Unique epigenomic signatures identify biologically significant subtypes of MDS and predict response to azacitidine

Unique epigenomic signatures identify biologically significant subtypes of MDS and predict response to azacitidine | bioRxiv /* */ /* */ <!-- <!-- /*! * yepnope1.5.4 * (c) WTFPL, GPLv2 */ (function(a,b,c){function d(a){return"[object Function]"==o.call(a)}function e(a){return"string"==typeof a}function f(){}function g(a){return!a||"loaded"==a||"complete"==a||"uninitialized"==a}function h(){var a=p.shift();q=1,a?a.t?m(function(){("c"==a.t?B.injectCss:B.injectJs)(a.s,0,a.a,a.x,a.e,1)},0):(a(),h()):q=0}function i(a,c,d,e,f,i,j){function k(b){if(!o&&g(l.readyState)&&(u.r=o=1,!q&&h(),l.onload=l.onreadystatechange=null,b)){"img"!=a&&m(function(){t.removeChild(l)},50);for(var d in y[c])y[c].hasOwnProperty(d)&&y[c][d].onload()}}var j=j||B.errorTimeout,l=b.createElement(a),o=0,r=0,u={t:d,s:c,e:f,a:i,x:j};1===y[c]&&(r=1,y[c]=[]),"object"==a?l.data=c:(l.src=c,l.type=a),l.width=l.height="0",l.onerror=l.onload=l.onreadystatechange=function(){k.call(this,r)},p.splice(e,0,u),"img"!=a&&(r||2===y[c]?(t.insertBefore(l,s?null:n),m(k,j)):y[c].push(l))}function j(a,b,c,d,f){return q=0,b=b||"j",e(a)?i("c"==b?v:u,a,b,this.i++,c,d,f):(p.splice(this.i++,0,a),1==p.length&&h()),this}function k(){var a=B;return a.loader={load:j,i:0},a}var l=b.documentElement,m=a.setTimeout,n=b.getElementsByTagName("script")[0],o={}.toString,p=[],q=0,r="MozAppearance"in l.style,s=r&&!!b.createRange().compareNode,t=s?l:n.parentNode,l=a.opera&&"[object Opera]"==o.call(a.opera),l=!!b.attachEvent&&!l,u=r?"object":l?"script":"img",v=l?"script":u,w=Array.isArray||function(a){return"[object Array]"==o.call(a)},x=[],y={},z={timeout:function(a,b){return b.length&&(a.timeout=b[0]),a}},A,B;B=function(a){function b(a){var a=a.split("!"),b=x.length,c=a.pop(),d=a.length,c={url:c,origUrl:c,prefixes:a},e,f,g;for(f=0;f<d;f++)g=a[f].split("="),(e=z[g.shift()])&&(c=e(c,g));for(f=0;f<b;f++)c=x[f](c);return c}function g(a,e,f,g,h){var i=b(a),j=i.autoCallback;i.url.split(".").pop().split("?").shift(),i.bypass||(e&&(e=d(e)?e:e[a]||e[g]||e[a.split("/").pop().split("?")[0]]),i.instead?i.instead(a,e,f,g,h):(y[i.url]?i.noexec=!0:y[i.url]=1,f.load(i.url,i.forceCSS||!i.forceJS&&"css"==i.url.split(".").pop().split("?").shift()?"c":c,i.noexec,i.attrs,i.timeout),(d(e)||d(j))&&f.load(function(){k(),e&&e(i.origUrl,h,g),j&&j(i.origUrl,h,g),y[i.url]=2})))}function h(a,b){function c(a,c){if(a){if(e(a))c||(j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}),g(a,j,b,0,h);else if(Object(a)===a)for(n in m=function(){var b=0,c;for(c in a)a.hasOwnProperty(c)&&b++;return b}(),a)a.hasOwnProperty(n)&&(!c&&!--m&&(d(j)?j=function(){var a=[].slice.call(arguments);k.apply(this,a),l()}:j[n]=function(a){return function(){var b=[].slice.call(arguments);a&&a.apply(this,b),l()}}(k[n])),g(a[n],j,b,n,h))}else!c&&l()}var h=!!a.test,i=a.load||a.both,j=a.callback||f,k=j,l=a.complete||f,m,n;c(h?a.yep:a.nope,!!i),i&&c(i)}var i,j,l=this.yepnope.loader;if(e(a))g(a,0,l,0);else if(w(a))for(i=0;i (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0];var j=d.createElement(s);var dl=l!='dataLayer'?'&l='+l:'';j.src='//www.googletagmanager.com/gtm.js?id='+i+dl;j.type='text/javascript';j.async=true;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-M677548'); Skip to main content Home About Submit ALERTS / RSS Search for this keyword Advanced Search New Results Unique epigenomic signatures identify biologically significant subtypes of MDS and predict response to azacitidine Qin Yang , Miguel Torres-Martin , Masataka Taguchi , Alice Brogi , Irene Casalin , Eleonora Ceneri , Stephanie Halene , Amy E. DeZern , Elizabeth A. Griffiths , Matilde Y. Follo , Carlo Finelli , Jerald P. Radich , Michael J. Rauh , Rafael Bejar , Mikkael A. Sekeres , Valeria Santini , Maria E. Figueroa doi: https://doi.org/10.1101/2025.04.23.650336 Qin Yang 1 Dept. of Biochemistry & Molecular Biology, University of Miami Miller School of Medicine , Miami, FL, USA 2 Sylvester Comprehensive Cancer Center, University of Miami Miller School of Medicine , Miami, FL, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Miguel Torres-Martin 2 Sylvester Comprehensive Cancer Center, University of Miami Miller School of Medicine , Miami, FL, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Masataka Taguchi 1 Dept. of Biochemistry & Molecular Biology, University of Miami Miller School of Medicine , Miami, FL, USA 2 Sylvester Comprehensive Cancer Center, University of Miami Miller School of Medicine , Miami, FL, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Alice Brogi 3 MDS Unit, Hematology, DMSC University of Florence , AOU Careggi, Florence, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Irene Casalin 4 Dept. of Biomedical and Neuromotor Sciences, University of Bologna , Bologna, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Eleonora Ceneri 4 Dept. of Biomedical and Neuromotor Sciences, University of Bologna , Bologna, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Stephanie Halene 5 Dept. of Internal Medicine, Yale University School of Medicine , New Haven, CT, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Amy E. DeZern 6 Sidney Kimmel Comprehensive Cancer Center, Johns Hopkins Medicine , Baltimore, MD, United States Find this author on Google Scholar Find this author on PubMed Search for this author on this site Elizabeth A. Griffiths 7 Roswell Park Comprehensive Cancer Center , Buffalo, NY, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Matilde Y. Follo 4 Dept. of Biomedical and Neuromotor Sciences, University of Bologna , Bologna, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Carlo Finelli 8 Dept. of Experimental, Diagnostic and Specialty Medicine, Institute of Hematology L. and A. Seràgnoli, University of Bologna , S. Orsola-Malpighi Hospital, Bologna, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Jerald P. Radich 9 Clinical Research Division, Fred Hutchinson Cancer Center , Seattle, WA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Michael J. Rauh 10 Department of Pathology and Molecular Medicine, Queen’s University , Kingston, Canada Find this author on Google Scholar Find this author on PubMed Search for this author on this site Rafael Bejar 11 Division of Hematology and Oncology, University of California San Diego Moores Cancer Center , La Jolla, CA, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Mikkael A. Sekeres 2 Sylvester Comprehensive Cancer Center, University of Miami Miller School of Medicine , Miami, FL, USA 12 Div. of Hematology, University of Miami Miller School of Medicine , Miami, FL, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site Valeria Santini 3 MDS Unit, Hematology, DMSC University of Florence , AOU Careggi, Florence, Italy Find this author on Google Scholar Find this author on PubMed Search for this author on this site Maria E. Figueroa 1 Dept. of Biochemistry & Molecular Biology, University of Miami Miller School of Medicine , Miami, FL, USA 2 Sylvester Comprehensive Cancer Center, University of Miami Miller School of Medicine , Miami, FL, USA 12 Div. of Hematology, University of Miami Miller School of Medicine , Miami, FL, USA 13 Dept. of Human Genetics, University of Miami Miller School of Medicine , Miami, FL, USA Find this author on Google Scholar Find this author on PubMed Search for this author on this site For correspondence: mefigueroa{at}med.miami.edu Abstract Full Text Info/History Metrics Supplementary material Data/Code Preview PDF Abstract Myelodysplastic syndromes (MDS) are characterized by aberrant DNA methylation, and mutations in epigenetic modifiers are frequently found in these patients. Although DNA methyltransferase inhibitors (DNMTi) are used to treat MDS, response variability remains a challenge in the clinic, with limited predictive markers. Through comprehensive genomic, epigenomic, and transcriptomic analyses, we have gained valuable insights into the intricate interplay between genetic and epigenetic alterations in MDS. We describe aberrantly hyper and hypomethylated regions in MDS, extending beyond promoter regions and affecting long-distance regulatory elements. Using these aberrant DNA methylation patterns, we classified MDS patients into epigenetic subtypes correlated with known molecular drivers. This epigenetic classification includes a novel group of patients characterized only by their shared DNA methylation profile and lacking any genetic drivers. Furthermore, we identified a robust DNA methylation signature capable of distinguishing DNMTi responders from non-responders prior to receiving treatment. Leveraging these DMRs, we developed