A PRC2-Dependent Repressive Role of PRDM14 in Human Embryonic Stem Cells and Induced Pluripotent Stem Cell Reprogramming§


  • Yun-Shen Chan,

    1. Gene Regulation Laboratory, Genome Institute of Singapore, Singapore, Singapore
    2. Graduate School for Integrative Sciences & Engineering , National University of Singapore, Singapore, Singapore
    Search for more papers by this author
  • Jonathan Göke,

    1. Gene Regulation Laboratory, Genome Institute of Singapore, Singapore, Singapore
    2. Department of Computational Molecular Biology, Max Planck Institute for Molecular Genetics, Berlin, Germany
    Search for more papers by this author
  • Xinyi Lu,

    1. Gene Regulation Laboratory, Genome Institute of Singapore, Singapore, Singapore
    2. Department of Biological Sciences, National University of Singapore, Singapore, Singapore
    Search for more papers by this author
  • Nandini Venkatesan,

    1. Division of Molecular Genetics and Cell Biology, School of Biological Sciences, Nanyang Technological University, Singapore, Singapore
    Search for more papers by this author
  • Bo Feng,

    1. School of Biomedical Sciences, The Chinese University of Hong Kong, Hong Kong, China
    Search for more papers by this author
  • I-Hsin Su,

    Corresponding author
    1. Division of Molecular Genetics and Cell Biology, School of Biological Sciences, Nanyang Technological University, Singapore, Singapore
    • Division of Molecular Genetics and Cell Biology, School of Biological Sciences, Nanyang Technological University, SBS-02n-46, 60 Nanyang Drive, Singapore 637551, Singapore
    Search for more papers by this author
    • Telephone: +65-65138687; Fax: +65-67913856

  • Huck-Hui Ng

    Corresponding author
    1. Gene Regulation Laboratory, Genome Institute of Singapore, Singapore, Singapore
    2. Graduate School for Integrative Sciences & Engineering , National University of Singapore, Singapore, Singapore
    3. Department of Biological Sciences, National University of Singapore, Singapore, Singapore
    4. Division of Molecular Genetics and Cell Biology, School of Biological Sciences, Nanyang Technological University, Singapore, Singapore
    5. Department of Biochemistry, National University of Singapore, Singapore, Singapore
    • Genome Institute of Singapore, 60 Biopolis Street, #02-01, Genome Building, Singapore 138672, Singapore
    Search for more papers by this author
    • Telephone: +65-68088145; Fax: +65-68089004

  • Author contributions: Y.C.: conception and design, collection and/or assembly of data, data analysis and interpretation, and manuscript writing; J.G.: collection and/or assembly of data, data analysis and interpretation, and manuscript writing; X.L., N.V., and B.F.: collection and/or assembly of data and data analysis and interpretation; I.S.: provision of study material, data analysis and interpretation, manuscript writing, and final approval of manuscript; H.N.: conception and design, data analysis and interpretation, manuscript writing, and final approval of manuscript.

  • Disclosure of potential conflicts of interest is found at the end of this article.

  • §

    First published online in STEM CELLS EXPRESS December 28, 2012.


PRDM14 is an important determinant of the human embryonic stem cell (ESC) identity and works in concert with the core ESC regulators to activate pluripotency-associated genes. PRDM14 has been previously reported to exhibit repressive activity in mouse ESCs and primordial germ cells; and while PRDM14 has been implicated to suppress differentiation genes in human ESCs, the exact mechanism of this repressive activity remains unknown. In this study, we provide evidence that PRDM14 is a direct repressor of developmental genes in human ESCs. PRDM14 binds to silenced genes in human ESCs and its global binding profile is enriched for the repressive trimethylation of histone H3 lysine 27 (H3K27me3) modification. Further investigation reveals that PRDM14 interacts directly with the chromatin regulator polycomb repressive complex 2 (PRC2) and PRC2 binding is detected at PRDM14-bound loci in human ESCs. Depletion of PRDM14 reduces PRC2 binding at these loci and the concomitant reduction of H3K27me3 modification. Using reporter assays, we demonstrate that gene loci bound by PRDM14 exhibit repressive activity that is dependent on both PRDM14 and PRC2. In reprogramming human fibroblasts into induced pluripotent stem cells (iPSCs), ectopically expressed PRDM14 can repress these developmental genes in fibroblasts. In addition, we show that PRDM14 recruits PRC2 to repress a key mesenchymal gene ZEB1, which enhances mesenchymal-to-epithelial transition in the initiation event of iPSC reprogramming. In summary, our study reveals a repressive role of PRDM14 in the maintenance and induction of pluripotency and identifies PRDM14 as a new regulator of PRC2. STEM CELLS 2013;31:682–692


