The Epigenetic Modifier Ubiquitin-specific Protease 22 (USP22) Regulates Embryonic Stem Cell Differentiation via Transcriptional Repression of Sex-determining Region Y-box 2 (SOX2)*

Background: Ubiquitin-specific protease 22 (USP22) is a deubiquitylating enzyme with established biological functions in cancer cells. Results: USP22 drives differentiation of embryonic stem cells (ESCs) and represses sex-determining region Y-box 2 (SOX2) transcription. Conclusion: USP22 is induced during ESC differentiation to repress SOX2 transcription. Significance: Understanding the epigenetic programs that control changes in gene expression during the transition from self-renewal to differentiation. Pluripotent embryonic stem cells (ESCs) undergo self-renewal until stimulated to differentiate along specific lineage pathways. Many of the transcriptional networks that drive reprogramming of a self-renewing ESC to a differentiating cell have been identified. However, fundamental questions remain unanswered about the epigenetic programs that control these changes in gene expression. Here we report that the histone ubiquitin hydrolase ubiquitin-specific protease 22 (USP22) is a critical epigenetic modifier that controls this transition from self-renewal to differentiation. USP22 is induced as ESCs differentiate and is necessary for differentiation into all three germ layers. We further report that USP22 is a transcriptional repressor of the locus encoding the core pluripotency factor sex-determining region Y-box 2 (SOX2) in ESCs, and this repression is required for efficient differentiation. USP22 occupies the Sox2 promoter and hydrolyzes monoubiquitin from ubiquitylated histone H2B and blocks transcription of the Sox2 locus. Our study reveals an epigenetic mechanism that represses the core pluripotency transcriptional network in ESCs, allowing ESCs to transition from a state of self-renewal into lineage-specific differentiation programs.


Pluripotent embryonic stem cells (ESCs)
have the ability to differentiate into any of the over 200 distinct cell types present in adult mammals. This pluripotent state is regulated by a set of core transcription factors, and combinations of these factors, including SOX2, OCT4, LIN28, NANOG, KLF4, and MYC, can reprogram somatic cells into induced pluripotent stem cells (1)(2)(3)(4). The balance between the core pluripotency factors OCT4, SOX2, and NANOG maintains pluripotency, with loss of OCT4 or NANOG promoting neuroectoderm differentiation and loss of SOX2 promoting mesendoderm differentiation (5,6). In addition to this transcriptional circuit, a second level of regulation exists to control the pluripotent state, in the form of epigenetic modifications deposited by either PcG (polycomb group) complexes (7) or the MLL (mixed lineage leukemia)trithorax complex (8). These complexes control gene expression by depositing the repressive histone H3 lysine 27 trimethyl mark or the activating histone H3 lysine 4 trimethyl mark, respectively. Furthermore, many of the loci in ESCs encoding genes involved in early differentiation programs contain both of these marks in a configuration termed "bivalent." These bivalent genes are transcriptionally repressed but are poised for rapid activation upon receipt of a differentiation stimulus (9 -12). Both histone modifications that comprise the bivalent mark are tied to histone ubiquitylation; trimethylated histone H3 lysine 4 is deposited at gene loci where histone H2B has been ubiquitylated (uH2B) by RNF20/40 (13)(14)(15)(16), and histone H3 lysine 27 methylation by PRC2 can trigger PRC1 (polycomb repressive complex 1) to ubiquitylate histone H2A (uH2A) via its RING1B/RNF2 subunit (8,17,18). Ultimately, these histone modification signaling cascades provide the functional link between histone ubiquitylation status and the transcriptional programs that dictate cell fate.
The remarkable plasticity that allows reprogramming of cell identity requires dynamic modulation of the epigenetic landscape, including patterns of histone ubiquitylation. Unlike many other epigenetic marks, factors regulating histone ubiquitylation remain poorly understood, with even the ubiquityla-tion sites themselves not fully characterized (19). Our group and others recently identified USP22, a cysteine protease that can hydrolyze monoubiquitin from either uH2A, to antagonize PcG, or from uH2B, in order to regulate MLL-trithorax-mediated trimethylation of histone H3 lysine 4 (20 -24). By removing monoubiquitin from the repressive uH2A or the activating uH2B, Usp22 might act as either a transcriptional activator or repressor. USP22 is a component of a deubiquitylase module within SAGA, a multiprotein transcriptional cofactor that also has acetyltransferase activity directed against histones and other substrates. In addition to its interaction with SAGA, USP22 binds and deubiquitylates the class III NAD-dependent deacetylase SIRT1 (25)(26)(27).
Along with its ability to regulate histone ubiquitylation patterns linked to PcG and trithorax-MLL, a variety of other lines of evidence suggested that Usp22 might play a role in stem cell function. First, Usp22 is part of an 11-gene cancer stem cell signature, where it regulates aggressive phenotypes, such as metastatic potential and resistance to therapy (26,28). Second, USP22 is required for embryonic development in mice (26,29). Third, the Drosophila USP22 ortholog Nonstop is required for proper neuronal development and the tissue-specific expression of SAGA-bound genes (30,31). Fourth, consistent with a role in regulating epigenetic patterns linked to pluripotency and differentiation, the USP22 locus is actively transcribed in both human ESCs and induced pluripotent stem cells (32). Fifth, the activating histone H3 lysine 4 trimethyl epigenetic mark is deposited along the USP22 promoter, which is also occupied by the core pluripotency factor KLF4 in both cell types (32). Finally, USP22 is an essential co-factor for the core pluripotency factor MYC and is required for transcription of MYC target genes (22). Collectively, these aspects of USP22 expression and function prompted the hypothesis that this epigenetic modifier might participate in controlling transcriptional programs that dictate stem cell identity.
