antibodies were used at a dilution of 1 in 5000. LacZ Expression Tissues from 7- to 12-week-old heterozygote mice were analyzed for LacZ expression using X-Gal staining followed by Nuclear Fast Red counterstaining. For whole-mount brain staining, the brain was cut longitudinally, fixed, and stained using X-Gal. Tissues were fixed with buffered formaldehyde. fMRI and BOLD Imaging Male mice were anesthetized with isoflurane and placed inside of a MR machine (Bruker BioSpec 47/40, quadratur head coil) under extensive physiological monitoring. The contact heat stimuli (40 C, 45 C, 50 C, and 55 C, plateau for 5 s after 15 s of heat increase) were applied at the right hind paw (presented at 3 min 25 s intervals, three times each temperature) using a custom-made computer-controlled Peltier heating device. For tactile stimulation, the C1 vibrissa of the mice was moved with an air-driven device integrated into a cradle shifting the vibrissa by an inverted comb with an amplitude of 5 mm at 7 Hz. A series of 750 sets of functional images (matrix 64 3 64, field of view 15 3 15 mm, slice thickness 0.5 mm, axial, 22 slices) were collected using the Echo Planar Technique (EPI, single shot: TR = 4000 ms, TEef = 24 and 38 ms). SNP Mapping in Humans We genotyped four single-nucleotide polymorphisms (SNPs) spaced evenly through a2d3 using the 50 exonuclease method (Tegeder et al., 2006). For acute pain studies, we genotyped 189 normal volunteers who had previously been phenotyped for ratings of experimental pain (Diatchenko et al., 2005). All subjects gave informed consent following protocols approved by the UNC Committee on Investigations Using Human Subjects. Volunteers were phenotyped with respect to temporal summation of heat pain (i.e., wind-up). For chronic pain studies, we collected DNA from 169 Caucasian adults who participated in a prospective observational study of surgical discectomy for persistent lumbar root pain (Atlas et al., 2001). The primary endpoint was persistent leg pain over the first postoperative year. Genotype-phenotype associations for each SNP were sought by regression analysis. The collection of DNA and genetic analyses were carried out with the approval of the National Institute of Dental and Craniofacial Research institutional review board, and informed consent was obtained from all subjects. Statistical Analyses For analysis of adult heat dose avoidance responses between and within control and painless flies, a two-way ANOVA was performed, followed by Tukey’s post hoc test. For analysis of adult Drosophila avoidance response and RNAi knockdown efficiency, a Student’s t test (with correction for multiple comparisons) was performed. For analysis of larval pain behavior, we have performed the Kruskal-Wallis nonparametric test for median comparison followed by the Dunn’s post hoc test. For mouse behavior, a Student’s t test was used. For fMRI, the mean activity of each activated brain structure was averaged across all significant activated voxels and subjected to t tests comparing a2d3 mutant and control mice. At the second-level group analysis, a standard analysis of variance (ANOVA) was performed for Z score maps
between the different mice genotypes. Areas of significant group activation differences (p < 0.05) were used as masks to only show the calculated peak activation maps in these regions. For human studies, genotype-phenotype associations for each SNP were sought by regression analysis. Unless otherwise indicated, data are represented as mean values ± SEM.
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.2010.09.047. ACKNOWLEDGMENTS We thank all members of our laboratories and the VDRC for helpful discussions and excellent technical support. We thank Ricardo de Matos Simoes for support with statistical analysis. We thank B.J. Dickson for elav/UAS Dicer 2 stocks. a2d3 mutant mice were generated by Deltagen (San Mateo, CA). J.M.P. is supported by grants from IMBA, the Austrian Ministry of Sciences, the Austrian Academy of Sciences, GEN-AU (AustroMouse), ApoSys, and an EU ERC Advanced Grant. A.C.K. is supported by National Institute of Health NRSA 1 F32GM086207-01. C.J.W. is supported by NIH NS039518 and NS038253. G.G.N. was supported by a Mary Curie IIF Fellowship and EuroThymaide. A.H. is supported by DFG 661/TP4 and BMBF (01EM0514, 01GQ0731, 0314102) and K.B. by the Doerenkamp Foundation for Innovations in Animal and Consumer Protection. P.A.G. is supported by NIH NS044232. Received: March 8, 2010 Revised: August 9, 2010 Accepted: September 24, 2010 Published: November 11, 2010 REFERENCES Aleman, A., Rutten, G.J., Sitskoorn, M.M., Dautzenberg, G., and Ramsey, N.F. (2001). Activation of striate cortex in the absence of visual stimulation: an fMRI study of synesthesia. Neuroreport 12, 2827–2830. Arthurs, O.J., and Boniface, S. (2002). How well do we understand the neural origins of the fMRI BOLD signal? Trends Neurosci. 25, 27–31. Atlas, S.J., Keller, R.B., Chang, Y., Deyo, R.A., and Singer, D.E. (2001). Surgical and nonsurgical management of sciatica secondary to a lumbar disc herniation: five-year outcomes from the Maine Lumbar Spine Study. Spine 26, 1179–1187.
