RUNX1 Mutations in Acute Myeloid Leukemia Are Associated with Distinct Clinico-Pathologic and Genetic Features

ORIGINAL ARTICLE RUNX1 mutations in acute myeloid leukemia are associated with distinct clinico-pathologic and genetic features

VI Gaidzik1,13, V Teleanu1,13,EPapaemmanuil2,DWeber1,PPaschka1,JHahn1, T Wallrabenstein1, B Kolbinger1,CHKöhne3,HAHorst4, P Brossart5,GHeld6, A Kündgen7, M Ringhoffer8, K Götze9,MRummel10, M Gerstung2, P Campbell2,JMKraus11, HA Kestler11,FThol12, MHeuser12, B Schlegelberger12,AGanser12, L Bullinger1,RFSchlenk1, K Döhner1 and H Döhner1 for the German-Austrian Acute Myeloid Leukemia Study Group (AMLSG)14

We evaluated the frequency, genetic architecture, clinico-pathologic features and prognostic impact of RUNX1 mutations in 2439 adult patients with newly-diagnosed acute myeloid leukemia (AML). RUNX1 mutations were found in 245 of 2439 (10%) patients; were almost mutually exclusive of AML with recurrent genetic abnormalities; and they co-occurred with a complex pattern of gene mutations, frequently involving mutations in epigenetic modifiers (ASXL1, IDH2, KMT2A, EZH2), components of the spliceosome complex (SRSF2, SF3B1) and STAG2, PHF6, BCOR. RUNX1 mutations were associated with older age (16–59 years: 8.5%; ⩾ 60 years: 15.1%), male gender, more immature morphology and secondary AML evolving from myelodysplastic syndrome. In univariable analyses, RUNX1 mutations were associated with inferior event-free (EFS, Po0.0001), relapse-free (RFS, P = 0.0007) and overall survival (OS, Po0.0001) in all patients, remaining significant when age was considered. In multivariable analysis, RUNX1 mutations predicted for inferior EFS (P = 0.01). The effect of co-mutation varied by partner gene, where patients with the secondary genotypes RUNX1mut/ASXL1mut (OS, P = 0.004), RUNX1mut/SRSF2mut (OS, P = 0.007) and RUNX1mut/PHF6mut (OS, P = 0.03) did significantly worse, whereas patients with the genotype RUNX1mut/IDH2mut (OS, P = 0.04) had a better outcome. In conclusion, RUNX1-mutated AML is associated with a complex mutation cluster and is correlated with distinct clinico-pathologic features and inferior prognosis.

Leukemia (2016) 30, 2160–2168; doi:10.1038/leu.2016.126

INTRODUCTION ‘AML with mutated CEBPA’.3 These have also been endorsed by Over the last few decades, the discovery of recurrent structural the European LeukemiaNet (ELN) recommendations, which in balanced and unbalanced chromosome abnormalities have largely addition to CEBPA and NPM1 mutations propose that FLT3 internal contributed to the clinical management of patients with acute tandem duplications (ITDs) are also included in a standardized myeloid leukemia (AML).1 These chromosomal abnormalities are reporting system of genetic markers.2 Although other gene among the most important prognostic markers and in part define mutations have been shown to provide prognostic information, – specific clinico-pathologic entities of the disease.1–4 In more recent none of these markers have to date entered clinical practice.1,6 8 years, the development of novel genomics technologies, including Beyond the prognostic and predictive value of a molecular next-generation sequencing,5 has greatly contributed to deciphering biomarker, one important question is whether there are specific the molecular genetic changes associated with the development of gene mutations other than NPM1 or CEBPA that define a disease AML. With the advent of these technologies, it has become evident entity that is correlated with distinct morphologic, immunopheno- that AML is characterized by remarkable genetic heterogeneity, with typic and clinical features. The planned update of WHO classification individual patients presenting with a distinct and almost unique considers AML with RUNX1 mutation as a new provisional entity. combination of structural genomic changes and somatically acquired RUNX1 is a transcriptional factor widely expressed in hema- gene mutations. topoietic cells and indispensable for the establishment of Although great progress has been made in unraveling the AML definitive hematopoiesis. In mouse models, lack of the Runx1 genome, to date only few molecular markers have been shown gene impairs definitive hematopoiesis and causes embryonic to have clinical relevance.5–7 In the WHO 2008 classification, for death. In adult hematopoiesis, disruption of the RUNX1 gene by the first time two provisional entities defined by the presence intragenic mutations leads to a preleukemic state that predis- – of gene mutations were added: ‘AML with mutated NPM1’ and poses to AML.9 12

