Mutational architecture and clonal heterogeneity in myelofibrosis: Prognostic role of variant type and acquisition sequence

Abstract BACKGROUND: The spectrum of somatic mutations of myeloid-associated genes in patients (pts) with myelofibrosis (MF) has been extensively studied, however a definite correlation between specific mutational patterns, clinical phenotype and outcome, including blast phase (BP), is still uncertain. AIM and METHODS To such an end, we analyzed 756 fully-annotated MF pts, with diagnosis (dx) according to WHO/ICC 2022 criteria, followed in our Institution. NGS was performed at dx using a 40-gene panel. To infer the order of mutation acquisition, genes were classified into mutational functional categories (MFC), including epigenetic modifiers (TET2, EZH2, ASXL1, DNMT3A, IDH2; cat. 1), RAS pathway genes (NRAS, KRAS, CBL; cat.2) and spliceasome (SF3B1, U2AF1, ZRSR2, SRSF2; cat. 3). For each MFC, we selected pts who concurrently harbored a driver mutation (DM). We then compared variant allele frequency (VAF) of the mutations and categorized pts as “DM-first” (DM VAF > MFC VAF) or “MFC-first” (MFC VAF > DM VAF).RESULTS: Dx was pMF in 62% (50.9% pre-PMF,49.1% overt-PMF) and sMF in 38% (47.4% post-PV,52.6% post-ET). Median age at dx 59.4y (range 17.5-93.5). BP occurred in 62 pts (8.2%); 300 pts died (40%). Median overall survival (OS) was 6.2y (0.05–36y). Driver mutations were: JAK2 V617F in 72.3%, CALR 22.6% (68.2% type 1/like, 27.3% type 2/like, 4.5% atypical), MPLW515K/L 8.1%; 3.3% were triple-negative and 4.1% had 2 driver mutations. Among myeloid genes, most frequently mutated were: ASXL1 (28.9%), TET2 (19.6%) and EZH2 (8.1%). Significant co-occurrences were observed between IDH2 and SRSF2 (p<0.001), CBL and U2AF1 (p=0.039), and ASXL1 with EZH2 (p<0.001), U2AF1 (p<0.001), or RAS MFC (NRAS p=0.002; KRAS p=0.006; CBL p=0.03). Contrarily, CALR and SRSF2 mutations were mutually exclusive (p=0.005). Variants in TP53 (p=0.013), EZH2 (p<0.001), TET2 (p=0.025), SRSF2 (p<0.001), NRAS (p=0.013), KRAS (p=0.007), CBL (p=0.001), and ASXL1 (p<0.001) resulted significantly associated with shorter OS. These were confirmed in multivariable analysis for TET2 (p=0.012; HR 1.4), SRSF2 (p<0.001; HR 2.4), ASXL1 (p=0.01; HR 1.4) and EZH2 (p=0.003; HR 1.8). The mere occurrence of mutations at dx in IDH1 (p=0.045), IDH2 (p=0.005), ASXL1 (p=0.003), SRSF2 (p<0.001), CBL (p=0.006), NRAS (p<0.001), RUNX1 (p<0.001), and SETBP1 (p<0.001) were associated with increased risk of BP. However, univariate analysis showed SRSF2 (p=0.008) and RUNX1 (p=0.016) to be associated with shorter time to BP, maintained in multivariable analysis only for SRSF2 (p=0.013; HR 3.73). Variant consequences analysis (missense, frameshift; fs, stop gain; sg and splicing) revealed prognostic variability depending on the mutation type. EZH2 sg and missense mutations were associated with shorter OS (p=0.026 sg; p=0.001 missense) unlike the fs ones. Conversely, harbouring a fs ASXL1MUT was strongly associated with an adverse OS (p<0.001) unlike missense and sg variants (p=0.42 and p=0.38 respectively). TET2 sg variants were associated with worse OS (p=0.009). As expected, CALR fs mutations were protective (HR=0.78 p = 0.002).Interaction analysis between variant consequence and BP revealed the latter to be enriched in SRSF2(p=0.01) and RUNX1(p=0.03) missense variants. The acquisition order of mutated genes was analyzed in a total of 416 patients (55%). MFC-first status was associated with shorter OS only for epigenetic modifiers (p=0.003). 218/344 pts with epigenetic mutations (63.4%) were classified as DM-first and 126 (36.6%) MFC first. Specifically, acquisition of ASXL1 or TET2 first conferred reduced OS (p=0.02 and p=0.03 respectively). Conversely, no significant association was found between mutation order and BP. Epigenetic-first patients were more likely to present with pre-PMF (p=0.008; OR 2.00) and at the same time, fibrosis was less frequent in these patients (p=0.04; OR 0.59), suggesting that early epigenetic alterations may influence disease phenotype, On the other hand, spliceosome-first was mainly associated with splenomegaly (p=0.02 OR 2.55). CONCLUSIONS This large MF cohort highlights the prognostic importance of variant effects and mutation timing. Functional annotation, especially of early mutations in epigenetic regulators like ASXL1 and TET2, enhances risk stratification. VAF-based clonal architecture analysis supports clinical application, with further studies on clonal evolution ongoing.