robust classifiers capable of predictive response to DNMTi by integrating DNA methylation, gene expression, mutations, and laboratory parameters. Our findings highlight the potential of epigenetic-based classifiers for personalized treatment approaches for MDS patients. Key Points DNA methylation patterns define biologically meaningful MDS subtypes and uncover a new group lacking known mutations. A methylation-based signature at diagnosis predicts azacitidine response, supporting its use in guiding personalized MDS therapy. INTRODUCTION Myelodysplastic syndromes (MDS) encompass a heterogeneous group of disorders characterized by ineffective hematopoiesis, dysplastic changes in the bone marrow (BM), and the potential to progress to acute myeloid leukemia (AML) 2 – 4 . Aberrant DNA methylation (DNAme) is a hallmark of MDS; however, its role remains incompletely understood. Early reports focused on aberrant hypermethylation of tumor suppressor gene promoters, e.g., p15, E-cadherin 5 – 7 , while the advent of DNAme microarrays revealed widespread aberrant hypermethylation at diagnosis 8 , which directly correlated with disease aggressiveness and progression 9 . Later molecular studies reported the existence of recurrent mutations in genes involved in epigenetic regulation, such as TET2, ASXL1, DNMT3A , and EZH2 in MDS 10 – 12 . However, these groundbreaking early studies used smaller cohorts from single centers, no CD34+ selection, and analysis restricted to promoter methylation and gene expression microarrays. Unlike de novo AML, MDS is more likely to be resistant to conventional chemotherapy 13 , 14 . For higher-risk MDS patients, while hematopoietic cell transplantation (HCT) offers the only curative option, it is usually available to only a small fraction of this older patient population. The remaining higher-risk patients rely on treatment with DNMTi – azacitidine (AZA) or decitabine (DAC)– the only available treatment option that has demonstrated a survival advantage compared to conventional care 15 – 19 . However, the overall response rate to these agents is between 30-40%, responses may take longer than six months 18 , and the duration lasts a median of 17 months. Consequently, for the majority of patients in real-world settings, these agents are ineffective, expose them to side effects, and result in significant time and economic burden 15 . Once DNMTi have failed, survival in the absence of an allogeneic HCT is generally less than six months 20 . It is thus critical to identify MDS patients who are unlikely to benefit from DNMTi to accelerate access to HCT, offer up-front investigational studies, and, for frailer patients, offer the option of supportive care alone. Despite some studies suggesting associations between the methylation reversal at select loci and clinical response to DNMTis 21 – 24 , overall, prior epigenetic studies have failed to identify robust associations between specific baseline DNAme profiles and DNMTi response 21 , 23 , 25 , 26 . The absence of response classifiers impedes the customization of treatment plans for individual MDS patients. To address this critical gap, we conducted a multicenter study of a cohort of AZA-treated MDS patients. Through genomic, epigenomic, and transcriptomic analyses on CD34 + BM cells, we identified widespread aberrant DNAme, targeting distal regulatory elements at key biological pathways. These aberrant epigenomic patterns can be used to classify MDS into distinct epigenetic subtypes linked to specific genetic, molecular, and clinical characteristics. Furthermore, we harnessed these epigenetic profiles to develop robust, epigenetic-based classifiers that correlate with accurate pre-treatment prediction of therapeutic response to AZA. These classifiers offer the potential to improve personalized treatment strategies and enhance outcomes for patients with MDS. METHODS MDS samples Clinical BM specimens were collected from 98 de-identified MDS specimens classified as Intermediate or higher-risk MDS by the International Prognostic Scoring System (IPSS) 27 . Mononuclear cells (MNC) were isolated through Ficoll density centrifugation at diagnosis and frozen for later use. Institutional Review Board (IRB) approval was obtained at the University of Miami Miller School of Medicine, Yale University School of Medicine, Roswell Park Cancer Institute, University of Florence, University of Bologna, and the Southwestern Oncology Group (SWOG) central IRB and local IRBs of participating institutions. Written informed consent was obtained from all patients at the time of collection in accordance with the Declaration of Helsinki. CD34⁺ cells were isolated using MACS magnetic beads. DNA and RNA were extracted with the QIAGEN AllPrep kit. Further details are provided in the Supplementary Methods. Statistical analysis All statistical analyses were performed using R statistical software 28 (version 4.1, R-Core Team, http://www.R-project.org/ ). For sequencing-based assays, significance was defined as q ≤ 0.05; for clinical comparisons, p ≤ 0.05 was used. Significance levels are denoted in figures as follows: *p ≤ 0.05, **p ≤ 0.01, ***p ≤ 0.001. Additional statistical details are provided in the Supplementary Methods. Software availability Code used for the analyses in this study is available on Github: https://github.com/santur90/MDS_AZA RESULTS MDS displays extensive aberrant DNA hypomethylation affecting distal regulatory elements To assess epigenetic deregulation in MDS, we analyzed pre-treatment CD34+ BM specimens from a multicenter cohort of 98 IPSS intermediate- or higher-risk MDS patients. 27 The cohort consisted of cases from the North American Intergroup MDS trial S1117 (n=37 cases), augmented with cases from US (n=24) and Italian (n=37) academic centers. All patients received AZA-based regimens and had documented responses. Patients were studied with targeted gene mutation profiling of 112 genes frequently mutated in MDS (n=92 patients), DNAme by enhanced reduced representation bisulfite sequencing (ERRBS; n=90 cases), and gene expression by RNA-seq (n=77 cases) ( Table 1 , Supplemental Table 1) . We first compared gene expression profiles between CD34+ hematopoietic stem and progenitor cells (HSPC) from MDS patients and age- matched healthy individuals (n=14). Unsupervised analysis using principal components (PC) revealed modest discrimination of the two groups ( Figure 1A , left ). This was confirmed by orthogonal approaches using hierarchical clustering and correspondence analysis (Supplemental Figure 1A) . Furthermore, a direct comparison identified only 53 differentially expressed genes (DEGs; absolute fold change (FC) ≥ 1.5 and a false discovery rate (FDR) ≤ 0.05; Supplemental Table 2 ) between MDS and controls (Supplemental Figure 1B) . Gene set enrichment analysis (GSEA) identified upregulation of pathways related to interferon responses, IL6 signaling, oxidative phosphorylation, reactive oxygen species (ROS), cell cycle, and mitophagy in MDS (Supplemental Figure 1C, Supplemental Table 3) . Download figure Open in new tab Figure 1. Comprehensive analysis of DNA methylation alterations in MDS A: Principal Component Analysis of RNA-seq data ( left) from healthy controls (green) and MDS patients (yellow), along with Enhanced Reduced Representation Bisulfite Sequencing (right) data from healthy controls (blue) and MDS patients (red). B: Volcano plot depicting methylation differences versus the -log 10 ( P -value) between healthy controls and MDS patients. Hypomethylated and hypermethylated regions in MDS are shown in blue and red, respectively (absolute methylation difference ≥ 25%, FDR ≤ 0.05). C: Heatmap with hierarchical clustering illustrating methylation levels of differentially methylated regions (without missing values) in healthy controls (blue bar) and MDS patients (red bar), annotated with WHO 2022 subtypes. D: Stacked bar plot showing the distribution of differentially methylated regions across genomic regions, CpG islands, and enhancers. P values were calculated using a binomial test. ***, P ≤ 0.001. E: Bubble plot depicting enriched pathways (FDR ≤ 0.1) for MDS-associated differentially methylated regions, using all CpG tiles as the background. Bubble size corresponds to the number of genes in each pathway, color indicates statistical significance. F: UCSC Genome Browser tracks of two representative loci illustrating hypermethylated ( top ) and hypomethylated ( bottom ) regions in MDS. Bar height represents the methylation level of each region. View this table: View inline View popup Table 1: Patient Characteristics Next, we investigated DNAme patterns in the same cohort of MDS