Dickman, D.K., Kurshan, P.T., and Schwarz, T.L. (2008). Mutations in a Drosophila alpha2delta voltage-gated calcium channel subunit reveal a crucial synaptic function. J. Neurosci. 28, 31–38. Dietzl, G., Chen, D., Schnorrer, F., Su, K.C., Barinova, Y., Fellner, M., Gasser, B., Kinsey, K., Oppel, S., Scheiblauer, S., et al. (2007). A genome-wide transgenic RNAi library for conditional gene inactivation in Drosophila. Nature 448, 151–156. Dudycha, G.J., and Dudycha, M.M. (1935). A Case of Synesthesia: Visual-Pain and Visual-Audition. J. Abnorm. Soc. Psychol. 30, 57–69. Dworkin, R.H., O’Connor, A.B., Backonja, M., Farrar, J.T., Finnerup, N.B., Jensen, T.S., Kalso, E.A., Loeser, J.D., Miaskowski, C., Nurmikko, T.J., et al. (2007). Pharmacologic management of neuropathic pain: evidence-based recommendations. Pain 132, 237–251. Eroglu, C., Allen, N.J., Susman, M.W., O’Rourke, N.A., Park, C.Y., Ozkan, E., Chakraborty, C., Mulinyawe, S.B., Annis, D.S., Huberman, A.D., et al. (2009). Gabapentin receptor alpha2delta-1 is a neuronal thrombospondin receptor responsible for excitatory CNS synaptogenesis. Cell 139, 380–392. Field, M.J., Cox, P.J., Stott, E., Melrose, H., Offord, J., Su, T.Z., Bramwell, S., Corradini, L., England, S., Winks, J., et al. (2006). Identification of the alpha2-delta-1 subunit of voltage-dependent calcium channels as a molecular target for pain mediating the analgesic actions of pregabalin. Proc. Natl. Acad. Sci. USA 103, 17537–17542. Hess, A., Stiller, D., Kaulisch, T., Heil, P., and Scheich, H. (2000). New insights into the hemodynamic blood oxygenation level-dependent response through combination of functional magnetic resonance imaging and optical recording in gerbil barrel cortex. J. Neurosci. 20, 3328–3338. Hubbard, E.M., and Ramachandran, V.S. (2005). Neurocognitive mechanisms of synesthesia. Neuron 48, 509–520. Hucho, T., and Levine, J.D. (2007). Signaling pathways in sensitization: toward a nociceptor cell biology. Neuron 55, 365–376. Kernan, M., Cowan, D., and Zuker, C. (1994). Genetic dissection of mechanosensory transduction: mechanoreception-defective mutations of Drosophila. Neuron 12, 1195–1206. Knabl, J., Witschi, R., Ho¨sl, K., Reinold, H., Zeilhofer, U.B., Ahmadi, S., Brockhaus, J., Sergejeva, M., Hess, A., Brune, K., et al. (2008). Reversal of pathological pain through specific spinal GABAA receptor subtypes. Nature 451, 330–334. Koester, S.E., and Insel, T.R. (2007). Mouse maps of gene expression in the brain. Genome Biol. 8, 212. Kurshan, P.T., Oztan, A., and Schwarz, T.L. (2009). Presynaptic alpha2delta-3 is required for synaptic morphogenesis independent of its Ca2+-channel functions. Nat. Neurosci. 12, 1415–1423.
Basbaum, A.I., Bautista, D.M., Scherrer, G., and Julius, D. (2009). Cellular and molecular mechanisms of pain. Cell 139, 267–284.
Lee, T., and Luo, L. (1999). Mosaic analysis with a repressible cell marker for studies of gene function in neuronal morphogenesis. Neuron 22, 451–461.
Beauchamp, M.S., and Ro, T. (2008). Neural substrates of sound-touch synesthesia after a thalamic lesion. J. Neurosci. 28, 13696–13702.
Lumpkin, E.A., and Caterina, M.J. (2007). Mechanisms of sensory transduction in the skin. Nature 445, 858–865.
Catterall, W.A. (2000). Structure and regulation of voltage-gated Ca2+ channels. Annu. Rev. Cell Dev. Biol. 16, 521–555.
Ly, C.V., Yao, C.K., Verstreken, P., Ohyama, T., and Bellen, H.J. (2008). straightjacket is required for the synaptic stabilization of cacophony, a voltage-gated calcium channel alpha1 subunit. J. Cell Biol. 181, 157–170.
Costigan, M., Scholz, J., and Woolf, C.J. (2009). Neuropathic pain: A maladaptive response of the nervous system to damage. Annu. Rev. Neurosci. 32, 1–32. Cox, J.J., Reimann, F., Nicholas, A.K., Thornton, G., Roberts, E., Springell, K., Karbani, G., Jafri, H., Mannan, J., Raashid, Y., et al. (2006). An SCN9A channelopathy causes congenital inability to experience pain. Nature 444, 894–898. Crawley, J.N. (2008). Behavioral phenotyping strategies for mutant mice. Neuron 57, 809–818. Diatchenko, L., Slade, G.D., Nackley, A.G., Bhalang, K., Sigurdsson, A., Belfer, I., Goldman, D., Xu, K., Shabalina, S.A., Shagin, D., et al. (2005). Genetic basis for individual variations in pain perception and the development of a chronic pain condition. Hum. Mol. Genet. 14, 135–143.
Manev, H., and Dimitrijevic, N. (2004). Drosophila model for in vivo pharmacological analgesia research. Eur. J. Pharmacol. 491, 207–208. Melzack, R. (1999). From the gate to the neuromatrix. Pain Suppl. 6, S121–S126. Ogawa, S., Lee, T.M., Kay, A.R., and Tank, D.W. (1990). Brain magnetic resonance imaging with contrast dependent on blood oxygenation. Proc. Natl. Acad. Sci. USA 87, 9868–9872. Petersen, C.C. (2007). The functional organization of the barrel cortex. Neuron 56, 339–355. Pitcher, G.M., Yashpal, K., Coderre, T.J., and Henry, J.L. (1995). Mechanisms underlying antinociception provoked by heterosegmental noxious stimulation in the rat tail-flick test. Neuroscience 65, 273–281.
Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc. 637
Premkumar, L.S. (2010). Targeting TRPV1 as an alternative approach to narcotic analgesics to treat chronic pain conditions. AAPS J. 12, 361–370.
tetrahydrobiopterin regulate pain sensitivity and persistence. Nat. Med. 12, 1269–1277.
Price, D.D. (2000). Psychological and neural mechanisms of the affective dimension of pain. Science 288, 1769–1772.
Thulborn, K.R., Waterton, J.C., Matthews, P.M., and Radda, G.K. (1982). Oxygenation dependence of the transverse relaxation time of water protons in whole blood at high field. Biochim. Biophys. Acta 714, 265–270.
Shmuel, A., Yacoub, E., Pfeuffer, J., Van de Moortele, P.F., Adriany, G., Hu, X., and Ugurbil, K. (2002). Sustained negative BOLD, blood flow and oxygen consumption response and its coupling to the positive response in the human brain. Neuron 36, 1195–1210. Silva, A.C., Lee, J.H., Aoki, I., and Koretsky, A.P. (2004). Manganeseenhanced magnetic resonance imaging (MEMRI): methodological and practical considerations. NMR Biomed. 17, 532–543. Sokabe, T., Tsujiuchi, S., Kadowaki, T., and Tominaga, M. (2008). Drosophila painless is a Ca2+-requiring channel activated by noxious heat. J. Neurosci. 28, 9929–9938. Stuurman, N., Delbecque, J.P., Callaerts, P., and Aebi, U. (1999). Ectopic overexpression of Drosophila lamin C is stage-specific lethal. Exp. Cell Res. 248, 350–357. Tegeder, I., Costigan, M., Griffin, R.S., Abele, A., Belfer, I., Schmidt, H., Ehnert, C., Nejim, J., Marian, C., Scholz, J., et al. (2006). GTP cyclohydrolase and
