Clinical Outcomes and Co-Occurring Mutations in Patients with RUNX1-Mutated Acute Myeloid Leukemia

(1) Runt-related transcription factor 1 (RUNX1) mutations in acute myeloid leukemia (AML) are often associated with worse prognosis. We assessed co-occurring mutations, response to therapy, and clinical outcomes in patients with and without mutant RUNX1 (mRUNX1); (2) We analyzed 328 AML patients, including 177 patients younger than 65 years who received intensive chemotherapy and 151 patients >65 years who received hypomethylating agents. RUNX1 and co-existing mutations were identified using next-generation sequencing; (3) RUNX1 mutations were identified in 5.1% of younger patients and 15.9% of older patients, and were significantly associated with increasing age (p = 0.01) as well as intermediate-risk cytogenetics including normal karyotype (p = 0.02) in the elderly cohort, and with lower lactate dehydrogenase (LDH; p = 0.02) and higher platelet count (p = 0.012) overall. Identified co-occurring mutations were primarily ASXL1 mutations in older patients and RAS mutations in younger patients; FLT3-ITD and IDH1/2 co-mutations were also frequent. Younger mRUNX1 AML patients treated with intensive chemotherapy experienced inferior treatment outcomes. In older patients with AML treated with hypomethylating agent (HMA) therapy, response and survival was independent of RUNX1 status. Older mRUNX1 patients with prior myelodysplastic syndrome or myeloproliferative neoplasms (MDS/MPN) had particularly dismal outcome. Future studies should focus on the prognostic implications of RUNX1 mutations relative to other co-occurring mutations, and the potential role of hypomethylating agents for this molecularly-defined group.


Introduction
The genetic diversity of acute myeloid leukemia (AML) and the complexity of its multi-step pathogenesis now allow the classification of AML based on molecular events. The evolution of AML has traditionally been proposed to follow the "two-hit model" [1], in which two classes of mutations are required for cancer development. Class I mutations are activating mutations that stimulate cell survival and proliferation, while class II mutations are inactivating mutations that interfere with hematopoietic Table 1 summarizes the clinicopathologic variables evaluated with respect to the impact of RUNX1 mutational status. In the elderly cohort, a significant correlation was observed between the presence of mutated RUNX1 and increasing age (p = 0.01); mRUNX1 was rare in the younger patient cohort (5%). Along with an increasing age, RUNX1 mutations occurred more frequently in older patients with intermediate-risk cytogenetics (particularly those with a normal karyotype), as compared to those with complex cytogenetic abnormalities (p = 0.02). Lower lactate dehydrogenase (LDH) levels and higher platelet counts were also found to be significant in RUNX1-mutated patients within the entire cohort overall (p = 0.02 and p = 0.012 respectively). RUNX1 mutational status did not significantly correlate with WBC count, hemoglobin, or peripheral blood and bone marrow blasts, or history of prior myelodysplastic syndrome or myeloproliferative neoplasms (MDS/MPN) or therapy-related AML (t-AML). Among the mutant RUNX1 patients, 15.2% (n = 20) of the patients classified as FAB M0 subtype.
A number of other gene mutations were analyzed and showed varying association with mutated RUNX1 (Figure 1) [15]. The frequencies of these co-mutations with mutant RUNX1 in the younger and the elderly cohorts were, respectively, as follows: FLT3-ITD, 33  We additionally identified a number of mutations that were negatively associated with mutant RUNX1. Mutations in the NPM1 gene were significantly lower in both the older (4.2%) and younger cohorts (0%) with RUNX1 mutations, compared to those with wild-type RUNX1 (older, 15.0%; younger, 26.2%). A similar negative correlation was identified between RUNX1 and CEBPα mutations-particularly among the younger cohort, wherein CEBPα incidence with wild-type RUNX1 was 16.1%, and was not identified (0%) in the setting of mutated RUNX1.
FLT3-ITD mutations were frequently identified in both cohorts, irrespective of mRUNX1: frequencies in the presence of mutated RUNX1 versus wild-type RUNX1 (older, 16.7% vs. 15.7%; younger, 30.0% vs. 20.8%). Of interest, co-occurring FLT3-ITD in the setting of mRUNX1 appeared to confer a more favorable prognosis (median OS with mutated FLT3-ITD vs. wild-type FLT3-ITD, NR vs. 11.5 months; p = 0.034) with a notable improved OS in FLT3-ITD and RUNX1 dual-mutated patients (Figure 2), although these differences were not significant when separated by age (younger, We additionally identified a number of mutations that were negatively associated with mutant RUNX1. Mutations in the NPM1 gene were significantly lower in both the older (4.2%) and younger cohorts (0%) with RUNX1 mutations, compared to those with wild-type RUNX1 (older, 15.0%; younger, 26.2%). A similar negative correlation was identified between RUNX1 and CEBPα mutations-particularly among the younger cohort, wherein CEBPα incidence with wild-type RUNX1 was 16.1%, and was not identified (0%) in the setting of mutated RUNX1.

