Dynamic Status of REST in the Mouse ESC Pluripotency Network

Background REST is abundantly expressed in mouse embryonic stem cells (ESCs). Many genome-wide analyses have found REST to be an integral part of the ESC pluripotency network. However, experimental systems have produced contradictory findings: (1) REST is required for the maintenance of ESC pluripotency and loss of REST causes increased expression of differentiation markers, (2) REST is not required for the maintenance of ESC pluripotency and loss of REST does not change expression of differentiation markers, and (3) REST is not required for the maintenance of ESC pluripotency but loss of REST causes decreased expression of differentiation markers. These reports highlight gaps in our knowledge of the ESC network. Methods Employing biochemical and genome-wide analyses of various culture conditions and ESC lines, we have attempted to resolve some of the discrepancies in the literature. Results We show that Rest+/− and Rest−/− AB-1 mutant ESCs, which did not exhibit a role of REST in ESC pluripotency when cultured in the presence of feeder cells, did show impaired self-renewal when compared with the parental cells under feeder-free culture conditions, but only in early passage cells. In late passage cells, both Rest+/− and Rest−/− AB-1 ESCs restored pluripotency, suggesting a passage and culture condition-dependent response. Genome-wide analysis followed by biochemical validation supported this response and further indicated that the restoration of pluripotency was associated by increased expression of the ESC pluripotency factors. E14Tg2a.4 ESCs with REST-knockdown, which earlier showed a REST-dependent pluripotency when cultured under feeder-free conditions, as well as Rest−/− AB-1 ESCs, showed no REST-dependent pluripotency when cultured in the presence of either feeder cells or laminin, indicating that extracellular matrix components can rescue REST's role in ESC pluripotency. Conclusions REST regulates ESC pluripotency in culture condition- and ESC line-dependent fashion and ESC pluripotency needs to be evaluated in a context dependent manner.


Introduction
Pluripotent mouse embryonic stem cells (mESCs) derived from the inner cell mass of the mammalian blastocyst are capable of forming all the tissues of the organism and possess the ability to self-renew in an undifferentiated manner. Understanding the mechanisms that control the self-renewal and pluripotency of ESCs has far-reaching implications for the fields of developmental biology, regenerative medicine, and oncology [1,2,3]. The recent successes with reprogramming somatic cells into induced pluripotent stem-like cells have brought intense enthusiasm to this area of research [4,5,6,7,8,9,10,11]. These findings suggest that factors such as Oct4, Nanog, and Sox2 are core components of a large, interconnected network that regulates self-renewal and pluripotency in both mouse and human ESCs [2,12,13,14]. This network is likely regulated by maintenance factors that are triggered by cellular and environmental signals [15]; these factors both protect the self-renewal state from spurious signals and cause differentiation of the cell when required.
REST is a transcriptional repressor that was originally discovered to be a repressor of many neuronal differentiation genes [16,17]. Experiments with Rest mutant mice supported REST's role in neurogenesis [18]. However, later reports indicated that REST can potentially repress about a thousand genes and affect various cellular functions in a context-dependent fashion [19,20,21,22,23]. Although REST expression is higher in ESCs than in most other cell type [24], its function in the selfrenewal network is unclear. REST has been reported to maintain mESC self-renewal and pluripotency by directly suppressing miR-21 [25]. In these studies, loss of REST in ESCs accompanied expression of many differentiation markers, including Gata4. In contrast, others found REST to have no role in the maintenance of mESC pluripotency [26,27,28] and loss of REST did not change expression of differentiation markers including Gata4 [27,28]. Yet, another report found that REST was not required to maintain ESC pluripotency and, in contrast, loss of REST actually caused decreased expression of differentiation markers, including Gata4 [29]. However, genome-wide promoter occupancy assays have shown that REST indeed is an integral part of the ESC pluripotency network [30,31,32,33,34], raising the interesting question whether such genome-wide assays truly reflect the biology of ESC pluripotency. Thus, these conflicting results have created a gap in our knowledge of the mESC network, particularly with respect to REST. Using wild-type, Rest +/2 and Rest 2/2 ESCs and conditions used by different laboratories, we determined how REST functions in the mESC pluripotency network in coordination with other factors. The work presented here resolves some of the apparent discrepancies in the field.

