Genomic Classification and Prognosis in Acute Myeloid Leukemia

Authors

Elli Papaemmanuil

Moritz Gerstung

Lars Bullinger

Verena I. Gaidzik

Peter Paschka

Nicola D. Roberts

Nicola E. Potter

Michael Heuser

Felicitas Thol

Niccolo Bolli

Gunes Gundem

Peter Van Loo

Inigo Martincorena

Peter Ganly

Laura Mudie

Stuart McLaren

Sarah O’Meara

Keiran Raine

David R. Jones

Jon W. Teague

Adam P. Butler

Mel F. Greaves

Arnold Ganser

Konstanze Döhner

Richard F. Schlenk

Hartmut Döhner

Peter J. Campbell

Doi

10.1056/NEJMoa1516192

PMID: 27276561 · DOI: 10.1056/NEJMoa1516192 · Journal: New England Journal of Medicine (2016)

TL;DR

Papaemmanuil and colleagues sequenced 111 cancer genes plus cytogenetics in 1540 adults with acute myeloid leukemia (AML) enrolled in three German–Austrian AMLSG intensive-therapy trials, identifying 5234 driver mutations across 76 genes/regions and partitioning the cohort into 11 mutually exclusive genomic subgroups. Beyond established WHO categories, three new categories emerged — chromatin–spliceosome AML (18%), TP53–aneuploidy AML (13%), and provisionally IDH2 R172 AML (1%) — with the chromatin–spliceosome and TP53–aneuploidy groups carrying poor prognoses. Co-occurring driver mutations and gene–gene interactions (notably the three-way NPM1–DNMT3A–FLT3 ITD interaction) substantially modified outcome, and the classification replicated in the independent TCGA AML cohort (laml_tcga_pub).

Cohort & data

  • N = 1540 adults with AML enrolled in three prospective AMLSG intensive-therapy trials: AML-HD98A (ages 18–65, ICE induction), AMLSG-07-04 (ages 18–61, ICE ± all-trans retinoic acid), and AML-HD98B (ages 58–84, ICE ± ATRA). Median follow-up 5.9 years. ClinicalTrials.gov NCT00146120.
  • Cancer type: AML (de novo and secondary).
  • Genomic data: sequencing of 111 leukemia genes plus cytogenetics (targeted-dna-seq); driver criteria per Lawrence et al. 2013. Sequencing data deposited at EGA accession EGAS00001000275.
  • Validation cohort: TCGA AML (laml_tcga_pub) — independent cohort with many older patients used to replicate subgroup structure and prognostic findings.
  • Treatments studied: cytarabine, idarubicin, and etoposide (ICE induction); tretinoin (ATRA) added in two trial arms; allogeneic stem-cell transplant for high-risk patients.

Key findings

  • 5234 driver mutations across 76 genes or genomic regions in 1540 patients; ≥1 driver in 1478/1540 (96%) and ≥2 drivers in 86% of patients. Point mutations accounted for 73% (3824/5234) of drivers (PMID:27276561).
  • Eleven mutually exclusive genomic subgroups were defined by Bayesian Dirichlet-process modeling. 1236/1540 (80%) patients fell unambiguously into one subgroup, 56 (4%) met criteria for ≥2 categories, and 166 (11%) with driver mutations remained unclassified (PMID:27276561).
  • NPM1-mutated AML was the largest subgroup (27% of cohort, 418 patients); 73% of NPM1-mutated cases (319/436) carried co-mutations in DNA methylation/hydroxymethylation genes (DNMT3A, IDH1, IDH2 R140, or TET2).
  • Chromatin–spliceosome subgroup (18%, n=275) — newly defined; mutations in RNA-splicing (SRSF2, SF3B1, U2AF1, ZRSR2), chromatin (ASXL1, STAG2, BCOR, KMT2A PTD, EZH2, PHF6), or transcription (RUNX1). Patients were older with lower WBC/blast counts, lower induction response, higher relapse, and poor long-term outcomes; 84% (232/275) would be classified intermediate-risk under ELN guidelines despite outcomes resembling adverse-risk groups.
  • TP53–aneuploidy subgroup (13%, n=199) — newly defined; characterized by TP53 mutations, complex karyotype, or chromosomal aneuploidies (−5/5q, −7/7q, −17/17p, +8/8q, etc.). TP53 mutation and complex karyotype contributed independently and additively to poor survival.
  • Provisional IDH2 R172 subgroup (1%, n=18) — mutually exclusive with NPM1 (OR 0.06, P=4×10⁻⁵), unlike IDH2 R140 which co-occurs strongly with NPM1 (OR 3.6, P=5×10⁻¹⁰). Long-term outcomes broadly similar to NPM1-mutated AML.
  • Three-way NPM1–DNMT3A–FLT3 ITD interaction: the deleterious effect of FLT3 ITD on survival was strongest only when NPM1 and DNMT3A were also mutated (n=93, 6% of cohort; HR 1.5 [1.2–1.9], q=0.004 for three-way interaction). Independent of FLT3 mutant-allele ratio.
  • NPM1–DNMT3A–NRAS G12/13 carried an unexpectedly favorable prognosis (n=45, P=0.0007 three-way), refining prior reports that “NPM1–NRAS” is favorable.
  • KMT2A PTD × FLT3 TKD co-occurrence (n=10) was strongly adverse (q=0.008); DNMT3A × IDH2 R140 co-occurrence (n=19) was also adverse (q=0.05).
  • Survival concordance ~71% with the full multivariate model vs. ~64% using only ELN variables. Genomic features explained ~two-thirds of explained variation; gene–gene interactions accounted for 11% of explained variance.
  • BRAF mutations were independently associated with worse prognosis (P=0.009, q=0.06); the authors suggest BRAF inhibitors as a candidate therapy in this small subgroup.
  • Clonal-evolution patterns: epigenetic-modifier mutations (DNMT3A, ASXL1, IDH1, IDH2, TET2) were typically founding/early events; RTK–RAS pathway mutations were typically late and frequently subclonal/recurrent within a patient. NPM1 was a late driver, often secondary to DNMT3A, IDH1, or NRAS.
  • WHO 2008 fusion-defined classes (PML–RARA, CBFB–MYH11, RUNX1–RUNX1T1, DEK–NUP214, inv(3)/MECOM, KMT2A fusions including MLLT3–KMT2A) were each ≤5% of the cohort and confirmed; CEBPA biallelic was a distinct subgroup (4%, n=66).

