Dissecting the dynamic changes of 5-hydroxymethylcytosine in T-cell development and differentiation

Abstract

The discovery of Ten Eleven Translocation proteins, enzymes that oxidize 5-methylcytosine (5mC) in DNA, has revealed novel mechanisms for the regulation of DNA methylation. We have mapped 5-hydroxymethylcytosine (5hmC) at different stages of T-cell development in the thymus and T-cell differentiation in the periphery. We show that 5hmC is enriched in the gene body of highly expressed genes at all developmental stages and that its presence correlates positively with gene expression. Further emphasizing the connection with gene expression, we find that 5hmC is enriched in active thymus-specific enhancers and that genes encoding key transcriptional regulators display high intragenic 5hmC levels in precursor cells at those developmental stages where they exert a positive effect. Our data constitute a valuable resource that will facilitate detailed analysis of the role of 5hmC in T-cell development and differentiation.

Keywords: epigenetics; thymic development.

Fig. 1.

5hmC is enriched in the gene body of highly expressed genes. (A) Proportion of the mm9 genome covered by HERGs in each indicated cell type. (B) Proportions of HERGs that fall into annotated genomic regions for each indicated cell type. TTS indicates transcription termination site. A magnified view of promoters and intragenic regions (other than introns) is shown at right. Comparison with the top bar (representation of the annotated genomic regions in the mm9 reference genome) shows that 5hmC is enriched at promoters, exons, and introns. (C) Average 5hmC enrichment over the gene body, categorized based on gene expression levels in the indicated cell types (SI Appendix, Table S4). The arrows indicate differentiation pathways, from precursor to progeny cell type. (D) Average 5hmC enrichment over extremely highly expressed genes (RPKM ≥ 100) in DP, CD4 SP, and CD8 SP cells. Gene body 5hmC levels in highly expressed genes are higher in CD4 SP and CD8 SP cells compared with DP thymocytes. E, as in D, but here the average 5hmC enrichment profiles are calculated over extremely highly expressed genes (RPKM ≥ 100) in naive CD4, Th1, and Th2 cells. Even in highly expressed genes, gene-body 5hmC levels drop in Th1 and Th2 cells compared with their precursor naive CD4 T cells.

Fig. 2.

Intragenic 5hmC levels correlate with RNA Pol II occupancy and with histone marks associated with active transcription. (A) Density plots depicting gene expression (log10 RPKM) and intragenic 5hmC (log2 CMS-IP/Input) in DP, CD4 SP, CD8 SP, naive CD4, naive CD8, and Th1 and Th2 cells. The Spearman correlation coefficient ρ is positive in all cases, reflecting the positive correlation of the two examined parameters; the weak correlation observed for Th1 and Th2 cells (Right panels) most likely reflects the low level of 5hmC in these cells. Arrows indicate the directions of lineage specification and differentiation. (B) Heat map depicting gene expression (log10 RPKM), intragenic 5hmC (log2 IP/Input) evaluated by GLIB and CMS-IP, and intragenic H3K4me3, H3K36me3, RNA polymerase II (Pol II), and H3K27me3 in DP T cells. Each row represents a gene, ordered from top to bottom by gene expression in DP cells. The cutoff RPKM of 1 was used to distinguish expressed and nonexpressed genes. (C) Density plots depicting the correlation of gene-body 5hmC (GLIB) (y axis, Upper) and gene expression (y axis, Lower) with gene-body H3K27me3, H3K4me3, H3K36me3 marks and RNA Pol II (Pol II). All values are represented as log10 RPKM; RPKM values are calculated from TSS to TTS. For each panel, the Spearman rank correlation coefficient ρ is shown (exact permutation test for testing the null hypothesis that there is no correlation, two tailed, *P < 2.2 × 10⁻¹⁶).

Fig. 3.

Portraits of genes in DP and ES cells demonstrating the intragenic distribution of 5hmC and marks of active transcription. Genome browser views of the distribution of 5hmC (GLIB), H3K36me3, and RNA polymerase II (Pol II) around the Bcl11b, Ets1, and Tcf7 genes in DP thymocytes and ES cells. 5hmC (GLIB) is shown in black, H3K4me3 in red, H3K36me3 in blue, and RNA polymerase II in green.

Fig. 4.