Mouse ESC lines and culture conditions
E14Tg2a.4 were purchased from Bay Genomics (MMRRC item #015890-UCD-CELL) and parental AB1 (N6: Wild-type), N9 (Rest +/2 ) and N8 (Rest 2/2 ) were obtained as gifts from Amanda Fisher Laboratory [27] (originally obtained in presence of feeder cells and were maintained in presence of feeder layer and LIF; passage numbers in the manuscript represent passage in absence of feeder cells), were cultured without feeder cells in presence of 1000 units ml 21 LIF (ESGRO) on gelatin-coated tissue culture dishes [35]. Inactivated (Mitomycin C-treated) mouse embryonic fibroblasts (MEFs) were used as feeder layer in addition to LIF in some experiments. MEFs were obtained from ATCC (SCRC-1040 TM , ATCC) and were used at passage no more than four passages (P4). Laminin coated plates were made by incubating the plates with 5 mg/ml Laminin in PBS (invitrogen, # 23017-015) at 4uC overnight.
shRNA and RNAi-mediated knockdown and self-renewal assay RNAi-mediated knockdown of REST was carried out per the manufacturer's protocol (Dharmacon). siREST (400 pmole), and nontargeting siRNA (400 pmole) were used in nucleofection using AMAXA. Approximately 1000 cells post-transfection were seeded per well in triplicates of 6-well plates for self-renewal assay. Colonies were assayed for alkaline phosphatase activity 3 days after transfection. shRNA mediated knockdown was performed using 1 MOI of non-targeting control (RHS4372, Thermo scientific) and REST (V3LMM_509606, Thermo scientific) specific lentiviruses. Puromycin selection (1.5 mg/ml) was performed 48 hours post transduction. Puromycin-resistant and GFP positive cells were selected and expanded for several generations for further experiments after FACS analysis. Percentages of selfrenewing colonies from three experimental replicates were plotted. Standard error of means was calculated and is depicted as error bars in the figures.

Microarray and bioinformatic analysis
One sample from each cell line (N6 (WT), N9 (Rest +/2 ), and N8 (Rest 2/2 )) cultured in parallel for passage 2 and passage 10 (after removal of feeder cells) were used for microarray analysis. Total RNA was isolated using Trizol method from N6 (WT), N9 (Rest +/2 ), and N8 (Rest 2/2 ) cells grown for various passages. Cleanup was performed for all samples using Qiagen RNeasy mini kit. Sample RNA concentration was determined with a Nanodrop ND-1000 Spectrophotometer and RNA quality was verified with an Agilent 2100 Bioanalyzer using a RNA Nano Chip. RNA samples displaying no visible degradation in the Bioanalyzer analysis with two sharp ribosomal peaks were deemed acceptable for further processing. Affymetrix's GeneChip IVT Express kit was used for cDNA synthesis and in vitro transcription. Affymetrix GeneChip Mouse Genome 430 2.0 was used in this study. A minimum of greater than or equal to 2.0 fold change was established as the criteria during analyses. All data was normalized using the RMA expression statistical analysis (Robust Multi-array Analysis. Venn Diagram between N6 (WT), N9 (Rest +/2 ), and N8 (Rest 2/2 ) was plotted after calculating the fold changes for genes from passage 2 to 10 for each cell types (N6, N9 and N8) and selected those with their absolute fold changes more than two as differential expressed ones. For overlap and its significance with FunGenES database, we selected 50 most up-regulated genes and 50 most down-regulated genes in N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells during the transition from passage 2 to passage 10 and compared them with FunGenES database: http://biit.cs.ut.ee/ fungenes/ for their overlap (concordance) in the regulation direction. The statistical significance (p-value) of the overlap was further evaluated using hypergeometric distribution.

Ingenuity Pathway Analysis (IPA)
The dataset obtained from microarray analysis of samples from passages 2 and 10 for each cell line was uploaded to IPA server. The genes with more than or equal to 2 fold change in expression were included for analysis. For all the IPA analyses, the significance value associated with Functional Analysis for the dataset is a measure of the likelihood that the association between a set of Functional Analysis genes in experimental samples and a given process or pathway is due to random chance. The smaller the p-value the less likely that the association is random and the more significant the association. In general, p-values less than 0.05 indicate a statistically significant, non-random association. The pvalue associated with a biological process or pathway annotation is a measure of its statistical significance with respect to the Functions/Pathways/Lists Eligible molecules for the dataset and a Reference Set of molecules (which define the molecules that could possibly have been Functions/Pathways/Lists Eligible). The p-value is calculated with the right-tailed Fisher's Exact Test.

