Machine learning integrates genomic signatures for subclassification beyond primary and secondary acute myeloid leukemia.

Although genomic alterations drive the pathogenesis of acute myeloid leukemia (AML), traditional classifications are largely based on morphology, and prototypic genetic founder lesions define only a small proportion of AML patients. The historical subdivision of primary/de novo AML and secondary AML has shown to variably correlate with genetic patterns. The combinatorial complexity and heterogeneity of AML genomic architecture may have thus far precluded genomic-based subclassification to identify distinct molecularly defined subtypes more reflective of shared pathogenesis. We integrated cytogenetic and gene sequencing data from a multicenter cohort of 6788 AML patients that were analyzed using standard and machine learning methods to generate a novel AML molecular subclassification with biologic correlates corresponding to underlying pathogenesis. Standard supervised analyses resulted in modest cross-validation accuracy when attempting to use molecular patterns to predict traditional pathomorphologic AML classifications. We performed unsupervised analysis by applying the Bayesian latent class method that identified 4 unique genomic clusters of distinct prognoses. Invariant genomic features driving each cluster were extracted and resulted in 97% cross-validation accuracy when used for genomic subclassification. Subclasses of AML defined by molecular signatures overlapped current pathomorphologic and clinically defined AML subtypes. We internally and externally validated our results and share an open-access molecular classification scheme for AML patients. Although the heterogeneity inherent in the genomic changes across nearly 7000 AML patients was too vast for traditional prediction methods, machine learning methods allowed for the definition of novel genomic AML subclasses, indicating that traditional pathomorphologic definitions may be less reflective of overlapping pathogenesis.

Machine learning demonstrates that somatic mutations imprint invariant morphologic features in myelodysplastic syndromes.

Morphologic interpretation is the standard in diagnosing myelodysplastic syndrome (MDS), but it has limitations, such as varying reliability in pathologic evaluation and lack of integration with genetic data. Somatic events shape morphologic features, but the complexity of morphologic and genetic changes makes clear associations challenging. This article interrogates novel clinical subtypes of MDS using a machine-learning technique devised to identify patterns of cooccurrence among morphologic features and genomic events. We sequenced 1079 MDS patients and analyzed bone marrow morphologic alterations and other clinical features. A total of 1929 somatic mutations were identified. Five distinct morphologic profiles with unique clinical characteristics were defined. Seventy-seven percent of higher-risk patients clustered in profile 1. All lower-risk (LR) patients clustered into the remaining 4 profiles: profile 2 was characterized by pancytopenia, profile 3 by monocytosis, profile 4 by elevated megakaryocytes, and profile 5 by erythroid dysplasia. These profiles could also separate patients with different prognoses. LR MDS patients were classified into 8 genetic signatures (eg, signature A had TET2 mutations, signature B had both TET2 and SRSF2 mutations, and signature G had SF3B1 mutations), demonstrating association with specific morphologic profiles. Six morphologic profiles/genetic signature associations were confirmed in a separate analysis of an independent cohort. Our study demonstrates that nonrandom or even pathognomonic relationships between morphology and genotype to define clinical features can be identified. This is the first comprehensive implementation of machine-learning algorithms to elucidate potential intrinsic interdependencies among genetic lesions, morphologies, and clinical prognostic in attributes of MDS.

Molecular landscape and clonal architecture of adult myelodysplastic/myeloproliferative neoplasms.

More than 90% of patients with myelodysplastic/myeloproliferative neoplasms (MDSs/MPNs) harbor somatic mutations in myeloid-related genes, but still, current diagnostic criteria do not include molecular data. We performed genome-wide sequencing techniques to characterize the mutational landscape of a large and clinically well-characterized cohort including 367 adults with MDS/MPN subtypes, including chronic myelomonocytic leukemia (CMML; n = 119), atypical chronic myeloid leukemia (aCML; n = 71), MDS/MPN with ring sideroblasts and thrombocytosis (MDS/MPN-RS-T; n = 71), and MDS/MPN unclassifiable (MDS/MPN-U; n = 106). A total of 30 genes were recurrently mutated in ≥3% of the cohort. Distribution of recurrently mutated genes and clonal architecture differed among MDS/MPN subtypes. Statistical analysis revealed significant correlations between recurrently mutated genes, as well as genotype-phenotype associations. We identified specific gene combinations that were associated with distinct MDS/MPN subtypes and that were mutually exclusive with most of the other MDSs/MPNs (eg, TET2-SRSF2 in CMML, ASXL1-SETBP1 in aCML, and SF3B1-JAK2 in MDS/MPN-RS-T). Patients with MDS/MPN-U were the most heterogeneous and displayed different molecular profiles that mimicked the ones observed in other MDS/MPN subtypes and that had an impact on the outcome of the patients. Specific gene mutations also had an impact on the outcome of the different MDS/MPN subtypes, which may be relevant for clinical decision-making. Overall, the results of this study help to elucidate the heterogeneity found in these neoplasms, which can be of use in the clinical setting of MDS/MPN.

