Page 83 - 2019_08-Haematologica-web

ZFP36L1 enhancer DNA methylation in MF
this supervised analysis, we detected 35,215 differentially methylated CpG (FDR≤0.05) corresponding to 10,253 coding genes. Among all of these differentially methylated CpG, 65.3% were hypomethylated (corresponding to 22,998 CpG) and the remaining 34.7% were hypermethy- lated (a total of 12,217 CpG), suggesting that loss of DNA methylation is the predominant alteration in MF. Global DNA hypomethylation has also been a common finding in other hematologic malignancies such as chronic lym- phocytic leukemia, multiple myeloma and acute myeloid leukemia.13,26,27
Analysis of the genomic location of differentially methylated CpG showed that both hyper- and hypo- methylated CpG were underrepresented in classical CGI and significantly enriched outside CpG islands (Figure 1B). This is an interesting finding, because traditionally, neo- plasms acquire hypomethylation outside CGI and hyper- methylation inside the islands,13,27 and suggests that pat- terns of methylation gain in MF might differ from those of other neoplasms. To shed light onto the specific function of the differentially methylated CpG, the chromatin state of each CpG was categorized adapting a publicly available annotation of chromatin immunoprecipitation-sequenc- ing data from CD34+ hematopoietic progenitor cells, in which four distinct states were defined: promoter (with H3K4me3), active enhancer (with H3K4me1 and H3K27ac), transcribed regions (showing H3K36me3) and
16 heterochromatin (including H3K9me3 and H3K27me3).
Both hyper- and hypo-methylated CpG showed signifi- cant enrichment in enhancer regions, together with a striking underrepresentation in promoter regions (Figure 1B). Unsupervised clustering of differentially methylated CpG located exclusively in enhancer regions (Online Supplementary Table S2) displayed a clear segregation of the majority of MF patients from healthy controls (Figure 1C) identifying 4,182 hypermethylated and 10,935 hypomethylated probes. These results suggest that patients with MF show an intrinsic aberrant pattern of DNA methylation preferentially located in enhancer regions of the genome.
To further characterize the aberrant DNA methylation of enhancer regions in MF, GO-PANTHER enrichment analysis was performed separately in differentially methy- lated genes. GO terms with an adjusted FDR<0.05 were selected, showing in the case of hypermethylated enhancers relevant cellular processes such as cellular defense response or induction of apoptosis (Figure 1D).
DNA methylation of enhancer regions is associated with gene expression profile in myelofibrosis
DNA methylation levels of enhancer regions were cor- related with the expression of host and adjacent coding genes using publicly available gene expression data of an independent cohort of MF patients and healthy donors (GSE26049).21 Fold increases in gene expression values were grouped according to the hypermethylated (Δβ>0.4) or hypomethylated (Δβ<-0.4) enhancer status in MF versus controls. This analysis showed that enhancer DNA hyper- methylation was associated with decreased gene expres- sion of host/adjacent coding genes. In contrast, hypomethylated enhancer regions were not associated with increased gene expression (Figure 2A).
Next, we designed a more stringent approach to identi- fy the set of genes underlying the most significant and substantial changes in enhancer DNA methylation (FDR
<0.01, Δβ>0.4), coupled with downregulation of their expression (logFC<0) (Figure 2B). After identifying a num- ber of potential candidates (27 genes), we focused on ZFP36L1, which codes for a RNA-binding protein that mediates the decay of unstable mRNA with AU rich ele- ments in the 3’ untranslated region.28,29 Interestingly, the enhancer region associated with this candidate gene was located in its intragenic region, presumably acting as a cis- regulatory element of ZFP36L1 transcription. It is worth noting that this regulatory element was consistently hypermethylated in the cohort of MF patients and showed the largest number of hypermethylated enhancer-related CpG probes among the final 27-gene list.
ZFP36L1 enhancer hypermethylation correlated with downregulation of expression in MF as compared to con- trols (Figure 2B and Online Supplementary Figure S1B), which was further confirmed in an independent cohort of MF patients and myeloid cell lines (Figure 2C). Bisulfite sequencing confirmed that DNA methylation of the enhancer region of ZFP36L1 was consistently higher in all MF samples and myeloid cell lines than in control samples, whereas the promoter region remained unmethylated (Figure 2D,E and Online Supplementary Figure S1C). Results obtained from luciferase-reporting assays demonstrated that the exogenous DNA methylation significantly reduced ZFP36L1 enhancer activity (Figure 2F). Moreover, 5’-́azacytidinehypomethylatingtreatmentwasableto reverse the DNA methylation levels of the enhancer region in vitro, partially restoring the gene expression levels of ZFP36L1 in the SET-2 cell line (Figure 2G,H).
ZFP36L1 acts as a tumor suppressor gene and potentially affects the myelofibrosis transcription profile
We hypothesized that ZFP36L1 downregulation could lead to upregulation of its putative targets in MF. We used DREME, a motif discovery algorithm specifically designed to find short, core DNA-binding motifs enriched in the 3’ untranslated region of genes. We found that the GTATTTDT motif (E-value=4.5x10-15) was in fact over- represented in transcripts upregulated in MF patients (Figure 3A). Subsequently, an analysis of motif enrich- ment was performed, revealing a significant enrichment of upregulated genes in MF patients among the group containing the mentioned motif (P=7.69x10-20; logFC >1; P<0.05).
To complement DREME analysis, we searched the AREsite30 database for AU-rich elements to determine whether we could detect, among the genes differentially expressed (B-value>10) between MF and controls, an enrichment of these sequences in the upregulated subset. Of all the possible AU motifs, we focused on the most restricted 9, 11 and 13-mer motifs. Interestingly, we were able to identify an enrichment of a 9-mer sequence WTATTTATW (P=0.01) and a 13-mer sequence WWW- TATTTATWWW (P=0.03) exclusively among the upregu- lated genes in MF patients (Figure 3A). Remarkably, both AU motifs strongly resemble the ZFP36L1 core-binding motif predicted by the DREME algorithm.
Re-expression of ZPF36L1 was achieved through lentiviral infection of the SET-2 cell line. Seventy-two hours after infection, the levels of EGFP-positive cells used as the positive control confirmed successful infec- tion, and the level of expression of ZFP36L1 confirmed
haematologica | 2019; 104(8)
1575