miR-294/miR-302 Promotes Proliferation, Suppresses G1-S Restriction Point, and Inhibits ESC Differentiation through Separable Mechanisms

The miR-294 and miR-302 microRNAs promote the abbreviated G1 phase of the embryonic stem cell cell cycle and suppress differentiation induced by let-7. Here, we evaluated the role of the retinoblastoma (Rb) family proteins in these settings. Under miR-294 the G1-S of the Rb In contrast, miR-294 the further accumulation of cells in G1 in response to nutrient deprivation and cell-cell contact in an Rb-dependent We uncovered ﬁve additional miRNAs (miR-26a, miR-99b, miR-193, miR-199a-5p, that silenced ESC self-renewal in the absence of other miRNAs, all of which were antagonized by miR-294 and miR-302. Four of the six differentia-tion-inducing miRNAs induced an Rb-dependent G1 accumulation.


INTRODUCTION
Embryonic stem cells (ESCs) have rapid and unlimited growth potential while retaining the ability to differentiate into any cell type of the adult (http://stemcells.nih.gov/info/basics/pages/ basics3.aspx). In contrast, somatic cells of adult tissues have limited growth and developmental potential. Because of the unique properties of ESCs, there is much promise in their use to study and treat disease, yet the basis of their potential remains incompletely understood.
Somatic cells have an extended G1 phase enabling them to respond to their environment (Blomen and Boonstra, 2007). Somatic cells like fibroblasts arrest in G0/G1 when nutrient starved or when in contact with neighboring cells (contact inhibition). This arrest, which occurs at a checkpoint called the restriction (R) point in the G1 phase, is governed by the Rb family of proteins (Blagosklonny and Pardee, 2002). When Rb proteins are phosphorylated by cyclin/cyclin-dependent kinase (CDK) complexes, cells exit the R point and become committed to completing the cell cycle. The cells will not respond again to the external environment until the next G1 phase. ESCs lack an R point, presumably due to constitutive cyclin/CDK activity that keeps Rb proteins in a hyperphosphorylated state (Orford and Scadden, 2008;Savatier et al., 1994;Stead et al., 2002;. As a result, ESCs have a short G1 phase and fail to respond to nutrient deprivation and contact inhibition. However, shortly after the initiation of differentiation, the R point is established (Orford and Scadden, 2008;Savatier et al., 1996;. It has been proposed that the cell-cycle structure of ESCs in part underlies their potential to remain undifferentiated and self-renew indefinitely (Burdon et al., 2002;Neganova and Lako, 2008;Singh and Dalton, 2009). Importantly, many transformed somatic cell lines also lack the R point (Blagosklonny and Pardee, 2002). Therefore, understanding the molecular basis of this unique cell-cycle structure is important to both stem cell and cancer biology.
MicroRNAs (miRNAs) play an important role in regulating the cell cycle in ESCs Blelloch, 2009, 2011). miRNAs are short noncoding RNAs that repress protein translation and mRNA stability (Huntzinger and Izaurralde, 2011). Most mature miRNAs arise following two processing steps: DGCR8/DROSHA cleavage of a long pri-miRNA to a hairpin pre-miRNA and DICER cleavage of the pre-miRNA to the mature miRNA (Kim et al., 2009;Winter et al., 2009). Removal of DGCR8, DROSHA, or DICER results in a loss of miRNAs. Dgcr8 and Dicer-null mouse ESCs have a reduced proliferation rate and an altered cell-cycle structure with a slight increase in the fraction of cells in the G1 phase of the cell cycle (Murchison et al., 2005;Wang et al., 2007Wang et al., , 2008. Introduction of individual members from a large family of miRNAs highly expressed in pluripotent stem cells can partially rescue the proliferation defect and reverse the accumulation of cells in G1 (Wang et al., 2008). The family shares the seed sequence (AAGUGCU), a sequence near the 5 0 end of the miRNA that is thought to largely determine a miRNA's downstream targets (Bartel, 2009). The family has eight members, including miR-294 and miR-302a-d, and because of their role in influencing the ESC cell cycle, they have been called the ESCC family of miRNAs (Wang et al., 2008).