Based on the rationale outlined above, studies were conducted to define the potential role of USP22 in ESC function and the maintenance of pluripotency. This analysis revealed that USP22 is both necessary and sufficient for the proper differentiation of ESC into all three germ layers. USP22 represses Sox2 transcription, and epistasis experiments suggest that Sox2 derepression may be responsible for the effects of USP22 depletion, because blocking the increase in SOX2 reversed the USP22 phenotype. Mechanistically, USP22 was found to directly occupy the Sox2 locus, where it controls the relative level of histone H2B ubiquitylation. USP22-mediated changes in H2B ubiquitylation at Sox2 probably explain its effects on Sox2 transcription and pluripotency because we find that RNF20, the E3 ligase responsible for H2B ubiquitylation, is essential for SOX2 expression in ESCs.

EXPERIMENTAL PROCEDURES
Cell Lines, Proliferation, and Differentiation Assays-R1 mouse embryonic stem cells were obtained from ATCC. E14 mouse embryonic stem cells were a gift from Carlisle Landel. Mouse ESCs were maintained in feeder-free conditions on gelatin-coated plates in 20% defined FBS-DMEM supplemented with 1% L-glutamine, 1% HEPES, 1% non-essential amino acids, 0.001% ␤-mercaptoethanol, and fresh LIF. The MEK inhibitor PD0325901 (1 M) and GSK3 inhibitor CH99021 (3 M), together known as 2i, were added fresh along with LIF. H9 human embryonic stem cells were obtained from WiCell and were grown on Matrigel-coated plates in mTeSR1 (STEMCELL Technologies). Cell cycle analysis was performed with a 1-h pulse of BrdU followed by propidium iodide staining, as described previously (33). Differentiation was achieved by embryoid body formation in the medium described without 2i/LIF for mouse ESCs. Human embryoid bodies (EBs) were grown in STEMdiff APEL medium (STEMCELL Technologies). Retinoic acid was used at 2 M and added to regular medium without 2i/LIF following incubation in N2B27 medium as described (5). Alkaline phosphatase expression was detected on cells fixed with 4% paraformaldehyde using an alkaline phosphatase detection kit (Millipore) or by colorimetric assay from whole cell lysate (Cell Biolabs). Optic density was measured at 405 nm and normalized to total protein concentration in the lysate (as measured by a BCA assay).
mRNA Analysis, shRNA Treatment, Ectopic Protein Expression, siRNA Treatment, and Western Blotting-mRNA was analyzed by quantitative RT-PCR as described (34). Primer sequences are provided in supplemental Table 2. In all cases, mRNA levels were normalized to ELF1 levels, and error bars represent S.D. values of technical triplicates. Lentiviral shRNA plasmids corresponding to USP22, SOX2, and RNF20 were obtained from the TRC library (Sigma and Openbiosystems). siRNA was obtained from the Silencer Select collection (Invitrogen) and was transfected into cells using Lipofectamine 2000 (Invitrogen). FLAG-USP22 was cloned into pLenti6.3/TO/V5-DEST using the T-REx TM system (Invitrogen). Western blots were performed as described (34), using antibodies to USP22 (Novus), ACTIN (Sigma), SOX2 (Santa Cruz Biotechnology, Inc., Santa Cruz, CA), OCT4 (Santa Cruz Biotechnology, Inc.), and RNF20 (Novus). ACTIN protein levels are used as a loading control.
Microarray Analysis-mRNA was harvested as described and repurified using the RNAeasy column (Qiagen). RNA quality assessment was performed on an Agilent 2100 bioanalyzer (Agilent, Palo Alto, CA). Amplification of cDNA was performed using the Ovation Pico WTA-system V2 RNA amplification system (NuGen Technologies, Inc.). Purified cDNA was then amplified with ribo-SPIA enzyme and SPIA DNA/RNA primers (NuGEN Technologies, Inc.). Sense target cDNA was fragmented and chemically labeled with biotin to generate biotinylated ST-cDNA using FL-Ovation cDNA biotin module V2 (NuGen Technologies, Inc.).
Mouse gene 1.0 ST array (Affymetrix, Santa Clara, CA) was hybridized with fragmented and biotin-labeled target (2.5 mg) in 110 l of hybridization mixture. Chips were scanned on an Affymetrix Gene Chip Scanner 3000, using Command Console software. Background correction and normalization were done using Iterative Plier 16 with GeneSpring V12.0 software (Agilent, Palo Alto, CA). A 1.5-fold (p Ͻ 0.12) differentially expressed gene list was generated. The differentially expressed gene list was loaded into Ingenuity Pathway Analysis (IPA) version 8.0 software (Ingenuity Systems, Inc.) to perform biological network and functional analyses.
