AbstractCell identity relies on the cross-talk between genetics and epigenetics and their impact on gene expression. Oxidation of 5-methylcytosine (5mC) into 5-hydroxymethylcytosine (5hmC) is the first step of an active DNA demethylation process occurring mainly at enhancers and gene bodies and, as such, participates in processes governing cell identity in normal and pathological conditions. Although genetic alterations are well documented in multiple myeloma (MM), epigenetic alterations associated with this disease have not yet been thoroughly analyzed. To gain insight into the biology of MM, genome-wide 5hmC profiles were obtained and showed that regions enriched in this modified base overlap with MM enhancers and super enhancers and are close to highly expressed genes. Through the definition of a MM-specific 5hmC signature, we identified FAM72D as a poor prognostic gene located on 1q21, a region amplified in high risk myeloma. We further uncovered that FAM72D functions as part of the FOXM1 transcription factor network controlling cell proliferation and survival and we evidenced an increased sensitivity of cells expressing high levels of FOXM1 and FAM72 to epigenetic drugs targeting histone deacetylases and DNA methyltransferases.
MM is a genetically and clinically heterogeneous hematological cancer associated with a limited number of gene translocations into the immunoglobulin heavy chain locus of plasma cells (PC). In particular, CCND1, CCND3, c-MAF, MAFB and MMSET translocations influence prognosis and are used to classify patients into molecular subgroups.1 Genome sequencing studies have revealed considerable heterogeneity and genomic instability, a complex mutational landscape and a branching pattern of clonal evolution.32 Epigenetic modifications including DNA methylation and histone modifications have been also related to MM pathophysiology.64 Patients with highly proliferative PC can also show genetic instability of the chromosome 1q arm and specially of the pericentromeric region 1q12 and of its immediate neighbor 1q21.87 Amplification of 1q21, and possibly overexpression of genes lying in 1q21, parallel disease progression.8 However, no causal link between proliferation and 1q21 instability has yet been demonstrated, although overexpression of the histone chaperone gene ANP32E, in 1q21.2 and the cyclin-dependent kinase regulator CKS1B, in 1q21.3, is of poor prognosis in MM.9 More recently, ILF2, in 1q21.3, was proposed to be involved in the pathogenic role of 1q21 amplification by increasing DNA damage resistance.10 Nonetheless, other yet unidentified genes might participate in the pathogenicity of 1q21 gain.
Tumor PC clones show different levels of differentiation,11 suggesting that MM could originate either from B cells that do not fulfill a complete differentiation program, or from PC that partially dedifferentiate. Cell differentiation relies on the selective engagement of small genomic regions called enhancers which are bound by transcription factors (TF) controlling cell-specific transcriptional programs. As an early step of activation, enhancers undergo active DNA demethylation through iterative oxidation of 5mC into 5hmC, 5-formylcytosine (5fC) and 5-carboxyl-cytosine (5caC) by Ten-Eleven-Translocation (TET) enzymes and repair by the base excision repair machinery, including the T:G mismatch DNA glycosylase TDG which cleaves 5fC and 5caC.12 5hmC has been mapped genome-wide in several cell differentiation models, including in vitro differentiation of human naive B cells (NBC) into plasmablasts (PB), where 5hmC accumulates at PC identity genes, as well as in mouse germinal center B cells.1513 These studies showed enrichment in 5hmC at poised/active enhancers as well as in the body of highly transcribed genes. Despite the wealth of information on the genetics of MM, the epigenetics of this disease is still poorly described. Nonetheless, a recent genome-wide investigation of active chromatin regions showed that opening of heterochromatin is a hallmark of MM.16 In addition, interrogation of DNA methylation in MM cells revealed that, despite a global hypomethylation, their genome shows specific hypermethylation of enhancers that normally undergo complete demethylation during B-cell commitment and are bound by B-cell TF.17 Interestingly, the methylation levels of these enhancers were anti-correlated with expression levels of B-cell-specific TF in MM patients,17 suggesting that variations in tumor PC differentiation states could indeed be controlled through DNA methylation/demethylation mechanisms guided by specific TF. Here, we investigated the genome-wide distribution of 5hmC in tumor PC and, through the identification of MM-specific hydroxymethylated regions, evidenced new prognosis genes that might contribute to the understanding of this disease.
