Genetic lesions such as BCR–ABL1, E2A–PBX1, and MLL rearrangements (MLLr) are associated with unfavorable outcomes in adult B-cell precursor acute lymphoblastic leukemia (B-ALL). Leukemia oncoproteins may directly or indirectly disrupt cytosine methylation patterning to mediate the malignant phenotype. We postulated that DNA methylation signatures in these aggressive B-ALLs would point toward disease mechanisms and useful biomarkers and therapeutic targets. We therefore conducted DNA methylation and gene expression profiling on a cohort of 215 adult patients with B-ALL enrolled in a single phase III clinical trial (ECOG E2993) and normal control B cells. In BCR–ABL1-positive B-ALLs, aberrant cytosine methylation patterning centered around a cytokine network defined by hypomethylation and overexpression of IL2RA(CD25). The E2993 trial clinical data showed that CD25 expression was strongly associated with a poor outcome in patients with ALL regardless of BCR–ABL1 status, suggesting CD25 as a novel prognostic biomarker for risk stratification in B-ALLs. In E2APBX1-positive B-ALLs, aberrant DNA methylation patterning was strongly associated with direct fusion protein binding as shown by the E2A–PBX1 chromatin immunoprecipitation (ChIP) sequencing (ChIP-seq), suggesting that E2A–PBX1 fusion protein directly remodels the epigenome to impose an aggressive B-ALL phenotype. MLLr B-ALL featured prominent cytosine hypomethylation, which was linked with MLL fusion protein binding, H3K79 dimethylation, and transcriptional upregulation, affecting a set of known and newly identified MLL fusion direct targets with oncogenic activity such as FLT3 and BCL6. Notably, BCL6 blockade or loss of function suppressed proliferation and survival of MLLr leukemia cells, suggesting BCL6-targeted therapy as a new therapeutic strategy for MLLr B-ALLs.

Significance: We conducted the first integrative epigenomic study in adult B-ALLs, as a correlative study to the ECOG E2993 phase III clinical trial. This study links for the first time the direct actions of oncogenic fusion proteins with disruption of epigenetic regulation mediated by cytosine methylation. We identify a novel clinically actionable biomarker in B-ALLs: IL2RA(CD25), which is linked with BCR–ABL1 and an inflammatory signaling network associated with chemotherapy resistance. We show that BCL6 is a novel MLL fusion protein target that is required to maintain the proliferation and survival of primary human adult MLLr cells and provide the basis for a clinical trial with BCL6 inhibitors for patients with MLLr. Cancer Discov; 2(11); 1004–23. ©2012 AACR.

Read the Commentary on this article by Cimmino and Aifantis, p. 976.

This article is highlighted in the In This Issue feature, p. 961

Adult B-cell precursor acute lymphoblastic leukemia (B-ALL) is an aggressive disease with a less than 40% long-term survival rate (1). This relatively poor outcome compared with childhood B-ALLs is partly explained by an increased frequency of high-risk molecular lesions such as BCR–ABL1 (20%–40% in adults vs. 2%–5% in children), MLL rearrangements (MLLr, 10%–20%), and E2A–PBX1 fusions (5%; refs. 1, 2). The molecular mechanisms underlying poor outcome in these adult B-ALLs are only partially understood. However, each of these B-ALL subtypes features distinct and perturbed gene expression profiles as compared with each other and to normal pre-B cells (3–6). A deeper understanding of the mechanisms driving aberrant gene expression as well as improved biomarkers and therapeutic targets are needed to improve risk stratification and therapy.

Transcriptional regulation and hence cellular phenotypes are increasingly understood to be programmed by epigenetic modifications of chromatin (7, 8). Epigenetic information is encoded, in large part, by patterns of cytosine methylation and histone modifications. An increased abundance of cytosine methylation at gene promoters, especially those containing CpG islands, is generally associated with transcriptional silencing, whereas decreased cytosine methylation may facilitate transcriptional activation. Importantly, perturbations in cytosine methylation patterning are epigenetically retained during cell division, enabling dividing cells to transmit transcriptional programming to their progeny (7, 8). This process allows cells within tissues including tumors to retain their specific phenotypes. Along these lines, DNA methylation patterning has been shown to shift during normal hematopoiesis and is believed to play an essential role in lineage specification (9). Accordingly, disruption of the function of DNA methyltransferases (DNMT) perturbs normal hematopoiesis (10, 11). By the same token, DNA methylation patterning of terminally differentiated cells must be erased for them to return to a pluripotent and self-renewal state (12, 13). Similar to normal tissues, tumors may be dependent on specific DNA methylation patterns to acquire their unique phenotypes (14). Characterization of cytosine methylation patterning in tumors may thus provide important insights into how gene expression is perturbed in different tumor types.

Identification of epigenetically modified genes may be highly informative, as deregulation of key signaling or transcriptional regulatory genes can alter entire downstream pathways and have significant biologic effects. For example, aberrant epigenetic silencing of WNT signaling suppressor genes can result in WNT pathway hyperactivity in colon cancers (15). In acute leukemia, a number of genetic lesions encode for proteins that can potentially indirectly or directly alter epigenetic regulatory states (16, 17). Hence, combined genetic, transcriptional, and epigenetic profiling studies may provide a mechanistic link between genetically altered proteins, their impact on chromatin, and their associated transcriptional profiles.

Several lines of investigation suggest that epigenomic programming is globally disrupted in B-ALLs. For example, many of the common MLL fusion proteins can recruit the DOT1L histone methyltransferase which dimethylates H3K79, an event that is associated with transcriptional activation and elongation (18–23). Infants with MLLr B-ALLs were shown to feature prominent cytosine hypermethylation of many genes and to be susceptible to DNMT inhibitors in vitro (24, 25). Other studies generally in limited numbers of B-ALLs and CpGs identified aberrantly methylated genes in B-ALLs (26–29). Differential cytosine methylation of CpG islands was noted in subtypes of childhood B-ALLs such as those with high hyperdiploidy and t(12;21) (30). To better understand the contribution of aberrant epigenetic gene regulation to the pathogenesis of adult B-ALLs, we conducted integrative genome-wide cytosine methylation and transcriptional profiling of a large cohort of adult patients with B-ALL all enrolled in a single multicenter phase III clinical trial (ECOG E2993). We focused primarily on B-ALL subtypes associated with poor outcome. When coupled with chromatin immunoprecipitation (ChIP) sequencing (ChIP-seq) of leukemic fusion proteins and histone modifications and computational and functional assays, the resulting DNA methylation profiles provided insight into mechanisms driving aberrant gene expression in adult B-ALLs as well as new biomarkers and therapeutic targets.

Specific Promoter DNA Methylation Patterning in Genetically Defined B-ALL Subtypes

We reasoned that cytosine methylation patterning would provide biologically and clinically significant information about B-ALLs in adult patients. Therefore, we conducted HELP (HpaII tiny fragment enrichment by ligation-mediated PCR) DNA methylation assays (31) on a cohort of 215 newly diagnosed adult patients with B-ALLs with available diagnostic specimens enrolled in the ECOG E2993 multicenter phase III clinical trial (Supplementary Tables S1 and S2 for detailed patient description). Eighty-three of these patients featured BCR–ABL1 translocations, 7 had E2A–PBX1 fusions, and 28 harbored MLL rearrangements. We focused our attention principally on these B-ALLs because of their defined genetic background and association with poor clinical outcome. We also profiled normal pre-B cells (CD19+ and VpreB+) isolated from the bone marrows of 12 healthy adults as normal counterpart for our B-ALL tumor samples (Supplementary Table S2). HELP was conducted using a customized microarray design covering more than 50,000 CpGs annotated to 14,000 gene promoters and our standard quality control and normalization algorithms (31–33). We conducted technical validation using base pair resolution quantitative DNA methylation assessment by MassArray EpiTyper to confirm the accuracy of methylation values derived from HELP arrays. This procedure indicated that our HELP-derived DNA methylation values were highly concordant (correlation coefficient [r] = 0.87, Supplementary Fig. S1 and Supplementary Table S3) and that based on the regression line in Supplementary Fig. S1, a log ratio difference (dx) of 1 or 1.5 from HELP data corresponds to a methylation difference of 20% or 30%, respectively, by MassArray. Consistent with previous reports (14, 33), these data confirm that HELP allows DNA methylation to be accurately assessed as a continuous variable.

We next conducted a supervised analysis comparing each of these 3 cytogenetically defined B-ALL subtypes to normal pre-B cells. This procedure identified a total of 1,740 probesets differentially methylated in at least 1 of the 3 B-ALL subtypes as compared with pre-B cell controls with DNA methylation difference >30% (dx > 1.5 by HELP) and adjusted P < 0.01 [ANOVA followed by Dunnett's test; (refs. 33, 34) and Benjamini–Hochberg correction (ref. 35)]. These probesets corresponded to 1,493 gene promoters, with a general trend toward genes being hypomethylated in these B-ALL subtypes compared with normal pre-B cells (Fig. 1A; Supplementary Table S4). Among hypomethylated probesets, 13.7% overlap with CpG islands, 55.3% with CpG shores (the 2 kb flanking CpG islands; ref. 36), and 31% were outside of CpG islands or shores. B-ALL DNA methylation signatures were queried for their association with gene sets relevant to lymphoid tumor biology (37). All 3 B-ALL subtypes were significantly enriched with MYC target genes (P < 1e-3, P < 1e-11, and P < 1e-7, respectively, for BCR–ABL1, E2A–PBX1, and MLLr; Fisher exact test with Benjamini–Hochberg correction; Fig. 1B; Supplementary Tables S5 and S6), suggesting that MYC transcriptional networks are disrupted epigenetically in all 3 leukemia subtypes consistent with the critical role of MYC in hematopoietic neoplasms. BCR–ABL1 and MLLr DNA methylation signatures were enriched in BCL6 target genes (P = 0.002 and P = 0.049, respectively). BCR–ABL1 B-ALLs also featured epigenetic deregulation of p53-induced genes (P = 0.012), interleukin (IL)10-responsive genes (P = 0.013), cell-cycle genes (P = 0.022), and others. E2A–PBX1 methylation signature involved gene sets associated with glutamine and glucose starvation (P = 0.021), SREBP-1 and 2 activity (P = 0.016), and T-cell cytokines (P = 0.042; Fig. 1B; Supplementary Tables S5 and S6). Detailed explanation and references for those signatures are provided in Supplementary Table S5. A similar analysis using Gene Ontology (GO) terms enriched for signal transduction, cell proliferation, and gene regulation in all 3 subtypes (Supplementary Tables S5 and S6). These data indicate that cytogenetically distinct forms of adult B-ALLs display specific and functionally defined DNA methylation signatures as compared with normal pre-B cells.

Figure 1.

B-ALL subtypes display aberrant DNA methylation patterns as compared with normal pre-B cells. A, heatmap representation of promoter DNA methylation levels of differentially methylated genes in BCR–ABL1 (n = 83), E2A–PBX1 (n = 7), and MLLr (n = 28) B-ALLs versus normal pre-B samples (n= 12). Probesets (1,740; 1,493 gene promoters) were identified with adjusted P < 0.01 and methylation difference >30% (i.e., dx > 1.5 on HELP). Each row of the heatmap represents one probe set on the HELP array, and each column represents a patient with B-ALL or a healthy bone marrow. The color scale bar represents the methylation levels with purple for hypomethylation and yellow for hypermethylation. B, the heatmap represents enrichment of B-cell and lymphoid-specific gene sets with the DNA methylation signatures of the 3 B-ALL subtypes versus normal pre-B cells. The statistical significance is provided in the color key. P values were calculated by Fisher exact test with Benjamini–Hochberg correction.

Figure 1.

B-ALL subtypes display aberrant DNA methylation patterns as compared with normal pre-B cells. A, heatmap representation of promoter DNA methylation levels of differentially methylated genes in BCR–ABL1 (n = 83), E2A–PBX1 (n = 7), and MLLr (n = 28) B-ALLs versus normal pre-B samples (n= 12). Probesets (1,740; 1,493 gene promoters) were identified with adjusted P < 0.01 and methylation difference >30% (i.e., dx > 1.5 on HELP). Each row of the heatmap represents one probe set on the HELP array, and each column represents a patient with B-ALL or a healthy bone marrow. The color scale bar represents the methylation levels with purple for hypomethylation and yellow for hypermethylation. B, the heatmap represents enrichment of B-cell and lymphoid-specific gene sets with the DNA methylation signatures of the 3 B-ALL subtypes versus normal pre-B cells. The statistical significance is provided in the color key. P values were calculated by Fisher exact test with Benjamini–Hochberg correction.

