Spectrum of ASXL1 mutations in primary myelofibrosis: prognostic impact of the ASXL1 p.G646Wfs*12 mutation

TO THE EDITOR:

The discovery of prognostically informative mutations in patients with primary myelofibrosis (PMF)¹  prompted the development of mutation-enhanced risk scores.^2-6  Among these mutations, those in ASXL1 are invariably associated with adverse outcome, either as single variant or as part of a high-mutation-risk category (HMR) including SRSF2, EZH2, and IDH1/2.^7,8  A frequently occurring ASXL1 variant, p.G646Wfs*12 c.1934dup (G646Wfs*12), was long considered an artifact due to sequencing errors owing to the presence of a homopolymer consisting of 8 guanine nucleotides⁹ ; therefore, this variant may go largely undetected and/or filtered-out in next-generation sequencing (NGS) analysis, possibly resulting in patients’ underscoring. To address this point, we evaluated the prognostic impact of ASXL1G646Wfs*12 in a large population of patients with PMF.

Consecutive cases of PMF, according to 2016 World Health Organization (WHO)¹⁰  criteria, for which an NGS sequence (Personal Genome Machine platform) of the entire ASXL1 exon 12 was available, were recruited from the Centro di Ricerca e Innovazione per le Malattie Mieloproliferative (CRIMM) database (n = 333). The study received institutional review board approval and was conducted in accordance with the Declaration of Helsinki; patients provided written informed consent. Mutation analysis was performed on purified granulocytes, as described.^2,8,11  All samples were analyzed for ASXL1G646Wfs*12 by bidirectional Sanger sequencing. Survival was calculated from diagnosis to death or last follow-up, or censored at transplantation. The cumulative probability of overall survival (OS) and leukemia-free survival (LFS) was calculated by the Kaplan-Meier method; differences were estimated by the log-rank test. Continuous variables were analyzed by the Mann-Whitney U or Kruskal-Wallis test with the Dunn method for multiple comparison. A P < .05 was considered statistically significant.

The list and gene localization of ASXL1 mutations are presented in Figure 1A. ASXL1 nonsynonymous variants were detected in 119 patients (35.7%), including 11 (9.2%) with ≥2 variants. Among these, 72 (56.0%) were frameshift, 49 (38.0%) nonsense, and 8 (6.0%) missense; 45.8% mapped in the steroid receptor coactivator 1 (SRC1)-binding domain and 54.2% in interdomain regions. By NGS, ASXL1G646Wfs*12 was called in 4 patients; after Sanger sequencing, 29 more such cases were identified. Therefore, ASXL1G646Wfs*12 was the most frequent variant, accounting for 27.7% of all ASXL1 mutations, followed by E635fs*15 (9.2%) and R693* (8.4%) (Figure 1A). ASXL1 mutations were enriched in overt PMF compared with pre-PMF (43.8% vs 24.4%; P = .02); a similar trend was noted for ASXL1G646Wfs*12 (11.8% vs 7.2%; P = .07). All mutations were heterozygous (median [range] variant allele frequency (VAF), 42% [5% to 50%]), except in 1 patient each with Q575* (VAF 71%) and E635fs*15 (VAF, 60%). The VAF of ASXL1G646Wfs*12 (37.5% [9% to 49%]) was similar to other ASXL1 variants (42.0% [5% to 71%]). ASXL1 mutations were more represented in triple-negative patients (50%) compared with patients with JAK2V617F and CALR mutations (33% for both) and MPLW515x mutations (31.5%) (P = .001). In particular, ASXL1G646Wfs*12 was found in 47.6% of triple-negative patients compared with JAK2V617F-mutated patients (26.8%), CALR type 1–mutated patients (7.1%), CALR type 2–mutated patients (14.2%), and MPLW515x–mutated patients (16.7%) (P = .02; Table 1). Additional nondriver mutations were enriched in ASXL1-mutated patients (71.2%) compared with ASXL1 wild-type patients (52.3%; P = .001), with no difference between ASXL1G646Wfs*12 and other ASXL1 variants, other than more TET2 mutations in the former (27.3% vs 9.3%; P = .02).