Chromatin Immunoprecipitation-ChIP analysis was conducted as described previously (34). shRNA viral infection (targeting USP22 or luciferase) was performed 7 days prior to cross-linking and lysis to capture the kinetic induction of SOX2 mRNA. Antibodies used in ChIP experiments included USP22 (Novus), H2B (Santa Cruz Biotechnology, Inc.), and uH2B (Milipore). All immunoprecipitations were normalized to IgG (Santa Cruz Biotechnology, Inc.) levels. After immunoprecipitation, precipitated DNA fragments were analyzed by qPCR. Primer sequences are provided in supplemental Table 2. Error bars represent standard deviations of technical triplicates.

RESULTS
USP22 Is Induced as ESCs Differentiate-Expression of the epigenetic regulator USP22 is altered as cells transition from a self-renewing pluripotent ESC to a state of differentiation. For example, USP22 mRNA is induced between days 4.5 and 6.5 of mouse embryogenesis (29). Furthermore, in genomewide assays in multiple cells lines, USP22 mRNA expression increases as ESCs differentiate in EBs (35). EBs are spherical aggregates of ESCs grown in suspension. While in these aggregates, ESCs differentiate into cells representing all three embryonic germ layers. Our studies validate this induction of USP22 as ESCs differentiate either through EB formation or by treatment with retinoic acid (RA) (Fig. 1). Three ESC lines from separate genetic backgrounds were used in these experiments: E14 and R1 mouse ESCs and H9 human ESCs. Pluripotent ESCs were harvested at day 0 from 2i/LIF-conditioned medium, and EBs were then collected at days 3, 6, 9, and 12 following withdrawal from 2i/LIF medium. As anticipated from previous studies (36,37), Oct4 protein levels rapidly decrease as E14 and R1 mouse ESCs differentiate (Fig. 1, A and B, respectively). SOX2 protein levels display a similar pattern but rebound slightly at day 9, presumably due to the previously reported role of Sox2 in neuronal differentiation (38 -40). USP22 mRNA levels increase in both cell lines after 3 days in EBs and then decrease again by day 6. USP22 protein levels follow the mRNA, increasing between days 3 and 9 of EB-induced differentiation and decreasing again by day 12 (Fig. 1, A and B). USP22 and SOX2 protein levels were quantified and normalized to ACTIN protein levels. This analysis revealed that the magnitude of change in the protein levels of USP22 and SOX2 inversely correlate (e.g. in Fig. 1B between days 0 and 3, USP22 protein levels increase ϳ2-fold, and SOX2 protein levels decrease ϳ2-fold). These findings were conserved in H9 human ESCs as well (data not shown).
In order to assess changes in USP22 levels in response to a lineage-specific differentiation signal, RA was used to induce neuronal differentiation in mouse ESCs (5,41,42). Cells were withdrawn from 2i/LIF and differentiated with RA over the course of 5 days. Results of this experiment demonstrated that USP22 protein increases after 3 days of RA-induced differentiation (Fig. 1C), displaying a pattern similar to that observed in EB differentiation assays. As in the embryoid body experiments above, this change in USP22 expression coincides with decreased SOX2 protein levels (Fig. 1C). This inverse correlation between USP22 and pluripotency factors, such as SOX2, during either spontaneous or lineage-directed differentiation, FIGURE 1. USP22 is induced as ESCs differentiate. E14 (A) and R1 (B) mouse ESCs grown in 2i/LIF medium were differentiated in 400-cell hanging drop EBs over the course of 12 days. Pluripotent ESCs were harvested at day 0 from 2i/LIF medium, and EBs were harvested at days 3, 6, 9, and 12 from medium without 2i/LIF. USP22, SOX2, and OCT4 protein expression was analyzed by Western blotting. qRT-PCR was used to define levels of USP22 mRNA during EB-induced differentiation. *, p Ͻ 0.05. C, E14 ESCs were withdrawn from 2i/LIF-conditioned medium and treated with 0.2 M RA. Whole cell lysate was harvested before the removal of 2i/LIF and following 2, 3, 4, and 5 days of RA treatment, and USP22, SOX2, and ACTIN protein levels were analyzed by Western blotting (top panels). USP22 and SOX2 protein levels were quantified and normalized to ACTIN (bottom). Error bars, S.D.
suggested the possibility of a conserved role for USP22 in ESC differentiation.