Extended experience with a non-cytotoxic DNMT1-targeting regimen of decitabine to treat myeloid malignancies.

The nucleoside analogue decitabine can deplete the epigenetic regulator DNA methyltransferase 1 (DNMT1), an effect that occurs, and is saturated at, low concentrations/doses. A reason to pursue this molecular-targeted effect instead of the DNA damage/cytotoxicity produced with high concentrations/doses, is that non-cytotoxic DNMT1-depletion can cytoreduce even p53-null myeloid malignancies while sparing normal haematopoiesis. We thus identified minimum doses of decitabine (0·1-0·2 mg/kg) that deplete DNMT1 without off-target anti-metabolite effects/cytotoxicity, and then administered these well-tolerated doses frequently 1-2X/week to increase S-phase dependent DNMT1-depletion, and used a Myeloid Malignancy Registry to evaluate long-term outcomes in 69 patients treated this way. Consistent with the scientific rationale, treatment was well-tolerated and durable responses were produced (~40%) in genetically heterogeneous disease and the very elderly.

Impact of germline CTC1 alterations on telomere length in acquired bone marrow failure.

Compound heterozygous germline mutations in CTC1 gene have been found in patients with atypical dyskeratosis congenita (DC), whereas heterozygous carriers are unaffected. Through screening of a large cohort of adult patients with acquired bone marrow failure syndromes, in addition to a DC case, we have also found extremely rare or novel heterozygous deleterious germline variants of CTC1 in patients with aplastic anaemia (AA; n = 5), paroxysmal nocturnal haemoglobinuria (PNH; n = 3) and myelodysplastic syndrome (MDS; n = 2). A compound heterozygous case of AA showed clonal evolution. Our results suggest that some of the inherited CTC1 variants may represent predisposition factors for acquired bone marrow failure.

Amplified EPOR/JAK2 Genes Define a Unique Subtype of Acute Erythroid Leukemia.

Acute erythroid leukemia (AEL) is a unique subtype of acute myeloid leukemia characterized by prominent erythroid proliferation whose molecular basis is poorly understood. To elucidate the underlying mechanism of erythroid proliferation, we analyzed 121 AEL using whole-genome, whole-exome, and/or targeted-capture sequencing, together with transcriptome analysis of 21 AEL samples. Combining publicly available sequencing data, we found a high frequency of gains and amplifications involving EPOR/JAK2 in TP53-mutated cases, particularly those having >80% erythroblasts designated as pure erythroid leukemia (10/13). These cases were frequently accompanied by gains and amplifications of ERG/ETS2 and associated with a very poor prognosis, even compared with other TP53-mutated AEL. In addition to activation of the STAT5 pathway, a common feature across all AEL cases, these AEL cases exhibited enhanced cell proliferation and heme metabolism and often showed high sensitivity to ruxolitinib in vitro and in xenograft models, highlighting a potential role of JAK2 inhibition in therapeutics of AEL. SIGNIFICANCE: This study reveals the major role of gains, amplifications, and mutations of EPOR and JAK2 in the pathogenesis of pure erythroleukemia. Their frequent response to ruxolitinib in patient-derived xenograft and cell culture models highlights a possible therapeutic role of JAK2 inhibition for erythroleukemia with EPOR/JAK2-involving lesions. This article is highlighted in the In This Issue feature, p. 369.

Somatic mutations in lymphocytes in patients with immune-mediated aplastic anemia.