patients and samples from 21 healthy individuals 29 using ERRBS. PC analysis based on DNAme profiles demonstrated a clear separation between healthy individuals and MDS patients ( Figure 1A , right) . Notably, unlike prior studies focusing solely on promoter regions that captured only aberrant DNA hypermethylation in MDS 8 , 9 , supervised analysis comparing the two groups identified almost twice as many hypomethylated regions (n=15,398 differentially methylated regions (DMRs); absolute methylation difference ≤ 25% and FDR ≤ 0.05) in MDS than hypermethylated ones (n= 8,805) ( Figure 1B-C ; Supplemental Table 4) . Since ERRBS targets not only promoters but also intronic, exonic, and distal intergenic regions, we were able to explore the impact of aberrant DNAme within specific genomic features. Overall, DMRs were enriched at intergenic regions (DMRs 41% vs. background 34%, binomial test p-value ≤0.001) while being less likely to target promoter regions (DMRs 14% vs. 24%, p-value ≤0.001). Hypermethylated DMRs were depleted from intronic regions (hyper-DMRs 26% vs. 33%, p-value ≤0.001), whereas hypomethylated DMRs displayed the opposite trend (hypo-DMRs 40% vs. 33%, p-value ≤0.001; Figure 1D , top ). Furthermore, we observed a significant enrichment of hypermethylated DMRs at CpG islands (hyper-DMRs 46% vs. 28%, p-value ≤0.001) while hypomethylated DMRs were enriched at CpG island shores (hypo-DMRs 24% vs. 15%, p-value ≤0.001; Figure 1D , middle ). Notably, both hypermethylated and hypomethylated DMRs were significantly enrichment at both gene body enhancers (DMRs 30% vs. 20%, p-value ≤ 0.001) and intergenic enhancers (DMRs 22% vs. 15%, p-value ≤0.001; Figure 1D , bottom ). Pathway enrichment analysis annotating these DMRs to neighboring transcripts showed enrichment in various leukemia pathways and signaling pathways known to play a role in hematopoiesis and malignant transformation, such as the Hippo 28 , 30 , WNT 31 , RAP1 32 , 33 , mTOR 34 , 35 , and MAPK 36 , 37 signaling pathways. ( Figure 1E-F ; Supplemental Table 3) . Taken together, these findings underscore the magnitude of aberrant gains and losses in DNAme in MDS beyond promoter regions, identifying crucial DNAme changes targeting distal regulatory elements. Furthermore, they provide insight into the potential functional implications of DNAme alterations in MDS, suggesting their involvement in critical signaling cascades and molecular processes known to regulate HSPC biology. DNAme profiles capture the underlying molecular and clinical heterogeneity of the disease We hypothesized that DNAme profiles could serve to stratify MDS patients into distinct epigenetic subtypes reflective of the disease’s molecular and biological heterogeneity. Using hierarchical clustering, we classified MDS patients in our cohort into seven clusters and analyzed each cluster with respect to associated clinical features (age, sex), laboratory parameters (complete blood counts, blast percentages), molecular and cytogenetic information, WHO 2022 classification and AZA response. Specific molecular or cytogenetic features were considered cluster-defining when they showed both statistically significant enrichment and were associated with more than 40% of patients in a given cluster. Table 2 describes the clinical, cytogenetic, and classification details for each of the 7 clusters. View this table: View inline View popup Table 2: Summary of Patient Characteristics within each of the 7 DNA methylation Clusters As expected, mutations in splicing factor proteins and epigenetic modifiers were very frequently seen in the cohort, and most clusters included patients carrying some of these mutations. However, similar to what we had previously observed in the context of AML 38 , the DNAme cluster identity correlated strongly with specific clinical and molecular features ( Figure 2A-B ) . Cluster II (n=13) was enriched for patients carrying ASXL1 (Fisher’s test p-value≤0.01), STAG2 (p-value≤0.05), and RUNX1 (p-value≤0.05) mutations. Cluster III (n=6) was defined by the presence of U2AF1 (p-value≤0.01) and BCOR (p-value≤0.01) mutations. Cluster IV (n=6) was enriched for the presence of SF3B1 mutations (p- value≤0.001) co-occurring with RUNX1 mutations in female patients (p-value≤0.05). In contrast, 5/6 patients with SF3B1 outside of Cluster IV were male and failed to cluster together. Notably, 10/19 patients in Cluster V carried mutations in TP53 (p-value≤0.001). In addition, patients in this cluster were almost devoid of splicing factor mutations, with the few positive cases carrying almost exclusively U2AF1 mutations. Moreover, Cluster V was strongly enriched in male patients, with 16/19 male cases (p-value≤0.05). Notably, the remaining 8 cases with TP53 mutations were scattered among the other clusters and were mainly female (5/8). Cluster VII was enriched in mutations in the DNAme machinery (8/14), most of them (6/8) affecting DNMT3A (p-value≤0.05) ( Supplemental Table 5 ). Download figure Open in new tab Figure 2. DNA methylation captures molecular heterogeneity in MDS A: Heatmap with hierarchical clustering illustrating the seven MDS patient clusters identified based on DNA methylation data, with corresponding mutation and clinical annotations displayed below. Patients harboring gene mutations are color-coded. B: Heatmap depicting the mutation profiles of each identified cluster. P values were calculated using Fisher’s exact test. *, P ≤ 0.05; **, P ≤ 0.01; ***, P ≤ 0.001. C: Box plot showing the number of mutations per patient across the identified clusters. P values were calculated using the Wilcoxon rank sum test. *, P ≤ 0.05; **, P ≤ 0.01. D: Bar plot displaying the number of clinical responders and non-responders within each cluster. P values were calculated using Fisher’s exact test. **, P ≤ 0.01. Two of the clusters identified could not be classified based on any specific mutation, but instead, patients in those clusters shared robust epigenetic profiles and clinical features. Patients in Cluster I (n=13) were characterized by older age (median 78 years old, range=69-85 vs. median 68, range = 37-86; p-value≤0.05) and intermediate-risk cytogenetics (p-value≤0.05), the majority of whom (n=10) did not respond to DNMTi treatment, though this did not achieve statistical significance. Finally, unlike patients in Cluster II, all of whom carried at least one mutation (with most patients carrying three or more), Cluster VI was striking because of its much lower mutation frequency than other clusters ( Figure 2C ) . However, despite this overall paucity of mutations, it was noteworthy that 5 out of 7 patients harboring DDX41 mutations (4 germline and 1 somatic) in our cohort were located in this cluster p-value≤0.001). Despite this, DDX41 mutations were not considered cluster-defining as they affected only 5/14 patients (36%) in the cluster. In addition to a low frequency of mutations, patients in cluster VI showed a higher response rate to AZA than patients in other clusters (defined as a response of hematologic improvement, partial or complete response within 6 months of beginning treatment), with 71% (10 out of 14) responding to the drug (p-value≤0.01; Figure 2D ) . In line with this paucity of mutations, 7 out of 12 patients from our MDS cohort that fell under the "Very Low," "Low," or "Moderate Low" 2022 IPSS-M risk categories were classified together in this cluster (p-value≤0.01) (Supplemental Table 5) . Notably, and in line with our PC analysis findings, a comparable hierarchical clustering analysis based on gene expression data failed to classify patients into robust clusters correlated with molecular and clinical features ( Supplemental Figure 2 ). Taken together, these findings highlight the robustness of DNAme in capturing the molecular and clinical heterogeneity of MDS. Cluster VI patients have mild epigenetic deregulation and shared clinical and laboratory features To further characterize the epigenetic deregulation of the epigenetic clusters, we compared them to healthy controls. We first conducted a PC analysis to visualize the epigenetic distances between controls and MDS patient clusters ( Figure 3A ). Notably, cluster VI, which carried very few mutations and responded well to AZA treatment, was epigenetically closer to the healthy controls than any of the other clusters, indicating the DNAme profiles of these patients are less perturbed. Supervised analysis comparing each cluster to the controls revealed that cluster VI displayed a lower number of DMRs than the other clusters. While the other six clusters all displayed widespread epigenetic abnormalities that clearly distinguished them from