1Klinik für Innere Medizin III, Universitätsklinikum Ulm, Ulm, Germany; 2Cancer Genome Project, Wellcome Trust Sanger Institute, Cambridge, UK; 3Klinik für Hämatologie und Onkologie, Klinikum Oldenburg, Oldenburg, Germany; 4Klinik für Innere Medizin II, Universitätsklinikum Schleswig-Holstein Campus Kiel, Kiel, Germany; 5Medizinische Klinik und Poliklinik III, Universitätsklinikum Bonn, Bonn, Germany; 6Klinik für Innere Medizin I, Universitätskliniken des Saarlandes, Homburg, Germany; 7Klinik für Hämatologie, Onkologie und Klinische Onkologie, Universitätsklinikum Düsseldorf, Düsseldorf, Germany; 8Medizinische Klinik III, Städtisches Klinikum Karlsruhe gGmbH, Karlsruhe, Germany; 9III. Medizinische Klinik, Klinikum Rechts der Isar, Technische Universität München, München, Germany; 10Medizinische Klinik IV, Universitätsklinikum Gießen, Gießen, Germany; 11Medical Systems Biology, Universität Ulm, Ulm, Germany and 12Klinik für Hämatologie, Hämostaseologie, Onkologie und Stammzelltransplantation, Medizinische Hochschule Hannover, Hannover, Germany. Correspondence: Professor H Döhner, Department of Internal Medicine III, University Hospital of Ulm, Albert-Einstein-Allee 23, Ulm 89081, Germany. E-mail: [email protected] 13These authors contributed equally to this work. 14A complete list of the members of the German-Austrian Acute Myeloid Leukemia Study Group (AMLSG) appears in the 'Supplementary information'. Received 10 January 2016; revised 13 April 2016; accepted 21 April 2016; accepted article preview online 3 May 2016; advance online publication, 10 June 2016 RUNX1 mutations in AML VI Gaidzik et al 2161 In human acute leukemia, RUNX1 is involved in recurrent (n =1381) and SF3B1 (n = 1381); secondary genotypes were evaluated for chromosomal translocations, such as t(8;21)(q22;q22); RUNX1- their impact on response to therapy as well as survival. RUNX1T1 and t(3;21)(q26.2;q22); MECOM(EVI1)-RUNX1 in AML, or In 16 cases, we analyzed paired BM samples from diagnosis and relapse. t(12;21)(p13;q22); TEL-RUNX1 in B-lineage acute lymphoblastic In another 10 cases, germline material (DNA obtained from buccal swabs leukemia. In addition to these balanced rearrangements, recurrent or from PB in complete remission (CR)) was studied for the presence of fi RUNX1 germline mutations. intragenic mutations have been identi ed in AML, myelodysplastic Furthermore, PB samples from 29 healthy volunteers were analyzed for the syndrome (MDS), chronic myelomonocytic leukemia and T-cell 13–28 presence of RUNX1 polymorphisms. All RUNX1 sequence variations were acute lymphoblastic leukemia. aligned to different SNP databases (http://www.ncbi.nlm.nih.gov/sites/snp; From a clinical point of view, there are interesting aspects to http://genome.ucsc.edu/cgi-bin/hgGateway; http://www.ensembl.org) to detect myeloid neoplasms with RUNX1 mutations. First, RUNX1 mutations known polymorphisms. have been frequently found in radiation-exposed patients with MDS/AML; furthermore, RUNX1 mutations have been linked to Statistical analyses therapy-related MDS and within this group a significant associa- 29–31 Statistical analyses for clinical outcome were performed according to tion was found with monosomy 7 or 7q deletions. Second, previous reports.42 The median follow-up for survival was calculated MDS/AML developing in patients with inherited disorders, such as according to the method of Korn.46 The definition of CR, event-free survival Fanconi anemia or congenital neutropenia, have been shown to (EFS), relapse-free survival (RFS) and overall survival (OS), as well as genetic frequently carry RUNX1 mutations.32,33 Finally, there are rare categorization into favorable-, intermediate-I/II and adverse-risk groups germline RUNX1 mutations that are associated with the autosomal followed the recommended criteria.2 Pairwise comparisons between dominant familial platelet disorder predisposing the affected patient characteristics (covariates) were performed by using the Mann– individuals to AML.34–37 Whitney test for continuous variables and by using Fisher’s exact test for – We and others previously reported on the frequency and clinical categorical variables. The Kaplan Meier method was used to estimate the 47 fi significance of RUNX1 mutations in adult AML patients.19–22,23,25 In distribution of EFS, RFS and OS. Estimation of con dence intervals (CIs) for the survival curves was based on Greenwood’s formula for the the present study of 2439 clinically and genetically annotated adult standard error estimation. A logistic regression model was used to analyze patients with AML, we show that RUNX1 mutations are associated associations between baseline characteristics and the achievement of CR. with characteristic clinico-pathologic features and inferior prognosis. Cox models were used to identify prognostic variables.48,49 Cox models for Furthermore, we show that in the context of RUNX1-mutated AML the entire cohort were stratified for age group (⩽60 years vs 460 years). patterns of co-mutations identify subsets with particularly poor Explanatory variables in the regression analyses included age, sex, prognosis. hemoglobin level, logarithm of white blood cell (WBC), type of AML (de novo, secondary AML, therapy-related AML), percentage of PB and BM blasts, genetic risk group,1,2 and mutational status of RUNX1, NPM1, FLT3 PATIENTS AND METHODS (ITD and TKD), CEBPA (CEBPA double mutated, CEBPAdmut), ASXL1, IDH1, Patient samples IDH2 and DNMT3A. There was no further variable selection to allow estimation of the relative impact of RUNX1 in the context with already Diagnostic bone marrow (BM) or peripheral blood (PB) samples from 2439 known prognostic markers. For multivariable analyses, a missing value AML patients (16.3–84.5 years) were analyzed. Patients were enrolled on 50 imputation technique for missing at random situations was used. In four consecutive AMLSG multicenter treatment trials for younger adult addition, we introduced interaction terms of RUNX1 mutation with patients (AML HD98A (n = 804)38 and AMLSG 07-04 (n = 885)39) and older mutations of ASXL1, DNMT3A, IDH1 and IDH2 (those mutations for which adult patients (AML HD98B (n = 307)40 and AMLSG 06-04 (n = 443)).41 the data set is nearly complete) for the end points achievement of CR, OS, Of these, 945 patients (AML HD98A, n = 651; AMLSG 07-04, n = 294) were RFS and EFS for younger and older patients. Data on EZH2, SRSF2, SF3B1, published previously with regard to RUNX1 mutations and their prognostic STAG2, PHF6 and BCOR were not included in multivariable analyses impact.21 The backbone of induction therapy was homogeneous across because data were only available for about half of the cohort. We the four consecutive AMLSG clinical trials; all trials included two cycles estimated missing data for covariates by using 50 multiple imputations in of ‘3+7’-based induction chemotherapy. In the AML HD98A trial, patients chained equations that incorporated predictive mean matching.51 All received two induction cycles of ICE (idarubicin, cytarabine, etoposide). statistical analyses were performed with the statistical software environ- Induction therapy in the AMLSG 07-04 trial also consisted of two cycles ment R version 3.0.2, using the R packages rms version 4.1-1, survival of ICE; in this trial, there was a randomization for all-trans retinoic acid and version 2.37-4 and cmprsk version 2.2-6.52 in a subset of patients also for valproic acid. In the AML HD98B trial, To study the organization of RUNX1 mutations within the WHO category induction therapy also consisted of two induction cycles of ICE, again with of ‘AML with recurrent genetic abnormalities’ and the association with a randomization for all-trans retinoic acid, and in the AMLSG 06-04 trial, additional gene mutations, we performed constraint analyses of gene patients received two induction cycles with idarubicin, cytarabine, all-trans signatures. Constraint sorting is based on the greedy algorithm for the retinoic acid, and were randomized for valproic acid. All patients gave minimal Set Covering Problem (Supplementary Statistical Method).53 informed consent for treatment and genetic analysis according to the Declaration of Helsinki. RESULTS Molecular studies Frequency and types of RUNX1 mutations In all patients, diagnostic samples were prospectively studied for the Overall, 280 RUNX1 mutations were found in 245 of 2439 (10.0%) presence of the recurring gene fusions RUNX1-RUNX1T1, CBFB-MYH11, KMT2A-MLLT3 and PML-RARA (either by fluorescence in situ hybridization or patients. Mutations clustered as follows: exon 3 (n = 48), exon 4 PCR), and for gene mutations in RUNX1, FLT3 (ITD and tyrosine kinase domain (n = 81), exon 5 (n = 42), exon 6 (n = 23), exon 7 (n = 23) and exon 8 (TKD) mutations at codons D835 and I836), NPM1, CEBPA, KMT2A (partial (n = 63) (Supplementary Table S1); there were 146 frameshift, 96 tandem duplication, PTD), IDH1, IDH2, ASXL1 and DNMT3A genes.21,42–45 missense and 38 non-sense mutations. One patient had two In a subset of our patient cohort, we took advantage of data from a large clones with a frameshift mutation in the first clone and with two targeted sequencing study on 111 myeloid cancer genes performed in 1540 other frameshift mutations in the second clone. In 33 patients, patients (manuscript in press) to evaluate the frequency of additional gene 2 RUNX1 mutations were found. mutations co-occurring with RUNX1 mutations. To this end, mutational data Testing for germline mutations was done in 10 patients, n =8 from the following genes co-mutated with a frequency of 45% were mut mut dmut mut considered: NRAS, TET2, PTPN11, TP53, SRSF2, KMT2A, WT1, KRAS, KIT, STAG2, with NPM1 /RUNX1 ; n = 1 with CEBPA /RUNX1 and n =1 with t(8;21)/RUNX1mut. Only one of the patients with NPM1mut/ RAD21, EZH2, PHF6, SF3B1, CBL, U2AF1, BCOR, GATA2, NF1, MYC, EP300 and mut ETV6. For further exploratory subset analyses, six genes that tightly correlated RUNX1 had a germline RUNX1 mutation (as c.G991A; p.M267I). with RUNX1 mutations were selected, that is, SRSF2 (data available for The analysis of 16 paired diagnosis and relapse samples n = 1378 patients), STAG2 (n =1376), BCOR (n = 1381), EZH2 (n = 1381), PHF6 revealed a stability of RUNX1 mutations in 13/16 of the cases,

© 2016 Macmillan Publishers Limited, part of Springer Nature. Leukemia (2016) 2160 – 2168 RUNX1 mutations in AML VI Gaidzik et al 2162 with 2 patients losing their second RUNX1 mutation in relapse; (n = 13), biallelic CEBPA mutations (n = 2), t(8;21) (n = 1), inv(16) 3 patients were RUNX1 wild type at the time of relapse. (n = 1), t(9;11) (n = 1) and inv(3) (n = 3) (Figure 2a).

