Leukemia (2013) 27, 82–91 & 2013 Macmillan Publishers Limited All rights reserved 0887-6924/13 www.nature.com/leu

ORIGINAL ARTICLE ASXL1 exon 12 mutations are frequent in AML with intermediate risk karyotype and are independently associated with an adverse outcome

S Schnittger1, C Eder1, S Jeromin1, T Alpermann1, A Fasan1, V Grossmann1, A Kohlmann1, T Illig2,3, N Klopp2,3, H-E Wichmann4, K-A Kreuzer5, C Schmid6, P Staib7, R Peceny8, N Schmitz9, W Kern1, C Haferlach1 and T Haferlach1

We aimed at evaluating ASXL1mut in 740 AML with intermediate risk karyotype for frequency, association with other mutations and impact on outcome. Five hundred fifty-three cases had a normal karyotype (NK) and 187 had intermediate risk aberrant cytogenetics. Overall, ASXL1mut were detected in 127/740 patients (17.2%). ASXL1mut were more frequent in males than in females (23.5% vs 9.9%, Po0.001). They were associated with higher age (median: 71.8 vs 61.8, Po0.001), a history of preceding myelodysplastic syndromes, and with a more immature immunophenotype compared with patients with wild-type ASXL1 (ASXL1wt). ASXL1mut were more frequent in patients with aberrant karyotype (58/187; 31.0%), especially in cases with trisomy 8 (39/ 74; 52.7%), than in those with NK (69/553; 12.5%; Po0.001). ASXL1mut were observed more frequent in RUNX1mut (Po0.001), and less frequent in NPM1mut (Po0.001), FLT3-internal tandem duplication (ITD) (Po0.001), FLT3-TKD (P ¼ 0.001) and DNMT3Amut (Po0.001). Patients with ASXL1mut had a shorter overall survival (OS) (Po0.001) and event free survival (P ¼ 0.012) compared with ASXL1wt. In multivariable analysis, ASXL1mut was an independent adverse factor for OS (P ¼ 0.032, relative risk: 1.70). In conclusion, ASXL1mut belong to the most frequent mutations in intermediate risk group AML. Their strong and independent dismal prognostic impact suggests the inclusion into the diagnostic work-up of AML.

Leukemia (2013) 27, 82–91; doi:10.1038/leu.2012.262 Keywords: ASXL1 mutations; AML; prognosis; intermediate karyotype

INTRODUCTION Several studies indicate that ASXL1 mutations occur frequently (AML) patients can be classified into in various myeloid malignancies, including myelodysplastic syn- different prognostic subgroups according to presence or absence dromes (MDS), AML, chronic myeloid leukemia, chronic myelo- 9,13–19,19 of distinct cytogenetic abnormalities. In the past years, various monocytic leukemia and myeloproliferative neoplasms, novel molecular genetic markers have been identified enabling and published data points to a poor prognostic impact in patients 14,18,19 further stratification of this heterogeneous disease. Screening for with these mutations. mutations in such as FLT3, NPM1, CEBPA, IDH1, IDH2, and In AML, the results regarding frequency and associations with RUNX1 allow a better prognostic prediction, in particular in AML karyotype abnormalities are quite diverse. In different studies, 15–17,20 with normal karyotype (NK) or intermediate cytogenetic risk ASXL1 mutations have been detected in about 6 to 30% of AML. profile.1–8 Furthermore, mutual exclusiveness of NPM1 mutations was 16 Recently, another promising candidate , ASXL1 (additional described. In a previous study, ASXL1 mutations occurred with sex combs-like 1), has been identified to be mutated in a similar frequency both in patients with NK (8.9%) and with 18 myeloproliferative neoplasms.9 The gene is located in the cytogenetic abnormalities (12.9%). Interestingly, there was not chromosomal region 20q11 encoding a of the polycomb only an inverse association observed with NPM1 mutations, but group and trithorax complex family. Mutations of ASXL1 can be also with FLT3-ITD and WT1 mutations. In addition, an association found particularly in exon 12 and virtually all are heterozygous.9 with RUNX1 mutations was found. Moreover, patients with ASXL1 Mainly frameshift and stop mutations were found that are pre- mutations had a shorter overall survival (OS), but the significance 18 dicted to lead to loss of the carboxyterminal plant homeodomain was lost in a multivariable analysis. A further study showed that finger on the protein level.10 This motif can be found in nuclear ASXL1 mutations identify a high-risk group of older patients within 21 involved in chromatin modifications. Indeed, ASXL1 can the ELN ‘favorable’ genetic category. interact with retinoic acid receptor and seems to be involved in To evaluate the impact and frequency of ASXL1 mutations in a chromatin remodeling, though the exact function remains large cohort of adult AML not selected for age but only for unknown thus far.11,12 karyotype, we here analyzed 740 cases with cytogenetically