638 Cell 143, 628–638, November 12, 2010 ª2010 Elsevier Inc.
Tracey, I., and Mantyh, P.W. (2007). The cerebral signature for pain perception and its modulation. Neuron 55, 377–391. Tracey, W.D., Jr., Wilson, R.I., Laurent, G., and Benzer, S. (2003). painless, a Drosophila gene essential for nociception. Cell 113, 261–273. Treede, R.D., Kenshalo, D.R., Gracely, R.H., and Jones, A.K. (1999). The cortical representation of pain. Pain 79, 105–111. Venken, K.J., Carlson, J.W., Schulze, K.L., Pan, H., He, Y., Spokony, R., Wan, K.H., Koriabine, M., de Jong, P.J., White, K.P., et al. (2009). Versatile P[acman] BAC libraries for transgenesis studies in Drosophila melanogaster. Nat. Methods 6, 431–434. Xu, S.Y., Cang, C.L., Liu, X.F., Peng, Y.Q., Ye, Y.Z., Zhao, Z.Q., and Guo, A.K. (2006). Thermal nociception in adult Drosophila: behavioral characterization and the role of the painless gene. Genes Brain Behav. 5, 602–613.
Resource
Extensive In Vivo Metabolite-Protein Interactions Revealed by Large-Scale Systematic Analyses Xiyan Li,1,2 Tara A. Gianoulis,3 Kevin Y. Yip,4 Mark Gerstein,3,4,5 and Michael Snyder1,2,4,* 1Department
of Genetics, Stanford University School of Medicine, Stanford, CA 94305-5120, USA of Molecular, Cellular and Developmental Biology 3Program in Computational Biology and Bioinformatics 4Department of Molecular Biophysics and Biochemistry 5Department of Computer Science Yale University, New Haven CT 06520, USA *Correspondence:
[email protected] DOI 10.1016/j.cell.2010.09.048 2Department
SUMMARY
Natural small compounds comprise most cellular molecules and bind proteins as substrates, products, cofactors, and ligands. However, a large-scale investigation of in vivo protein-small metabolite interactions has not been performed. We developed a mass spectrometry assay for the large-scale identification of in vivo protein-hydrophobic small metabolite interactions in yeast and analyzed compounds that bind ergosterol biosynthetic proteins and protein kinases. Many of these proteins bind small metabolites; a few interactions were previously known, but the vast majority are new. Importantly, many key regulatory proteins such as protein kinases bind metabolites. Ergosterol was found to bind many proteins and may function as a general regulator. It is required for the activity of Ypk1, a mammalian AKT/ SGK kinase homolog. Our study defines potential key regulatory steps in lipid biosynthetic pathways and suggests that small metabolites may play a more general role as regulators of protein activity and function than previously appreciated. INTRODUCTION During the past decade, considerable effort has been devoted to analyzing biological networks, particularly protein-protein, expression, transcription factor binding, and even protein phosphorylation networks (reviewed in Snyder and Gallagher, 2009). These studies have provided a wealth of information for understanding protein function, which components work together, and the basic principles of regulatory network organization. In total numbers, small metabolites comprise the vast majority of cellular components, and like proteins, they are present in a broad range of cellular concentrations and participate in
a wide variety of biochemical and regulatory functions. They serve as metabolic components, cofactors for enzymes, forms of energy for biochemical reactions, and regulators of protein function (Fo¨rster et al., 2003). As regulators of protein function, metabolites can act globally to control many proteins or specifically target a limited number of proteins. Examples of the wide variety of small metabolite-protein associations include the binding of galactose to a sensor protein (Yano and Fukasawa, 1997), steroid hormones to transcription factors (Evans, 1988), and second messengers, including phospholipids, cyclic nucleotides, and arachidonic acids to specific cellular targets. In spite of their importance in mediating protein function and regulation, systematic approaches for analyzing in vivo interactions have not been performed. Such information is expected to be valuable not only for elucidating the biochemical activities and regulation of individual proteins, but also for assembling and understanding regulatory networks and connections between biological pathways. Furthermore, because metabolite levels can be adjusted by dietary intake of nutrients, understanding the regulation of cellular processes by metabolites has potential therapeutic value in correcting defects in biochemical pathways. Saccharomyces cerevisiae has served as an important model organism for many large-scale studies, including analysis of protein-protein interactions, phenotypes, genetic interactions, protein localization, gene expression, and transcription factor binding (reviewed in Horak and Snyder, 2002; Snyder and Gallagher, 2009). To date, more than 682 metabolic compounds have been identified in yeast (Fo¨rster et al., 2003), and many are known to be hydrophobic; 52% have a logP greater than methanol (Figure S1A). Many models of yeast metabolism have been generated and are capable of predicting key regulatory metabolic steps (Cascante and Marin, 2008; Herrga˚rd et al., 2008). Several previous studies have analyzed small molecules and small molecule-protein interactions in yeast. Metabolites have been profiled from yeast extracts using mass spectrometry (Allen et al., 2003), and limited studies to identify metabolites that bind proteins have been performed (Lee et al., 2007). Protein and small molecule microarrays have been used to discover several in vitro interactions (Beloqui et al., 2009; Cell 143, 639–650, November 12, 2010 ª2010 Elsevier Inc. 639
Sample
Control UPLC-APCI-MS TIC
LC
%
Lysis + IgG beads
bead
bead
**
time MS
* m/z 100100-1500
Wash
SDSPAGE
1) Methanol extraction IgG 2) SDS extraction Figure 1. Flowchart for the Identification of Small Metabolites Bound to Proteins Molecules bound to a strain expressing a protein of interest relative to a control stain are identified using the scheme presented. See also Figure S1 and Table S1.
Kuruvilla et al., 2002; Morozov et al., 2003; Zhu et al., 2001). Assays to examine small molecule-protein interactions have been developed (Maynard et al., 2009; Tagore et al., 2008); however, a systematic effort to identify the small metabolites that bind to large numbers of proteins in vivo has not been performed. Thus, the number and types of proteins that bind small molecules in the cell are not known. Such information is expected to both help inform potential regulatory interactions and elucidate the function and regulation of proteins and pathways. Here, we present a systematic large-scale investigation of the endogenous protein-metabolite interactome in yeast. We focused on the interaction of hydrophobic metabolites with components of the ergosterol biosynthesis pathway and protein kinases. Ergosterol biosynthetic enzymes were studied because we expected that these might bind hydrophobic metabolites, and protein kinases were chosen because of their importance in global regulation of protein function. We found that a large number of proteins bound to hydrophobic metabolites and described many new interactions. Further analysis has revealed that the yeast sterol, ergosterol, binds to many protein kinases, often with 1:1 stoichiometry, and is important for the activity of a highly conserved kinase, Ypk1, a member of the AKT/SGK family, and for the protein levels of Ssk22, a MAPKKK involved in osmotic responses. Ergosterol is the major sterol in yeast, and analogous to cholesterol in mammals, it is an abundant component of plasma membranes. Overall, our results demonstrate that a variety of small metabolite-protein interactions occur in eukaryotes and suggest an extensive role in the global regulation of protein activities. 640 Cell 143, 639–650, November 12, 2010 ª2010 Elsevier Inc.