USP22 Is Necessary and Sufficient for ESC Differentiation-Ubiquitylated histones H2A and H2B, the best characterized substrates of USP22, are both implicated in transcriptional reg-ulatory processes (20 -22, 43-45). In order to assess any potential role for USP22 in ESC differentiation, USP22 was depleted from E14 mouse ESCs ( Fig. 2A). Because shRNA 2 provided a more efficient knockdown of USP22 in this particular experiment, it was utilized in the subsequent differentiation assay FIGURE 2. USP22 is required and sufficient for differentiation of ESCs. A, USP22 knockdown was confirmed by Western blotting (top left). shUsp22 and shLuc E14 cells were stained to detect AP expression (pink). Pictures were taken at ϫ10 zoom; the top right picture at ϫ4 zoom shows clonal density of cells grown in 2i/LIF-conditioned medium. The middle panels show typical colonies of cells following EB-induced differentiation. The bottom panels represent examples of less common and more extreme cells, with some AP expression in control cells and some loss of AP in USP22-depleted cells. Scale bar, relative size. The number of cells staining positive for AP is quantified in B. This effect was repeated Ͼ5 times. C, from a separate experiment, a quantitative colorimetric activity assay was used to detect the amount of AP activity in USP22-depleted and control cells. D, from a separate experiment, AP-positive cells and all cells were quantified as a percentage of total cells. ****, p Ͻ 0.0001; E, propidium iodide was used to label DNA content in USP22-depleted and control cells maintained in 2i/LIF. Student's t test was used to analyze differences between the cell cycle profile of control and USP22-depleted cells over four biological repeats, p ϭ 0.57. F, qRT-PCR was used to define levels of markers for early lineage commitment after 2 days of EB differentiation. *, p Ͻ 0.05. G, qRT-PCR was used to define levels of markers for late lineage commitment following EB differentiation for 1 week and adherent growth for 5 days. *, p Ͻ 0.05; ****, p Ͻ 0.001. H, E14 mouse ESCs were transfected with the FLAG-USP22 or FLAG-Vector as a control. USP22 overexpression is shown by Western blot and qRT-PCR (top panels). ***, p Ͻ 0.005. qRT-PCR was used to define levels of markers for late lineage commitment (bottom). *, p Ͻ 0.05; **, p Ͻ 0.01; ****, p Ͻ 0.001. I, after 6 days, FLAG-USP22 cells began to spontaneously form EBs. Error bars, S.D. (Fig. 2, A and B). USP22-depleted and control ESCs were allowed to spontaneously differentiate as EBs for 7 days. Cells were then reseeded at equal densities and grown under adherent conditions for 5 days, followed by staining to detect alkaline phosphatase (AP) activity. These assays detect a placental isoform of AP that is unique to pluripotent ESCs and epiblast stem cells (46 -48). Parental cells from both groups were maintained in 2i/LIF-conditioned medium over the course of the experiment as controls. Both USP22-depleted and control cells retained AP activity and grew clonally when maintained in 2i/LIF-conditioned medium ( Fig. 2A, pre-EB panels), consistent with the ability to grow in 2i/LIF-conditioned media to promote self-renewal. In contrast, after EB-induced differentiation, USP22-depleted ESCs showed increased AP activity compared with control cells (Fig. 2, A-C) and when quantified as a percentage of all cells (Fig. 2D). As an independent measure, AP was quantified using a colorimetric activity assay (Fig.  2C), and these results were also consistent with an essential role for USP22 in the successful differentiation of ESCs. To confirm that this defect in differentiation is not an off-target effect of the shRNA, the experiments shown in Fig. 2, C and D, were done using a different shRNA (shRNA 3 and 4, respectively).
USP22 is important for normal cell cycle progression in certain human cancer cells, with USP22 depletion causing a G 1 cell cycle arrest (22). To determine if the defect in differentiation observed in USP22-depleted ESCs is a consequence of a cell cycle defect, cells were stained for DNA content. From this analysis, no significant changes were found in the ESC cell cycle profile when USP22 was depleted from pluripotent ESCs (Fig. 2E).
As an epigenetic regulator, USP22 is potentially capable of modulating broad transcriptional programs (20 -22, 49, 50). Based on this, the role of USP22 in regulating the transcription of genes involved in the differentiation/self-renewal decision was evaluated. mRNA from differentiating EBs (Fig. 2F) or from post-EB differentiated cells (Fig. 2G) was analyzed for lineagespecific markers by qRT-PCR. In differentiating EBs, specific genes were used as markers for early lineage commitment: primitive endoderm (GATA6), definitive endoderm (SOX17, GSC), trophectoderm (Eomes), mesoderm (T), and ectoderm (OCT6). In post-EB differentiated cells, a distinct panel of genes served as late markers of commitment to the three embryonic germ layers: endoderm (␣-fetoprotein), mesoderm (ACTA and LY6A), and ectoderm (NESTIN). When compared with control ESCs, USP22-depleted ESCs express less mRNA for markers of the three germ layers both during and following differentiation (Fig. 2, F and G), suggesting that USP22-depleted ESCs retain a more stemlike transcriptional profile. Furthermore, there was no apparent bias for blocking a specific lineage in USP22-depleted cells, suggesting an effect very early in the onset of differentiation.
To test whether USP22 is not only necessary for the differentiation of ESCs, but also sufficient to induce differentiation, human USP22 was ectopically expressed in E14 mouse ESCs (Fig. 2H, top). After 3 days, RNA was harvested from these cells, and markers for late germ layer formation were assayed: mesoderm (ACTA, LY6A), endoderm (AFP), and ectoderm (NES-TIN). USP22-overexpressing cells had higher mRNA levels for the markers of all three germ layers, even while being maintained under conditions specifically designed to promote selfrenewal and pluripotency (Fig. 2H, bottom panels). The increased expression of these markers suggests that these cells might be more competent for differentiation than control cells. Remarkably, 6 days after infection, USP22-overexpressing ESCs began to spontaneously form EBs, again while being maintained in 2i/LIF-conditioned medium (Fig. 2I, bottom  panels). The ability of ectopic USP22 expression to drive differentiation was observed in both E14 and R1 mESCs and using both transfection and lentiviral transduction to ectopically express USP22 (Figs. 2 and 4E).