The prevalence and functional impact of somatic mutations in nonleukemic T cells is not well characterized, although clonal T-cell expansions are common. In immune-mediated aplastic anemia (AA), cytotoxic T-cell expansions are shown to participate in disease pathogenesis. We investigated the mutation profiles of T cells in AA by a custom panel of 2533 genes. We sequenced CD4+ and CD8+ T cells of 24 AA patients and compared the results to 20 healthy controls and whole-exome sequencing of 37 patients with AA. Somatic variants were common both in patients and healthy controls but enriched to AA patients' CD8+ T cells, which accumulated most mutations on JAK-STAT and MAPK pathways. Mutation burden was associated with CD8+ T-cell clonality, assessed by T-cell receptor beta sequencing. To understand the effect of mutations, we performed single-cell sequencing of AA patients carrying STAT3 or other mutations in CD8+ T cells. STAT3 mutated clone was cytotoxic, clearly distinguishable from other CD8+ T cells, and attenuated by successful immunosuppressive treatment. Our results suggest that somatic mutations in T cells are common, associate with clonality, and can alter T-cell phenotype, warranting further investigation of their role in the pathogenesis of AA.

Clinical and basic implications of dynamic T cell receptor clonotyping in hematopoietic cell transplantation.

TCR repertoire diversification constitutes a foundation for successful immune reconstitution after allogeneic hematopoietic cell transplantation (allo-HCT). Deep TCR Vß sequencing of 135 serial specimens from a cohort of 35 allo-HCT recipients/donors was performed to dissect posttransplant TCR architecture and dynamics. Paired analysis of clonotypic repertoires showed a minimal overlap with donor expansions. Rarefied and hyperexpanded clonotypic patterns were hallmarks of T cell reconstitution and influenced clinical outcomes. Donor and pretransplant TCR diversity as well as divergence of class I human leukocyte antigen genotypes were major predictors of recipient TCR repertoire recovery. Complementary determining region 3-based specificity spectrum analysis indicated a predominant expansion of pathogen- and tumor-associated clonotypes in the late post-allo-HCT phase, while autoreactive clones were more expanded in the case of graft-versus-host disease occurrence. These findings shed light on post-allo-HCT adaptive immune reconstitution processes and possibly help in tracking alloreactive responses.

Complex landscape of alternative splicing in myeloid neoplasms.

Myeloid neoplasms are characterized by frequent mutations in at least seven components of the spliceosome that have distinct roles in the process of pre-mRNA splicing. Hotspot mutations in SF3B1, SRSF2, U2AF1 and loss of function mutations in ZRSR2 have revealed widely different aberrant splicing signatures with little overlap. However, previous studies lacked the power necessary to identify commonly mis-spliced transcripts in heterogeneous patient cohorts. By performing RNA-Seq on bone marrow samples from 1258 myeloid neoplasm patients and 63 healthy bone marrow donors, we identified transcripts frequently mis-spliced by mutated splicing factors (SF), rare SF mutations with common alternative splicing (AS) signatures, and SF-dependent neojunctions. We characterized 17,300 dysregulated AS events using a pipeline designed to predict the impact of mis-splicing on protein function. Meta-splicing analysis revealed a pattern of reduced levels of retained introns among disease samples that was exacerbated in patients with splicing factor mutations. These introns share characteristics with "detained introns," a class of introns that have been shown to promote differentiation by detaining pro-proliferative transcripts in the nucleus. In this study, we have functionally characterized 17,300 targets of mis-splicing by the SF mutations, identifying a common pathway by which AS may promote maintenance of a proliferative state.

A Therapeutic Strategy for Preferential Targeting of TET2 Mutant and TET-dioxygenase Deficient Cells in Myeloid Neoplasms.

TET2 is frequently mutated in myeloid neoplasms. Genetic TET2 deficiency leads to skewed myeloid differentiation and clonal expansion, but minimal residual TET activity is critical for survival of neoplastic progenitor and stem cells. Consistent with mutual exclusivity of TET2 and neomorphic IDH1/2 mutations, here we report that IDH1/2 mutant-derived 2-hydroxyglutarate is synthetically lethal to TET-dioxygenase deficient cells. In addition, a TET-selective small molecule inhibitor decreased cytosine hydroxymethylation and restricted clonal outgrowth of TET2 mutant, but not normal hematopoietic precursor cells in vitro and in vivo. While TET-inhibitor phenocopied somatic TET2 mutations, its pharmacologic effects on normal stem cells were, unlike mutations, reversible. Treatment with TET inhibitor suppressed the clonal evolution of TET2 mutant cells in murine models and TET2-mutated human leukemia xenografts. These results suggest that TET inhibitors may constitute a new class of targeted agents in TET2 mutant neoplasia.

Influence of Killer Immunoglobulin-Like Receptors and Somatic Mutations on Transplant Outcomes in Acute Myeloid Leukemia.

