p27Kip1 Directly Represses Sox2 during Embryonic Stem Cell Differentiation

1Differentiated cells can be converted into induced pluripotent stem cells (iPSCs) through the combined action of transcription factors, most notably OCT4, KLF4, and SOX2. Importantly, the mechanisms involved in this process might provide clues about the molecular mechanisms governing stem cell biology and cancer. Recently, we and others have shown that tumor suppressors, such as those encoded by the p53 gene and the Ink4a/Arf locus, oppose reprogramming and limit the efficiency of the process.

2The tumor suppressor p27Kip1 binds and inhibits multiple cyclin-dependent kinases. Importantly, low protein levels of p27 constitute a poor prognosis marker for several types of cancer and germline mutations of the p27 gene (also known as CDKN1B) are responsible for a subset of human multiple endocrine neoplasia (MEN) syndromes, notably characterized by pituitary tumors. The cyclin-dependent kinase 2 (CDK2) is one of the main CDKs inhibited by p27. Paradoxically, however, the main phenotypes of p27 null mice, namely, increased body size, organ hyperplasia, pituitary tumors, and retinal dysplasia, are not rescued by concomitant deletion of Cdk2, thus suggesting that these p27 null phenotypes are not primarily caused by uncontrolled CDK2 activity.

3In the context of investigating the role of tumor suppressors during reprogramming, we studied p27 null cells and we noticed that these cells can be reprogrammed into iPSCs without ectopic expression of Sox2. This observation led us to explore the potential link between these two previously unrelated proteins, p27 and SOX2.

4While investigating the effect of tumor suppressor genes on the process of reprogramming to induced pluripotent stem cells (iPSCs) by the three Yamanaka factors (Oct4, Klf4, and Sox2), we tested the three possible combinations of two factors (abbreviated as 2F-OK, 2F-OS, and 2F-KS in reference to Oct4, Klf4, and Sox2) in a series of primary mouse embryo fibroblasts (MEFs) lacking cell cycle regulators and tumor suppressors. After repeated attempts, we were unable to obtain alkaline-phosphatase-positive (AP+) colonies in any of the tested MEFs using 2F-OS or 2F-KS. Interestingly, however, p27 null MEFs and, to a lesser extent, p130 null MEFs gave rise to AP+ colonies with 2F-OK (Figures 1A and 1B). Absence of the p27-related protein p21 also produced AP+ colonies and further increased the number of AP+ colonies when combined with p27 deficiency, thus suggesting some degree of functional redundancy between p27 and p21. In all these MEFs, the emergence of visible AP+ colonies was delayed compared to the standard three-factor cocktail (3F-OKS) (4 weeks versus 2 weeks) and the average efficiency was about 100-fold lower (9 × 10−5 in p27 null/2F-OK versus 8 × 10−3 in WT/3F-OKS). In contrast to this, WT MEFs or MEFs deficient in p53, Arf, or Ink4a/Arf could not be reprogrammed by 2F-OK despite the fact that these cells are reprogrammed with very high efficiency by 3F-OKS. Also, absence of p27 had a modest stimulatory effect on 3F-OKS reprogramming (Figure S1A available online). Together, these observations suggest that the absence of p27 selectively renders cells susceptible to reprogramming in the absence of ectopic Sox2. The p27 null/2F-OK AP+ colonies were confirmed to be bona fide iPSCs based on their expression of endogenous pluripotency genes (Nanog, Sox2, and Oct4; Figure S1B), production of teratomas (Figure S1C), and efficient contribution to chimeric mice (Figure 1C). To further validate the use of alkaline phosphatase as a marker of reprogramming under our experimental conditions (most notably characterized by the absence of ectopic c-Myc and by the use of serum-free medium), we obtained WT and p27 null MEFs carrying a transgenic GFP reporter under the Sox2 promoter, and we observed that >90% of the AP+ colonies were GFP+. All together, these results indicate that the absence of p27 eliminates the absolute requirement for ectopic Sox2 in reprogramming.

5Mouse fibroblasts express low, but detectable, levels of Sox2, and therefore, we wondered whether p27 deficiency affected Sox2 expression. Indeed, p27 null MEFs had a significant increase in Sox2 mRNA levels compared to WT controls (7-fold), although these levels were still about 64-fold lower than in embryonic stem cells (ESCs) (Figure 1D). We also detected SOX2 protein by immunofluorescence. Consistent with the mRNA data, quantitative image analyses indicated that SOX2 protein levels were globally increased in p27 null MEFs (Figure 1E). Of note, the distribution of SOX2 fluorescence intensity in p27 null MEFs is broad, and therefore, it is conceivable that only those p27 null cells with the highest SOX2 levels are the ones susceptible of 2F-OK reprogramming. It is also worth mentioning that WT and p27 null MEFs have the same proliferative rate, thus implying that their different SOX2 levels are not secondary to a different proliferative activity. We tested whether the increased levels of SOX2 observed in p27 null MEFs could be reverted by ectopic overexpression of p27. Interestingly, quantitative immunofluorescence of p27-overexpressing p27 null MEFs indicated a downregulation of SOX2 levels (Figure 1E). Finally, we also observed higher Sox2 mRNA levels in the retina, brain, and lung of adult p27 null mice (Figure 1F). Together, these data indicate that p27 contributes to the silencing of Sox2 in differentiated cells and tissues.