Expression Profiling in ESCs Links USP22 to the Repression of Pluripotency Genes-To determine the mechanism by which USP22 controls differentiation, genome-wide expression profiling was performed in USP22-depleted and control (shLuc) ESCs (Fig. 3A). This profiling revealed USP22-dependent expression of 2270 genes. Lending confidence to the data set, Usp22 itself was the second most repressed mRNA in the USP22-depleted sample. Approximately 150 total genes were compiled from gene ontology groups generated in previous studies that were linked to differentiation or pluripotency (supplemental Table 1) (36,39,48,(51)(52)(53)(54). Consistent with the biological phenotype, ontology analysis revealed that USP22 controls many genes involved in the decision between self-renewal and differentiation (Fig. 3B). Among the genes in this list that had a -fold change of 1.5 or more, which were associated with lineage-specific differentiation, no bias was observed for USP22 dependence (Fig. 3C). This is consistent with the observations from Fig. 2 that all three embryonic germ layers are affected by USP22 depletion, suggesting that the defect in differentiation lies upstream of lineage commitment. This gene expression analysis also revealed that a majority of differentiation-associated genes were activated by USP22, whereas a majority of pluripotency genes were repressed (Fig. 3D). Among the high confidence, USP22-dependent, core pluripotency genes to be validated, Sox2 was the most robustly and reproducibly repressed by USP22.
USP22 Represses Sox2 Transcription-Based on their inverse expression patterns during ESC differentiation (Fig. 1), the dominant role of USP22 in controlling SOX2 mRNA levels (Fig.  3), and their antagonist effects on ESC pluripotency, the relationship between USP22 and SOX2 was explored further. Following SOX2 mRNA levels over time in USP22-depleted ESCs revealed a dramatic increase in SOX2 transcript levels as USP22 protein levels decreased (Fig. 4A). This transcriptional relationship was seen using a panel of five individual shRNA constructs targeting USP22 (Fig. 4B). USP22 depletion leads to SOX2 transcriptional induction in both the mouse E14 (Fig. 4, A and B) and R1 (Fig. 4F) ESCs as well as the human H9 ESCs (Fig. 4C), suggesting a conserved regulatory mechanism. Interestingly, the defect in differentiation observed in USP22-depleted mouse ESCs was conserved in H9 human ESCs as well (Fig. 4D). USP22 and SOX2 protein levels change inversely during human ESC differentiation as well (data not shown). To test whether USP22 can repress Sox2 when ectopically expressed, human USP22 was expressed in R1 mESCs. These studies showed that ectopic USP22 is sufficient to decrease Sox2 transcription by 50% (Fig.

Control of Differentiation by USP22
4E). This has been observed in multiple experiments, and the decrease in SOX2 mRNA can be detected as early as 3 days following ectopic USP22 expression, well before the onset of spontaneous EB formation observed in these cells (Fig. 2). Furthermore, following the increase in SOX2 mRNA levels in USP22-depleted cells, there is a kinetic increase in SOX2 protein (Fig. 4, C and F). Finally, SOX2 protein induction results in the up-regulation of the Sox2 target gene Nanog, providing further evidence that the increase in SOX2 protein expression observed in USP22-depleted cells is functionally relevant (Fig.  4F, bottom). NANOG mRNA was not significantly (Ͼ1.5-fold) changed in our expression profiling studies (Fig. 3), which were performed prior to the induction of SOX2 protein. In contrast to SOX2, mRNA levels for the pluripotency factor OCT4 were not affected by USP22 depletion in our expression profiling studies, and little to no effect on OCT4 mRNA was observed when examined in ESCs treated with our panel of USP22 shRNA constructs (Fig. 4B). Consistent with this, OCT4 protein levels were not affected by USP22 depletion in human ESCs (Fig. 4C). -Fold changes of this magnitude (ϳ2-fold) in SOX2 levels are sufficient to induce phenotypic changes in ESCs, depending on the levels of the other pluripotency factors (55,56).
Collectively, these findings suggest that failure to repress Sox2 might contribute to the "enhanced pluripotency" phenotype observed in USP22-depleted ESCs. These data are differ-ent from a previous study in which ectopic SOX2 overexpression promoted differentiation and down-regulated Nanog transcription (55), in that the endogenous induction of SOX2 leads to NANOG up-regulation in this system. Together, these data implicate a transcriptional circuit in USP22-depleted ESCs in which increased levels of SOX2 and its target NANOG, along with unchanged OCT4 levels, create a defect in differentiation along mesendoderm (ME), neural ectoderm (NE), and neural crest (NC) lineages (Fig. 4G). A broad differentiation defect such as this is consistent with our experimental observations in USP22-depleted ESCs (Figs. 2 and 4D).

Loss of USP22 Occupancy at the Sox2 Locus Is Linked to Elevated uH2B Levels and Increased Sox2
Transcription-USP22, as part of SAGA, is known to occupy the promoter of many genes and can move through the body of these genes with RNA polymerase II as they are transcribed (22,30). In order to determine how USP22 chromatin occupancy functions to repress transcription of Sox2, chromatin immunoprecipitation (ChIP) experiments were performed at this locus in ESCs (Fig. 5A). ChIP was performed in control (shLuc) and USP22-depleted ESCs at a time point that captured the peak of Sox2 transcription in USP22-depleted cells (Fig. 5B). Results of ChIP using antibodies directed against USP22 demonstrated that it primarily occupies the Sox2 promoter region immediately upstream of the transcription start site (Fig. 5C). Confirming the specificity of this signal, USP22 levels at the Sox2 promoter FIGURE 3. Usp22 depletion promotes a pluripotent gene signature. A, RNA from shUsp22 and shLuc E14 cells was subjected to expression profiling (GSE42934). B, heat map depicting changes to all genes known to contribute to the maintenance of pluripotency or the derivation of each of the three embryonic germ layers. C, heat map depicting genes with Ͼ1.5-fold change and p Ͻ 0.12 that contribute to the decision between pluripotency and differentiation. Red and green, increased and decreased expression, respectively (relative to the mean value of a gene). Of the genes involved in maintaining pluripotency, Sox2 was up-regulated (red arrow). D, a total of 2270 genes were significantly (fold change Ͼ 1.5, p Ͻ 0.12) changed upon USP22 depletion. The percentage of genes involved in the commitment to the three embryonic germ layers or the maintenance of pluripotency that are up-and down-regulated is shown in red and green, respectively.