Natural killer (NK) cells are regulated by killer immunoglobulin-like receptor (KIR) interactions with human leukocyte antigen class I ligands. Various models of NK cell alloreactivity have been associated with outcomes after allogeneic hematopoietic cell transplant (alloHCT), but results have varied widely. We hypothesized that somatic mutations in acute myeloid leukemia (AML) in the context of KIR profiles may further refine their association with transplant outcomes. In this single-center, retrospective, observational study, 81 AML patients who underwent matched-related donor alloHCT were included. Post-HCT outcomes were assessed based on mutational status and KIR profiles with the Kaplan-Meier method and log-rank test. On multivariable analysis those with any somatic mutations and C1/C2 heterozygosity had less acute graft-versus-host disease (GvHD) (hazard ratio [HR], 0.32; 95% confidence interval [CI], 0.14-0.75; P = .009), more relapse (HR, 3.02; 95% CI, 1.30-7.01; P = .010), inferior relapse-free survival (RFS; HR, 2.22; 95% CI, 1.17-4.20; P = .014), and overall survival (OS; HR, 2.21; 95% CI, 1.17-4.20; P = .015), whereas those with a missing KIR ligand had superior RFS (HR, 0.53; 95% CI, 0.30-0.94; P = .031). The presence of a somatic mutation and donor haplotype A was also associated with less acute GvHD (HR, 0.38; 95% CI, 0.16-0.92; P = .032), more relapse (HR, 2.72; 95% CI, 1.13-6.52; P = .025), inferior RFS (HR, 2.11; 95% CI, 1.07-4.14; P = .030), and OS (HR, 2.20; 95% CI, 1.11-4.38; P = .024). Enhanced NK cell alloreactivity from more KIR activating signals (donor B haplotype) and fewer inhibitory signals (recipient missing KIR ligand or C1 or C2 homozygosity) may help mitigate the adverse prognosis associated with some AML somatic mutations. These results may have implications for improving patient risk stratification prior to transplant and optimizing donor selection.

Context dependent effects of ascorbic acid treatment in TET2 mutant myeloid neoplasia.

Loss-of-function TET2 mutations (TET2MT) are common in myeloid neoplasia. TET2, a DNA dioxygenase, requires 2-oxoglutarate and Fe(II) to oxidize 5-methylcytosine. TET2MT thus result in hypermethylation and transcriptional repression. Ascorbic acid (AA) increases dioxygenase activity by facilitating Fe(III)/Fe(II) redox reaction and may alleviate some biological consequences of TET2MT by restoring dioxygenase activity. Here, we report the utility of AA in the prevention of TET2MT myeloid neoplasia (MN), clarify the mechanistic underpinning of the TET2-AA interactions, and demonstrate that the ability of AA to restore TET2 activity in cells depends on N- and C-terminal lysine acetylation and nature of TET2MT. Consequently, pharmacologic modulation of acetyltransferases and histone deacetylases may regulate TET dioxygenase-dependent AA effects. Thus, our study highlights the contribution of factors that may enhance or attenuate AA effects on TET2 and provides a rationale for novel therapeutic approaches including combinations of AA with class I/II HDAC inhibitor or sirtuin activators in TET2MT leukemia.

Subclonal STAT3 mutations solidify clonal dominance.

T large granular lymphocyte leukemia (T-LGLL) is a clonal lymphoproliferative disorder that can arise in the context of pathologic or physiologic cytotoxic T-cell (CTL) responses. STAT3 mutations are often absent in typical T-LGLL, suggesting that in a significant fraction of patients, antigen-driven expansion alone can maintain LGL clone persistence. We set out to determine the relationship between activating STAT3 hits and CTL clonal selection at presentation and in response to therapy. Thus, a group of patients with T-LGLL were serially subjected to deep next-generation sequencing (NGS) of the T-cell receptor (TCR) Vß complementarity-determining region 3 (CDR3) and STAT3 to recapitulate clonal hierarchy and dynamics. The results of this complex analysis demonstrate that STAT3 mutations produce either a sweeping or linear subclone within a monoclonal CTL population either early or during the course of disease. Therapy can extinguish a LGL clone, silence it, or adapt mechanisms to escape elimination. LGL clones can persist on elimination of STAT3 subclones, and alternate STAT3-negative CTL clones can replace therapy-sensitive CTL clones. LGL clones can evolve and are fueled by a nonextinguished antigenic drive. STAT3 mutations can accelerate this process or render CTL clones semiautonomous and not reliant on physiologic stimulation.

Distinct clinical and biological implications of CUX1 in myeloid neoplasms.