RESULTS A Large-Scale Assay to Identify Hydrophobic Compounds Associated with Proteins: Application to Ergosterol Biosynthetic Enzymes We developed a sensitive and scalable method to systematically identify small metabolites bound to proteins using affinity protein purification and mass spectrometry (Figure 1). In brief, proteins tagged with an IgG-binding protein domain (Gelperin et al., 2005) were isolated from lysates using magnetic beads. After washing, the small metabolites were then extracted in methanol and analyzed using a reverse-phase C18 column-equipped ultra performance liquid chromatography (UPLC) column coupled to a quadrupole time-of-flight (Q-TOF) mass spectrometer. As a negative control, parallel experiments were performed using a yeast strain lacking the fusion protein (Y258). The metabolites significantly enriched in the presence of the fusion relative to the control strains and methanol solvent were identified. We focused on hydrophobic molecules because they are less likely to be removed from proteins during washes and can readily be detected by atmospheric pressure chemical ionization (APCI). The resultant mass spectra are comprised mostly of the protonated precursor ions ([M+H]+), which readily allow small metabolite identification. The mass spectrometry assay was first developed and standardized using 12 diverse compounds, including lipid-soluble vitamins from A, D, E, and K families, and two sterols (ergosterol and lanosterol), which allowed us to optimize the sensitive detection (200 femtomole in a mixture in profile scan mode) and separation of each of these compounds (Figure S1B). We also found that this method
detected more than 340 features in a methanol extract of yeast cells (an LC profile and list of peaks enriched in the extract relative to the solvent are in Figure S1C and Table S1), indicating that its scope is sufficiently broad to cover at least hundreds of metabolites in a single experiment. We first established the profiling assay using a group of 21 enzymes involved in ergosterol biosynthesis (Parks and Casey, 1995), whose known substrates and products, most of which are nonpolar hydrophobic molecules, are readily detectable by LC-APCI-Q-TOF (Figure 2A). The assays for the Erg proteins were performed using two to three separate protein preparations (biological replicates), each containing 0.5–5 picomoles of protein; for each sample preparation, four to six technical replicates were analyzed in parallel. The mass spectra data were analyzed in MarkerLynx or XCMS to identify molecules based on retention times and accurate molecular mass (see Experimental Procedures and Figure S1D for details). Because the background for the peak regions is very low, the correspondence between both technical and biological replicates was extremely high (mean of relative standard deviation = 7.5% ± 4% for three technical replicates; for biological replicates, see next section). We therefore used a stringent threshold for calling positive signals. Finally, protein purity was examined using SDSgel electrophoresis after metabolite extraction. In general, a single or major band of the expected size is present in the strain expressing the fusions relative to the negative control, although 14% of preparations contained more than one band indicative of either associated proteins or degradation products (Figure S2A and Figure S3A). Analysis of the liquid chromatography-mass spectrometry (LC-MS) results revealed that 16 of the 21 purified Erg proteins associated with small metabolites (Table 1). One example shown in Figure 2 contains several compounds associated with Erg6 that eluted from the LC column at retention time 10.84 min (Figure 2B); this peak contained three mass peaks, which were significantly lower (t test p value < 0.01, >10-fold signal/control) in the Y258 yeast control or methanol solvent (Figure 2C). These three peaks were identified as episterol (381.353 atomic mass unit [amu]), dimethylzymosterol (395.368 amu), and lanosterol (409.384 amu), respectively, by elemental composition analysis and hydrophobicity matching in retention time with known chemicals (see Figure S1B). Nine other proteins also bound small metabolites related to ergosterol biosynthesis. A large number of Erg and other proteins analyzed in this study that were just as abundant in the protein preparations as the metabolite-binding protein did not bind any metabolites (Figure S3D). For each of the 10 proteins that bound ergosterol-related metabolites, a specific set of associating molecules were observed, and similarly, each metabolite had a distinct profile. For example, (S)-2,3-epoxysqualene and 5a-cholesta-8,24-dien-3-one specifically associated only with Erg1 (Figure 2D), whereas lanosterol, ergosterol, and episterol consistently copurified with 5, 5, and 3 Erg proteins, respectively, above the control (Table 1). One Erg protein was found to bind known substrate (e.g., Erg3 bound episterol), and three other Erg proteins bound known products (e.g., [S]-2,3-epoxysqualene for Erg1), suggesting that these substrates and products are tightly associated with their metabolizing/biosynthetic enzymes.
Importantly, the majority of metabolite-protein interactions detected in our assay were new (e.g., dimethylzymosterol for Erg6) (Table 1). Of particular interest were lanosterol and ergosterol, which each bound five Erg proteins. Ergosterol bound its natural synthesizing enzyme Erg4 and four other enzymes that control the last five steps in ergosterol biosynthesis, starting from zymosterol, suggesting a multistep feedback regulatory mechanism (Figure S2B and Table 1). Ergosterol was not detected with the known ergosterol-regulated enzyme Hmg1, probably due to a low level of protein in the protein preparation (Figure S2A). Although low levels of protein may be an issue in several instances, it is unlikely to be a major problem overall, as the distribution of protein levels of the 37 metabolite-binding proteins identified in our entire study (Erg proteins and protein kinases) is similar to the distribution of the level of the majority of the 124 proteins analyzed (Figure S3D). Lanosterol also bound five enzymes; these enzymes are located at different points in the biosynthetic pathway. Of interest, unlike other erg mutant strains, yeast lacking Erg7, the enzyme that produces lanosterol, fails to grow in the absence of lanosterol (Karst and Lacroute, 1977), suggesting a major role for this lipid in yeast that might include modulation of protein function. Overall, these results raise the possibility that, in addition to the known inhibitory regulation of Hmg1 by ergosterol and of Erg13 by acetoacetyl-CoA (Parks and Casey, 1995), many steps in the ergosterol biosynthesis pathway may be regulated by biosynthetic products of the pathway (Figure 2A). New Metabolites Were Discovered to Bind Yeast Erg Proteins In addition to well-characterized sterols and other lipids that bound Erg proteins was an unexpected metabolite, pentaporphyrin I. Pentaporphyrin is a heme-related intermediate that may bind noncovalently to proteins due to the lack of peripheral methyl groups. It was detected in 7 of 21 Erg enzymes and was identified using a known standard (Figures 2E and 2F and Table 1). It is not clear whether this metabolite is ‘‘free’’ porphyrin or is derived from a bound form after loss of the central coordinated metal ion during LC-MS detection. Nevertheless, measurements of the binding affinity and stoichiometry revealed dissociation constants (Kd) of 8–34 mM and pentaporphyrin:protein stoichiometries of 1:1 (for Idi1 and Erg6) to 2:1 (for Erg27) (Figure 2G). The discovery that pentaporphyrin is associated with several ergosterol biosynthetic components may explain the observation that elimination of heme synthesis results in ergosterol auxotrophy (Parks and Casey, 1995) and suggests that pentaporphyrin-Erg protein interactions are important for protein function. Thus, our systematic analyses of small-protein interactions reveal new interactions important for protein function and further help to explain phenotypes for yeast strains lacking these different proteins. Large-Scale Analysis of Hydrophobic Small Metabolites Bound to Yeast Protein Kinases Protein kinases control cellular processes and regulate protein function at many levels. We next examined which of the yeast protein kinases bound hydrophobic metabolites. 103 protein kinases representing all functional branches in yeast (Hunter Cell 143, 639–650, November 12, 2010 ª2010 Elsevier Inc. 641