Association of RUNX1 mutations with clinical characteristics Subset analyses. Data from the targeted sequencing study (manuscript in press) were used for exploratory subset analyses RUNX1 mutations were significantly associated with older age to further investigate gene mutations associated with RUNX1mut. (Po0.0001) (Figure 1), male gender (P = 0.02), secondary AML Beside the already described significant associations with muta- evolving from MDS (Po0.0001), higher platelet counts (P = 0.007), tions in ASXL1, IDH2 and KMT2A, we found additional significant lower lactate dehydrogenase (LDH) levels (Po0.0001) and with co-occurrence with mutations in SRSF2 (Po0.0001), BCOR French-American-British (FAB) M0 morphology (P = 0.05; analysis (Po0.0001), PHF6 (P = 0.008), STAG2 (P = 0.02), SF3B1 (P=0.03) of morphology according to FAB classification only for a subset of and in trend with EZH2 (P = 0.08) (Table 2; Figures 2b and c). patients available) (Table 1). Data for specific patients' subsets (younger vs older patients; patients with cytogenetically normal (CN)-AML; patients with RUNX1mut/ASXL1mut, RUNX1mut/SRSF2mut, Response to induction therapy RUNX1mut/IDH2mut, RUNX1mut/KMT2A-PTD, RUNX1mut/FLT3-ITD, Clinical correlation analyses could be done in 2404 patients RUNX1mut/DNMT3Amut, RUNX1mut/EZH2mut, RUNX1mut/PHF6mut, (missing follow-up data, n = 35). The CR rate was significantly RUNX1mut/SF3B1mut genotypes; subgroups according to mutation lower in patients with RUNX1 mutations than in patients with types and localization of mutations in the functional domains as well as type of AML) are provided in Supplementary Tables S2–16. Table 1. Pretreatment clinical characteristics according to RUNX1 mutational status Association of RUNX1 mutations with cytogenetic and molecular genetic changes RUNX1mut RUNX1wt P Genetic data in the entire study cohort. Cytogenetic data were (n = 245) (n = 2194) available in 2231/2439 (91.5%) patients (Table 2). RUNX1 muta- fi Age, years tions were rare in ELN favorable-risk AML (4%), signi cantly Median; range 59.2; 19.2–79.1 53.6; 16.3–84.5 o0.0001 enriched in intermediate-I (47.1%) and intermediate-II-risk AML (30.7%), and 18.2% in adverse-risk AML. With regard to specific Male sex, n (%) 147 (60.0) 1137 (51.8) 0.02 cytogenetic abnormalities, RUNX1 mutations were associated with the presence of -7/7q- (P = 0.04) and +13 (P = 0.001). AML history, n (%) RUNX1 mutations were inversely correlated with NPM1 De novo 194 (79.5) 1920 (88.5) o0.0001 o (Po0.0001) and biallelic CEBPA (P = 0.02) mutations, as well as Secondary 38 (15.6) 119 (5.5) 0.0001 with the recurrent abnormalities t(15;17); PML-RARA, inv(16); CBFB- Therapy-related 12 (4.9) 131 (6.0) 0.57 MYH11 and t(8;21); RUNX1-RUNX1T1 (Table 2). RUNX1 mutations WBC count, × 109/l were significantly associated with mutations in ASXL1 (Po0.0001), Median; range 13.7; 0.3–532.7 13; 0.1–439.5 0.77 KMT2A (Po0.0001) and IDH2 (P = 0.02). Compared with RUNX1- Missing 3 45 mutated AML presenting de novo, those with secondary AML 9 more frequently had concurrent ASXL1 mutations (15.1 vs 52.6%; Platelet count, × 10 /l – – Po0.0001) (Supplementary Table S16A). Median; range 67; 4 575 53; 2 933 0.007 Missing 3 46 For visualization of RUNX1 mutations within the WHO category ‘AML with recurrent genetic abnormalities’, constraint sorting of Hemoglobin, g/dl gene signatures was performed in 1506 cases of which a full data Median; range 9.2; 2.7–14.6 9.1; 2.5–17.6 0.56 set was available (Figure 2a). RUNX1 mutations were almost Missing 3 48 entirely mutually exclusive of these recurrent genetic abnormalities, that is, they form a distinct cluster within this category of Blood blasts, % – – AML. RUNX1 mutations rarely co-occurred with NPM1 mutations Median; range 35; 0 100 35; 0 100 0.75 Missing 17 201

Bone marrow blasts, % Median; range 75; 3–100 75; 0–100 0.74 Missing 12 211

LDH, U/l Median; range 322; 110–5406 418; 40–15098 o0.0001 Missing 8 72

FAB classification, n (%)a M0 11 (13.8) 49 (5.1) 0.05 M1 14 (17.5) 164 (17.2) 0.37 M2 19 (23.8) 250 (26.2) 0.08 M3 0 95 (10.0) o0.0001 M4 22 (27.5) 241 (25.3) 0.39 M5 12 (15.0) 114 (12.0) 1.00 M6 2 (2.5) 27 (2.8) 0.76 M7 0 13 (1.4) 0.63 Not classified by FAB 165 1241 Figure 1. Frequency of RUNX1 mutations by age group in 2439 AML patients. In younger adult patients (16–59 years), the frequency Abbreviations: AML, acute myeloid leukemia; FAB, French-American-British; ⩾ LDH, lactate dehydrogenase; n, number of patients; P, P-value; WBC, white ranges between 6 and 10%; in patients 60 years of age, the a frequency is in the order of 15%. Given a median age at diagnosis of blood cell count; WHO, World Health Organization. Restricted to AML – HD98A and AML HD98B trials; since 2001 AML was categorized according AML of 70 72 years, RUNX1 mutation is one of the most frequent fi genetic lesions in AML. to the WHO classi cation.

Table 2. Association of RUNX1 mutations with cytogenetic and Table 2. (Continued ) molecular genetic changes RUNX1mut RUNX1wt P RUNX1mut RUNX1wt P (n = 245) (n = 2194) (n = 245) (n = 2194) KMT2A-PTD, n (%) o0.0001 Cytogenetic data Absent 205 (86.1) 1487 (95.6) t(15;17), n (%) 0 77 (3.8) 0.0003 Present 33 (13.9) 68 (4.4) t(8;21), n (%) 1 (0.4) 102 (5.1) 0.0003 Missing 7 639 inv(16)/t(16;16), n (%) 1 (0.4) 124 (6.2) o0.0001 t(6;9), n (%) 0 15 (0.8) 0.39 BCOR, na (%) o0.0001 inv(3)/t(3;3), n (%) 3 (1.3) 38 (1.9) 0.79 Wild type 107 (89.2) 1242 (98.5) t(9;11), n (%) 1 (0.5) 33 (1.7) 0.25 Mutated 13 (10.8) 19 (1.5) t(v;11q23), n (%) 2 (0.9) 29 (1.5) 0.76 Missing 125 933 − 5/5q-, n (%) 13 (5.8) 135 (6.7) 0.67 − 7/7q-, n (%) 24 (10.6) 137 (6.8) 0.04 STAG2, na (%) 0.02 +8, n (%) 19 (8.4) 202 (10.1) 0.48 Wild type 110 (91.7) 1211 (96.4) 9q-, n (%) 6 (2.7) 66 (3.3) 0.84 Mutated 10 (8.3) 45 (6.3) +11, n (%) 3 (1.3) 36 (1.8) 0.79 Missing 125 938 +13, n (%) 11 (4.9) 28 (1.4) 0.001 +21, n (%) 8 (3.5) 55 (2.7) 0.52 SRSF2, na (%) o0.0001 − 17/abnl(17p), n (%) 9 (4.0) 114 (5.7) 0.36 Wild type 91 (75.8) 1208 (96.0) Complex karyotype, n (%) 22 (9.7) 248 (12.4) 0.28 Mutated 29 (24.2) 50 (4.0) Monosomal karyotype, n (%) 21 (9.3) 211 (10.5) 0.65 Missing 125 936 Normal karyotype, n (%) 113 (50.0) 974 (48.6) 0.73 Others, n (%) 27 (11.0) 123 (5.6) 0.002 EZH2, na (%) 0.08 Missing, n (%) 19 (7.8) 189 (8.6) Wild type 113 (94.2) 1227 (97.3) Mutated 7 (5.8) 34 (2.7) Molecular genetic data Missing 125 933 NPM1, n (%) Wild type 230 (94.7) 1509 (69.9) o0.0001 PHF6, na (%) 0.008 Mutated 13 (5.4) 651 (30.1) Wild type 111 (92.5) 1228 (97.4) Missing 2 34 Mutated 9 (7.5) 33 (2.6) Missing 125 933 CEBPA, n (%) 0.02 Wild type 225 (94.9) 1890 (92.4) SF3B1, na (%) 0.03 Mutated monoallelic 10 (4.2) 70 (3.4) Wild type 114 (95.0) 1240 (98.3) Mutated biallelic 2 (0.8) 86 (4.2) Mutated 6 (5.0) 21 (1.7) Missing 8 148 Missing 125 933