Somatic mutations of the CUT-like homeobox 1 (CUX1) gene (CUX1 MT) can be found in myeloid neoplasms (MNs), in particular, in myelodysplastic syndromes (MDSs). The CUX1 locus is also deleted in 3 of 4 MN cases with -7/del(7q). A cohort of 1480 MN patients was used to characterize clinical features and clonal hierarchy associated with CUX1 MT and CUX1 deletions (CUX1 DEL) and to analyze their functional consequences in vitro. CUX1 MT were present in 4% of chronic MNs. CUX1 DEL were preferentially found in advanced cases (6%). Most MDS and acute myeloid leukemia (AML) patients with -7/del(7q) and up to 15% of MDS patients and 5% of AML patients diploid for the CUX1 locus exhibited downmodulated CUX1 expression. In 75% of mutant cases, CUX1 MT were heterozygous, whereas microdeletions and homozygous and compound-heterozygous mutations were less common. CUX MT/DEL were associated with worse survival compared with CUX1 WT Within the clonal hierarchy, 1 of 3 CUX1 MT served as founder events often followed by secondary BCOR and ASXL1 subclonal hits, whereas TET2 was the most common ancestral lesion, followed by subclonal CUX1 MT Comet assay of patients' bone marrow progenitor cells and leukemic cell lines performed in various experimental conditions revealed that frameshift mutations, hemizygous deletions, or experimental CUX1 knockdown decrease the repair of oxidized bases. These functional findings may explain why samples with either CUX1 MT or low CUX1 expression coincided with significantly higher numbers of somatic hits by whole-exome sequencing. Our findings implicate the DNA repair dysfunction resulting from CUX1 lesions in the pathogenesis of MNs, in which they lead to a mutator phenotype.

Genomic Biomarkers to Predict Resistance to Hypomethylating Agents in Patients With Myelodysplastic Syndromes Using Artificial Intelligence.

PURPOSE: We developed an unbiased framework to study the association of several mutations in predicting resistance to hypomethylating agents (HMAs) in patients with myelodysplastic syndromes (MDS), analogous to consumer and commercial recommender systems in which customers who bought products A and B are likely to buy C: patients who have a mutation in gene A and gene B are likely to respond or not respond to HMAs. METHODS: We screened a cohort of 433 patients with MDS who received HMAs for the presence of common myeloid mutations in 29 genes that were obtained before the patients started therapy. The association between mutations and response was evaluated by the Apriori market basket analysis algorithm. Rules with the highest confidence (confidence that the association exists) and the highest lift (strength of the association) were chosen. We validated our biomarkers in samples from patients enrolled in the S1117 trial. RESULTS: Among 433 patients, 193 (45%) received azacitidine, 176 (40%) received decitabine, and 64 (15%) received HMA alone or in combination. The median age was 70 years (range, 31 to 100 years), and 28% were female. The median number of mutations per sample was three (range, zero to nine), and 176 patients (41%) had three or more mutations per sample. Association rules identified several genomic combinations as being highly associated with no response. These molecular signatures were present in 30% of patients with three or more mutations/sample with an accuracy rate of 87% in the training cohort and 93% in the validation cohort. CONCLUSION: Genomic biomarkers can identify, with high accuracy, approximately one third of patients with MDS who will not respond to HMAs. This study highlights the importance of machine learning technologies such as the recommender system algorithm in translating genomic data into useful clinical tools.

Invariant patterns of clonal succession determine specific clinical features of myelodysplastic syndromes.

Myelodysplastic syndromes (MDS) arise in older adults through stepwise acquisitions of multiple somatic mutations. Here, analyzing 1809 MDS patients, we infer clonal architecture by using a stringent, the single-cell sequencing validated PyClone bioanalytic pipeline, and assess the position of the mutations within the clonal architecture. All 3,971 mutations are grouped based on their rank in the deduced clonal hierarchy (dominant and secondary). We evaluated how they affect the resultant morphology, progression, survival and response to therapies. Mutations of SF3B1, U2AF1, and TP53 are more likely to be dominant, those of ASXL1, CBL, and KRAS are secondary. Among distinct combinations of dominant/secondary mutations we identified 37 significant relationships, of which 12 affect clinical phenotypes, 5 cooperatively associate with poor prognosis. They also predict response to hypomethylating therapies. The clonal hierarchy has distinct ranking and the resultant invariant combinations of dominant/secondary mutations yield novel insights into the specific clinical phenotype of MDS.