Journal list menu

Volume 591, Issue 18 p. 2879-2889
Research Letter
Free Access

Epigenetic barrier against the propagation of fluctuating gene expression in embryonic stem cells

Yuki Saito

Yuki Saito

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

These authors contributed equally to this workSearch for more papers by this author
Akira Kunitomi

Akira Kunitomi

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

These authors contributed equally to this workSearch for more papers by this author
Tomohisa Seki

Tomohisa Seki

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Shugo Tohyama

Shugo Tohyama

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Dai Kusumoto

Dai Kusumoto

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Makoto Takei

Makoto Takei

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Shin Kashimura

Shin Kashimura

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Hisayuki Hashimoto

Hisayuki Hashimoto

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Gakuto Yozu

Gakuto Yozu

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Chikaaki Motoda

Chikaaki Motoda

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Masaya Shimojima

Masaya Shimojima

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Toru Egashira

Toru Egashira

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Mayumi Oda

Mayumi Oda

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Keiichi Fukuda

Keiichi Fukuda

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Search for more papers by this author
Shinsuke Yuasa

Corresponding Author

Shinsuke Yuasa

Department of Cardiology, Keio University School of Medicine, Tokyo, Japan

Correspondence

S. Yuasa, Department of Cardiology, Keio University School of Medicine, 35 Shinanomachi Shinjuku-ku, Tokyo 160-8582, Japan

Fax: +81 3 5363 3875

Tel: +81 3 5363 3373

E-mail: [email protected]

Search for more papers by this author
First published: 14 August 2017
Citations: 1
Edited by Didier Stainier

Abstract

The expression of pluripotency genes fluctuates in a population of embryonic stem (ES) cells and the fluctuations in the expression of some pluripotency genes correlate. However, no correlation in the fluctuation of Pou5f1, Zfp42, and Nanog expression was observed in ES cells. Correlation between Pou5f1 and Zfp42 fluctuations was demonstrated in ES cells containing a knockout in the NuRD component Mbd3. ES cells containing a triple knockout in the DNA methyltransferases Dnmt1, Dnmt3a, and Dnmt3b showed correlation between the fluctuation of Pou5f1, Zfp42, and Nanog gene expression. We suggest that an epigenetic barrier is key to preventing the propagation of fluctuating pluripotency gene expression in ES cells.

Abbreviations

Dnmt TKO, Dnmt triple knock out

ES, embryonic stem

LIF, leukemia inhibitory factor

Mbd3 KO, Mbd3 knock out

SCNT, somatic cell nuclear transfer technique

The control of pluripotent stem cells is of major concern in basic science [1]. Although embryonic stem (ES) cells and induced pluripotent stem cells have come into use in many types of basic research, variability in experimental results is often observed. To reduce this variability, how it is created in pluripotent stem cells must be understood. Variability itself has been recognized in many biological situations. Variability is also observed as uncertainty, noise, and heterogeneity. Variability may be required in order to adapt to different situations. Our inability to predict such complicated scenarios is due to our insufficient understanding of variability.

Single-cell analyses have revealed heterogeneity in the expression level and epigenetic status of pluripotency genes in ES cells [2-4]. Subpopulations differentially expressing pluripotency genes show biased differentiation propensities [4, 5]. It has been challenging to understand and control heterogeneity in ES cells. In a population of ES cells, cell-to-cell heterogeneity would be masked due to averaging. Mouse ES cells are self-renewing pluripotent stem cells that show the same characteristics following serial passage. However, we often observed some variations within each experiment. Therefore, we hypothesized that ES cells have different characteristics as a population at each serial passage. In this study, we focused on the fluctuation of pluripotency gene expression in a population of ES cells. As the expression level of pluripotency genes is highly important for several characteristics in ES cells, we investigated the relationship between the fluctuating expression of pluripotency genes and how those relationships are formed.