Primary multiple myeloma cells
Bone marrow samples were collected after patients’ written informed consent in accordance with the Declaration of Helsinki and institutional research board approval from Montpellier University hospital. Patients’ MM cells were purified using anti-CD138 MACS microbeads (Miltenyi Biotech, Bergisch Gladbach, Germany). RNA and genomic DNA were extracted using Qiagen kits (Qiagen, Hilden, Germany) and their gene expression profile (GEP) obtained using Affymetrix U133 plus 2.0 microarrays as described.18 Plasma cell labeling index (PCLI)19 was investigated using BrdU incorporation and flow cytometry in 101 patients at diagnosis. Correlation between gene expression and PCLI was determined with a Spearman’s test. We used publicly available Affymetrix GEP (Gene Expression Omnibus, accession number GSE2658) of a cohort of 345 purified MM cells from previously untreated patients from the University of Arkansas for Medical Sciences (UAMS, Little Rock, AR), termed in the following UAMS-TT2 cohort. These patients were treated with total therapy 2 including high-dose melphalan (HDM) and autologous stem cell transplant (ASCT).20 We also used Affymetrix data from the total therapy 3 cohort (UAMS-TT3; n=158; E-TABM-1138)21 of 188 relapsed MM patients subsequently treated with bortezomib (GSE9782) from the study by Mulligan et al.22
FDI-6 treatment of primary MM cells from patients
Bone marrow of patients presenting with previously untreated MM (n=6) at the University Hospital of Montpellier was obtained after patients’ written informed consent in accordance with the Declaration of Helsinki and agreement of the Montpellier University Hospital Centre for Biological Resources (DC-2008-417). Mononuclear cells were treated with or without graded concentrations of FDI-6 and MM cells cytotoxicity was analyzed using anti-CD138-phycoerythrin monoclonal antibody (Immunotech, Marseille, France) as described previously.5
Genome-wide mapping of 5hmC and bioinformatics
5hmC was mapped by selective chemical labeling,23 coupled or not with exonuclease digestion,2524 of 10 μg of sonicated (Bioruptor, Diagenode) genomic DNA from MM patients or from MCF-7 cells. Libraries were obtained using the TruSeq ChIP Sample Prep Kit (Illumina), quantified using the KAPA library quantification kit (KAPA Biosystems) and 50 bp single end sequenced with HiSeq 1500 (Illumina). Reads were mapped to hg19 and processed as described.25 SCL-exo signal was normalized to the input signal. Principal component analyses (PCA) were run online with Galaxy (http://deeptools.ie-freiburg.mpg.de/) with 5hmC signal bined by 10 kb windows. Heatmap clustering of hydroxymethylated CpG was run online (http://cistrome.org/).26 Search for transcription factor binding site (TFBS) motifs surrounding hydroxymethylated CpG used the online Centdist tool27 in 600 bp windows centered on 5hmCpG. Annotation of 5hmCpG used GREAT.28 Oxidative bisulfite modification of gDNA and hybridization to Illumina 450K arrays were run as previously described.29 ChIP-seq data for H3K27ac in primary MM cells16 were downloaded from the European Nucleotide Archive database (PRJEB25605). Data were deposited in the Gene Expression Omnibus database under accession number GSE124188.
A detailed description of additional methods is available in the Online Supplementary Information.
5hmC-enrichment partly recapitulates the molecular classification of MM
To better understand the relationship between DNA demethylation processes and cell identity in MM, we generated genome-wide maps of 5hmC in MM cells isolated from 11 patients belonging to different molecular subgroups, as well as for three human myeloma cell lines (HMCL), either by SCL-exo2524 or SCL-seq23 (Figure 1A). In parallel, 10 MM samples were processed through the oxidative-bisulfite modification procedure and hybridized to Illumina 450K arrays (oxBS-450K).29 Annotation of 5hmC positive regions (40,586 CpG for oxBS-450K; 86,591 CpG for SCL-exo; 64,424 regions for SCL-seq) aggregated from all patients included in each procedure was run using GREAT.28 Results indicated that oxBS-450K did not generate meaningful information whereas SCL-exo and SCL-seq highlighted characteristics of PC biology such as endoplasmic reticulum stress, immune response, but also pathways which could be linked to MM, such as "IRF4 target genes"30 (Online Supplementary Figure S1A). Based on these annotation data, only SCL-exo and SCL-seq identified regions were further analyzed. Results were first compared to genome-wide 5hmC maps of NBC and PB previously generated by SCL-seq.14 PCA of the signal showed dispersion of the samples, suggesting variability between tumor hydroxymethylomes. Nonetheless, most MM hydroxymethylomes grouped closer to plasmablasts (PB than to NBC) (Figure 1B). When running PCA only with MM patients from the MMSET, CCND1 and Proliferation groups, 3 clusters were observed (C1 to C3 Figure 1B). These clusters gathered together patients from similar molecular groups, although two patients did not follow this rule (E12097 and E6068), indicating that molecular groups are probably heterogeneous in nature and that 5hmC maps can help refine molecular clustering. We next generated the union of significantly hydroxymethylated CpG (P< 5e) overlapping between samples within each cluster. These 6,385 individual CpG were clusterized according to their 5hmC levels (Figure 1C). The resulting heatmap evidenced two groups of CpG that were selectively more hydroxymethylated either in C1 patients (C'1 cluster) or in C2 patients (C'2 cluster). Analysis of motif enrichment for transcription factor binding sites in C'1 and C'2 and their comparison with motifs found in NBC and PB hydroxymethylated regions suggested that MM cells from C2 patients remain more differentiated than those from C1 patients. Indeed, the BLIMP1 (a major regulator of PC differentiation) motif was significantly enriched in C'2 but not in C'1 regions (Online Supplementary Figure S1B). Accordingly, functional annotation through GREAT showed that C'2 regions associated with endoplasmic reticulum stress gene signature or IL-6 signaling, features of mature PC, whereas C'1 regions did not (Figure 1C). Conversely, NOTCH, MYC, Cell Cycle and BCR signaling, which are more characteristic of proliferating and/or undifferentiated B cells, were significantly more associated with C'1 regions than with C'2 regions (Figure 1C). Of note, the SUH and GLI motifs were also selectively enriched in C'1 regions (Online Supplementary Figure S1B). Hence, C1 patients from the proliferation group probably have active NOTCH and SHH pathways, both important for proliferation of CD138 MM cells.3231 In addition, the binding motif for MEIS1, a known regulator of hematopoietic progenitor self-renewal,33 was also selectively enriched in C'1 regions. Collectively, these data indicate that MM cells from the proliferation group have an under-differentiated phenotype and proliferation is likely to rely on a MEIS1/SUH/GLI transcription factor cocktail as well as MYC activity. These results also show that 5hmC is indeed indicative of the biological and clinical traits of patients and could be used to delineate groups with different characteristics.