The ESCC miRNAs are also promoters of the pluripotent state. When introduced together with the transcription factors OCT4, SOX2, and KLF4 into human or mouse somatic cells, they dramatically enhance the dedifferentiation to generate induced pluripotent stem cells (iPSCs) (Judson et al., 2009;Subramanyam et al., 2011). Indeed, it has been proposed that together with just one or two other miRNAs they induce pluripotency in the absence of any exogenously introduced coding genes (Anokye- Danso et al., 2011;Miyoshi et al., 2011). Consistent with this role, the ESCC miRNAs inhibit another family of miRNAs, the let-7 family, from silencing the pluripotency program of ESCs, thereby promoting their self-renewal . In contrast to the ESCC miRNAs, let-7 is a suppressor of cell-cycle progression (Johnson et al., 2007); however, it is unclear whether the cell-cycle targets alone can explain the ability of ESCC miRNAs to antagonize the effects of let-7. It also remains unknown whether the ESCC miRNAs can suppress other somatic miRNAs from inducing ESC differentiation.
In this study, we make the surprising finding that G1 accumulation seen in Dgcr8 knockout ESCs under normal growth conditions and reversed by the ESCC miRNAs occurs independently of Rb family proteins and, therefore, is not secondary to the classic G1-S check or R point. However, under cytostatic conditions (serum starvation and cell confluency), the cells do show evidence of an R point that is absent in wild-type ESCs. In particular, the Dgcr8 knockout ESCs show a dramatic increase in G1, which is reversed by either the addition of miR-294/ miR-302 or the removal of all three Rb family members. Through a miRNA functional screen, we identify five miRNAs that in addition to let-7 can silence the pluripotency program of Dgcr8 knockout ESCs, but not wild-type cells. All these miRNAs are antagonized by the simultaneous addition of miR-294 and four of them induce an Rb-dependent accumulation in G1 suggesting a direct link between the R point and the silencing of self-renewal. Surprisingly though, the simultaneous deletion of the Rb family genes along with Dgcr8 did not block the six miRNAs from silencing the pluripotency program. Therefore, the ESCC miRNAs are acting through mechanistically separable pathways to promote passage through G1 under normal growth conditions, suppress the R point under cystostatic conditions, and maintain ESC pluripotency.

ESCC miRNAs Promote Passage through G1 Independent of the G1-S Restriction Point
We previously hypothesized that miRNA-deficient ESCs accumulate in the G1 phase of the cell cycle due to low-level activation of the G1-S check/restriction point under normal growth conditions (Wang et al., 2008). To test this hypothesis, we removed all Rb family members (Rb1, Rbl1, and Rbl2) along with Dgcr8 (Figures S1A-S1D). The Rb family proteins are obligate regulators of the G1-S R point (Blagosklonny and Pardee, 2002), and, therefore, their loss should remove any accumulation of cells in G1 associated with activation of the R point. Surprisingly, the proliferation rate and the accumulation of cells in G1 was unaffected by the loss of all three Rb family members ( Figures S1E and 1A). Furthermore, introduction of the ESCC miRNAs still increased proliferation and decreased the fraction of cells in G1 in the Dgcr8, Rb1, Rbl1, Rbl2 quadruple-knockout cells (Figures 1A and S1F). Therefore, the ability of the ESCC miRNAs to promote proliferation and suppress G1 accumulation under normal growth conditions is independent of their action on the G1 R point.

ESCC miRNAs Repress the Ability of ESCs to Respond to Nutrient Starvation and Cell-Cell Contact
To more directly evaluate the role of miRNAs in regulating the R point in ESCs, we evaluated the response of Dgcr8-null ESCs to nutrient-starvation and cell-confluency conditions, well-known activators of this checkpoint. The cell-cycle profiles of wildtype and Dgcr8-null cells were compared in high and low concentrations of fetal bovine serum (FBS) as well as in increasing density of plated cells. As previously reported (Schratt et al., 2001), wild-type ESCs did not accumulate in G0/G1 in response to low FBS or to increasing cell number (Figures 1B and 1C). In contrast, Dgcr8-null ESCs showed a striking increase in cells in G0/G1 under both conditions ( Figures 1B and 1C).