controls, with total DMRs ranging from 22,000 to 71,000, Cluster VI had only 5,855 hypomethylated DMRs and 1,654 hypermethylated DMRs (absolute methylation difference ≥ 25% and FDR ≤ 0.05; Figure 3B , Supplemental Figure 3 & 4 and Supplemental Table 6 ). Furthermore, Cluster VI exhibited significantly higher levels of hemoglobin (Hb) compared to the other clusters (Mean: 10.9 g/dL; range: 7.7-13.7 g/dL; ANOVA p-value≤0.05), once again indicating that patients in this cluster display more benign features than the rest of the cohort. Notably, there were no significant differences in BM blast percentage, red blood cells, platelets, white blood cells, or absolute neutrophil counts between Cluster VI and the other clusters ( Figure 3C ). Download figure Open in new tab Figure 3. Epigenetic and clinical characteristics associated with each DNA methylation cluster A: Principal Components Analysis of healthy controls (blue) and MDS patient clusters. B: Bar plot showing number of hypermethylated (red) and hypomethylated (blue) regions in each MDS cluster compared with healthy controls. C: Box plot illustrating the distribution of clinical features across MDS patient clusters. P values were calculated using ANOVA with a post hoc Tukey HSD test. *, P ≤ 0.05; **, P ≤ 0.01. To understand the distinct epigenetic deregulation associated with MDS patient cluster VI, we annotated the DMRs identified ( Figure 4A ) to neighboring coding and non-coding transcripts. Pathway analysis revealed they targeted genes involved in IL-2 and IL-6 signaling and other inflammation-related pathways ( Figure 4B ; Supplemental Table 7) . Comparing the transcriptional profiles between controls and cluster VI patients, we identified 930 down-regulated and 1,253 up-regulated DEGs ( Supplemental Table 8 ). GSEA revealed that while a few of the same inflammation-related pathways seen in the entire cohort of MDS cases ( Supplemental Figure 1C, Supplemental Table 3 ) were also upregulated in cluster VI, this cluster was characterized by a less inflammatory profile overall, showing instead distinct enrichment of pathways involved in DNA replication, cell cycle, DNA repair (mismatch repair, base excision repair, and homologous recombination), cysteine and methionine metabolism, and phosphatidylinositol signaling ( Figure 4C ; Supplemental Table 7) . Download figure Open in new tab Figure 4. Cluster VI patients display mild epigenetic deregulation and shared clinical and laboratory features A: Volcano plot illustrating methylation difference versus the -log 10 ( P -value) between healthy controls and MDS Cluster VI. Significant hypomethylated (blue) and hypermethylated (orange) regions in MDS Cluster VI are highlighted (absolute methylation difference ≥ 25%, FDR ≤ 0.05). B: Bubble plot depicting enriched pathways (FDR ≤ 0.1) associated with differentially methylated regions in MDS Cluster VI. Bubble size corresponds to the number of genes within each gene set, while color indicates statistical significance. C: Bar plot showing the normalized enrichment score (NES) for pathways enriched in MDS Cluster VI compared with healthy controls. Pathways with increased activity are shown in red, while those with decreased activity are shown in blue (FDR ≤ 0.1). These findings demonstrate how epigenetic profiles reveal novel biology in MDS. Namely, identified a distinct subgroup of MDS patients, defined by a unique DNAme signature and characterized by shared clinical and laboratory features as well as upregulation of cell cycle, DNA repair, and methionine pathways. This may help explain both the paucity of mutations in this group and its less aberrant DNAme profile, as well as an increased rate of incorporation of the nucleoside analog into the DNA, resulting in a more benign clinical presentation and improved response to DNMTi. Sensitivity to AZA is associated with specific epigenetic profiles at diagnosis Using International Working Group criteria 39 , we classified patients as responders (if they achieved a hematologic improvement, partial or complete response within 6 months of treatment initiation), and non-responders (those with stable or progressive disease). PC analysis based on gene expression profiles could not effectively differentiate between AZA responders and non-responders ( Figure 5A ) , while a supervised analysis approach identified only 25 DEGs between the two groups ( Figure 5B ; Supplemental Table 9). Despite this, GSEA revealed distinct pathway enrichment, with sensitivity to DNMTi associated with an expression profile characterized by a lower degree of inflammation ( Figure 5C ; Supplemental Table 10 ). By contrast, while unsupervised analysis of DNAme profiles also failed to robustly segregate overall AZA responders or even best- responders (i.e. those achieving a complete response) from non-responders (Supplemental Figure 5A-B ), supervised analysis revealed marked DNAme differences associated with response to AZA, consisting of 1,505 hypomethylated and 1,067 hypermethylated DMRs (absolute methylation difference ≥25% and FDR≤0.05; Figure 5D ; Supplemental Table 11) . These DMRs were enriched at intergenic regions (DMRs 48% vs. 36% p-value≤0.001), and strongly depleted from promoter regions (DMRs 5% vs. 21%, p-value≤0.001) and CpG islands (DMRs 7% vs. 25%, p-value≤0.001), highlighting a potential role for distal regulatory regions in helping determine response to AZA (Supplemental Figure 5C) . When annotated to neighboring genes, like gene expression, these DMRs were enriched in gene sets associated with cell adhesion and TGFb responses (Supplemental Figure 5D; Supplemental Table 10). Download figure Open in new tab Figure 5. Epigenetic differences distinguish AZA Responders and Non-Responders prior to treatment A: Principal Components Analysis of RNA-seq data from AZA non-responders (brown) and responders (orange). B: Volcano plot of the log 2 fold-change (AZA responders/non-responders) gene expression versus the -log 10 ( P -value). Significant downregulated and upregulated genes in AZA responders are shown in brown and orange, respectively (absolute fold change ≥ 1.5, FDR ≤ 0.05). C: Bar plot displaying the normalized enrichment score (NES) for pathways differentially regulated in AZA responders versus AZA non-responders. Upregulated pathways are shown in red, and downregulated pathways are shown in blue (FDR ≤ 0.05). D: Volcano plot showing methylation difference in methylated regions versus the -log 10 ( P - value) between AZA responders and AZA non-responders. Significant hypomethylated and hypermethylated regions in AZA responders are colored in green and light purple, respectively (absolute methylation difference ≥ 25%, FDR ≤ 0.05). E: Volcano plot depicting methylation difference versus the -log 10 ( P -value) between AZA complete responders and AZA non-responders. Significant hypomethylated (green) and hypermethylated (dark purple) regions in AZA complete responders are highlighted (absolute methylation difference ≥ 25% and FDR ≤ 0.05). F: Venn diagram ( top ) illustrating the overlap of differentially methylated regions between AZA non-responders vs. responders and AZA non-responders vs. complete responders. The bottom panel shows a bubble plot of enriched pathways (FDR ≤ 0.1) associated with shared hypomethylated and hypermethylated regions. Bubble size corresponds to the number of genes within each gene set, while color indicates statistical significance. To further refine the epigenetic profiles associated with sensitivity to AZA, we repeated the analysis, restricting the AZA responders’ group to those who achieved complete remission (CR) and compared them to non-responders. This analysis identified 1,198 hypomethylated and 608 hypermethylated DMRs (absolute methylation difference ≥ 25% and FDR ≤ 0.05) associated with achieving a CR ( Figure 5E ; Supplemental Table 12) . Once again, DMRs were enriched at intergenic regions (DMRs 47% vs. 35%, p- value≤0.001), and were strongly depleted from promoter regions (DMRs 6% vs. 22%, p- value≤0.001), and CpG islands (DMRs 9% vs. 26%, p-value≤0.001) (Supplemental Figure 5E) . Pathway analysis revealed that while these DMRs were enriched in inflammation pathways as before, they were also enriched in signaling pathways such as cAMP, RAS, calcium signaling, and cell adhesion pathways (Supplemental Figure 5F; Supplemental Table 13) . Notably, a subset of DMRs overlapped between the two comparisons, and these overlapping DMRs were enriched in signaling pathways such as cAMP, TGFb, and calcium signaling pathways, indicating that these may play a central role in response to AZA ( Figure 5F ; Supplemental Table 14 ). In summary, comprehensive