Genes & alterations

  • NPM1 — mutations in 28% (n=436); class-defining for largest subgroup; favorable as monotherapy but outcome strongly modified by co-occurring FLT3, DNMT3A, IDH2, NRAS, PTPN11, and chromatin–spliceosome mutations.
  • FLT3 — ITD in 22% (n=341); deleterious effect strongest in NPM1+/DNMT3A+ context. FLT3 TKD distinct from FLT3 ITD in co-mutation patterns and interactions (e.g. with KMT2A PTD).
  • DNMT3A — frequent founding-clone driver; modifies effect of FLT3 ITD, NRAS G12/13, IDH2 R140, and RAD21 mutations.
  • TP53 — 6% (n=98); core of TP53–aneuploidy subgroup; HR 1.7 (1.4–2.2), P=7×10⁻⁶, q=0.0002. Independent and additive with complex karyotype.
  • IDH2 — R140 (co-occurs with NPM1, OR 3.6) vs. R172 (mutually exclusive with NPM1, OR 0.06) define functionally distinct hotspots; R172 forms a provisional subgroup with metabolic/methylation profile distinct from other IDH mutants.
  • IDH1 — early/founding driver in epigenetic axis; co-mutes with NPM1.
  • TET2, ASXL1 — early epigenetic drivers; ASXL1 part of chromatin–spliceosome group with independent adverse effect.
  • SRSF2 — splicing factor in chromatin–spliceosome group; HR 1.4 (1.1–1.7), P=0.003, q=0.03. Additive adverse effect with ASXL1.
  • SF3B1, U2AF1, ZRSR2 — additional splicing factors defining chromatin–spliceosome subgroup.
  • STAG2, BCOR, KMT2A PTD, EZH2, PHF6 — chromatin regulators in chromatin–spliceosome subgroup.
  • RUNX1 — transcription regulator in chromatin–spliceosome subgroup.
  • CEBPA — biallelic mutations define favorable subgroup (4%, n=66); HR 0.6 (0.4–0.7), P=4×10⁻⁵.
  • NRAS — late driver in RTK–RAS pathway; G12/13 vs. Q61 hotspots have distinct co-mutation patterns; G12/13 in NPM1/DNMT3A context paradoxically favorable.
  • BRAF — 1% (n=9); independently adverse (P=0.009, q=0.06); proposed candidate for BRAF inhibitor therapy.
  • MYC — novel hotspot mutations described in the cohort (Supplementary Results S1).
  • PML–RARA t(15;17) — 4% (n=60); class-defining; HR 0.3 (0.2–0.4).
  • CBFB–MYH11 inv(16)/t(16;16) — 5% (n=81); HR 0.3 (0.2–0.4).
  • RUNX1–RUNX1T1 t(8;21) — 4% (n=60).
  • DEK–NUP214 t(6;9) — 1% (n=15).
  • inv(3)/MECOM — 1% (n=20); among the strongest adverse main effects (HR 2.9 [1.8–4.7], P=9×10⁻⁶, q=0.0003).
  • KMT2A fusions — 3% (n=44); multiple partners; MLLT3–KMT2A (t(9;11)) appears favorable relative to other KMT2A fusions.

