Continual removal of H3K9 promoter methylation by Jmjd2 demethylases is vital for ESC self-renewal and early development

Abstract

Chromatin-associated proteins are essential for the specification and maintenance of cell identity. They exert these functions through modulating and maintaining transcriptional patterns. To elucidate the functions of the Jmjd2 family of H3K9/H3K36 histone demethylases, we generated conditional Jmjd2a/Kdm4a, Jmjd2b/Kdm4b and Jmjd2c/Kdm4c/Gasc1 single, double and triple knockout mouse embryonic stem cells (ESCs). We report that while individual Jmjd2 family members are dispensable for ESC maintenance and embryogenesis, combined deficiency for specifically Jmjd2a and Jmjd2c leads to early embryonic lethality and impaired ESC self-renewal, with spontaneous differentiation towards primitive endoderm under permissive culture conditions. We further show that Jmjd2a and Jmjd2c both localize to H3K4me3-positive promoters, where they have widespread and redundant roles in preventing accumulation of H3K9me3 and H3K36me3. Jmjd2 catalytic activity is required for ESC maintenance, and increased H3K9me3 levels in knockout ESCs compromise the expression of several Jmjd2a/c targets, including genes that are important for ESC self-renewal. Thus, continual removal of H3K9 promoter methylation by Jmjd2 demethylases represents a novel mechanism ensuring transcriptional competence and stability of the pluripotent cell identity.

Keywords: Kdm4; development; epigenetics; histone demethylation; transcription.

© 2016 The Authors.

Figures

Figure 1. Jmjd2a and Jmjd2b are individually dispensable for ESC self‐renewal and embryogenesis

1. A, B

   Schematic representation of the KO strategies used for the conditional deletion of Jmjd2a (A) and Jmjd2b (B). FRT: flippase recognition site, LoxP: Cre recombinase recognition site.

2. C–E

   2a and 2b ESCs maintained in 2i medium were cultured in the absence (Ctrl) or presence of OHT and subsequently used for WB (C), growth curves and RT‐qPCR (D, E).

3. F, G

   2a and 2b ESCs were adapted to grow in serum‐containing ESC medium, exposed to OHT as indicated and subsequently used for growth curves and RT‐qPCR analyses.

4. H, I

   Jmjd2a(+/−); Jmjd2a(+/−) (H) or Jmjd2b(+/−); Jmjd2b(+/−) (I) mice were intercrossed and the number of pups alive at time of weaning are reported. P‐values were calculated using a chi‐square test.

Data information: Data in panels (C–G) are representative of results obtained with at least two independently derived ESC lines of each genotype. Graphs show mean ± standard deviation (SD) for three cultures counted in parallel (growth curves) or three technical replicates (RT‐qPCR).Source data are available online for this figure.

Figure 2. The combined functions of Jmjd2a and Jmjd2c are essential for ESC self‐renewal

1. A–D

   WB and growth curves of 2ac (A), 2abc (B), 2ab (C) and 2bc (D) ESCs cultured in 2i medium and exposed to OHT as indicated.

2. E

   Flow cytometric profiles of 2ac ESCs pulsed with BrdU in 2i medium. The upper table shows the percentages of living cells (the subG1 fraction is omitted from the analysis), which can be classified as being in S (BrdU+), G1 (2N DNA content, BrdU−) or G2/M phase (4N DNA content, BrdU−). The percentage of dead cells (subG1, events with a < 2N DNA content) are shown in red.

3. F

   RT‐qPCR analyses for ESCs cultured in 2i medium.

4. G–J

   2ac and 2abc ESCs were adapted to grow in serum‐containing ESC medium, exposed to OHT as indicated and subsequently used for growth curves (G, H) and RT‐qPCR analyses (I, J). The RT‐qPCR data are also presented in Appendix Fig S2C and D with a zoom‐in on the lower region of the y‐axis.

Data information: Data are representative of results obtained with at least two independently derived ESC lines of each genotype, and graphs show mean ± SD for three cultures counted in parallel (growth curves) or three technical replicates (RT‐qPCR).Source data are available online for this figure.

Figure EV1. Lack of both Jmjd2a and Jmjd2c impairs mitotic progression in ESCs

1. A–C

   2ac and 2abc ESC clones (A) and 2a and 2c ESC clones (B) expressing H2B‐Venus were cultured in 2i medium in the absence or presence of OHT and subsequently subjected to time‐lapse imaging. Y‐axes show the duration of mitosis (from peak chromosome condensation to anaphase onset) for individual cells. Red lines show median values. Data are representative of results obtained in at least two different experiments. (C) Representative images demonstrating proper chromosome alignment after 20 min in mock, but not OHT‐treated 2ac ESCs.