1MLL Munich Leukemia Laboratory, Munich, Germany; 2Research Unit of Molecular Epidemiology, Helmholtz Zentrum Mu¨nchen, Munich, Germany; 3Hannover Unified Biobank, Hannover Medical School, Hannover, Germany; 4Institute of Epidemiology I, Helmholtz Zentrum Mu¨nchen, Munich, Germany; 5Department I of Internal Medicine, University at Cologne, Cologne, Germany; 6Department of Internal Medicine II, Klinikum Augsburg, Augsburg, Germany; 7Clinic for Hematology and Oncology, St Antonius Hospital, Eschweiler, Germany; 8Department for Hematology and Oncology, Klinikum Osnabru¨ck, Osnabru¨ck, Germany and 9Department for Hematology and Stemcell Transplantation, Asklepios Klinik St Georg, Hamburg, Germany. Correspondence: Dr S Schnittger, MLL Munich Leukemia Laboratory, Max-Lebsche-Platz 31, 81377 Munich, Germany. E-mail: [email protected] or URL: www.mll.com Received 20 August 2012; revised 3 September 2012; accepted 4 September 2012; accepted article preview online 11 September 2012; advance online publication, 28 September 2012 ASXL1 mutations in AML S Schnittger et al 83 intermediate-risk AML and detected a frequency of 17.2% ASXL1 cultures. Karyotypes, analyzed after G-banding, were described according mutations. There was an association to male sex, higher age, more to the International System for Human Cytogenetic Nomenclature.36 immature phenotype, aberrant karyotype, a strong positive Cytogenetic classification as ‘intermediate’ risk group was performed 24 correlation to RUNX1 mutations, and a negative correlation to according to the refined MRC criteria. Cytogenetic results were available NPM1, FLT3-ITD, FLT3-TKD and DNMT3A mutations. For the first for all patients in the study. Immunophenotyping was performed in 388 cases as described previously.37,38 time we could show that ASXL1 mutations have strong independent negative impact on survival. In larger subsets than previously reported, we could show that ASXL1 mutations are Statistical analysis stable mutations in paired diagnostic/relapse samples and are Survival curves were calculated for OS and event free survival (EFS) highly correlated with trisomy 8. In addition, we could show that according to Kaplan–Meier and compared using the two-sided log rank only frameshift and stop mutations in ASXL1 are somatic test. OS was the time from diagnosis to death or last follow-up. EFS was mutations. defined as the time from diagnosis to treatment failure, relapse, death or last follow-up in complete remission. Relapse was defined according to Cheson et al.39 Cox regression analysis was performed for OS and EFS with different parameters as covariates. Median follow-up was calculated taking PATIENTS, CONTROLS AND METHODS the respective last observations in surviving cases into account and Patients censoring non-surviving cases at the time of death. Results were All 740 patient samples were referred to our laboratory for first diagnosis of considered significant at Po0.05. Parameters that were significant in AML between September 2005 and September 2010. AML was diagnosed univariable analyses were included into multivariable analyses. according to the FAB (French-American-British) and WHO (World Health Dichotomous variables were compared between different groups using 2 Organization) classifications.22,23 Three hundred forty-five patients were the w -test and continuous variables by Student´s t-test. All reported P- female, 395 male and the median age was 66.9 years (range 18.5–100.4 values are two-sided. No adjustments for multiple comparisons were years). Five hundred fifty-three cases had a NK and 187 carried non- performed. SPSS (version 19.0.0) software (IBM Corporation, Armonk, recurrent intermediate risk aberrant cytogenetics (according to the refined NY, USA) was used for statistical analysis. MRC (United Kingdom Medical Research Council) classification24). Six hundred ninety-seven (94.2%) patients showed de novo AML, whereas 26 (3.5%) patients presented with secondary AML following either MDS or RESULTS myeloproliferative neoplasms, and 17 (2.3%) showed therapy-related AML. Frequency and characterization of ASXL1 alterations Data on the molecular markers NPM1, FLT3-ITD, MLL-partial tandem Overall, 135 ASXL1 alterations were detected in 134/740 patients duplication (PTD), CEBPA, and RUNX1 was available in all cases. In addition, data on other molecular markers were available for: FLT3-TKD: n ¼ 692, (18.1%). The majority of these alterations were frameshift IDH1R132: n ¼ 598 and IDH2R142 þ IDH2R172: n ¼ 534, WT1: n ¼ 587, NRAS: mutations caused by deletion or duplication of a nucleotide n ¼ 475; DNMT3A: n ¼ 204, TET2: n ¼ 166. Clinical follow-up data were (n ¼ 100; 74.1%). Further, 28 mutations (20.7%) were base available in 639 patients, but for prognostic analyses only de novo AML exchanges leading to a premature stop of translation. Seven with intensively treatment strategies (like standard protocols including alterations were single-base exchanges leading to missense ‘7 þ 3’ or combinations of chemotherapeutics, such as TAD (thioguanine, mutations (5.2%). cytarabine and daunorubicin) and HAM (high-dose cytarabine and mitoxantrone) were included (n ¼ 481). All patients gave their informed consent for scientific evaluations, for example, molecular studies. The study Detailed evaluation of molecular variants was approved by the Internal Review Board of the MLL Munich Leukemia To evaluate whether the detected alterations were somatic Laboratory and adhered to the tenets of the Declaration of Helsinki. mutations or even rare constitutional polymorphisms, we did (1) in silico analysis, (2) analysis of remission samples and (3) Healthy controls evaluation of healthy controls. (4) In addition, to assure the The KORA (Cooperative Health Research in the Region of Augsburg, validity of the detected muations in homopolymeric regions, we Germany) participants were selected from the F4 visit (2006–2008), the performed repeated testing to exclude sequencing artefatcs. follow-up survey for the KORA S4 cohort sample, recruited between 1999 and 2001.25 The KORA F4 visit population comprises 3080 male and female Repeated testings. G646WfsX12 (Gly646TrpfsX12) has repeatedly residents of the city and region of Augsburg in southern Germany. been discussed not to be a somatic mutation but more likely a Altogether 491 individuals from KORA were analyzed. polymorphism or even a sequencing artefact due to an 8-bp guanine homopolymere at that site.40 In this study, we excluded a Molecular analysis technical problem as all G646WfsX12 cases remained positive and Isolation of mononuclear cells, DNA extraction and mRNA extraction as all G646wt samples remained negative upon repeated testing of well as random primed cDNA synthesis were performed as described 20 samples up to 10 times. In addition, this aberration disappears previously.26 In 611 cases bone marrow and in 129 cases peripheral blood in remission (see below). were used for the molecular analysis. Screening for ASXL1 mutations in exon 12 was performed at the DNA In silico analysis. For in silico analysis we used two different level by direct Sanger sequencing of six different amplicons using BigDye algorithms: SIFT (http://sift.jcvi.org),41 and PolyPhen-2 (http:// terminator v1.1 cycle sequencing chemistry (Applied Biosystems, Weiter- genetics.bwh.harvard.edu/pph2/index.shtml).42 All frameshift and stadt, Germany). The primers for PCR and sequencing were described nonsense mutations were predicted to confer a damaging previously.13 For PCR Qiagen Master Mix (Qiagen, Hilden, Germany) was used; solely for amplicon 12.4, the GC-rich-Kit was used (Roche Applied character for the protein structure of ASXL1. Results of the Science, Mannheim, Germany). prediction analysis were ambigious: whereas the missense Analyses for mutations of NPM1, FLT3-TKD, NRAS, KRAS, DNMT3A, RUNX1, mutations were predicted to be damaging in some instances they TET2, WT1, IDH1, IDH2, CEBPA as well as MLL-PTD and FLT3-ITD were were predicted to be tolerated in others and thus in part was described previously.5,26–34 inconsistent between the different methods (Supplementary Table 1).

Cytomorphology, cytogenetics and immunophenotyping Analysis of remission samples. In five cases with a missense Cytomorphologic assessment was based on May–Gru¨nwald–Giemsa stains, mutation, remission material was available. All these five cases myeloperoxidase reaction and non-specific esterase using alpha-naphtyl- were also positive for an NPM1 mutation at diagnosis. For the acetate as described before and was performed according to the criteria detection of the NPM1 mutation a highly sensitive real-time PCR À 5 À 6 43 defined in the French-American-British and the World Health Organization assay with a sensitivity of 10 –10 was available. Although all classifications.22,23,35 Cytogenetic studies were performed after short-term cases were negative for the NPM1mut in the respective remission