Association of RUNX1 with Response to Frontline AML Therapy
In the elderly AML cohort, response to frontline therapy was generally unaffected by RUNX1 mutational status (Table 2). There were no significant differences between patients with mutated and wild-type RUNX1 for complete remission (CR) rates, overall response rate (ORR);(CR + complete remission with incomplete platelet recovery (CRp) + hematological improvement (HI) + partial remission (PR), HI, and/or early death (ED; death within 28 days of initiating treatment) in the cohort as a whole. Notably, a significant difference was observed for the subset of older AML patients with prior MDS/MPN (Table 3). Overall, ORR was obtained in 80% of MDS/MPN patients in the older cohort with wild-type RUNX1, but was achieved in only 25% of the patients identified with mutant RUNX1 (p = 0.004).

Association of RUNX1 with Response to Frontline AML Therapy
In the elderly AML cohort, response to frontline therapy was generally unaffected by RUNX1 mutational status (Table 2). There were no significant differences between patients with mutated and wild-type RUNX1 for complete remission (CR) rates, overall response rate (ORR);(CR + complete remission with incomplete platelet recovery (CRp) + hematological improvement (HI) + partial remission (PR), HI, and/or early death (ED; death within 28 days of initiating treatment) in the cohort as a whole. Notably, a significant difference was observed for the subset of older AML patients with prior MDS/MPN (Table 3). Overall, ORR was obtained in 80% of MDS/MPN patients in the older cohort with wild-type RUNX1, but was achieved in only 25% of the patients identified with mutant RUNX1 (p = 0.004).   In the younger cohort (Table 3), while CR rates were significantly reduced in patients with mutated versus wild-type RUNX1 (33.3% vs. 76.2%; p = 0.004), no significant difference in overall response rate was observed.

Discussion
In our study, RUNX1 mutations were detected at an overall incidence of 10.1%. This is consistent with previous studies that reported mutated RUNX1 alleles in 5.6% to 16.1% of patients with AML [3,4,7,10,12].
Mutations in RUNX1 associate with increasing age, also consistent with previous studies examining RUNX1 mutation status and age [4,7,12,16]. Accumulation of cellular stress and impaired repair of double-stranded DNA breaks (e.g., after radiation exposure) may contribute to genomic instability [17] and increase the frequency of genetic mutations, including RUNX1 [18][19][20]. Furthermore, RUNX1 mutations in our elderly cohort clustered among patients identified with intermediate cytogenetic risk including cytogenetically-normal AML, consistent with previous studies [7,10]. Other clinical characteristics (white blood cell (WBC) count, hemoglobin, peripheral blood and bone marrow blasts, t-AML, and AML secondary to MDS/MPN) were not associated with RUNX1 mutational status. Our analysis additionally confirmed a relationship between RUNX1 mutation and platelet count and LDH level, as has been previously reported [4,7].
The most commonly co-occurring mutation overall and particularly prevalent in the older cohort was ASXL1. Of interest, clonal expansion of ASXL1 is one of the most frequently identified genes in age-related clonal hematopoiesis, and subsequent RUNX1 mutations may be implicated in malignant progression in these individuals [21,22]. Among younger patients, co-occurring RAS mutations (44%) and FLT3-ITD (33%) were also frequent. Their classification as class I mutations represents their capacity to promote stem cell proliferation and thereby cooperate with RUNX1 mutations in AML pathogenesis [2,23].
Co-occurring FLT3 and RUNX1 mutations have been previously described, with a hypothesis that loss of function RUNX1 mutations may predispose hematopoietic cells to overexpress activating mutations of FLT3 [24]. When examining the clinical implications of this hypothesis in patients with mutant RUNX1, our analysis interestingly found superior OS and EFS in mRUNX1 associated with FLT3 mutations. Previous studies have identified that the adverse effect of RUNX1 is independent of FLT3-ITD [25]. Evaluation in future studies will be important to see if this is a reproducible finding. It is also important to note that NPM1 and biallelic CEBPα mutations confer a favorable prognosis in terms of disease-free survival and OS in the absence of FLT3 mutations [2], and both are inversely related to the incidence of RUNX1 mutations [7]. In our cohort, the incidence of NPM1 mutations declined from 26.2% to 0% in younger patients and from 15.0% to 4.2% in older patients in the absence and presence of RUNX1 mutations, respectively. Conversely, although partial tandem duplications

