Remodeling of epigenome and transcriptome landscapes with aging in mice reveals widespread induction of inflammatory responses

Abstract

Aging is accompanied by the functional decline of tissues. However, a systematic study of epigenomic and transcriptomic changes across tissues during aging is missing. Here, we generated chromatin maps and transcriptomes from four tissues and one cell type from young, middle-aged, and old mice-yielding 143 high-quality data sets. We focused on chromatin marks linked to gene expression regulation and cell identity: histone H3 trimethylation at lysine 4 (H3K4me3), a mark enriched at promoters, and histone H3 acetylation at lysine 27 (H3K27ac), a mark enriched at active enhancers. Epigenomic and transcriptomic landscapes could easily distinguish between ages, and machine-learning analysis showed that specific epigenomic states could predict transcriptional changes during aging. Analysis of data sets from all tissues identified recurrent age-related chromatin and transcriptional changes in key processes, including the up-regulation of immune system response pathways such as the interferon response. The up-regulation of the interferon response pathway with age was accompanied by increased transcription and chromatin remodeling at specific endogenous retroviral sequences. Pathways misregulated during mouse aging across tissues, notably innate immune pathways, were also misregulated with aging in other vertebrate species-African turquoise killifish, rat, and humans-indicating common signatures of age across species. To date, our data set represents the largest multitissue epigenomic and transcriptomic data set for vertebrate aging. This resource identifies chromatin and transcriptional states that are characteristic of young tissues, which could be leveraged to restore aspects of youthful functionality to old tissues.

Figure 1.

A genome-wide epigenomic and transcriptomic landscape in four tissues and one cell type during mouse aging. (A) Experimental setup (see Supplemental Table S1). (B) Example genome browser region showing tracks of data sets in cerebellum tissue at different ages. (Chr) Chromosome. (C–F) Multidimensional scaling analysis results across data sets based on RNA expression (C), H3K4me3 peak intensity (D), H3K4me3 peak breadth (E), or H3K27ac peak intensity at all peaks (F). For RNA-seq data, the input was a matrix of log₂-transformed DESeq2 1.6.3 normalized counts. For chromatin marks, the most intense or broadest peak associated with a gene was used when more than one peak was present, and the log₂-transformed DESeq2 1.6.3 normalized intensity or breadth was used as input.

Figure 2.

Separation of samples across tissues and cell types as a function of age. Multidimensional scaling analysis results across samples derived from specific tissues, liver and cerebellum, based on RNA expression (A,B), H3K4me3 peak intensity (C,D), H3K4me3 peak breadth (E,F), H3K27ac peak intensity (all peaks: G,H; super-enhancers only: I,J). For RNA-seq data, the input was log₂-transformed DESeq2 1.6.3 normalized counts. For chromatin marks, the most intense or broadest peak associated with a gene was used when more than one peak was present, and the log₂-transformed normalized intensity or breadth was used as input.

Figure 3.

Machine-learning analysis reveals that changes in enhancer score and H3K4me3 domain breadth with age can predict transcriptional aging. (A) Scheme of the three-class machine-learning pipeline. (NNET) Neural network, (SVM) support vector machine, (RF) random forest, (GBM) gradient boosting machine. (B,C) Balanced classification accuracy over the three classes across tissues for random forest models (B) or gradient boosting machine models (C). The accuracy of the model trained in a specific tissue on the same tissue (e.g., the liver-trained model on liver data) is measured using held-out validation data. For cross-tissue validation, the entire data of the tested tissue were used. ‘Random’ accuracy illustrates the accuracy of a meaningless model (∼50%). All tests were more accurate than random. The robustness of the prediction is supported by the fact that samples for RNA and chromatin profiling were collected from independent mice at two independent times (Supplemental Table S1A). Balanced accuracy across the three classes is reported. (D,E) Feature importance from random forest models (D; Gini score and mean decrease in accuracy) or gradient boosting machine models (E; Gini score). High values indicate important predictors. See two-class models in Supplemental Figure S3. Note that two-class models, though containing less biological information, outperformed three-class models, which is consistent with the increased complexity of a classification problem with the number of classes to discriminate.

Figure 4.

Misregulated pathways during aging reveal the activation of an inflammatory innate immune signature. (A,B) Venn diagram for the overlap of significantly up-regulated (A) or down-regulated (B) genes with aging in each tissue called by DESeq2 1.6.3 at FDR < 5%. (C–F) Functional pathway enrichments (C,E–G) and transcription factor (TF) target enrichments (D) using the minimum hypergeometric test for differential RNA expression (C,D), H3K4me3 intensity (E), H3K4me3 breadth (F), and H3K27ac intensity (all enhancers) (G). Enriched pathways were plotted if four out of the six tests (RNA) or three out of the five tests (chromatin marks) were significant (FDR < 5%). (H) Heat maps of expression for all repetitive elements with significant differential expression with aging (TEtranscripts quantification and DESeq2 1.16.1 statistical test at FDR < 5%). Analysis of repetitive elements using HOMER, and overlap with TEtranscripts, is reported in Supplemental Table S6A–E.

Figure 5.

Age-related transcriptional signatures are overall conserved across vertebrate species. (A) Functional enrichments using the minimum hypergeometric test for differential RNA expression with aging in mouse, rat, human, and killifish samples. The mouse data are a subset of Figure 4C and are plotted as a reference. (B) DESeq2 1.6.3 normalized log₂ fold changes per unit of time for genes orthologous to differentially expressed mouse genes in rat, human, and killifish samples. The mouse data are plotted for comparison. P-values were calculated using a one-sample, one-sided Wilcoxon test to test the differences between observed fold changes and 0 (i.e., no change with age). Only male data are plotted. Data with the contribution of females (when available) are in Supplemental Figure S8B.