& 2013 Macmillan Publishers Limited Leukemia (2013) 82 – 91 ASXL1 mutations in AML S Schnittger et al 84 sample, in all cases the missense exchange in ASXL1 was retained for heterogeneity between the three groups). With respect to with a load of 50%, which is highly suggestive of a constitutional morphology, ASXL1mut were more frequent in French-American- polymorphism. Thus, these cases were assigned to the ASXL1 wild- British M2 (58/240; 24.2%) compared with all other subtypes (65/ type group. For the remaining two missense mutations no 474; 13.7%; P ¼ 0.001) as well as in M5a (8/18, 44.4%) compared remission samples were available, but these were clearly with all other subtypes (115/696; 16.5%; P ¼ 0.006), but less assigned as tolerated with in silico analysis. In patients with the frequent in M1 (19/221, 8.6%) compared with all other subtypes p.G646WfsX12 mutation who achieved a complete remission after (104/493, 21.1%; Po0.001). chemotherapy, this variant disappeared (n ¼ 9) or its load In 388 cases, immunophenotyping data were available. Cases diminished in the respective remission sample (n ¼ 7). Thus, this with ASXL1mut (n ¼ 66) had a stronger expression of CD13 (% variant was proven to be a somatic mutation. An example is positive cells, mean±s.d., 51±26 vs 43±25%, P ¼ 0.025), CD34 depicted in Supplementary Figure 1. (44±28% vs 29±29%, Po0.001), CD133 (27±24% vs 20±25%, P ¼ 0.047) and HLA-DR (40±24% vs 33±24%, P ¼ 0.034) as well as Evaluation of healthy controls from KORA (n ¼ 491). We analyzed a weaker expression of CD33 (68±25% vs 75±23%, P ¼ 0.014) 491 age and sex matched individuals from a population-based and thus had a more immature immunophenotype as compared cohort panel (KORA ¼ Cooperative Health Research in the Region with ASXL1wt. of Augsburg, Germany). In KORA, only one p.G646WfsX12 positive sample (0.2%) with a mutation load of only 10% was identified. Association with karyotype This was in contrast to the leukemia samples that all had a mutation load of around 50%. This incidence in the KORA cohort is The total cohort was comprised of 553 cases with NK and 187 significantly below the one observed in myeloid malignancies.44 cases with intermediate risk karyotype aberrations comprising the Because of the low mutation load the one positive case was following recurrent aberrations: trisomy 8 (n ¼ 74/187 39.6%); loss interpreted as presumably having a small pre-malignant clone or of Y (n ¼ 13/109; 11.9%), trisomy 13 (n ¼ 10/187; an early yet undetected clonal disease. From this data we 5.3%) and trisomy 21 (n ¼ 10/187; 4.9%). ASXL1mut were more conclude that the p.G646WfsX12 is a somatic mutation and was frequent in patients with aberrant karyotype than in those with NK regarded as true mutation throughout the paper. (Po0.001, Table 1). Particularly, a strong correlation to trisomy 8 Besides the common silent p.Ser1253Ser (302/491, 61.5%) with was observed as 39 of these 74 cases (52.7%) had an ASXL1 14.5% (71/491) homozygotes and the intronic c.*22 A4G (292/ mutation compared with only 19 of 113 (16.8%) in other aberrant 491, 59.5%) with 11.4% homozygotes, 26 different rare missense karyotypes (Po0.001). mutations were detected in 44 individuals of the KORA cohort with frequencies between 0.1 and 0.4% (Supplementary Table 2). Association with other molecular mutations Only few of them have been assigned as polymorphisms before Generally, ASXL1mut were observed together with all other (Supplementary Table 2). In addition, 11 further rare variants were molecular mutations. There was a strong correlation with RUNX1mut described in the literature that were not detected in the KORA (Po0.001) and a trend to increased frequency in IDH2mut cases cohort (Supplementary Table 3). This data suggest that a number (P ¼ 0.079). A negative correlation was found for NPM1mut of rare inborn variants exist in ASXL1. (Po0.001), FLT3-ITD (Po0.001), FLT3-TKD (P ¼ 0.001), DNMT3Amut In conclusion, this data suggest that all missense mutations in (Po0.001) and a negative trend for WT1mut (P ¼ 0.068). No ASXL1 are inborn polymorphisms and in the following only frameshift significant associations were observed for CEBPA, RUNX1, IDH1R132, and stop mutations in ASXL1 were regarded as somatic mutations. NRAS and TET2 mutations. A detailed description of the mutation coincidences is given in Table 2 as well as in Figure 2. Frequency and characterization of somatic ASXL1 mutations After exclusion of the missense alterations, a total of 128 somatic Stability during follow-up mutations were observed in 127 cases. All had a deleterious effect Paired samples of diagnosis and relapse time points were on the protein structure due to their character as stop or available in 16 cases with an ASXL1 mutation at diagnosis. At frameshift mutations. In detail, the most frequent mutation was diagnosis, nine of these cases had a NK, six had trisomy 8 and one p.Gly646TrpfsX12 (n 69, 53.9%). The p.Gly646TrpfsX12 at the ¼ had trisomy 11. In addition, in 15/16 patients one or two protein level was a result of c.1934dupG (n 65), c.1927_1928insA ¼ additional molecular mutations were detected (in 5 and 10 cases, (n 2) or c.1935dupT (n 2) at the DNA level. The second most ¼ ¼ respectively). At relapse, all ASXL1 mutations and all other frequent mutation was p.Glu635ArgfsX15 (n 18), followed by ¼ molecular mutations (with the exception of one BCOR mutation) p.Tyr591X (n 5), p.Arg693X (n 4) and p.Gln733X (n 2). The ¼ ¼ ¼ were retained. Thus, ASXL1 is a stable mutation. However, this remaining 30 mutations were non-recurrent consisting of 14 pattern does not allow any conclusions on the hierarchy of all frameshift and 16 nonsense mutations. The majority of the these mutations. mutations were detected with a mutation/wild-type load of In contrast, in 5 of 14 cases (with available cytogenetics at 40–50%. One of the patients had two ASXL1 mutations: a p.Gln829X relapse) additional chromosomal aberrations were detected at with a mutation/wild-type load of 10% and a p.Ala1172LeufsX2 relapse, which were not present at diagnosis (Table 3). We would with a mutation/wild-type load of 50%. The positions of the like to outline that within this cohort of 16 relapsed AML 11 cases mutations within the gene are indicated in Figure 1. also had a RUNX1mut, one a CEBPAmut, and two an NPM1mut. Only two cases did not reveal any of these three mutations and Association with biological characteristics one of these two even relapsed with a t(8;21)(q22;q22)/RUNX1- ASXL1mut were more frequent in males than in females (93/395, RUNX1T1. In addition to ASXL1mut, this particular case also was 23.5% vs 34/345, 9.9%, Po0.001) and were associated with higher IDH2R140 mutated at both time points. Thus, this represents a age (mean±s.d. 71.8±9.4 vs 61.8±14.9 years, Po0.001) and very unusual case with ASXL1 and IDH2R140 mutations at lower white blood cell (WBC) counts (mean±s.d. 34.2±49.6 vs diagnosis and additional t(8;21)(q22;q22) at relapse 17 months 46.8±61.9 Â 109/l, P ¼ 0.025) (Table 1). There was no association later. The RUNX1-RUNX1T1 was backtracked with highly sensitive of ASXL1mut to platelet counts or hemoglobin levels. ASXL1mut real-time PCR and nested PCR but was not present at diagnosis. were detected more frequently in s-AML after MDS/myeloproli- In addition, in four cases with s-AML paired samples from the ferative neoplasms (11/26; 42.3%) compared with de novo AML MDS phase were analyzed. All four cases were ASXL1 mutated (114/697; 16.4%) and therapy-related AML (2/17; 11.8%) (P ¼ 0.002 already at the MDS phase of the disease. One case was in addition

Leukemia (2013) 82 – 91 & 2013 Macmillan Publishers Limited ASXL1 mutations in AML S Schnittger et al 85

Figure 1. Localization of all 128 ASXL1 mutations within the coding region. Each single mutation is indicated as a dot (frameshifts in red, nonsense in blue). Missense mutations (polymorphisms) are indicated below the diagram. ASXN, conserved domain at the N-terminus; ASXM, conserved domain in the middle part; Gly, glycine-rich region; Rb, Rb interacting motif; NR box, domain interacting with RAR and RXR; PHD, plant homedomain finger.