Pluripotent stem cells can be derived from the inner cell mass of the early blastocyst [1, 2] or can be induced from somatic cells through various approaches [3]. Their unique ability to self-renew indefinitely and differentiate into cells of all three embryonic germ layers is invaluable for therapeutic applications [4]. However, these cells' unique capacity to differentiate into multiple somatic tissue types is also a major hurdle for deriving homogenous population of somatic cells. Thus, a detailed understanding of how cellular plasticity is regulated would be essential for the precise control of the differentiation process. The pluripotent cell state is maintained by a transcriptional regulatory network that activates pluripotency-associated genes and suppresses lineage-specific genes [5]. Sequence-specific transcription factors such as the core embryonic stem cell (ESC) regulators OCT4, SOX2, and NANOG play key roles in regulating a number of downstream pluripotent regulators in maintaining the pluripotency network [6, 7]. Ectopic expression of these pluripotent regulators in somatic cells can reactivate this ESC transcriptional circuit to regain pluripotency; highlighting the critical role of transcription factors in regulating cell fate [3]. In addition to transcription factors, chromatin regulators also play an integral role in the ESC transcriptional network by maintaining the epigenetic landscape of the genome [5]. The best studied of such epigenetic factors are the polycomb group complexes (PcG). PcG are implicated in repressing lineage-specific genes in ESCs [8, 9] and PcG-null mouse ESCs do not activate differentiation program appropriately [10].

Apart from analyzing the transcriptional regulatory networks of ESCs, epigenetic profiling experiments have revealed regulatory regions with distinctive histone modifications [11–15]. Many of the promoters of lineage-specific genes in human and mouse ESCs are marked by the active trimethylated histone H3 lysine 4 (H3K4me3) and the repressive trimethylated histone H3 lysine 27 (H3K27me3), termed bivalent domains [16, 17]. Genes marked by the bivalent domain are thought to be in a poised state, allowing rapid induction of lineage-specific genes when ESCs differentiate. In human ESCs, mapping of coactivator p300 and various histone modifications identified two classes of enhancers [12, 15]. Enhancers that are defined by the presence of p300 binding, monomethylation of histone H3 lysine 4 (H3K4me1), and low nucleosome levels are modified by two different histone H3 lysine 27 modifications. Class I enhancers, which are the active enhancers, are marked by acetylation of histone H3 at lysine 27 (H3K27ac); these enhancers are associated with genes expressed in human ESCs. The class II enhancers, described as “poised” enhancers, lack H3K27ac but are trimethylated at this residue (H3K27me3). Although the genes harboring class II enhancers are inactive in human ESCs, they are involved in early stages of human embryogenesis and development [12]. Hence, understanding the intricate interplay between transcription factors and chromatin modifiers in establishing these unique epigenomic signatures will shed light on how different regulators in ESC cooperatively maintain the pluripotency network.

We have previously conducted a whole-genome RNAi screen to identify novel regulators required to maintain the pluripotent state of human ESCs. PRDM14, a PR domain-containing transcriptional regulator, was identified to be one of the major determinants of the human ESC identity [18]. Depletion of PRDM14 in human ESCs results in the increased expression of developmental genes [18, 19] and the ectopic expression of PRDM14 suppresses the expression of differentiation genes in embryoid bodies [19]. These findings suggest that PRDM14 may act as a transcriptional repressor to regulate pluripotency in human ESCs. In mouse, the repressive activity of Prdm14 was also highlighted in ESCs and germ cells. Prdm14 was shown to be a critical germ cell regulator with a role in establishing pluripotency and epigenetic reprogramming in primordial germ cells (PGCs) [20]. Specifically, PGC-like cells in Prdm14 null mice fail to regain the repressive H3K27me3 mark during development. Prdm14 also has a potential repressive function in mouse ESCs; the overexpression of Prdm14 prevented the induction of extraembryonic endoderm (ExEn) fate in embryoid bodies [21]. Prdm14 was also found to bind and repress target genes when overexpressed in epiblast stem cells (EpiSC) [22].

In this study, we uncover a polycomb repressive complex 2 (PRC2)-dependent repressive function of PRDM14. In human ESCs, PRDM14 binds to many silenced genes and its binding intensity highly correlates with H3K27me3 levels. PRDM14 is found to interact directly with PRC2. PRC2 is detected at PRDM14 binding sites near developmental genes in human ESCs and its recruitment is regulated by PRDM14. The PRDM14/PRC2 bound locus exhibits repressive activity that is dependent on both PRDM14 and PRC2. In the process of induced pluripotent stem cell (iPSC) reprogramming, PRDM14 was found to enhance the suppression of its target developmental genes expressed in fibroblasts and to directly repress a key mesenchymal gene during the initial phase of reprogramming, thereby enhancing the reprogramming process. In summary, we identified PRDM14 as a new regulator of PRC2 and provided insights to how a key ESC regulator functions with chromatin modifiers to maintain and induce pluripotency in human cells.


Cell Culture

The human ESC line H1 (WiCell, Madison, WI) was used in this study. The cells were maintained as feeder-free culture on Matrigel (BD, Franklin Lakes, NJ) [23]. The medium containing 20% KnockOut serum replacement, 1 mM L-glutamine, 1% nonessential amino acids, 0.1 mM 2-mercaptoethanol, and 4 ng/ml basic fibroblast growth factor (Invitrogen) in Dulbecco's modified Eagle's medium (DMEM)/F12 (Invitrogen, Carlsbad, CA) was conditioned with mitotically inactivated mouse embryonic fibroblast for 24 hours.Additional 8 ng/ml of basic fibroblast growth factor (Invitrogen, Carlsbad, CA) is supplemented to conditioned medium before use. Medium was changed daily. The human ESCs were passaged with 1 mg/ml collagenase IV (Gibco, Carlsbad, CA) upon confluence. 293T cell and human embryonic lung fibroblasts MRC-5 (ATCC, Manassas, VA) were cultured in DMEM with 15% fetal bovine serum. All cell cultures are maintained at 37°C with 5% CO2.