A
B
C
E
D
G
F
642 Cell 143, 639–650, November 12, 2010 ª2010 Elsevier Inc.
Table 1. Summary of Identified Small Metabolites Associated with Ergosterol Biosynthetic Proteins
RT (min)
Ergosterol
5,7,24(28)Ergostatrienol
Lanosterol
4,4-DimethylZymosterol
5a-Cholesta-8, 24-Dien-3-One
Episterol
(S)-2,3Epoxylsqualene
Pentaporphyrin
10.84
13.47
11.06
11.03
11.89
10.99
9.95
9.40
Mass (amu)
379.337
379.337
409.384
395.368
383.328
381.353
425.379
311.100
Element composition
C28H43*
C28H43*
C30H49*
C29H47*
C27H43O
C28H45*
C30H49O
C20H15N4
Erg1
P
Erg2
P
Erg3 Erg4 Erg5
P
S
P S
Erg6 Erg7
P
S
Erg9 Erg11
S
Erg24
P
Erg25
S
Erg26
P
Erg27 Hmg1 Idi1 Mvd1
S
Erg enzymes not bound to any known intermediates or pentaporphyrin: Erg 8, Erg10, Erg12, Erg13, Erg20. Gray denotes a small metabolite-protein association identified in this study. ‘‘P’’ and ‘‘S’’ indicate known product and substrate, respectively. Bound proteins for ergosterol and pentaporphyrin had an enrichment greater than 3-fold and 1000-fold, respectively, and the remainder were enriched greater than 5-fold. All bound metabolites had a t test (two-tailed unequal variance in relative to the negative control) p value less than 0.05; for ergosterol and pentaporphyrin p values of less than 0.01 and 1E10 was used. Asterisks indicate a dehydrated form of the expected formula.
and Plowman, 1997) (Figure 3A) were purified and analyzed as described above (Figure S3A). Two biological replicates were performed for all 103 kinases, with at least three technical replicates per biological replicate. A total of 95 metabolite peaks (background and specific peaks) were identified in both experiments, and these peaks were highly correlated in both retention time and exact mass from the two batches (R2 = 0.78) (Figure S3B). Using a stringent threshold, a total of 10 different
peaks were found to be associated with 21 protein kinases, but not with negative (Y258) or methanol-solvent controls (Table S2 and Figures S3C and S3E) or with many other abundant yeast proteins (Figure S3D). The specific peaks fell into two classes. One major class (11 of 14 analyzed) yielded reproducible peak intensity signals (R2 > 0.9), whereas another set (3 of 14) exhibited less correlation of peak intensities (R2 < 0.75). It is likely that the compounds bound to the highly correlated peaks
Figure 2. Identification of Small Metabolites Associated with Ergosterol Biosynthetic Proteins (A) An overview of ergosterol biosynthesis pathway. Substrates and products of the yeast ergosterol biosynthetic pathway (retrieved from MetaCyc with modification) are in blue, whereas protein enzymes are in black (included in this study) or gray (not included in this study). Known interactions are linked by a red curve labeled with Q for inhibitory effects. Interactions discovered in this study are indicated by green arrows from a metabolite to a binding protein. (B) LC plots of the small metabolites extracted from a protein (Erg6, red), the negative control (Y258, purple), and the methanol solvent (green), respectively. Base peak intensity (BPI, %) is plotted with retention time (in minutes) of corresponding mass spectra (shifted by 1% for clarity). Note that BPI peaks are composite, not a good indicator of the intensity of single molecular masses. The 100% BPI in counts is indicated on the graph. All traces were smoothed by the Savitzky-Golay method using two passes of window size of three scans. (C) Combined average mass spectra of the 10.80–11.20 min region in (B) (indicated by a blue block arrow). The masses of three Erg5-bound small metabolites are indicated along with their chemical identities. The x axis is the peak mass (amu); the y axis is the peak intensity (%). (D) Summary of the average peak intensity of two small metabolites listed in Table 1 extracted from each of the 21 ergosterol biosynthetic proteins (n = 5). Asterisk indicates statistical significance (two-tailed t test, p < 0.01) in comparison with the negative control Y258. Error bars = SEM. (E) An LC plot showing detection of pentaporphyrin I (311.100 amu at 9.40 min) from Mvd1 (red), but not from Y258 (purple) or methanol samples (green). Indent shows the LC of pure pentaporphyrin I. (F) Combined mass spectra of the LC peak region in (E). Color labels are as in (E). The x axis is shifted by 0.05 amu for clarity. The indent profile shows the mass spectrum of pure pentaporphyrin. (G) In vitro binding curves of pentaporphyrin and Erg proteins. Each binding curve was subject to fitting comparison (p < 0.01) to a saturable binding curve (specific) or a straight line (nonspecific). Binding constants Kd, Bmax, and curve-fitting R2 are indicated (in mM) on each graph. Stoichiometry (metabolite:protein) is also indicated. Error bars = SD; n = 2. See also Figure S2
Cell 143, 639–650, November 12, 2010 ª2010 Elsevier Inc. 643
B
PRR2
HO
% 0
MeOH
100
%
TOF MS AP+ BPI 3.42e3 8.00
10.99
80.5
% 3.31e4
%
%
[C28H44O+H] +
0 Ergosterol 100 379.339
0 397.35
3.85e3
0 Ypk1 100 379.339 0
TOF MS AP+ 583
100
397.348
[C28H42-H2O+H ] +
%
681
379.339
Y258
MS
%
% %
TOF MS AP+ 583
0
11.01
%
379.339
MeOH
TOF MS AP+ 583
100
%
M E C C V P TK 1 S 2 SSK PK 15 SSK 2 P2 2 2 SN F4 VPS34
1 B U 3 B TK C
STE
2.69e3
Time (min)
C Kin4
379.34 0.01Da Ergocalciferol
0
12.00
10.00
Time (min)
100
10.91 TOF MS AP+ 100
0 Ypk1 100
STE
0
1 PSK 2 PSK
LC
0 Ergosterol 100
TOF MS AP+ BPI 3.42e3
TA
TOR1 TOR2 3 L 82 A G SIP
Ergocalciferol 100
100
K2
P TE KP L1 1
D