Table 1. Patient characteristics

Total cohort ASXL1 wild-type ASXL1 mutated P value

All patients n ¼ 740 n ¼ 613 (82.8%) n ¼ 127 (17.2%) Female 345 (46.6%) 311 (90.1%) 34 (9.9%) o0.001 Male 395 (53.4%) 302 (76.5%) 93 (23.5%) Age (years) Mean: 63.6, s.d.±(14.6) Mean: 61.8, s.d.±(14.9) Mean: 71.8, s.d.±(9.4) o0.001 WBC count ( Â 109/l) Mean: 44.6, s.d.±(60.1) Mean: 46.8, s.d.±(61.9) Mean: 34.2, s.d.±(49.6) 0.025 Platelet count ( Â 109/l) Mean: 89.7, s.d.±(90.8) Mean: 89.6, s.d.±(92.0) Mean: 90.4, s.d.±(85.9) NS Hb (g/dl) Mean: 9.3, s.d.±(2.0) Mean: 9.3, s.d.±(2.0) Mean: 9.4, s.d.±(1.8) NS

FAB subtype n ¼ /percenta n ¼ /percentb n ¼ /percentb Total available 714 (100%) 591 (82.8%) 123 (17.2%) AML M0 36 (5.0%) 26 (72.2%) 10 (27.8%) NSc AML M1 221 (31.0%) 202 (91.4%) 19 (8.6%) o0.001c AML M2 240 (33.6%) 182 (75.8%) 58 (24.2%) 0.001c AML M4 168 (23.5%) 145 (86.3%) 23 (13.7%) NSc AML M5a 18 (2.5%) 10 (55.6%) 8 (44.4%) 0.006c AML M5b 11 (1.5%) 9 (81.8%) 2 (18.2%) NSc AML M6 18 (2.5%) 15 (83.3%) 3 (16.7%) NSc AML M7 2 (0.3%) 2 (100.0%) 0 NSc

History of disease De novo AML 697 (94.2%) 583 (83.6%) 114 (16.4%) 0.002 Secondary AML 26 (3.5%) 15 (57.7%) 11 (42.3%) Therapy related AML 17 (2.3%) 15 (88.2%) 2 (11.8%)

Cytogenetics Normal karyotype 553 (74.7%) 484 (87.5%) 69 (12.5%) o0.001 Aberrant karyotype 187 (25.3%) 129 (69.0%) 58 (31.0%) Abbreviations: AML, acute myeloid leukemia; FAB, French-American-British; NS, not significant; WBC, white blood cell. a% of total cohort with FAB available. b% wild-type and mutated, respectively, within FAB subtype. cFAB subtype vs all other FAB subtypes.

& 2013 Macmillan Publishers Limited Leukemia (2013) 82 – 91 ASXL1 mutations in AML S Schnittger et al 86 Table 2. Coincidence of ASXL1mut with other mutations Po0.001) and EFS (median: 9.1 vs 16.3 months in ASXL1wt, P ¼ 0.012) (Figures 3a and b). Mutation ASXL1 wild- ASXL1 P In a next step, patients were subdivided according to age X60 (n ¼ cases analyzed) type n (%) mutated years (ASXL1wt: n ¼ 217, ASXL1mut: n ¼ 47) and o60 years n (%) (ASXL1wt: n ¼ 213; ASXL1mut: n ¼ 10). In the younger as well as in the older cohort, OS was shorter in the ASXL1mut compared NPM1 (n ¼ 740) with the ASXL1wt subset (median: 11.5 vs 36.3 months, P ¼ 0.040 Wt 285 (46.5%) 119 (93.7%) ko0.001 Mut 328 (53.5%) 8 (6.3%) in the older and not reached for both in the younger (median 2 years survival: 60% vs 78%, P ¼ 0.049)) (Figures 3c and d). FLT3-ITD (n ¼ 740) Furtheron, also within the cohort of patients with NK (n ¼ 376) Negative 449 (73.2%) 114 (89.9%) ko0.001 patients harboring an ASXL1mut had shorter OS (median: 9.8 vs Positive 164 (26.8%) 13 (10.2%) 62.2 months in ASXL1wt, Po0.001) and EFS (median: 7.5 vs 17.7 months in ASXL1wt, P ¼ 0.001) (Figures 3e and f). In contrast, in the MLL-PTD (n ¼ 740) cohort with aberrant intermediate karyotypes, the difference of OS Wt 566 (92.3%) 120 (94.5%) NS between ASXL1mut (n ¼ 24) and ASXL1wt (n ¼ 81) (median: 20.0 vs Mut 47 (7.7%) 7 (5.5%) 36.7 months) was not significant. CEBPA (n ¼ 740) As we observed a high coincidence of ASXL1 with RUNX1 Wt 544 (88.7%) 110 (86.6%) NS mutations, which were previously shown to have a negative Mut 69 (11.3%) 17 (13.4%) impact on prognosis, we also investigated the prognostic impact of ASXL1mut according to RUNX1 mutational status. The prog- RUNX1 (n ¼ 740) nostically adverse effect of ASXL1mut was seen within the Wt 515 (84.0%) 71 (55.9%) mo0.001 RUNX1wt cohort (n ¼ 408; median OS: 10.1 vs 62.2 months, Mut 98 (16.0%) 56 (44.1%) P ¼ 0.001) (Figure 4a), but there was only a trend toward an adverse effect in the RUNX1mut subgroup (n ¼ 73; median OS: FLT3-TKD (n ¼ 692) Wt 514 (88.9%) 112 (98.2%) k0.001 15.3 vs 24.9 months, NS) (Figure 4b). Mut 64 (11.1%) 2 (1.8%) The following parameters were tested in univariable Cox regression analyses for impact on OS and EFS: sex, age, WBC IDH1R132 (n ¼ 598) count, platelet count, hemoglobin level, cytogenetics (normal vs Wt 449 (88.6%) 83 (91.2%) NS aberrant karyotype), and mutational status of ASXL1, NPM1, FLT3- Mut 58 (11.4%) 8 (8.8%) ITD, MLL-PTD, CEBPA, RUNX1, FLT3-TKD, IDH1, IDH2, WT1, NRAS, DNMT3A and TET2. A significant negative impact on OS was shown IDH2R140 or for higher age (Po0.001, relative risk (RR) per decade: 1.50), higher IDH2R172 (n ¼ 534) WBC count (P 0.001, RR per 10 Â 109/l: 1.07), ASXL1 mutations Wt 380 (84.1%) 62 (75.6%) k0.079 o Mut 72 (15.9%) 20 (24.4%) (Po0.001; RR: 2.23), FLT3-ITD (P ¼ 0.002, RR: 1.69), MLL-PTD (P ¼ 0.006, RR: 2.06), and RUNX1 mutations (P ¼ 0.010, RR: 1.64). WT1 (n ¼ 587) A favorable impact was found for biallelic CEBPA mutations Wt 470 (94.0%) 86 (89.9%) k0.068 (P ¼ 0.009, hazard ratio: 0.55). No impact was found for the other Mut 30 (6.0%) 1 (1.1%) parameters. A significant negative impact on EFS was found for higher age NRAS (n ¼ 475) (Po0.001, RR per decade: 1.03), higher WBC count (Po0.001, RR per Wt 345 (83.9%) 51 (79.7%) NS 10 Â 109/l: 1.06), ASXL1 mutations (P¼ 0.013; RR: 1.59), FLT3-ITD Mut 66 (16.1%) 13 (20.3%) (P¼ 0.021, RR: 1.38), MLL-PTD (P¼ 0.010, RR: 1.79) and RUNX1 mutations DNMT3A (n ¼ 204) (P¼ 0.030, RR: 1.42). No impact was found for the other parameters. Wt 109 (66.9%) 39 (95.1%) ko0.001 In multivariable analysis, ASXL1mut revealed an independent Mut 54 (33.1%) 2 (4.9%) prognostic impact on OS (P ¼ 0.028, RR: 1.73) besides age (Po0.001, RR per decade: 1.51), WBC count (Po0.001, RR per TET2 (n ¼ 166) 10 Â 109/l: 1.06) and FLT3-ITD status (P ¼ 0.049, RR: 1.40). In Wt 103 (67.2%) 10 (76.9%) NS multivariable analysis for EFS, ASXL1mut revealed no independent Mut 50 (32.7%) 3 (23.1%) impact and only age (Po0.001, RR per decade: 1.33), WBC count 9 Abbreviations: FLT3-ITD, FLT3-internal tandem duplication; MLL-PTD, MLL- (Po0.001, RR per 10 Â 10 /l: 1.05) and FLT3-ITD (P ¼ 0.046, RR: partial tandem duplication; mut, mutated; NS, not significant; wt, 1.39) were associated with outcome (Table 4). wildtype;mASXL1 mutations more frequent in cases mutated in the respective gene, kASXL1 mutations less frequent in cases mutated in the respective gene. DISCUSSION Mutations in ASXL1 have been reported in various myeloid malignancies but have not been intensively studied in AML. Still, CEBPAmut and IDH2R140mut at both time points, one gained an the incidence, associations with other molecular markers and IDH2R140mut and a trisomy 8 at the time point of diagnosis of associations with biologic characteristics were reported variably, AML. The third case was RUNX1mut at both time points and mainly because of selected cohorts or different ethnical back- gained an FLT3-ITD at the time point of transformation to AML, grounds of the analyzed cohorts.16–18 We concentrated on adult while in the fourth case, the ASXL1 mutation was the sole AML with intermediate risk karyotype independent of age. In mutation detected at both time points. addition, as mutations have been shown to cluster in exon 12 and were detected very rarely outside this region,45 we sequenced only exon 12, which actually comprises B50% of the whole Prognostic relevance of ASXL1 mutations coding region of the gene. We show that ASXL1 mutations Only patients with de novo AML who received intensive treatment occurred in 17.2% and therewith belong to the most common (n ¼ 481) were included into the prognostic analyses. Patients with molecular markers mutated in the cytogenetic intermediate risk ASXL1mut had shorter OS (11.0 vs 62.2 months in ASXL1wt, group AML. They are associated with distinct clinical and