Informatics Analysis

ChIP-Seq data were downloaded from the European Nucleotide Archive (Table 1). Class I and class II elements were obtained from Rada-Iglesias et al. [12]. The reads were mapped to the human genome (hg19) using Bowtie (0.12.5) [27]. Peak calling was done with MACS (1.4.0, p value <1e-05, false discovery rate <2%, 4,960 peaks for PRDM14) [28]. Afterward we called peaks on the control only and removed all peaks which could be identified in the control datasets from the results [29]. Peaks were associated with their nearest transcription start site using Peak Analyzer. ChIP-Seq enrichment plots were created by extending the genomic loci of interest to a range of 4,000 bp. We used SamTools [30] to extract the tag number at every position. The plots show the average number of reads (depth) for all positions (±2,000 bp around the center) over all genomic loci of interest, using a sliding window of size 30 bp. Clustering of ChIP-Seq experiments (Fig. 1C) was done on all PRDM14 peaks. We first calculated the sum of tags (±2,000 bp around the peak) for all peaks and all ChIP-Seq experiments. Then we calculated the Pearson correlation coefficients on the log-transformed sums. The heatmap was created using the R function heatmap for symmetric data, which is based on hierarchical clustering of the Euclidean distances calculated from the matrix of all correlation coefficients. Gene expression data were obtained from GEO Profiles (GSE22792). We performed a gene set enrichment analysis (GSEA) using the GSEA software from Subramanian et al. [31] and Mootha et al. [32]. We compared gene expression data 4 days after knockdown of PRDM14. We selected all genes near the 500 most significant PRDM14 binding sites as the gene set for which we calculated the enrichment score.

Table 1. List of ChIP_Seq datasets used in this study
  1. Included is the corresponding accession number (European Nucleotide Archive) and references for each dataset. The EZH2 ChIP_Seq dataset is generated in this study and made available in the public database.

inline image
Figure 1.

PRDM14 binding correlates with H3K27me3 and represses transcription of nearby genes. (A): Expression of genes near the 500 highest scoring PRDM14 peaks, sorted by expression ratio between control and PRDM14 knockdown after 4 days. Colors show expression values from lowest (blue) to highest (red). (B): Gene set enrichment analysis of PRDM14 target genes. Shown is the running-sum statistic for all genes on the array; PRDM14 target genes increase the statistic, nontarget genes decrease it. The maximum deviation from zero indicates the enrichment score (ES). PRDM14 target genes are significantly enriched at the top of the list (KD > control, ES 0.59, p-value = .0, false discovery rate = 0.0). (C): Heatmap showing correlations of PRDM14 binding with binding sites of histone modifications and other transcription factors. The heatmap shows two main clusters; a repressive cluster consisting of H3K9me3, PRDM14, and H3K27me3 and an active cluster consisting of NANOG, OCT4, P300, and active histone marks. Both clusters also show overlapping binding sites, likely reflecting a poised state (black box). (D): Average ChIP-Seq signal for PRDM14 (top) and H3K27me3 (bottom) in human embryonic stem cells. PRDM14 peaks were divided into four equally sized quartiles according to the binding strength (top). H3K27me3 levels are enriched at peaks with strong PRDM14 binding signal (bottom).


The human ESCs and transfected 293T cells were lysed with lysis buffer (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 10 μM ZnCl, 0.5% Nonidet P40, and 5% glycerol with protease inhibitor) for 30 minutes at 4°C. The whole cell lysate was precleared, collected, and incubated overnight with protein G beads coated with antibodies at 4°C. The beads were washed four to six times with the cell lysis buffer and boiled 10 minutes for elution. The interacting protein bands are resolved with 10% SDS-PAGE gel and transferred to the polyvinylidene fluoride membrane, followed by detection with an appropriate primary antibody, a horseradish peroxidase-conjugated second antibody, and an enhanced chemiluminescence reagent. Anti-GST (Santa Cruz, Santa Cruz, CA), anti-PRDM14 (custom-made [18]), anti-hemagglutinin (HA) (sc-7392, Santa Cruz, Santa Cruz, CA), anti-cMyc (sc-40, Santa Cruz, Santa Cruz, CA), and anti-Enhancer of Zeste homolog 2 (EZH2) (active motif, Carlsbad, CA) antibodies were used to pull down the protein complexes.

Glutathione S-Transferase Pulldown Assay

The Baculovirus system was used to obtain the PRC2 complex consisting of four subunits—EZH2, Suppressor of Zeste 12 homolog (SUZ12), PHD finger protein 1 (PHF1), and embryonic ectoderm development protein (EED). The Baculovirus constructs of the four members of the PRC2 complex (EZH2, SUZ12, PHF1, and Flag-EED) were transfected into Sf9 insect cells individually and low-titer virus stocks were obtained. Initial low-titer virus stocks (P1) were amplified to subsequently produce P2 and then P3 stocks. The final P3 virus stock was assayed for the virus titer by the plaque assay method. All P3 stocks obtained were 1–2 × 108 pfu/ml. The virus stocks were then used to infect Sf9 cells at a multiplicity of infection of 10 and at 48 hours postinfection, cells were harvested. They were then lysed and the PRC2 complex was immunoprecipitated using an anti-Flag antibody. Full length PRDM14 was cloned into pet42b vector (Novagen, Darmstadt, Germany) and expressed in BL21 Eshcerichia coli. The GST tagged PRDM14 protein was first purified using glutathione (GSH)-Sepharose beads (Amersham, Piscataway, NJ) followed by nickel-nitrilotriacetic acid beads (Qiagen, Valencia, CA). The purified PRDM14 were bound to the GSH beads and incubated with the purified FLAG tagged PRC2 complex and vice versa purified PRC2 complex is bound to the FLAG beads and incubated with the purified PRDM14, for 2 hours in 4°C. The beads were washed four to six times and eluents were analyzed by Western blot.