Results
Rest +/2 and Rest 2/2 AB-1 ESCs cultured in the absence of feeder cells show impaired self-renewal that is restored upon prolonged culture Previously, using two Rest +/2 mutant and their parental E14Tg2a.4 mESC cell lines cultured in the absence of mouse embryonic fibroblast feeder cells, we found that REST maintains the self-renewal and pluripotency of ESCs [25]. We cultured these cells without feeder cells because we wanted to use a system in which the ESC pluripotency pathways are not influenced by the feeder cells. In addition, we chose the E14Tg2a.4 mESC line because these cells were generated to grow without feeder cells. However, using Rest mutant (N9 (Rest +/2 ) and N8 (Rest 2/2 )) and their parental AB-1 N6 (WT) cell lines cultured in the presence of feeder cells, others found that REST does not have a role in the maintenance of mESC self-renewal and pluripotency [27]. To gain insight into this discrepancy, we obtained these cell lines [N6 (WT), N9 (Rest +/2 ) and N8 (Rest 2/2 )] from the investigators and maintained them in the presence of feeder cells plus LIF. We then took aliquots of the N6 (WT), N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells, cultured them with LIF, but in the absence of feeder cells, and measured their self-renewal capacity using alkaline-phosphatase (AP) assays.
The control N6 (WT) line morphologically showed minimal differentiation, whereas both the mutant N9 (Rest +/2 ) and N8 (Rest 2/2 ) AB-1 lines showed extensive differentiation within 2 passages after they were shifted to feeder-free culture conditions (data not shown), similar to what was reported previously with the E14Tg2a.4 cell line and its derivative REST +/2 mutant cells [25]. Surprisingly, upon prolonged culture (10 passages), both the mutant cell population morphologically looked similar to the wildtype cells. To analyze this phenomenon further, we performed AP assays. As shown in Figure 1A, the self-renewal efficiencies of both the mutant lines were significantly lower than that of the parental AB-1 ESCs at passage 2 but not at passage 10 ( Figure 1A, quantification of self-renewal assays; Figure 1B, representative pictures of colonies from self-renewal assays) (see also Figure S1 for picture of representative dishes and calculations of percent selfrenewal). At passage 2, the N9 (Rest +/2 ) and N8 (Rest 2/2 ) lines had fewer compact colonies and less self-renewal efficiency compared with the parental line. In contrast, at passage 10, the self-renewal efficiency of the parental line stayed the same (no statistical difference), whereas both the mutant lines recovered their self-renewal capacity with strong statistical significance and were now equivalent to the wild-type line ( Figure S1). These results indicated that REST mutant AB-1 ESCs cultured in the absence of feeder cells produced impaired self-renewal during early passage and this deficiency is restored upon prolonged culture. Thus, these results explain the discrepancy seen between the contrasting reports reported earlier [25,27]. Interestingly, although both Rest +/2 and Rest 2/2 cells show drastic reduction in self-renewal at passage 2 and subsequent restoration at passage 10 when compared with WT cells, Rest 2/2 cells show higher ability of selfrenewal when compared with Rest +/2 cells at both passage 2 and passage 10. One of the possibilities to explain this phenomenon is that complete loss of REST protein renders these cells predisposed to faster restoration of self-renewal by a yet-unknown mechanism (see Discussion).
