AML with RUNX1 Mutations and Loss of RUNX1 Wild-Type - a Distinct Subset?

Background: Mutations in RUNX1 have been reported in 5 to 20% of AML. RUNX1 mutated AML is associated with a myeloid rather than monocytic differentiation, shows a typical pattern of cytogenetic abnormalities with a high frequency of trisomy 8 or 13, has a typical pattern of additional molecular mutations with a high frequency of accompanying ASXL1 and SF3B1 mutations and is nearly mutually exclusive of NPM1 and CEBPA double mutations and other entity-defining genetic abnormalities. In a subset of patients with RUNX1 mutations loss of the wild-type allele can be assumed due to a high mutation load. The aim of this study was the detailed analysis of a subset of RUNX1 mutated AML with RUNX1 wild-type loss with respect to accompanying cytogenetic and molecular genetic abnormalities and prognostic impact.

Patients and Methods: A cohort of 467 AML with RUNX1 mutations (mut) at diagnosis identified during diagnostic work-up in our laboratory were the basis of this study. Median age was 72 years (yrs) (range 18-91 yrs), and male:female ratio 296: 171. 366 patients had de novo AML, 77 s-AML following MDS, 24 t-AML. For all patients (pts) cytogenetics and for 341 data on FAB subtype was available. Mutation data was available for NPM1 (n=456), MLL-PTD (n=453), CEBPA (n=449), FLT3-ITD (n=457), FLT3-TKD (n=457), WT1 (n=398), ASXL1 (n=313), TP53 (n=231), DNMT3A (n=177), TET2 (n=174), NRAS (n=305), KRAS (n=213) and SF3B1 (n=119). 64 patients with a mutation load of RUNX1 mutation >70% evaluated by sequencing analysis were selected for further analysis. All 64 cases were analysed by genomic arrays (SurePrint G3 ISCA CGH+SNP Microarray, Agilent, Waldbronn, Germany) to determine the copy number state and copy neutral loss of heterozygosity (CN-LOH). Median age was 73 yrs (range 24-87 yrs), and male:female ratio was 27: 37. 50 patients had de novo AML, 11 s-AML following MDS, 3 t-AML.

Results: Array CGH revealed a cytogenetically cryptic deletion on the long arm of chromosome 21 encompassing the RUNX1 gene in 5/64 (8%) patients while a CN-LOH on 21q including the RUNX1 gene was observed in 45 cases (70%). Thus in 50 cases (78%) with a high RUNX1 mutation load a RUNX1 wild-type loss (wt-loss) was detected by array CGH. In 43% (6/14) of the remaining cases the high RUNX1 mutation load was caused by amplification of the long arm of chromosome 21 either due to gain of whole chromosomes 21 or to an isochromosome 21q. First we focused on the characterization of RUNX1 mutated cases with RUNX1 wt-loss. In 22/50 cases (44%) an aberrant karyotype was observed with a distinct aberration pattern. 11 cases harbored +13, 5 had +8 and 6 cases a loss of 7q. No other recurrent abnormalities were observed. With respect to concurrent mutations the following frequencies were found: ASXL1 (42%), FLT3 -ITD (34%), TET2 (21%), KRAS (11%), MLL-PTD (8%), NRAS (7%), and FLT3-TKD (6%). No NPM1 mutation or CEBPA double mutations were identified. Comparison of those cases with RUNX1 wt-loss to all other RUNX1 mutated AML (n=417) revealed a significantly higher frequency of +13 (22% vs 9%, p=0.01) and FLT3 -ITD (34% vs 19%, p=0.015). FAB subtypes M0 and M1 were more frequent (46% vs 12%, p<0.001; 35% vs 22%, n.s.) and M2 and M4 less frequent (14% vs 46%, p<0.0001; 5% vs 17%, n.s.). Survival analyses were restricted to 212 de novo AML pts with RUNX1 mut who received intensive chemotherapy (median overall survival (OS): 20 months (mo), median event-free survival (EFS): 12 mo). Median OS and EFS was shorter in patients with RUNX1 wt-loss compared to those without (15 vs 20 mo, n.s., 10 vs 12 mo, p=0.04). In univariate Cox regression analysis a negative impact on OS was observed for RAS mut (relative risk (RR): 2.2, p=0.005), male gender (RR: 1.6, p=0.02), and age (RR: 1.3 per decade, p<0.001). Shorter EFS was associated with RUNX1 wt-loss (RR: 1.7, p=0.04), RAS mut (RR: 1.9, p=0.02) and age (RR: 1.2 per decade, p<0.001). In multivariate analysis RAS mut (OS: RR: 2.4, p=0.002; EFS: RR: 2.0, p=0.008) and age (OS: RR: 1.3 per decade, p<0.001; EFS: RR: 1.2 per decade, p<0.001) had independent prognostic impact.

Conclusions:RUNX1 mutated AML with wild-type loss is a distinct AML subset that does not overlap with any of the genetically defined WHO categories and is characterized by an immature phenotype (81% FAB Subtype M0 and M1) and a higher frequency of +13 and FLT3-ITD as compared to RUNX1 mutated AML without wild-type loss. Wild-type loss and RAS mutations are associated with inferior outcome in RUNX1 mutated AML.