DNA methylation is dispensable for the initiation of differentiation programs As we found that hypomethylated cells are capable of partially down regulating pluripotency

associated genes, we next asked whether these cells are able to initiate differentiation

Results

56

programs. To address this question, we performed genome- wide expression analysis of wt, dnmt1^-/- and TKO cells at three different time points: i) in the undifferentiated state (day 0), ii) during early differentiation (day 4) and iii) during a late differentiation (day 16) stage in duplicate experiments. The Microarray hybridizations were performed by our collaborators Kerstin Maier and Dietmar Martin (Laboratory of Patrick Cramer) and the initial processing of the microarray data was done by our collaborators Benedikt Zacher and Achim Tresch (Laboratory of Achim Tresch). They performed the Principal Component Analysis (Fig. 14), identified differentially regulated genes, compared differentially expressed genes in undifferentiated TKO ESCs identified in our study to published data (Fig. 16B) and did the analysis of bivalent genes and all gene changes between d0-4 (Fig. 21, 23). I performed all other analyses, including GO, Kegg pathway, chromosomal location and tissue expression analysis using DAVID software (Huang et al., 2008, 2009; for details see chapter 2.2.2.4).

Differentially regulated genes identified at the various time points are listed in chapter 6.

The Principal Component Analysis (PCA) revealed that the microarray data can be separated according to day of differentiation and genotype (Fig. 14).

Figure 14. Principal Component Analysis of genome- wide expression data from wt, dnmt1^-/- and TKO ESCs during differentiation

RNA samples from undifferentiated ESCs and respective EBs at day 4 and 16 of differentiation were subjected to microarray expression analysis. Data from independent biological duplicates per cell line and per time point were processed by two dimensional principal component analysis. Two principal components (PC), PC1 and PC2, were identified (see also text for details).

In the undifferentiated state, all cell lines were closely clustered, suggesting that loss of DNA methylation in the undifferentiated state affects the expression of relatively few genes.

However, principal component 1 provides a clear differentiation signature, as it increases continuously with culture time. This increase was less pronounced among the knockout genotypes, reflecting a crucial role of DNA methylation during differentiation. Intriguingly, at both differentiation time points, dnmt1^-/- EBs took an intermediate position on the median line between TKO and wt EBs, again emphasizing the less severe phenotype of dnmt1^-/- EBs compared to TKO EBs.

Results

57

Using a two- fold cutoff and false discovery rate below 5 %, we calculated gene expression changes in undifferentiated dnmt1^-/- and TKO ESCs relative to wt ESC lines (Fig. 15A).

Furthermore, we determined gene expression changes for each genotype between the undifferentiated ESC state and day 4 EBs (Fig. 15B), as well as between day 4 and day 16 EBs (Fig. 15C). As suggested by PCA, the expression of only few genes was altered in undifferentiated hypomethylated ESCs (54 genes in dnmt1^-/- ESCs, 82 genes in TKO ESCs).

As DNA methylation is mostly considered a repressive mark involved in silencing of genes (Siegfried et al., 1999), we reasoned that genes being down regulated in the knock outs are most likely the results of indirect effects and not due to the lack of DNA methylation.

Therefore we focused most of the analysis on the set of up regulated genes in the knock out ESCs and EBs.

Figure 15. Analysis of genome- wide expression data from wt, dnmt1^-/- and TKO ESCs during differentiation.

A) Venn diagram of differentially expressed genes in dnmt1^-/- (blue circles) and TKO (green circles) ESCs compared to wt ESCs. (B, C) Venn diagram of gene expression changes occurring between day 0-4 (C) and day 4-16 (C) in knockout and wt (red circles) EBs. Expression changes occurring in wt EBs between day 0-4 and day 4-16 were identified and compared to expression patterns in dnmt1^-/- and/ or TKO EBs. Larger numbers indicate total numbers of gene changes, whereas smaller numbers refer to up- (↑) and down- (↓) regulated transcripts in the respective sectors. Data were averaged from independent biological duplicates.

Gene ontology (GO) enrichment analysis of the 57 genes up regulated in TKO ESCs revealed that the most enriched categories are involved in reproductive processes like gamete generation, oogenesis and spermatogenesis (Fig. 16A).

Results

58

Figure 16. Gene ontology enrichment of transcriptome analysis of TKO ESCs compared to wt ESCs (A) 57 genes were up regulated in TKO ESCs compared to wt ESCs (Fig. 15) and can be grouped in 12 GO categories. (B) Venn diagram showing the comparison of differentially expressed genes in TKO ESCs identified in the current study and deregulated genes identified by Fouse et al. (Fouse et al., 2008) in compound dnmt3a/3b -/-ESCs with stable Dnmt1 knockdown (TKO*; black circles). To directly compare differentially expressed genes in the two studies, the same cut off criteria based solely on a fold change >2 were used. These less stringent cut off criteria lead to a significant higher number of deregulated genes in our TKO ESCs (274 genes compared to 82 with additional cut off criteria of p<0.05). However, only 60 commonly differentially regulated genes between our TKO ESCs and the 336 genes identified by Fouse et al. can be identified. Those genes can also be grouped within the same GO categories comparable to the ones in (A). A comparison of the GO terms associated with the individual genes in the lists from the two studies shows a significant functional overlap (red numbers in red shaded circle). 159 genes identified from our study are associated with the same GO terms as genes from Fouse et al. and vice verse, 183 genes from their data are linked to the same GO terms as genes from our data set.