Genome-wide analysis of Rest+/2 and Rest2/2 AB-1 ESCs cultured in the absence of feeder cells shows impaired self-renewal that is restored upon prolonged culture To understand the mechanisms behind the restoration of selfrenewal in the mutant cells, we determined the mRNA expression profiles of N6 (WT), N9 (Rest +/2 ), and N8 (Rest 2/2 ) cells that were cultured after either passage 2 or passage 10. A comparison of gene expression profiles between N9 (Rest +/2 ), N8 (Rest 2/2 ) and N6 (WT) cells cultured for passage 2 and published lineage specific markers (the lists are from microarray profiling studies of lineage specific differentiation of ES cells) [37]. We then used Ingenuity Pathway Analysis (IPA) to overlay N9 (Rest +/2 ), N8 (Rest 2/2 ) and N6 (WT) expression profiles with known lineage specific markers). Such analysis revealed that N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells showed lower expression of many ESC-specific genes and higher expression of ectoderm-, mesoderm-, and endoderm-specific lineage genes (Table 1). These results indicated that N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells were less similar to ESCs than the wild-type cells at passage 2 and supported the earlier AP assays described in Figures 1A and 1B.
To determine what happens to the N9 (Rest +/2 ), N8 (Rest 2/2 ) and N6 (WT) cells during the transition between passage 2 and passage 10, we compared their gene expression profiles [37]. Genes that showed a significant change in expression during the transition between passage 2 and passage 10 were scarce in wildtype cells (about 65 genes) but plentiful in N9 (Rest +/2 ) (965 genes) and N8 (Rest 2/2 ) (1296 genes) ( Figure 1C), which supports the idea that the N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells undergo a strong change in global gene expression upon prolonged culture.
We then compared alterations in expression profiles of N9 (Rest +/2 ), N8 (Rest 2/2 ) and N6 (WT) cells over passages (from passage 2 to passage 10) ( Table 2). The wild-type cells showed minimal alteration in either ESC-, ectoderm-, mesoderm-or endoderm-specific genes during the transition between passage 2 and 10. In contrast, the N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells showed significant increase in ESC specific genes and significant decrease in all three lineage-specific genes, again supporting our AP-based evidence that while the wild-type cells represent pluripotent state and remain largely unchanged during the prolonged culture, the impaired self-renewal in Rest mutant cells seen in early passage cells is restored in late passage cells.
We then selected the 50 genes with the highest fold changes that were either up-regulated or down-regulated in N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells during the transition from passage 2 to passage 10 (Table S1 (for N9 (Rest +/2 )) and Table S2 (N8 (Rest 2/2 )) for specific fold-changes in gene expression levels) and compared them to an additional independent database (FunGenES database: http://biit.cs.ut.ee/fungenes/) [38] for their concordancy in upor down-regulation (Table S1 and Table S2). The wild-type ESC expression profile was not included in this analysis because these cells did not show major changes in their gene expression profile from passage 2 to passage 10. Results indicated that the upregulated genes corresponded mostly to genes expressed in selfrenewing ESCs (the concordancy is 37 out of 50; P = 0.00999 for N9 (Rest +/2 ) and 39 out of 50; P = 0.00163 for N8 (Rest 2/2 )) ( Figure 1D and Figure 1F, see Table S3), whereas the downregulated genes corresponded mostly to genes expressed in differentiating embryoid bodies (EB6) [similarly, the concordancy is 46 out of 50; P = 3.30E-13 for N9 (Rest +/2 ) and 45 out of 50; P = 4.12E-12 for N8 (Rest 2/2 )] ( Figure 1E and Figure 1G, see Table S3). Thus, supporting the AP assays seen above, the genome-wide analysis using both IPA and FunGenES databases indicated that, while the wild-type cells showed similarity to pluripotent ESCs and remain unchanged from passage 2 to passage 10, the N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells showed more similarity to differentiating cells at passage 2 while they showed more similarity to self-renewing ESCs in passage 10.
In genome-wide analysis, restoration of the impaired selfrenewal and pluripotency of Rest+/2 and Rest2/2 AB-1 ESCs associates, at least in part, with increased ESC pluripotency pathways To determine whether there were changes in the ESC-specific canonical pathways in N6 (WT), N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells during the transition between passage 2 and passage 10, we used the IPA program to analyze the potential pathway changes in these cells. The results indicated that the wild-type cells did not have any significant changes (except three involving regulation of actin-based motility by Rho, role of tissue factor in cancer and Bcell receptor signaling) in cellular pathways during the transition from passage 2 to passage 10 ( Figure 2A). In contrast, several pathways changed during the prolonged culture of the N9 (Rest +/2 ) ( Figure 2B) and N8 (Rest 2/2 ) cells ( Figure 2C). Importantly, the pathway representing the human embryonic stem cells in N9 (Rest +/2 ) cells (p = 1.55E-06) and the role of Nanog in mammalian embryonic stem cell pluripotency was the second most-altered pathway in the Rest 2/2 cells (p = 2E-06) (the top 5 changed pathways, with p-values, are shown in Table S4).