Figure 3. The combined functions of Jmjd2a and Jmjd2c are essential for early embryonic development

1. A, B

   2ac ESCs expressing GFP were cultured in the absence or presence of OHT and subsequently used for morula injections. (A) Average volume of the GFP‐positive area in chimeric embryos at 24 and 72 h after morula injection. Number of embryos used for quantification: Ctrl 24 h n = 18; +OHT 24 h n = 15; Ctrl 72 h n = 5; +OHT 72 h n = 5. Data are presented as mean ± SD. (B) Representative images of chimeric embryos. The white bars indicate a length of 32 μm. Relative intensities of the GFP signal are indicated for each image.

2. C, D

   Jmjd2a(+/−);Jmjd2c(+/−) mice were intercrossed. The number of (C) pups alive at weaning or (D) embryos recovered at E6.5 is presented. P‐values were calculated using exact binomial test with the most conservative estimate of expected (−/−;−/−) embryos (6.25%, assuming no linkage between the Jmjd2a and Jmjd2c loci even though they are located on the same chromosome).

Figure 4. Jmjd2a and Jmjd2c both localize to H3K4me3‐marked TSSs

1. A

   Examples of ChIP‐seq data obtained for Jmjd2a, Jmjd2c (Pedersen et al, 2014) and H3K4me3.

2. B

   Diagram illustrating the distribution of Jmjd2a peaks with regard to TSSs and open reading frames. TES: transcription end site.

3. C

   Unsupervised k‐means clustering of ChIP‐seq tags over all TSSs (± 5 kb).

4. D, E

   Bar graphs showing the number of Jmjd2a peaks displaying at least 1 bp overlap with (D) a H3K4me3 or (E) a Jmjd2c peak. Jmjd2a peaks were classified as TSS associated (“TSS”) if overlapping a region of ± 1 kb of a TSS.

5. F

   Venn diagram illustrating the number of TSSs containing binding sites for Jmjd2a and/or Jmjd2c within ± 1 kb. The P‐value was calculated using hypergeometric test.

6. G

   ChIP‐qPCR validations. Graphs show mean ± SD for three technical replicates and are representative of results obtained in at least two independent experiments.

Data information: All data were obtained using ESCs cultured in 2i medium.

Figure EV2. 2ac and 2abc KO ESCs show reproducible and comparable changes in H3K9me3 and H3K36me3 patterns

1. A

   WB analyses of 2ac and 2abc ESC lines. Results are representative of at least three independent experiments.

2. B

   Quantification of the histone modification WB presented in (A). Each membrane was reprobed with an antibody detecting H4 for normalization. Data are presented as log2 fold change (OHT versus control treated).

3. C, D

   Results of (C) principal component analyses and (D) Pearson's correlation coefficient analyses of H3K9me3 and H3K36me3 read counts across all TSS regions (± 1 kb). ChIP‐seq data were generated in two different experiments involving a total of four independently derived ESC lines cultured in the absence or presence of OHT.

Data information: All data were obtained using ESCs cultured in 2i medium.Source data are available online for this figure.

Figure 5. Jmjd2a and Jmjd2c redundantly regulate histone methylation levels

1. Density plots showing average ChIP‐seq signals with 95% confidence intervals of the mean indicated in grey. Y‐axes show mean tags per million (TPM). Data were obtained using the 2ac ESC line #4. TSSs were classified as “Bound” if containing binding sites for both Jmjd2a and Jmjd2c within ± 1 kb (see Fig 4F), or “Not bound” if neither protein binds within the same region.

2. Heat map showing H3K9me3 ChIP‐seq data for Jmjd2a/c bound TSSs (± 5 kb) sorted according to read number in 2abc #13+OHT.

3. ChIP‐qPCR validations for selected Jmjd2a/c targets (see Fig 4G). Graphs show mean ± SD for technical triplicates and are representative of results obtained in at least two independent experiments.

Data information: All data were obtained using ESCs cultured in 2i medium.

Figure EV3. 2ac and 2abc KO ESCs show consistent and comparable changes in H3K9me3 and H3K36me3 distributions

1. Density plots showing average ChIP‐seq signals with 95% confidence intervals of the mean indicated in grey. Y‐axes show mean tags per million (TPM). Data were obtained using the indicated ESC lines. TSSs were classified as “Bound” if containing binding sites for both Jmjd2a and Jmjd2c within ± 1 kb. Note that H3K4me3 and H3 ChIP‐seq data were only obtained for the ESC lines 2ac #5 and 2abc #9.

