Genomic analysis of primary and secondary myelofibrosis redefines the prognostic impact of ASXL1 mutations: a FIM study

Abstract

We aimed to study the prognostic impact of the mutational landscape in primary and secondary myelofibrosis. The study included 479 patients with myelofibrosis recruited from 24 French Intergroup of Myeloproliferative Neoplasms (FIM) centers. The molecular landscape was studied by high-throughput sequencing of 77 genes. A Bayesian network allowed the identification of genomic groups whose prognostic impact was studied in a multistate model considering transitions from the 3 conditions: myelofibrosis, acute leukemia, and death. Results were validated using an independent, previously published cohort (n = 276). Four genomic groups were identified: patients with TP53 mutation; patients with ≥1 mutation in EZH2, CBL, U2AF1, SRSF2, IDH1, IDH2, NRAS, or KRAS (high-risk group); patients with ASXL1-only mutation (ie, no associated mutation in TP53 or high-risk genes); and other patients. A multistate model found that both TP53 and high-risk groups were associated with leukemic transformation (hazard ratios [HRs] [95% confidence interval], 8.68 [3.32-22.73] and 3.24 [1.58-6.64], respectively) and death from myelofibrosis (HRs, 3.03 [1.66-5.56] and 1.77 [1.18-2.67], respectively). ASXL1-only mutations had no prognostic value that was confirmed in the validation cohort. However, ASXL1 mutations conferred a worse prognosis when associated with a mutation in TP53 or high-risk genes. This study provides a new definition of adverse mutations in myelofibrosis with the addition of TP53, CBL, NRAS, KRAS, and U2AF1 to previously described genes. Furthermore, our results argue that ASXL1 mutations alone cannot be considered detrimental.

Graphical abstract

Figure 1.

Mutational landscape of the entire cohort. Proportion of patients with ≥1 mutation per gene according to the classification of mutations (A), the type of myelofibrosis (B), and the driver mutation (JAK2 or CALR) (C). The distribution of allelic burden of additional mutations is represented by violin plots (D), and the association between mutations on the genes of the 4-tier classification is shown in a circos plot (E). *P < .05; **P < .01; ***P < .001. Only pathogenic and likely pathogenic mutations were retained for the circos plot. VAF, variant allele frequency.

Figure 2.

Bayesian network of the molecular landscape. Driver genes are in blue. A green link represents a positive association, and a red one represents 2 mutually exclusive genes. Red, yellow, and blue genes represent the genomic groups identified in the following order: TP53 mutations, high-risk mutation (without TP53 mutation), and ASXL1-only mutation (ie, without TP53 or high-risk mutation).

Figure 3.

Prognosis according to the 4-tier genomic classification. Kaplan-Meier curve representing OS (A) or cumulative incidence of leukemic transformation (B). Follow-up was right censored at 72 months for Kaplan-Meier.

Figure 4.

Prognosis performance comparison. The performance (time-AUC; A and C) and accuracy (Brier score; B and D) for death prediction were measured over the time for usual prognosis scoring systems, the 4-tier genomic classification (NGS) alone, or in combination for primary (A-B) and secondary (C-D) myelofibrosis patients.