Discussion
In our study, RUNX1 mutations were detected at an overall incidence of 10.1%. This is consistent with previous studies that reported mutated RUNX1 alleles in 5.6% to 16.1% of patients with AML [3,4,7,10,12].
Mutations in RUNX1 associate with increasing age, also consistent with previous studies examining RUNX1 mutation status and age [4,7,12,16]. Accumulation of cellular stress and impaired repair of double-stranded DNA breaks (e.g., after radiation exposure) may contribute to genomic instability [17] and increase the frequency of genetic mutations, including RUNX1 [18][19][20]. Furthermore, RUNX1 mutations in our elderly cohort clustered among patients identified with intermediate cytogenetic risk including cytogenetically-normal AML, consistent with previous studies [7,10]. Other clinical characteristics (white blood cell (WBC) count, hemoglobin, peripheral blood and bone marrow blasts, t-AML, and AML secondary to MDS/MPN) were not associated with RUNX1 mutational status. Our analysis additionally confirmed a relationship between RUNX1 mutation and platelet count and LDH level, as has been previously reported [4,7].
The most commonly co-occurring mutation overall and particularly prevalent in the older cohort was ASXL1. Of interest, clonal expansion of ASXL1 is one of the most frequently identified genes in age-related clonal hematopoiesis, and subsequent RUNX1 mutations may be implicated in malignant progression in these individuals [21,22]. Among younger patients, co-occurring RAS mutations (44%) and FLT3-ITD (33%) were also frequent. Their classification as class I mutations represents their capacity to promote stem cell proliferation and thereby cooperate with RUNX1 mutations in AML pathogenesis [2,23].
Co-occurring FLT3 and RUNX1 mutations have been previously described, with a hypothesis that loss of function RUNX1 mutations may predispose hematopoietic cells to overexpress activating mutations of FLT3 [24]. When examining the clinical implications of this hypothesis in patients with mutant RUNX1, our analysis interestingly found superior OS and EFS in mRUNX1 associated with FLT3 mutations. Previous studies have identified that the adverse effect of RUNX1 is independent of FLT3-ITD [25]. Evaluation in future studies will be important to see if this is a reproducible finding. It is also important to note that NPM1 and biallelic CEBPα mutations confer a favorable prognosis in terms of disease-free survival and OS in the absence of FLT3 mutations [2], and both are inversely related to the incidence of RUNX1 mutations [7]. In our cohort, the incidence of NPM1 mutations declined from 26.2% to 0% in younger patients and from 15.0% to 4.2% in older patients in the absence and presence of RUNX1 mutations, respectively. Conversely, although partial tandem duplications in the MLL gene have been previously associated with mutated RUNX1 and implicated in shorter survival [2,7], this association was not identified in our patient cohort, as only four patients were identified with the MLL rearrangement 11q23. An association of co-occurring IDH1/IDH2 mutations with RUNX1 mutations was additionally identified, consistent with Gaidzik et al. [10]. Furthermore, mutations in RUNX1 and IDH1/2 have been collectively described as the most frequent genetic lesions in M0-AML [26], the prognostic impact of the latter remaining controversial [27].
While RUNX1 mutations are frequently associated with inferior responses to AML therapy, our analysis suggests that the poor outcomes seen in RUNX1 mutations occur primarily within younger patients treated with intensive therapeutic strategies, as well as the subgroup of RUNX1 mutant AML with prior MDS/MPN. The similar responses and outcome of the larger cohort of older mRUNX1 treated with HMA approaches was a notable finding of this analysis.
A previous study in patients with MDS identified that treatment with HMAs may partially mitigate the adverse prognostic impact of gene mutations in patients with RUNX1, ASXL1, EZH2, or ETV6 mutations [28]. Differences in outcomes between patient populations may also be linked with the distribution of co-existing mutations and cytogenetic abnormalities that may exacerbate or mitigate the effect of mutated RUNX1, [2] and our study is limited by a lack of complete co-mutational data, most notably for splicing factor mutations, as whole exome sequencing was not possible. The importance of RUNX1 as a marker of MRD in responding patients was additionally not able to be assessed in our study.