2. Heat map showing H3K36me3 ChIP‐seq data for Jmjd2a/c bound TSSs (± 5 kb) sorted according to read number in 2abc #13+OHT.

3. ChIP‐qPCR validations exemplifying differential effects on histone methylation. Als2, Mfn1, Ska1, Tcl1 and Pim3 are examples of down‐regulated target genes, Gadd45g is induced, while the expression levels of Gstz1 and Kif15 are unaltered upon loss of Jmjd2a/c expression. Graphs show mean ± SD for three technical replicates and are representative of results obtained in at least two independent experiments.

Data information: All data were obtained using ESCs cultured in 2i medium.

Figure EV4. TSS regions with substantial gain of H3K9me3 show reduced levels of H3K4me3

1. Density plots are presented for the 100 Jmjd2a/c bound TSSs showing the greatest increase in average H3K9me3 levels. Fold change in H3K9me3 read counts was calculated for each ESC line (OHT versus Ctrl) for regions of ± 1 kb of TSSs. Plots show average ChIP‐seq signals with 95% confidence intervals of the mean.

2. Independent ChIP‐qPCR validations. Note that the target genes show impaired expression upon loss of Jmjd2a/c expression. Graphs show mean ± SD for three technical replicates and are representative of results obtained in at least two independent experiments.

Data information: All data were obtained using ESCs cultured in 2i medium.

Figure 6. H3K9me3 accumulation in Jmjd2a/c KO ESCs impairs gene expression

1. A, B

   Independent RT‐qPCR validations of genes identified as being (A) down‐regulated or (B) up‐regulated by expression arrays. Graphs show mean ± SD for technical triplicates and are representative of results obtained with at least two different ESC lines of each genotype in independent experiments.

2. C, D

   Density plots showing average ChIP‐seq signals for (C) H3K9me3 or (D) H3K36me3 with 95% confidence intervals of the mean indicated in grey. Plots are shown for genes, which are down‐ or up‐regulated according to the microarray analyses (FDR < 0.05, absolute fold change log2(|FC|) > 0.5) for both 2ac and 2abc KO ESCs (overlaps presented in Appendix Fig S6A), or do not show strong expression changes (log2(|FC|) < 0.2). Genes are further classified as “Bound” if containing binding sites for both Jmjd2a and Jmjd2c within ± 1 kb of a TSS. ChIP‐seq data are presented for the 2ac #4 ESC line.

3. E

   Heat map presenting array data in the form of log2(fold change) values comparing OHT with control treated ESCs. Genes are shown for which TSS regions (± 1 kb) show the most substantial increase in H3K9me3 levels in OHT‐treated cells according to the ChIP‐seq analyses (defined in Fig EV4).

Data information: All data were obtained using ESCs cultured in 2i medium.

Figure EV5. Consistent accumulation of H3K9me3 at Jmjd2a/c targets with impaired expression

1. A, B

   Density plots showing average ChIP‐seq signals with 95% confidence intervals of the mean. Plots are shown for the gene subsets defined for Fig 6C and D, and data were obtained using antibodies specific for (A) H3K9me3 or H3K36me3 or (B) H3K4me3 or H3.

Data information: All data were obtained using ESCs cultured in 2i medium.

Figure 7. The catalytic activity is essential for Jmjd2a/c functions

1. A–C

   2ac clones established after transfection with an empty vector (empty) or plasmids expressing wild‐type Jmjd2c (2c wt) or a catalytic mutant (2c mut) were exposed to OHT as indicated and subsequently used for (A) WB, (B) RT‐qPCR analyses or (C) serial plating for the generation of growth curves. Data are representative of results obtained in at least two independent experiments, and graphs show mean ± SD for three cultures counted in parallel (growth curves) or three technical replicates (RT‐qPCR).

Data information: All data were obtained using ESCs cultured in 2i medium. Source data are available online for this figure.

Figure 8. Model

We propose the following model: Jmjd2a and Jmjd2c are both recruited through their double Tudor domains to H3K4me3‐marked nucleosomes, where they continuously and redundantly prevent accumulation of H3K9me3 and H3K36me3. In the absence of Jmjd2a/c catalytic activity, substantial increases in H3K9me3 levels at specific loci impair gene expression, which perturb the transcriptional programme required for embryonic stem cell identity, proliferation and embryonic development.