were dramatically decreased in USP22-depleted ESCs. This depletion in Usp22 occupancy coincided with decreased histone H2B at the Sox2 promoter, consistent with partial displacement of nucleosomes during active transcription. As assessed with a modification-specific antibody, uH2B accumulated along the Sox2 locus in USP22-depleted cells. uH2B accumulated both at the promoter and within the body of the Sox2 gene, consistent with previous studies (42,50). Furthermore, USP22-mediated control of uH2B levels appears selective because no change was seen at the Oct4 TSS or at an intergenic region of chromosome 9 (Ch9 D2) (Fig. 5D). When analyzed as a ratio of uH2B to total H2B occupancy, the accumulation of uH2B coincided geographically with USP22 occupancy at the locus.
To interrogate whether the accumulation of uH2B at the Sox2 locus might play a causative role in the induction of Sox2 transcription, E14 mESCs were depleted of the H2B E3 ligase RNF20. RNF20-depleted and control (Luc) cells were maintained in 2i/LIF-conditioned medium, similar to the USP22 depletion studies above. RNF20 depletion resulted in a sharp decrease in SOX2 expression in these cells (Fig. 5E), suggesting that uH2B is a critical determinant of Sox2 transcription. Importantly, USP22 levels were unaffected by RNF20 depletion. These data support a model in which USP22 represses Sox2 transcription by deubiquitylating H2B at the Sox2 locus, thereby antagonizing the positive effects of RNF20 on Sox2 transcription.
The USP22 Interacting Partners SIRT1 and ATXN7l3 Display Similar Phenotypic Effects with USP22 by Repressing Sox2 Transcription-USP22, as part of the SAGA acetyltransferase complex, is a transcriptional co-activator, and USP22 catalytic activity is dependent on the interaction between USP22 and the deubiquitylase module members within SAGA (20,22,23). USP22 also forms a functional complex with the deacetylase FIGURE 4. USP22 represses Sox2 transcription. A, shUsp22 or shLuc E14 mESCs were maintained in 2i/LIF medium and were harvested daily from days 4 -8 following infection. Western blotting was used to confirm USP22 knockdown (bottom panels), and qRT-PCR was used to show the temporal correlation between USP22 protein depletion and Sox2 transcriptional induction (top). **, p Ͻ 0.01. B, a panel of five shRNA constructs was used in E14 mESCs to deplete USP22 as confirmed by Western blotting (bottom panels), and qRT-PCR was used to measure SOX2 mRNA induction (top). ***, p Ͻ 0.005, C, USP22 depletion in H9 human ESCs causes SOX2 induction but not OCT4 induction, as measured by Western blotting. D, shUSP22 and shLUC H9 human ESCs were stained to detect AP expression following EB-induced differentiation. AP-positive cells were counted and quantified as a percentage of total cells. *, p Ͻ 0.05. E, ectopic expression of FLAG-USP22 or FLAG-Vector as a control was achieved in R1 mESCs by transfection. After 5 days, SOX2 and OCT4 mRNA were measured by quantitative RT-PCR. ***, p Ͻ 0.005. F, Sox2 transcriptional induction in R1 mESCs subsequently results in increased SOX2 protein levels (top panels). This increase in SOX2 protein expression leads to up-regulation of the downstream SOX2 transcriptional target NANOG (bottom panels). G, a schematic representation of the transcriptional circuit in ESCs in which SOX2, OCT4, and NANOG promote self-renewal by blocking lineage-specific differentiation pathways. ME, mesendoderm; NE, neural ectoderm; NC, neural crest. Error bars, S.D.
SIRT1 (25)(26)(27). SIRT1 functions in part as a transcriptional repressor and could therefore facilitate repression of Sox2 transcription by USP22. To test this possibility, RNAi was utilized to deplete mESCs of either USP22, SIRT1, or the SAGA deubiquitylase module subunit ATXN7l3 (Fig. 6, A  and B). Remarkably, depletion of either of the USP22 partners ATXN7l3 or SIRT1 induced Sox2 at both the mRNA and protein level (Fig. 6, A and B), similar to what is observed upon depletion of USP22 itself. As for USP22, derepression of Sox2 by SIRT1 depletion was complemented by studies in which ectopic expression of SIRT1 caused repression of Sox2 (Fig. 6C). SIRT1 directly occupies the Sox2 promoter, and this occupancy is lost upon SIRT1 depletion, as expected (Fig. 6D, top). Furthermore, the association with SIRT1 and the Sox2 locus is dependent upon USP22 expression (Fig. 6D, middle). Finally, transcriptional repression of other loci by SIRT1 involves the deacetylation of H4K16 (57,58), and depletion of USP22 results in an accumulation of this activating chromatin mark at the Sox2 locus (Fig. 6D, bottom).