FLT3-ITD, n (%) 0.74 ELN classiﬁcation, n (%) Absent 194 (80.5) 1704 (79.3) Favorable 9 (4.0) 587 (30.5) o0.0001 Present 47 (19.5) 446 (20.7) Intermediate-I 106 (47.1) 612 (31.8) o0.0001 Missing 4 44 Intermediate-II 69 (30.7) 326 (17) o0.0001 Adverse 41 (18.2) 401 (20.8) 0.38 FLT3-TKD, n (%) 0.06 Missing 20 268 Wild type 220 (96.1) 1929 (92.7) Mutated 9 (3.9) 153 (7.4) Abbreviations: ELN, European LeukemiaNet2;ITD, internal tandem duplica- Missing 16 112 tion; n, number of patients; P, P-value; PTD, partial tandem duplication; TKD, tyrosine kinase domain. aMutation data of BCOR, STAG2, SRSF2, EZH2, DNMT3A, n (%) 0.67 PHF6 and SF3B1 were obtained from a comprehensive targeted sequencing Wild type 193 (81.8) 1663 (80.3) study (manuscript in press). Mutated 43 (18.2) 409 (19.8) Missing 9 122 o IDH1, n (%) 0.29 RUNX1 wild type (48.4 vs 68.1%; P 0.0001; Table 3a); this was Wild type 222 (91.4) 2008 (93.1) attributable to a higher rate of resistant disease (40.6 vs 23.4%; Mutated 21 (8.6) 149 (6.9) P = 0.03). A similar pattern was found in the subsets of younger Missing 2 37 (16–60 years) and older (460 years) patients (Table 3a, Supplementary Tables S2C and S3C), as well as in patients with IDH2, n (%) 0.02 CN-AML (Supplementary Table S4B). With regard to different Wild type 205 (84.3) 1926 (89.3) Mutated 38 (15.6) 230 (10.7) treatment trials, the following CR rates were observed for AML mut Mutated (R140) 29 (11.9) 171 (7.9) with RUNX1 vs RUNX1 wild type, respectively: AML HD98A, 57.9 Mutated (R172) 9 (3.7) 54 (2.5) vs 71.8%; AMLSG 07-04, 64.0 vs 77.0%; AML HD98B, 35.0 vs 48.6%; Other mutation 0 5 (0.2) AMLSG 06-04, 31.9 vs 54.1%. We found no difference in response Missing 2 38 rate among the trials when stratiﬁed for age (data not shown). ASXL1, n (%) o0.0001 Among RUNX1-mutated patients, those with secondary AML had Wild type 193 (79.4) 2016 (94.0) particularly poor response to induction therapy (CR rate 15.8%, Mutated 50 (20.6) 129 (6.0) rate of resistant disease 71.1%; vs de novo AML, 54.4 and 35.2%, Missing 2 49 respectively; Po0.0001; Supplementary Table S16B). The majority of AML patients have two or more acquired mutations, and the number of acquired mutations has been associated with worsening outcomes. We explored how frequent secondary

© 2016 Macmillan Publishers Limited, part of Springer Nature. Leukemia (2016) 2160 – 2168 RUNX1 mutations in AML VI Gaidzik et al 2164 RUNX1 genotypes associated with response to therapy. Responses vs 64.4% in RUNX1mut/SFRS2wt; P = 0.02) and RUNX1mut/PHF6mut were lowest for the genotypes RUNX1mut/ASXL1mut (30 vs (22.2 vs 60.9% in RUNX1mut/PHF6wt; P = 0.03) (Supplementary 52.6% in RUNX1mut/ASXL1wt; P = 0.007), RUNX1mut/SRSF2mut (37.9 Tables S5B–S13B).

Figure 2. (a) Organization of RUNX1 mutations within the WHO category ‘AML with recurrent genetic abnormalities’ (n = 1506); vertical lines represent individual patients. To allow an overview on all abnormalities, we omitted fractions of patients (patients with NPM1 mutation between nos. 50 and 660, and patients with RUNX1 mutation between nos. 740 and 910) showing no overlap with the other abnormalities. Only CEBPAdmut is included. (b) Organization of concurrent gene mutations in RUNX1-mutated AML by functional classes (n = 108). Only patients with a full data set of the respective genes were included. (c) Conditional probabilities of gene mutation frequencies for the RUNX1mut subgroup. On top axis, molecular data are shown for all patient samples with RUNX1 mutations. In blue patterned, RUNX1 mutation is present in 100% of the samples in this subgroup. Each dark blue barplot indicates the proportion of patients harboring each of the genetic alterations in legend. On the lower axis, light blue bar plots represent mutation frequencies for the same lesions in RUNX1 wild type patients. Significantly or in trend co-mutated lesions (as defined by Fisher's exact test) are marked with red asterisks or a black asterisk on the top axis, respectively, whereas on the lower panel, red asterisks mark a significant mutual exclusivity.

Table 3a. Univariable outcome analyses according to RUNX1 mutation Table 3b. Multivariable analyses for all AML patients (APL excluded), status stratiﬁed analysis according to age