5mC oxidation targets MM plasma cell enhancers
Investigating the genomic location of high confidence MM 5hmCpG (86,591 CpG, p<1e) showed that they were mostly distributed in introns and distal intergenic regions and, as previously observed in NBC, PB and mouse activated B cells,1514 genes which were close to a 5hmC enriched region showed higher expression than genes which had no such region in their vicinity (Figure 2A). We next considered the presence of these 5hmCpG in regions harboring different chromatin states (ChromHMM) and defined for the lymphoblastoid cell line GM12878 (a cell line used as a model of normal B cells) by the ENCODE consortium. In accordance with a proposed global opening of chromatin in MM cells and as already described for open chromatin regions in MM,16 and for MM hypomethylated CpG,17 MM 5hmCpG mainly reside in regions of heterochromatin in GM12878 cells (Online Supplementary Figure S2A). During differentiation of NBC into PB, genomic sites undergoing 5mC oxidation often bear histone marks of either commissioned (H3K4me1) or active (both, H3K27ac and H3K4me1) enhancers.1514 To investigate the relationship between 5mC oxidation and enhancers in MM, we next analyzed enrichment in H3K4me1 and H3K27ac from NBC, PC and MM cells around 5hmCpG detected in MM patients. Most MM 5hmCpG were found in PC and NBC H3K4me1-premarked genomic sites that become active in MM (Figure 2B). We next investigated the presence of MM 5hmCpG in active enhancers and super-enhancers (SE) identified in the MM1.S HMCL.34 Most notably, 89.3% (275 of 308) of the active MM1.S SEs and 36% (2949 of 8285) of all active enhancers were enriched in 5hmCpG in MM (Online Supplementary Figure S2B), strongly suggesting a role for 5mC oxidation in the control of SE activity. Similar results were obtained when analyzing the overlap between 5hmCpG and U-266 enhancers and SE defined on the H3K27ac ChIP-seq signal from the Blueprint consortium (not shown). However, as exemplified for the KLF13 SE (Figure 2C), the vast majority (98.5%, 271 of 275) of these MM1.S active SE were already marked with 5hmC in PB, indicating that these SE probably maintain a cell-of-origin identity in MM cells. Nonetheless, four SE showed a MM-specific hydroxymethylation profile, including 2 DEPTOR SE, MYC SE and GAS2 SE (Figure 2C). DEPTOR and MYC are known to promote MM cell proliferation and survival354 and, in support of our data, their SE were found to be specifically active in MM cells.16 GAS2 has not been associated with MM yet but it has been shown to be upregulated and to favor survival in chronic myeloid leukemia cells in correlation with hypomethylation of its promoter region.3736 When examined during NBC to PC differentiation (Online Supplementary Figure S2C) as well as in patients of the Arkansas MM cohort (Online Supplementary Figure S2D), GAS2 mRNA levels were detected in bone marrow plasma cells (BMPC) but found to be higher in MM cells. Furthermore, GAS2 overexpression impacted patient survival (Figure 2C). Since GAS2 inhibits calpain proteases which degrade TET enzymes,38 it could thus play a central role in sustaining 5mC oxidation levels in MM cells. Overall, this analysis indicates that 5mC oxidation in MM occurs mainly in intronic enhancers and participates in the establishment of a myeloma-specific gene expression program.