%
N1
1 DU
C SKLA4 M1
K1
SM
K1 3 2 BCU D B 1 K1 N 1 CAIP1 SL LK S A 1 F
2 C SLTL161 YK
AL
10.98
TOF MS AP+ BPI 3.42e3
Y258
CK1
K1 SAOS3 T
R SK IO2 Y RI 1 O1
x4
0
L1 SN F
S1 SPTE20 S CDC15 STE11 3MKK1 S HR PBTE7MKK2 R2 YC S K3 YCK2 5 EL E 1 Y M NV7 CK2 1
1 W ISR L150 YP IRE1
1
CAMK
K1 N4 ARRK1 G I CC4 P K L1 K A K1 R1 CMMK2 PR C MEK1 K IN 1 Y K L KI N 2 171W
CMGC
Kin4
100
C
KIN
IM K N E 2 YAK1 S1 YGK3 MCK1 1 GCN2 1 M I R RK1 M PHO85 8 2 C D CSN3 S V1 8 SGI N 2 K S1 KSUS3 F G1 C TK
K1 2 5 HRDL0 Y
HS
RAD 53 SWE 1 CDC RC 7 C D RC K1 C 5 K2
C1 PK A 1 TD H3
PK
AGC
A MPTG S1 1
IPL1 RIM1 5
9
YP YPKK1 2 P FPK1 PKKH K IN 8 H21 2 YBR028C DBF2 DBF20 K1 CB KSP1
K1 1 V H SS1 SK K1 2 PTTK P
A1 2 CKKA C
H SC K2 TP
TPK TPK 1 3
CH
CAMK
L5 H A Q8 KK
‘Yeast’
SAT4 K YP I N L14 41
C
A
425 m/z (amu)
0
385.219 250
275
300
325
350
375
m/z (amu)
400
425
450
Figure 3. S. cerevisiae Protein Kinase-Small Metabolite Interaction (A) A total of 103 of 129 kinases were analyzed in this study. Kinases not tested are indicated in gray. The 21 kinases that bound small metabolites are in red. (B) An LC plot of the small metabolites extracted from a kinase (Kin4, red), the negative control (Y258, purple), and the methanol (green). The focused retention time region (in minutes, zoomed in 43) is indicated above the trace. Graph labels are as in Figure 2B. (C) Combined average mass spectra of the 10.9–11.1 min region in (B). Graph label is as in Figure 2C (n = 3 for Kin4; n = 9 for Y258 and methanol). The peak corresponding to ergosterol is marked by an asterisk. (D) An example showing identification of a bound metabolite as ergosterol. The mass spectra of pure ergosterol, pure ergocalciferol, and one of the small metabolites extracted from protein kinase Ypk1 are shown on the right and their respective UPLC on the left. The elemental composition is indicated along with respective mass peaks. Graph label is as in (B) and (C). See also Figure S3 and Table S2.
represent strong steady-state interactions, and those with differing levels of interacting metabolites interact transiently and/or weakly (Morozov et al., 2003). An example of specific binding is shown in Figures 3B and 3C for Kin4, a protein kinase regulating mitotic exit (Caydasi and Pereira, 2009). The methanol extract from Kin4 contained a singly charged ion of 379.337 amu at retention time 10.98 min. Analysis of standards revealed that this compound is ergosterol (dehydrated state) rather than ergocalciferol, a compound of identical molecular mass (Figure 3D). The 10 specific binding metabolites represent different lipids and sterols; however, one kinase, Ste20, was found to bind pentaporphyrin, albeit at reduced levels relative to the ergosterol 644 Cell 143, 639–650, November 12, 2010 ª2010 Elsevier Inc.
enzymes, raising the possibility that Ste20 is associated with this molecule. The compound identified most often among different kinases was ergosterol, which was associated with 15 different protein kinases. None of these proteins was previously known to bind ergosterol, and, except for Yck2, none were known to be membrane-associated. KEGG pathways analysis reveals an enrichment of ergosterol-bound kinases in metabolism of inositol phosphate, starch and sucrose, nicotinate and nicotinamide, and sphingoglycolipids (Table S3). The different ergosterol-binding kinases belong to different families of protein kinases (Figure 3A), suggesting a potential regulatory role of ergosterol in the regulation of many types of protein kinases and many different aspects of yeast biology.
A
B
D
C
Figure 4. Detailed Analysis of Several Kinase-Small Metabolite Interactions (A) In vitro binding analysis of ergosterol and several protein kinases. The curve-fitting was done in GraphPad Prism 5. Error bars = SEM; n = 3. Statistical comparison of curve fitting between a straight line for nonspecific binding (null) versus one-site specific binding was used to determine specific binding pattern (p < 0.05). The R2 (unweighted) was 0.918, 0.908, and 0.933 for Hal5, Rck2, and Ypk1, respectively. The ergosterol-binding characteristics of their protein kinases are listed below. (B) Protein kinase activity of Ypk1 is stimulated by the addition of ergosterol. Ypk1 protein was purified from wild-type (BY4741) cells grown in the presence or absence of 2 mM ergosterol during galactose induction of protein expression or from ergosterol-deficient yeast (erg4D). Equal amounts of purified protein were tested in each assay. The relative activity was determined using a Sgk1-specific kinase assay. Error bars = SEM; n = 4. (C) Levels of Ssk22 and Ypk1 in yeast. (a) Ssk22 and Ypk1 were purified from equal amounts of wild-type (BY4741) and ergosterol-lacking mutant (erg4D) cells with or without 0.4 mM ergosterol. In three independent experiments, Ssk22 cannot be detected in erg4D. (b) Immunoblot of Ssk22 and Ypk1 from yeast cell lysates for 7 hr after galactose induction. Proteins were probed with rabbit IgG (1:10,000 dilution of 10 mg/ml stock). Equal amounts of protein were loaded; erg4D strains produce less protein, as indicated by relative abundance listed below (percentage of wild-type). (D) Cell growth (absorption at 600 nm) is affected by the ergosterol-repressing drug fluconazole in mutants of ergosterol-binding protein kinases. Error bars = SEM; n = 8. Dotted lines and gray legends are mutants of protein kinases that did not bind ergosterol in this study. See also Figure S4.