Restoration of the impaired self-renewal and pluripotency of Rest+/2 and Rest2/2 AB-1 ESCs parallels increased REST expression in Rest+/2 and increased pluripotency factors in both Rest+/2 and Rest2/2 cells To validate the genome-wide analysis, we used Western blotting assays to measure REST protein levels ( Figure 2D: left panel:   Western blot, right panels: quantification of western blot) and the core pluripotency factors Oct4, Sox2, and Nanog in cells cultured for 2, 5, and 10 passages ( Figure 2D; left panel: Western blot, right panels: quantification of western blot). Surprisingly, REST protein levels were elevated in N9 (Rest +/2 ) cells with increased passage whereas, it remained unchanged in the N6 (WT) and N8 (Rest 2/2 ) cells ( Figure 2D and Figure S2), as expected. Consistent with previous reports [30,31,32,33,34], even though no REST protein can be made in N8 (Rest 2/2 ) cell line, we observed an increase in the transcript levels of the truncated Rest gene during this recovery of impaired self-renewal ( Figure 2E). As described above, one of the possibilities is that there is a yet-unknown homeostatic mechanism that attempts to upregulate Rest expression when the Rest transcript levels are lower as in Rest+/2 cells compared with wild-type cells.
Whereas the wild-type cells maintained high levels of the core self-renewal factors throughout the passages, the protein levels of Nanog ( Figure S3), Oct4 ( Figure S4) and Sox2 ( Figure S5) were significantly lower during the early passages of N9 (Rest +/2 ) and N8 (Rest 2/2 ) cells compared to the parental control ESCs ( Figure 2D). However, the levels of these factors (Nanog, Oct4 and Sox2) in the N9 (Rest +/2 ) and in N8 (Rest 2/2 ) cells increased with passage, such that by passage 10, their levels were equivalent to those in the wild-type control ESCs. Interestingly, the levels of Nanog in the N8 (Rest 2/2 ) cells significantly increased during the transition from early to late passage, to a level about 2-fold higher than that in the control cells. The increased protein levels of the core pluripotency factors during their transition from passage 2 to passage 10 in Rest 2/2 cells also paralleled in the increase in their corresponding transcript levels ( Figure 2F). Thus, these results suggested that the elevated levels of pluripotency factors, particularly Nanog, played a critical role in the recovery of selfrenewal of ESCs in the complete absence of REST protein. Thus, there appeared to be an intricate interaction between REST and the core pluripotency factors, including the Nanog pathway, in which the loss of REST is compensated by the pluripotency factors (see Discussion).

Extracellular cues provided by feeder cells compensate for REST-mediated loss of self-renewal and pluripotency in AB-1 and E14Tg2a.4 ESCs
The previous reports using Rest-null AB-1 cells cultured in the presence of feeder cells indicated that REST plays no role in the maintenance of ESC self-renewal and pluripotency [27]. To determine whether the presence of feeder cells in these experiments was responsible for the difference between their results and the results of our experiments performed without feeder cells (described above), we cultured the N6 (WT), N9 (Rest +/2 ) and N8 (Rest 2/2 ) AB-1 cells in the presence and absence of feeder cells. In contrast to our results obtained in the absence of feeder cells (Figure 1 A and B), the presence of feeder cells resulted in no large difference in self-renewal efficiency between the N6 (WT), N9 (Rest +/2 ) and N8 (Rest 2/2 ), as determined by AP assay ( Figure 3A: quantification of AP self-renewal assay; Figure 3B: representative pictures of AP-stained colonies). Further Western blotting assays to detect the core self-renewal regulators Oct4, Sox2, and Nanog ( Figure 3C; left panel: Western blot, right panels: quantification of western blot) confirmed that the feeder cells compensated for the loss of self-renewal and pluripotency in N9 (Rest +/2 ) and N8 (Rest 2/2 ) AB-1 ESCs. Interestingly, REST protein levels in N9 (Rest +/2 ) cells when cultured in presence of feeder cells is similar to that of N9 (Rest+/2) cells cultured without feeder at passage 10.