Entire cohort RUNX1mut RUNX1wt P End point Variables HR 95% CI P Clinical end point (n = 245) (n = 2194) EFS CR rate, % 48.4 68.1 o0.0001 RUNX1 mutation 1.22 1.04–1.42 0.01 s-AML 1.21 1.00–1.46 0.05 RD rate, % 40.6 23.4 0.03 t-AML 1.12 0.92–1.37 0.27 Missing, n 1 34 FLT3-ITD 1.32 1.13–1.54 0.0004 FLT3-TKD mutation 0.93 0.75–1.15 0.51 EFS o0.0001 Age (10 years difference) 1.17 1.10–1.24 o0.0001 Median, mo (95% CI) 2.0 (1.6–2.8) 8.2 (7.3–8.9) Gender (female) 0.87 0.78–0.96 0.004 5-year EFS (%) 9 (6–13) 24 (22–26) NPM1 mutation 0.66 0.58–0.76 o0.0001 DNMT3A mutation 1.13 0.99–1.29 0.07 RFS 0.0007 ASXL1 mutation 1.04 0.87–1.24 0.68 – – Median, mo (95% CI) 12.1 (10.4 14.7) 16.0 (14.5 18.4) ELN Intermediate-I 1.77 1.51–2.07 o0.0001 – 5-year RFS (%) 22 (16 to 30) 36 (34 39) ELN Intermediate-II 1.81 1.54–2.14 o0.0001 – o o ELN Adverse 3.17 2.70 3.73 0.0001 OS 0.0001 BM blasts 1.00 1.00–1.00 0.51 Median, mo (95% CI) 13.3 (11.5–16.0) 21.1 (19.1–23.5) – o – – WBC (log10) 1.25 1.13 1.37 0.0001 5-year OS (%) 22 (27 28) 37 (35 39) LDH (log10) 1.10 0.92–1.33 0.30 mut wt RFS Younger adults RUNX1 RUNX1 P RUNX1 mutation 1.03 0.84–1.27 0.75 Clinical end point (n = 132) (n = 1557) s-AML 1.05 0.80–1.40 0.71 – CR rate, % 61.4 74.5 0.002 t-AML 1.49 1.16 1.91 0.002 FLT3-ITD 1.49 1.23–1.80 o0.0001 RD rate, % 28.8 17.8 0.003 FLT3-TKD mutation 0.86 0.66–1.11 0.24 Missing, n 26 221 Age (10 years difference) 1.18 1.10–1.26 o0.0001 Gender (female) 0.96 0.85–1.09 0.54 EFS o0.0001 NPM1 mutation 0.75 0.63–0.89 0.0007 Median, mo (95% CI) 3.1 (1.4–6.8) 10.1 (9.1–11.3) DNMT3A mutation 1.11 0.95–1.31 0.20 5-year EFS (%) 12 (8–20) 30 (28–32) ASXL1 mutation 1.18 0.93–1.50 0.18 ELN Intermediate-I 1.50 1.25–1.81 o0.0001 RFS 0.007 ELN Intermediate-II 1.42 1.17–1.73 0.0004 Median, mo (95% CI) 14.1 (10.8–21.3) 23.4 (18.8–28.8) ELN Adverse 2.16 1.77–2.64 o0.0001 5-year RFS (%) 26 (19–37) 42 (39–45) BM blasts 1.00 1.00–1.00 0.69 WBC (log10) 1.23 1.09–1.39 o0.001 OS 0.002 LDH (log10) 1.29 1.03–1.63 0.03 Median, mo (95% CI) 18.0 (13.6–25.3) 38.6 (30.3–50.2) OS 5-year OS (%) 33 (26–43) 46 (43–48) RUNX1 mutation 1.10 0.93–1.30 0.26 s-AML 1.228 1.05–1.56 0.02 mut wt Older adults RUNX1 RUNX1 P t-AML 1.20 0.96–1.48 0.11 Clinical end point (n = 113) (n = 637) FLT3-ITD 1.53 1.29–1.80 o0.0001 FLT3-TKD mutation 1.04 0.82–1.32 0.72 CR rate, % 33.0 51.8 0.0003 Age (10 years difference) 1.38 1.29–1.47 o0.0001 Gender (female) 0.94 0.84–1.05 0.28 RD rate, % 54.5 37.6 0.001 NPM1 mutation 0.88 0.75–1.02 0.10 Missing, n 62 280 DNMT3A mutation 1.05 0.91–1.21 0.51 ASXL1 mutation 1.14 0.95–1.38 0.17 EFS 0.0009 ELN Intermediate-I 1.73 1.44–2.07 o0.0001 Median, mo (95% CI) 1.7 (1.3–2.2) 3.2 (2.5–4.6) ELN Intermediate-II 1.81 1.50–2.19 o0.0001 5-year EFS (%) 4 (2–10) 8 (6–11) ELN Adverse 3.41 2.85–4.09 o0.0001 – RFS 0.43 BM blasts 1.00 1.00 1.00 0.99 – – WBC (log10) 1.21 1.10–1.35 0.0002 Median, mo (95% CI) 9.9 (6.4 13.1) 9.8 (8.9 11.0) – 5-year RFS (%) 14 (7–27) 15 (12–19) LDH (log10) 1.39 1.13 1.71 0.002 Abbreviations: AML, acute myeloid leukemia; APL, acute promyelocytic OS 0.09 leukemia; BM, bone marrow; CI, conﬁdence interval; EFS, event-free Median, mo (95% CI) 9.6 (6.3–12.8) 11.1 (9.9–12.5) 2 – – survival; ELN, European LeukemiaNet ; HR, hazard ratio; ITD, internal 5-year OS (%) 8 (4 15) 15 (13 19) tandem duplication; LDH, lactate dehydrogenase serum levels; OS, overall Abbreviations: CR, complete remission; EFS, event-free survival; mo, survival; P, p-value; PTD, partial tandem duplication; RFS, relapse-free months; n, numbers; OS, overall survival; P, p-value; RD, refractory disease; survival; s-AML, secondary AML evolving from myelodysplastic syndrome; RFS, relapse-free survival. t-AML, therapy-related AML; TKD, tyrosine kinase domain; WBC, white blood cell count. Cox regression model on event-free, relapse-free and overall survival.

In multivariable analyses, RUNX1 mutation was an independent poor prognostic factor for achievement of CR in the whole cohort (odds ratio, 0.70; 95% CI, 0.51–0.96; P = 0.03). This effect was even more pronounced in the older patients (odds ratio, 0.48; 95% CI, ‘RUNX1 mutation*DNMT3A mutation’, ‘RUNX1 mutation*IDH1 0.28–0.81; P = 0.006). Data by age group (⩽60 years vs 460 yrs) mutation’ and ‘RUNX1 mutation*IDH2 mutation’ in multivariable are provided in Supplementary Tables S2D and S3D. We also analyses, but they had no signiﬁcant impact (Supplementary included the interaction terms ‘RUNX1 mutation*ASXL1 mutation’, Tables S2D and S3D).

© 2016 Macmillan Publishers Limited, part of Springer Nature. Leukemia (2016) 2160 – 2168 RUNX1 mutations in AML VI Gaidzik et al 2166 Supplementary Table S12C, Supplementary Figure S11A); in contrast, outcome for RUNX1mut/IDH2mut patients was significantly better (OS, P = 0.04; RFS, P = 0.02; Supplementary Table S7C, Supplementary Figure S6). In multivariable analysis, RUNX1 mutation was an independent prognostic marker for inferior EFS (hazard ratio (HR), 1.22; P = 0.01), but not RFS (HR, 1.03; P = 0.75) and OS (HR, 1.10; P = 0.26; Table 3b). In younger patients, RUNX1 mutation only had a moderate impact on EFS (HR, 1.20; P = 0.09), and not on RFS (HR, 1.09; P = 0.48) and OS (HR, 1.13; P = 0.28; Supplementary Table S2E). In patients 460 years, RUNX1 mutation had no independent prognostic impact (EFS: HR, 1.22, P = 0.08; RFS: HR, 0.88, P = 0.49; OS: HR, 1.08, P = 0.53; Supplementary Table S3E). Furthermore, we performed multivariable analyses including the interaction terms ‘RUNX1 mutation*ASXL1 mutation’, ‘RUNX1 mutation*DNMT3A mutation’, ‘RUNX1 mutation*IDH1 mutation’ and ‘RUNX1 mutation*IDH2 mutation’ to evaluate a possible effect of combined genotypes for the end points OS, EFS and RFS. Neither in younger patients nor in older patients, we found a significant interaction for OS, RFS or EFS (Supplementary Table S2E). Among 81 younger (16–60 years) patients with RUNX1 mutations achieving a CR, 36 patients received an allogeneic hematopoietic cell transplantation in first CR; there was a significant impact by Mantel-Byar analysis of allogeneic hematopoietic cell transplantation on RFS (P = 0.01), but not on OS (P = 0.53; Supplementary Figure S1E).