DNAme analysis provided valuable insights into the response of MDS patients to AZA treatment. Moreover, the genomic distribution pattern of these response-associated DMRs explains why promoter-centric analyses did not identify any correlation between DNAme profiles and response to DNMTi and underscores the biological importance of epigenetic deregulation at distal genomic elements. DNAme profiles can be harnessed for the robust prediction of therapeutic response Given the existence of baseline DNAme differences in responders and non-responders, we hypothesized that these unique methylation profiles could be harnessed to predict which patients would be sensitive or resistant to AZA treatment. To test this, we first used stratified randomization to divide patients into a training and independent validation cohort while ensuring a balanced inclusion of responders and non-responders in both groups. The training and independent validation cohorts consisted of 63 and 27 patients, respectively. To avoid any potential overfitting of the model, percent cytosine methylation at 154 DMRs identified between responders and non-responders using only the 63 randomly selected patients in the training cohort was used as input ( Supplemental Table 15 ) for a machine-learning approach using random forest (RF) to build a classifier. Five- fold cross-validation was performed, and input features were ranked and selected based on their importance. Ninety-eight selected features were then used to train the final RF model with the best parameters and applied to the 27 patients in the independent blinded test set. The reported area under the receiver operating characteristic curve (ROC-AUC) on the test set was used to evaluate the classifier’s performance ( Figure 6A ). This DNAme RF classifier demonstrated the ability to predict the response to AZA with an AUC score of 0.82, with a sensitivity of 0.7 and a specificity of 0.77 ( Figure 6B ). Download figure Open in new tab Figure 6. Machine learning models predict response to AZA at baseline A: Schematic representation of the machine learning model used to identify predictive features for AZA response at diagnosis. B: Workflow of machine learning predictor utilizing DNA methylation features to distinguish AZA responders from non-responders. The receiver operating characteristic curve of biomarker performance is shown. C: Protein-protein interaction network of genes associated with differentially methylated regions within the same topologically associating domain. Line thickness represents connection confidence levels. D: Workflow of an integrated machine learning predictor incorporating DNA methylation, gene expression, mutations, and clinical features. The receiver operating characteristic curve of biomarker performance is shown. E: Pie chart illustrating the distribution of selected features from the model in D ( top ), and bar plot depicting the importance of each selected feature ( bottom ). Despite this robustness, we hypothesized that incorporating clinical, mutation, and gene expression information would further improve its performance. Since previous results showed only a few genes with differential expression between responders and non- responders, we first sought to identify genes whose expression is likely epigenetically regulated by DNAme. We previously reported that analyzing gene expression- methylation correlations within functional genomic units determined by topologically associating domains (TADs) leads to the identification of genes regulated via DNAme. 40 Therefore, we used 2,711 TAD regions identified in CD34+ HSPC and, for each sample for which we had gene expression, mutation, clinical information, and DNAme data available (n=69), we calculated the Pearson correlation between the DNAme level of all the DMRs between responders and non-responders and the gene expression value of each gene transcript within the same TAD. As a result, 4,612 genes were found to be highly correlated with 1,092 DMRs within 820 TADs, with an absolute r≥0.7. Next, we performed network analysis to identify relevant networks of epigenetically regulated genes. The top network identified consisted of 15 genes forming a highly interconnected network with CCL5 as the hub node ( Figure 6C ) . Notably, several genes in this network were regulated via DMRs located at long distances from their loci, again underscoring the importance that distal events play on epigenetic deregulation in MDS (Supplemental Figure 6A). Using this cohort of patients with available DNAme, RNA-seq, mutation, and clinical information, we randomly assigned patients to a training cohort (n=48) and an independent validation cohort (n=21). Using DMRs (n=192), the expression from our CCL5 network (n=25), mutation data (n=69 genes), and clinical parameters (n=7) previously shown to be relevant to MDS biology and prognosis 41 – 45 ( Supplemental Table 16 ), we performed feature selection as before. Fifty-three features were selected, consisting of a combination of clinical, mutation, expression, and DNAme data. Among the clinical data, hemoglobin, age, platelet count, BM blast percentage, and absolute neutrophil counts (ANC) were included in the final model. Twenty-three gene mutations were included in the RF model. While some of them have been shown to have prognostic impact in MDS (e.g. NPM1 46 , TP53 47 , FLT3 48 , 49 , and CUX1 50 , 51 ), most of them have not been previously implicated as either prognostic or predictive of response to therapy in MDS. Notably, TET2 mutations, for which conflicting findings regarding their correlation with AZA response have been reported 52 – 54 , had the lowest feature importance in our model, while ASXL1 and DNMT3A mutations, which have also been reported to correlate with response 55 – 57 , were not selected at all. When applied to the 21 patients in the blinded independent test set, this new prediction model resulted in a higher AUC score of 0.93, with a sensitivity of 0.83 and specificity of 0.93, demonstrating the increased robustness of this integrative classifier for predicting AZA response ( Figure 6D-E ) . To determine the relative importance of the different types of features included, we also tested predictive models, excluding each of the different data types one at a time. Exclusion of the DNAme data resulted in the classifier’s lowest performance, with an AUC score of 0.54 ( Supplemental Figure 6B ). Conversely, including DNAme while excluding gene expression features from the same model improved performance to an AUC score of 0.79, with a sensitivity of 0.8 and a specificity of 0.83. However, this was still significantly lower than the complete classifier’s performance ( Supplemental Figure 6C ). Exclusion of either mutation or clinical data also resulted in a decrease in the classifier’s performance, with AUC scores of 0.78 and 0.79, respectively (Supplemental Figure 6D- E). These findings indicate that DNAme data significantly contribute to the predictive power of classifier models, even to a greater extent than gene expression, mutation, and clinical and laboratory data. It is important to note that the highest-performing model resulted when integrating all four types of information, underscoring the synergism among the different features. DISCUSSION The slow DNMTi response and the necessity for ongoing treatment present significant challenges in assessing treatment effectiveness and adjusting therapeutic regimens promptly. By integrating genomic, epigenomic, and transcriptomic data, we gained insights into the complex biological networks underlying MDS heterogeneity, identifying those likely to respond to AZA. Understanding critical genomic regions, gene expression patterns, mutations, and clinical parameters related to AZA sensitivity is essential for developing more targeted and effective therapies. Our study offers initial insights into long-range epigenetic gene regulation mechanisms in MDS pathogenesis. We found that genomic regions affected by DNAme primarily reside in distal regulatory areas. Hi-C sequencing revealed how DMRs control the expression of distal genes within the same TADs, confirming previous findings in AML 40 and emphasizing the need to focus on promoter and distal regulatory elements in epigenetic dysregulation research in MDS. Recent efforts have been made to classify MDS patients on the basis of its molecular make-up, resulting in the creation of the International Consensus Classification (ICC) 58 and the 5th edition of the WHO Classification 59 , and the IPSS-M for prognostication. However, these classifications overlook gene expression and epigenetic information, crucial for understanding the disease. The epigenetic classification in this study identified a novel MDS subgroup (Cluster VI), characterized by a unique phenotype, few mutations, near-normal hemoglobin, lower-risk IPSS-M, and a higher proportion of responders to AZA treatment. Though it shares features with the CCUS-like group described by Bernard et al., that group included both MDS and CMML patients, which are distinct disorders with different molecular profiles and therapy responses. Furthermore, DDX41 mutations, linked to AML and MDS predisposition and favorable outcomes 60 , were more prevalent in this cluster, contrasting with the Bernard classification that separates DDX41 mutation patients despite their shared features with others, as demonstrated in our study. While prior studies have linked specific mutations to an increased response rate to DNMTi’s, this did not consistently correlate with improved survival, nor were they reproducible across other studies. 