An early event in wild-type ESC differentiation is an accumulation of cells in G1 ( Figure S2A) (Savatier et al., 1996;. To rule out the possibility that nutrient starvation was inducing differentiation of Dgcr8 knockout cells and hence secondarily leading to G1 accumulation, we evaluated multiple markers of pluripotency. Oct4, Nanog, and Klf4 mRNA levels remained high in wild-type and Dgcr8-null ESCs in low serum conditions for 2 days ( Figure S2B). Indeed, Oct4 and Nanog mRNA were slightly up in the knockouts as previously described  likely secondary to the absolute block in differentiation of these cells as well as low-level expression of miRNAs in WT ESCs that normally repress these factors to a small degree (Tay et al., 2008a(Tay et al., , 2008b. Similarly, OCT4 and NANOG immunohistochemistry as well as alkaline phosphatase activity remained positive in low serum ( Figures S2C and S2D). Another alternative explanation for the accumulation of cells in G0/G1 is selective apoptosis of Dgcr8-null cells in non-G0/G1 phases of the cell cycle. Indeed, both wild-type and Dgcr8-null cells showed increased levels of apoptosis in 1% FBS (Figure S3). To remove apoptosis, we combined the Dgcr8 knockout with null alleles of Bax and Bak (Wei et al., 2001). Deletion of both alleles of Bak and one allele of Bax blocked apoptosis in both Dgcr8 heterozygous and Dgcr8 homozygous-null ESCs ( Figures  S3A-S3E). However, the percentage of Dgcr8 -/-/Bak -/-/Bax -/flox ESCs in the G0/G1 phase still showed an increase in response to serum starvation, unlike their Dgcr8 -/flox /Bak -/-/Bax -/flox counterparts ( Figure 1D), showing that apoptosis cannot explain the G0/G1 accumulation.
To rule out non-miRNA roles for the Dgcr8-null phenotype, we evaluated Dicer knockout cells. Similar to Dgcr8, Dicer knockout ESCs accumulated in G0/G1 in response to low serum (Figure S4A). Furthermore, acute deletion of Dgcr8 led to a similar increase in the percentage of cells in G0/G1 upon serum starvation, ruling out adaption to Dgcr8 loss as the underlying cause ( Figure S4B). Together, these data suggest that miRNAs normally suppress the ability of ESCs to pause in G0/G1 in response to external cues such as nutrient starvation and cellcell contact.

The ESCC miRNAs Act through the Rb Pathway to Suppress the G1 Restriction Point in ESCs
The ESCC miRNAs target activators of the G1/S R point including Cdkn1a, Rb1, and Rbl2 (Wang et al., 2008). To directly evaluate whether they can suppress the G1 restriction point, we measured the impact of the ESCC miRNA miR-294 on G0/G1 accumulation under serum starvation and increasing cell density. Introduction of miR-294 mimic into Dgcr8 knockout ESCs blocked the increase in the fraction of cells in G0/G1 under both conditions (Figures 2A and 2B). This block was dependent on the ESCC family seed sequence. Mutation of the seed or the introduction of other ESC expressed miRNAs that have a different seed sequence had no effect on the accumulation of G0/G1 cells ( Figure S5A). The effect was stable for the lifespan of the miRNA mimic ( Figure S5B). These findings show that the ESCC miRNAs suppress the accumulation of cells in G0/G1 in response to cytostatic growth conditions.
To confirm that the ESCC miRNAs are targeting the cyclin/Cdk pathway under cystostatic conditions, we evaluated the previously described targets Cdkn1a, Rb1, and Rbl2 (Wang et al., 2008). All three were elevated in Dgcr8 knockout relative to wild-type ESCs under standard culture conditions ( Figure 3A, p < 0.0002). Upon serum starvation, the three genes remained repressed in wild-type ESCs (p < 0.01). In Dgcr8 knockout cells, Rb1 and Rbl2 remained elevated, whereas Cdkn1a was further elevated (p < 0.002). Rbl1 was also elevated in the knockout cells in both culture conditions, but this effect was independent of the 3 0 UTR as luciferase assays on the 3 0 UTRs of the three Rb genes showed that only the 3 0 UTRs of Rb1 and Rbl2 were suppressed in the wild-type relative to Dgcr8 knockout cells ( Figure S6A). We confirmed that protein levels encoded by the three Rb genes are higher in Dgcr8 knockout cells under both standard culture and serum-starvation conditions ( Figure S6B). The reintroduction of miR-294 was able to suppress Cdkn1a, Rbl1, and Rbl2, but not Rb1, under serum-starvation conditions, which is consistent with indirect effects on Rb1 control ( Figure 3B). These data show that miR-294 suppresses Cdkn1a and Rbl2 in Dgcr8 knockout ESCs under nutrient-starvation conditions.