Close modal

BCR–ABL1 B-ALL DNA Methylation Signature Features Dysregulation of IL2RA

To investigate epigenetic programming linked to BCR–ABL1 signaling, we sought to identify genes differentially methylated in BCR–ABL1-positive ALLs. We conducted a supervised analysis comparing DNA methylation profiles of the 83 BCR–ABL1-positive to the 132 BCR–ABL1-negative patients using the Student t test with Benjamini–Hochberg correction for multiple testing. This yielded a set of 192 probesets (170 unique gene promoters) with methylation difference >20% (dx > 1 by HELP) and adjusted P < 0.01 (Fig. 2A; Supplementary Table S7A). Gene expression profiling was also conducted in these E2993 patients to define potential links between DNA methylation and transcriptional regulation. Supervised analysis of gene expression profiling in the same patients yielded 605 probesets (417 unique genes) with an adjusted P < 0.01 and log2 fold change >1 (Fig. 2B; Supplementary Table S7B). There was a tendency for DNA methylation to be inversely correlated with gene expression (i.e., the average expression fold changes (log2 fold) of BCR–ABL1-positive versus -negative ALLs in the hypomethylated signature genes is higher than that in the hypermethylated signature genes, P = 0.013, one-tailed Wilcoxon test). More than 80% of the genes in the methylation signature were hypomethylated and 80% of the genes were overexpressed in BCR–ABL1-positive B-ALLs, suggesting strong gene activation functions downstream of the fusion protein. Using a Bayesian predictor (38) and leave-one-out cross-validation (LOOCV), the methylation and expression signatures re-classified 87% and 93%, respectively, of these B-ALL cases according to BCR–ABL1 status (Fig. 2A and B). To identify a core set of highly dysregulated genes, we examined the overlap between the DNA methylation and gene expression signatures. This analysis yielded 13 genes, 11 of which featured inversely correlated methylation and expression levels (Fig. 2C). Among the 11 core signature genes, the IL2RA gene (interleukin receptor 2 alpha chain), which encodes the CD25 protein, featured the highest overexpression and the second highest hypomethylation in BCR–ABL1 ALLs (Fig. 2C). Differential expression of some of these genes were also captured in a previous gene expression profiling effort conducted in 54 E2993 patients with B-ALLs (6). Differential expression and methylation was confirmed for all 11 genes by quantitative PCR (qPCR) and MASSArry EpiTyper (Supplementary Fig. S2A and S2B and Supplementary Tables S8 and S9). The DNA methylation and gene expression microarray data are available at GEO database (SuperSeries: GSE34941; gene expression: GSE34861; DNA methylation: GSE34937).

Figure 2.

BCR–ABL1 B-ALLs feature DNA methylation and expression signatures centered aroundIL2RA(CD25). Methylation signature (A) and expression signature (B) of BCR–ABL1-positive versus BCR–ABL1-negative B-ALLs. Each row represents one probeset and each column one patient. The color scale bars represent the methylation or expression levels. The probability of signatures to reclassify cases according to BCR–ABL1 status is shown on the top. The actual case labels: red, BCR–ABL1-positive; blue, BCR–ABL1-negative. The numbers of correctly classified and misclassified cases are given below the heatmap. C, the difference of mean methylation or expression values in BCR–ABL1-positive versus -negative B-ALL is depicted by the color scale on the 11 core genes overlapping between methylation and expression signatures. D, the gene network most highly enriched by methylation and expression signatures in BCR–ABL1 B-ALLs is centered around IL2RA (CD25). E, the degree of IL2RA hypomethylation and expression from E2993 cases and from an independent cohort of 132 patients with St. Jude B-ALLs (4, 39) in BCR–ABL1-positive versus BCR–ABL1-negative and versus normal pre-B cells. MassArray (F) and qPCR (G) validation for IL2RA (CD25). The color key represents % methylation of each promoter CpG (rows). Columns represent individual cases. Location of HELP probesets and EpiTyper primers on IL2RA are shown (UCSC genome browser hg19). GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Figure 2.

BCR–ABL1 B-ALLs feature DNA methylation and expression signatures centered aroundIL2RA(CD25). Methylation signature (A) and expression signature (B) of BCR–ABL1-positive versus BCR–ABL1-negative B-ALLs. Each row represents one probeset and each column one patient. The color scale bars represent the methylation or expression levels. The probability of signatures to reclassify cases according to BCR–ABL1 status is shown on the top. The actual case labels: red, BCR–ABL1-positive; blue, BCR–ABL1-negative. The numbers of correctly classified and misclassified cases are given below the heatmap. C, the difference of mean methylation or expression values in BCR–ABL1-positive versus -negative B-ALL is depicted by the color scale on the 11 core genes overlapping between methylation and expression signatures. D, the gene network most highly enriched by methylation and expression signatures in BCR–ABL1 B-ALLs is centered around IL2RA (CD25). E, the degree of IL2RA hypomethylation and expression from E2993 cases and from an independent cohort of 132 patients with St. Jude B-ALLs (4, 39) in BCR–ABL1-positive versus BCR–ABL1-negative and versus normal pre-B cells. MassArray (F) and qPCR (G) validation for IL2RA (CD25). The color key represents % methylation of each promoter CpG (rows). Columns represent individual cases. Location of HELP probesets and EpiTyper primers on IL2RA are shown (UCSC genome browser hg19). GAPDH, glyceraldehyde-3-phosphate dehydrogenase.

Close modal

Pathway analysis combining the full DNA methylation and gene expression signatures from Fig. 2A and B identified the heavy involvement of one particular gene network, centered around IL2RA(CD25) and other cytokines (Fig. 2D). Collectively, these analyses pointed toward IL2RA as a potentially significant gene in BCR–ABL1-positive B-ALLs. Indeed, we observed that that IL2RA is significantly more hypomethylated in BCR–ABL1-positive B-ALLs not only versus BCR–ABL1-negative B-ALLs but also versus normal pre-B cells (P < 1e-6 for both, t test; Fig. 2E, top). IL2RA mRNA in gene expression profiling of the same patients was also significantly more abundant in BCR–ABL1-positive versus BCR–ABL1-negative B-ALLs or normal pre-B cells (Fig. 2E, middle). The same differential expression of IL2RA was detected upon examining an independent set of 132 B-ALL gene expression profiles (refs. 4, 39; Fig. 2E, bottom). Significant hypomethylation and upregulation of IL2RA in BCR–ABL1-positive ALLs was validated by MassArray EpiTyper and qPCR on randomly selected E2993 patient samples with available DNA (P < 1e-6 for all, one-tailed t-test; Fig. 2F and G).

CD25 Positivity Confers Poor Clinical Outcomes

BCR–ABL1 translocation is an indicator of unfavorable outcome in adults with B-ALLs (2). Examination of a total of 465 patients with B-ALLs enrolled in the E2993 trial including those profiled with microarrays plus additional patients with complete molecular and clinical annotation [median age, 37 years, interquartile range (IQR), 25–47 years; median follow-up, 7.9 years; IQR, 5–11 years] confirmed the inferior overall survival (OS) of BCR–ABL1-positive (n = 113) versus BCR–ABL1-negative B-ALL cases (n = 352, log-rank, P < 0.0001; Fig. 3A). The subset of imatinib-treated patients with B-ALLs was excluded from the survival analysis. Consistent with a previous report (40), CD25 positivity (measured by flow cytometry) was strongly associated with BCR–ABL1 (61.1% BCR–ABL1-positive patients are also CD25 positive, 69 of 113) but was rare in BCR–ABL1-negative patients (only 5.1% BCR–ABL1-negative patients are CD25 positive, 18 of 352). This difference was highly significant (P < 1e-6, Fisher exact test; Supplementary Fig. S3). When comparing expression profiles of CD25-positive versus CD25-negative cases among BCR–ABL1-positive B-ALLs, IL2RA was the most differentially expressed gene (>6-fold overexpressed in CD25-positive cases, P = 0.005, t test) and also notably hypomethylated (P = 0.01, t test; Supplementary Fig. S4A). The overexpression and hypomethylation of IL2RA was validated by qPCR and MassArray (Supplementary Fig. S4B). Those data suggest that IL2RA (CD25) might be an important factor in BCR–ABL1-positive ALLs. Indeed CD25 expression was clinically relevant, as BCR–ABL1-positive patients can be further stratified into 2 groups with significantly different clinical outcome according to CD25 positivity. CD25-positive patients have significantly worse OS [median OS time, 0.8 years; 95% confidence interval (CI), 0.6–1.1, n = 69] than the CD25-negative patients (median OS time, 1.3 years; 95% CI, 0.90–2.5, n = 44, log-rank, P = 0.0064; Fig. 3B). Although the CD25-negative BCR–ABL1-positive patients tended to display worse OS than the CD25-negative BCR–ABL1-negative patients, the difference did not reach statistical significance (log-rank, P = 0.08; Fig. 3C). Collectively, the data indicate that hypomethylation and corresponding overexpression of CD25 is strongly associated with BCR–ABL1 fusion protein and may be linked with processes that account, at least in part, for a particularly unfavorable clinical outcome in BCR–ABL1 B-ALLs. CD25 can serve as a new prognostic biomarker for risk stratification in B-ALLs.

Figure 3.

CD25 expression is associated with unfavorable clinical outcome. A, BCR–ABL1-positive B-ALL (n = 113, median OS, 1.0 years; 95% CI, 0.8–1.2) have a worse OS than the BCR–ABL1-negative B-ALLs (n = 352, median OS, 2.4 years; 95% CI, 1.8–3.8) in the E2993 cases (P < 0.0001, log-rank test). B, CD25 positivity further stratifies BCR–ABL1-positive B-ALLs into 2 groups with significantly different clinical outcome: CD25-positive patients have significantly inferior OS (n = 69, median OS, 0.8 years; 95% CI, 0.6–1.1) than the CD25-negative patients (n = 44, median OS, 1.3 years; 95% CI, 0.9–2.5; P = 0.0064, log-rank test). C, the CD25-negative BCR–ABL1-positive patients (n = 44) did not have significantly different OS than the BCR–ABL1-negative CD25-negative patients (n = 334, P = 0.08, log-rank test).

Figure 3.

CD25 expression is associated with unfavorable clinical outcome. A, BCR–ABL1-positive B-ALL (n = 113, median OS, 1.0 years; 95% CI, 0.8–1.2) have a worse OS than the BCR–ABL1-negative B-ALLs (n = 352, median OS, 2.4 years; 95% CI, 1.8–3.8) in the E2993 cases (P < 0.0001, log-rank test). B, CD25 positivity further stratifies BCR–ABL1-positive B-ALLs into 2 groups with significantly different clinical outcome: CD25-positive patients have significantly inferior OS (n = 69, median OS, 0.8 years; 95% CI, 0.6–1.1) than the CD25-negative patients (n = 44, median OS, 1.3 years; 95% CI, 0.9–2.5; P = 0.0064, log-rank test). C, the CD25-negative BCR–ABL1-positive patients (n = 44) did not have significantly different OS than the BCR–ABL1-negative CD25-negative patients (n = 334, P = 0.08, log-rank test).