Here, we demonstrate that fluctuating expression of pluripotency genes occurs in a population of ES cells and that the fluctuating expression of some genes is correlated. There is no correlation in the fluctuating expression of Pou5f1, Zfp42, and Nanog in wild-type ES cells and the ES cells established by somatic cell nuclear transfer technique (SCNT-ES cells). Moreover, cell seeding concentration slightly affected the correlation between the fluctuating expression of Pou5f1 and Sox2. ES cells in which the NuRD component Mbd3 is knocked out (Mbd3 KO) show correlation between the fluctuating expression of Pou5f1, Sox2, and Zfp42. Furthermore, we confirmed that the rescue of Mbd3 recovers the fluctuating expression of these pluripotency genes. ES cells containing a triple knockout in the DNA methyltransferase genes Dnmt1, Dnmt3a, and Dnmt3b (Dnmt TKO) show correlation between the fluctuating expression of Pou5f1, Sox2, Nanog, and Zfp42 and the wild-type ES cells with 2i demonstrated similar correlation profiles. These results suggest that epigenetic modification creates a barrier against the propagation of fluctuating expression of pluripotency genes in ES cells.

Materials and methods

ES cell culture

Murine embryonic fibroblast-free ES cells were used. Undifferentiated ES cells (R1) [6] were maintained on gelatin-coated dishes in GMEM that was supplemented with 10% FBS (Equitech-Bio, Kerrville, TX, USA), 2 mm l-glutamine, 0.1 mm nonessential amino acids, 1 mm sodium pyruvate, and 1000 U·mL−1 murine leukemia inhibitory factor (LIF; MERCK Millipore, Billerica, MA, USA) [7]. ES cells were passaged every 72 h (100 000 cells/10 cm dish). The Dnmt TKO ES cell line was provided by RIKEN BRC through the National Bio-Resource Project of MEXT, Japan. The Mbd3 KO ES cell line was kindly provided by Brian D. Hendrich of Wellcome Trust – MRC Stem Cell Institute, Department of Biochemistry, University of Cambridge.

Transfection and selection of stable Mbd3 overexpressed clones

The expression vector for mouse Mbd3 was kindly provided by Brian D. Hendrich of Wellcome Trust – MRC Stem Cell Institute, University of Cambridge. Mbd3 KO ES cell line was cultured under the same condition of R1 ES cells. Cells were cultured in 6 cm dish to 50% confluence and were then transfected with 10 μg of plasmid and 10 μL of Lipofectamine 2000 (Invitrogen, Carlsbad, CA, USA). One day after transfection, cells were replated and cultured in the presence of 1 μg·mL−1 Puromycin (InvivoGen, San Diego, CA, USA). Puro-resistant colonies were selected and expanded. We confirmed significant Mbd3 expression by quantitative RT-PCR.

Quantitative RT-PCR

Total RNA was extracted using TRIzol reagent (Life Technologies, Carlsbad, CA, USA). cDNA was produced using ReverTra Ace qPCR RT Master Mix with gDNA Remover (Toyobo Biochemicals, Osaka, Japan). RT-PCR was performed as previously described [8]. The PCR primers used are listed in Table  S1. At least three replicates were carried out for each time point. The Gapdh gene was used as an internal control.

Bisulfite genomic sequencing

Bisulfite treatment was performed using the EZ DNA Methylation-Gold kit (Zymo Research, Orange, CA, USA) according to the manufacturer's recommendations. The PCR primers used are listed in the Table  S2. Amplified products were cloned into the pGEM-T Easy Vector (Promega, Madison, WI, USA). Randomly selected clones were sequenced with the SP6 forward primer.

Statistical analysis

Values are presented as the mean ± SEM. Correlations between different genes were analyzed using several statistical methods. Spearman's correlation test was performed to compare the simultaneous expression correlations of different genes. Cross-correlation analysis was performed to describe the time-delayed gene expression correlations. A value of P < 0.05 was considered significant. Each value is shown in Tables S1–S13. The experiments included in Figs 1 and 2 were performed simultaneously, as were the experiments described in Figs S2,S4andS12. Likewise, the experiments described in Figs 3 and S13 were simultaneously performed, as were the experiments described in Figs 4 and S14. Other experiments were performed independently.