The above data suggested that the R point is activated in Dgcr8-null cells in cytostatic conditions. To evaluate the role of the R point directly, we compared the Dgcr8 knockout to the Dgcr8, Rb1, Rbl1, Rbl2 quadruple-knockout ESCs. Deletion of all three Rb family genes blocked the response of the Dgcr8 knockout cells to serum starvation ( Figure 3C). Notably, the presence of even one Rb WT allele was enough to maintain the R point. Similar results were obtained under increased cell-confluency conditions ( Figure 3D). These findings show that unlike the G1 accumulation seen in normal growth conditions, the accumulation seen in cytostatic conditions is Rb dependent and, therefore, is due to activation of an R point, which is normally suppressed by the ESCC miRNAs in wild-type ESCs.

Multiple miRNAs Can Induce Differentiation and G1
Accumulation in miRNA-Deficient Cells Let-7 is able to silence self-renewal of Dgcr8 knockout cells but is antagonized by the miR-294/miR-302 family in wild-type embryonic stem cells . Let-7 levels increase during ESC differentiation coinciding with a decrease in the ESCC miRNAs and an increase in the number of cells in G1 ) ( Figure S2A). Therefore, we predicted that let-7 may in part function by lengthening the G1 phase. Indeed, the introduction of let-7 led to an accumulation of Dgcr8 knockout cells in the G1 phase, which was reversed by cointroduction of miR-294 ( Figures 4A and 6A). Based on this finding, we hypothesized that there may be additional miRNAs that can induce an accumulation of cells in G1 and silence self-renewal in Dgcr8 knockout cells and whose function is normally antagonized by the ESCC miRNAs in wild-type ESCs.
To determine if additional miRNAs are able to silence selfrenewal, we performed a screen reintroducing 256 different miRNAs individually into the knockout cells ( Figure 4B; Table  S1). To follow the silencing of self-renewal, we qualitatively scored the degree of loss of alkaline phosphatase (AP) activity on a scale from 1 to 8 with 8 being complete loss of staining (Figure 4C). We uncovered 32 miRNAs that decreased the number of cells showing AP activity by approximately 75% or greater (score R6). To narrow down the miRNAs followed up from the screen, we performed microarray analysis of miRNAs under two differentiation conditions (minus LIF and retinoic acid) ( Figure 4D; Table S2). Fourteen miRNAs with a differentiation score R6 were upregulated under both differentiation conditions. A subset of these miRNAs was confirmed by reverse tran- scriptase-quantitative PCR (RT-qPCR) ( Figure S7). To focus on a small number, we combined our profiling data with published profiling data for human ESC embryoid body differentiation, mouse embryonic fibroblasts, and mouse neural progenitor cells ( Figure 4E) (Bar et al., 2008;Marson et al., 2008). We selected five miRNAs in addition to let-7 that had a differentiation score R6 and were up in two or more differentiation conditions. These were miR-26a, miR-99b, miR-193, miR-199a-5p, and miR-218. To further evaluate the role of these miRNAs in silencing selfrenewal, we tested resynthesized mimics for a representative let-7 family member (let-7c) and five of the newly identified miRNAs. miR-134 was also tested because it had been previously described in inducing differentiation of wild-type mouse ESCs (Tay et al., 2008b). All six miRNAs identified in our screen robustly silenced AP activity, as did miR-134 (Figure 5Ai). RT-qPCR for three additional markers of the pluripotency program, Oct4, Sox2, and Nanog, also showed robust silencing with the screen-positive miRNAs and miR-134 ( Figure 5Bi). Together, these experiments show that multiple miRNAs in addition to let-7, which are upregulated upon ESC differentiation, can silence the self-renewal program in Dgcr8 knockout ESCs.
