Clinical and Molecular Characteristics of RAS and RAS-like Mutational Signature in Acute Myeloid Leukemia

Background: NRAS and KRAS activating mutations are present in ~15-20% of adult AML cases and are associated with poor clinical outcomes. Although the independent prognostic impact of it remains controversial, survival patterns may be more heterogeneous on a granular scale among patient clusters with similar clinical/molecular profiles. We sought to externally validate this by analyzing the clinical outcomes and the mutational/transcriptomic landscape of RAS-mutated AML patients. Moreover, given the sub-clonal nature of RAS mutations and many downstream shared oncogenic pathways among different driver mutations in AML, we explored the gene expression data for a RAS-like signature which may hold therapeutic implications using the novel RAS inhibitors.

Methods: We analyzed a cohort of NRAS^MT and KRAS^MT AML patients treated at Karmanos Cancer Center from 2016-23, data from cBioPortal (Cerami et al., 2012), AACR GENIE (v 16.0), and a cohort of various sub-studies (Awada et al., Blood 2021, Kewan et al., Nat. Com., 2023). We analyzed the molecular data in the context of overall survival (OS), clinical parameters, coexisting mutations, karyotype, AML subtypes, and RNA-seq expression data. Chi-square and t-tests for univariate statistics, and Kaplan-Meier and log-rank methods for OS were used.

Results: From a total of 18,200 adult AML cases, 4,132 eligible patients were separated into 4 cohorts: 1433 (35%) with NRAS^MT/KRAS^WT, 781 (19%) with NRAS^WT/KRAS^MT, 128 (3%) with NRAS^MT/ KRAS^MT, and 1790 (43%) with NRAS^WT/ KRAS^WT. The median age of the double-mutated cohort was 67 ys. [25.2-85], compared to the NRAS^MT and KRAS^MT cohorts at 63.6 and 65.5 ys. [23-85 ys. p=0.01; vs 27-85 ys. p=0.04 resp.]. Females were more represented in the co-mutated cohort than NRAS^MT and KRAS^MT AML [61% vs 25% and 43%; p =<0.001, <0.001]. Abnormal karyotype enriched the NRAS^MT cohort more than the KRAS^MT cohort [54% vs 27%, p =<0.001], whereas normal karyotype was more prominent in the double mutated cohort [45% vs 28% and 31%; p =<0.001 and 0.0243 resp.]. FLT3-ITD^NEG status dominated all cohorts; KRAS^MT and the double-mutated cohorts had similar FLT3-ITD^NEG mutants [58% vs 55%, p =0.59], and NRAS^MT AML had the highest FLT3-ITD^POS hits (14%). All cohorts were profoundly hyperproliferative in bone marrow blast response (47% vs 46.5% vs 48.5%). The degree of pancytopenia was comparable in NRAS^MT, KRAS^MT and double mutant patients.

The double mutated cohort had the highest OS compared to NRAS^MT, and KRAS^MT patients [31.6 mo. vs 21.2 mo. vs 19.2 mo., p=<0.001 and 0.02 resp.]. Most of the NRAS^MT patients had G12 S/A/C (42%) or G13 D/R/V/C (24%) canonical hits, whereas most of the KRAS^MT patients had G12 D/V/A (53%) or G13 D/R/C (25%) mutations. In terms of co-mutational profile, the NRAS^MT cohort was uniquely enriched in NPM1 (50%) and DNMT3A (41%), whereas the KRAS^MT cohort was enriched in worse prognosticating myelodysplasia related gene mutations such as ASXL1 (14%), U2AF1 (10%) and STAG2 (8%).

To define a RAS-like signature, differentially expressed genes (DEGs) from KRAS^MT, NRAS^MT, and double-mutated patients were collated. The expression levels were normalized, and the top 20 genes (out of 35 total) with expression z-scores > 2.0 (termed high expressors) were selected. We examined the presence of high expressors in KRAS^WT and NRAS^WT patient profiles using Elastic Net regularization and found that a RAS-like gene signature based on DNMT3A, CALR, CNPY3, PTPRC, and PTPN11 mutations was highly correlated with gene expression similar to KRAS^MT and NRAS^MT patients (Spearman correlation coefficient > 0.85).

Conclusion: While NRAS^MT and KRAS^MT may not independently affect prognosis, the patterns of cooperativity with other driver mutations may causally influence disease progression and treatment of myeloid malignancies. Double mutant RAS conveying a better prognosis needs further investigation possibly supporting using a pan RAS inhibitor for translational applications. The presence of a RAS-like signature in wildtype patients further highlights more shared components based on the leukemic disease biology in the RAS gene family and beyond. Ongoing work is focused on using Weighted Gene Co-expression Network Analysis (WGCNA) to rigorously qualify DEGs, dissect the shared transcriptional programming for further preclinical evaluation and identify other therapeutic vulnerabilities in the same pathway.