Details are in the caption following the image
Sequential expression patterns of core pluripotency genes in ES cells. (A–D) Quantitative RT-PCR analyses of Pou5f1, Sox2, Nanog, and Zfp42 expression. Bars show mean values of n = 3 replicate samples ± SD. (E) Overlay of Pou5f1, Sox2, Nanog, and Zfp42 expression patterns. (F) Coefficient of variation representing the ratio of the standard deviation to the mean for each gene. (G) Spearman's correlation coefficients for these genes (Table  S3). Light gray columns represent weakly significant (P < 0.05) correlations.
Details are in the caption following the image
Sequential expression patterns of Pou5f1-related genes in ES cells. (A–F) Quantitative RT-PCR analyses of Sall4, Dnmt3a, Dnmt3b, Tcf3, Ctr9, and G9a expression. Bars show mean values of n = 3 replicate samples ± SD. (G) Coefficient of variation representing the ratio of the standard deviation to the mean for each gene. (H) Spearman's correlation coefficients for these genes (Table  S10). Light gray colored column was considered weak significant (P < 0.05). Dark gray columns represent significant (P < 0.01) correlations.
Details are in the caption following the image
Sequential expression patterns of core pluripotency genes in Mbd3 KO ES cells. (A–D) Quantitative RT-PCR analyses of Pou5f1, Sox2, Nanog, and Zfp42 expression. Bars show mean values of n = 3 replicate samples ± SD. (E) Overlay of Pou5f1, Sox2, Nanog, and Zfp42 expression patterns. (F) Spearman's correlation coefficients for these genes (Table  S11). Light gray columns represent weakly significant (P < 0.05) correlations. Dark gray columns represent significant (P < 0.01) correlations.
Details are in the caption following the image
Sequential expression patterns of core pluripotency genes in Dnmt TKO ES. (A–D) Quantitative RT-PCR analyses of Pou5f1, Sox2, Nanog, and Zfp42 expression. Bars show mean values of n = 3 replicate samples ± SD. (E) Overlay of Pou5f1, Sox2, Nanog, and Zfp42 expression patterns. (F) Spearman's correlation coefficients for these genes (Table  S13). Dark gray columns represent significant (P < 0.01) correlations.

Results

Independent fluctuation in core pluripotency gene expression in ES cells

To elucidate whether we can observe fluctuation in pluripotency gene expression in ES cells, we serially examined the expression of pluripotency genes. Murine ES cells were maintained with serum and LIF in feeder-free conditions and passaged every three days [9]. We confirmed that the ES cells maintain stable pluripotent state through every passage by alkaline phosphatase staining (Fig. S1). When passaging, we collected the remaining cells and examined the expression patterns of pluripotency genes. Pou5f1 (also known as Oct4) is a member of the octamer-binding subgroup of the POU family of transcription factors. This gene plays a major role in the core transcriptional regulatory network in pluripotent stem cells [10]. The expression of Pou5f1 fluctuated within a range of approximately +50% to −40% of the average expression level (Fig. 1A). Sox2 is a member of the SRY-related high-mobility group domain-containing transcriptional regulators. The Sox2 protein forms a heterodimer with Pou5f1 on DNA to activate the transcription of pluripotency genes [11]. The expression of Sox2 fluctuated within a range of approximately +100% to −40% of the average expression level (Fig. 1B). Nanog is a homeodomain transcription factor gene and is a critical regulator that blocks the differentiation of ES cells [12-14]. The expression of Nanog fluctuated within a range of approximately +30% to −30% of the average expression level (Fig. 1C). Zfp42 (also known as Rex1) is a member of the YY1 subfamily of transcription factors and is widely used as a pluripotency marker [15, 16]. The expression of Zfp42 fluctuated within a range of approximately +20% to −30% of the average expression level (Fig. 1D). The fluctuation ranges are relatively high for Pou5f1 and Sox2 when compared to the fluctuations in Nanog and Zfp42 expression (Fig. 1E,F). Next, we examined correlations between the expression levels of pluripotency genes. Interestingly, fluctuation in the expression of Pou5f1 was correlated with fluctuations in the expression of Sox2 (Fig. 1G). However, the expression levels of the pluripotency genes Nanog and Zfp42 were not correlated with the expression of Pou5f1. Additionally, we confirmed that we have same correlation profiles when normalizing to different housekeeping genes (Fig. S2) [17, 18]. Next, we investigated the expression levels of pluripotency genes in different cell seeding concentration into 5.0 × 104, 1.0 × 105, 1.5 × 105 cells per 10 cm dish (Fig. S3A–G,S4A–G, S5A–G). They indicated different fluctuation range patterns of pluripotency genes among different concentration levels (Figs S3F, S4F, S5F). Significant correlation of expression level was only observed between Pou5f1 and Sox2 in all concentration levels and interestingly, correlation ratio was highest in the most concentrated culture condition (Fig. S5G). We also verified the expression levels of pluripotency genes in the ES cells established by SCNT-ES cells as other wild-type ES cells (Fig. S6A–G). Similar to the R1 wild-type ES cells, Sox2 expression showed highest fluctuation among four genes. The correlations between the expression levels of pluripotency genes were almost same profile to the R1 wild-type ES cells (Fig. S6G). Furthermore, we examined the fluctuation in pluripotency gene expression using wild-type ES cells with 2i (CHIR99021 and PD0325901) condition instead of LIF (Fig.  S7 A–G). The fluctuation ranges are relatively suppressed than wild-type ES cells without 2i except of Pou5f1 (Figs 1F and S7F). Intriguingly, the fluctuation in the expression of all pluripotency genes is significantly correlated with each other (Fig. S7G).