6To further define the repressive role of p27 on Sox2, we analyzed the differentiation of pluripotent stem cells upon treatment with retinoic acid (RA). This differentiation protocol efficiently reduces SOX2 protein and mRNA concomitantly with a dramatic upregulation of p27 (Figure S2A). The upregulation of p27 during RA-induced differentiation is in agreement with a previous report on differentiating human embryonic carcinoma cells. We wondered whether the absence of p27 would affect the kinetics of Sox2 repression. For this, we generated iPSCs derived from WT or p27 null MEFs after reprogramming them with Oct4, Klf4, and Sox2. The levels of Sox2 mRNA were similar in undifferentiated WT or p27 null iPSCs (Figure 2A). However, upon RA-induced differentiation, Sox2 levels in p27 null iPSCs were not reduced as efficiently as in WT iPSCs (Figure 2A). Similarly, we tested Sox2 expression during iPSC differentiation into embryoid bodies (EBs). Again, Sox2 mRNA levels were abnormally high in p27-deficient EBs compared to WT ones (Figure 2B). These observations are in agreement with a previous study reporting numerous abnormalities in EBs from p27 null ESCs. We considered the possibility that the increased Sox2 levels in p27 null EBs could reflect a skewed neural differentiation, but the levels of Nestin mRNA were similar in WT and p27 null EBs (Figure S2B). To directly test the repressive activity of p27, we retrovirally transduced ESCs with a pMSCV vector expressing p27, and interestingly, p27 overexpression was able to reduce the levels of Sox2 to an extent comparable to RA (Figure 2C). Previous studies have demonstrated that Sox2 null ESCs spontaneously differentiate into trophectoderm-like cells. We asked whether p27 overexpression in ESCs promotes the trophectoderm-like differentiation characteristic of Sox2 null ESCs. Indeed, this was the case and p27 overexpression selectively induced trophectoderm markers to a similar extent as RA differentiation, whereas ectoderm, endoderm, and mesoderm markers were not induced by p27 but were induced by RA (Figure 2C). Moreover, p27-overexpressing ESCs produced cells with giant trophoblast-like morphology, while this type of cell was absent in RA-differentiated ESCs (Figure S2C). To extend the generality of these findings to other pluripotent cells, we overexpressed p27 in murine P19 embryonal carcinoma (P19EC) cells at levels that did not affect proliferation and found a significant decrease of Sox2 mRNA levels (Figure S2D). Together, these results indicate that p27 exerts a repressive effect on Sox2 that is relevant during differentiation.

7The main regulatory element responsible for the expression of Sox2 in pluripotent stem cells is located ∼4 kb downstream of the single Sox2 coding exon and it is named SRR2. Based on our above observations in MEFs and in differentiating pluripotent cells, we asked whether the presence or absence of p27 had an effect on the repressive epigenetic marks on the Sox2-SRR2 enhancer. Interestingly, we observed that p27 null MEFs present lower levels of the H3K9me3 and H3K27me3 repressive marks at the Sox2-SRR2 enhancer as evidenced by chromatin immunoprecipitation (ChIP) analysis (Figure 3A). Similarly, RA-induced differentiation of WT iPSCs dramatically increased the levels of H3K9me3 and H3K27me3 at the Sox2-SRR2 enhancer, while these epigenetic marks were modestly increased in p27 null iPSCs (Figure 3B). These results indicate that the absence of p27 leads to a defective epigenetic remodeling of the Sox2-SRR2 enhancer both in differentiated cells (MEFs) and during the differentiation of pluripotent cells.