To determine if a similar mechanism operated in the E14Tg2a.4 ESC line, we performed knock-down experiments in these cells using siRest [or nontargeting (NT) control siRNA], cultured them in the presence or absence of feeder cells, and determined their self-renewal efficiency with AP assay ( Figure 4A). The results indicated that the lower self-renewal observed in the siRest-treated cells cultured in the absence of feeder cells could be countered by culturing the cells with feeder cells. Further analysis indicated that laminin, an extracellular matrix component, could counter lowered self-renewal in ESCs that were treated with siRest ( Figure 4B). Analysis by quantitative RT-PCR (qRT-PCR) of the transcript levels of the cells cultured without feeder cells or with laminin further indicated that while the siRest treatment lowered Rest levels in cells cultured in either condition, the levels of the self-renewal regulators Oct4, Sox2, and Nanog were lowered extensively in the cells cultured without feeder cells and to a lesser extent in the cells cultured with laminin ( Figure 4C). Previous studies showed that REST maintained pluripotency of E14Tg2a.4 cultured on gelatin by repressing the transcription of miRNA-21 [25]. Therefore, we evaluated the status of miRNA-21 levels in cells after siRest treatment and cultured either on gelatin or on laminin surfaces. MiRNA-21 levels were elevated in siRest-treated cells but only when the cells were cultured on gelatin coated surfaces and not when cells were cultured on laminin coated surfaces (supplementary Figure S6). These results suggest that culturing cells on laminin triggered a yet unknown mechanism that uncoupled the REST-miRNA-21 axis.
To determine whether laminin could counter the RESTmediated loss of self-renewal and pluripotency in long-term culture of E14Tg2a.4 ESCs, as compared to the immediate effects seen in the transient siRest-mediated knock-down experiments described above, we purchased new stock of E14Tg2a.4 ESCs to verify that our earlier results did not arise because of unforeseen problems    with our cell line during laboratory culture. Further analysis indicated that the new cells were of normal karyotype ( Figure S7). We then generated stable shRest and NT control ESC lines. After selecting puromycin-resistant and GFP-positive (by FACS) clones, we performed self-renewal assays ( Figure 4D) and determined the protein levels of REST, Oct4, Sox2, and Nanog ( Figure 4E; left panel: Western blot, right panels: quantification of western blot).
The results supported the conclusion reached by siRest experiments and showed that while shRest-treated cells lowered the REST protein levels in cells cultured either on gelatin or on laminin, the self-renewal efficiency and the levels of corresponding self-renewal and pluripotenmcy regulators were substantially lower when the cells were cultured on gelatin, but a minimal reduction was observed when cells were cultured on laminin. Thus, the lowered self-renewal caused by lowered REST levels in E14Tg2a. 4 ESCs, as in AB-1 ESCs, could be countered by culturing the cells with laminin. The results indicate that the loss of REST-mediated ESC pluripotency can be countered by feeder cells or extracellular matrix components and may be seen only when the ESCs are cultured in the absence of these extracellular cues.

Stable knockdown of REST in E14Tg2a.4 ESCs causes decreased self-renewal and pluripotency and increased expression of differentiation markers under feeder-free conditions
Previously, we had used two different Rest+/2 lines and siRNA-mediated knockdown of REST in E14Tg2a.4 ESCs. To determine whether stable knockdown of REST in these cells made any difference in the role of REST in ESC pluripotency, we generated stable shRest and NT control ESC lines of E14Tg2a.4 cells. First, we performed self-renewal assays. The E14Tg2a.4 cells showed decreased self-renewal, as reported earlier [25] (Figure 4D) (see Figure S8 for representative AP-stained colonies and calculations for self-renewal efficiency). The western blot analysis results ( Figure 4E; left panel: Western blot, right panels: quantification of western blot) is consistent with that of selfrenewal assay shown in Figure 4D. The REST levels showed distinct decreases due to shRest treatment. In the shRest E14Tg2a.4 cells, loss of REST resulted in lowered levels of Sox2, Oct4, and Nanog protein as compared to control.
