Unique patterns of trimethylation of histone H3 lysine 4 are prone to changes during aging in Caenorhabditis elegans somatic cells

Abstract

Tri-methylation on histone H3 lysine 4 (H3K4me3) is associated with active gene expression but its regulatory role in transcriptional activation is unclear. Here we used Caenorhabditis elegans to investigate the connection between H3K4me3 and gene expression regulation during aging. We uncovered around 30% of H3K4me3 enriched regions to show significant and reproducible changes with age. We further showed that these age-dynamic H3K4me3 regions largely mark gene-bodies and are acquired during adult stages. We found that these adult-specific age-dynamic H3K4me3 regions are correlated with gene expression changes with age. In contrast, H3K4me3 marking established during developmental stages remained largely stable with age, even when the H3K4me3 associated genes exhibited RNA expression changes during aging. Importantly, the genes associated with changes in H3K4me3 and RNA levels with age are enriched for functional groups commonly implicated in aging biology. Therefore, our findings suggested divergent roles of H3K4me3 in gene expression regulation during aging, with important implications on aging-dependent pathophysiologies.

Fig 1. H3K4me3 markings on gene-bodies are enriched for age-dependent H3K4me3 changes.

(A) PCA plot showing normalized H3K4me3 data from three biological replicates. The H3K4me3 peak regions at D2 and D12 were determined using the MACS2 narrow peak calling method. (B) Scatter plot showing normalized H3K4me3 levels at D2 and D12 for stable (grey) or dynamic (brown) H3K4me3 peaks. The plot shows the average normalized H3K4me3 signals from three biological replicates calculated using Homer. (C) Protein-coding genes associated with H3K4me3 peaks were grouped into 25 clusters (a-y) using k-means clustering based on their normalized H3K4me3 levels at D2 and D12 time points. The heatmaps show the log2 ratio of H3K4me3 levels normalized to H3 levels. (D) The bar chart shows the percentage of genes associated with age-dependent dynamic H3K4me3 peaks identified by DiffBind for each of the 25 clusters shown in (C). Clusters k, l, m, n, s and y are significantly enriched for genes associated with H3K4me3 peaks that dynamically changed with age. (*) p-value<0.05, (**) p-value<0.01, Fisher’s exact test. (E.) H3K4me3 peaks spanning gene-bodies are enriched for dynamic changes with age. The percentage of genes associated with age-dependent H3K4me3 change for clusters l to q, where H3K4me3 markings spanned gene-bodies, were compared with that for all the other clusters using Fisher’s exact test.

Fig 2. The age-dynamic H3K4me3 markings are mainly deposited in adults.

(A) The boxplots show the normalized H3K4me3 levels in the 25 clusters at L3, D2 or D12 stages. (B) Average plots of the H3K4me3 levels at larval stage 3 (L3) (blue), D2 (red) and D12 (green) for stable peaks or peaks that showed significant dynamic changes with age identified by DiffBind. Normalized H3K4me3 levels within the peak regions and 2kb upstream and downstream are shown for the peaks that showed significantly decreased (left), increased (middle), or stable (right) H3K4me3 levels with age. (C) Age-dynamic H3K4me3 peaks are enriched for peaks marked with higher levels of H3K4me3 at D2 relative to L3. GSEA analysis was performed as described in Methods. The enrichment scores were computed using the GSEAPreranked tool. An enrichment score of 0.51 represents a statistically significant enrichment for peaks with a high D2/L3 H3K4me3 ratio. (D) H3K4me3 markings deposited during adult stage are enriched for age-dynamic H3K4me3 peaks. DiffBind was used to identify the H3K4me3 peaks with statistically significant differences between D2 and L3 stages. The differential (D2>L3, or D2

Fig 3. Age-dynamic H3K4me3 peaks accompanied by RNA expression change are more likely to span gene-body regions and be deposited during adult-stage.