5hmC is enriched in active tissue-specific enhancers. (A) Average enrichment of 5hmC (log2 GLIB/Input) (y axis) from DP (red line) and mouse embryonic stem cells (blue line) in thymus-specific (n = 5,605) (Left), mESC-specific (n = 8,552) (Center), and MEF-specific (n = 8,230) (Right) enhancers. The enrichment is quantified ±1.5 kb of the reported enhancer center positions (35). The shaded regions depict the SDs of the means. (B) Scatter density plot evaluating the coenrichment of 5hmC, H3K4me1, and H3K27ac in all thymus enhancers (n = 39,071). The enrichments are quantified over the regions defined as ±1.5 kb of the reported enhancer center positions (35). (C) Histogram showing enrichment of 5hmC in thymic enhancers during the specification of DP T cells to CD4 and CD8 SP T cells. The enrichments are quantified over the regions defined as ±1.5 kb of the reported enhancer center positions (35).

Fig. 5.

5hmC changes dynamically over the course of T-cell lineage specification. (A) Heat map representation of 6,509 genes hierarchically clustered according to their 5hmC patterns in gene sets (Experimental Procedures). Genes (≥1 kb) with intragenic 5hmC (log2 CMS-IP/Input > 1) in at least one of the five studied subsets (DP, CD4 SP, CD8 SP, naive CD4, naive CD8) were considered in the clustering analysis. The six identified clusters are numbered and differently colored, and the number of genes per cluster is shown on the left. Six transcriptional regulators in the second cluster that exert a significant role in shaping T-cell identity are indicated (Right). (B) Close-up view of the six transcriptional regulators (Tcf7, Bcl11b, Satb1, Gata3, Runx3, Zbtb7b) highlighted in A and clearly showing the differential intragenic distribution of 5hmC in these different T-cell subsets. (C) Scatter plots depicting the change in gene expression (y axis) versus the change in intragenic 5hmC (x axis) in CD4 SP and DP cells (Left, n = 236) and CD8 SP and DP cells (Right, n = 242). Genes with intragenic 5hmC (log2 CMS-IP/Input > 2) in at least one of the two cell types examined in each plot were considered; in addition, the logarithmic fold change of 5hmC (5hmC log2 fold change) greater than 0.5 and the logarithmic fold change in gene expression (RPKM log2 fold change) greater than 2 were required. Black dots depict genes that show the same direction of change in gene expression and intragenic 5hmC, whereas red dots indicate genes that show opposite changes. The depicted genes are listed in SI Appendix, Table S7. In each plot, the Spearman rank correlation coefficient ρ is shown (exact permutation test for testing the null hypothesis that there is no correlation, two tailed, *P < 2.2 × 10⁻¹⁶).

Fig. 6.

Portraits of intragenic 5hmC in T-cell lineage-specifying transcription factors. (Left) Genome browser views of 5hmC (evaluated by CMS-IP, gray) in (Top) the Zbtb7b gene, encoding the transcription factor ThPOK that regulates the CD4 lineage; (Middle) Gata3; and (Bottom) Runx3, the lineage-determining factor of the CD8 lineage. A black arrow above the gene symbol indicates the direction of transcription. Bar graphs depict RPKM values evaluated over the gene body (y axis) for gene expression and 5hmC in each of seven cell subtypes (DP in pink, CD4 SP in white, CD8 SP in purple, naive CD4 in gray-blue, naive CD8 in yellow, Th1 in green, and Th2 in bright blue).

Fig. 7.

Genome browser views of 5hmC distribution at additional genes with key roles in T-cell biology. Arrows show the direction of transcription. Bar graphs depict RPKM values summed over the gene body (TSS to TTS, y axis) for gene expression and 5hmC in each of the seven cell subtypes. (Left) The genes encoding the transcription factors Tcf7 and Bcl11b and the chromatin-associated protein Satb1 are most highly expressed in DP T cells, and their expression is considerably diminished in progeny CD4 and CD8 SP cells; notably, however, 5hmC levels either diminish more slowly (Tcf7) or increase slightly (Bcl11b, Satb1) from DP to SP T cells. (Right) Genes encoding proteins implicated in T-cell receptor signaling during T-cell development also show gene-body 5hmC at early but not developmental stages that is almost extinguished in terminally differentiated Th1 and Th2 despite continuing expression (Cd4, Ets1, Itk).