Similar to let-7, none of the newly uncovered miRNAs were able to silence the pluripotency program in wild-type cells (Figures 5A and 5Bii). miR-134 was able to suppress AP activity and to a small degree Oct4, Sox2, and Nanog levels in wildtype cells. Let-7 and the screen-positive miRNAs were all antagonized by miR-294 and miR-302b. Specifically, the cointroduction of miR-294 or miR-302b, but not mutant miR-294, blocked the ability of miRs-26a, 99b, 193, 199a-5p, and 218 to downregulate AP activity or Oct4, Sox2, and Nanog mRNA levels ( Figures 5A and 5Bi). Although miR-294 and miR-302b were unable to suppress miR-134's ability to silence AP activity, it did inhibit, albeit to a smaller degree, the downregulation of Oct4, Sox2, and Nanog, consistent with the findings when miR-134 was introduced into wild-type cells. These findings suggest that the ESCC miRNAs are general inhibitors of miRNA-induced differentiation rather than being a specific antagonist of let-7.
Next, we asked whether the screen-positive miRNAs, like let-7, were able to induce the accumulation of cells in G1. Indeed, four of the five newly uncovered differentiation-inducing miRNAs increased the fraction of cells in G1 when introduced into Dgcr8 knockout cells ( Figure 6A, p < 0.02 except miR-199a-5p, p = 0.7). Similar to let-7, the effects of these miRNAs on G1 accumulation were reversed by the introduction of the ESCC miRNA, miR-294 ( Figure 6A). Therefore, differentiation-associated miRNAs both silence the pluripotency program and induce accumulation of cells in G1, both of which are antagonized by the ESCC miRNAs.
The ESCC miRNAs Act Independently of the Rb Pathway to Antagonize Other miRNAs from Silencing Self-Renewal Based on the above findings, we hypothesized that wild-type ESCs were resistant to the effects of the differentiation-inducing miRNAs because of the capacity of miR-294/miR-302 family to suppress the R point. To test this hypothesis, we introduced All results shown as mean ± SD, n = 3. See also Figure S6. the differentiation-inducing miRNAs into Dgcr8, Rb1, Rbl1, Rbl2 quadruple-knockout ESCs and evaluated their effects on both the accumulation of cells in G1 and the silencing of self-renewal. The loss of all three Rb proteins blocked the ability of the differentiation-inducing miRNAs to cause an accumulation of cells in G1 ( Figure 6B), consistent with these miRNAs activating the R point. In striking contrast, the loss of all three Rb proteins did not inhibit the ability of these miRNAs to silence self-renewal. In particular, the six differentiation inducing miRNAs were equally effective at suppressing Oct4, Nanog, and Sox2 levels in the quadruple knockouts as the Dgcr8 alone knock out ESCs ( Figures 5B and 6C). These findings demonstrate that the ability of the differentiation-inducing miRNAs to silence self-renewal is independent of their capacity to induce the R point.

DISCUSSION
ESCs have a unique cell-cycle structure lacking a G1 restriction point, which has been hypothesized to play an important role in the maintenance of pluripotency (Burdon et al., 2002;Neganova and Lako, 2008;Orford and Scadden, 2008;Singh and Dalton, 2009). Here, we show that a family of miRNAs, the ESCC miRNAs, suppresses the R point as measured by the response of ESCs to cytostatic conditions (serum starvation and increased cell confluency). Furthermore, using genetic tools, we show that the miRNAs are acting upstream of the Rb family of proteins. Surprisingly, the previously reported G1 accumulation seen in Dgcr8 knockout cells grown under normal growth conditions occurs independently of the G1-S restriction point. Similarly, the ability of the ESCC miRNAs to antagonize other miRNAs (C) miRNA screen data plotted for individual miRNAs with the error representing the range of scores for n = 3. (D) A scatterplot depicting the results for individual miRNAs based on miRNA array data in mouse ESCs 4 days after LIF withdrawal or in 1 mM all-trans-retinoic acid (n = 3 for each condition). Red dots show miRNAs with a screen score greater than or equal to 6. (E) A heatmap depicting miRNA expression changes in mouse NPC and MEF relative to mouse ESC, human EB differentiation, and mouse-LIF and RA differentiation. miRNAs labeled in green were previously implicated in ESC differentiation, whereas those in red were chosen for further investigation in this study. See also Figure S7 and Tables S1 and S2. from silencing ESC self-renewal occurs independently from their ability to suppress the R point. Therefore, the abilities of the ESCC miRNAs to promote the G1-S transition under normal growth conditions, suppress the R point under cystostatic conditions, and block ESC differentiation are mechanistically separable (Figure 7).