The correlated fluctuation of pluripotency genes in ES cells

To further explore fluctuation and regulation of pluripotency gene expression, we examined several factors that positively or negatively regulate Pou5f1 expression. Sall4, a spalt–like zinc finger protein, mediates the positive regulation of Pou5f1 in cooperation with other transcription factors [19]. The expression of Sall4 fluctuated within a range of approximately +170% to −60% of the average expression level (Fig. 2A). The mammalian de novo DNA methyltransferases Dnmt3a and Dnmt3b play a role in Pou5f1 silencing during differentiation. The expression of Dnmt3a and Dnmt3b also fluctuated (Fig. 2B,C). Tcf3 is a β-catenin partner in Wnt signaling and a negative regulator of Pou5f1. The expression of Tcf3 fluctuated within a range of approximately +200% to −60% of the average expression level (Fig. 2D). The PAF1/RNA polymerase II complex component Ctr9 negatively regulates Pou5f1 expression. The expression of Ctr9 fluctuated. The fluctuation range for this gene was small, within a range of approximately +20% to −20% of the average expression level (Fig. 2E). The histone H3 lysine 9 methyltransferase G9a regulates Pou5f1 expression through histone methylation and DNA methylation. The expression of G9a fluctuated within a range of approximately +100% to −60% of the average expression level (Fig. 2F). The fluctuation ranges were relatively high for Sall4 and Tcf3 expression compared to the fluctuation ranges for the expression of Dnmt3a, Dnmt3b, and Ctr9 (Fig. 2G). We speculated that correlation occurred solely between Sall4, Tcf3, and G9a due to similarity in the shape of the fluctuation (Fig. 2A,D,F). However, our results showed that the expression level of Pou5f1 was correlated with the expression of all genes (Fig. 2H). The expression level of Sall4 is not correlated with the expression of Dnmt3a and Ctr9.

Correlation exists between Pou5f1 and Zfp42 expression and between Sox2 and Zfp42 expression in Mbd3 KO ES cells

