AUTHORS
Giorgio Mattiuz, PhD, Valeria Visconte, PhD2, Gloria Andreossi3, Elettra Celoni4, Carmelo Gurnari, MD,PhD5, Giulia Maggioni, MD6, Marco Gabriele Gabriele Raddi, MD7, Angela Consagra, MD8, Matteo Giovanni Della Porta, MD9, Jaroslaw P. Maciejewski, MD, PhD, FACP10, Valeria Santini, MD11
AFFILIATIONS
1 Università degli Studi di Firenze, Firenze, Italy
2 Taussig Cancer Center, Cleveland Clinic, Cleveland, Ohio, United States
3 Università degli Studi di Firenze, Firenze, Italy
4 Università degli Studi di Firenze, Firenze, Italy
5 Taussig Cancer Center, Cleveland Clinic, Cleveland, Ohio, United States, Tor Vergata University of Rome, Rome, Italy
6 IRCCS Humanitas Research Hospital, Rozzano, Milan, Italy
7 Università degli Studi di Firenze, Firenze, Italy
8 Università degli Studi di Firenze, Firenze, Italy
9 IRCCS Humanitas Research Hospital, Rozzano, Milan, Italy
10 Taussig Cancer Center, Cleveland Clinic, Cleveland, Ohio, United States
11 Università degli Studi di Firenze, Firenze, Italy
ABSTRACT
Introduction: The Molecular International Prognostic Scoring System (IPSS-M) has significantly improved prognostic stratification in MDS; however, it encodes most genes using a binary mutated/wild-type status, thereby agnostic of specific characteristics of the mutations (i.e. the altered function, protein domain involvement, different hotspots, clonal burden as per variant allele frequency (VAF)). Given the marked clinical and molecular heterogeneity of MDS, one approach will be to incorporate the above genomic features to further refine the current IPSS-M and improve risk-adapted decision-making. Thus, we focus on 3 most mutated genes (ASXL1, SF3B1, TET2) in MDS as proof of principle.
Methods: We analyzed a pooled cohort of 3,008 patients (pts) with MDS from The Cleveland Clinic cohort (n=258), GenoMed4All/Humanitas (n=1010), and IPSS-M development cohort (n=1740). Pts were classified according to the 2016 World Health Organization criteria (bone marrow blasts <20%; white blood cell count <13 × 10?/L). Median age was 71 years (63.0–78.0), 64.5% of patients were male, the median number of mutations was 2 (1-4) per patient; median follow-up was 3.9 (3.7-3.9) years. IPSS-R and IPSS-M scores were available. Mutations were annotated according to functional type (gain-of-function [GOF] or loss-of-function [LOF]), occurrence in protein domain, recurrent hotspots, and VAF. Truncating ASXL1 hotspots were annotated as GOF lesions, whereas TET2 mutations were classified according to involvement of the catalytic double-stranded β-helix (DSBH) domain. VAF cutoff was set per gene as the median VAF of its mutations. OS and leukemia-free survival (LFS) were assessed by Kaplan-Meier and log-rank tests within merged lower-risk/higher-risk (LR/HR) and single IPSS-M categories. Comparisons were restricted to groups including ≥15 patients, and p values were adjusted using the Benjamini-Hochberg procedure. Multivariable Cox models adjusted for the continuous IPSS-M score assessed independent prognostic effects. Co-mutation patterns were evaluated using Fisher’s exact test.
Results: ASXL1 mutations were present in 26.7% pts. Among HR-MDS patients, G646fs*12 frameshift hotspot was associated with shorter OS (median 1.5 vs 2.0 years; p=0.0022) and LFS (1.2 vs 2.0 years; p<0.001; q=0.017) compared with other ASXL1 mutations. Its prognostic effect was independent of the IPSS-M score (OS: HR, 1.37; p=0.005; LFS: HR, 1.49; p<0.001). G646fs*12 was enriched for co-mutations in RUNX1, STAG2, and TP53, and was mutually exclusive with SF3B1 mutations. In LR-MDS, ASXL1 mutation was prognostically negative for OS (median 5.1 vs 7.2 years; p=0.0032) and LFS (5.7 vs 8.9 years; p=0.0101) compared to the wild-type group.
SF3B1 mutations were detected in 28.1% pts. SF3B1 mutation status was associated with favorable outcomes even in HR-MDS (median OS 2.7 vs 1.7 years, p=0.0041; LFS 2.7 vs 1.6 years, p=0.0041) but did not provide additional prognostic information beyond the IPSS-M score (OS: HR, 0.96; p=0.54). High SF3B1 VAF was associated with longer survival both in LR-MDS (median OS 7.7 vs 6.8 years, p=0.0072; LFS 8.4 vs 7.7 years, p=0.0283) and within the individual IPSS-M Low category (OS 7.7 vs 5.5 years, p=0.0063). In cases with High IPSS-M score, K700E was associated with better outcome respect to other LOF lesions (median OS 2.5 vs 1.5 years, p=0.032; LFS 3.0 vs 1.4 years, p=0.009). LOF mutations (OS: HR, 1.60; p=0.015) showed IPSS-M score-independent prognostic effects.
TET2 mutations were detected in 27.2% of cases. Although TET2 mutations status did not refine IPSS-M score prognostic value (OS: HR, 0.98; p=0.73) a composite classification combining mutations in the DSBH-domain and VAF succeeded in stratifying HR-MDS patients for OS (p=0.0236) and LFS (p=0.0086). Conversely, low-VAF lesions outside the DSBH domain showed the most favorable outcomes (median OS 2.6 vs 1.4 years and LFS 2.8 vs 1.4 years for low- vs high-VAF out-of-domain lesions).
Conclusions: Mutation features refine IPSS-M prognostic assessment differently according to mutated gene and risk score category. ASXL1 G646fs*12 conferred worst prognosis for HR-MDS patients. SF3B1 mutation clonal burden and lesion type provided additional prognostic information on the top of the ones derived from the mere mutation status. Combined domain and VAF annotation revealed prognostic heterogeneity among TET2 mutations and impact on outcomes based on mutation location and VAF.
DISCLOSURES
Giorgio Mattiuz, PhD: No relevant conflicts of interest to declare.