Page 131 - 2020_01-Haematologica-web

Intraclonal heterogeneity in CMML
Epigenetic heterogeneity among induced pluripotent stem cell clones from the patient with chronic myelomonocytic leukemia
Epigenetic intraclonal heterogeneity has been reported in hematologic malignancies2,19 as well as in solid tumors,20 and co-dependency between epigenetic and genetic evolution has been questioned, e.g., in large B-cell lymphoma21 and acute myeloid leukemia.3 We performed DNA methylation analysis on CD34+CD43+ cells generat- ed from iPSC to investigate their epigenetic state. Unsupervised analysis by correspondence analysis sepa- rated control- from CMML-derived cells. The latter showed a much greater diversity (Figure 6A). We then used a β binomial model implemented in methylSig22 to perform a supervised analysis of differentially methylated regions (DMR). We identified 5,651 hypermethylated DMR in CMML-derived cells compared to controls. In contrast, only 874 DMR were identified as more methy- lated in control iPSC-derived cells (Figure 6B, C). Genomic annotation revealed that, in CMML samples, DMR were depleted at promoter regions and CpG islands while being enriched at gene body and intergenic enhancers (P<0.001 in all cases) (Figure 6D). DMR detect- ed at CpG islands, gene body enhancers and intergenic enhancers were significantly more often hypermethylat- ed regions (P<0.001) (Online Supplementary Figure S5A). In accordance with changes in DNA methylation affecting enhancers, we observed significant enrichment of DMR within enhancers23 compared to background (36.8% at total DMR vs. 15.2% background, P-value <2.2x10-16). Motif enrichment analysis suggested that the main sequences targeted by DMR were motifs recognized by transcription factors of the ETS family (Online Supplementary Figure S5B). Gene ontology analysis of bio- logical processes of DMR hypermethylated in CMML- derived cells showed enrichment of differentiation and
hematopoietic development categories (Figure 6E), sug- gesting that DNA methylation differences between CMML- and control-derived cells may capture an epige- netic memory related to the biology of the disease, still present after having been reprogrammed. Finally, focus- ing on CMML iPSC-derived hematopoietic cells, unsuper- vised hierarchical clustering based on their DNA methy- lation profiles revealed high concordance with their phy- logenetic background (Figure 6F).
In order to correlate changes in DNA methylation with those seen at the expression level, we first looked at the expression status of genes closest to DMR using a nearest gene annotation approach. Using this approach, only 114 genes showed overlapping changes in expression and DNA methylation. However, since focusing on DMR- nearest gene correlations may not correctly capture the three-dimensional nature of gene regulation, we next explored the role of DMR within specific topologically associated domains (TAD) identified using publicly avail- able coordinates. We thus localized each DMR into a given TAD. For every gene within a TAD, we correlated gene expression and methylation levels across the samples using Pearson correlation. With this method, of 196 DMR identified in 66 TAD, we detected changes in gene expres- sion in 72 genes (Online Supplementary Table S2).
To further explore the epigenetic differences between subclones of the original disease and the potential contri- bution of specific mutations to the epigenetic program- ming, we also compared DNA methylation profiles in KRAS wildtype (A1, A2, A4) and KRAS(G12D) (A3, A5) clones (Figure 7). Globally, clones that had acquired KRAS(G12D) seemed to be relatively hypomethylated compared to the KRAS wildtype clones (Figure 7A, B). Acquisition of KRAS(G12D) correlated with a new repar- tition of DMR, including a significant decrease in methy- lation at intronic regions (P<0.001), CpG shores (P<0.001),
Figure 8. Graphical summary of intratumor heterogeneity layers detected by analysis of five patient-derived induced pluripotent stem cells.
121