Close modal

Deregulated DNA Methylation Is Associated with Binding of the E2A–PBX1 Fusion Protein

We next examined whether E2A–PBX1 translocations were also associated with specific perturbations of DNA methylation patterning. Supervised analysis comparing 7 E2A–PBX1-positive versus 97 E2A–PBX1-negative ALLs identified 94 differentially methylated probesets (69 genes) with adjusted P < 0.01 and methylation difference more than 20% (Fig. 4A; Supplementary Table S10A) and 1,312 differentially expressed probesets (863 genes) with adjusted P < 0.01 and log2 fold change >1 (Fig. 4B; Supplementary Table S10B). Most genes in the methylation signatures (58 of 69) were hypomethylated, and the hypomethylation was associated with overexpression [P < 0.01, one-tailed Wilcoxon test by comparing the means of the log2 expression fold changes (E2A–PBX1-positive vs. -negative cases) in the hypo- and hypermethylated gene groups]. When applied to a Bayesian predictor (38) using LOOCV, the DNA methylation and gene expression signatures reclassified E2A–PBX1 ALLs with 100% and 98% accuracy, respectively. Eight genes overlapped between methylation and expression signatures and 7 showed inverse correlation, including MLKL, SERINC5, and CTDSP2 (hypermethylated and underexpressed) and CALD1, ARL4C, ST6GALNAC3, and EXTL3 (hypomethylated and overexpressed; Fig. 4C). E2A–PBX1 is reported to act as an aberrant transcriptional activator that cooperates with the p300 histone acetyltransferase (41). We wondered whether the fusion protein itself might be linked to hypomethylation in these B-ALLs. We therefore conducted ChIP-seq using antibodies recognizing the N terminus of E2A (E2AN), the C terminus of PBX1 (PBX1C), as well as p300 in an E2A–PBX1-positive B-ALL cell line, 697 cells (Fig. 4D; Supplementary Table S11). Global analysis of these datasets identified 3,358 E2AN peaks, located predominantly at intronic and upstream regulatory regions (83%: broken down into 41% intronic, 23% distal at 2–50 kb upstream, and 19% intergenic >50 kb from RefSeq TSS) versus only 11% at promoters (ChIPseeqer software; ref. 42). The 19,108 PBX1C peaks were located at a much higher frequency at promoters (36%) but were also at putative regulatory regions (28% introns, 17% distal, and 13% intergenic; Supplementary Fig. S5A and Supplementary Table S14). Much of the intergenic/intronic PBX1C sites may be accounted for by E2A–PBX1 fusion protein, as there was an 83% overlap between E2AN sites with PBX1C sites (2,971 peaks, z-score: 741.57, P < 1e-10, ChIPseeqer software; ref. 42; Supplementary Fig. S5B). p300 ChIP-seq identified 15,501 binding sites, which similar to E2AN were preferentially to regulatory regions (85%, with 43% in introns, 22% distal, and 20% intergenic). Remarkably, 96% of the putative E2A-PBX1 fusion protein sites overlapped with p300 (2,684 peaks, z-score: 960.58, P < 1e-10, ChIPseeqer software; ref. 42), consistent with the known complex formation between these proteins (Supplementary Fig. S5B). Fully 87% of these joint binding sites were located at intronic/distal and intergenic regions and only 8% at promoters (Supplementary Fig. S5C). Altogether the data place a minority of E2A–PBX1–p300 complexes at traditional promoters and the majority at both intra- and intergenic regulatory regions.

Figure 4.

Specific DNA methylation and expression signatures associated with binding of E2A–PBX1 fusion protein. Methylation signature (A), expression signature (B), and overlapping (C) between methylation and expression signatures of E2A–PBX1-positive versus E2A–PBX1-negative B-ALL. Rows are probesets and columns are patients. The color scale bars represent the methylation or expression levels. The probability of signatures to reclassify E2A–PBX1 cases is shown on the top. C, the difference of mean methylation and expression values in E2A–PBX1-positive versus E2A–PBX1-negative B-ALLs is depicted by the color scale. D, graphical representations of ChIP for E2A (orange), PBX1 (green), and E2A–PBX1 fusion proteins. ChIP-seq tracks (E) and qChIP (F) results for E2AN (green), PBX1C (orange), and p300 (blue) antibodies versus input (black) in 697 cells on the 4 hypomethylation signature genes in (C). E, y-axis: number of reads for peak summits normalized by the total number of reads per track (set to 1 Gb for each track). Locations of CpG islands, HELP probesets, qChIP primers, and an enlarged view of binding peaks are shown. G, qPCR was carried out in 697 cells transfected with E2A-shRNA or scramble-shRNA. y-axis, relative expression normalized to the scramble-shRNA. Data represent means ± SEM (n = 3).

Figure 4.

Specific DNA methylation and expression signatures associated with binding of E2A–PBX1 fusion protein. Methylation signature (A), expression signature (B), and overlapping (C) between methylation and expression signatures of E2A–PBX1-positive versus E2A–PBX1-negative B-ALL. Rows are probesets and columns are patients. The color scale bars represent the methylation or expression levels. The probability of signatures to reclassify E2A–PBX1 cases is shown on the top. C, the difference of mean methylation and expression values in E2A–PBX1-positive versus E2A–PBX1-negative B-ALLs is depicted by the color scale. D, graphical representations of ChIP for E2A (orange), PBX1 (green), and E2A–PBX1 fusion proteins. ChIP-seq tracks (E) and qChIP (F) results for E2AN (green), PBX1C (orange), and p300 (blue) antibodies versus input (black) in 697 cells on the 4 hypomethylation signature genes in (C). E, y-axis: number of reads for peak summits normalized by the total number of reads per track (set to 1 Gb for each track). Locations of CpG islands, HELP probesets, qChIP primers, and an enlarged view of binding peaks are shown. G, qPCR was carried out in 697 cells transfected with E2A-shRNA or scramble-shRNA. y-axis, relative expression normalized to the scramble-shRNA. Data represent means ± SEM (n = 3).

Close modal

Although promoter sites were the minority of fusion protein binding sites, we focused on these sites to more firmly link them to our studies on promoter DNA methylation. Focusing down on E2A–PBX1 signature genes, all 4 hypomethylated and upregulated genes in the core signature (Fig. 4C) showed enriched binding of E2AN, PBX1C, and p300 in their promoter regions (Fig. 4E). Ten additional hypomethylation signature genes were enriched by E2AN, PBX1C, and p300 antibodies, including ACOXL, BLK, COMT, MAST4, MYBPH, SLAMF1, SMPD3, TERT, TNS4, and WAPAL. A total of 14 genes overlap between the E2A–PBX1 hypomethylation signature (n = 58 genes) and those identified as putative E2A–PBX1 and p300 direct targets (n = 2,684 overlapping peaks annotated to n = 1,642 genes). This level of overlap was indicative of overrepresentation of E2A/PBX1 target genes among hypomethylated genes (P = 0.006, 2-tailed χ2 test with Yates' correction). We further validated these targets by conducting qChIP with E2AN, PBX1C, and p300 antibodies (as positive ChIP's) and PBX1N antibody (as a negative control ChIP) in 697 and Nalm6 cells (Nalm6 is an E2A–PBX1-negative B-ALL cell line). PBX1N qChIP serves as a negative control, as PBX1N binds to only wild-type PBX1 but not E2A–PBX1 fusion and wild-type PBX1 is not expressed in lymphoid precursor B cells (43), therefore we would expect no PBX1N binding enrichment on those loci in neither 697 nor Nalm6 cells. This procedure confirmed binding enrichment of E2AN, PBX1C, and p300 to the 4 core hypomethylation signature genes in 697 but not Nalm6 cells and no binding enrichment of PBX1N to those loci in either cell type (Fig. 4F; Supplementary Table S12). To determine whether these genes are directly controlled by E2A–PBX1, we conducted lentiviral short hairpin RNA (shRNA) knockdown in 697 cells which targeted the middle coding region (amino acids 340–348) of E2A (E12/E47) and therefore knockdown both wild-type E2A and E2A–PBX1 fusion. Figure 4G shows the shRNA effectively knocked down E2A, E2A–PBX1, as well as the 4 core signature genes. Collectively, these data identify a set of genes specifically deregulated epigenetically by E2A–PBX1 fusion, and moreover, directly implicate the fusion protein as driving this epigenetic and transcriptional signature.

Adult MLLr B-ALL Displays a Unique DNA Methylation Signature with Predominant Hypomethylation

Chromosomal translocations involving MLL are frequent in B-ALLs, can involve more than 50 partner genes, and are generally associated with a poor clinical outcome (44). Among the 28 MLLr cases in our cohort, there were 20 MLL–AF4, 6 MLL–ENL, 1 MLL–AF9, and 1 MLL–EPS15 cases. Unsupervised analysis of DNA methylation and gene expression profiles using hierarchical clustering showed that MLLr cases do not segregate according to fusion partners (data not shown). Moreover, a supervised analysis between the 20 MLL–AF4 and the other 8 MLLr cases showed no genes differentially methylated or expressed with adjusted P < 0.01 (data not shown). Many of the common MLL fusion partners such as AF9, ENL, and AF4 form part of transcriptional elongation complexes and directly or indirectly recruit the DOT1L histone 3 lysine 79 (H3K79) methyltransferase (21, 23). This biochemical commonality may explain why these MLLr DNA methylation profiles are highly similar to each other. Supervised analysis comparing the 28 MLLr versus the other 97 MLLr-negative ALLs identified 430 differentially methylated probesets (379 genes) with adjusted P < 0.01 and methylation difference > 20% (Fig. 5A; Supplementary Table S13A) and 1,581 differentially expressed probesets (1,030 genes) with adjusted P < 0.01 and log2 fold change >1 (Fig. 5B; Supplementary Table S13B). Sixty percent (231 of 379) differentially methylated genes were hypomethylated and displayed higher expression levels than the hypermethylated genes in MLLr B-ALL (P < 1e-6, one-tailed Wilcoxon test by comparing the means of the log2 expression fold changes of MLLr-positive vs. -negative cases in the hypo- and hypermethylated gene groups). When applying a Bayesian predictor (38) to these cases, the methylation and expression signatures each reclassified MLLr cases with 97% accuracy, (Fig. 5A and B). A core set of 32 genes overlapped between the methylation and expression signatures, among which 25 showed inverse correlation between methylation and expression. Those genes include ITGAE, IGF2BP2 (IMP-2), PCDHGB3, PCDHGA5, FUT4, PARP8, MAP7, RBKS, FLT3, ANXA5, MRPL33, TCP11, MAP1A, and ACSL1 as hypomethylated and overexpressed and PRKCH, LTB, GAB1, AFF3, LST1, ZAP70, CCR6, ITIH3, MLXIP, NINJ1, and PIK3IP1 as hypermethylated and downregulated in MLLr B-ALLs (Fig. 5C). MassArray EpiTyper and qPCR confirmed the differential methylation and expression on 14 randomly selected signature genes from this list (Supplementary Fig. S6A and B). One of the core genes was FLT3 (hypomethylated and overexpressed in MLLr B-ALLs, Fig. 5D and E), which encodes a class III receptor tyrosine kinase that plays key roles in proliferation of hematopoietic cells (45). FLT3 was previously shown to be highly expressed in MLLr leukemias and a potential therapeutic target in this B-ALL subtype (46, 47). Aberrant DNA hypomethylation may thus contribute to or facilitate increased expression of FLT3 in MLLr B-ALLs.

Figure 5.

DNA methylation and expression signatures are linked to MLL fusion protein binding patterns. Methylation signature (A), expression signature (B), and overlapping genes between methylation and expression signatures (C) of MLLr versus non-MLLr B-ALLs. Rows for probesets and columns for patients. The probability of a Bayesian predictor to reclassify MLLr ALLs is shown on top. The numbers of correctly classified and misclassified cases are given below the heatmap. C, the color scale depicts the difference of mean methylation or expression values in MLLr versus non-MLLr cases. qPCR (D) and MassArray (E) validation of FLT3 in 5 MLLr, 5 non-MLLr B-ALLs, and 5 normal pre-B samples. Data present means ± SEM (n = 3). P values from one-tailed t-test. The heatmap represents the % methylation of CpGs (rows) in samples (columns). GAPDH, glyceraldehyde-3-phosphate dehydrogenase. F, graphical representation of ChIP for MLL (red), AF4 (blue), and MLL-AF4 fusion. G, qChIP using MLLN, AF4C, H3K79me2, or H3 antibodies in RS4;11 and CCRF-CEM cells, enriching for 4 positive control and 7 MLLr hypomethylated and overexpressed loci. Data represent means ± SEM (n = 3). H, FLT3 locus ChIP-seq enrichment profiles in RS4;11 cells. Locations of TSS, CpG islands, HELP probesets, and MassArray EpiTyper primers are shown below the gene model in the UCSC genome browser view.

Figure 5.

DNA methylation and expression signatures are linked to MLL fusion protein binding patterns. Methylation signature (A), expression signature (B), and overlapping genes between methylation and expression signatures (C) of MLLr versus non-MLLr B-ALLs. Rows for probesets and columns for patients. The probability of a Bayesian predictor to reclassify MLLr ALLs is shown on top. The numbers of correctly classified and misclassified cases are given below the heatmap. C, the color scale depicts the difference of mean methylation or expression values in MLLr versus non-MLLr cases. qPCR (D) and MassArray (E) validation of FLT3 in 5 MLLr, 5 non-MLLr B-ALLs, and 5 normal pre-B samples. Data present means ± SEM (n = 3). P values from one-tailed t-test. The heatmap represents the % methylation of CpGs (rows) in samples (columns). GAPDH, glyceraldehyde-3-phosphate dehydrogenase. F, graphical representation of ChIP for MLL (red), AF4 (blue), and MLL-AF4 fusion. G, qChIP using MLLN, AF4C, H3K79me2, or H3 antibodies in RS4;11 and CCRF-CEM cells, enriching for 4 positive control and 7 MLLr hypomethylated and overexpressed loci. Data represent means ± SEM (n = 3). H, FLT3 locus ChIP-seq enrichment profiles in RS4;11 cells. Locations of TSS, CpG islands, HELP probesets, and MassArray EpiTyper primers are shown below the gene model in the UCSC genome browser view.