Leukemia (2013) 82 – 91 & 2013 Macmillan Publishers Limited ASXL1 mutations in AML S Schnittger et al 87

Figure 2. Distribution and frequency of ASXL1mut and other molecular mutations in the total cohort of 740 patients. Red indicates a mutation within the respective gene, gray indicates no mutation. White cells indicate that the respective gene mutation was not analyzed for this patient. Patients are depicted vertically.

Table 3. Molecular and cytogenetic analysis of paired diagnostic and relapse samples

ASXL1 status Second mutation Third mutation Cytogenetics

1D ASXL1mut RUNX1mut FLT3-ITD 47,XY, þ 8 1R ASXL1mut RUNX1mut FLT3-ITD 46,XY, þ 14,der(14;18) (q10;q10) 2D ASXL1mut CEBPAmut IDH2R140 47,XY, þ 11 2R ASXL1mut CEBPAmut IDH2R140 47,XY, þ 11 3D ASXL1mut 47,XY, þ 8 3R ASXL1mut 46,XY,del(5)(q12q34), del(11)(q13), þ r(11)(p14q13), der(17)t(11;17) (q13;p13) 4D ASXL1mut NPM1_A IDH2R140 46,XY 4R ASXL1mut NPM1_A IDH2R140 46,XY 5D ASXL1mut RUNX1mut IDH2R140 46,XY 5R ASXL1mut RUNX1mut IDH2R140 46,XY 6D ASXL1mut RUNX1mut 47,XY, þ 8 6R ASXL1mut RUNX1mut 47,XY, þ 8 7D ASXL1mut RUNX1mut 47,XX, þ 8 7R ASXL1mut RUNX1mut 47,XX, þ 8 8D ASXL1mut RUNX1mut BCORmut 46,XY 8R ASXL1mut RUNX1mut BCORwt NA 9D ASXL1mut IDH2R140 46,XY 9R ASXL1mut IDH2R140 46,XY,t(8;21)(q22;q22) 10 D ASXL1mut RUNX1mut IDH1R132 47,XY, þ 8 10 R ASXL1mut RUNX1mut IDH1R132 47,XY, þ 8 11 D ASXL1mut RUNX1mut 46,XX 11 R ASXL1mut RUNX1mut 46,XX,del(6)(q14q27) 12 D ASXL1mut RUNX1mut IDH2R140 46,XY 12 R ASXL1mut RUNX1mut IDH2R140 46,XY 13 D ASXL1mut RUNX1mut NRAS 47,XY, þ 8 13 R ASXL1mut RUNX1mut NRAS 47,XY, þ 8 14 D ASXL1mut NPM1_A 46,XY 14 R ASXL1mut NPM1_A 46,XY 15 D ASXL1mut RUNX1mut NRAS 46,XY 15 R ASXL1mut RUNX1mut NRAS NA 16 D ASXL1mut RUNX1mut IDH1R132 46,XY 16 R ASXL1mut RUNX1mut IDH1R132 47,XY, þ 11 Abbreviations: D, diagnosis; FLT3-ITD, FLT3-internal tandem duplication; ID, patient ID; R, relapse.

biological features like male sex, higher age, immature to the high coincidence with RUNX1mut, which is also a highly immunophenotype, concomitant RUNX1 and IDH2 mutations, adverse prognostic marker in AML.7,8 We confirmed a high aberrant intermediate risk karyotype, especially trisomy 8, and correlation of ASXL1 mutations and RUNX1 mutations. In a with adverse prognosis. multivariable analysis on OS, taking age, WBC count, ASXL1, Similar findings have been reported in a Taiwanese popula- FLT3-ITD, MLL-PTD and RUNX1 status into account, however, we tion.18 However, in this study, there was no independent found ASXL1 mutations to be an independent adverse prognostic prognostic effect of ASXL1mut and this was discussed to be due factor for OS (P ¼ 0.032).

& 2013 Macmillan Publishers Limited Leukemia (2013) 82 – 91 ASXL1 mutations in AML S Schnittger et al 88 Total Cohort

ASXL1wt (n=430; 62.2 months)

ASXL1wt (n=430; 16.3 months)

ASXL1mut ASXL1mut (n=51; 11.0 months) (n=51; 9.1 months)

p<0.001 p=0.012

< 60 years ≥60 years

ASXL1wt (n=213; 2-year survival 78%)

ASXL1wt ASXL1mut (n=217; 36.3 months) (n=10; 2-year survival 60%)

ASXL1mut p=0.049 p=0.040 (n=41; 11.0 months)

Normal Karyotype Cohort

ASXL1wt (n=349; 62.2 months)

ASXL1wt (n=349; 17.7 months)

ASXL1mut (n=27; 9.8 months) ASXL1mut (n=27; 7.5 months) p<0.001 p=0.001

Figure 3. Outcome within the total cohort with survival data (n ¼ 481). Kaplan–Meier plot showing inferior (a) overall and (b) event free survival of the ASXL1 mutated cases (red) compared with ASXL1 wild-type cases (gray). Median values for OS and EFS are indicated. (c)OS restricted to patients o60 years (n ¼ 223). (d) OS restricted to patients X60 years (n ¼ 258). (e) Survival within the normal karyotype subcohort (n ¼ 376). Kaplan–Meier plot showing inferior overall and (f) event free survival of the ASXL1 mutated cases (red) compared with ASXL1 wild- type cases (gray). Median values for OS and EFS are indicated.