The role of Rb family of proteins in the regulation of the G1/S transition has been well studied but remains complex. Knockout of the three Rb genes removes the R point in mouse embryonic fibroblasts in response to serum starvation and cellular conflu-  -7c, miR-26a, miR-99b, miR-193, miR-199a-5p, and miR-218 alone or in combination with miR-294, mutant-miR-294, or miR-302b. (B) qRT-PCR for Pou5f1/Oct4, Sox2, and Nanog normalized first to beta-actin then to mock transfection after miRNA introduction as in (A). Results shown as mean ± SD, n = 2. ency (Sage et al., 2000). However, their loss has little effect on ESCs (Wirt et al., 2010), presumably because the R point is already suppressed by the ESCC miRNAs (this study). Interestingly, loss of the Rb family members does not block the accumulation of many cell types in G1 during ESC differentiation (Wirt et al., 2010). That is, cells can still exit the cell cycle, and early organogenesis proceeds largely undisturbed following injection of triple-knockout ESCs into early embryos. Therefore, there must be multiple mechanisms associated with G1 accumulation seen during ESC differentiation. In particular, considering their differing dependencies on the Rb family, the mechanism that controls the R point that responds to serum starvation or cellular confluency must be different from the mechanisms that mediate G1 accumulation in normal growth conditions and during ESC differentiation. It will be important in future studies to determine if the accumulation seen in Dgcr8 knockout cells under normal growth conditions share common mechanisms with the Rb-independent pathways seen with differentiation.
Previously, we had found that the ESCC and let-7 miRNAs antagonize one another in the switch between ESC self-renewal and differentiation. Here, we describe five additional miRNAs that promote the silencing of ESC self-renewal and are antagonized by the ESCC miRNAs. Interestingly, each of these miRNAs has a distinct seed sequence from let-7 with very different predicted or experimentally determined downstream targets (TargetScan) (C.M. and R.B., unpublished data). Furthermore, unlike let-7, these additional miRNAs have more restricted expression patterns, some being expressed in a small number of tissues (Landgraf et al., 2007). Therefore, their functional targets in silencing self-renewal are likely quite different from one another, which will be the focus of future studies. The ability of the ESCC miRNAs to suppress these alternative mechanisms would suggest a more global role for the ESCC miRNA in promoting the pluripotent state. Such a conclusion fits well with the ability of the ESCC miRNAs to dramatically enhance the dedifferentiation of somatic cells to induced pluripotent stem cells (Judson et al., 2009). How the ESCC miRNAs are able to maintain self-renewal in the presence of differentiation-inducing miRNAs or promote the dedifferentiation of somatic cells remains an open question. Several genomic studies have shown that these miRNAs target hundreds of mRNAs (Hanina et al., 2010;Leung et al., 2011;Sinkkonen et al., 2008). Functional analysis of a small number of targets chosen based on known roles for the encoded proteins has begun to give some insight into their impact on reprogramming to iPSCs (Liao et al., 2011;Subramanyam et al., 2011). However, knockdown of individual members thus far has failed to recapitulate the full effects of the ESCC miRNAs (Subramanyam et al., 2011). Further, our attempts to recapitulate ESCC function by knocking down a small number of individual targets failed to block differentiation by let-7 and the other miRNAs (C.M. and R.B., data not shown). A more systematic dissection of the ESCC miRNA targets including combinatorial suppression of multiple targets will be required to understand how the ESCC miRNAs can have such a powerful impact on stabilizing and promoting the pluripotent state. Figure 6. ESCC miRNAs Act Independently of the Rb Pathway to Antagonize Other miRNAs from Silencing ESC Self-Renewal (A) A bar graph depicting fraction of cells in the G0/G1 phase after transfection of Dgcr8 knockout ESCs with let-7c, miR-26a, miR-99b, miR-193, miR-199a-5p, and miR-218 alone or in combination with miR-294. Mean ± SD for n = 2-5. **p < 0.001. (B) A bar graph depicting fraction of cells in the G0/G1 phase after transfection of Dgcr8 knockout and Rb family/Dgcr8 quadruple knockout ESCs with let-7c, miR-26a, miR-99b, miR-193, miR-199a-5p, and miR-218. Mean ± SD for n = 4. *p < 0.05. (C) qRT-PCR for Pou5f1/Oct4, Sox2, and Nanog normalized first to beta-actin then to mock transfection after miRNA introduction in (i) near Rb family and Dgcr8 knockout and (ii) Rb family and Dgcr8 knockout ESCs. Mean ± SD for n=3.