Close modal

MLL-AF4 Binding Is Associated with DNA Hypomethylation and Overexpression of Genes, Several with Known or Suspected Oncogenic Functions

Because MLL fusion proteins are chromatin modifiers and transcriptional activators, we hypothesized that genes aberrantly hypomethylated and highly expressed in MLLr cases might be direct targets of MLL fusions. To determine whether this is the case, we conducted qChIP using MLL N-terminal antibody (MLLN), AF4 C-terminal antibody (AF4C), H3K79me2 antibody (as a chromatin mark induced by MLL fusions through recruitment of DOT1L), and histone H3 antibody (as a negative control) in a MLL-AF4–positive B-ALL cell line RS4;11 and an MLL-wild-type ALL cell line CCRF-CEM (Fig. 5F; Supplementary Table S11). We assessed enrichment at promoters of (i) 4 positive control genes: HOXA7, HOXA9, HOXA10, and MEIS1 (Fig. 5G; Supplementary Figs. S7 and S8A, and S8B); (ii) 4 genes from the core signature in Fig. 5B: FLT3, IGF2BP2/IMP-2, FUT4 and ITGAE; (iii) 2 randomly selected genes from among the top 100 most highly overexpressed genes (IGFBP7 and VAT1L/KIAA1576), which were also hypomethylated although did not reach the threshold for the signature in Fig. 5A: IGFBP7 (dx = 1.18, P = 0.091) and VAT1L (dx = 0.94, P = 0.0002); and (iv) BCL6, which was of interest because BCL6 target genes were one of the few pathways differentially methylated in patients with MLLr (Fig. 1B), and BCL6 was differentially methylated in MLLr B-ALLs versus normal pre-B cells in Fig. 1A; Supplementary Table S4 (dx = 1.59, P = 0.00001). We observed enrichment for MLLN and AF4C at all 7 of these loci and the 4 positive controls in RS4;11 but not CCRF-CEM cells, consistent with binding of the MLL-AF4 fusion protein (Fig. 5G). In accordance with this notion, we also noted increased H3K79me2 signal in RS4;11 but not CCRF-CEM cells at 9 of those 11 loci (except at ITGAE and VAT1L), and histone 3 enrichment was relatively even in both cell lines (Fig. 5G). Similar enrichment for MLLN and AF4C was observed in a second MLL-AF4–positive B-ALL cell line, SEM, at all those 11 loci (Supplementary Fig. S8). We next conducted ChIP-seq with MLLN, AF4C, and H3K79me2 antibodies in RS4;11 cells (GEO accession numbers: GSE38403). MLLN ChIP-seq identified 3,312 binding peaks, annotated to 1,928 promoters (promoter regions defined as 2 kb up- and downstream to TSS); AF4C ChIP-seq identified 6,253 peaks (1,387 promoters); and H3K79me2 ChIP-seq identified 28,476 peaks (5,395 promoters, Supplementary Tables S14 and S15). 603 gene promoters showed binding overlap with all 3 antibodies (MLLN, AF4C, and H3K79me2) and were hence identified as MLL fusion direct targets by ChIP-seq data (Supplementary Table S15). These 603 MLL fusion target genes showed higher levels of hypomethyation and expression than the nontarget genes in the MLLr B-ALL cases (P < 1e-6 for both, one-tailed Wilcoxon test; Supplementary Fig. S9A and S9B), indicating MLL-AF4 binding is associated with DNA hypomethylation and overexpression of the target genes. Nine of the 11 loci queried by QChIP were also in this 603 gene list (except ITGAE and VAT1L). ChIP-seq read densities at the FLT3 locus are shown as an example in Fig. 5H. Our results showed that the hypomethylation signatures are linked with MLL fusion protein binding and H3K79me2, which induce overexpression of these genes. These are first examples of DNA methylation signatures directly mechanistically linked to binding patterns of aberrant fusion oncoproteins. Our data also revealed that FLT3 is in fact a direct target of MLL fusion proteins in B-ALLs, and consequent hypomethylation and H3K79 dimethylation of the FLT3 promoter in these tumors may contribute to direct transcriptional activation of this oncogenic tyrosine kinase.

BCL6 Is a Transcriptional Target of MLL Fusion Proteins

Among novel MLLr fusion targets identified in this study, BCL6 captured our particular attention because of its known role as a potent oncoprotein in B-cell lymphomas. As noted above, direct binding of MLL-AF4 as well as increased H3K79me2 was observed at the BCL6 locus by qChIP (Fig. 5G). ChIP-seq by MLLN, AF4C, and H3K79me2 antibodies also confirmed enriched binding at the BCL6 locus (Fig. 6A). We showed that hypomethylation of BCL6 is indeed an aberrant feature of MLLr by comparing the degree of cytosine methylation in MLLr B-ALLs with normal pre-B cells as captured by our HELP assays (P < 1e-6, t test; Fig. 6B). Hypomethylation of BCL6 in MLLr versus non-MLLr B-ALLs and normal pre-B cells was again confirmed by MassArray EpiTyper (P < 1e-3 and P < 1e-5, respectively, t test; Fig. 6C). Immunoblots using BCL6 antibody in a panel of 9 B-ALL cell lines, 11 primary B-ALL patient specimens, and 2 normal pre-B samples revealed that expression of BCL6 protein was elevated in MLLr as compared with other B-ALL subtypes and normal pre-B cells (Fig. 6D). BCL6 mRNA abundance was largely consistent with protein in these cells (Supplementary Fig. S10). Finally, to determine whether MLL fusion protein not only binds but also induces BCL6 expression, we conducted MLL-AF4 siRNA knockdown using fusion transcript–specific siRNA sequences (48) in RS4;11 and SEM cells. The siRNAs effectively downregulated MLL-AF4 transcripts and proteins within 48 hours, and this was accompanied by a reduction in expression of HOXA9 (a known MLL fusion target) and BCL6 at both mRNA and protein levels (Fig. 6E and F). MLL fusion proteins thus bind to the BCL6 promoter and contribute to its upregulation in MLLr B-ALLs.

Figure 6.

BCL6 is upregulated in MLL-AF4 ALL with its promoter hypomethylated and bound by MLL-AF4 fusion. A, ChIP-seq profiles at BCL6 enriched by MLLN (red), AF4C (green), and H3K79me2 (blue) antibodies versus input (black) in RS4;11 cells. The locations of CpG islands, HELP probesets, and MassArray EpiTyper primers are shown. B, DNA methylation values on BCL6 derived from HELP assays in MLLr (n = 28) versus normal pre-B (n = 12) samples. P value from t test. C, heatmap representation of CpG methylation (rows) at the BCL6 locus measured by MassArray on 10 MLL-AF4–positive, 10 MLL-AF4–negative, and 10 normal pre-B samples (columns). The color key indicates percent DNA methylation. P values from t test comparing the average methylation level of CpGs in each group. D, BCL6 Western blot analyses conducted in ALL cell lines (black labels) and primary specimens (blue labels). E and F, RS4;11 and SEM cells were transduced with MLL-AF4 fusion–specific siRNA as described (48) or nontargeted control siRNA, followed 48 hours later by (E) qPCR or (F) immunoblots for MLL-AF4, HOXA9, and BCL6. E, y-axis represents the relative expression of MLL-AF4 siRNA knockdown normalized to the control siRNA. Data represent means ± SEM from 6 independent knockdown experiments. F, RS4;11 Western blot bands were quantified relative to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) using ImageJ (80). *Values from densitometry and normalized to GAPDH and control siRNA.

Figure 6.

BCL6 is upregulated in MLL-AF4 ALL with its promoter hypomethylated and bound by MLL-AF4 fusion. A, ChIP-seq profiles at BCL6 enriched by MLLN (red), AF4C (green), and H3K79me2 (blue) antibodies versus input (black) in RS4;11 cells. The locations of CpG islands, HELP probesets, and MassArray EpiTyper primers are shown. B, DNA methylation values on BCL6 derived from HELP assays in MLLr (n = 28) versus normal pre-B (n = 12) samples. P value from t test. C, heatmap representation of CpG methylation (rows) at the BCL6 locus measured by MassArray on 10 MLL-AF4–positive, 10 MLL-AF4–negative, and 10 normal pre-B samples (columns). The color key indicates percent DNA methylation. P values from t test comparing the average methylation level of CpGs in each group. D, BCL6 Western blot analyses conducted in ALL cell lines (black labels) and primary specimens (blue labels). E and F, RS4;11 and SEM cells were transduced with MLL-AF4 fusion–specific siRNA as described (48) or nontargeted control siRNA, followed 48 hours later by (E) qPCR or (F) immunoblots for MLL-AF4, HOXA9, and BCL6. E, y-axis represents the relative expression of MLL-AF4 siRNA knockdown normalized to the control siRNA. Data represent means ± SEM from 6 independent knockdown experiments. F, RS4;11 Western blot bands were quantified relative to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) using ImageJ (80). *Values from densitometry and normalized to GAPDH and control siRNA.

Close modal

BCL6 Is a Therapeutic Target in MLLr B-ALL

Constitutive expression of BCL6 is known to maintain the proliferation and survival of B-cell lymphomas (49). The fact that BCL6 is hypomethylated, overexpressed, and a direct transcriptional target of MLL fusion proteins suggests that it might contribute to the proliferation and survival of MLLr B-ALLs. To determine whether this is the case, we transduced MLL-AF4 B-ALL cells (BEL-1) with a retrovirus expressing a 4-hydroxytamoxifen (4OHT)-inducible dominant-negative form of BCL6 consisting of its zinc finger domain and also expressing GFP through an internal ribosomal entry site (DNBCL6-ERT2–IRES-GFP; ref. 50) or ERT2–IRES-GFP control vectors. We observed that while the percentage of control ER-GFP–transduced cells remained constant after 3-day exposure to 4-OHT, there was a 55% decrease after 4-OHT induction in GFP-positive cells transduced with dominant negative BCL6 (15.91% vs. 35.00%; Fig. 7A), suggesting that BCL6 is important in maintaining cell proliferation in MLLr B-ALLs. Specific inhibitors of BCL6 were developed to block its transcriptional repressor activity (49, 51, 52) including the widely used peptidomimetic inhibitor RI-BPI, which specifically blocks BCL6 by preventing its binding to co-repressors (52). We tested the ability of 2 MLL-AF4–positive B-ALL cell lines (RS4;11 and SEM) to form colonies after exposure to 5 μmol/L RI-BPI or vehicle. RI-BPI resulted in approximately 90% reduction in colonies in both SEM and RS4;11 cells (Fig. 7B; P < 0.001 for both, t-test). In contrast, RI-BPI treatment of 2 MLLr wild-type ALL cell lines (Nalm6 and REH) did not suppress colony formation (Supplementary Fig. S11A and S11B), suggesting that BCL6 dependence is a feature of MLLr. Similarly, 5 μmol/L RI-BPI induced an approximately 40% reduction in viability of RS4;11 and an approximately 60% reduction in SEM cells after 48 hours as compared with vehicle (P = 0.003 and P < 0.001, respectively) or a control peptide (CP; P = 0.004 and P < 0.001, respectively, t test; Fig. 7C). Seven viably frozen primary human MLLr B-ALL specimens from the E2993 clinical trial were thawed and placed in liquid culture. Expression of BCL6 was verified in each of these cases by qPCR (Supplementary Fig. S12). The cells were exposed to 5 μmol/L RI-BPI, 5 μmol/L CP, or vehicle for 48 hours and assayed for colony forming potential by plating cells in methylcellulose in the presence of IL7. RI-BPI caused a 30% to 70% reduction in colony formation of MLLr specimens (Fig. 7D). Three patient specimens had sufficient cells to conduct dose–response viability assays by flow cytometry using Annexin V and 7-aminoactinomycin D (7AAD) co-staining. Cells were treated with RI-BPI or CP at doses ranging from 1.25 to 20 μmol/L to cover the range of IC50 values reported in BCL6-dependent lymphomas (52). In all 3 cases, RI-BPI, but not CP, induced apoptosis of these primary human MLLr B-ALL within a similar dose range as B-cell lymphomas (ref. 52; Fig. 7E). In contrast, when we ran cell viability experiments on 3 independent B-ALL primary samples (not from ECOG) with the phenotype labels blinded during the experiments, the RI-BPI suppression was only observed in the 2 MLLr cases but not the MLLr wild-type B-ALLs (Supplementary Fig. S13). These data suggest that BCL6 plays a crucial role in maintaining the survival and proliferation of MLLr B-ALL cells and is a bona fide therapeutic target. Collectively, the integration of DNA methylation patterning studies in tandem with gene expression profiling and ChIP-seq provide an enhanced ability to understand the mechanism of action of oncogenic fusion proteins, establish putative prognostic biomarkers, and identify new therapeutic targets in adult patients with B-ALLs.