A MM-specific 5hmC signature identifies prognosis genes
We next delineated a MM-specific 5hmC signature, independently of molecular subgroups, through first calling CpG that were hydroxymethylated (P<1e) in at least 33% of the myeloma samples and not falling into genomic regions hydroxymethylated in PB (p<1e), and second by iterative clustering of these CpG according to their hydroxymethylation signal both in MM patients and in PB. Hence, a cluster of 415 CpG uniquely hydroxymethylated in MM was obtained (Online Supplementary Figure S3A). A list of 29 genes, including BMP6 and FAM72D (Online Supplementary Figure S3B), associating with at least 3 of these hydroxymethylated CpG was next established and investigated for prognostic value in the UAMS-TT2 (n=256) and UAMS-TT3 (n=158) Arkansas cohorts (http://genomicscape.com/). Interestingly, around a quarter (7/29) of the 29 genes significantly associated with 5hmCpG were located at 1q21.1 (P=7.5e) whereas none were at 1q21.2 (Figure 3A and Online Supplementary Figure S3C-D). Results showed that among these 29 genes, only six had no prognostic value (P>0.05) in both analyzed cohorts (Figure 3A). Others could be classified as either of good (n=4) or of poor (n=7) prognosis in both cohorts, as exemplified for FAM72D (Figure 3B). Investigating the expression of these genes during the differentiation of NBC, poor prognostic genes were mostly expressed in pre-PB, whereas good prognostic genes were mainly expressed in non-proliferating cells (Online Supplementary Figure S4). Among the poor prognostic genes located on 1q21.1, FAM72D is a gene of unknown function which implication in MM biology has not yet been addressed. Of note, recent duplication events of the ancestral FAM72A gene led to the presence of four highly conserved FAM72 genes in human (FAM72A ,B , C, D)39 which cannot be discriminated at the mRNA level (99% identity). Accordingly, "FAM72" is used thereafter when referring to expression data. FAM72 expression levels correlated with 1q21 copy number, which is associated with poor outcome in MM (Online Supplementary Figure S5A-B). Importantly, FAM72 expression levels correlated with 5hmC enrichment at the FAM72D locus (Online Supplementary Figure S5C) which was not biased by 1q21 amplification (Online Supplementary Figure S5D) and was associated with the presence of the enhancer histone modification H3K4me1 in normal plasma cells (Online Supplementary Figure S5E). Conversely to BMP6 (Online Supplementary Figure S4 and S5F), expression of FAM72 was found to be higher in proliferative B cells, i.e. centroblasts and pre-PB (Online Supplementary Figure S4), as well as in the proliferation group (Online Supplementary Figure S5G). In accordance with a putative role of FAM72 in stimulating MM cell proliferation, plasma cell labeling index (PCLI)19 was significantly correlated with FAM72 expression in a cohort of 101 newly diagnosed MM patients (Figure 3C). PCLI was also significantly correlated to FOXM1 and CKS1B expression, but at a lower extent. No correlation was found between MM cell proliferation and ANP32E or ILF2 expression (Figure 3C). These data suggest that amplification of the 1q21 region, together with its hydroxymethylation, might affect cell proliferation in MM through enhancing FAM72 expression.
FAM72D Is involved in MM cell proliferation
Examination of the top-12 FAM72 co-expressed (P<0.05) genes during NBC differentiation into PC, highlighted FOXM1, a TF known to play a key role in MM cell proliferation40 (Figure 4A). Furthermore, a strong correlation between FOXM1 and FAM72 expression was evidenced in MM samples and derived HMCL (Figure 4A). ChIP-seq data from GM12878 lymphoblastoid cells indicated that FOXM1 binds to the FAM72 promoter (Online Supplementary Figure S6A), strongly suggesting that this TF directly regulates FAM72 expression. In support of this hypothesis, FAM72 hydroxymethylation levels were positively correlated with FOXM1 expression (Online Supplementary Figure S6B). Comparison of FOXM1 and FAM72 coexpressed genes in patients from the proliferation molecular MM subgroup revealed a significant overlap (86% of common genes) (Online Supplementary Figure S6C). Moreover, FAM72 coexpressed genes were significantly associated with M phase cell cycle annotations (Online Supplementary Figure S6D). FAM72 was significantly overexpressed in HMCL and MM cells compared to BMPC and monoclonal gammopathy of undetermined significance (MGUS), underlining the link with MM cell proliferation (Online Supplementary Figure S6E). Furthermore, gene set enrichment analysis (GSEA) of expression data from patients with high FAM72 expression and a poor outcome revealed a significant enrichment of genes related to proliferation, overexpressed in proliferating PB compared to mature BMPC and stem cell genes (Online Supplementary Figure S7). To study the biological function of FAM72 overexpression in MM, the XG21 and XG23 HMCL were selected for their low level of endogenous FAM72 expression (Online Supplementary Figure S8A-C). In fetal calf serum (FCS) free medium, FAM72D overexpression resulted in significant growth advantage and response to IL-6, a key MM growth factor (Figure 4B). Similar results were obtained with overexpression of FAM72 in XG23 (Online Supplementary Figure S8D). These data support the recent characterization of FAM72B as an S/G2-M phase gene whose inactivation reduces cell proliferation in human fibroblasts.41 To test for a FOXM1-dependency of primary cancer cells as well as derived cell lines, FDI-6, a DNA binding inhibitor of FOXM1,42 was used. Primary MM cells were highly sensitive to FOXM1 inhibition compared to bone marrow microenvironment cells (Figure 4C). Overexpression of FAM72D in XG21 cells partially counteracted FDI-6-induced cell death (Online Supplementary Figure S8E), suggesting that FAM72D mediates part of FOXM1 effects on cell growth and survival.