In contrast to the data obtained from TKO ESCs, no enriched categories could be identified for the 49 up regulated genes in dnmt1^-/- ESCs. However, in both mutant cell lines, deregulated genes were mainly located on the X chromosome (dnmt1^-/- 13.1 %; TKO 40.4

%). In this context, it is important to note that both wt and knock out ESCs are derived from male (XY) mice and hence lack X chromosome inactivation. Therefore, the relatively high amount of X- linked, upregulated genes in both hypomethylated ESCs is not linked to a deregulated X inactivation process. Furthermore, in dnmt1^-/- ESCs many deregulated genes are also located on the Y chromosome (13.3 %). The up regulated transcripts of both cell lines were mostly genes known to exhibit testis-specific expression (dnmt1^-/- 57.7 %; TKO 61.5 %). These data are consistent with previous results from Fouse and colleagues (Fouse et al., 2008), who performed genome-wide expression analysis on a dnmt3 double knock ESC line (dnmt3a^-/-, dnmt3b^-/-) with constitutive knockdown of dnmt1 (TKO*). Surprisingly though, using the same cut off criteria (fold change >2) as in Fouse et al., only 60 out of 274 differentially expressed genes identified in our study for TKO were also found to be deregulated in the previous study (Fig. 16B). This may at least in part be due to the less

Results

59

stringent cut off criteria used in Fouse et al. However, when we compare the GO terms associated with the differentially regulated genes, we found a significant functional overlap between these two studies.

To investigate the role of DNA methylation during differentiation, we first determined the gene changes occurring in wt EBs between day 0-4 and day 4-16 and used this as a baseline of expression changes normally occurring during ESC differentiation. We then compared these changes to the gene changes occurring in the knock outs at the respective days. This strategy allowed us to identify concordant gene changes of wt and knock out EBs which are regulated independently of DNA methylation (Fig. 15B, C). In the first differentiation period (d0-4), approximately the same number of genes in wt (1255) and dnmt1^-/- EBs (1230) were altered and two thirds (833) of the genes concordantly changed in the two EB populations. In comparison, 819 genes changed their transcript level in TKO EBs, which translates to about two thirds of gene changes compared to those in wt EBs. However, only one third of the genes (376) changed concertedly in wt and TKO EBs (Fig. 15B).

Strikingly, in the second differentiation period (d4-16) total transcript level changes in dnmt1 -/-(879) and TKO EBs (406) were roughly half and one fifth of those in wt EBs (1808), respectively (Fig. 16C). Furthermore, one third (593) of the genes changing in dnmt1^-/- but only one tenth (187) of transcripts altered in TKO are concordantly regulated compared to wt EBs. The observation of a high proportion of concerted gene expression changes in mutant and wt EBs during the early differentiation stage, suggests that hypomethylated cells are able to induce differentiation programs to some extent. However, the decreased amount of concordant transcription level changes between day 4-16 implies that the progression of differentiation programs in hypomethylated EBs is impaired, in particular in TKO EBs, indicating that the presence of Dnmt3 proteins could enable hypomethylated dnmt1^-/- EBs the execution of differentiation programs to similar extents as in wt EBs.

Intriguingly, we detected 365 concerted gene changes in EBs of all three genotypes during the first four days of differentiation (Fig. 15B). GO analysis of the 198 conjointly up regulated genes yielded categories involved in developmental processes like anatomical structure development, system and organ development and cell differentiation being the most enriched (Fig. 17A).

Results

60

Figure 17. Gene ontology enrichment and cell type specific expression of concomitantly regulated genes in wt, dnmt1^-/- and TKO EBs during day 0-4 of differentiation

From the 365 concordant genes changes (Fig. 15), 198 genes were upregulated and are associated with 47 GO terms (A) involved in various developmental processes. On the contrary, 167 genes were concordantly down regulated and are connected to stem cell maintenance and development GO categories (B). The latter set of genes is specifically expressed in ESCs and EGCs specifically (C).

Furthermore, genes playing roles in transcription, metabolism and signaling pathways like the wnt receptor and tyrosine kinase pathways were identified by Kegg pathway analysis (not shown). By contrast, the expression of 167 genes was down regulated in all three cell lines and they could be categorized into genes involved in stem cell maintenance and development and are known to be mainly expressed in ESCs and embryonic germ cells (EGCs) (Fig. 17B, C).

Results

61

These data clearly demonstrate that all cell lines independent of their genotype were able to activate differentiation programs by down regulating genes associated with stem cell fate and up regulating genes required for developmental processes.

3.1.4 Cells lacking Dnmt1 possess a greater differentiation potential than TKO cells