In a previous study, instability of ASXL1mut was reported as thus a hierarchical pattern of mutation could not be identified. In two of six patients lost ASXL1mut at relapse or even in primary contrast, in four s-AML cases that were backtracked to the MDS refractory disease.18 In our cohort, 16 combined diagnosis/relapse phase the ASXL1mut were already present at the MDS stage of the samples were available and all these cases retained the same disease and additional aberrations (FLT3-ITD, IDH2R140 or trisomy 8) ASXL1 mutation at relapse. Almost all of these cases had two were gained in AML transformation suggesting that ASXL1 additional mutations of different classes: (1) RUNX1mut (11/16), mutations are an early event in transformation. NPM1mut (2/16), CEBPAmut (1/16). (2) IDH1R132 (2/16), IDH2R140 A previous study has shown that ASXL1 mutations are five time (5/16), FLT3-ITD (1/16), NRAS (2/16), BCOR (1/3). With the exception as frequent in patients older than 60 years and are associated with of the BCOR mutation, all mutations were stable at relapse and high risk in the ‘favorable’ cytogenetic category according to ELN

Leukemia (2013) 82 – 91 & 2013 Macmillan Publishers Limited ASXL1 mutations in AML S Schnittger et al 89 RUNX1wt RUNX1mut

ASXL1wt (n=380; 62.2 months)

ASXL1wt (n=50; 24.9 months)

ASXL1mut

(n=28; 10.1 months) ASXL1mut (n=23; 15.3 months)

p<0.001 n.s.

Figure 4. Survival according to RUNX1 mutation status. (a) Within the subgroup of RUNX1wt patients (n ¼ 408) an inferior OS was observed for ASXL1 mutated cases (red) compared with ASXL1 wild-type cases (gray). (b) Within the subgroup of RUNX1mut (n ¼ 73) ASXL1 mutation status (red) resulted only in a non-significantly inferior OS as compared with ASXL1 wild-type status (gray). Median values for OS and EFS are indicated.

Table 4. Cox regression analysis

Overall survival Event free survival

Univariable Multivariable Univariable Multivariable

P HR P HR P HR P HR

Sex NS - - - NS - - - Age o0.001 1.50a o0.001 1.52a o0.001 1.03a o0.001 1.33a WBC count o0.001 1.07b o0.001 1.06b o0.001 1.06b o0.001 1.05b Platelet count NS - - - NS - - - Hemoglobin level NS - - - NS - - - Karyotype (NK vs AK) NS - - - NS - - - ASXL1 mutation o0.001 2.23 0.028 1.73 0.013 1.59 NS - NPM1 mutation NS NS - - - FLT3-ITD 0.002 1.69 0.049 1.49 0.021 1.38 0.046 1.39 MLL-PTD 0.006 2.06 NS - 0.010 1.79 NS - CEBPA biallelic 0.009. 0.55 NS - NS - - - RUNX1 mutation 0.010 1.64 NS - 0.030 1.42 NS FLT3-TKD mutation NS - - - NS - - - IDH1R132 NS - - - NS - - - IDH2R140 or IDH2R172 NS - - - NS - - - WT1 mutation NS - - - NS - - - NRAS mutation NS - - - NS - - - DNMT3A mutation NS - - - NS - - - TET2 mutation NS - - - NS - - - Abbreviations: AK, aberrant karyotype; FLT3-ITD, FLT3-internal tandem duplication; HR, hazard ratio; MLL-PTD, MLL-partial tandem duplication; NK, normal karyotype; NS, not significant; WBC, white blood cell. aPer 10 years of increase; bPer 10 Â 109/l. criteria.21 We confirmed the considerably higher frequency in relapse. In this line, during the past years it has become clear older patients. In addition, we showed that ASXL1 mutations have that the concept of a two-hit event in leukemia with a classical a negative impact on outcome in the AML with intermediate risk ‘type 1’ (proliferation) and ‘type 2’ (differentiation) mutation46 karoytpe and also in the subset with NK. In addition, ASXL1 cannot fully explain all recent findings on molecular mutations. mutations were also correlated with adverse outcome in AML More and more mutations have been shown to be important for o60 years. the development of leukemia including alterations in genes ASXL1 and NPM1 mutations have been suggested to be relevant for genomic stability like TP53,47 metabolic enzymes mutually exclusive.16 Different routes of leukemogenesis rather (IDH1, IDH2, ND4)5,48,49 or proteins with effects on epigenetic than two alternate hits on the same route were discussed. We modification (TET2, EZH2, DNMT3A)50,51.52 This leads to a high confirmed a negative correlation of ASXL1 and NPM1, however, in probability of multiple mutations from different pathways are our cohort in 8 ASXL1mut cases, we also detected NPM1 randomly combined and thereby underlie the pathogenesis of mutations. This leads to the more likely hypothesis that certain AML. routes of gene mutations are more prevalent than others but do All ASXL1 mutations detected in this study were heterozygous, not exclude each other. This is supported by one of our cases which is consistent with the hypothesis of a dominant negative demonstrating both ASXL1mut and RUNX1mut at diagnosis and at effect of truncated ASXL1 proteins. All mutations were either (1) relapse who additionally gained a t(8;21)/RUNX1/RUNX1T1 at frameshift mutations caused by deletion or duplication of one