In order to extend our findings to other cancer types, publicly available data were analyzed (http://www.cbioportal.org/,43 https://xenabrowser.net/heatmap/) and indicated (i) that amplification of 1q21 is a highly prevalent event in breast cancer (Online Supplementary Figure S9A) and (ii) a higher expression of FAM72D within mutated TP53 patients (Online Supplementary Figure S9B). In agreement with FAM72D genes being targets of FOXM1, inhibition of FOXM1 in MCF-7 breast cancer cells has been shown to significantly reduce FAM72D, FAM72A and FAM72B expression.44 Hence, we next generated genome-wide maps of 5hmC in MCF-7 cells and observed that, as in MM cells, the upstream region of FAM72D was highly hydroxymethylated (Online Supplementary Figure S9C), suggesting a similar regulation of this gene between MM and breast cancer cells. To investigate a putative function of FAM72D in mitosis, FAM72D was knocked-down by transfection of siRNAs in MCF-7 cells and mitotic anomalies were analyzed. Data revealed that a reduction in FAM72D levels led to a higher occurrence of defects such as misaligned chromosomes, lagging chromatids and micronuclei (Online Supplementary Figure S9D), suggesting that FAM72D helps maintain mitotic fidelity.
High FAM72 expression is associated with resistance to bortezomib and sensitivity to histone deacetylase/decitabine inhibitors (HDACi/DNMTi)
Since FAM72D expression is associated with a poor outcome in MM, we investigated whether high FAM72 expression could be related with drug resistance in MM. Correlating FAM72 gene expression with response to conventional drugs (bortezomib, melphalan, lenalidomide, dexamethasone and panobinostat) in HMCL, we identified that high FAM72 expression levels are associated with resistance to bortezomib (n=12, P=0.049) (Figure 5A). These observations are consistent with the fact that high FAM72 expression is associated with a poor prognosis in a cohort of patients at relapse treated by bortezomib monotherapy (Mulligan cohort, Online Supplementary Figure S10A). Conversely, no significant correlation was identified between FAM72 expression and survival in a cohort of patients at relapse treated by dexamethasone monotherapy (Online Supplementary Figure S10B) and between FAM72 expression and in vitro HMCL response to melphalan, lenalidomide or dexamethasone (not shown). Of particular interest, high FAM72 expression tended to be significantly correlated with a better response to panobinostat (n=10, P=0.061; Online Supplementary Figure S10C), suggesting that HDACi could have a therapeutic interest in FAM72 myeloma patients associated with poor survival. Since a combination of both HDACi and DNMTi has been shown to reprogram HMCL cells,45 we next investigated the ability of the combined DNMTi and HDACi decitabine (5aza-dC) and trichostatin A (TSA) to regulate FAM72 and FOXM1 expression in HMCL. Data indicated that the combined treatment reduced, although to different extent, the expression of both genes in a majority of investigated cell lines (Figure 5B). Finally, primary MM cells were treated with a similar combination (5aza-C and the HDACi SAHA) and their resistance to these drugs was inversely correlated to the expression of FAM72 and FOXM1 (Figure 5C and Online Supplementary Figure S10D). When considering FAM72/FAM72 and FOXM1/FOXM1 expressing cells, only FAM72 samples remained significantly correlated with higher sensitivity to the HDCAi/DNMTi combination (Figure 5C). These results suggest that high-risk patients overexpressing FAM72 could benefit from treatment by a HDACi/DNMTi combination.
Collectively, our data point to a prominent role of DNA demethylation events occurring at 1q21.1, and specially at the FAM72D locus, in MM biology and malignancy. One of the most common genetic features in MM linked to high-risk prognosis is the gain (three copies) or amplification (four and more copies) of part or all of the q arm of chromosome 1.46 Among the partial gains of 1q, 1q21 is particularly detected both in newly diagnosed (30%) and relapse (70%) cases. However, there is still no unifying picture explaining the functional role of the 1q arm amplification and its wide occurrence among high-risk patients. Interestingly, MM-like 1q alterations can be experimentally triggered by inhibiting DNA methylation. Indeed, 5aza-dC treatment of activated B cells leads to a decondensation of the 1q12 pericentromeric chromatin, a process that might enable local rearrangements of the 1q arm.47 Although DNA methylation levels are generally low in MM cells, residual methylation accumulates at specific sites17 and can still be erased by active demethylation through TET enzymes. Such mechanisms could play a role in disease progression provided that TET are targeted to the pericentromeric region of chromosome 1 and its adjacent region 1q21. Consistent with this idea, we show that one fourth of the genes associated with MM-specific hydroxymethylated regions lie within the 1q21.1 region. Among them, FAM72D, a gene encoding for a protein of unknown function, was shown here to enhance proliferation of MM cells. Whereas most mammals have only one copy of the FAM72 gene, due to recent duplication events, four genes encode the human FAM72 proteins A to D.39 The FAM72 genes are known to be overexpressed in cancer and favor tumorigenesis induced by the Epstein-Barr virus-derived latent membrane protein 1.4948 FAM72 genes are highly expressed in TP53 mutated cancer cells, whose growth has been shown to be dependent on FAM72D.5150 Interestingly, cancer cell growth dependency on FAM72D has also been demonstrated for cells with mutated or copy number variants of p300, p19Arf and CDKI2A,51 suggesting that FAM72D is required for the growth of different cancer types. Corroborating the tumorigenic potential of FAM72 proteins, two insertional mutagenesis screens in mouse identified the mouse SRGAP2/FAM72A locus as a driver of chronic myeloid leukemia progression and growth factor-independent leukemogenesis.5352 Hence, TET-mediated demethylation of the FAM72D upstream region is likely coupled to the proliferation potential of MM cells.