Ergosterol-Protein Kinase Interactions Have Binding Affinities and Stoichiometries in Ranges Expected for Physiological Relevance To determine whether the binding coefficients observed are likely to be physiologically relevant, the stoichiometry and affinity of the ergosterol-protein kinase association was determined for three kinases, Ypk1, Hal5, and Rck2, along with three nonbinding controls, Atg1, Psk2, and denatured Ypk1, using an in vitro binding assay that we developed (Figure 4A). For a fixed amount of the ergosterol-bound protein kinases, ergosterol binding exhibited a saturable curve over an increasing concentration of free ergosterol, indicating specific binding. In contrast, the binding curve was close to linear for the same amount of denatured protein or control proteins
(Figure S4), indicating nonspecific adsorption. The Kd for each ergosterol-binding kinase was between 4.7 and 17.9 mM (Figure 4A), figures significantly lower than the endogenous concentration of 4.8 mM for ergosterol in yeast, assuming a uniform distribution throughout the cell (Ejsing et al., 2009). Although neither the kinase nor ergosterol is likely to be uniform in its cellular distribution, the 4.7–17.9 mM binding constant found for Ypk1 and the other kinases is well within a plausible range for biological relevance. The binding ratio of ergosterol to protein was found to be close to 1 (0.98–1.1, 95% confidence interval) for each protein kinase, suggesting that one small metabolite binds to one protein (Figure 4A). This ratio is consistent with an in vivo biological role for ergosterol in regulating kinase activity. Cell 143, 639–650, November 12, 2010 ª2010 Elsevier Inc. 645
Ergosterol Regulates the Activity of a Highly Conserved Kinase, Ypk1, and Influences the Ssk22 Levels To determine whether ergosterol binding is important for kinase function, we tested the effect of ergosterol on Ypk1 activity using in vitro kinase assays (Figure 4B). Ypk1 is a yeast homolog of the mammalian SGK/AKT protein kinases, which are involved in many important cellular processes and human disease (Brazil and Hemmings, 2001); yeast Ypk1 has been implicated in receptor-mediated endocytosis and sphingolipid-mediated signaling (Jacquier and Schneiter, 2010). Ypk1 was purified from wild-type cells grown in the absence and presence of 0.4 mM ergosterol and was tested for stimulation of in vitro kinase activity in the presence of increasing concentrations of ergosterol. Ypk1 activity from cells grown in the absence of ergosterol is significantly (and reproducibly) elevated in the presence of increasing amounts of ergosterol (Figure 4B). Ypk1 activity was even higher when cells were grown in the presence of ergosterol and could be stimulated to a similar extent. Because Ypk1 was isolated from wild-type cells that contain ergosterol, it might already contain bound ergosterol. We therefore purified Ypk1 protein from an erg4D strain that lacks ergosterol (Parks and Casey, 1995). Ypk1 protein levels were similar to preparations from wild-type cells (Figure 4Ca); however, Ypk1 activity was very low in erg4D strains (at least 5-fold lower) relative to wild-type cells and could not be stimulated (Figure 4B). These results demonstrate that ergosterol stimulates Ypk1 kinase activity. Because Ypk1 isolated from erg4D strains was low and could not be stimulated, it is likely that some ergosterol must be present during Ypk1 synthesis and activation. Overall, these results demonstrate that ergosterol is critical for Ypk1 activity. We also attempted to analyze the activity of Ssk22 (a kinase involved in osmosensing [Posas et al., 1996]) in wild-type and erg4D cells. In multiple independent experiments, we found that we could not purify Ssk22 from erg4D strains lacking ergosterol (Figure 4Ca). Addition of exogenous ergosterol to the medium restored Ssk22 levels in the mutant strain. To explore this further, Ssk22 levels were examined for up to 7 hr after expression was induced from a GAL promoter. As shown in Figure 4Cb, copious amounts of Ssk22 protein are detected in wild-type cells, but levels are substantially reduced (6- to 20fold) in erg4D cells (Figure 4Cb). Although the levels of Ssk22 were too low to measure kinase activity, our results demonstrate that ergosterol is important for maintaining Ssk22 protein levels in wild-type yeast. Growth of Strains Deleted for Ypk1 and Other Ergosterol-Binding Proteins Is Affected by Ergosterol Inhibitors We next determined whether ergosterol levels are important for the growth of yeast strains lacking ergosterol-binding kinases. Yeast strains lacking Ypk1 were shown previously to be sensitive to nystatin and fluconazole, two inhibitors of ergosterol biosynthesis (Gupta et al., 2003; Hillenmeyer et al., 2008). Six strains deleted for different ergosterol-binding kinases, as well as four strains deleted for kinases not found to bind ergosterol, were grown in the presence of different concentrations of fluconazole, and cell density was determined. As shown in Figure 4D, growth 646 Cell 143, 639–650, November 12, 2010 ª2010 Elsevier Inc.