Figure 7.

BCL6-targeted therapy kills MLLr B-ALL cells. A, BEL-1 cells (MLL-AF4–positive) were transduced with 4-OHT–inducible dominant-negative BCL6 (DN-BCL6-ERT2) or ERT2 control vectors (expressing GFP) and exposed or not exposed to 4-OHT for 3 days. Percentage of GFP-positive cells is indicated in the flow cytometric plots before and after 4-OHT for each construct. B, SEM and RS4;11 cells and (D) 7 E2993 MLL-AF4 patient samples were exposed to 5 μmol/L RI-BPI, 5 μmol/L CP, or vehicle for 48 hours, followed by colony formation assay. The y-axis shows percent colony formation of RI-BPI–treated cells versus vehicle. Experiments were carried out in triplicate. C, RS4;11 and SEM cells were exposed to 5 μmol/L RI-BPI, 5 μmol/L CP, or vehicle for 48 hours, followed by flow cytometry for Annexin V/7-AAD. Percent viability mol/L vehicle is shown on the y-axis. Experiments were carried out in triplicate. E, three E2993 MLL-AF4 patient samples were treated with increasing concentrations of RI-BPI or CP for 48 hours, and viability determined by Annexin V/7-AAD flow cytometry. Percent viability relative to CP is shown on the y-axis. Data represent means ± SEM (n = 3). Student t test was used to calculate P values.

Figure 7.

BCL6-targeted therapy kills MLLr B-ALL cells. A, BEL-1 cells (MLL-AF4–positive) were transduced with 4-OHT–inducible dominant-negative BCL6 (DN-BCL6-ERT2) or ERT2 control vectors (expressing GFP) and exposed or not exposed to 4-OHT for 3 days. Percentage of GFP-positive cells is indicated in the flow cytometric plots before and after 4-OHT for each construct. B, SEM and RS4;11 cells and (D) 7 E2993 MLL-AF4 patient samples were exposed to 5 μmol/L RI-BPI, 5 μmol/L CP, or vehicle for 48 hours, followed by colony formation assay. The y-axis shows percent colony formation of RI-BPI–treated cells versus vehicle. Experiments were carried out in triplicate. C, RS4;11 and SEM cells were exposed to 5 μmol/L RI-BPI, 5 μmol/L CP, or vehicle for 48 hours, followed by flow cytometry for Annexin V/7-AAD. Percent viability mol/L vehicle is shown on the y-axis. Experiments were carried out in triplicate. E, three E2993 MLL-AF4 patient samples were treated with increasing concentrations of RI-BPI or CP for 48 hours, and viability determined by Annexin V/7-AAD flow cytometry. Percent viability relative to CP is shown on the y-axis. Data represent means ± SEM (n = 3). Student t test was used to calculate P values.

Close modal

The focus of this study was to identify promoter DNA methylation profiles associated with unfavorable B-ALL subtypes in adult patients and determine their functional and clinical significance. This effort was facilitated by profiling a large and unique cohort of clinically well-annotated adult patients with B-ALL enrolled in a single multicenter phase III clinical trial (E2993). We observed that aberrant epigenetic regulation occurs universally and is distributed to specific gene sets in genetically defined B-ALL subtypes associated with poor outcomes. Integrative analysis of epigenetic and transcriptional profiles identified core gene signatures, associated with the mechanism of action of aberrant fusion proteins and pointing to key molecular mechanisms and potential therapeutic targets. We also conducted unsupervised hierarchical clustering of the 97 normal karyotype B-ALLs in our cohort, which yielded 2 robust clusters (1-Pearson, Ward method, ×1,000 bootstrapping, data not shown). A supervised analysis revealed 1,512 genes differentially methylated (methylation difference >20%, adjusted P < 0.01; Student t test) between the 2 clusters, enriched in hematopoietic cell quiescence genes (ref. 53; Benjamini–Hochberg adjusted P = 0.03) as well as MYC (Benjamini–Hochberg adjusted P = 0.02), and IKAROS target genes (ref. 54; Benjamini–Hochberg adjusted P = 0.03). These data suggest differences in the underlying pathogenesis of these cases and point toward a rationale for further study of mechanisms driving aberrant gene regulation in translocation-negative patients.

In the case of BCR–ABL1-positive B-ALLs, we identified the most deregulated gene network captured by integrative analysis of DNA methylation and gene expression profiles was centered around IL2RA, which is the most hypomethylated and overexpressed gene in BCR–ABL1-positive B-ALLs. In general, IL2RA (CD25) positivity was strongly associated with BCR–ABL1 translocation in this large cohort of patients with B-ALLs, which confirms a previous report on a smaller set of E2993 patients (40). BCR–ABL1-positive patients with ALLs experienced significantly worse outcome than other patients with B-ALLs in the E2993 clinical trial. However, much of the difference appears related to CD25 expression, as it was the subset of BCR–ABL1-positive CD25-positive B-ALLs that featured the lowest OS, whereas BCR–ABL1-positive CD25-negative cases were not statistically worse than BCR–ABL1-negative B-ALLs. Further research will be required to determine how and why IL2RA and BCR–ABL1 are functionally linked. It is notable that CD25 expression signature in our B-ALLs (data not shown) was significantly associated with a leukemia stem cell (LSC) signature (ref. 55; Benjamini–Hochberg adjusted P < 1e-5) and a hematopoietic stem cell (HSC) signature (56) in leukemia (Benjamini–Hochberg adjusted P < 1e-6; Fisher exact test). Because LSCs are generally considered to be more chemoresistant (57), it is possible that CD25 positivity might be an indicator of an enlarged LSC-like fraction. The IL2RA network also featured aberrant regulation of numerous cytokines and interleukins. Notably, paracrine cytokine signaling by tumor cell microenvironment was recently shown to also confer chemoresistance (58). Along these lines, quiescent AML leukemia stem cells localized to the endosteal niche within the bone marrow strongly expressed CD25 (59). These endosteal localized AML stem cells were also particularly chemotherapy-resistant (60), providing another possible link between CD25 positivity and poor outcome. Collectively, these data raise the possibility that CD25 positivity may be linked in some way to cross-talk between BCR–ABL1-positive B-ALL cells and the bone marrow microenvironment. Along these lines, we noted that when placed in in vitro culture, CD25 levels on B-ALL cells markedly decline (data not shown). Importantly, expression of CD25 provides an opportunity for therapeutic targeting. Monoclonal antibodies against CD25 such as RFT5.dgA, which is conjugated to deglycosylated ricin A-chain, have specific activity against CD25-positive cells in vitro and in vivo, and are currently in clinical trials for CD25-positive lymphoid malignancies (61, 62). Using such antibodies, it seems reasonable to attempt to eradicate CD25-positive B-ALL cells in patients to determine whether these cells contribute to chemoresistance and relapse. CD25 immunotoxins are also able to kill infrequently dividing cells and hence may be able to eradicate more quiescent cells such as LSCs. Meanwhile, the underlying cause of aberrant hypomethylation and upregulation of CD25 needs to be further investigated and may provide clues to explain the etiology of CD25-positive B-ALLs.

DNA methylation profiling coupled with ChIP-seq suggested that hypomethylation signatures in B-ALLs with t(1;19) and MLL rearrangements may be functionally linked to the function of these fusion oncoproteins. In E2A–PBX1 B-ALLs, 14 of the 58 hypomethylation signature genes were also identified as E2A–PBX1 fusion protein complex targets by ChIP-seq using E2AN, PBX1C, and p300 antibodies. E2A–PBX1 recruits p300 and perhaps other co-activators to the target genes, which could potentially lead to loss of regulatory cytosine methylation and may epigenetically facilitate transcriptional activation. A number of these hypomethylated genes appear functionally relevant to leukemogenesis including glutamine metabolic pathways and factors such as BLK (B lymphocyte tyrosine kinase), which is an SRC kinase that can be targeted by tyrosine kinase inhibitors such as dasatinib (63). In MLLr B-ALLs, the MLL-AF4 fusion targets identified by the overlap of 3 ChIP-seq datasets (MLLN, AF4C, and H3K79me2) also showed increased hypomethyation and gene expression compared with nontarget genes. All MLL fusions retain the wild-type MLL CXXC motif, which mediates direct binding to unmethylated CpG sequences (64). Binding of wild-type MLLs to CpG-rich regions in the HOXA9 locus has been shown to prevent these genomic regions from becoming methylated (65). MLL depletion resulted in gain of cytosine methylation at these HOXA9 sites and conversely re-expression of MLLs rescued the locus back to a hypomethylated state (65). A similar effect was observed with MLL-AF4 and MLL-AF9 (64, 65). In fact, mutation of the CXXC motifs that disrupt binding to hypomethylated CpGs abrogate the leukemogenic functions of MLL fusion proteins (64), suggesting that maintenance of hypomethylation is a critical component of MLLr leukemogenesis. Our data suggest that this CXXC mechanism may extend to additional genes actively regulated by MLL fusion proteins (an illustration cartoon for a potential relationship between MLL-AF4 binding and DNA hypomethylation was shown in Supplementary Fig. S14A and S14B). Recent data from our group also suggest a link between HOXA9 and enhancer hypomethylation in myeloid MLLr leukemias (66). At the current time, there is no evidence for direct biochemical cross-talk between DNMTs and MLL fusion proteins. However, given the previously described cooperation between Polycomb complexes and DNMTs (67) and the generally antagonistic actions of MLL/trithorax proteins with Polycomb (68), it remains possible that competition between MLL and DNMTs might occur in at least an indirect manner (Supplementary Fig. S14). Additional studies will be needed to explore whether aberrant DNA hypomethylation of loci occurring during leukemogenesis facilitates the recruitment of MLL fusions and induces abnormal transcriptional activation or whether MLL fusion proteins in some way contribute or cause the initial aberrant hypomethylation of these loci. Whichever the case it seems clear that MLL binding to hypomethylated DNA is important to its leukemic functions and perhaps facilitates assembly of MLL fusion complexes with DOT1L, thus accounting for the association we observed between DNA hypomethylation, MLL fusion protein binding, and H3K79me2. Perhaps CXXC-induced DNA hypomethylation and DOT1L work together or form a positive feedback loop to highly upregulate certain gene sets important for leukemogenesis in MLLr cells, including not only HOX genes but also factors such as FLT3 and BCL6. In fact, FLT3 was one of the most hypomethylated and highly expressed genes in MLLr B-ALLs. FLT3 expression in MLLr leukemia was previously reported (69) as functionally relevant, as FLT3 inhibitors could suppress the growth of MLLr cells (46). We show herein that FLT3 is in fact a bona fide direct target gene of MLL-AF4 and perhaps an example of a leukemogenic gene in which MLL fusion protein associated DNA hypomethylation and H3K79me2 epigenetically deregulate its expression.