Short Hairpin RNA-Mediated Knockdown


ChIP and RNA Expression Analysis

For ChIP experiments where PRDM14 is depleted, human ESCs transfected with shRNA targeting PRDM14 were harvested 40 hours post-transfection to ensure that total protein levels of target molecules in the ChIP experiments are largely unchanged when PRDM14 level decreased. ChIP was performed as described previously [6] with PRDM14 antibody (custom-made [18]), NANOG antibody (R&D, Minneapolis, MN), SUZ12 and EZH2 antibody (Active motif, Carlsbad, CA), H3K27me3 antibody (Millipore, Billerica, MA), and histone H3 antibody (Santa Cruz, Santa Cruz, CA). For the EZH2 ChIP-Seq library, 5–15 ng of ChIP enriched DNA was modified with the ChIP-Seq DNA Sample Prep Kit (IP-102-1001, Illumina, San Diego, CA). Briefly, ends of the ChIP DNA were repaired with exonucleases and an adaptor was ligated, followed by 15 cycles of polymerase chain reaction (PCR) amplification. The 200–300 bp fragments of the amplified DNA were gel purified for subsequent SOLEXA sequencing (Illumina, San Diego, CA). Total RNA extraction, reverse transcription, and quantitative real-time PCR were performed as described previously [6]. The Student's t test was used in the statistical analysis of the results. The results are considered significant if the p value <.05.

Luciferase Assay

Human ESCs were transfected with the reporter and knockdown constructs with TransIT (Mirus, Madison, WI) following the manufacturers protocol. DNA fragments analyzed were cloned into the pGL4.23 vector (Promega, Madison, WI). Primers used for cloning genomic HES7 DNA fragment; forward—GGACCAGGTCAGTCCCTCCGC and reverse—ATCGCATTTGCGCACTGCCCA. A Renilla luciferase plasmid pRL-SV40 (Promega, Madison, WI) was cotransfected as an internal control. Cells were harvested 48 hours post-transfection and the luciferase activity of the cell lysate was analyzed using the dual-luciferase reporter assay system (Promega, Madison, WI).

Retroviral Production and Viral Induction

pMXs retroviral plasmids containing the human OCT4, SOX2, KLF4, and c-MYC cDNA were obtained from Addgene (plasmids 17217, 17218, 17219, and 17220. S. Yamanaka). The pMXs retroviral plasmids containing cDNA of human PRDM14 gene were obtained as described previously [18]. Retroviruses were packaged using Pantropic Retroviral Expression System (Clontech, Mountain View, CA) and concentrated with centrifugal filter devices (Millipore, Billerica, MA). Confluent MRC-5 cells were split into 24 wells 16–24 hours before transduction with the retrovirus stock in the presence of 4 μg/ml polybrene (Sigma, St. Louis, MO). The virus medium was removed 16–24-hour postinfection and the cells were split from one 24-well into two six wells preseeded with CF-1 feeders 48-hour postinfection.


PRDM14 Binds to Repressed Genes

To investigate the potential repressive function of PRDM14 in human ESCs, we first analyzed the influence of PRDM14 binding on the expression of nearby genes. Putative target genes of PRDM14 are taken to be genes whose transcription start sites are nearest to individual PRDM14 binding regions. Analysis of the global expression profile of these genes in human ESCs revealed that the expression of the majority of genes near PRDM14 binding sites was upregulated upon PRDM14 depletion (Fig. 1A). In order to quantify this effect, we performed a GSEA, which tests for enrichment of a subset of genes in a ranked expression dataset. All genes were ranked according to their signal ratio between PRDM14 knockdown and control data and we tested for enrichment of PRDM14 target genes. PRDM14 target genes are enriched in the set of upregulated genes after PRDM14 depletion (p-value <.001, enrichment score −0.69) (Fig. 1B). This suggests that PRDM14 plays a predominant role in transcriptional repression in human ESCs.

PRDM14 Binding Correlates with H3K27me3

Next, we calculated the correlation of PRDM14 binding intensity (as estimated by the number of sequencing tags) with the major histone modifications and key transcription factors (Fig. 1C). We observed two clusters: a repressive cluster that includes H3K9me3, PRDM14, and H3K27me3, and an active cluster with OCT4, NANOG, and histone modifications associated with active transcription. Interestingly, in addition to NANOG and OCT4 [18], PRDM14 binding overlaps significantly with H3K27me3, H3K4me1, and p300, suggesting that PRDM14 potentially binds to poised regulatory elements. To verify this, we calculated the binding intensity (defined by tagged counts) of PRDM14 at known poised and active enhancers in human ESCs [12]. Indeed, PRDM14 binding is strongly enriched at poised enhancers, as compared to active enhancers (supporting information Fig. S1A).