We have defined here a MM-specific hydroxymethylome that favors a survival/proliferation program relying, at least in part, on the cooperative actions of enhancers and super-enhancers controlling FAM72D, GAS2, DEPTOR and MYC expression (Figure 6). In accordance with the observation that FAM72 expression is predictive to the sensitivity of MM cells to different treatments, its evaluation in patients could help to tailor therapy.
The authors thank the BioGenouest GEH sequencing plate-form (https://geh.univ-rennes1.fr/) from the UMS Biosit.
- ↵$ FC, AP and AAS are co-first authors.
- FundingGS was supported by a grant from La Ligue Contre le Cancer (Grand Ouest committee). The work was partially funded by the hematology laboratory of Rennes University Hospital. AP was supported by PhD training grants from Région Bretagne and Ligue Nationale Contre le Cancer. JM was supported by grants from French INCA (Institut National du Cancer) Institute (PLBIO15-256), PLBIO2018-PIT-MM, LR-FEDER Hemodiag, Fondation de France (201400047510), ITMO Cancer (MM&TT), SIRIC Montpellier (INCa-DGOS-Inserm 6045) and Labex EpiGenMed.
- Check the online version for the most updated information on this article, online supplements, and information on authorship & disclosures: www.haematologica.org/content/105/3/774
- Received March 15, 2019.
- Accepted June 19, 2019.
- Zhan F, Huang Y, Colla S. The molecular classification of multiple myeloma. Blood. 2006; 108(6):2020-2028. PubMedhttps://doi.org/10.1182/blood-2005-11-013458Google Scholar
- Bolli N, Avet-Loiseau H, Wedge DC. Heterogeneity of genomic evolution and mutational profiles in multiple myeloma. Nat Commun. 2014; 5:2997. PubMedhttps://doi.org/10.1038/ncomms3997Google Scholar
- Lohr JG, Stojanov P, Carter SL. Widespread genetic heterogeneity in multiple myeloma: implications for targeted therapy. Cancer Cell. 2014; 25(1):91-101. PubMedhttps://doi.org/10.1016/j.ccr.2013.12.015Google Scholar
- Delmore JE, Issa GC, Lemieux ME. BET bromodomain inhibition as a therapeutic strategy to target c-Myc. Cell. 2011; 146(6):904-917. PubMedhttps://doi.org/10.1016/j.cell.2011.08.017Google Scholar
- Moreaux J, Reme T, Leonard W. Gene expression-based prediction of myeloma cell sensitivity to histone deacetylase inhibitors. Br J Cancer. 2013; 109(3):676-685. Google Scholar
- Herviou L, Kassambara A, Boireau S. PRC2 targeting is a therapeutic strategy for EZ score defined high risk multiple myeloma patients and overcome resistance to IMiDs. Clin Epigenetics. 2018; 10(1):121. Google Scholar
- Sawyer JR, Tricot G, Mattox S, Jagannath S, Barlogie B. Jumping translocations of chromosome 1q in multiple myeloma: evidence for a mechanism involving decondensation of pericentromeric heterochromatin. Blood. 1998; 91(5):1732-1741. PubMedGoogle Scholar
- Hanamura I, Stewart JP, Huang Y. Frequent gain of chromosome 1q21 in plasma-cell dyscrasias detected by fluorescence in situ hybridization: incidence increases from MGUS to relapsed myeloma and is related to prognosis and disease progression following tandem stem-cell transplantation. Blood. 2006; 108(5):1724-1732. PubMedhttps://doi.org/10.1182/blood-2006-03-009910Google Scholar
- Walker BA, Leone PE, Chiecchio L. A compendium of myeloma-associated chromosomal copy number abnormalities and their prognostic value. Blood. 2010; 116(15):e56-65. PubMedhttps://doi.org/10.1182/blood-2010-04-279596Google Scholar
- Marchesini M, Ogoti Y, Fiorini E. ILF2 is a regulator of RNA splicing and DNA damage response in 1q21 amplified multiple myeloma. Cancer Cell. 2017; 32(1):88-100. https://doi.org/10.1016/j.ccell.2017.05.011Google Scholar
- Paiva B, Puig N, Cedena MT. Differentiation stage of myeloma plasma cells: biological and clinical significance. Leukemia. 2017; 31(2):382-392. https://doi.org/10.1038/leu.2016.211Google Scholar
- Wu X, Zhang Y. TET-mediated active DNA demethylation: mechanism, function and beyond. Nat Rev Genet. 2017; 18(9):517-534. PubMedhttps://doi.org/10.1038/nrg.2017.33Google Scholar
- Sérandour AA, Avner S, Oger F. Dynamic hydroxymethylation of deoxyribonucleic acid marks differentiation-associated enhancers. Nucleic Acids Res. 2012; 40(17):8255-8265. PubMedhttps://doi.org/10.1093/nar/gks595Google Scholar
