Aging of the hematopoietic system in humans is associated with increased incidence of myeloid malignancies. Epigenetic machinery such as DNA methylation is known to change with age, and is disproportionately impacted by recurrent genetic mutations in acute myeloid leukemia (AML). We hypothesized that epigenetic changes in CD34+ hematopoietic stem and progenitor cells (HSPCs) may precede recurrent genetic changes in AML, and might be detected in normal aging HSPCs with increasing frequency. We also hypothesized that areas with increased variability in methylation may be hot-spots for the emergence of epigenetically distinct HSPC clones.

In order to characterize these changes, we performed methylome-wide profiles of human HSPCs from different age groups (20-25 years (10), 40-45 years (10) and >60 years (9)).Adult HSPCs were obtained from Leukocyte Reduction System cones from healthy platelet donors. CD34+ cells were then isolated by Fluorescence Assisted Cell Sorting. 200 ng of DNA extracted from the CD34+ cells was processed using the Infinium Methylation 450k Beadchip (Illumina). Differentially methylated regions (DMRs) were identified using the bump hunting procedure of Jaffe (2011) to pool information across CpG loci into regions of consistent change and to quantify statistical significance.

893 differentially methylated regions (DMRs) varied linearly with age in HSPCs; a set of 31 such regions yielded an accurate predictor of age in lineage-sorted cells (N=48, Reinius et al., 2012) and whole blood (N=656, Hannum et al., 2011), with a root-mean-squared error of 5.3 years. While age-related lymphopenia has previously been reported, DNA methylation marks for lineage commitment (Houseman et al., 2012) were nearly uniform within our subjects’ CD34+ cells, and exhibited no relationship with age. However, regional summaries of methylation provided more accurate age predictions than specific CpG loci. We reasoned that differential variability at individual loci might be the cause.

We thus investigated regions where methylation variability increased with age. Known imprinted clusters and allelically methylated regions (AMRs) identified by Fang (2012, PNAS) were disproportionately represented among these; 27% of known imprinting regions and 33.3% of allelically methylated regions in the genome coincided with at least one such region, while comprising only 0.3% of the genome and 0.7% of loci assayed. Among these, the H19 imprinting control region has been shown to crucially regulate long-term HSPC homeostasis in mice via IGF2, and the allelically methylated WT1/WT1-AS region on chromosome 11p is a highly recurrent hotspot for disordered methylation in AML, as well as sequential epigenetic defects in Wilms’ tumor. The allelically methylated vault RNA VTRNA2-1 (recently shown to predict survival in AML) on chromosome 5q, and the monoallelically expressed TP73 and DIRAS3 genes on chromosome 1p, were also sites of greater methylation variability with age in normal HSPCs. Wu et al. (1997) showed that loss of imprinting at H19/IGF2 is common in AML, seemingly conferring a selective metabolic advantage, and global loss of imprinting in mice leads to widespread tumorigenesis (Holm et al., 2005). Recurrent methylation aberrations in induced pluripotent stem cells favor imprinted clusters (Nazor, 2012), and epigenetic polymorphisms arise in these regions over time in cultured cells (Tanay et al, 2012). However, to our knowledge, ours is the first report of this type of heterogeneity with age in normal human adult HSPCs.

Clonal hematopoiesis has previously been documented in healthy elderly adults (Levine 2012), and the majority of patients in the Cancer Genome Atlas (TCGA) AML study exhibited mutations in one or more genes regulating epigenetic machinery. We propose that increased epigenetic heterogeneity in aging HSPCs, particularly at regions with allele-specific methylation (such as H19/IGF2), may precede malignant evolution in some leukemias, and warrants further investigation.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.