& 2013 Macmillan Publishers Limited Leukemia (2013) 82 – 91 ASXL1 mutations in AML S Schnittger et al 90 nucleotide or (2) base exchanges leading to stop mutations. The 3 Swerdlow SH, Campo E, Harris NL, Jaffe ES, Pileri SA, Stein H et al. WHO Classifi- most common mutation was p.Gly646TrpfsX12, which accounted cation of Tumours of Haematopoietic and Lymphoid Tissues. 4th edn., Lyon: for 53.9% of all deleterious mutations, followed by International Agency for Research on Cancer (IARC); 2008. p.Glu635ArgfsX15 in 14.2%. Despite previous suggestions that 4 Preudhomme C, Sagot C, Boissel N, Cayuela JM, Tigaud I, De Botton S et al. these mutations40 may be germline or even technical artifacts, we Favorable prognostic significance of CEBPA mutations in patients with de novo could clearly show that these are true somatic mutations. This acute myeloid leukemia: a study from the Acute Leukemia French Association conclusion was based on analysis of a large healthy control cohort, (ALFA). Blood 2002; 100: 2717–2723. analysis of remission samples and repeated testings. In constrast, 5 Schnittger S, Haferlach C, Ulke M, Alpermann T, Kern W, Haferlach T. IDH1 mutations are detected in 6.6% of 1414 AML patients and are associated with we could show that missense mutation in ASXL1 are highly likely intermediate risk karyotype and unfavorable prognosis in adults younger than 60 to be always rare inborn polymorphisms. years and unmutated NPM1 status. Blood 2010; 116: 5486–5496. 17 As has been suggested before the majority of mutations 6 Boissel N, Nibourel O, Renneville A, Gardin C, Reman O, Contentin N et al. (59%) in our cohort are localized to one particular region within Prognostic impact of isocitrate dehydrogenase enzyme isoforms 1 and 2 muta- exon 12, around the Gly-rich domain spanning amino acids 642– tions in acute myeloid leukemia: a study by the Acute Leukemia French Asso- 685 (Figure 1). Three mutations that lead to a truncated protein ciation group. J Clin Oncol 2010; 28: 3717–3723. are located upstream of the C-terminal nuclear receptor box 7 Tang JL, Hou HA, Chen CY, Liu CY, Chou WC, Tseng MH et al. AML1/RUNX1 (amino acid 1107–1112), which is predicted to interact with the mutations in 470 adult patients with de novo acute myeloid leukemia: prognostic retinoic acid receptor. The predicted truncated protein would lack implication and interaction with other gene alterations. Blood 2009; 114: 5352–5361. its plant homeodomain, thus compromising the function of the 9 8 Schnittger S, Dicker F, Kern W, Wendland N, Sundermann J, Alpermann T et al. associated chromatin modifiers. Although the function of ASXL1 is RUNX1 mutations are frequent in de novo AML with noncomplex karyotype and not completely understood, the presented defects suggest an confer an unfavorable prognosis. Blood 2011; 117: 2348–2357. important role in pathogenesis of AML. 9 Carbuccia N, Murati A, Trouplin V, Brecqueville M, Adelaide J, Rey J et al. Mutations In conclusion, our findings indicate that ASXL1 mutations are of ASXL1 gene in myeloproliferative neoplasms. Leukemia 2009; 23: 2183–2186. one of the most commonly occurring molecular mutations in 10 Vainchenker W, Delhommeau F, Constantinescu SN, Bernard OA. New mutations intermediate risk AML and they have to be considered to and pathogenesis of myeloproliferative neoplasms. Blood 2011; 118: 1723–1735. significantly contribute to leukemogenesis. There is a strong 11 Cho YS, Kim EJ, Park UH, Sin HS, Um SJ. Additional sex comb-like 1 (ASXL1), in association of ASXL1 mutations with male sex, MDS prephase, cooperation with SRC-1, acts as a ligand-dependent coactivator for retinoic acid receptor. J Biol Chem 2006; 281: 17588–17598. higher age, immature immunophenotype and mutations in 12 Lee SW, Cho YS, Na JM, Park UH, Kang M, Kim EJ et al. ASXL1 represses retinoic RUNX1. Given their strong and independent dismal prognostic acid receptor-mediated transcription through associating with HP1 and LSD1. impact, ASXL1 mutations should be included in the diagnostic J Biol Chem 2010; 285: 18–29. work-up of patients with cytogenetically intermediate-risk AML. 13 Gelsi-Boyer V, Trouplin V, Adelaide J, Bonansea J, Cervera N, Carbuccia N et al. Mutations of polycomb-associated gene ASXL1 in myelodysplastic syndromes and chronic myelomonocytic leukaemia. Br J Haematol 2009; 145: 788–800. CONFLICT OF INTEREST 14 Gelsi-Boyer V, Trouplin V, Roquain J, Adelaide J, Carbuccia N, Esterni B et al. ASXL1 mutation is associated with poor prognosis and acute transformation in chronic SS, WK, CH, and TH are part owners of the MLL Munich Leukemia Laboratory. CE, SJ, myelomonocytic leukaemia. Br J Haematol 2010; 151: 365–375. TA, AF, VG and AK are employed by the MLL Munich Leukemia Laboratory. 15 Abdel-Wahab O, Manshouri T, Patel J, Harris K, Yao J, Hedvat C et al. Genetic analysis of transforming events that convert chronic myeloproliferative neo- plasms to leukemias. Cancer Res 2010; 70: 447–452. ACKNOWLEDGEMENTS 16 Carbuccia N, Trouplin V, Gelsi-Boyer V, Murati A, Rocquain J, Adelaide J et al. Mutual exclusion of ASXL1 and NPM1 mutations in a series of acute myeloid We thank all coworkers in our laboratory for their excellent technical assistance. We leukemias. Leukemia 2010; 24: 469–473. also thank Hubert Serve, Johann Wolfgang Goethe-University, Frankfurt; Dietrich 17 Boultwood J, Perry J, Pellagatti A, Fernandez-Mercado M, Fernandez-Santamaria Braumann, Asklepios Klinik Altona, Hamburg; Hermann-Josef Pielken, St Johannes C, Calasanz MJ et al. Frequent mutation of the polycomb-associated gene ASXL1 Hospital, Dortmund; Clemens-Martin Wendtner, Klinikum Schwabing, Munich; Tanja in the myelodysplastic syndromes and in acute myeloid leukemia. Leukemia 2010; Hesse, Klinikum Lippe, Lemgo; Hans-Jo¨ rg Weh, Franziskus Hospital, Bielefeld; Ju¨rgen 24: 1062–1065. Wehmeyer, Gemeinschaftspraxis fu¨r Ha¨matologie und Onkologie, Mu¨ nster; Heinz- 18 Chou WC, Huang HH, Hou HA, Chen CY, Tang JL, Yao M et al. Distinct clinical and Gert Ho¨ffkes, Klinikum Fulda; Michael Flasshove, Krankenhaus Du¨ren, Du¨ren; Michael biological features of de novo acute myeloid leukemia with additional sex comb- Rummel, Justus Liebig University, Gieen; Christian Peschel, Klinikum Rechts der Isar like 1 (ASXL1) mutations. Blood 2010; 116: 4086–4094. der Technischen Universita¨tMu¨nchen, Munich; Andreas Neubauer, Philipps 19 Thol F, Friesen I, Damm F, Yun H, Weissinger EM, Krauter J et al. Prognostic University, Marburg and all other physicians for referring samples to our center. significance of ASXL1 mutations in patients with myelodysplastic syndromes. J Clin Oncol 2011; 29: 2499–2506. 20 Rocquain J, Carbuccia N, Trouplin V, Raynaud S, Murati A, Nezri M et al. Combined AUTHOR CONTRIBUTIONS mutations of ASXL1, CBL, FLT3, IDH1, IDH2, JAK2, KRAS, NPM1, NRAS, RUNX1, TET2 SS was the principal investigator of this study, analyzed the data and wrote the paper. and WT1 genes in myelodysplastic syndromes and acute myeloid leukemias. BMC CE and AF did sequence analysis of ASXL1. AK and VG performed next-generation Cancer 2010; 10: 401. sequencing. SJ contributed to writing of the paper. KAK, CS, PS, RP, and NS provided 21 Metzeler KH, Becker H, Maharry K, Radmacher MD, Kohlschmidt J, Mrozek K et al. patient samples and clinical data. CH was responsible for chromosome banding ASXL1 mutations identify a high-risk subgroup of older patients with primary analysis. WK was responsible for immunophenotyping and was involved in the cytogenetically normal AML within the ELN Favorable genetic category. Blood statistical analysis. TH was responsible for cytomorphologic analysis. TA collected and 2011; 118: 6920–6929. analyzed clinical data. All authors read and contributed to the final version of the 22 Bennett JM, Catovsky D, Daniel MT, Flandrin G, Galton DA, Gralnick HR et al. paper. Proposals for the classification of the acute leukaemias. French-American-British (FAB) co-operative group. Br J Haematol 1976; 33: 451–458. 23 Arber DA, Brunning RD, Le Beau MM, Falini B, Vardiman J, Porwit A et al. Acute myeloid leukemia with recurrent genetic abnormalities. In: Swerdlow SH, Campo REFERENCES E, Harris NL, Jaffe ES, Pileri SA, Stein H et al. (eds). WHO Classification of Tumours of 1 Falini B, Bolli N, Liso A, Martelli MP, Mannucci R, Pileri S et al. Altered nucleo- Haematopoietic and Lymphoid Tissues. Lyon: International Agency for Research on phosmin transport in acute myeloid leukaemia with mutated NPM1: molecular Cancer (IARC), 2008, pp 110–123. basis and clinical implications. Leukemia 2009; 23: 1731–1743. 24 Grimwade D, Hills RK, Moorman AV, Walker H, Chatters S, Goldstone AH et al. 2 Dohner H, Estey EH, Amadori S, Appelbaum FR, Buchner T, Burnett AK et al. Refinement of cytogenetic classification in acute myeloid leukemia: determina- Diagnosis and management of acute myeloid leukemia in adults: recommenda- tion of prognostic significance of rare recurring chromosomal abnormalities tions from an international expert panel, on behalf of the European LeukemiaNet. among 5876 younger adult patients treated in the United Kingdom Medical Blood 2010; 115: 453–474. Research Council trials. Blood 2010; 116: 354–365.