Since H3K27me3 is associated with transcriptional repression [12], we asked whether there is a direct correlation between PRDM14 binding and H3K27me3 levels. We divided all PRDM14 binding sites into four quartiles based on the observed binding strength (Fig. 1D). For every category, we tabulated the average H3K27me3 profiles (Fig. 1D). Indeed, the level of H3K27me3 modifications increases with PRDM14 binding intensity. This highlights a potential functional link between PRDM14 and H3K27me3. As p300 is similarly found at poised enhancers, we perform similar correlation studies between p300 binding and H3K27me3 modification (supporting information Fig. S1B). Although p300 binding was used to identify poised enhancers [12], we did not observe similar correlations as compared to PRDM14 (supporting information Fig. S1B). Corresponding to its role in active enhancers in ESC [33], p300 binding intensity highly correlates with the active H3K27ac modification compared to PRDM14 (supporting information Fig. S1B). This result suggests that the H3K27me3 association is specific for PRDM14 and extends beyond the PRDM14 bound poised enhancers. To further validate the specificity of the association between H3K27me3 and PRDM14, we calculated the correlations with other histone modifications at PRDM14 binding sites (supporting information Fig. S1C). The correlation between PRDM14 and H3K27me3 is the highest of all 11 histone modifications tested (Pearson's correlation coefficient 0.59); further supporting a functional connection between PRDM14 binding and the repressive H3K27me3 histone mark.

PRDM14 Interacts with PRC2

The H3K27me3 histone mark plays an essential role in PRC2-mediated gene silencing [34–36]. PRC2 binds extensively to developmental genes in both mouse and human ESCs [8, 9]. The strong enrichment of the H3K27me3 modification in PRDM14 binding sites thus prompted us to investigate whether PRDM14 interacts with PRC2 to mediate gene silencing. We found EZH2, the enzymatic component of PRC2 [36], coimmunoprecipitates with PRDM14 from human ESCs whole cell extract and vice versa (Fig. 2A). We next confirmed this interaction in 293T cells expressing HA-tagged PRDM14 and MYC-tagged EZH2 proteins (Fig. 2B). To test whether PRDM14 protein directly interacts with PRC2, we expressed and purified recombinant GST-tagged PRDM14 for pull-down assays with PRC2 purified from Sf9 cells transduced with a virus-expressing components of PRC2 (FLAG-tagged EED, SUZ12, EZH2, and PHF1). We detected PRDM14 in the immunoprecipitation (IP) with a FLAG-specific antibody, along with other components of PRC2 (Fig. 2C). Similarly, we were able to detect EZH2, SUZ12, and FLAG-tagged EED in the IP with the GST-specific antibody (Fig. 2D). Taken together, these results support a direct interaction between PRDM14 and PRC2.

Figure 2.

PRDM14 interacts with PRC2. (A): Endogenous protein co-IP assays using human embryonic stem cell whole cell lysate. Antibodies against EZH2 or PRDM14 were used for the IP. EZH2 is detected in the PRDM14 IP (left) and vice versa (right). (B): Co-IP assay with Myc-tagged EZH2 and HA-tagged PRDM14 ectopically expressed in 293T cells. IP was performed with antibodies against MYC and HA. MYC-tagged EZH2 is detected in the HA IP (left) and vice versa (right). (C): Direct interaction between PRDM14 and the PRC2 complex. A GST-tagged recombinant PRDM14 was used for the pull-down assay with the purified PRC2 complex which contains a Flag-tagged EED component. GST-tagged PRDM14 was incubated with the purified PRC2 complex and subjected to IP with anti-FLAG antibody. GST-PRDM14 as well as other components of the PRC2 complex was identified in the Flag-IP. * A specific background band that is likely to be degraded GST-PRDM14. (D): Similar to above pull-down assay, IP was performed with anti-GST antibodies. Flag-EED and other members of the PRC2 complex, EZH2 and SUZ12, were detected in the GST-IP. Abbreviations: GST, glutathione S-transferase; IP, immunoprecipitation.

PRDM14 Modulates PRC2 Binding at Its Loci in Human ESC

We next explored if PRC2 binds to PRDM14-bound genes in human ESCs. We mapped the genome-wide binding profile of EZH2 in human ESCs and analyzed the correlation between PRDM14 and EZH2 binding. We observed a clear enrichment of EZH2 binding at PRDM14 peaks compared to that of NANOG (Supporting Information Fig. S2A). This corresponds with a high level of H3K27me3 and low level of H3K27ac enriched at PRDM14-bound loci compared to NANOG-bound loci (supporting information Fig. S2A). We next proceed to validate the binding of PRC2 to PRDM14-bound loci in human ESCs. Using antibodies specific to EZH2 and SUZ12 for ChIP assays, we detected EZH2 and SUZ12 binding at multiple PRDM14 bound loci (Fig. 3A, 3B). To determine whether EZH2 and SUZ12 binding at these loci are dependent on the presence of PRDM14, we repeated the ChIP assays in PRDM14-depleted human ESCs. The binding of EZH2 and SUZ12 was significantly reduced with the knockdown of PRDM14 (Fig. 3A, 3B). Importantly, this decrease in binding was not due to a reduction in total EZH2 and SUZ12 protein level (supporting information Fig. S2B); neither was it due to ESC differentiation following PRDM14 depletion, given that NANOG protein level and binding are unchanged (supporting information Fig. S2B, S2C). Corresponding to the loss of PRC2 binding, H3K27me3 modification at these loci was found to be markedly reduced (Fig. 3C), with global H3K27me3 levels remaining largely unchanged in PRDM14-depleted human ESCs (supporting information Fig. S2B). As the level of histone H3 is largely unchanged at these loci in PRDM14-depleted human ESCs (supporting information Fig. S2D), we concluded that the loss of H3K27me3 is not due to the loss of nucleosomes in these regions. In summary, all the data suggest that the observed reduction of PRC2 and H3K27me3 in PRDM14 binding sites is a direct effect of PRDM14 depletion in human ESCs.