Integrative epigenomic analysis allowed us to identify the BCL6 transcriptional repressor as a therapeutic target in MLLr B-ALLs. The BCL6 locus displayed all the hallmarks of MLL fusion protein–mediated epigenetic deregulation including direct MLL-AF4 binding, DNA hypomethylation, H3K79 dimethylation, and transcriptional upregulation. BCL6 is required for germinal center B cells to undergo proliferation and immunoglobulin affinity maturation. BCL6 mediates these effects by suppressing replication and DNA damage checkpoints (70–72). More recently, BCL6 was shown to play a similar role in maintaining survival of pre-B cells (73). Specifically, pre-B cells transiently express BCL6 during the transition from IL7-dependent to IL7-independent stages, whereupon BCL6 protects cells from genotoxic stress during immunoglobulin light chain recombination (73). These actions link BCL6 to processes associated with development of B-ALLs. In the same way, that constitutive expression of BCL6 in germinal center cells contributes to formation of diffuse large B-cell lymphomas (74), it is reasonable to hypothesize that MLL fusion–driven constitutive expression of BCL6 might also contribute to B-ALL leukemogenesis. Importantly, our data indicate that MLLr B-ALLs are addicted to BCL6 and require its presence to maintain proliferation and survival. This scenario is quite different than the recently described role of BCL6 as an imatinib treatment–inducible factor that protects BCR–ABL1-positive B-ALL cells from the toxic effects of ABL kinase blockade (75). This current study expands the paradigm of BCL6-targeted therapy to an aggressive form of B-ALLs and supports the rationale for developing clinical trials of BCL6 inhibitors specifically for MLLr B-ALLs. Altogether, it appears that MLL fusion proteins drive expression of FLT3, BCL6, and other key downstream target genes through the dual epigenetic activating effects of inducing and maintaining aberrant DNA hypomethylation and aberrant H3K79 dimethylation.

A small previous study combining DNA methylation profiling, gene expression profiling, and ChIP-on-chip in 5 patients with leukemia illustrated the proof of principle that triangulating between these platforms allowed more depth characterization of gene deregulation than single platforms alone (76). In manner similar to recent studies in AML showing that DNA methylation patterns can be used to identify new pathogenic mechanisms (14), this current analysis of adult patients with B-ALLs enrolled in ECOG E2993 phase III clinical study shows the power of the integrative epigenomic approach to identify disease mechanisms and new prognostic biomarkers and therapeutic targets. These data point the way for future functional studies to explore in depth how BCR–ABL1, E2A–PBX1, and MLL fusions directly or indirectly alter DNA methylation during malignant transformation and to test the role of IL2RA signaling in mediating chemotherapy-resistant disease. Most importantly, we expect these studies to trigger therapeutic trials of BCL6 inhibitors in MLL leukemia and the application of CD25 as a putative biomarker for risk stratification in B-ALLs.

Patient Samples, Human Cells, and Cell Lines

Pretreatment bone marrow or peripheral blood samples were obtained at diagnosis (before any treatment) from 215 patients enrolled in the Medical Research Council (MRC) UKALLXII/Eastern Cooperative Oncology Group (ECOG) E2993 phase III trial in adult B-lineage ALLs. Clinical details, including age, white blood cell count (WBC), and survival time, were collected by the Clinical Trial Service Unit (CTSU), University of Oxford, Oxford, UK. All patient specimens used in this study were accrued by ECOG. The diagnosis was established by central morphology review and confirmed by multiparameter flow cytometry and reverse transcription-PCR assays in ECOG's Leukemia Translational Studies Laboratory. Cytogenetic information was reviewed by the ECOG Cytogenetics Committee. The study was approved by the Institutional Review Board of each treatment center. Normal human pre-B or pro-B cells (CD19+ and VpreB+) were sorted from the bone marrows of 12 healthy donors by flow cytometry using antibodies (BD Biosciences) and a FACSVantage SE cell sorter (BD Biosciences). The use of human tissue was in agreement with research ethics board of Weill Cornell Medical College (New York, NY). Patient characteristics are given in Supplementary Tables S1 and S2. The human ALL cell lines RS4;11, SEM, BEL-1, CCRF-CEM, 697, Nalm6, REH, BV173, NALM1, SUP-B15, TOM1, and MUTZ5 were obtained fresh from DSMZ (Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH, Germany). The expression of aberrant fusion proteins was verified by Western blot analysis and qPCR.

Array-Based Methylation Analysis Using HELP

The HELP (HpaII tiny fragment enrichment by ligation-mediated PCR) assay was conducted as previously published (refs. 31, 33, 77; Supplementary Data). HELP data quality control (QC) and analysis were conducted as described previously (32) using R software (78). Basically, raw data (.pair) files were generated using NimbleScan software. Signal intensities at each HpaII-amplifiable fragment were calculated as a robust (25% trimmed) mean of their component probe-level signal intensities. Any fragments found within the level of background MspI signal intensity, measured as 2.5 mean absolute deviation (MAD) above the median of random probe signals, were considered as “failed” probes and removed. After QC processing, a median normalization was conducted on each array by subtracting the median log ratio (HpaII/MspI), resulting in median log ratio of 0 for each array.

HELP Array Data Analysis

Identification of the aberrantly methylated genes for each B-ALL subtype compared with the normal pre-B cells was conducted using an ANOVA test followed by Dunnett post hoc test (34) using the normal pre-B samples as the reference group and multiple testing corrected by the Benjamini–Hochberg method (35). R Package “multcomp” Version 1.2-6 was used (34). The comparison between the positive and negative groups of BCR–ABL1, E2A–PBX1, and MLLr was determined by the Student t test and the Benjamini–Hochberg adjusted P values. As the methylation difference between ALL and normal pre-B cells is more apparent than the methylation difference among the ALL subtypes, to capture and prioritize the top differentially methylated genes, we used a higher threshold dx > 1.5 (which corresponds to 30% methylation difference) in the ALL versus normal pre-B comparison, and a lower threshold dx > 1 (which corresponds to 20% methylation difference) in the ALL subtype comparison. On the basis of the linear regression line in MassArray EpiTyper validation (Supplementary Fig. S1, equation: y = −4.9484x + 3.3232, where x is MassArray readout and y is HELP readout), a log2 ratio difference (dx) of 1 and 1.5 from HELP data corresponds to methylation difference of 20% and 30%, respectively.

Single Locus Quantitative DNA Methylation Assays

MassArray EpiTyper assays (Sequenom) were conducted on bisulfite-converted DNA, as previously described (79). The primers were designed using Sequenom EpiDesigner beta software (Sequenom, Inc.), and data were analyzed using EpiTyper software version 1.0 (Sequenom, Inc.). Note that MassArray and qPCR validation studies were conducted in different sets of randomly selected specimens with available DNA and RNA, from among the cohort of 215 profiled E2993 patients.

Cell Culture

The human B-ALL cell lines were maintained in Roswell Park Memorial Institute medium (RPMI-1640, Invitrogen) with 10% FBS, 100 IU/mL penicillin, and 100 μg/mL streptomycin. Primary B-ALL cells were cultured in Iscove's modified Dulbecco's medium (IMDM, Invitrogen) with 10% FBS, LDL, 1× β-mercaptoethanol, 2 mmol/L l-glutamine, 50 U/mL penicillin, 50 μg/mL streptomycin, recombinant human FLTG ligand (rhFLT3) (20 ng/mL), rhSCF (50 ng/mL), IL3, (20 ng/mL), IL6 (20 ng/mL), and IL7 (10 ng/mL; Invitrogen). Cell cultures were kept at 37°C in a humidified incubator under a 5% CO2 atmosphere.

MLL-AF4 and E2A–PBX1 Knockdown

The MLL-AF4 knockdown in RS4;11 and SEM cells was carried out as described in the work of Thomas and colleagues (ref. 48; Supplementary Data). Synthetic sense and antisense oligoribonucleotides specific for the MLL-AF4 fusion sites were synthesized by Dharmacon RNAi Technologies. RNA was extracted 48 hours after transduction, and the levels of genes were determined by qPCR. The E2A–PBX1 knockdown in 697 cells was carried out using an E2A-shRNA that targets the middle coding region (AA 340–348) of E2A (E12/E47) and therefore affects both the E2A and E2A fusions (E2A–PBX1a and E2A–PBX1b). The shRNA vectors were obtained from Open Biosystems, clone ID: TRCN0000017534. The lentivirus was prepared as described in the Addgene website (Addgene). RNA was extracted 96 hours postinfection and transcript abundance determined by qPCR.

Western Blot

For BCL6 Western blot analysis, rabbit antibody raised against BCL6-N3 (sc-858) and anti-actin-horseradish peroxidase (HRP) conjugated (sc-1615) were purchased from Santa Cruz Biotechnology. For the Western blot experiments in MLL-AF4 siRNA knockdown, SEM and RS4;11 cells were treated with control and MLL-AF4 siRNAs as described. A total of 4 × 106 cells were collected, and whole-cell extracts were made via resuspension in 40 μL of a modified radioimmunoprecipitation assay (RIPA) buffer (50 mmol/L HEPES-KOH, pH 7.6, 500 mmol/L LiCl, 1 mmol/L EDTA, 1% NP-40, 0.7% Na-Deoxycholate, ddH2O) for 30 minutes at 4°C. Whole-cell extracts were split in half and loaded onto either a 3%–8% Tris-Acetate gel (NuPAGE, Life Sciences) or a 4%–12% Bis-Tris gel (NuPAGE, Life Sciences). Gels were blotted onto PVDF membranes and probed with BCL6 antibodies (as described), anti-HOXA9 (Millipore, 07-178), anti-AF4-C (Abcam, ab31812), or anti-GAPDH (Bethyl, A300-641A). RS4;11 Western blot bands were quantified relative to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) using a standard protocol on ImageJ (80).

Cell Viability Assay

One hundred thousand MLL-AF4 B-ALL cells from the cell lines or 200,000 primary cells were seeded in a volume of 100 μL medium (as described in “Cell Culture”) per well. RI-BPI or CPs were diluted in medium and added at the indicated concentration in a total culture volume of 120 μL. After culturing for 48 hours, cell viability was measured by flow cytometry using Annexin V–fluorescein isothiocyanate (FITC; BD Biosciences) and 7-AAD (Molecular Probes-Invitrogen) to detect phosphatidylserine exposition and cell permeability, respectively. At least 50,000 events were recorded per condition on an LSR II flow cytometer (BD Biosciences). Data analysis was conducted using FlowJo 8.2 software for Mac OS X (TreeStar). Cells that were negative for Annexin V and 7-AAD were scored as viable. Fold changes were calculated using baseline values of untreated or the CP-treated cells as a reference.

Colony-Forming Assay

MLL-AF4 ALL cell lines were plated with 10,000 cells per plate in semisolid agar in the presence of 5 μmol/L retro-inverso BCL6 peptide-inhibitor (RI-BPI) or vehicle. Colonies were counted after 10 days. For MLL-AF4 B-ALL primary samples, 25,000 cells per sample were treated with RI-BPI, CP, or vehicle for 48 hours and plated in M3434 methylcult (Stem Cell Technologies) with IL-7 (10 ng/mL; Invitrogen). Colonies were counted after 14 days.

ChIP and qChIP

RS4;11, SEM, CCRF-CEM, 697, and Nalm6 cells were fixed with 1% formaldehyde, lysed, and sonicated to generate fragments less than 500 bp. Sonicated lysates were incubated with antibodies overnight, and after increasing stringency, washes immunocomplexes were recovered and DNA was isolated. The ChIP protocol was conducted as in the work of Milne and colleagues (81). A fraction of the ChIP products was used as template in qPCR reactions using TaqMan or SYBR Green and a 7900 ABI qRT-QPCR machine (Applied Biosystems). The qChIP results were quantified relative to inputs as explained in detail in the work of Milne and colleagues (81).

ChIP-seq

See Supplementary Data for details. Peaks from ChIP-seq data were called using the ChIPseeqer program (42) with parameters indicated in Supplementary Table S14.

Survival Analysis

Kaplan–Meier survival analysis was used to estimate OS, and log-rank test was used to compare survival differences between patient groups. R package “survival” version 2.35-8 (78) and “proc lifetest” in SAS was used for the survival analysis.

No potential conflicts of interest were disclosed.