52 , 61 , 62 , 53 , 63 – 65 We report the development of robust predictive classifiers by integrating DNAme, gene expression, mutations, and clinical and laboratory parameters. This approach was critical in enhancing the predictive power of our model. Specifically, mutations in 23 genes -including NPM1, RAD21, and TP53 - along with the expression patterns of 16 genes, such as C5AR2 , C3, and TNFRSF9 , significantly contributed to the model’s performance. While some of these mutations and gene expression patterns have been reported in previous studies on myeloid disorders 66 – 69 , they have not been collectively used as biomarkers for AZA response. Notably, excluding DNAme from the model resulted in a significant decline in performance, highlighting the importance of DNAme information in capturing disease biology and improving prediction accuracy. Our findings highlight the potential of epigenetic data as predictive biomarkers and emphasize the importance of integrating multiple molecular features for improved predictive accuracy. Prospective validation in larger cohorts will be necessary to fully validate these classifiers for their incorporation into clinical practice. In conclusion, our integrative approach sheds light on the complexity of MDS and underscores how this molecular interplay influences treatment response. This holistic perspective can offer valuable insights into the molecular mechanisms underlying AZA response in MDS and may ultimately lead to more effective and personalized treatment strategies. Further exploration of the roles of biological pathways identified will be critical for advancing our understanding of MDS pathogenesis and therapeutic approaches. AUTHOR CONTRIBUTIONS MEF conceived and designed the study and supervised all aspects of the analysis and data interpretation; QY and MTM analyzed genome-wide data; RB analyzed mutational data; MT contributed to clinical data organization and annotation; AB, IC, EC, SH, EAG, MYF, CF, JPR, MJR, AED, MAS, and VS contributed clinically annotated samples; MAS and VS provided critical feedback and contributed to data interpretation; QY and MEF wrote the manuscript with approval from all co-authors. DISCLOSURE OF CONFLICTS OF INTEREST RB reports employment and equity from Aptose Biosciences, consulting fees from Bristol-Myers Squibb (BMS), Servier, Geron, Ipsen, Keros, Gilead, BigSur, and SAB membership for NeoGenomics. EAG reports Advisory Board/Honoraria fees from AbbVie, Alexion Pharmaceuticals/ AZ rare disease, Apellis, Celgene/BMS, CTI Biopharma, Genentech, Novartis, Picnic Health, Servier, Takeda Oncology, Taiho Oncology and Research Funding from Astex Pharmaceuticals, AstraZeneca Rare Disease, Alexion Pharmaceuticals, Apellis Pharmaceuticals, Blueprint Medicines, Genentech Inc, NextCure. MAS serves on the advisory board for BMS. AED reports consulting fees from BMS, Novartis, Takeda, Pfizer, Agios. QY, MTM, AB, MT, IC, EC, SH, MYF, CF, JPR, MJR, VS, and MEF have no significant conflicts to declare. Data sharing statement The ERRBS and bulk RNA-seq data have been deposited in the NCBI’s Gene Expression Omnibus (GEO) database under the accession code GSE281562 (RNA-seq) and GSE281610 (ERRBS). Additional ERRBS data for healthy donors was obtained from GSE52945, samples GSM1536373, GSM1536374 and GSM1536375. For normal RNA data, 8 of them were from publications. 1 ACKNOWLEDGMENTS We thank all members of the Figueroa Lab for thoughtful discussions and suggestions. This study was supported by award R01HL126947 from the NIH NHLBI, a Leukemia and Lymphoma Society (LLS) scholar award (13-5719), and an LLS SCOR award (7028-22) to MEF. Research reported in this publication was performed in part at the Onco- Genomics Shared Resource (OGSR) of the Sylvester Comprehensive Cancer Center at the University of Miami, RRID: SCR_022502, which is supported by the National Cancer Institute (NCI) of the National Institutes of Health (NIH) under award number P30CA240139. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH. Library construction and sequencing services were provided by the University of Miami’s John P. Hussman Institute for Human Genomics Core Facility (RRID:SCR_017828). Samples for this work were provided by the Roswell Park Comprehensive Cancer Centers Biorepository and Laboratory Services Shared Resource supported by National Cancer Institute (NCI) grant P30CA016056. Funding National Institutes of Health, , R01HL126947 , P30CA016056 , P30CA240139 Leukemia and Lymphoma Society, https://ror.org/04m0xav37 , 13-5719 , 7028-22 Footnotes ↵ § Centre for Genomic Regulation, Barcelona Institute of Science and Technology, Barcelona, Spain ↵ $ Dept. of Hematology, Nagasaki University, Nagasaki, Japan https://github.com/santur90/MDS_AZA References 1. ↵ Li S , Garrett-Bakelman FE , Chung SS , et al. Distinct evolution and dynamics of epigenetic and genetic heterogeneity in acute myeloid leukemia . Nat Med . 2016 ; 22 ( 7 ): 792 – 799 . doi: 10.1038/nm.4125 OpenUrl CrossRef PubMed 2. ↵ Nimer SD. Myelodysplastic syndromes . Blood . 2008 ;111(10):4841-4851. doi: 10.1182/blood-2007-08-078139 OpenUrl Abstract / FREE Full Text 3. Garcia-Manero G. Myelodysplastic syndromes: 2014 update on diagnosis, risk- stratification, and management . Am J Hematol . 2014 ; 89 ( 1 ):97-108. doi: 10.1002/ajh.23642 OpenUrl CrossRef PubMed 4. ↵ Menssen AJ , Walter MJ . Genetics of progression from MDS to secondary leukemia . Blood . 2020 ; 136 ( 1 ): 50 – 60 . doi: 10.1182/blood.2019000942 OpenUrl CrossRef PubMed 5. ↵ Melki JR , Vincent PC , Brown RD , Clark SJ . Hypermethylation of E-cadherin in leukemia . Blood . 2000 ; 95 ( 10 ): 3208 – 3213 . OpenUrl Abstract / FREE Full Text 6. Christiansen DH , Andersen MK , Pedersen-Bjergaard J . Methylation of p15INK4B is common, is associated with deletion of genes on chromosome arm 7q and predicts a poor prognosis in therapy-related myelodysplasia and acute myeloid leukemia . Leukemia . 2003 ; 17 ( 9 ): 1813 – 1819 . doi: 10.1038/sj.leu.2403054 OpenUrl CrossRef PubMed Web of Science 7. ↵ Tessema M , Länger F , Dingemann J , Ganser A , Kreipe H , Lehmann U . Aberrant methylation and impaired expression of the p15INK4b cell cycle regulatory gene in chronic myelomonocytic leukemia (CMML) . Leukemia . 2003 ; 17 ( 5 ): 910 – 918 . doi: 10.1038/sj.leu.2402891 OpenUrl CrossRef PubMed Web of Science 8. ↵ Figueroa ME , Skrabanek L , Li Y , et al. MDS and secondary AML display unique patterns and abundance of aberrant DNA methylation . Blood. Published online 2009 . 9. ↵ Jiang Y , Dunbar A , Gondek LP , et al. Aberrant DNA methylation is a dominant mechanism in MDS progression to AML . Blood . 2009 ; 113 ( 6 ): 1315 – 1325 . doi: 10.1182/blood-2008-06-163246 OpenUrl Abstract / FREE Full Text 10. ↵ Bejar R , Stevenson KE , Caughey BA , et al. Validation of a Prognostic Model and the Impact of Mutations in Patients With Lower-Risk Myelodysplastic Syndromes . J Clin Oncol . 2012 ; 30 ( 27 ): 3376 – 3382 . doi: 10.1200/JCO.2011.40.7379 OpenUrl Abstract / FREE Full Text 11. Bejar R , Stevenson K , Abdel-Wahab O , et al. Clinical Effect of Point Mutations in Myelodysplastic Syndromes . N Engl J Med . 2011 ; 364 ( 26 ): 2496 – 2506 . doi: 10.1056/NEJMoa1013343 OpenUrl CrossRef PubMed Web of Science 12. ↵ Palomo L , Meggendorfer M , Hutter S , et al. Molecular landscape and clonal architecture of adult myelodysplastic/myeloproliferative neoplasms . Blood . 2020 ; 136 ( 16 ): 1851 – 1862 . doi: 10.1182/blood.2019004229 OpenUrl CrossRef PubMed 13. ↵ Burnett AK , Milligan D , Prentice AG , et al. A comparison of low-dose cytarabine and hydroxyurea with or without all-trans retinoic acid for acute myeloid leukemia and high-risk myelodysplastic syndrome in patients not considered fit for intensive treatment . Cancer . 