DISCUSSION In the present study, we assessed the frequency, genetic relationship and clinical impact of RUNX1 mutations in adult patients with newly-diagnosed AML. Overall, we detected RUNX1 mutations in 10% of patients, 24.2% in secondary AML and 9.2% in de novo AML. Consistent with previous reports,19,24,25 we found a significant increase in the mutation frequency with older age (Figure 1). Given a median age of diagnosis in AML of 70–72 years,54 RUNX1 mutations are thus among the most frequent genetic lesions in AML. With regard to the genetic distribution, we found that AML with Figure 3. Univariable survival analysis. EFS (a), RFS (b) and OS (c). RUNX1 mutations was almost entirely mutually exclusive of the current WHO category ‘AML with recurrent genetic abnormalities’ (Figure 2a).3,4 In only rare cases, there was co-occurrence of the mutation with t(8;21), inv(16), inv(3), NPM1 or biallelic CEBPA mutation. With regard to secondary cytogenetic abnormalities, the Survival analysis strongest association of RUNX1 mutations was found with +13 as The median follow-up time for survival in the entire cohort was previously reported;17,19,22 there was no significant correlation – 5.79 years (95% CI, 5.57 5.98). In univariable analysis, RUNX1 with +8 and +21.19,21 In addition, we could confirm that RUNX1 mutations were significantly associated with inferior EFS mutations were associated with -7/7q-, a finding that is consistent (Po0.0001), RFS (P = 0.0007) and OS (Po0.0001) (Figure 3; with the higher incidence of RUNX1 mutations in patients with a Table 3a). The effect was similar in patients with CN-AML history of MDS. (Supplementary Figure S3, Supplementary Table S4C), as well as Importantly, RUNX1 mutations were significantly correlated with in the subset of younger patients (OS, P = 0.002; RFS, P = 0.007; EFS, a characteristic pattern of concurrent gene mutations. Associa- Po0.0001; Supplementary Figure S1); in patients 460 years, tions were found with mutations in the epigenetic modifiers RUNX1 mutations were associated with inferior EFS (P = 0.0009) ASXL1, IDH2, KMT2A and EZH2, in the splicing factors SRSF2 and and in trend with poor OS (P = 0.09), there was no effect on SF3B1, in the cohesin complex gene STAG2, in the tumor RFS (P = 0.43) (Supplementary Figure S2). RUNX1-mutated patients suppressor PHF6 and in the transcriptional regulator BCOR presenting with secondary AML had significantly inferior outcome (Figures 2b and c). Similar associations have previously been compared to patients with RUNX1-mutated de novo AML described in high-risk MDS, suggesting that these genotypes can (OS, P = 0.007; RFS, P = 0.05; EFS, P = 0.007; Supplementary Table cross conventional diagnostic boundaries between high-risk S16C; Supplementary Figure S15). chronic and acute myeloid neoplasms.26,27 Recently, Thota In exploratory analyses, we looked at outcome according to et al.55 also described a significant enrichment of RUNX1 secondary genotypes. Survival end points were poorest for the mutations in patients with concurrent cohesin mutations.55 genotypes RUNX1mut/ASXL1mut (OS, P = 0.004; RFS, P = 0.05; EFS, The complex molecular architecture associated with RUNX1 P = 0.03, compared with RUNX1mut/ASXL1wt; Supplementary Table mutation highlights the complexity in delineating class-defining S5C; Supplementary Figure S4), RUNX1mut/SRSF2mut (OS, P = 0.007; biomarkers with both clinical and prognostic associations. RFS, P = 0.06; EFS, P = 0.001; Supplementary Table S6C, In addition to the complex molecular landscape that often Supplementary Figure S5) and RUNX1mut/PHF6mut (OS, P = 0.03, incorporates adverse prognostic biomarkers (that is, ASXL1 and