8Based on the above data and the recent report that p27 can associate to gene promoters in association with the repressive complex p130-E2F4-SIN3A, we hypothesized that p27 might be recruited in this manner to the Sox2-SRR2 enhancer. To directly test this, we performed ChIP with anti-p27 antibodies in MEFs and we detected p27 associated to the Sox2-SRR2 enhancer (Figure 3C). As before, we wondered whether this was also the case in differentiating pluripotent cells. Interestingly, RA-induced differentiation of iPSCs was accompanied by a strong recruitment of p27 to the Sox2-SRR2 enhancer (Figure 3C). To further extend these observations, we used ESCs and, as in the case of iPSCs, p27 was immunoprecipated at the Sox2-SRR2 enhancer upon RA-induced differentiation, but not at the Nanog promoter used here as a control (Figure 3D). We sought additional proof by performing ChIP from ESCs transfected with flag-tagged p27 and we also found p27 bound to the Sox2-SRR2 enhancer after immunoprecipitation with antibodies against p27 or against the flag tag (Figure 3E). Similar results were obtained in P19EC cells, both upon RA-induced differentiation (Figure S3A) and upon transfection of flag-tagged p27 (Figure S3B).

9To examine the presence of a p130-E2F4-SIN3A repressive complex at the Sox2-SRR2 enhancer, we performed ChIP assays using antibodies against p130, E2F4, and SIN3A in WT or p27 null iPSCs undergoing RA-induced differentiation. Interestingly, the three proteins were detected in the Sox2-SRR2 enhancer of RA-differentiated cells regardless of the presence or absence of p27 (Figure 3F). We also detected binding of p130, E2F4, and SIN3A to the Sox2-SRR2 enhancer in WT and p27 null MEFs (Figure S3C) and RA-differentiated P19EC cells (Figure S3D). These results indicate that a repressive p130-E2F4-SIN3A complex is assembled at the Sox2-SRR2 enhancer upon differentiation and independently of p27.

10Finally, we asked whether the direct inhibition of the p130-E2F4-SIN3A complex would also result in derepression of Sox2. In agreement with previous reports, depletion of SIN3A had a dramatic effect on cell viability that precluded us from further examining the effect on Sox2 expression. Interestingly, however, knockdown of p130 or E2f4 with shRNAs resulted in a severe reduction of their expression (Figure S3E) and, importantly, this was accompanied in both cases by a significant upregulation of Sox2 expression (Figure 3G). In summary, we conclude that p27 associates to the Sox2-SRR2 enhancer together with the repressive complex p130-E2F4-SIN3A, and together contribute to the repression of Sox2 upon cell differentiation.

11Based on our above data, we wondered whether the incomplete repression of Sox2 observed in p27-deficient cells and tissues could mediate some of the phenotypes characteristic of p27 null mice. To evaluate this in a genetic manner, we generated compound Sox2 heterozygous (Sox2-het) and p27 null mice. Sox2 null mice are not viable, but Sox2-het are viable and display a moderate reduction in body size and hypopituitarism. Interestingly, the characteristic gigantism of p27 null mice was normalized by deletion of one Sox2 allele (Figure 4A; Figure S4A), as was also the case for the pituitary mass in young 3- to 6-month-old mice (at this age, p27 null mice present pituitary hyperplasia, but do not have pituitary tumors yet) (Figure 4B). However, given the fact that Sox2-het mice have a modest, but detectable, defect in growth, the above results leave open the possibility that p27 and Sox2 simply have opposite effects on growth that balance each other. For this reason, we decided to focus in the progenitor cell layer that surrounds the pituitary cleft and which is formed by SOX2-positive (SOX2+) cells. The progenitor layer in Sox2-het pituitaries had the same thickness as in WT pituitaries (Figures 4C and 4D; Figure S4B). Interestingly, the thickness of the progenitor layer was significantly increased in p27 null mice compared to WT or to Sox2-het mice, and this defect was absent in Sox2-het/p27-null mice (Figures 4C and 4D). The progenitor cells of the pituitary have been proposed to constitute the origin of pituitary adenomas. In this regard and in line with our above observations, the incidence of pituitary tumors was significantly reduced in Sox2-het/p27-null mice compared to p27 null littermates (Figure 4E).