2007 ; 109 ( 6 ): 1114 – 1124 . doi: 10.1002/cncr.22496 OpenUrl CrossRef PubMed Web of Science 14. ↵ Lin P , Ren Y , Yan X , et al. The high NRF2 expression confers chemotherapy resistance partly through up-regulated DUSP1 in myelodysplastic syndromes . Haematologica . 2019 ; 104 ( 3 ): 485 – 496 . doi: 10.3324/haematol.2018.197749 OpenUrl Abstract / FREE Full Text 15. ↵ Griffiths EA , Gore SD . DNA Methyltransferase and Histone Deacetylase Inhibitors in the Treatment of Myelodysplastic Syndromes . Semin Hematol . 2008 ; 45 ( 1 ): 23 – 30 . doi: 10.1053/j.seminhematol.2007.11.007 OpenUrl CrossRef PubMed Web of Science 16. Fenaux P , Mufti GJ , Hellstrom-Lindberg E , et al. Efficacy of azacitidine compared with that of conventional care regimens in the treatment of higher-risk myelodysplastic syndromes: a randomised, open-label, phase III study . Lancet Oncol . 2009 ; 10 ( 3 ): 223 – 232 . doi: 10.1016/S1470-2045(09)70003-8 OpenUrl CrossRef PubMed Web of Science 17. Platzbecker U , Wermke M , Radke J , et al. Azacitidine for treatment of imminent relapse in MDS or AML patients after allogeneic HSCT: results of the RELAZA trial . Leukemia . 2012 ; 26 ( 3 ): 381 – 389 . doi: 10.1038/leu.2011.234 OpenUrl CrossRef PubMed Web of Science 18. ↵ Gore SD , Fenaux P , Santini V , et al. A multivariate analysis of the relationship between response and survival among patients with higher-risk myelodysplastic syndromes treated within azacitidine or conventional care regimens in the randomized AZA-001 trial . Haematologica . 2013 ; 98 ( 7 ): 1067 – 1072 . doi: 10.3324/haematol.2012.074831 OpenUrl Abstract / FREE Full Text 19. ↵ Sekeres MA , Othus M , List AF , et al. Randomized Phase II Study of Azacitidine Alone or in Combination With Lenalidomide or With Vorinostat in Higher-Risk Myelodysplastic Syndromes and Chronic Myelomonocytic Leukemia: North American Intergroup Study SWOG S1117 . J Clin Oncol . 2017 ; 35 ( 24 ): 2745 – 2753 . doi: 10.1200/JCO.2015.66.2510 OpenUrl CrossRef PubMed 20. ↵ Prébet T , Gore SD , Esterni B , et al. Outcome of High-Risk Myelodysplastic Syndrome After Azacitidine Treatment Failure . J Clin Oncol . 2011 ; 29 ( 24 ): 3322 – 3327 . doi: 10.1200/JCO.2011.35.8135 OpenUrl Abstract / FREE Full Text 21. ↵ Daskalakis M , Nguyen TT , Nguyen C , et al. Demethylation of a hypermethylated P15/INK4B gene in patients with myelodysplastic syndrome by 5-Aza-2′- deoxycytidine (decitabine) treatment . Blood . 2002 ; 100 ( 8 ): 2957 – 2964 . doi: 10.1182/blood.V100.8.2957 OpenUrl Abstract / FREE Full Text 22. Mund C , Hackanson B , Stresemann C , Lübbert M , Lyko F . Characterization of DNA Demethylation Effects Induced by 5-Aza-2′-Deoxycytidine in Patients with Myelodysplastic Syndrome . Cancer Res . 2005 ; 65 ( 16 ): 7086 – 7090 . doi: 10.1158/0008-5472.CAN-05-0695 OpenUrl Abstract / FREE Full Text 23. ↵ Shen L , Kantarjian H , Guo Y , et al. DNA Methylation Predicts Survival and Response to Therapy in Patients With Myelodysplastic Syndromes . J Clin Oncol . 2010 ; 28 ( 4 ): 605 – 613 . doi: 10.1200/JCO.2009.23.4781 OpenUrl Abstract / FREE Full Text 24. ↵ Follo MY , Finelli C , Mongiorgi S , et al. Reduction of phosphoinositide-phospholipase C beta1 methylation predicts the responsiveness to azacitidine in high-risk MDS . Proc Natl Acad Sci . 2009 ; 106 ( 39 ): 16811 – 16816 . doi: 10.1073/pnas.0907109106 OpenUrl Abstract / FREE Full Text 25. ↵ Issa JPJ , Garcia-Manero G , Giles FJ , et al. Phase 1 study of low-dose prolonged exposure schedules of the hypomethylating agent 5-aza-2′-deoxycytidine (decitabine) in hematopoietic malignancies . Blood . 2004 ; 103 ( 5 ): 1635 – 1640 . doi: 10.1182/blood-2003-03-0687 OpenUrl Abstract / FREE Full Text 26. ↵ Fandy TE , Herman JG , Kerns P , et al. Early epigenetic changes and DNA damage do not predict clinical response in an overlapping schedule of 5-azacytidine and entinostat in patients with myeloid malignancies . Blood . 2009 ; 114 ( 13 ): 2764 – 2773 . doi: 10.1182/blood-2009-02-203547 OpenUrl Abstract / FREE Full Text 27. ↵ Greenberg P , Cox C , LeBeau MM , et al. International Scoring System for Evaluating Prognosis in Myelodysplastic Syndromes . Blood . 1997 ; 89 ( 6 ): 2079 – 2088 . doi: 10.1182/blood.V89.6.2079 OpenUrl Abstract / FREE Full Text 28. ↵ Thongsa-ad U , Wongpan A , Wongkummool W , et al. Improving hematopoietic differentiation from human induced pluripotent stem cells by the modulation of Hippo signaling with a diarylheptanoid derivative . Stem Cell Res Ther . 2024 ; 15 ( 1 ): 60 . doi: 10.1186/s13287-024-03686-4 OpenUrl CrossRef PubMed 29. ↵ Rampal R , Alkalin A , Madzo J , et al. DNA Hydroxymethylation Profiling Reveals that WT1 Mutations Result in Loss of TET2 Function in Acute Myeloid Leukemia . Cell Rep . 2014 ; 9 ( 5 ): 1841 – 1855 . doi: 10.1016/j.celrep.2014.11.004 OpenUrl CrossRef PubMed 30. ↵ Kim I , Park T , Noh JY , Kim W . Emerging role of Hippo pathway in the regulation of hematopoiesis . BMB Rep . 2023 ; 56 ( 8 ): 417 – 425 . doi: 10.5483/BMBRep.2023-0094 OpenUrl CrossRef PubMed 31. ↵ Staal FJT , Clevers HC . WNT signalling and haematopoiesis: a WNT–WNT situation . Nat Rev Immunol . 2005 ; 5 ( 1 ): 21 – 30 . doi: 10.1038/nri1529 OpenUrl CrossRef PubMed Web of Science 32. ↵ Khattar E , Maung KZY , Chew CL , et al. Rap1 regulates hematopoietic stem cell survival and affects oncogenesis and response to chemotherapy . Nat Commun . 2019 ; 10 ( 1 ): 5349 . doi: 10.1038/s41467-019-13082-9 OpenUrl CrossRef PubMed 33. ↵ Imai T , Tanaka H , Hamazaki Y , Minato N . Rap1 signal modulators control the maintenance of hematopoietic progenitors in bone marrow and adult long-term hematopoiesis . Cancer Sci . 2019 ; 110 ( 4 ): 1317 – 1330 . doi: 10.1111/cas.13974 OpenUrl CrossRef PubMed 34. ↵ Kalaitzidis D , Sykes SM , Wang Z , et al. mTOR Complex 1 Plays Critical Roles in Hematopoiesis and Pten-Loss-Evoked Leukemogenesis . Cell Stem Cell . 2012 ; 11 ( 3 ): 429 – 439 . doi: 10.1016/j.stem.2012.06.009 OpenUrl CrossRef PubMed Web of Science 35. ↵ Fernandes H , Moura J , Carvalho E . mTOR Signaling as a Regulator of Hematopoietic Stem Cell Fate . Stem Cell Rev Rep . 2021 ; 17 ( 4 ): 1312 – 1322 . doi: 10.1007/s12015-021-10131-z OpenUrl CrossRef PubMed 36. ↵ Ito K , Hirao A , Arai F , et al. Reactive oxygen species act through p38 MAPK to limit the lifespan of hematopoietic stem cells . Nat Med . 2006 ; 12 ( 4 ): 446 – 451 . doi: 10.1038/nm1388 OpenUrl CrossRef PubMed Web of Science 37. ↵ Geest CR , Coffer PJ . MAPK signaling pathways in the regulation of hematopoiesis . J Leukoc Biol . 2009 ; 86 ( 2 ): 237 – 250 . doi: 10.1189/jlb.0209097 OpenUrl CrossRef PubMed Web of Science 38. ↵ Figueroa ME , Lugthart S , Li Y , et al. DNA Methylation Signatures Identify Biologically Distinct Subtypes in Acute Myeloid Leukemia . Cancer Cell. Published online 2010 . 39. ↵ Cheson BD . Clinical application and proposal for modification of the International Working Group (IWG) response criteria in myelodysplasia . Blood . 2006 ; 108 ( 2 ): 419 – 425 . doi: 10.1182/blood-2005-10-4149 OpenUrl Abstract / FREE Full Text 40. ↵ Telonis AG , Yang Q , Huang HT , Figueroa ME . MIR retrotransposons link the epigenome and the transcriptome of coding genes in acute myeloid leukemia . Nat Commun . 2022 ; 13 ( 1 ): 6524 . doi: 10.1038/s41467-022-34211-x OpenUrl CrossRef PubMed 41. ↵ Scott BL , Park JY , Deeg HJ , et al. Pretransplant Neutropenia Is Associated with Poor- Risk Cytogenetic Features and Increased Infection-Related Mortality in Patients with Myelodysplastic Syndromes . Biol Blood Marrow Transplant . 2008 ; 14 ( 7 ): 799 – 806 . doi: 10.1016/j.bbmt.2008.04.011 OpenUrl CrossRef PubMed 42. Malcovati L , Della Porta MG , Strupp C , et al. Impact of the degree of anemia on the outcome of patients with myelodysplastic syndrome and its integration into the WHO classification-based Prognostic Scoring System (WPSS) . Haematologica . 2011 ; 96 ( 10 ): 1433 – 1440 . doi: 10.3324/haematol.2011.044602 OpenUrl Abstract / FREE Full Text 43. Tong W , Quintás-Cardama A , Kadia T , et al. Predicting survival of patients with hypocellular myelodysplastic syndrome: Development of a disease-specific prognostic score system . Cancer . 2012 ; 118 ( 18 ): 4462 – 4470 . doi: 10.1002/cncr.27420 OpenUrl CrossRef PubMed 44. Greenberg PL , Tuechler H , Schanz J , et al. Revised International Prognostic Scoring System for Myelodysplastic Syndromes . Blood . 2012 ; 120 ( 12 ): 2454 – 2465 . doi: 10.1182/blood-2012-03-420489 OpenUrl Abstract / FREE Full Text 45. ↵ Karantanos T , Gondek LP , Varadhan R , et al. Gender-related differences in the outcomes and genomic landscape of patients with myelodysplastic syndrome/myeloproliferative neoplasm overlap syndromes . Br J Haematol . 