Patients
A total of 328 patients with treatment-naïve AML treated at the University of Texas MD Anderson Cancer Center between August 2013 and July 2016 were analyzed in this study. The presence of de novo AML, or history of prior myelodysplastic syndrome (MDS) or myeloproliferative neoplasms (MPNs) was collected. As treatment intensity at our institution during this time period was strongly and near-universally correlated to patient age at diagnosis, patients were categorized into two cohorts according to their age and therapeutic regimen: patients younger than 65 years who received intensive chemotherapy (younger cohort; n = 177) and patients ≥65 years who received HMA therapy (elderly cohort; n = 151).
Variables in our study included patient demographics (age and sex), relevant diagnostic tests including white blood cell (WBC) count, hemoglobin (Hb) levels, platelet count, lactate dehydrogenase (LDH) levels, and percentage of bone marrow blasts, cytogenetics, FAB classification, RUNX1 mutation, and co-occurrence of other mutations. Clinical outcomes were assessed using Kaplan-Meier estimates of OS, EFS, and RFS [29]. Response to therapy was evaluated using revised IWG criteria for AML [30].
All participating patients provided informed consent in compliance with the Declaration of Helsinki, and the study received ethical approval by the Institutional Review Board as part of a retrospective chart review protocol, PA11.0788 on 10 October 2011 within the Department of Leukemia at University of Texas MD Anderson Cancer Center.
The lower limit of detection of this assay (analytical sensitivity) for single nucleotide variations was determined to be 5% (one mutant allele in the background of nineteen wild type alleles) to 10% (one mutant allele in the background of nine wild type alleles). Testing of patients with active hematologic malignancies was limited to somatic mutations only. Known germline polymorphisms identified in >20% of our in-house patient population were excluded. Additionally, germline polymorphisms previously reported in population databases such as dbSNP and ExAC were classified as "variants of probable germline origin" and were excluded from this analysis.

Statistical Analysis
Continuous variables were described as means, medians, standard deviations, and ranges, whereas the categorical variables were presented as frequencies and percentages. The categorical variables in patients with mutated RUNX1 were compared with the wild-type using the Chi-squared test. Kaplan-Meier curves [26] were used to estimate unadjusted OS durations. OS was calculated from the start of treatment to the last follow up or death (censor = alive/dead). EFS was measured from the start of treatment to any adverse event or the last follow-up. Adverse events included either relapse (i.e., NR, ED) or death. RFS was determined from leukemia-free response (i.e., at the time of documented CR, CRp, CRi, PR) to relapse (loss), death, or last follow-up. For all analyses, p < 0.05 was considered statistically significant. All computations were carried out in Statistica 12.0 (StatSoftInc. Tulsa, OK, USA; 1984-2013).

Conclusions
In conclusion, RUNX1 mutations are among the most frequent recurring genetic abnormalities in patients with AML, often associated with specific clinical features and poor outcomes. ASXL1, DNMT3A, IDH1/2, FLT3-ITD, and RAS mutations frequently co-occur with RUNX1 mutations. FLT3-ITD mutations did not impart an inferior prognosis in mRUNX1 AML, and of note appeared to confer an improved OS in this molecularly defined cohort. While younger mRUNX1 patients treated with intensive chemotherapy and a subgroup of older mRUNX1 patients with prior MDS/MPN exhibited inferior treatment responses, the majority of mRUNX1 AML patients included older patients treated with HMA therapy, in whom treatment responses and clinical outcomes were not inferior compared to RUNX1 wild-type. Further studies investigating the treatment response to HMAs-in particular in younger patient populations with RUNX1-mutated AML-are required to fully establish their therapeutic role in the individualized treatment of these patients.