Regulation of SOX2 Expression by USP22 Controls the Transition from Self-renewal to Differentiation in ESCs-USP22 depletion from ESCs leads to both an increase in SOX2 expression and a defect in differentiation. To test whether the increase in SOX2 is causally related to the biological phenotype, siRNA targeting SOX2 was utilized to blunt the SOX2 increase caused by USP22 depletion. As shown in Fig. 7A, SOX2 mRNA levels increased in USP22-depleted cells, as expected. This induction was eliminated by simultaneous treatment with SOX2 siRNA (Fig. 7A). ESCs with individual or co-depletion of USP22 and SOX2 were differentiated by EB formation and subsequently analyzed for AP activity. Consistent with previous experiments, USP22-depleted cells retained AP activity (Fig. 7B), indicating a defect in differentiation potential. Remarkably, this defect in ESC differentiation was rescued by co-depletion of SOX2 because the dual depletion cells lost AP. The requirement for SOX2 induction in the differentiation defect caused by USP22 depletion was observed using both quantitative enzymatic assays and direct staining of individual ESCs (Fig. 7C). Collectively, these data suggest a model in which an epistatic relationship links repression of Sox2 transcription by USP22 to control of the transition from pluripotency to differentiation (Fig. 7D).

DISCUSSION
In embryonic stem cells, the transition from pluripotency to differentiation is controlled by epigenetic changes (59 -62). The data presented here establish the epigenetic modifier USP22 as a critical regulator of this transition. USP22 is induced as ESCs differentiate. More importantly, both mouse and human ESCs require USP22 for proper differentiation. Conversely, ectopic expression of USP22 is sufficient to trigger spontaneous differentiation of ESCs, even in the absence of other differentiation signals. Gene expression profiling in ESCs demonstrated broad defects in the transcription of genes linked to all three germ layers when USP22 levels were depleted.
Remarkably, this analysis also demonstrated that USP22 represses transcription at the locus encoding the pluripotency factor SOX2. The effect of USP22 on Sox2 transcription appears to be direct because USP22 is bound at the Sox2 promoter, where its presence results in increased nucleosome density and a decrease in the USP22 substrate uH2B. This role of USP22 does not extend to loci encoding other pluripotency factors, including OCT4. Deubiquitylation of H2B as the primary biochemical mechanism by which USP22 controls Sox2 transcription is supported by evidence that depletion of the H2B E3 ligase RNF20 opposes the effect of USP22 on Sox2 transcription. The USP22 interacting partners ATXN7l3 and SIRT1 are also required for the repression of this locus. Functionally, the repression of Sox2 by USP22 is required for the transition from self-renewal to differentiation because ESCs depleted of USP22 fail to respond appropriately to differentiation cues. This defect in differentiation is rescued by artificially reestablishing Sox2 repression using RNAi. Remarkably, ectopic expression of USP22 is sufficient to drive ESCs to initiate differentiation, even in culture conditions designed to promote pluripotency. Collectively, these findings implicate USP22 as a potent regulator of the early steps in differentiation via its modulation of the epigenetic landscape at the Sox2 locus.
ESC pluripotency is achieved in part by maintaining appropriate expression levels of the core set of transcription factors that includes SOX2, OCT4, and NANOG (5, 6, 63-65). Here, depletion of USP22 was found to elevate levels of SOX2 and subsequently its downstream target NANOG, without changes in OCT4 levels. Phenotypically, USP22 depletion leads to a defect in differentiation that encompasses all three germ layers. Our current understanding of the transcriptional circuitry controlling pluripotency predicts that the induction of SOX2 in USP22-depleted cells blocks mesendoderm differentiation (Fig.  4G). Increased NANOG then blocks neuroectoderm and neural crest differentiation (6). In the presence of steady-state OCT4 levels, these effects are predicted to render USP22-depleted ESCs unable to commit to any lineage and suspended in a state of self-renewal (6), consistent with our experimental observations. The stoichiometry of the core transcription factors is central to the biological effects observed here because previous reports show, for example, that elevated SOX2 expression in the absence of NANOG induction would instead lead to ESC differentiation (55).
A variety of lines of evidence support the model that H2B deubiquitylation is the functionally relevant biochemical signal controlled by USP22 at the Sox2 transcription unit. Although there are a limited number of non-histone proteins that have been established as substrates of USP22 (e.g. TRF1, FBP1, and SIRT1), these proteins typically have highly specialized roles on chromatin. FBP1 is a sequence-specific transcription factor with no known role at the Sox2 locus, and TRF1 maintains telomere integrity and is localized to telomeres (66,67). In ESCs, SIRT1 does not appear to be a direct substrate of USP22 (Fig. 6A). USP22 is also able to deubiquitylate histone H2A, as FIGURE 6. The USP22 interacting partners SIRT1 and ATXN7l3 display similar phenotypic effects with USP22 by repressing Sox2 transcription. USP22, SIRT1, and ATXN7L3 were depleted from R1 mESCs using shRNA-containing lentivirus. A, whole cell lysates were analyzed by Western blotting for changes in SOX2 protein expression. B, mRNA was analyzed by quantitative RT-PCR for changes in mRNA expression. **, p Ͻ 0.01; ***, p Ͻ 0.005; ****, p Ͻ 0.0001. C, ectopic expression of human FLAG-tagged USP22 or SIRT1 was achieved by transfection in R1 mESCs, and changes in SOX2 mRNA levels were analyzed by quantitative RT-PCR. pcDNA FLAG-Vector was used as a control. Error bars, S.D. **, p Ͻ 0.01; ***, p Ͻ 0.005. D, chromatin immunoprecipitation and quantitative RT-PCR were used to show relative SIRT1 binding along the Sox2 locus following SIRT1 or USP22 depletion and H4K16Ac at the Sox2 locus following USP22 depletion. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.005; ****, p Ͻ 0.0001. Error bars, S.D.