of ypk1D cells was inhibited by fluconazole, whereas ssk22D were resistant to fluconazole. Similarly, ypk1D cells were sensitive to nystatin (data not shown). The mutants of four other ergosterol-binding protein kinases behaved similarly to the wild-type cells and strains lacking nonergosterol binding kinases. These results indicate that genetic interactions are evident between ergosterol-bound kinases and the ergosterol pathway. An Integrated Global Small Metabolite-Protein Network To better view how our results might be connected with other regulatory interactions, we next integrated the small metabolite-protein binding results with protein-protein interaction data, genetic interaction data, and metabolite networks constructed by others. Our results suggest a highly connected network of interactions in which the small metabolites add an extra dimension of regulatory information. We found extensive interactions between the ergosterol-bound kinases, the ergosterol biosynthesis proteins, and a wide variety of other cellular components. Simplification of the network to only those interactions that directly interact with the Erg pathway and ergosterol-bound kinases reveals a bipartite pattern (Figure 5A). Of interest, the kinases are connected to their interacting partners, whereas ergosterol pathway proteins interact through genetic and phenotypic interactions. We suggest that Erg pathway components often operate in the same pathways as the affected kinases and/or small metabolite products of the pathway affect kinase regulators and/or substrates. Overall, our results demonstrate close functional connections between ergosterol biosynthetic pathway components and ergosterolbound kinases. Within the overall interaction network is a variety of interesting interactions. For example, Erg20, an essential enzyme for both isoprenoid and ergosterol (Daum et al., 1998), is indirectly phosphorylated by ergosterol-bound protein kinases Sat4 via Tpk1, whereas five other enzymes, Erg1, Erg4, Erg6, Erg7, and Erg26, are transcriptionally regulated by five proteins (in the middle circle, Figure 5A). In addition, 13 of 21 enzymes (62%) in the ergosterol pathway and three ergosterol-binding kinases (Ypk1, Yak1, and Mck1) have physical interactions with the ubiquitin Ubi4. This figure is statistically significant (p value = 1.5e5 by Fisher’s exact test), as only 18% of all yeast proteins have physical interaction with Ubi4 (Figure 5A, bottom). Perhaps many components involved in ergosterol biosynthesis are modified and/or degraded by the Ubi4-mediated ubiquitination pathway. We next conducted gene function enrichment analyses for the 137 yeast genes known to interact physically, genetically, or phenotypically with both 1 of the 21 ergosterol biosynthetic proteins and 1 of the 15 ergosterol-bound protein kinases (Figure 5A, top). Several categories of cell division and growth are particularly overrepresented, such as cell cycle, stress responses, transcription, and general metabolism of lipids, vitamins, and carbohydrates (Table S3 for KEGG pathway and Figure 5B for gene ontology). These results further suggest that ergosterol can act through modulation of protein kinase activation as a general regulator in coordinating various cellular biological processes.
A
DISCUSSION
MOB1 BEM2 CLB3GAS1GIM4 CDC37 MNN4 RAD52 GET2PLP2 SLX8 AFT1 RSC8 YPS1 CSF1 SHR3 YPT6 ERP2 HFI1 ERV25 RPC40 CLA4 RMI1 UFD4 ISC1 PDR3 MET18 ARP4 RAD27 ESA1 SEC28 RPN4 SHP1
UPC2
CTF8
PIL1
ABF1
SAC1
SWI4
TEX1
RTT109
GUP1
CSM3
VPS1 BST1 FUS3 YAF9 POP2 MMS22 UBX7 TAF12 RIM101 ARO1
ERG7
SUT1 RPB3
ERG8 ERG9
ERG6
OPI3
SSK22
HMG1
PAF1
YAK1
GZF3
SCH9
ERG5
MED8 INP54
YCK2
IDI1
RAD3 SEC14
SAT4
ERG4
RVB2 PDR5 MNN1
YPK1
MVD1
TLG2 PMT1
ERG3
SWI5
RCK2
ORM2 APM3
ERG1
GIM3
CBK1 RTG3
ERG27
SSD1 APS3
PRR2
ACE2
ERG10
APL6 HAL5
GET1
ERG26
ARV1 BRE1
PRR1
ERG11
SLA1 SLT2 KCC4
ERG25
RTN2 SAS10
MCK1
ERG12
AVT4 CAX4
KIN4
KIN82
ERG24
HCM1 HEK2
ERG20
SSA1
ERG13 ERG2
TPK1 BUL1 SWI6 INO2
(15)
(21)
SEC2 SNL1 MGA1
SPF1
CDC13
SCS7
MBP1
RGP1
AGE1
SUR4
SKN7
YBR159W
LEO1 RIC1 YAP6 INO4 YNL010W HSC82 MST27 STP1 PRP6 STE20 UBI4 PTK2 TEC1 PPH22 YPC1 TOR1 HSP82 NCS2 NAP1 STE12 PHO85 MSN2 BRE2 MSN4PCL9 FAR3SLX5 SNF1TOS3SOH1BRE5HAP4 PHO80
(137) ERG6
ERG7 ERG8 ERG9
ERG5 SSK22
YCK2
TPK1
HMG1
YAP6
IDI1 ERG4 MVD1
ACE2
ERG1
MGA1
HAL5
SAT4
ERG3 ERG27 ERG10 ERG26
MBP1
MCK1
PRR2
GZF3
ERG11
ERG25
ERG12 ERG24 ERG20
STP22STT4TEC1TOR2TPS2TRM3 TRP1 STE23STE5 UBP15 STE20 UBR1 STB2 UBX5 SSA1 UBX7 SRO7 UFD4 SPT16 UME6 SOG2 URA8 SNF1 VMA5 SMF1 VPS1 SLM1 VPS27
SLA1
ERG13 ERG2
YBR159W
SIT4
YCF1
SIP5
YCK1
SHE3 YDR266C SEC28 YGR012W SEC2 YGR126W SAC6 YGR130C RVS167 YLR241W RTN2 YMR115W RPN3 YMR295C SSK22
ERG6
RPN1 YAK1
ERG9
YNL045W
IDI1
ROD1
SCH9
YNL217W YCK2
ERG5
RET1 YPK2
SAT4
MVD1
PTP2 YPR091C PIL1
YPK1
ERG4
ZRC1 PHO13 ADP1
RCK2
ERG1
PGA2 AKR1 CBK1
PFK26
ERG3
ALY2 PFK2
PRR2
AMD1
UBI4
PDX3 ARP2
HAL5
PDR5 ATG9
PRR1
OPI1 AVT1 OLA1
KCC4
ERG11 ERG27 ERG12 ERG26
AVT3 MCK1
ERG13
NSR1 KIN82
BMH2
KIN4
MYO2 CAT8
(13)
MSN2 CBF5 MIH1
(15)
CDC12 MCM2 CDC28 KSP1 CDC39
KEM1
CHS1
IST2
CMK1
IRC6
CYC7
INP54
DCS1 HUL4 DCS2 HSP42 DNM1 HSL1 DRE2 HSF1 ECM10 HRT1 ERP2 HOG1 ERV25 HEF3 FOL2 HAS1 FZO1 GZF3GUS1 GCD7 GUP1GSY2GRE3GPH1GLY1GDB1
(119) B endoplasmic reticulum
Golgi apparatus
cytoplasm cell cycle
nucleus
transcription
CC membrane endomembrane system
response to stress signal transduction
BP GO
transport
trans cripti
vesicle−mediated transport protein modification process MF
y ctivit
tor a
gula
on re
DNA binding transferase activity protein kinase activity
1.00E-2