EXPERIMENTAL PROCEDURES
Tissue Culture, Cell-Cycle, Apoptosis, and Luciferase Reporter Analysis Mouse ESCs were grown on gelatin-coated plates or irradiated mouse embryonic fibroblast feeders as previously described . For serum-starvation experiments, 150,000 or 300,000 cells for wild-type and Dgcr8 knockout ESCs were plated in a well of a 6-well plate. After growing for $24 hr in standard culture media (15% FBS), for standard culture control, cells were then fixed for cell-cycle analysis as previously described (Wang et al., , 2008; for serum-starved samples, media was replaced with 1% FBS culture media, and cells were grown for another 24 or 48 hr before fixation for cell-cycle analysis. For serum starvation of miRNA-transfected cells, media were replaced with low serum media $24 hr after transfection. For contact inhibition experiments, 200,000-1,600,000 cells were plated in a well of a 24-well plate and grown in standard culture media for $24 hr before fixation for cell-cycle analysis. For contact inhibition of miRNA-transfected cells, cells were trypsinized and plated at increasing densities $24 hr after transfection. Notably, all cell-cycle analyses were internally controlled because there was slight variability between cell-cycle distribution under differing conditions including day of experiment, location of experiment, and specifics of experiment. The sequence for the control mimic in this study is UUCUCCGAACGUGUCACGUTT, which is obtained from Shanghai Genepharma and predicted to have no homology with mouse genes. For apoptosis analysis, cells were labeled with propidium iodide and FITC-Annexin V and analyzed by flow-cytometry. Fraction of PI-negative and Annexin-V-positive cells in total population was calculated to track early stage of apoptosis. miRNA transfection and luciferase reporter assay were performed essentially the same as previously described (Wang et al., 2008). For luciferase assay, 4,000 cells for wild-type and 8,000 cells for Dgcr8 knockout ESCs were plated in 96-well plate in either the standard media or 1% FBS media. After growing for $16 hr, reporter constructs were transfected and cells were grown in respective media for $36 hr before lysis. For alkaline phosphatase staining and qPCR of differentiation markers, 6,000 wild-type or 20,000 Dgcr8 knockout cells were plated in a 24-well plate, transfected the next day, and fixed or lysed in Trizol on the fourth day. For cell-cycle experiments with differentiation inducing miRNAs, 50,000 Dgcr8 knockout or quadruple-knockout cells were plated in a 12-well plate; the next day they were transfected and on the second day harvested for cellcycle analysis.
RNA Extraction, qPCR, and miRNA RT-PCR Total RNA was extracted according to standard Trizol protocol (Invitrogen). Samples were centrifuged at >12,000 3 g during washing steps to preserve small RNAs. qPCR and miRNA RT-PCR were performed using Sybr Green mix (Applied Biosystems) as previously described. Sequences for qPCR primers were previously reported Wang et al., 2008). miRNA qPCR was performed by polyadenylating the miRNAs and then using a modified oligodT reverse transcription primer as described previously (Shi and Chiang, 2005).

Alkaline Phosphatase Staining and Immunofluorescence Analysis
Cells were fixed in 4% paraformaldehyde for 10-15 min. Alkaline phosphatase staining was performed using kit from Vector Laboratories according to the manufacturer's protocol. For immunofluorescence analysis, primary antibodies for OCT4 (Santa Cruz Biotechnology, sc-9081) and NANOG (Calbiochem, sc-1000) were diluted 200-to 400-fold. Secondary antibodies were Alex-Fluor-conjugated antibodies from Invitrogen.