2021 ; 193 ( 6 ): 1142 – 1150 . doi: 10.1111/bjh.17534 OpenUrl CrossRef PubMed 46. ↵ Montalban-Bravo G , Kanagal-Shamanna R , Sasaki K , et al. NPM1 mutations define a specific subgroup of MDS and MDS/MPN patients with favorable outcomes with intensive chemotherapy . Blood Adv . 2019 ; 3 ( 6 ): 922 – 933 . doi: 10.1182/bloodadvances.2018026989 OpenUrl Abstract / FREE Full Text 47. ↵ Montoro MJ , Palomo L , Haferlach C , et al. Influence of TP53 gene mutations and their allelic status in myelodysplastic syndromes with isolated 5q deletion . Blood . 2024 ; 144 ( 16 ): 1722 – 1731 . doi: 10.1182/blood.2024023840 OpenUrl CrossRef PubMed 48. ↵ Bains A , Luthra R , Medeiros LJ , Zuo Z . FLT3 and NPM1 Mutations in Myelodysplastic Syndromes . Am J Clin Pathol . 2011 ; 135 ( 1 ): 62 – 69 . doi: 10.1309/AJCPEI9XU8PYBCIO OpenUrl CrossRef PubMed 49. ↵ Daver N , Strati P , Jabbour E , et al. FLT3 mutations in myelodysplastic syndrome and chronic myelomonocytic leukemia . Am J Hematol . 2013 ; 88 ( 1 ): 56 – 59 . doi: 10.1002/ajh.23345 OpenUrl CrossRef PubMed 50. ↵ Chronic Myeloid Disorders Working Group of the International Cancer Genome Consortium , Wong CC , Martincorena I , et al. Inactivating CUX1 mutations promote tumorigenesis . Nat Genet . 2014 ;46(1):33-38. doi: 10.1038/ng.2846 OpenUrl CrossRef PubMed 51. ↵ Dermawan JK , Wensel C , Visconte V , Maciejewski JP , Cook JR , Bosler DS . Clinically Significant CUX1 Mutations Are Frequently Subclonal and Common in Myeloid Disorders With a High Number of Co-mutated Genes and Dysplastic Features . Am J Clin Pathol . 2022 ; 157 ( 4 ): 586 – 594 . doi: 10.1093/ajcp/aqab157 OpenUrl CrossRef PubMed 52. ↵ Bejar R , Lord A , Stevenson K , et al. TET2 mutations predict response to hypomethylating agents in myelodysplastic syndrome patients . Blood . 2014 ; 124 ( 17 ): 2705 – 2712 . doi: 10.1182/blood-2014-06-582809 OpenUrl Abstract / FREE Full Text 53. ↵ Coombs CC , Sallman DA , Devlin SM , et al. Mutational correlates of response to hypomethylating agent therapy in acute myeloid leukemia . Haematologica . 2016 ; 101 ( 11 ): e457 – e460 . doi: 10.3324/haematol.2016.148999 OpenUrl FREE Full Text 54. ↵ Döhner H , Dolnik A , Tang L , et al. Cytogenetics and gene mutations influence survival in older patients with acute myeloid leukemia treated with azacitidine or conventional care . Leukemia . 2018 ; 32 ( 12 ): 2546 – 2557 . doi: 10.1038/s41375-018-0257-z OpenUrl CrossRef PubMed 55. ↵ Falconi G , Fabiani E , Piciocchi A , et al. Somatic mutations as markers of outcome after azacitidine and allogeneic stem cell transplantation in higher-risk myelodysplastic syndromes . Leukemia . 2019 ; 33 ( 3 ): 785 – 790 . doi: 10.1038/s41375-018-0284-9 OpenUrl CrossRef PubMed 56. Scheller M , Ludwig AK , Göllner S , et al. Hotspot DNMT3A mutations in clonal hematopoiesis and acute myeloid leukemia sensitize cells to azacytidine via viral mimicry response . Nat Cancer . 2021 ; 2 ( 5 ): 527 – 544 . doi: 10.1038/s43018-021-00213-9 OpenUrl CrossRef PubMed 57. ↵ Rahmani NE , Ramachandra N , Sahu S , et al. ASXL1 mutations are associated with distinct epigenomic alterations that lead to sensitivity to venetoclax and azacytidine . Blood Cancer J . 2021 ; 11 ( 9 ): 157 . doi: 10.1038/s41408-021-00541-0 OpenUrl CrossRef PubMed 58. ↵ Arber DA , Orazi A , Hasserjian RP , et al. International Consensus Classification of Myeloid Neoplasms and Acute Leukemias: integrating morphologic, clinical, and genomic data . Blood . 2022 ; 140 ( 11 ): 1200 – 1228 . doi: 10.1182/blood.2022015850 OpenUrl CrossRef PubMed 59. ↵ Khoury JD , Solary E , Abla O , et al. The 5th edition of the World Health Organization Classification of Haematolymphoid Tumours: Myeloid and Histiocytic/Dendritic Neoplasms . Leukemia . 2022 ; 36 ( 7 ): 1703 – 1719 . doi: 10.1038/s41375-022-01613-1 OpenUrl CrossRef PubMed 60. ↵ Sébert M , Passet M , Raimbault A , et al. Germline DDX41 mutations define a significant entity within adult MDS/AML patients . Blood . 2019 ; 134 ( 17 ): 1441 – 1444 . doi: 10.1182/blood.2019000909 OpenUrl Abstract / FREE Full Text 61. ↵ Traina F , Visconte V , Elson P , et al. Impact of molecular mutations on treatment response to DNMT inhibitors in myelodysplasia and related neoplasms . Leukemia . 2014 ; 28 ( 1 ): 78 – 87 . doi: 10.1038/leu.2013.269 OpenUrl CrossRef PubMed Web of Science 62. ↵ Kim M , Yahng SA , Kwon A , et al. Mutation in TET2 or TP53 predicts poor survival in patients with myelodysplastic syndrome receiving hypomethylating treatment or stem cell transplantation . Bone Marrow Transplant . 2015 ; 50 ( 8 ): 1132 – 1134 . doi: 10.1038/bmt.2015.110 OpenUrl CrossRef PubMed 63. ↵ Cedena MT , Rapado I , Santos-Lozano A , et al. Mutations in the DNA methylation pathway and number of driver mutations predict response to azacitidine in myelodysplastic syndromes . Oncotarget . 2017 ; 8 ( 63 ): 106948 – 106961 . doi: 10.18632/oncotarget.22157 OpenUrl CrossRef PubMed 64. Itzykson R , Kosmider O , Cluzeau T , et al. Impact of TET2 mutations on response rate to azacitidine in myelodysplastic syndromes and low blast count acute myeloid leukemias . Leukemia . 2019 ; 25 ( 7 ): 1 – 6 . doi: 10.1038/leu.2011.71 OpenUrl CrossRef 65. ↵ Li X , Xu F , Zhang Z , et al. Dynamics of epigenetic regulator gene BCOR mutation and response predictive value for hypomethylating agents in patients with myelodysplastic syndrome . Clin Epigenetics . 2021 ; 13 ( 1 ): 169 . doi: 10.1186/s13148-021-01157-8 OpenUrl CrossRef PubMed 66. ↵ Mullighan CG , Zhang J , Kasper LH , et al. CREBBP mutations in relapsed acute lymphoblastic leukaemia . Nature . 2011 ; 471 ( 7337 ):235-239. doi: 10.1038/nature09727 OpenUrl CrossRef PubMed Web of Science 67. Eckstein OS , Wang L , Punia JN , et al. Mixed Phenotype Acute Leukemia (MPAL) Has a High Frequency of Mutations in Epigenetic Regulatory Genes: Results from Whole Exome Sequencing . Blood . 2014 ; 124 ( 21 ): 3560 – 3560 . doi: 10.1182/blood.V124.21.3560.3560 OpenUrl CrossRef 68. Chen Y , Chun-Fung S , Tu L , et al. Pathogenetic Role of RNA Splicing and NRAS Mutation Combination in Chronic Myelomonocytic Leukemia . Blood . 2022 ; 140 (Supplement 1 ): 11491 – 11491 . doi: 10.1182/blood-2022-168466 OpenUrl CrossRef 69. ↵ Huang Y , Zheng H , Zhu Y , et al. Loss of CD28 expression associates with severe T- cell exhaustion in acute myeloid leukemia . Front Immunol . 2023 ; 14 : 1139517 . doi: 10.3389/fimmu.2023.1139517 OpenUrl CrossRef View the discussion thread. Back to top Previous Next Posted April 26, 2025. Download PDF Supplementary Material Data/Code Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following Unique epigenomic signatures identify biologically significant subtypes of MDS and predict response to azacitidine Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share Unique epigenomic signatures identify biologically significant subtypes of MDS and predict response to azacitidine Qin Yang , Miguel Torres-Martin , Masataka Taguchi , Alice Brogi , Irene Casalin , Eleonora Ceneri , Stephanie Halene , Amy E. DeZern , Elizabeth A. Griffiths , Matilde Y. Follo , Carlo Finelli , Jerald P. Radich , Michael J. Rauh , Rafael Bejar , Mikkael A. Sekeres , Valeria Santini , Maria E. Figueroa bioRxiv 2025.04.23.650336; doi: https://doi.org/10.1101/2025.04.23.650336 Share This Article: Copy Citation Tools Unique epigenomic signatures identify biologically significant subtypes of MDS and predict response to azacitidine Qin Yang , Miguel Torres-Martin , Masataka Taguchi , Alice Brogi , Irene Casalin , Eleonora Ceneri , Stephanie Halene , Amy E. DeZern , Elizabeth A. Griffiths , Matilde Y. Follo , Carlo Finelli , Jerald P. Radich , Michael J. Rauh , Rafael Bejar , Mikkael A. Sekeres , Valeria Santini , Maria E. Figueroa bioRxiv 2025.04.23.650336; doi: https://doi.org/10.1101/2025.04.23.650336 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Cancer Biology Subject Areas All Articles Animal Behavior and Cognition (7629) Biochemistry (17660) Bioengineering (13881) Bioinformatics (41910) Biophysics (21436) Cancer Biology (18576) Cell Biology (25480) Clinical Trials (138) Developmental Biology (13368) Ecology (19887) Epidemiology (2067) Evolutionary Biology (24302) Genetics (15598) Genomics (22482) Immunology (17726) Microbiology (40360) Molecular Biology (17163) Neuroscience (88534) Paleontology (666) Pathology (2830) Pharmacology and Toxicology (4821) Physiology (7637) Plant Biology (15129) Scientific Communication and Education (2045) Synthetic Biology (4290) Systems Biology (9817) Zoology (2269)