Leukemia (2013) 82 – 91 & 2013 Macmillan Publishers Limited ASXL1 mutations in AML S Schnittger et al 91 25 Illig T, Gieger C, Zhai G, Romisch-Margl W, Wang-Sattler R, Prehn C et al. A 38 Kern W, Bacher U, Haferlach C, Schnittger S, Haferlach T. The role of multi- genome-wide perspective of genetic variation in human metabolism. Nat Genet parameter flow cytometry for disease monitoring in AML. Best Pract Res Clin 2010; 42: 137–141. Haematol 2010; 23: 379–390. 26 Schnittger S, Schoch C, Dugas M, Kern W, Staib P, Wuchter C et al. Analysis of FLT3 39 Cheson BD, Bennett JM, Kopecky KJ, Buchner T, Willman CL, Estey EH et al. length mutations in 1003 patients with acute myeloid leukemia: correlation to Revised recommendations of the international working group for diagnosis, cytogenetics, FAB subtype, and prognosis in the AMLCG study and usefulness as a standardization of response criteria, treatment outcomes, and reporting stan- marker for the detection of minimal residual disease. Blood 2002; 100:59–66. dards for therapeutic trials in acute myeloid leukemia. J Clin Oncol 2003; 21: 27 Bacher U, Haferlach C, Kern W, Haferlach T, Schnittger S. Prognostic relevance of 4642–4649. FLT3-TKD mutations in AML: the combination matters--an analysis of 3082 40 Abdel-Wahab O, Kilpivaara O, Patel J, Busque L, Levine RL. The most commonly patients. Blood 2008; 111: 2527–2537. reported variant in ASXL1 (c.1934dupG;p.Gly646TrpfsX12) is not a somatic 28 Schnittger S, Schoch C, Kern W, Mecucci C, Tschulik C, Martelli MF et al. alteration. Leukemia 2010; 24: 1656–1657. Nucleophosmin gene mutations are predictors of favorable prognosis in 41 Kumar P, Henikoff S, Ng PC. Predicting the effects of coding non-synonymous acute myelogenous leukemia with a normal karyotype. Blood 2005; 106: variants on protein function using the SIFT algorithm. Nat Protoc 2009; 4: 3733–3739. 1073–1081. 29 Bacher U, Haferlach T, Schoch C, Kern W, Schnittger S. Implications of NRAS 42 Adzhubei IA, Schmidt S, Peshkin L, Ramensky VE, Gerasimova A, Bork P et al. A mutations in AML: a study of 2502 patients. Blood 2006; 107: 3847–3853. method and server for predicting damaging missense mutations. Nat Methods 30 Dicker F, Haferlach C, Kern W, Haferlach T, Schnittger S. Trisomy 13 is strongly 2010; 7: 248–249. associated with AML1/RUNX1 mutations and increased FLT3 expression in acute 43 Schnittger S, Kern W, Tschulik C, Weiss T, Dicker F, Falini B et al. Minimal residual myeloid leukemia. Blood 2007; 110: 1308–1316. disease levels assessed by NPM1 mutation-specific RQ-PCR provide important 31 Kohlmann A, Grossmann V, Klein HU, Schindela S, Weiss T, Kazak B et al. Next- prognostic information in AML. Blood 2009; 114: 2220–2231. generation sequencing technology reveals a characteristic pattern of molecular 44 Gelsi-Boyer V, Brecqueville M, Devillier R, Murati A, Mozziconacci MJ, Birnbaum D. mutations in 72.8% of chronic myelomonocytic leukemia by detecting Mutations in ASXL1 are associated with poor prognosis across the spectrum of frequent alterations in TET2, CBL, RAS, and RUNX1. J Clin Oncol 2010; 28: malignant myeloid diseases. J Hematol Oncol 2012; 5:12. 3858–3865. 45 Abdel-Wahab O, Pardanani A, Patel J, Wadleigh M, Lasho T, Heguy A et al. Con- 32 Grossmann V, Schnittger S, Schindela S, Klein HU, Eder C, Dugas M et al. Strategy comitant analysis of EZH2 and ASXL1 mutations in myelofibrosis, chronic mye- for robust detection of insertions, deletions, and point mutations in CEBPA, a GC- lomonocytic leukemia and blast-phase myeloproliferative neoplasms. Leukemia rich content gene, using 454 next-generation deep-sequencing technology. J Mol 2011; 25: 1200–1202. Diagn 2011; 13: 129–136. 46 Gilliland DG, Jordan CT, Felix CA. The molecular basis of leukemia. Hematology Am 33 Schnittger S, Kinkelin U, Schoch C, Heinecke A, Haase D, Haferlach T et al. Soc Hematol Educ Program 2004, 80–97. Screening for MLL tandem duplication in 387 unselected patients with AML 47 Link DC, Schuettpelz LG, Shen D, Wang J, Walter MJ, Kulkarni S et al. Identification identify a prognostically unfavorable subset of AML. Leukemia 2000; 14: of a novel TP53 cancer susceptibility mutation through whole-genome sequen- 796–804. cing of a patient with therapy-related AML. JAMA 2011; 305: 1568–1576. 34 Grossmann V, Kohlmann A, Zenger M, Schindela S, Eder C, Weissmann S et al. 48 IDH1 Cazzola M.. and IDH2 mutations in myeloid neoplasms--novel paradigms A deep-sequencing study of chronic myeloid leukemia patients in blast crisis and clinical implications. Haematologica 2010; 95: 1623–1627. (BC-CML) detects mutations in 76.9% of cases. Leukemia 2011; 25: 557–560. 49 Damm F, Bunke T, Thol F, Markus B, Wagner K, Gohring G et al. Prognostic 35 Haferlach T, Kern W, Schoch C, Hiddemann W, Sauerland MC. Morphologic dys- implications and molecular associations of NADH dehydrogenase subunit 4 (ND4) plasia in acute myeloid leukemia: importance of granulocytic dysplasia. J Clin mutations in acute myeloid leukemia. Leukemia 2012; 26: 289–295. Oncol 2003; 21: 3004–3005. 50 Delhommeau F, Dupont S, Della V,V, James C, Trannoy S, Masse A et al. Mutation 36 ISCN (1995). Guidelines for cancer cytogenetics, Supplement to: an International in TET2 in myeloid cancers. N Engl J Med 2009; 360: 2289–2301. System for Human Cytogenetic Nomenclature Mitelman F, Karger S (ed., 1995. 51 Ernst T, Chase AJ, Score J, Hidalgo-Curtis CE, Bryant C, Jones AV et al. Inactivating 37 Kern W, Voskova D, Schoch C, Hiddemann W, Schnittger S, Haferlach T. Deter- mutations of the histone methyltransferase gene EZH2 in myeloid disorders. Nat mination of relapse risk based on assessment of minimal residual disease during Genet 2010; 42: 722–726. complete remission by multiparameter flow cytometry in unselected patients 52 Ley TJ, Ding L, Walter MJ, McLellan MD, Lamprecht T, Larson DE et al. DNMT3A with acute myeloid leukemia. Blood 2004; 104: 3078–3085. mutations in acute myeloid leukemia. N Engl J Med 2010; 363: 2424–2433.

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

& 2013 Macmillan Publishers Limited Leukemia (2013) 82 – 91