miRNA Screening
On day 0, 4,000 Dgcr8 knockout ESCs were plated per well of a 96-well plate in media without LIF. On day 1, miRNA mimics were transfected one per well at a concentration of 100 nM final volume 100 ml. On day 4, cells were fixed and stained for alkaline phosphatase activity using Vector Red substrate (Vector Labs). The screen was repeated in triplicate, and scoring was done manually and blind to the layout of the screening plate.
miRNA Microarray Analysis Cells (400,000) were plated in a 6 cm plate and differentiated either in media without LIF or equivalent media with 1 mM all-trans-retinoic acid (Sigma-Aldrich). RNA was isolated using Trizol (Invitrogen). Total RNA (2 mg) was labeled with an Exiqon miRCURY LNA microRNA Power Labeling Kit (Exiqon) following the manufacturer's protocol for manual hybridization. All hybridiza-tions were dual labeled using day 0 undifferentiated ESC RNA as a reference. One array in each case for (À)LIF and RA differentiation was repeated in reverse color. Arrays were scanned, and data were extracted using a GenePix Scanner (Molecular Devices) and associated software. Data were discarded for spots where at least one color was not 2-fold above background. Background was subtracted for individual spots and quadruplicate spots from each array were averaged. Data at this point were manually filtered to remove data for inconsistent values within quadruplicate spots. Arrays were median centered based on the median of probes with meaningful data on all three arrays (i.e., commonly unfiltered probes). The average of data for all probes passing analysis is given in Figure 4 for probes present both on the RA and (À)LIF arrays. See also Figure S7.
miRNA Expression Meta Analysis Human ES and EB deep sequencing data were downloaded from the supplementary materials of Bar et al. (2008). Mouse neural progenitor cell (NPC), mouse embryonic fibroblast (MEF), and ESC deep sequencing data were downloaded from the supplementary material of Marson et al. (2008). Data were manipulated in Microsoft Excel, clustered in Cluster (http://rana.lbl. gov/EisenSoftware.htm), and visualized in Java TreeView (http://jtreeview. sourceforge.net/).

Animal Use
All animal experiments described in this study were approved by the University of California, San Francisco's Institutional Animal Care and Use Committee.

SUPPLEMENTAL INFORMATION
Supplemental Information includes seven figures and two tables and can be found with this article online at http://dx.doi.org/10.1016/j.celrep.2013.05.027.

ACKNOWLEDGMENTS
We would like to thank members of the Blelloch laboratory, especially Raga Krishnakumar, for critical reading of the manuscript. Y.W. was supported by Figure 7. ESCC miRNAs Suppress the R Point and Silencing of Self-Renewal through Different Mechanisms ESCC miRNAs suppress the R point and G1 accumulation induced by differentiation-inducing miRNAs through targeting the Rb pathway. However, knocking out Rb family proteins is not sufficient to prevent silencing of self-renewal by differentiation-inducing miRNAs. Therefore, other pathways must be regulated by ESCC miRNAs to antagonize silencing of self-renewal by differentiation-inducing miRNAs. Additionally, ESCC miRNAs regulate Rb-independent pathways to promote G1/S transition and proliferation at standard culture conditions, because triple knockout of Rb family proteins neither prevent G1 accumulation nor promote proliferation rate in Dgcr8 knockout ESCs. funds from the Chinese Ministry of Science and Technology (2011CBA01100 and 2012CB966700), the Institute of Molecular Medicine (Peking University), and a California Institute of Regeneration Medicine (CIRM) postdoctoral training award (T1-00002). C.M. was supported by a National Science Foundation predoctoral training award, A.S. was supported by a CIRM predoctoral training award (TG2-01153), and R.B. was supported by the National Institutes of Health (R01 NS057221 and R01 GM101180) and CIRM (RN2-00906-1). Y.W., Y.-P.L., and X.-X.Z. performed experiments for Figures 1, 2, 3, and S1-S5. C.M. performed experiments for Figures 4, 5, 6, S6, and S7. D.S. helped with experiments in Figure 5. A.S. produced Dgcr8/Bax/Bak ESC lines and helped with experiments in Figure 6. All authors were involved in interpretation of data. R.B., Y.W., and C.M. conceived the project and wrote the manuscript.