Leukemia (2016) 2160 – 2168 © 2016 Macmillan Publishers Limited, part of Springer Nature. RUNX1 mutations in AML VI Gaidzik et al 2167 SRSF2), RUNX1-mutated AML was associated with characteristic clinical and pathologic features, and in univariable analyses with presenting clinical features, such as older age, male gender, more an inferior prognosis. immature morphology and secondary AML evolving from MDS.26,27 All these clinical parameters have well-recognized adverse effects on clinical outcome. CONFLICT OF INTEREST In previous studies, RUNX1 mutation has been shown to be a The authors declare no conflict of interest. significant predictor for resistance to standard induction therapy and for inferior survival.19,21,22,24,25 Similarly, in the present study ACKNOWLEDGEMENTS the mutation was associated with lower CR rates, due to a higher rate of resistant disease. Response rates were lowest in patients Supported in part by grants 01GI9981 and 01KG0605 from the German with the genotypes RUNX1mut/ASXL1mut, RUNX1mut/SRSF2mut, Bundesministerium für Bildung und Forschung (BMBF), grant 109675 from the RUNX1mut/PHF6mut, as well as in patients presenting with Deutsche Krebshilfe and the Sonderforschungsbereich (SFB) 1074 funded by the secondary RUNX1-mutated AML (being significantly enriched for Deutsche Forschungsgemeinschaft (SFB 1074, projects B3 and B4). VG is a grant recipient of the Medical Faculty of Ulm University; LB and MH are Heisenberg RUNX1mut/ASXL1mut). In univariable analyses, patients with RUNX1 fi Professors of the Deutsche Forschungsgemeinschaft (DFG, BU 1339/3-8 and HE 5240- mutation had signi cantly inferior EFS, RFS and OS. The effect was 6-1). HAK is in part supported by Deutsche Forschungsgemeinschaft (SFB 1074, most pronounced in younger patients, but was also evident in project Z1), German Bundesministerium für Bildung und Forschung (BMBF; older patients, as well as in the subset of patients with CN-AML. Gerontosys II, Forschungskern SyStaR, project ID 0315894); European Community's Our exploratory analyses of combined genotypes demonstrate that Seventh Framework Programme (FP7/2007-2013) under grant agreement nº 602783. there are significant interactions among RUNX1 and the tightly AMLSG treatment trials were in part supported by Pfizer and Amgen. We are grateful correlated gene mutations (see Supplementary Figures S7–S12). to all members of the German-Austrian AML Study Group (AMLSG) for their In line with the results on response to induction, survival of participation in this study and providing patient samples; a list of participating patients with RUNX1-mutated AML was poorest for those patients institutions and investigators appears in the Supplementary information. who in addition had ASXL1 or SRSF2 mutation, whereas outcome fi of patients with additional IDH2 mutations was signi cantly better. AUTHOR CONTRIBUTIONS The RUNX1mut/ASXL1mut genotype was previously shown by us to be associated with dismal outcome.45 VIG, VT, RFS, KD and HD designed the research; VIG, VT, EP, PP, JH, TW and BK In multivariable analyses, the negative prognostic effect was performed experiments; VIG, VT, EP, PP, JH, BS, TW, BK and KD analyzed results; only observed for EFS pointing out that RUNX1 mutations per DW, MG, JMK, HAK and RFS performed statistical analyses; CHK, HAH, PB, GH, se are not independently associated with a negative effect on AK, M Ringhoffer, KG, M Rummel, FT, MH, AG, LB, RFS, KD and HD accrued survival. In two larger studies, RUNX1 mutation has been shown to patients and provided material; VIG, VT, EP, RFS, KD and HD wrote the paper. be a strong independent predictor for inferior OS.19,24 However, patients in these studies were highly selected, in the Taiwanese REFERENCES study19 only de novo AML and in the CALGB study24 only de novo CN-AML were included in the analysis. When taking an unbiased 1 Döhner H, Weisdorf DJ, Bloomfield CD. Acute myeloid leukemia. N Engl J Med 373 – approach on all newly-diagnosed AML patients, RUNX1-mutated 2015; : 1136 1152. 2 Döhner H, Estey EH, Amadori S, Appelbaum FR, Büchner T, Burnett AK et al. AML comes together with a compendium of adverse events Diagnosis and management of acute myeloid leukemia in adults: recommenda- clinically and molecularly. Whether the observed associations with tions from an international expert panel, on behalf of the European LeukemiaNet. clinical outcome including attainment of CR, EFS and suggestive Blood 2010; 115:453–474. correlation with OS are causal is not clear. Nonetheless, RUNX1 3 Swerdlow SH, Campo E, Harris NL, Jaffe ES, Pileri SA, Stein H et al. (eds). WHO mutation identifies about 10% of AML patients with such distinct Classification of Tumours of Haematopoietic and Lymphoid Tissues. IARC Press: molecular and clinical presentation that in almost 80% of cases Lyon, France, 2008. would otherwise fall into the intermediate-I or -II ELN genetic 4 Vardiman JW, Thiele J, Arber DA, Brunning RD, Borowitz MJ, Porwit et al. The 2008 category. revision of the WHO Classification of Myeloid Neoplasms and Acute Leukemia: 114 – Given the poor outcome of patients with RUNX1 mutation, rationale and important changes. Blood 2009; :937 951. novel therapeutic approaches are warranted. Younger patients 5 Cancer Genome Atlas Research Network. Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med 2013; 368: 2059–2074. may benefit from allogeneic hematopoietic cell transplantation as 19,21 6 Döhner H, Gaidzik VI. Impact of genetic features on treatment decisions in AML. previously suggested by us and Tang et al. Also in our current Hematology Am Soc Hematol Educ Program 2011; 2011:36–42. analysis, younger RUNX1-mutated patients benefitted from 7 Meyer SC, Levine RL. Translational implications of somatic genomics in acute allogeneic hematopoietic cell transplantation in terms of RFS. myeloid leukaemia. Lancet Oncol 2014; 15: e382–e394. Older patients with RUNX1-mutated AML should be considered for 8 Ossenkoppele G, Löwenberg B. How I treat the older patient with acute myeloid investigational therapies. leukemia. Blood 2015; 125:767–774. In younger patients, RUNX1 mutations were predominantly 9 Okuda T, van Deuersen J, Hiebert SW, Grosveld G, Downing JR. AML1, the target located in exons 4 and 8, whereas exons 3 and 4 were more of multiple chromosomal translocations in human leukemia, is essential for 84 – often affected in older patients consistent with the results normal fetal liver hematopoiesis. Cell 1996; :321 330. 24 19 10 Ichikawa M, Asai T, Saito T, Seo S, Yamazaki I, Yamagata T et al. AML-1 is required reported by Mendler et al. Tang et al. found an equal for megakaryocytic maturation and lymphocytic differentiation, but not for distribution of the mutations between the two functional maintenance of hematopoietic stem cells in adult hematopoiesis. Nat Med 2004; domains, the Runt Homology Domain represented by exons 3–5 10:299–304. and the Transactivation Domain comprising exons 6–8. In contrast, 11 Mangan JK, Speck NA. RUNX1 mutations in clonal myeloid disorders: from we and others detected the majority of mutations in the Runt conventional cytogenetics to next generation sequencing, a story 40 years in – Homology Domain.20 22,25 No difference in outcome was the making. Crit Rev Oncog 2011; 16:77–91. found in our study according to the location of the mutations 12 Sakurai M, Kunimoto H, Watanabe N, Fukuchi Y, Yuasa S, Yamazaki S et al. (see Supplementary Figures S13 and S14). Impaired hematopoietic differentiation of RUNX1-mutated induced pluripotent 28 – In summary, RUNX1 mutation is among the most frequent stem cells derived from FPD/AML patients. Leukemia 2014; :2344 2354. 13 Osato M, Asou N, Abdalla E, Hoshino K, Yamasaki H, Okubo T et al. Biallelic and genetic lesions in AML, is mutually exclusive of the other recurrent fi heterozygous point mutations in the runt domain of the AML1/PEBP2alphaB genetic abnormalities as de ned in the current WHO category gene associated with myeloblastic leukemias. Blood 1999; 93: 1817–1824. ‘AML with recurrent genetic abnormalities’ and exhibits a 14 Imai Y, Kurokawa M, Izutsu K, Hangaishi A, Takeuchi K, Maki K et al. Mutations of characteristic architecture of concurrent genetic lesions. Further- the AML1 gene in myelodysplastic syndrome and their functional implications in more, RUNX1 mutation is associated with specific presenting leukemogenesis. Blood 2000; 96: 3154–3160.