We examined whether epigenetic regulation affects fluctuations in the expression of pluripotency genes. Methyl-CpG binding domain protein 3 (Mbd3) is an essential scaffold protein of the NuRD complex. In the absence of Mbd3, the complex is not assembled [20]. We used an Mbd3 knock out (Mbd3 KO) ES cell line, which cannot differentiate and can be maintained in LIF-free conditions [20, 21]. Mbd3 KO ES cells also show fluctuation in the expression of Pou5f1, Nanog, Sox2 and Zfp42 (Fig. 3A–E). The fluctuation ranges of Pou5f1 and Sox2 appeared to be smaller in Mbd3 KO ES cells, whereas the fluctuation ranges of Nanog and Zfp42 expression appeared to be unaffected (FigS8 A–D). We also looked for correlation between the expression levels of pluripotency genes in Mbd3 KO ES cells (Fig. 3F). We observed correlation between Pou5f1 and Sox2 expression in Mbd3 KO ES cells, as was observed in wild-type and in SCNT-ES cells. Correlation between Pou5f1 and Zfp42 expression was also observed in Mbd3 KO ES cells, which was not observed in wild-type and in SCNT-ES cells. Sox2 expression was also correlated with Zfp42 expression in Mbd3 KO ES cells. Nanog expression was not correlated with Pou5f1, Sox2, or Zfp42 expression; this was also found in wild-type and in SCNT-ES cells. Next, we investigated whether the rescue of Mbd3 recovers the fluctuation of the genes expression. Exogenous expression of Mbd3 by lipofection indicated significantly higher Mbd3 expression (Fig. S9). Mbd3-rescued ES cells showed more fluctuation in the expression of Pou5f1, Sox2, Nanog, and Zfp42 than Mbd3 KO ES cells and it became more similar to that of the wild-type ES cells (Fig.S10A–G). There remained the significant correlation between only Sox2 and Zfp42 in Mbd3-rescued ES cells (Fig.S10G).

Dnmt TKO ES cells show correlation in the expression of Pou5f1, Sox2, Nanog, and Zfp42

DNA methylation regulates epigenetic status in mammalian development [22] and plays a critical role in silencing gene expression [23]. In cultured ES cells, DNA methylation dynamics parallel early developmental processes [24]. Despite the absence of CpG methylation, ES cells can grow robustly and maintain their undifferentiated characteristics [25]. We examined whether DNA methylation affects fluctuation in pluripotency gene expression using ES cells containing a triple knockout in the DNA methyltransferase genes Dnmt1, Dnmt3a, and Dnmt3b (Dnmt TKO); in these cells, DNA methylation was totally absent. Dnmt TKO ES cells also showed fluctuated expression of Pou5f1, Nanog, Sox2, and Zfp42, and these fluctuation patterns were similar to each other (Fig. 4A–E). The fluctuation range of Sox2 expression appeared to be smaller in Dnmt TKO ES cells compared to that in wild-type ES cells (Fig.S8F). However, the fluctuation ranges of Pou5f1, Nanog, and Zfp42 expression appeared to be unaffected (Fig.S8E,G,H). We also examined the correlations between the expression levels of pluripotency genes in Dnmt TKO ES cells. Surprisingly, the expression levels of Pou5f1, Sox2, Nanog and Zfp42 were strongly correlated with each other (Fig. 4F). To understand the possible mechanism of these fluctuations, we examined the methylation status of the Zfp42 locus by bisulfite genomic sequencing. The Zfp42 locus was not completely unmethylated in wild-type ES cells, as previously described [26]. In Mbd3 KO ES cells and Dnmt TKO ES cells, this DNA methylation was completely absent (Fig.S11).

Cross-correlation analysis revealed there was no time-delayed correlation

Finally, we performed cross-correlation analysis to determine whether there were internal correlations or temporal patterns, meaning whether the expression patterns at a given passage would influence the patterns in the following passages. In wild-type ES cells, Pou5f1 and Sox2 correlated at zero lag (Fig. S12). However, there were no significant correlations at different lags. Likewise, in Mbd3 KO ES cells, significant correlations were observed only between Pou5f1 and Sox2, Pou5f1 and Zfp42, and Sox2 and Zfp42 at a lag of zero (Fig. S13). In Dnmt TKO ES cells, the correlation was strongest at zero lag in any combination (Fig. S14).