Figure 3.

PRDM14 modulates PRC2 binding at its loci in human embryonic stem cell (ESC). (A): PRDM14 regulates EZH2 binding at its bound loci. ChIP using anti-EZH2 antibody was performed with human ESCs transfected with shRNA targeting PRDM14 or LUCIFERASE (control). Results are presented in the form of fold enrichment against control regions 1 and 2 which only have background levels of enrichment. All values are means ± SEM from three independent experiments (n = 3). (B): PRDM14 regulates SUZ12 binding at its bound loci. Similar to above, ChIP was performed using anti-SUZ12 antibody. (C): PRDM14 regulates level of H3K27me3 at its loci. Similar to above, ChIP was performed using anti-H3K27me3 antibody.

PRDM14-Mediated Repressive Activity Is Dependent on PRC2

The presence of PRC2 and H3K27me3 at PRDM14 bound loci and the enrichment of targets among derepressed genes in PRDM14-depleted human ESCs (Fig. 1B) suggest that these genomic regions may have repressive function. We identified a ∼500 bp PRDM14-bound loci near the HES7 gene and cloned it into the upstream of a reporter gene driven by the highly active phosphoglycerate kinase (PGK) promoter (Fig. 4A). The presence of this DNA fragment resulted in a decrease of more than 50% in the PGK gene promoter activity in human ESCs and the mutation of the PRDM14 binding motif in this genomic fragment resulted in the loss of repressive activity (Fig. 4B). The depletion of PRDM14 also alleviates the repressive activity of this HES7 gene element (Fig. 4C), confirming the requirement of PRDM14 for repressive activity at this locus. Interestingly, depletion of EZH2 also results in a loss of repressive activity, suggesting that the PRDM14 repressive activity at this locus is dependent on PRC2. However, as the 500 bp HES7 DNA sequence may potentially include DNA binding motifs of other transcription factors that may unknowingly influence PRC2 activity, we replaced the entire 500 bp HES7 DNA sequence with the tandem repeats of the 16 bp PRDM14 binding motif previously identified (Fig. 4D) [18]. Similar to the HES7 DNA fragment, the PRDM14 binding motifs alone were sufficient to reduce the PGK promoter activity (Fig. 4E). This repressive activity is similarly lost when the PRDM14 binding motifs are mutated or when PRDM14 is depleted (Fig. 4E, 4F). Importantly, we found that the repressive activity of the PRDM14 binding motif is relieved when EZH2 was depleted (Fig. 4F); hence confirming the role of PRC2 in PRDM14-mediated repression. In summary, the reporter assays demonstrate that genomic regions bound by PRDM14 can have repressive activity mediated by PRDM14 in concert with PRC2.

Figure 4.

PRDM14 exhibits PRC2-dependent repressive activity. (A): Schematic representations of the reporter construct. A ∼500 bp PRDM14-bound locus in the HES7 gene was cloned before the PGK gene promoter driving the LUCIFERASE gene. (B): PRDM14-bound locus exhibits repressive activity. Addition of the HES7 genomic fragment containing the PRDM14 binding motif results in the repression of PGK promoter activity (WT). Point mutations in the PRDM14 binding motif led to a decrease in repressive activity (MUT). All values are means ± SEM from three independent experiments (n = 3). (C): Repressive activity of the PRDM14 motif is dependent on PRDM14 and EZH2. The reporter construct was cotransfected with shRNA construct targeting PRDM14 or EZH2. Luciferase activity is normalized against samples cotransfected with the control shRNA targeting green fluorescent protein (GFP). (D): Schematic representations of reporter construct with two copies of the PRDM14 binding motif in place of the HES7 DNA fragment. (E): Repressive activity of PRDM14 is dependent on its conserved binding element. Addition of two tandem sequences of the PRDM14 binding motif results in the repression of PGK promoter activity. Point mutations in the two motifs derepressed promoter activity. (F): Repressive activity of the PRDM14 binding motif is dependent on PRDM14 and EZH2. The reporter construct was cotransfected with a shRNA construct targeting PRDM14 or EZH2. Luciferase activity is normalized against samples cotransfected with the control shRNA targeting GFP. Abbreviation: PGK, phosphoglycerate kinase.