we and others have shown (20,21). However, deubiquitylation of H2A by USP22 at Sox2 is predicted to induce, rather than repress, Sox2 transcription. In addition, our genome-wide expression profiling provides some evidence against the model of uH2A as the relevant USP22 substrate. For example, if uH2A were the relevant USP22 substrate, we would expect widespread activation of differentiation genes in the USP22-depleted condition. Instead, the majority of differentiation genes were repressed when USP22 was depleted (Fig. 3). In contrast to the other USP22 substrates, the deubiquitylation of H2B is well documented as a mechanism of transcriptional repression (42, 44, 68 -70). uH2B accumulation at the Sox2 locus occurs in a USP22-dependent manner. Because uH2B is a direct substrate of USP22-mediated deubiquitylation, these data are consistent with USP22 having a direct effect at the Sox2 locus levels. However, additional studies will be required to formally assess this possibility. Perhaps the strongest evidence for H2B as the relevant USP22 substrate at the Sox2 locus comes from our observation that blocking H2B ubiquitylation by depleting ESCs of the H2B E3 ligase RNF20 results in decreased SOX2 expression. From a developmental standpoint, USP22-mediated repression of Sox2 may contribute to the embryonic lethality observed in USP22 null mice at day 9.5, a stage when Sox2 repression is critical for the differentiation of neural stem cells in the closing neuropore and developing brain (26,38,40,71,72).
USP22 associates with both the transcriptional co-activator complex SAGA and the transcriptional repressor SIRT1 (26,27). Additional studies will be required to gain an understanding of whether the ability of USP22 to repress Sox2 transcription is a function of its association with SIRT1, a novel co-repressor activity of the SAGA complex, or via an unknown partner or biochemical activity.
Finally, other recent studies are consistent with our model that controlling the deubiquitylation of uH2B, and thereby controlling gene repression, is critical in the transition from selfrenewal to differentiation (42,73). In one of these studies the Oren group reported that deubiquitylation of H2B at differentiation genes is essential for transcriptional repression (42) in pluripotent ESCs. This effect was catalyzed by the USP22-related enzyme USP44, whose levels decrease during the transition from self-renewal to differentiation, just as USP22 levels increase. This decrease in USP44 allows the accumulation of uH2B and enhances transcription at genes involved in this transition. Thus, as cells differentiate, USP44 expression is decreased, facilitating the induction of genes involved in differentiation, whereas USP22 is simultaneously increased to FIGURE 7. Regulation of SOX2 expression by USP22 controls ESC transition from self-renewal to differentiation. Following selection with puromycin, shLuc of shUsp22 E14 mESCs were co-transfected with GFP and siRNA targeting SOX2 or control siRNA (SiNeg). Flow cytometry was used to determine transfection efficiency, which was 60 -75%. SOX2 expression was analyzed by qRT-PCR (A, top) and by Western blotting (A, bottom). **, p Յ 0.01; ****, p Ͻ 0.0001. B, post-EB cells were harvested as whole-cell lysate and subjected to a colorimetric AP activity assay. **, p Յ 0.01. C, post-EB cells were fixed and stained to detect AP expression, and positive cells were quantified in triplicate. **, p Յ 0.01. D, schematic representation of how USP22 controls ESC transition from self-renewal to differentiation. Top, when ESCs are undergoing self-renewal, USP44 levels are high in order to repress genes involved in differentiation through the deubiquitylation of H2B at these gene loci. USP22 levels are low during self-renewal, and Sox2 is therefore expressed. When ESCs receive a signal to differentiate, such as RA or EB formation, USP44 levels are reduced, allowing differentiation genes to be derepressed and turn on. Simultaneously, USP22 levels increase and USP22 deubiquitylates H2B at Sox2. Bottom, a schematic depicting the balance of USP44 with differentiation genes and USP22 with Sox2. The changing levels of USP44 and USP22 act in concert to epigenetically control the transition from self-renewal to differentiation. Error bars, S.D.
repress the transcription of the pluripotency factor Sox2 (Fig.  7D). It remains unclear how this separation of function between USP22 and USP44 is achieved. It seems likely that a mechanism exists that allows these enzymes to distinguish their respective target gene loci, perhaps via differential interaction with the sequence-specific transcription factors that recruit them across the genome. Current efforts are focused on defining the mechanisms that distinguish the roles of USP22 and USP44 in the critical events controlling pluripotency and differentiation and on understanding how USP22 functions selectively as either a co-activator or co-repressor of gene transcription.