Discussion

The fact that Pou5f1, Nanog, and Zfp42 independently fluctuate suggests that the expression of these genes is regulated by independent networks in steady-state ES cells. Although there are reports that Pou5f1 and Nanog cross-regulate each other [11, 27], the two genes have distinct roles. Fluctuations in Nanog expression occur at the single-cell level in individual Pou5f1-positive ES cells [2, 4, 28]. These fluctuations transiently prime individual ES cells for differentiation without marking a definitive commitment. There have been many attempts to decipher the mechanisms of cell-to-cell variation and fluctuation in single ES cells. In this study, we found new fluctuations within a population of ES cells. Cross-correlation analysis revealed that this new fluctuation is synchronic and without time delays. It remains unclear whether these fluctuations are intrinsically regulated or if individual ES cells communicate with each other within a population. Synchronized gene expression could be intrinsically maintained in ES cells. In a different model, ES cells could communicate with each other via humoral factors, direct communication, or by other means. As the concentration of seeding ES cells increases, the correlation of Pou5f1 and Sox2 slightly increased in our study. It could indicate that increasing cell seeding concentration enhances the communication with each other through the culture or emphasizing the intrinsic gene expression fluctuation profile of the most major cell population.

DNA methylation is established primarily by Dnmt3a and Dnmt3b and is maintained by the related protein Dnmt1 [29]. In our study, correlation between Pou5f1, Sox2, Nanog, and Zfp42 expression arises in Dnmt TKO ES cells. These findings suggest that DNA methylation generates an epigenetic barrier (Fig. S15). Bisulfite sequencing showed that DNA methylation at the Zfp42 locus was completely absent in Mbd3 KO ES cells and Dnmt TKO ES cells, which supports this hypothesis. However, variance of pluripotency marker expression appeared to be constant or smaller in Mbd3 KO ES cells and Dnmt TKO ES cells (Fig.S8A–H). This finding cannot be explained solely by DNA methylation; several unknown mechanisms must exist in this system. Further study is required to understand the underlying molecular mechanisms. A longer observation period may provide a more detailed understanding by confirming the statistical difference in gene expression variations in different cell lines and by correcting the effect of autocorrelation of pluripotency genes in cross-correlation analysis.

It is of note that the expression of some epigenetic regulators, such as Dnmt3a and Dnmt3b, is also correlated with Pou5f1 and Nanog expression in wild-type ES cells [30]. It remains unclear what gene encodes the upstream regulator in this network. In our study, 2i contributed to suppressing the fluctuation in the expression of pluripotency genes, and significantly enhanced the correlation among Pou5f1, Sox2, Nanog, and Zfp42. 2i reduces Dnmt3a and Dnmt3b expression and diminishes global levels of DNA methylation in ES cells [31]. It is assumed that this function leads to suppression of fluctuation in the expression of pluripotency genes and reinforces the correlation of fluctuation among the pluripotency genes close to that of Dnmt TKO ES cells. Although the complete loss of DNA methylation has a deleterious role in the differentiation of ES cells [32], 2i might also affect expression fluctuations in a population of ES cells through mildly reduced DNA methylation. Independent fluctuations in the expression of essential pluripotency genes may be crucial to maintain ES cell characteristics, and the epigenome plays an important role in this maintenance mechanism.

The fluctuation in gene expression that we observed may be akin to a wave in the sea. The wave that has left will inevitably be back. As long as the characteristics of ES cells are maintained, fluctuation in pluripotency gene expression would be limited within a certain range. Complete steadiness in gene expression may make it difficult to maintain pluripotency. We revealed here the fluctuations in the expression of pluripotency genes and described the epigenetic barrier against the propagation of fluctuations in the expression of core pluripotency genes in ES cells.

Acknowledgements

The authors thank Dr Takayuki Abe (Department of Preventive Medicine and Public Health, Keio University School of Medicine) for his advice on biostatistics analysis and all laboratory members for their critical comments and helpful discussions. This study was supported, in part, by research grants from Grants-in-Aid for Scientific Research (JSPS KAKENHI Grant Number: 15H01521, 16H02651, 16H050304, 16K15415), and Keio University Medical Science Fund.

    Conflict of interest

    KF is a Founding Scientist and owns equity of Heartseed.