PRDM14 Represses Developmental Genes During iPSC Reprogramming

As PRDM14 was previously reported to enhance the induction of pluripotency in fibroblast cells [18], we investigated whether the repressive activity of PRDM14 is involved in the reprogramming of somatic cells back to the pluripotent state. We first determined the expression of PRDM14 during the reprogramming of human fibroblast using the transcription factors OCT4, SOX2, KLF4, and cMYC (4F) [37]. As PRDM14 transcript remains undetectable until 7 days postinfection (d.p.i.) (supporting information Fig. S3), we thus investigated the effect of ectopic PRDM14 expression in the early time points of reprogramming. We first looked at the expression of developmental genes that are bound by PRDM14 in human ESC and highly expressed in fibroblast to determine whether PRDM14 could repress these genes during reprogramming. We found that the transcripts of LMNA, OAF, IGFBP5, CDC92, USP3, and MYH9 were either weakly repressed before 7 d.p.i. or only downregulated upon 7 d.p.i. during reprogramming with the 4F (Fig. 5A). However, upon the ectopic expression of PRDM14, these genes become strongly repressed by 4 d.p.i. (Fig. 5A). This result suggests that PRDM14 is able to silence its target genes in fibroblasts in the process of reprogramming, possibly explaining how PRDM14 improves reprogramming [18].

Figure 5.

PRDM14 represses its target genes in fibroblasts during reprogramming. (A): Time point analysis of the expression of PRDM14 target genes during reprogramming. Fibroblasts were infected with retroviruses ectopically expressing the reprogramming factors OCT4 (O), SOX2 (S), KLF4 (K), and cMYC (M) with or without PRDM14. Expression of the PRDM14 target genes LMNA, OAF, IGFBP5, CDC92, USP3, and MYH9 was measured with quantitative real-time PCR 4 and 7 days postinfection (d.p.i.). Relative expression was obtained by normalizing against fibroblasts infected with retroviruses ectopically expressing GFP (control). All values are means ± SEM from three independent experiments (n = 3). (B): Downregulation of the expression of mesenchymal genes ZEB1, SNAIL, and SLUG during OSKM reprogramming with and without PRDM14. Cotransduction of PRDM14 further suppressed the expression of the three genes. Abbreviation: GFP, green fluorescent protein.

Two studies identified the mesenchymal-to-epithelial transition (MET) as a key initiation event during reprogramming of fibroblasts [38, 39]. This event entails the repression of key mesenchymal regulators SNAIL, SLUG, and ZEB1 [40]. Li et al. reported that bone morphogenetic protein (BMP) signaling increases reprogramming efficiency by inducing miRNAs that silence the key mesenchymal factors ZEB1 and ZEB2 [38] and Samavarchi-Tehrani et al. reported that transforming growth factor beta (TGF-β) signaling impedes reprogramming by inducing these genes [39]. We thus investigate if PRDM14 potentially enhances reprogramming by repressing the key mesenchymal genes during the early stages of iPSC reprogramming. These three genes are repressed by 4 d.p.i. and the ectopic expression of PRDM14 during reprogramming further suppresses their expression at similar time point (Fig. 5B). In human ESCs, PRDM14 binds strongly at the promoter of ZEB1 (Fig. 6A). To investigate whether PRDM14 can directly repress ZEB1 in fibroblast, we infected fibroblast with retrovirus to ectopically express PRDM14. ChIP with PRDM14 antibody shows that PRDM14 is able to bind to the ZEB1 promoter (Fig. 6B) in fibroblasts. We also detected significant PRDM14 binding at another target differentiation gene OAF (Fig. 6B), but not at the two control sites, confirming that the binding of PRDM14 is specific. Remarkably, the binding of PRDM14 at ZEB1 and OAF promoters also resulted in a modest increase in EZH2 recruitment and H3K27me3 levels at these loci (Fig. 6C, 6D). The presence of the repressive histone mark is consistent with decreased ZEB1 gene expression in the presence of ectopic PRDM14 (Fig. 6E). These data show that PRDM14 is able to repress its target genes in fibroblasts and potentially increases reprogramming efficiency by repressing a key mesenchymal gene for the initiation event.

Figure 6.

PRDM14 binds and represses ZEB1 in human fibroblast. (A): ChIP-Seq profile of PRDM14, OCT4, and H3K27me3 at the ZEB1 locus in human embryonic stem cells. Control ChIP-Seq library was obtained from sequencing input genomic DNA. (B): PRDM14 binds to target genes in MRC-5 human fibroblasts. Fibroblasts were infected with retroviruses ectopically expressing PRDM14 or GFP (control). Cells were harvested at 5 d.p.i. for ChIP assay. Relative enrichment of PRDM14 binding on ZEB1 and OAF loci is obtained via normalization against control regions 1 and 2 which only have background levels of enrichment. All values are means ± SEM from three independent experiments (n = 3). (C): Recruitment of EZH2 to PRDM14-bound loci in MRC-5 fibroblasts. EZH2 ChIP in fibroblasts infected with retroviruses ectopically expressing PRDM14 or GFP. (D): Increased H3K27me3 modifications at PRDM14-bound loci in fibroblasts. H3K27me3 ChIP in fibroblasts infected with retroviruses ectopically expressing PRDM14 or GFP. (E): PRDM14 represses ZEB1 expression in fibroblasts. ZEB1 expression was measured 5 d.p.i. Relative expression level was obtained by normalizing against fibroblasts infected with retroviruses ectopically expressing GFP. Abbreviation: GFP, green fluorescent protein.