To determine whether the self-renewal status of these cells corresponded to the expression of differentiation lineage markers, we performed qRT-PCR assays. E14Tg2a.4 cells treated with shRest showed significant increases in the expression of lineage markers representing the ectoderm, trophectoderm, mesoderm, and endoderm that were concomitant to their decreased selfrenewal and pluripotency ( Figure S9).

Conditional role of REST as an ESC pluripotency maintenance factor
This work attempts to resolve the current discrepancies in the various published reports about the role of REST in the maintenance of ESC self-renewal and pluripotency. Here we show that REST, a maintenance factor, maintains self-renewal and pluripotency in a culture condition and cell-dependent manner. Loss of self-renewal and pluripotency due to loss of REST in the shRest lines of E14Tg2a.4, which was originally generated as a feeder-independent line, can be seen in cells propagated for many generations. Loss of REST in AB-1 cells, which was generated as a feeder-dependent line, shows RESTdependent maintenance of self-renewal when they are shifted to feeder-free culture conditions, but only during early passages. This impaired self-renewal and pluripotency is then restored in late passage cells, presumably owing to a culture condition-dependent response. However, the exact mechanism or mechanisms that lead to this response is unknown. One mechanism could be through a compensatory response, in which majority of cells respond to the selective pressure by capturing alternate pathways, such as increased Nanog pathway, to regain their self-renewal and pluripotency. Such a potential connection between Nanog and REST is supported by the observation that the greatest increase in the Nanog protein levels was also seen in the V6.5 ESC line upon loss of REST (29). In addition, REST has been shown to be in complex with Nanog by protein-protein interaction [33,34]. And, loss of one component could be compensated by increased levels of another component of protein complexes that regulate important biological functions. For example, loss of Rpb4 can be countered by increased levels of another component Rpb7 in the RNA polymerase II protein complex [39]. Thus, it is conceivable that the decreased ESC self-renewal/pluripotency because of loss of REST is recovered, at least in part, due to a compensatory role of Nanog and the Nanog-pluripotency pathway. However, another mechanism could be through the selective expansion of minor population of cells containing high levels of pluripotency factors and high capacity of self-renewal. These cells, which constituted a minor population at early passage, are selected to become a major population after prolonged culturing. Thus, the explanation for the observed restoration of pluripotency and self-renewal may be due to expansion of specific cell population or a compensatory response or both.

Extracellular cues influence ESC Pluripotency
Our observations presented here support an interesting hypothesis that REST-mediated mESC pluripotency depends on both the intracellular regulatory network and the extracellular environment. This might explain why some studies have found that REST plays a role in the maintenance of mESC pluripotency while others have not (see Introduction). It is likely that the regulatory pathways of mESC pluripotency that are regulated by factors other than REST are also maintained in equilibrium with environmental cues.

Conclusions
Here we provide evidence that REST maintains self-renewal and pluripotency of ESCs in a context-dependent fashion and that ESC pluripotency should be evaluated in a context-dependent manner.

Supporting Information
Figure S1 Impaired self-renewal efficiency is restored after prolonged culturing. Alkaline phosphatase based selfrenewal assay for WT, Rest+/2 and Rest2/2 cells cultured in parallel for 2 and 10 passages. Top: Scanned image of self-renewal assay plates. Sample name are labeled on the left side and passage numbers are at the bottom of the images. Bottom: Table for total number of colonies in self-renewal plates, differentiating (white), self-renewing colonies (pink) and percent self-renewal is shown. (TIF) Figure S2 REST protein levels is elevated in Rest +/2 ESCs over prolonged culturing under self-renewing conditions. Western blot analysis for REST in whole cell lysate from N6 (WT), N9 (Rest +/2 ) and N8 (Rest 2/2 ) ESCs after passaging without feeder layer but in presence of LIF is shown. Actin for same membrane is shown on the left. Whole blot image of three membranes are shown. Lanes are numbered on top of each blot. The table at the bottom shows the sample names and their passage numbers. (TIF) Figure S3 Nanog protein levels is elevated in Rest +/2 and Rest 2/2 ESCs over prolonged culturing under selfrenewing conditions. Western blot analysis for Nanog in whole cell lysate from N6 (WT), N9 (Rest +/2 ) and N8 (Rest 2/2 ) ESCs after passaging without feeder layer but in presence of LIF is shown. Actin for same membrane is shown on the left.