(A) The Venn diagrams show the protein-coding genes that were associated with H3K4me3 markings in adult stage (blue), significant H3K4me3 changes with age (red), and significant RNA expression changes with age (green). Gene numbers for each group are shown. (B) Age-dependent H3K4me3 changes and gene expression changes are positively correlated. Spearman’s correlation coefficient was calculated by using log2 ratio of normalized H3K4me3 levels and RNA expression levels at D12 relative to D2 for the age-dynamic peak regions and their assigned genes. (S8 Table). rho, Spearman’s correlation coefficient. (C) Genes associated with both age-dependent H3K4me3 and RNA expression changes tend to gain H3K4me3 markings in the adult stage. Average plots show normalized H3K4me3 levels at L3, D2, and D12 for gene groups associated with only age-dependent H3K4me3 changes (left), with both age-dependent H3K4me3 and RNA expression changes (middle), and with only age-dependent RNA expression changes (right). (D) The bar graph shows the percentages of age-dynamic H3K4me3 peaks that were accompanied by RNA expression changes in the l to q clusters (with gene-body H3K4me3) vs the other clusters. Fisher’s exact test shows that the dynamic H3K4me3 peaks in clusters l to q are significantly more likely to be accompanied by age-dependent RNA expression changes. (E.)The lengths of the age-dynamic H3K4m3 peaks are positively correlated with the likelihood of RNA expression change with age. Age-dynamic H3K4me3 peaks assigned only to one gene were used for this analysis. The y-axis represents the fraction of genes exhibiting age-dependent RNA expression change for each group of the indicated peak length (x-axis). The size of the dots indicates the number of peaks (also the number of genes) for each peak length range.

Fig 4. Genes that show downregulated expression when ash-2 was knocked down are more likely to acquire H3K4me3 markings during the adult stage.

(A) For genes exhibiting significantly downregulated expression when ash-2 was knocked down, GSEA revealed a significant enrichment for genes associated with higher D2 H3K4me3 markings compared to L3 (top). For genes exhibiting significantly upregulated expression when ash-2 was knocked down, GSEA revealed no bias (bottom). Genes with detectable expression from the published ash-2 RNAi microarray experiment (GSM756690-GSM756695) were ranked according to the difference in H3K4me3 levels between D2 and L3 (S5 Table). The enrichment scores were computed using the GSEAPreranked tool. The genes that showed significant expression change (FDR <0.05) between ash-2 and control RNAi were identified using Limma in the R package (S9 Table). (B) Average plots show that the genes that exhibited significantly downregulated expression upon ash-2 RNAi (top) were marked with higher levels of H3K4me3 at D2 compared to L3. For the genes that exhibited significant upregulated expression upon ash-2 RNAi (bottom), their average H3K4me3 levels did not change at the different developmental time points. (C) Clusters k, l, w and y were enriched for genes with a downregulated expression upon ash-2 RNAi, and clusters h, i, and j were enriched for genes with an upregulated expression upon ash-2 RNAi. Genes in all of the 25 clusters were ranked according to their expression fold change as detected in the ash-2 microarray experiment. GSEA was used to determine the enrichment of downregulated or upregulated genes in each cluster. (**) FDR<0.01, (*) FDR<0.05.

Fig 5. Fatty acid metabolism genes are overrepresented for age-dynamic H3K4me3 and gene expression.

(A) A significant fraction of the genes annotated to act in the fatty acid biosynthetic (left) or fatty acid oxidation (right) pathways showed decreased RNA expression and/or H3K4me3 levels with age. Green: Genes with decreased RNA expression with age. Blue: Genes with decreased H3K4me3 levels with age. Red: Genes annotated to participate in fatty acid biosynthetic or fatty acid oxidation processes. Gene numbers of each group are shown in the Venn diagrams. (B) Oil-Red-O (ORO) staining of worms at D2, D12 and D16 aging time points (right). Quantification of the ORO signals (left) was done by normalizing the total signal counts in the red channel with the total signal counts in the green channel.

Fig 6. H3K4me3 markings established during adulthood are more likely to be age-dynamic and be accompanied by corresponding RNA expression change.

Our data revealed that the H3K4me3 markings established before adulthood generally remain stable with age. These H3K4me3 marks exhibit the canonical pattern of accumulating to high levels around TSS and correlate with the presence of H3K36me3 markings in gene-bodies. In contrast, the H3K4me3 markings established in adult stage tend to be of lower levels and distribute more evenly into gene-bodies that are marked with very low or non-detectable levels of H3K36me3. The width of the gray arrow indicates the proportion of genes with significant age-dependent H3K4me3 change for each H3K4me3 marking pattern. For the genes associated with significantly increased (blue) or decreased (red) H3K4me3 levels with age, ~35% or 28% were accompanied by corresponding RNA expression change.