Conception and design: H. Geng, W.-Y. Chen, R.G. Roeder, E. Paietta, O. Elemento, A.M. Melnick

Development of methodology: H. Geng, W.-Y. Chen, O. Elemento, A.M. Melnick

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.A. Milne, W.-Y. Chen, C. Hurtz, S.-M. Kweon, S. Shojaee, C. Huang, J. Racevskis, H. Lazarus, M.S. Tallman, J.M. Rowe, M.R. Litzow, M.L. Guzman, C.D. Allis, R.G. Roeder, E. Paietta

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Geng, C. Hurtz, S.-M. Kweon, L. Zickl, D. Neuberg, M.S. Tallman, M.L. Guzman, R.G. Roeder, M. Müschen, E. Paietta, O. Elemento, A.M. Melnick

Writing, review, and/or revision of the manuscript: H. Geng, S. Brennan, T.A. Milne, D. Neuberg, D. Biswas, R.P. Ketterling, S.M. Luger, H. Lazarus, M.S. Tallman, J.M. Rowe, M.R. Litzow, M. Müschen, E. Paietta, O. Elemento, A.M. Melnick

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S. Brennan, D. Biswas, J. Racevskis, R.P. Ketterling, A.M. Melnick

Study supervision: M.L. Guzman, C.D. Allis, E. Paietta, O. Elemento, A.M. Melnick

Sample processing: Y. Li, Y. Xin

Patient enrollment: S.M. Luger

The authors thank Dr. Peter H. Wiernik for his support of the E2993 clinical trial and correlative studies as former chair of the ECOG leukemia committee and Dr. Sue Richards and the MRC/ECOG trial management group supported by MRC grant (G8223452) for maintaining the trial database for ECOG. They also thank the assistance of the Weill Cornell Epigenomics Core Facility.

A.M. Melnick is supported by the Chemotherapy Foundation, Burroughs Wellcome Foundation, and the Sackler Center for Biomedical and Physical Sciences at the Weill Cornell Medical College. O. Elemento is supported in part by the NSF CAREER grant DBI 1054964. A.M. Melnick and M. Müschen are supported by LLS TRP 6097-10. A.M. Melnick, R.G. Roeder, C.D. Allis, and T.A. Milne were supported by LLS SCOR 7132-08. T.A. Milne was supported by the Medical Research Council UK. R.G. Roeder was supported by Starr Cancer Consortium and NIH0 grants. W.-Y. Chen and D. Biswas were supported by LLS fellowship 5230-09 and 5089-10.