- Caron G, Hussein M, Kullis M. Cell-cycle-dependent reconfiguration of the DNA methylome during terminal differentiation of human B cells into plasma cells. Cell Rep. 2015; 13(5):1059-1071. Google Scholar
- Lio C-W, Shukla V, Samaniego-Castruita D. TET enzymes augment activation-induced deaminase (AID) expression via 5-hydroxymethylcytosine modifications at the Aicda superenhancer. Sci Immunol. 2019; 4(34)Google Scholar
- Jin Y, Chen K, De Paepe A. Active enhancer and chromatin accessibility landscapes chart the regulatory network of primary multiple myeloma. Blood. 2018; 131(19):2138-2150. PubMedhttps://doi.org/10.1182/blood-2017-09-808063Google Scholar
- Agirre X, Castellano G, Pascual M. Whole-epigenome analysis in multiple myeloma reveals DNA hypermethylation of B cell-specific enhancers. Genome Res. 2015; 25(4):478-487. PubMedhttps://doi.org/10.1101/gr.180240.114Google Scholar
- Viziteu E, Klein B, Basbous J. RECQ1 helicase is involved in replication stress survival and drug resistance in multiple myeloma. Leukemia. 2017; 31(10):2104-2113. PubMedhttps://doi.org/10.1038/leu.2017.54Google Scholar
- Requirand G, Robert N, Boireau S. BrdU incorporation in multiparameter flow cytometry: A new cell cycle assessment approach in multiple myeloma. Cytometry B Clin Cytom. 2019; 96(3):209-214. Google Scholar
- Barlogie B, Tricot G, Rasmussen E. Total therapy 2 without thalidomide in comparison with total therapy 1: role of intensified induction and posttransplantation consolidation therapies. Blood. 2006; 107(7):2633-2638. PubMedhttps://doi.org/10.1182/blood-2005-10-4084Google Scholar
- Pineda-Roman M, Zangari M, Haessler J. Sustained complete remissions in multiple myeloma linked to bortezomib in total therapy 3: comparison with total therapy 2. Br J Haematol. 2008; 140(6):625-634. PubMedhttps://doi.org/10.1111/j.1365-2141.2007.06921.xGoogle Scholar
- Mulligan G, Mitsiades C, Bryant B. Gene expression profiling and correlation with outcome in clinical trials of the proteasome inhibitor bortezomib. Blood. 2007; 109(8):3177-3188. PubMedhttps://doi.org/10.1182/blood-2006-09-044974Google Scholar
- Song CX, Szulwach KE, Fu Y. Selective chemical labeling reveals the genome-wide distribution of 5-hydroxymethylcytosine. Nat Biotechnol. 2011; 29(1):68-72. PubMedhttps://doi.org/10.1038/nbt.1732Google Scholar
- Sérandour AA, Avner S, Mahé EA. Single-CpG resolution mapping of 5-hydroxymethylcytosine by chemical labeling and exonuclease digestion identifies evolutionarily unconserved CpGs as TET targets. Genome Biol. 2016; 17:56. Google Scholar
- Sérandour AA, Avner S, Salbert G. Coupling exonuclease digestion with selective chemical labeling for base-resolution mapping of 5-hydroxymethylcytosine in genomic DNA. Bio Protoc. 2018; 8(5):2747. Google Scholar
- Liu T, Ortiz JA, Taing L. Cistrome: an integrative platform for transcriptional regulation studies. Genome Biol. 2011; 12(8):R83. PubMedhttps://doi.org/10.1186/gb-2011-12-8-r83Google Scholar
- Zhang Z, Chang CW, Goh WL, Sung WK, Cheung E. Centdist: discovery of co-associated factors by motif distribution. Nucleic Acids Res. 2011; 39(Web Server issue):W391-399. PubMedhttps://doi.org/10.1093/nar/gkr387Google Scholar
- McLean CY, Bristor D, Hiller M. GREAT improves functional interpretation of cis-regulatory regions. Nat Biotechnol. 2010; 28(5):495-501. PubMedhttps://doi.org/10.1038/nbt.1630Google Scholar
- Stewart SK, Morris TJ, Guilhamon P. oxBS-450K: a method for analysing hydrox-ymethylation using 450K beadchips. Methods. 2015; 72:9-15. PubMedhttps://doi.org/10.1016/j.ymeth.2014.08.009Google Scholar
- Shaffer AL, Emre NC, Lamy L. 2008. IRF4 addiction in multiple myeloma. Nature. 2008; 454(7201):226-231. PubMedhttps://doi.org/10.1038/nature07064Google Scholar
- Liu Z, Xu J, He J. A critical role of autocrine sonic hedgehog signaling in human CD138+ myeloma cell survival and drug resistance. Blood. 2014; 124(13):2061-2071. PubMedhttps://doi.org/10.1182/blood-2014-03-557298Google Scholar
- Jundt F, Probsting KS, Anagnostopoulos I. Jagged1-induced Notch signaling drives proliferation of multiple myeloma cells. Blood. 2004; 103(9):3511-3515. PubMedhttps://doi.org/10.1182/blood-2003-07-2254Google Scholar