12Having established that Sox2 heterozygosity rescues the gigantism and the pituitary phenotypes of p27 null mice, we wondered whether the same was true for the retinal defects of p27 null mice. In agreement with previous reports, the retinas of Sox2-het mice were normal (Figure S4C), while p27 null retinas presented focal protrusions of the outer nuclear layer (ONL) (Figures 4F and 4G). Interestingly, these protrusions were absent in Sox2-het/p27-null retinas (Figures 4F and 4G). Also, we observed that p27 null retinas present an increased abundance of SOX2+ nuclei at the inner nuclear layer (INL) and mislocalized SOX2+ nuclei in the outer plexiform layer (OPL) (Figure S4C). We quantified the number of SOX2+ nuclei in complete retinal sections and confirmed that p27 null retinas present an excess of SOX2+ nuclei and, importantly, we found that this defect was absent in Sox2-het/p27-null retinas (Figure 4H). We wanted to corroborate the increased abundance of SOX2+ nuclei in the absence of p27, and for this we used transgenic mice with GFP under the control of a Sox2 promoter region that marks neural multipotent progenitors associated to Sox2 expression (Sox2-GFP mice). Reinforcing our above observations, p27-het/Sox2-GFP retinas presented a significant increase in GFP+ nuclei when compared with WT/Sox2-GFP retinas (Figure S4D). Together, we conclude that a decrease in the gene dosage of Sox2 rescues the main phenotypes associated with p27 deficiency, namely, gigantism, pituitary hyperplasia, pituitary adenomas, and retinal abnormalities. In addition to these phenotypes, p27 null mice also have adrenal gland hyperplasias and tumors (pheochromocytomas) and female sterility. We examined WT adrenal glands and p27 null pheochromocytomas by immunostaining for SOX2, but they were negative and for this reason we did not further pursue this phenotype. Regarding the sterility of p27 null females, Sox2-het/p27-null females remained sterile, suggesting that this phenotype is independent of Sox2. Collectively, these results provide genetic support to the concept that p27 is a negative regulator of Sox2 in the pituitary and in the retina.

13The mechanisms responsible for the transcriptional silencing of pluripotency genes in differentiated cells are poorly understood. The results reported here demonstrate that the tumor suppressor p27 contributes to the transcriptional repression of Sox2. We have observed that the absence of p27 leads to a defective repression of Sox2 in fibroblasts, lung, retina, and brain, and to a delayed and incomplete silencing of Sox2 during differentiation of pluripotent cells, including iPSCs, ESCs, and P19EC cells. These observations led us to identify p27 as a transcriptional regulator of Sox2 together with a repressive complex formed by p130, E2F4, and SIN3A at a critical enhancer responsible for Sox2 expression. These findings are in line with a recent report describing the capacity of p27 to interact with the p130/E2F4/SIN3A complex and contribute to its transcriptional repressive activity. We have found that p27 deficiency leads to an expansion of SOX2+ cells in the progenitor layer of the pituitary and in the retina, which results in pituitary hyperplasia and tumors and morphological defects of the retina. Importantly, these defects are rescued when p27 deficiency is combined with Sox2 heterozygosity. In humans, germline mutations in p27 and SOX2 also affect the pituitary and the retina. On one hand, loss-of-function mutations in p27 produce MEN syndrome, notably characterized by pituitary tumors. On the other hand, loss-of-function mutations in SOX2 produce syndromes characterized by anophthalmia and hypopituitarism. Our current findings unveil a mechanistic connection between p27 and SOX2, and thereby contribute to our understanding of the molecular basis of the human pathologies associated with the deregulation of these two proteins.

14Mice p27 null, Sox2-het, and Sox2-promoter/GFP transgenic have been previously described. All comparisons were made among mice derived from the same sets of crosses, and they therefore shared the same genetic background. Animal experimentation at the CNIO, Madrid was performed according to protocols approved by the CNIO-ISCIII Ethics Committee for Research and Animal Welfare (CEIyBA) and animal experimentation at the MRC-NIMR, Mill Hill, London was carried out in accordance with the UK Animals (Scientific Procedures) Act 1986.

15ChIP and quantitative PCR was performed following standard methods (detailed in the Supplemental Experimental Procedures). PCR primer sequences, shRNA encoding plasmids, and other DNA constructs, as well as antibodies and other standard molecular biology methods, are all detailed in Supplemental Experimental Procedures.

16Reprogramming of primary (passage 2–4) MEFs was performed as previously described by us using plasmids pMXs-Klf4, pMXs-Sox2, or pMXs-Oct4 (obtained from Addgene and previously described; ). For additional details, see Supplemental Experimental Procedures.

17Differentiation with RA was performed essentially as described. ESCs or iPSCs were adapted to grow on gelatin-coated plates (and in the absence of feeder cells). Cells were grown to near confluency in their corresponding complete medium (day 0) and then were trypsinized and seeded at lower density in the absence of LIF for 1 day (day 1). During the following 2 days (days 2 and 3), RA was added at a concentration of 10−6 M, and day 4 cells were without LIF and without RA. In the case of P19EC cells, differentiation was induced by addition of RA (10−6 M) for 4 days.

18For immunohistochemical stainings, quantifications were performed on representative fields at the same magnification, on a minimum of three different areas per sample and a minimum of three different samples per genotype. For immunofluorescence, cells were inspected under a Leica TCS-SP5 confocal microscope (AOBS) and analyzed using Definiens Developer XD 1.5 software, under the same exposure conditions. For additional details, see Supplemental Experimental Procedures.

19Unless otherwise specified (Figures 4E and 4G), quantitative data are presented as mean ± SD and significance was assessed by the two-tailed Student's t test.