Methylome of human senescent hematopoietic progenitors

Human in vivo senescent CD34+ HSPC were obtained and isolated from healthy subjects as previously described [16]. Briefly, cells were drained from leukocyte reduction system cones collected from healthy platelet donors. HSPCs were enriched using the RosetteSep human progenitor enrichment cocktail (StemCell Technologies). Following a 20-min incubation with the enrichment cocktail at room temperature, mononuclear cells were isolated by density gradient centrifugation using Ficoll-Paque premium (GE Healthcare). The mononuclear cells were suspended in IMDM containing 100 nM bafilomycin A1 for 1 h at 37 °C, followed by incubation with C₁₂FDG at 37 °C for 90 min. Then, the cells were washed with sorting buffer and stained with PE-conjugated anti-CD34, PE-Cy7-conjugated anti-CD38, and APC-conjugated anti-CD45 (eBiosciences) and subjected to fluorescence-activated cell sorting (FACS) (BD FACS Aria III). Dead cells were excluded based on propidium iodide staining. Samples were sorted for CD34+ and CD45dim+ cells to identify HSPCs and then gated for C₁₂FDG staining for senescence-associated beta-galatosidase (SA-βgal) expression [29]. To identify CD38+ and CD38− populations, we used PE-Cy7-conjugated anti-CD38 antibody (eBiosciences). The protocol was approved by the Institutional Review Board (Protocol # IRB-HS-12-00693).

Libraries for whole-genome bisulfite sequencing were generated from 5 ng of purified DNA from paired senescent and active CD34+ cells from 3 healthy human donors (33, 45, and 53 years of age) using the NuGen Ovation Ultralow MethylSeq Kit following the manufacturer’s protocol, for a total of 6 individual libraries. Samples were not pooled prior to library generation. Reads were aligned using Biscuit and Metilene was used for calling of differentially methylated regions. Motif analysis of DMRs was conducted using the PWMEnrich package with Hocomoco and Factorbook motif databases provided in the motifbreakR package [30].

Reads were aligned using biscuit and post-processed using biscuiteer [31] prior to compartment inference. Briefly, observed CpG loci were restricted to “open sea” regions and smoothed using a Dirichlet smoothing approach. Loci that had less than 3× coverage were set to values of NA and any locus that had greater than 50% NAs were removed from the dataset. Remaining NAs were imputed using k-nearest neighbors [32, 33]. Chromatin confirmation inference was performed as described previously [34] and implemented in compartmap [35].

In order to assess if epigenetic alterations such as changes in DNA methylation contributed to increased expression of TEs in senescent HSPCs, we subjected paired senescent and active HSPCs from 3 healthy humans to WGBS. We employed a modified FACS protocol [16] to identify and isolate senescent and non-senescent HSPCs (CD34+ CD45dim+ cells) from peripheral blood of healthy platelet donors. Senescence-associated β-galactosidase (SA-β-gal) was used as a senescence marker [12‐15, 36, 37] to identify and isolate circulating senescent HSPCs as previously described [16]. Using this technique, we isolated between 2800 and 4800 senescent HSPCs and 250,000 to 360,000 non-senescent HSPCs from each of our three individual donors. All samples were standardized to an input of 5 ng prior to WGBS library generation.

WGBS yielded 7.5× genome coverage. Using the 2-dimensional Komogorov-Smirnov approach implemented in Metilene with cutoffs of 10% minimum difference, 10 CpGs and 10% FDR, we identified 61 differentially methylated regions (DMRs) in senescent vs. active HSPCs, of which 51 were hypo-methylated (hypoDMRs) and 10 hyper-methylated (hyperDMRs) (Additional file 1: Table S1). Multi-resolution chromatin conformation inference for non-senescent and senescent HSPCs revealed representative genome-wide changes in open/closed compartments between cell types (Fig. 1). DNA methylation changes in the senescent cells were focal rather than global (Figs. 1 and 2). By mapping the hypoDMRs to chromatin states using ChIP-seq data of primary human CD34+ cells, we found the majority of DMRs to overlap with transcriptional enhancers (Fig. 3b). CCAAT/enhancer binding proteins (CEBPA/B/G) were the dominant motif (8/10 top hits) in hypoDMRs (Fig. 3b). CBFA2T2, TIMM44, and Myc-associated factor X were among the top hits in hyper DMRs (Fig. 3b).

Next we examined whether the hypoDMRs correlated with the repeat regions in the genome. Interestingly, all of the hypoDMRs (51/51) but only 4/10 of the hyperDMRs overlapped with repeat elements (Fig. 4, Additional file 2: Figure S1 and Additional file 3: Table S2). Based on the observation that ~ 50% of the genome is repetitive and that an overlap may equally likely affect hyper- and hypo-DMRs, Fisher’s Exact test showed a significantly increased occurrence of repeat elements in the hypoDMRs (P < 10⁻⁶). This led us to conclude that senescent HSPCs display focal loss of DNA methylation in the repetitive DNA-containing DMRs.