© 2016 Macmillan Publishers Limited, part of Springer Nature. Leukemia (2016) 2160 – 2168 RUNX1 mutations in AML VI Gaidzik et al 2168 15 Preudhomme C, Warot-Loze D, Roumier C, Grardel-Duflos N, Garand R, Lai JL et al. 34 Jongmans MC, Kuiper RP, Carmichael CL, Wilkins EJ, Dors N, Carmagnac et al. High incidence of biallelic point mutations in the Runt domain of the AML1/ Novel RUNX1 mutations in familial platelet disorder with enhanced risk for acute PEBP2 alpha B gene in M0 acute myeloid leukemia and in myeloid malignancies myeloid leukemia: clues for improved identification of the FPD/AML syndrome. with acquired trisomy 21. Blood 2000; 96: 2862–2869. Leukemia 2010; 24:242–246. 16 Chen CY, Lin LI, Tang JL, Ko BS, Tsay W, Chou WC et al. RUNX1 gene mutation in 35 Owen C. Insights into familial platelet disorder with propensity to myeloid primary myelodysplastic syndrome--the mutation can be detected early at malignancy (FPD/AML). Leuk Res 2010; 34:141–142. diagnosis or acquired during disease progression and is associated with poor 36 Preudhomme C, Renneville A, Bourdon V, Philippe N, Roche-Lestienne C, Boissel N outcome. Br J Haematol 2007; 139:405–414. et al. High frequency of RUNX1 biallelic alteration in acute myeloid leukemia 17 Dicker F, Haferlach C, Kern W, Haferlach T, Schnittger S. Trisomy 13 is strongly secondary to familial platelet disorder. Blood 2009; 113: 5583–5587. associated with AML1/RUNX1 mutations and increased FLT3 expression in acute 37 Nickels EM, Soddalter J, Churpek JE, Godley LA. Recognizing familial myeloid myeloid leukemia. Blood 2007; 110: 1308–1316. leukemia in adults. Ther Adv Hematol 2013; 4:254–269. 18 Gelsi-Boyer V, Trouplin V, Adélaïde J, Aceto N, Remy V, Pinson S et al. Genome 38 Schlenk RF, Döhner K, Mack S, Stoppel M, Király F, Götze K et al. Prospective profiling of chronic myelomonocytic leukemia: frequent alterations of RAS and evaluation of allogeneic hematopoietic stem-cell transplantation from matched RUNX1 genes. BMC Cancer 2008; 8:299. related and matched unrelated donors in younger adults with high-risk acute 19 Tang JL, Hou HA, Chen CY, Liu CY, Chou WC, Tseng MH et al. AML1/RUNX1 myeloid leukemia: German-Austrian trial AMLHD98A. J Clin Oncol 2010; 28: mutations in 470 adult patients with de novo acute myeloid leukemia: prognostic 4642–4648. implication and interaction with other gene alterations. Blood 2009; 114: 39 Schlenk RF, Döhner K, Krauter J, Gaidzik V, Paschka P, Heuser M et al. 5352–5361. All-trans retinoic acid improves outcome in younger adult patients with 20 Dicker F, Haferlach C, Sundermann J, Wendland N, Weiss T, Kern W et al. Mutation nucleophosmin-1 mutated acute myeloid leukemia – results of the AMLSG analysis for RUNX1, KMT2A-PTD, FLT3-ITD, NPM1 and NRAS in 269 patients with 07-04 randomized treatment trial [abstract]. Haematologica 2014; 99, MDS or secondary AML. Leukemia 2010; 24: 1528–1532. abstract 646. 21 Gaidzik VI, Bullinger L, Schlenk RF, Zimmermann AS, Röck J, Paschka P et al. 40 Schlenk RF, Fröhling S, Hartmann F, Fischer JT, Glasmacher A, del Valle F et al. RUNX1 mutations in acute myeloid leukemia: results from a comprehensive Phase III study of all-trans retinoic acid in previously untreated patients 61 years genetic and clinical analysis from the AML study group. J Clin Oncol 2011; 29: or older with acute myeloid leukemia. Leukemia 2004; 18:1798–1803. 1364–1372. 41 Tassara M, Döhner K, Brossart P, Held G, Götze K, Horst HA et al. Valproic acid in 22 Schnittger S, Dicker F, Kern W, Wendland N, Sundermann J, Alpermann T et al. combination with all-trans retinoic acid and intensive therapy for acute myeloid RUNX1 mutations are frequent in de novo AML with noncomplex karyotype and leukemia in older patients. Blood 2014; 123: 4027–4036. confer an unfavorable prognosis. Blood 2011; 117: 2348–2357. 42 Schlenk RF, Döhner K, Krauter J, Fröhling S, Corbacioglu A, Bullinger L et al. 23 Bejar R, Stevenson K, Abdel-Wahab O, Galili N, Nilsson B, Garcia-Manero G et al. Mutations and treatment outcome in cytogenetically normal acute myeloid Clinical effect of point mutations in myelodysplastic syndromes. N Engl J Med leukemia. N Engl J Med 2008; 358: 1909–1918. 2011; 364: 2496–2506. 43 Gaidzik VI, Schlenk RF, Paschka P, Stölzle A, Späth D, Kuendgen et al. 24 Mendler JH, Maharry K, Radmacher MD, Mrózek K, Becker H, Metzeler KH et al. Clinical impact of DNMT3A mutations in younger adult patients with acute RUNX1 mutations are associated with poor outcome in younger and myeloid leukemia: results of the AML Study Group (AMLSG). Blood 2013; 121: older patients with cytogenetically normal acute myeloid leukemia and with 4769–4777. distinct gene and MicroRNA expression signatures. J Clin Oncol 2012; 30: 44 Paschka P, Schlenk RF, Gaidzik VI, Habdank M, Krönke J, Bullinger L et al. IDH1 and 3109–3118. IDH2 mutations are frequent genetic alterations in acute myeloid leukemia and 25 Greif PA, Konstandin NP, Metzeler KH, Herold T, Pasalic Z, Ksienzyk B et al. RUNX1 confer adverse prognosis in cytogenetically normal acute myeloid leukemia with mutations in cytogenetically normal acute myeloid leukemia are associated with a NPM1 mutation without FLT3 internal tandem duplication. J Clin Oncol 2010; 28: poor prognosis and up-regulation of lymphoid genes. Haematologica 2012; 97: 3636–3643. 1909–1915. 45 Paschka P, Schlenk RF, Gaidzik VI, Herzig JK, Aulitzky T, Bullinger L et al. ASXL1 26 Papaemmanuil E, Gerstung M, Malcovati L, Tauro S, Gundem G, Van Loo P et al. mutations are associated with distinct clinical and genetic features and Clinical and biological implications of driver mutations in myelodysplastic confer poor prognosis in younger adult patients with acute myeloid leukemia: syndromes. Blood 2013; 122:3616–3627. A study of the German-Austrian AML Study Group. Haematologica 2015; 100: 27 Haferlach T, Nagata Y, Grossmann V, Okuno Y, Bacher U, Nagae G et al. Landscape 324–330. of genetic lesions in 944 patients with myelodysplastic syndromes. Leukemia 46 Korn EL. Censoring distributions as a measure of follow-up in survival analysis. 2014; 28: 241–247. Stat Med 1986; 5:255–260. 28 Grossmann V, Kern W, Harbich S, Alpermann T, Jeromin S, Schnittger S et al. 47 Kaplan E, Meier P. Nonparametric estimation from incomplete observations. JAm Prognostic relevance of RUNX1 mutations in T-cell acute lymphoblastic leukemia. Stat Assoc 1958; 53:457–481. Haematologica 2011; 96: 1874–1877. 48 Therneau TM, Grambusch PM. Modeling Survival Data: Extending the Cox Model. 29 Steensma DP, Gibbons RJ, Mesa RA, Tefferi A, Higgs DR. Somatic point mutations Springer Verlag: New York, NY, USA, 2000. in RUNX1/CBFA2/AML1 are common in high-risk myelodysplastic syndrome, but 49 Cox DR. Regression models and life tables (with discussion). J R Stat Soc B 1972; not in myelofibrosis with myeloid metaplasia. Eur J Haematol 2005; 74:47–53. 34:187–220. 30 Harada H, Harada Y, Tanaka H, Kimura A, Inaba T. Implications of somatic 50 Little RJ, D'Agostino R, Cohen ML, Dickersin K, Emerson SS, Farrar JT et al. mutations in the AML1 gene in radiation-associated and therapy-related myelo- The prevention and treatment of missing data in clinical trials. N Engl J Med 2012; dysplastic syndrome/acute myeloid leukemia. Blood 2003; 101: 673–680. 367:1355–1360. 31 Christiansen DH, Andersen MK, Pedersen-Bjergaard J. Mutations of AML1 are 51 Harrell FE. Regression Modeling Strategies: With Applications to Linear Models, Logistic common in therapy-related myelodysplasia following therapy with alkylating Regression, and Survival Analysis. Springer Verlag: New York, NY, USA, 2001. agents and are significantly associated with deletion or loss of chromosome arm 52 R Development Core Team. R: A Language and Environment for Statistical 7q and with subsequent leukemic transformation. Blood 2004; 104: 1474–1481. Computing. R Foundation for Statistical Computing: Vienna, Austria, 2009. 32 Quentin S, Cuccuini W, Ceccaldi R, Nibourel O, Pondarre C, Pagès MP et al. 53 Chvatal V. A greedy heuristic for the set-covering problem. Math Oper Res 1979; 4: Myelodysplasia and leukemia of Fanconi anemia are associated with a specific 233–235. pattern of genomic abnormalities that includes cryptic RUNX1/AML1 lesions. 54 Juliusson G, Lazarevic V, Hörstedt AS, Hagberg O, Höglund M. Swedish Acute Blood 2011; 117:e161–e170. Leukemia Registry Group. Acute myeloid leukemia in the real world: why 33 Skokowa J, Steinemann D, Katsman-Kuipers JE, Zeidler C, Klimenkova O, population-based registries are needed. Blood 2012; 119: 3890–3899. Klimiankou M et al. Cooperativity of RUNX1 and CSF3R mutations in severe 55 Thota S, Viny AD, Makishima H, Spitzer B, Radivoyevitch T, Przychodzen B et al. congenital neutropenia: a unique pathway in myeloid leukemogenesis. Blood Genetic alterations of the cohesin complex genes in myeloid malignancies. Blood 2014; 123: 2229–2237. 2014; 124:1790–1798.

Supplementary Information accompanies this paper on the Leukemia website (http://www.nature.com/leu)