Clinical implications

  • Disease classification reform: the authors argue genomic subgroups should supplement WHO morphology-based classification. Three new genomic categories (chromatin–spliceosome, TP53–aneuploidy, provisional IDH2 R172) capture 32% of the cohort that was previously unclassified by WHO 2008 lesions.
  • Prognostic stratification beyond ELN: the full genomic+clinical model achieves ~71% concordance for overall survival vs. ~64% with ELN variables alone. The authors recommend incorporating TP53, SRSF2, ASXL1, DNMT3A, and IDH2 into prognostic guidelines, and adding RUNX1, ASXL1, and KMT2A PTD evaluation at diagnosis to identify chromatin–spliceosome patients.
  • Risk reclassification: 84% of chromatin–spliceosome patients (232/275) currently classified as ELN intermediate-risk would be more appropriately treated as adverse-risk based on observed outcomes.
  • Context-dependent prognosis for NPM1-mutated AML: outcome depends on co-mutation context — favorable with NPM1+DNMT3A+NRAS G12/13, adverse with NPM1+DNMT3A+FLT3 ITD. Implies allogeneic transplant decisions in NPM1 AML should integrate co-mutation profiles.
  • Therapeutic candidates: BRAF inhibitors proposed for BRAF-mutant AML (small subgroup but independently adverse); FLT3 and RAS-pathway inhibitors expected to alter outcome predictions for those subgroups.
  • Boundary with chronic myeloid disorders: chromatin–spliceosome AML overlaps mutationally with high-risk myeloproliferative neoplasms and MDS, suggesting shared evolutionary trajectories and a possible reorganization of acute/chronic myeloid disease boundaries.

Limitations & open questions

  • Cohort skewed to younger, intensively treated patients: trials enrolled adults selected for intensive induction. Older AML patients ineligible for intensive therapy are underrepresented; the TCGA replication cohort partially addresses this.
  • Prospective validation needed: the authors explicitly state predictions from co-mutation contexts (especially within NPM1 AML) require validation in prospective clinical trials.
  • 166 unclassified patients (11%) with driver mutations — may harbor mutations in genes not on the 111-gene panel, or unrecognized class-defining lesions.
  • Provisional IDH2 R172 subgroup is small (n=18) and the case for it as an independent class rests on co-mutation structure and prior metabolic/methylation evidence; larger cohorts are needed.
  • Treatment-specific outcomes not modeled: the prognostic model averages over induction regimen and post-remission therapy choices, limiting ability to define class-specific optimal therapy.
  • Open question for cross-paper synthesis: how does the chromatin–spliceosome AML class relate genetically and clinically to MDS and high-risk MPN with the same mutated genes? See MDS.

Citations from this paper used in the wiki

  • “We identified 5234 driver mutations across 76 genes or genomic regions, with 2 or more drivers identified in 86% of the patients” (Abstract).
  • “Patterns of co-mutation compartmentalized the cohort into 11 classes, each with distinct diagnostic features and clinical outcomes” (Abstract).
  • “AML with mutations in genes encoding chromatin, RNA-splicing regulators, or both (in 18% of patients); AML with TP53 mutations, chromosomal aneuploidies, or both (in 13%); and, provisionally, AML with IDH2R172 mutations (in 1%)” (Abstract).
  • “the deleterious effect of FLT3ITD was most clinically relevant in patients with concomitant NPM1 and DNMT3A mutations (P=0.009 for three-way interaction in the univariate analysis, q=0.004 in the multivariate analysis)” (Results, Influence of Complex Gene Interactions).
  • “Using the full model, we could correctly rank approximately 71% of patients for overall survival (vs. 64% with models using only variables in the European LeukemiaNet criteria)” (Results, Influence of Co-occurring Mutations).
  • “BRAF mutations are independently associated with a worse prognosis (P=0.009, q=0.06), and BRAF inhibitors might be a useful therapeutic option for patients in this subgroup” (Results).
  • “84% of patients in this subgroup (232 of 275) would be classified as being at intermediate risk, whereas their outcomes are in fact similar to those for patients in subgroups of AML with adverse outcomes” (Results, Clinical Implications).
  • “IDH2R140 mutations, which show strong co-mutation with NPM1 (odds ratio for co-mutation, 3.6; P= 5×10−10), IDH2R172 mutations are mutually exclusive with NPM1 (odds ratio for co-mutation, 0.06; P= 4×10−5)” (Results).

This page was processed by crosslinker on 2026-05-14.