Previous studies have suggested that PRDM14 has repressive functions in human and mouse ESCs, mouse PGC, and EpiSC [18–22]. The ectopic expression of PRDM14 in both human and mouse ESC-derived embryoid bodies and in mouse EpiSC suppresses the expression of lineage-specific genes [19, 21, 22]. Genome-wide mapping studies in mice have shown that Prdm14 binds to regulators of the ExEn lineage and Prdm14 depletion results in the upregulation of these genes [21]. However, little is known about the mechanism of these repressive activities of PRDM14. Here, we found that PRDM14 plays a repressive role through its interaction with PRC2. Although PRDM14 contains a SET domain, there has been no report of enzymatic activity for PRDM14. The interaction with PRC2 potentially explains the high level of H3K27me3 modification present at PRDM14-bound loci. This correlation was also observed in mice where PGC-like cells in Prdm14-null mice fail to establish the H3K27me3 mark [20]. In addition, Yamaji et al. also showed that the mouse Prdm14 protein exhibits repressive activity using the GAL4-UAS reporter assay [20]. Whether this repressive activity of the mouse Prdm14 is dependent on PRC2 remains to be validated.

Since its discovery, the recruitment of PRC2 to its target genes has been of great interest [41]. In mouse ESCs, Jarid2, the founding member of the Jumonji family of histone demethylase, was identified as a potential recruiter of PRC2 binding to CpG-rich sequences [42–44]; whether the human JARID2 homolog plays a similar role in human ESC remains to be determined. Moreover, Jarid2 binds to almost all PRC2-bound loci and is proposed to be a core component of PRC2 [41]; how PRC2 is selectively recruited to target genes in various cell types remains unclear. Previous mapping studies suggested a potential interaction between the core ESC regulators and PRC2 based on the large overlap in their chromatin binding distributions [8, 9]. However, these interactions have yet to be functionally validated; direct interaction between the core regulators and PRC2 has not been reported and OCT4 and NANOG have not been shown to exhibit repressive activity. Other than the transcriptional repressor protein Yin Yang 1, there has been no strong evidence for transcription factor-mediated recruitment of the mammalian PRC2 [41]. In this study, we uncovered a novel interaction between an important pluripotency factor of human ESC and a major chromatin regulator which mediates epigenetic repression. The PRDM14 and PRC2 interaction further supports the notion that PRC2 can be recruited by transcription factors to cohesively regulate gene expression. PRDM14 directs PRC2 activity specifically to the developmental genes which it binds in human ESCs.

The use of transcription factors to induce pluripotent stem cells highlights a major breakthrough in stem cell studies [3]. Although human fibroblasts can similarly be reprogrammed by the 4F [37], the efficiency of generating human iPSCs is much lower and requires a longer period compared to mouse. The use of PRDM14 in concert with the 4F increases the number of human iPSC colonies obtained by more than threefolds [18]. Understanding how PRDM14 improves reprogramming efficiency may reveal insights to rate limiting events during iPSC reprogramming. The cotransduction of PRDM14 with the 4F resulted in the stronger induction of INADL, CDH1, and CLDN3, which are markers for MET during reprogramming [18]; it is possible that PRDM14 may enhance reprogramming by targeting MET processes. Previous studies have highlighted the importance of suppressing ZEB1 during reprogramming [39, 45]. Samavarchi-Tehrani et al. showed that in the initial stage of reprogramming, the 4F represses TGF-β signaling which in turn downregulates the mesenchymal gene ZEB1 [39]. The effect of ZEB1 in reprogramming was also highlighted by Onder et al. in a screen for chromatin modifying enzymes that regulates reprogramming efficiency [45]. In addition, the authors showed that the inhibition of DOT1L results in the increased MET efficiency and overall improvement in reprogramming efficiency, but overexpression of ZEB1 abolishes the effect of DOT1L inhibition [45]. In this study, we found that PRDM14 is able to directly repress ZEB1 when ectopically expressed in fibroblasts. The binding of PRDM14 to the ZEB1 promoter will recruit PRC2 to the locus, consequently resulting in the increase in H3K27me3 modification to mediate silencing. This potentially explains why PRDM14 is able to enhance reprogramming efficiency with the 4F.


In all, our study uncovered PRDM14 as a new regulator of PRC2, revealing insights to how key transcription factors cooperate with chromatin regulators to establish the pluripotency transcriptional network. PRC2 is essential for PRDM14's function in repressing developmental genes for maintaining the human ESC identity and also in enhancing the efficiency of converting human fibroblasts to pluripotent stem cells.


We are grateful to the Biomedical Research Council (BMRC). This work is supported by the Agency for Science, Technology and Research (A*STAR) of Singapore. Dr. Jonathan Göke is supported by a fellowship within the Postdoc-Programme of the German Academic Exchange Service, DAAD (J.G.). We are grateful to Lai-Ping Yaw and Kim-Jee Goh for technical assistance. We are grateful to Dr. Qiang Yu for providing the EZH2 plasmids. We thank Mei-Sheng Lau, Kevin Andrew Uy Gonzales, Jia-Chi Yeo, and Andrew Hutchins for critical comments to the manuscript.


The authors declare that there are no potential conflicts of interest.