1.
Pui
CH
,
Evans
WE
. 
Treatment of acute lymphoblastic leukemia
.
N Engl J Med
2006
;
354
:
166
78
.
2.
Armstrong
SA
,
Look
AT
. 
Molecular genetics of acute lymphoblastic leukemia
.
J Clin Oncol
2005
;
23
:
6306
15
.
3.
Yeoh
EJ
,
Ross
ME
,
Shurtleff
SA
,
Williams
WK
,
Patel
D
,
Mahfouz
R
, et al
Classification, subtype discovery, and prediction of outcome in pediatric acute lymphoblastic leukemia by gene expression profiling
.
Cancer Cell
2002
;
1
:
133
43
.
4.
Ross
ME
,
Zhou
X
,
Song
G
,
Shurtleff
SA
,
Girtman
K
,
Williams
WK
, et al
Classification of pediatric acute lymphoblastic leukemia by gene expression profiling
.
Blood
2003
;
102
:
2951
9
.
5.
Chiaretti
S
,
Li
X
,
Gentleman
R
,
Vitale
A
,
Wang
KS
,
Mandelli
F
, et al
Gene expression profiles of B-lineage adult acute lymphocytic leukemia reveal genetic patterns that identify lineage derivation and distinct mechanisms of transformation
.
Clin Cancer Res
2005
;
11
:
7209
19
.
6.
Juric
D
,
Lacayo
NJ
,
Ramsey
MC
,
Racevskis
J
,
Wiernik
PH
,
Rowe
JM
, et al
Differential gene expression patterns and interaction networks in BCR-ABL-positive and -negative adult acute lymphoblastic leukemias
.
J Clin Oncol
2007
;
25
:
1341
9
.
7.
Jones
PA
,
Baylin
SB
. 
The epigenomics of cancer
.
Cell
2007
;
128
:
683
92
.
8.
Esteller
M
. 
Epigenetics in cancer
.
N Engl J Med
2008
;
358
:
1148
59
.
9.
Ji
H
,
Ehrlich
LI
,
Seita
J
,
Murakami
P
,
Doi
A
,
Lindau
P
, et al
Comprehensive methylome map of lineage commitment from haematopoietic progenitors
.
Nature
2010
;
467
:
338
42
.
10.
Broske
AM
,
Vockentanz
L
,
Kharazi
S
,
Huska
MR
,
Mancini
E
,
Scheller
M
, et al
DNA methylation protects hematopoietic stem cell multipotency from myeloerythroid restriction
.
Nat Genet
2009
;
41
:
1207
15
.
11.
Trowbridge
JJ
,
Snow
JW
,
Kim
J
,
Orkin
SH
. 
DNA methyltransferase 1 is essential for and uniquely regulates hematopoietic stem and progenitor cells
.
Cell Stem Cell
2009
;
5
:
442
9
.
12.
Polo
JM
,
Liu
S
,
Figueroa
ME
,
Kulalert
W
,
Eminli
S
,
Tan
KY
, et al
Cell type of origin influences the molecular and functional properties of mouse induced pluripotent stem cells
.
Nat Biotechnol
2010
;
28
:
848
55
.
13.
Kim
K
,
Doi
A
,
Wen
B
,
Ng
K
,
Zhao
R
,
Cahan
P
, et al
Epigenetic memory in induced pluripotent stem cells
.
Nature
2010
;
467
:
285
90
.
14.
Figueroa
ME
,
Abdel-Wahab
O
,
Lu
C
,
Ward
PS
,
Patel
J
,
Shih
A
, et al
Leukemic IDH1 and IDH2 mutations result in a hypermethylation phenotype, disrupt TET2 function, and impair hematopoietic differentiation
.
Cancer Cell
2010
;
18
:
553
67
.
15.
Suzuki
H
,
Watkins
DN
,
Jair
KW
,
Schuebel
KE
,
Markowitz
SD
,
Chen
WD
, et al
Epigenetic inactivation of SFRP genes allows constitutive WNT signaling in colorectal cancer
.
Nat Genet
2004
;
36
:
417
22
.
16.
Di Croce
L
,
Raker
VA
,
Corsaro
M
,
Fazi
F
,
Fanelli
M
,
Faretta
M
, et al
Methyltransferase recruitment and DNA hypermethylation of target promoters by an oncogenic transcription factor
.
Science
2002
;
295
:
1079
82
.
17.
Ley
TJ
,
Ding
L
,
Walter
MJ
,
McLellan
MD
,
Lamprecht
T
,
Larson
DE
, et al
DNMT3A mutations in acute myeloid leukemia
.
N Engl J Med
2010
;
363
:
2424
33
.
18.
Krivtsov
AV
,
Feng
Z
,
Lemieux
ME
,
Faber
J
,
Vempati
S
,
Sinha
AU
, et al
H3K79 methylation profiles define murine and human MLL-AF4 leukemias
.
Cancer Cell
2008
;
14
:
355
68
.
19.
Bernt
KM
,
Zhu
N
,
Sinha
AU
,
Vempati
S
,
Faber
J
,
Krivtsov
AV
, et al
MLL-rearranged leukemia is dependent on aberrant H3K79 methylation by DOT1L
.
Cancer Cell
2011
;
20
:
66
78
.
20.
Jo
SY
,
Granowicz
EM
,
Maillard
I
,
Thomas
D
,
Hess
JL
. 
Requirement for Dot1l in murine postnatal hematopoiesis and leukemogenesis by MLL translocation
.
Blood
2011
;
117
:
4759
68
.
21.
Biswas
D
,
Milne
TA
,
Basrur
V
,
Kim
J
,
Elenitoba-Johnson
KS
,
Allis
CD
, et al
Function of leukemogenic mixed lineage leukemia 1 (MLL) fusion proteins through distinct partner protein complexes
.
Proc Natl Acad Sci U S A
2011
;
108
:
15751
6
.
22.
Lin
C
,
Smith
ER
,
Takahashi
H
,
Lai
KC
,
Martin-Brown
S
,
Florens
L
, et al
AFF4, a component of the ELL/P-TEFb elongation complex and a shared subunit of MLL chimeras, can link transcription elongation to leukemia
.
Mol Cell
2010
;
37
:
429
37
.
23.
Yokoyama
A
,
Lin
M
,
Naresh
A
,
Kitabayashi
I
,
Cleary
ML
. 
A higher-order complex containing AF4 and ENL family proteins with P-TEFb facilitates oncogenic and physiologic MLL-dependent transcription
.
Cancer Cell
2010
;
17
:
198
212
.
24.
Stumpel
DJ
,
Schneider
P
,
van Roon
EH
,
Boer
JM
,
de Lorenzo
P
,
Valsecchi
MG
, et al
Specific promoter methylation identifies different subgroups of MLL-rearranged infant acute lymphoblastic leukemia, influences clinical outcome, and provides therapeutic options
.
Blood
2009
;
114
:
5490
8
.
25.
Schafer
E
,
Irizarry
R
,
Negi
S
,
McIntyre
E
,
Small
D
,
Figueroa
ME
, et al
Promoter hypermethylation in MLL-r infant acute lymphoblastic leukemia: biology and therapeutic targeting
.
Blood
2010
;
115
:
4798
809
.
26.
Hogan
LE
,
Meyer
JA
,
Yang
J
,
Wang
J
,
Wong
N
,
Yang
W
, et al
Integrated genomic analysis of relapsed childhood acute lymphoblastic leukemia reveals therapeutic strategies
.
Blood
2011
;
118
:
5218
26
.
27.
Taylor
KH
,
Pena-Hernandez
KE
,
Davis
JW
,
Arthur
GL
,
Duff
DJ
,
Shi
H
, et al
Large-scale CpG methylation analysis identifies novel candidate genes and reveals methylation hotspots in acute lymphoblastic leukemia
.
Cancer Res
2007
;
67
:
2617
25
.
28.
Kuang
SQ
,
Tong
WG
,
Yang
H
,
Lin
W
,
Lee
MK
,
Fang
ZH
, et al
Genome-wide identification of aberrantly methylated promoter associated CpG islands in acute lymphocytic leukemia
.
Leukemia
2008
;
22
:
1529
38
.
29.
Kuang
SQ
,
Bai
H
,
Fang
ZH
,
Lopez
G
,
Yang
H
,
Tong
W
, et al
Aberrant DNA methylation and epigenetic inactivation of Eph receptor tyrosine kinases and ephrin ligands in acute lymphoblastic leukemia
.
Blood
2010
;
115
:
2412
9
.
30.
Davidsson
J
,
Lilljebjorn
H
,
Andersson
A
,
Veerla
S
,
Heldrup
J
,
Behrendtz
M
, et al
The DNA methylome of pediatric acute lymphoblastic leukemia
.
Hum Mol Genet
2009
;
18
:
4054
65
.
31.
Figueroa
ME
,
Melnick
A
,
Greally
JM
. 
Genome-wide determination of DNA methylation by Hpa II tiny fragment enrichment by ligation-mediated PCR (HELP) for the study of acute leukemias
.
Methods Mol Biol
2009
;
538
:
395
407
.
32.
Thompson
RF
,
Reimers
M
,
Khulan
B
,
Gissot
M
,
Richmond
TA
,
Chen
Q
, et al
An analytical pipeline for genomic representations used for cytosine methylation studies
.
Bioinformatics
2008
;
24
:
1161
7
.
33.
Figueroa
ME
,
Lugthart
S
,
Li
Y
,
Erpelinck-Verschueren
C
,
Deng
X
,
Christos
PJ
, et al
DNA methylation signatures identify biologically distinct subtypes in acute myeloid leukemia
.
Cancer Cell
2010
;
17
:
13
27
.
34.
Hothorn
T
,
Bretz
F
,
Westfall
P
. 
Simultaneous inference in general parametric models
.
Biom J
2008
;
50
:
346
63
.
35.
Benjamini
Y
,
Hochberg
Y
. 
Controlling the false discovery rate: a practical and powerful approach to multiple testing
.
J R Stat Soc Ser B
1995
;
57
:
289
300
.
36.
Wen
B
,
Wu
H
,
Shinkai
Y
,
Irizarry
RA
,
Feinberg
AP
. 
Large histone H3 lysine 9 dimethylated chromatin blocks distinguish differentiated from embryonic stem cells
.
Nat Genet
2009
;
41
:
246
50
.
37.
Shaffer
AL
,
Wright
G
,
Yang
L
,
Powell
J
,
Ngo
V
,
Lamy
L
, et al
A library of gene expression signatures to illuminate normal and pathological lymphoid biology
.
Immunol Rev
2006
;
210
:
67
85
.
38.
Wright
G
,
Tan
B
,
Rosenwald
A
,
Hurt
EH
,
Wiestner
A
,
Staudt
LM
. 
A gene expression-based method to diagnose clinically distinct subgroups of diffuse large B cell lymphoma
.
Proc Natl Acad Sci U S A
2003
;
100
:
9991
6
.
39.
van Zelm
MC
,
van der Burg
M
,
de Ridder
D
,
Barendregt
BH
,
de Haas
EF
,
Reinders
MJ
, et al
Ig gene rearrangement steps are initiated in early human precursor B cell subsets and correlate with specific transcription factor expression
.
J Immunol
2005
;
175
:
5912
22
.
40.
Paietta
E
,
Racevskis
J
,
Neuberg
D
,
Rowe
JM
,
Goldstone
AH
,
PH
W
. 
Expression of CD25 (interleukin-2 receptor alpha chain) in adult acute lymphoblastic leukemia predicts for the presence of BCR/ABL fusion transcripts: results of a preliminary laboratory analysis of ECOG/MRC Intergroup Study E2993
.
Eastern Cooperative Oncology Group
/Medical Research Council
.
Leukemia
1997
;
11
:
1887
90
.
41.
Bayly
R
,
Chuen
L
,
Currie
RA
,
Hyndman
BD
,
Casselman
R
,
Blobel
GA
, et al
E2A-PBX1 interacts directly with the KIX domain of CBP/p300 in the induction of proliferation in primary hematopoietic cells
.
J Biol Chem
2004
;
279
:
55362
71
.
42.
Giannopoulou
EG
,
Elemento
O
. 
An integrated ChIP-seq analysis platform with customizable workflows
.
BMC Bioinformatics
2011
;
12
:
277
.
43.
Aspland
SE
,
Bendall
HH
,
Murre
C
. 
The role of E2A-PBX1 in leukemogenesis
.
Oncogene
2001
;
20
:
5708
17
.
44.
Krivtsov
AV
,
Armstrong
SA
. 
MLL translocations, histone modifications and leukaemia stem-cell development
.
Nat Rev Cancer
2007
;
7
:
823
33
.
45.
Sanz
M
,
Burnett
A
,
Lo-Coco
F
,
Lowenberg
B
. 
FLT3 inhibition as a targeted therapy for acute myeloid leukemia
.
Curr Opin Oncol
2009
;
21
:
594
600
.
46.
Armstrong
SA
,
Kung
AL
,
Mabon
ME
,
Silverman
LB
,
Stam
RW
,
Den Boer
ML
, et al
Inhibition of FLT3 in MLL. Validation of a therapeutic target identified by gene expression based classification
.
Cancer Cell
2003
;
3
:
173
83
.
47.
Armstrong
SA
,
Mabon
ME
,
Silverman
LB
,
Li
A
,
Gribben
JG
,
Fox
EA
, et al
FLT3 mutations in childhood acute lymphoblastic leukemia
.
Blood
2004
;
103
:
3544
6
.
48.
Thomas
M
,
Gessner
A
,
Vornlocher
HP
,
Hadwiger
P
,
Greil
J
,
Heidenreich
O
. 
Targeting MLL-AF4 with short interfering RNAs inhibits clonogenicity and engraftment of t(4;11)-positive human leukemic cells
.
Blood
2005
;
106
:
3559
66
.
49.
Polo
JM
,
Dell'Oso
T
,
Ranuncolo
SM
,
Cerchietti
L
,
Beck
D
,
Da Silva
GF
, et al
Specific peptide interference reveals BCL6 transcriptional and oncogenic mechanisms in B-cell lymphoma cells
.
Nat Med
2004
;
10
:
1329
35
.
50.
Shaffer
AL
,
Yu
X
,
He
Y
,
Boldrick
J
,
Chan
EP
,
Staudt
LM
. 
BCL-6 represses genes that function in lymphocyte differentiation, inflammation, and cell cycle control
.
Immunity
2000
;
13
:
199
212
.
51.
Cerchietti
LC
,
Ghetu
AF
,
Zhu
X
,
Da Silva
GF
,
Zhong
S
,
Matthews
M
, et al
A small-molecule inhibitor of BCL6 kills DLBCL cells in vitro and in vivo
.
Cancer Cell
2010
;
17
:
400
11
.
52.
Cerchietti
LC
,
Yang
SN
,
Shaknovich
R
,
Hatzi
K
,
Polo
JM
,
Chadburn
A
, et al
A peptomimetic inhibitor of BCL6 with potent antilymphoma effects in vitro and in vivo
.
Blood
2009
;
113
:
3397
405
.
53.
Su
AI
,
Wiltshire
T
,
Batalov
S
,
Lapp
H
,
Ching
KA
,
Block
D
, et al
A gene atlas of the mouse and human protein-encoding transcriptomes
.
Proc Natl Acad Sci U S A
2004
;
101
:
6062
7
.
54.
Novershtern
N
,
Subramanian
A
,
Lawton
LN
,
Mak
RH
,
Haining
WN
,
McConkey
ME
, et al
Densely interconnected transcriptional circuits control cell states in human hematopoiesis
.
Cell
2011
;
144
:
296
309
.
55.
Gentles
AJ
,
Plevritis
SK
,
Majeti
R
,
Alizadeh
AA
. 
Association of a leukemic stem cell gene expression signature with clinical outcomes in acute myeloid leukemia
.
JAMA
2010
;
304
:
2706
15
.
56.
Eppert
K
,
Takenaka
K
,
Lechman
ER
,
Waldron
L
,
Nilsson
B
,
van Galen
P
, et al
Stem cell gene expression programs influence clinical outcome in human leukemia
.
Nat Med
2010
;
17
:
1086
93
.
57.
Konopleva
MY
,
Jordan
CT
. 
Leukemia stem cells and microenvironment: biology and therapeutic targeting
.
J Clin Oncol
2011
;
29
:
591
9
.
58.
Gilbert
LA
,
Hemann
MT
. 
DNA damage-mediated induction of a chemoresistant niche
.
Cell
2010
;
143
:
355
66
.
59.
Saito
Y
,
Kitamura
H
,
Hijikata
A
,
Tomizawa-Murasawa
M
,
Tanaka
S
,
Takagi
S
, et al
Identification of therapeutic targets for quiescent, chemotherapy-resistant human leukemia stem cells
.
Sci Transl Med
2010
;
2
:
17ra9
.
60.
Ishikawa
F
,
Yoshida
S
,
Saito
Y
,
Hijikata
A
,
Kitamura
H
,
Tanaka
S
, et al
Chemotherapy-resistant human AML stem cells home to and engraft within the bone-marrow endosteal region
.
Nat Biotechnol
2007
;
25
:
1315
21
.
61.
Schnell
R
,
Vitetta
E
,
Schindler
J
,
Borchmann
P
,
Barth
S
,
Ghetie
V
, et al
Treatment of refractory Hodgkin's lymphoma patients with an anti-CD25 ricin A-chain immunotoxin
.
Leukemia
2000
;
14
:
129
35
.
62.
Schnell
R
,
Borchmann
P
,
Staak
JO
,
Schindler
J
,
Ghetie
V
,
Vitetta
ES
, et al
Clinical evaluation of ricin A-chain immunotoxins in patients with Hodgkin's lymphoma
.
Ann Oncol
2003
;
14
:
729
36
.
63.
Montero
JC
,
Seoane
S
,
Ocana
A
,
Pandiella
A
. 
Inhibition of SRC family kinases and receptor tyrosine kinases by dasatinib: possible combinations in solid tumors
.
Clin Cancer Res
2011
;
17
:
5546
52
.
64.
Cierpicki
T
,
Risner
LE
,
Grembecka
J
,
Lukasik
SM
,
Popovic
R
,
Omonkowska
M
, et al
Structure of the MLL CXXC domain-DNA complex and its functional role in MLL-AF9 leukemia
.
Nat Struct Mol Biol
2010
;
17
:
62
8
.
65.
Erfurth
FE
,
Popovic
R
,
Grembecka
J
,
Cierpicki
T
,
Theisler
C
,
Xia
ZB
, et al
MLL protects CpG clusters from methylation within the Hoxa9 gene, maintaining transcript expression
.
Proc Natl Acad Sci U S A
2008
;
105
:
7517
22
.
66.
Akalin
A
,
Garrett-Bakelman
FE
,
Kormaksson
M
,
Busuttil
J
,
Zhang
L
,
Khrebtukova
I
, et al
Base-pair resolution DNA methylation sequencing reveals profoundly divergent epigenetic landscapes in acute myeloid leukemia
.
PLoS Genet
2012
;
8
:
e1002781
.
67.
Vire
E
,
Brenner
C
,
Deplus
R
,
Blanchon
L
,
Fraga
M
,
Didelot
C
, et al
The Polycomb group protein EZH2 directly controls DNA methylation
.
Nature
2006
;
439
:
871
4
.
68.
Schuettengruber
B
,
Chourrout
D
,
Vervoort
M
,
Leblanc
B
,
Cavalli
G
. 
Genome regulation by polycomb and trithorax proteins
.
Cell
2007
;
128
:
735
45
.
69.
Armstrong
SA
,
Staunton
JE
,
Silverman
LB
,
Pieters
R
,
den Boer
ML
,
Minden
MD
, et al
MLL translocations specify a distinct gene expression profile that distinguishes a unique leukemia
.
Nat Genet
2002
;
30
:
41
7
.
70.
Ranuncolo
SM
,
Polo
JM
,
Dierov
J
,
Singer
M
,
Kuo
T
,
Greally
J
, et al
Bcl-6 mediates the germinal center B cell phenotype and lymphomagenesis through transcriptional repression of the DNA-damage sensor ATR
.
Nat Immunol
2007
;
8
:
705
14
.
71.
Ranuncolo
SM
,
Wang
L
,
Polo
JM
,
Dell'Oso
T
,
Dierov
J
,
Gaymes
TJ
, et al
BCL6-mediated attenuation of DNA damage sensing triggers growth arrest and senescence through a p53-dependent pathway in a cell context-dependent manner
.
J Biol Chem
2008
;
283
:
22565
72
.
72.
Phan
RT
,
Dalla-Favera
R
. 
The BCL6 proto-oncogene suppresses p53 expression in germinal-centre B cells
.
Nature
2004
;
432
:
635
9
.
73.
Duy
C
,
Yu
JJ
,
Nahar
R
,
Swaminathan
S
,
Kweon
SM
,
Polo
JM
, et al
BCL6 is critical for the development of a diverse primary B cell repertoire
.
J Exp Med
2010
;
207
:
1209
21
.
74.
Cattoretti
G
,
Pasqualucci
L
,
Ballon
G
,
Tam
W
,
Nandula
SV
,
Shen
Q
, et al
Deregulated BCL6 expression recapitulates the pathogenesis of human diffuse large B cell lymphomas in mice
.
Cancer Cell
2005
;
7
:
445
55
.
75.
Duy
C
,
Hurtz
C
,
Shojaee
S
,
Cerchietti
L
,
Geng
H
,
Swaminathan
S
, et al
BCL6 enables Ph+ acute lymphoblastic leukaemia cells to survive BCR–ABL1 kinase inhibition
.
Nature
2011
;
473
:
384
8
.
76.
Figueroa
ME
,
Reimers
M
,
Thompson
RF
,
Ye
K
,
Li
Y
,
Selzer
RR
, et al
An integrative genomic and epigenomic approach for the study of transcriptional regulation
.
PLoS One
2008
;
3
:
e1882
.
77.
Khulan
B
,
Thompson
RF
,
Ye
K
,
Fazzari
MJ
,
Suzuki
M
,
Stasiek
E
, et al
Comparative isoschizomer profiling of cytosine methylation: the HELP assay
.
Genome Res
2006
;
16
:
1046
55
.
78.
2008 R Development Core Team
: 
A language and environment for statistical computing
.
Vienna (Austria)
; 
2008
.
[cited 2012 Oct 17]. Available from
: http://www.R-project.org.
79.
Ehrich
M
,
Nelson
MR
,
Stanssens
P
,
Zabeau
M
,
Liloglou
T
,
Xinarianos
G
, et al
Quantitative high-throughput analysis of DNA methylation patterns by base-specific cleavage and mass spectrometry
.
Proc Natl Acad Sci U S A
2005
;
102
:
15785
90
.
80.
Collins
TJ
. 
ImageJ for microscopy
.
Biotechniques
2007
;
43
(
1 Suppl
)
25
30
.
81.
Milne
TA
,
Zhao
K
,
Hess
JL
. 
Chromatin immunoprecipitation (ChIP) for analysis of histone modifications and chromatin-associated proteins
.
Methods Mol Biol
2009
;
538
:
409
23
.