- Cai L, Langer EM, Gill JG. Dual actions of Meis1 inhibit erythroid progenitor development and sustain general hematopoietic cell proliferation. Blood. 2012; 120(2):335-346. PubMedhttps://doi.org/10.1182/blood-2012-01-403139Google Scholar
- Lovén J, Hoke HA, Lin CY. Selective inhibition of tumor oncogenes by disruption of super-enhancers. Cell. 2013; 153(2):320-334. PubMedhttps://doi.org/10.1016/j.cell.2013.03.036Google Scholar
- Peterson TR, Laplante M, Thoreen CC. DEPTOR is an mTOR inhibitor frequently overexpressed in multiple myeloma cells and required for their survival. Cell. 2009; 137(5):873-886. PubMedhttps://doi.org/10.1016/j.cell.2009.03.046Google Scholar
- Zhou H, Ge Y, Sun L. Growth arrest specific 2 is up-regulated in chronic myeloid leukemia cells and required for their growth. PLoS One. 2014; 9(1):e86195. PubMedhttps://doi.org/10.1371/journal.pone.0086195Google Scholar
- Maupetit-Methouas S, Court F, Bourgne C. DNA methylation profiling reveals a pathological signature that contributes to transcriptional defects of CD34+ CD15-cells in early chronic-phase chronic myeloid leukemia. Mol Oncol. 2018; 12(6):814-829. https://doi.org/10.1002/1878-0261.12191Google Scholar
- Wang Y, Zhang Y. Regulation of TET protein stability by calpains. Cell Rep. 2014; 6(2):278-284. PubMedhttps://doi.org/10.1016/j.celrep.2013.12.031Google Scholar
- Kutzner A, Pramanik S, Kim PS, Heese K. All-or-(N)One - an epistemological characterization of the human tumorigenic neuronal paralogous FAM72 gene loci. Genomics. 2015; 106(5):278-285. Google Scholar
- Gu C, Yang Y, Sompallae R. FOXM1 is a therapeutic target for high-risk multiple myeloma. Leukemia. 2016; 30(4):873-882. Google Scholar
- Giotti B, Chen S-H, Barnett MW. Assembly of a parts list of the human mitotic cell cycle machinery. J Mol Cell Biol. 2018. Google Scholar
- Gormally MV, Dexheimer TS, Marsico G. Suppression of the FOXM1 transcriptional programme via novel small molecule inhibition. Nat Commun. 2014; 5:5165. PubMedhttps://doi.org/10.1038/ncomms6165Google Scholar
- Gao J, Aksoy BA, Dogrusoz U. Integrative analysis of complex cancer genomics and clinical profiles using cBioPortal. Sci Signal. 2013; 6(269):pl1. PubMedhttps://doi.org/10.1126/scisignal.2004088Google Scholar
- Sanders DA, Ross-Ines CS, Beraldi D, Carroll JS, Balasubramanian S. Genome-wide mapping of FOXM1 binding reveals co-binding with estrogen receptor alpha in breast cancer cells. Genome Biol. 2013; 14(1):R6. PubMedhttps://doi.org/10.1186/gb-2013-14-1-r6Google Scholar
- Bruyer A, Maes K, Herviou L. DNMTi/HDACi combined epigenetic targeted treatment induces reprogramming of myeloma cells in the direction of normal plasma cells. Br J Cancer. 2018; 118(8):1062-1073. Google Scholar
- Pawlyn C, Morgan GJ. Evolutionary biology of high-risk multiple myeloma. Nat Rev Cancer. 2017; 17(9):543-556. Google Scholar
- Sawyer JR, Tian E, Heuck CJ. Evidence of an epigenetic origin for high-risk 1q21 copy number alterations in multiple myeloma. Blood. 2015; 125(24):3756-3759. PubMedhttps://doi.org/10.1182/blood-2015-03-632075Google Scholar
- Guo C, Zhang X, Fink SP. Ugene, a newly identified protein that is commonly overexpressed in cancer and binds uracil DNA glycosylase. Cancer Res. 2008; 68(15):6118-6126. PubMedhttps://doi.org/10.1158/0008-5472.CAN-08-1259Google Scholar
- Wang LT, Lin CS, Chai CY, Liu KY, Chen JY, Hsu SH. Functional interaction of Ugene and EBV infection mediates tumorigenic effects. Oncogene. 2011; 30(26):2921-2932. PubMedhttps://doi.org/10.1038/onc.2011.16Google Scholar
- Wang X, Sun Q. TP53 mutations, expression and interaction networks in human cancers. Oncotarget. 2017; 8(1):624-643. PubMedGoogle Scholar
- Tshemiak A, Vazquez F, Montgomery PG. Defining a cancer dependency map. Cell. 2017; 170(3):564-576. PubMedhttps://doi.org/10.1016/j.cell.2017.06.010Google Scholar
- Giotopoulos G, van der Weyden L, Osaki H. A novel mouse model identifies cooperating mutations and therapeutic targets critical for chronic myeloid leukemia progression. J Exp Med. 2015; 212(10):1551-1569. PubMedhttps://doi.org/10.1084/jem.20141661Google Scholar
- Guo Y, Updegraff BL, Park S. Comprehensive Ex vivo transposon mutagenesis identifies genes that promote growth factor independence and leukemogenesis. Cancer Res. 2016; 76(4):773-786. PubMedhttps://doi.org/10.1158/0008-5472.CAN-15-1697Google Scholar