Hematologic and clinical features of the patients, 139 with prefibrotic PMF (41.7%) and 194 overt PMF (58.3%),¹⁰  are shown in Table 1. Median follow-up was 4.5 years (range, 0.1-33.1 years), during which time 150 deaths (45%) and 35 leukemic transformations (10.5%) occurred. The median OS of the entire cohort was 8.1 years (95% confidence interval [CI], 5.9-10.3), 12.4 years (95% CI, 10.3-14.5) for prefibrotic PMF, and 6.5 years (95% CI, 5.2-7.8) for overt PMF (P = .001). Compared with the wild-type counterpart, patients harboring any ASXL1 mutation differed for leukocyte, platelet, and blast count; constitutional symptoms; splenomegaly; male sex; diagnosis of overt PMF; bone marrow fibrosis grade ≥ 2; unfavorable karyotype¹² ; and more advanced International Prognostic Scoring System (IPSS) and mutation-enhanced International Prognostic Scoring System for transplantation-age patients with PMF (MIPSS70) categories (Table 1). No notable difference was found by comparing ASXL1G646Wfs*12 and other ASXL1 mutations.

Of the 150 patients who died, 78.8%, 66.3%, and 31.3% had ASXL1G646Wfs*12, ASXL1 others, and wild-type genotype, respectively (P < .0001 vs wild-type patients). There was no differential impact on survival of ASXL1 mutation type (ie, frameshift, including ASXL1G646Wfs*12, nonsense and missense), whereas each of them conferred adverse prognosis compared with wild-type genotype (supplemental Figure 1, available on the Blood Web site; P < .0001 for all), confirming the report of others.¹³  In particular, the median (range) survival of ASXL1G646Wfs*12-mutated patients was 3.2 years (2.2-4.2 years), significantly shorter than other ASXL1 mutations (4.5 years [3.7-5.3 years]; P = .03) and wild-type genotype (12.4 years [9.1-15.7 years]; P < .0001) (Figure 1B). The hazard ratio (HR) (95% CI) for survival vs wild-type patients was 4.3 (95% CI, 2.7-6.9) and 2.7 (95% CI, 1.9-3.8) for ASXL1G646Wfs*12 and other ASXL1 mutations, respectively.

The adverse impact of ASXL1G646Wfs*12 was also confirmed by separate analysis of pre-PMF and overt PMF patients (Figure 1C-D). Median (range) survival of patients harboring ASXL1G646Wfs*12 was 2.1 years (1.5-2.8 years) and 3.6 years (2.6-4.7 years) in pre-PMF and overt PMF, respectively, compared with 7.1 years (2.1-12.1 years) and 4.6 years (2.6-4.7 years) for other ASXL1 mutations, and not reached and 10.9 years (7.0-14.8 years) for wild-type patients (P < .0001), respectively. The HR for ASXL1G646Wfs*12 and other ASXL1 mutations was 7.2 (95% CI, 3.0-17.2) and 3.6 (95% CI, 1.8-7.0) in pre-PMF, and 3.1 (95% CI, 1.8-5.4) and 2.1 (95% CI, 1.3-2.1) in overt PMF (P < .0001). We then performed multivariable analysis to validate the prognostic significance of ASXL1G646Wfs*12 in the context of MIPSS70 and MIPSS70plusV2.0. Results indicate that ASXL1G646Wfs*12 remained independently associated with shortened survival, as well as any other ASXL1 mutation (supplemental Table 1).

These findings were validated in an independent cohort of 271 patients, 81% with overt PMF, collected in 2 other institutions. NGS analysis identified ASXL1 mutations in 70 patients (25.8%) including ASXL1G646Wfs*12 in 2 (2.8%). However, after Sanger sequencing, ASXL1G646Wfs*12 was detected in 20 additional cases (7.4%), to a total of 90 ASXL1-mutated patients (33.2%); ASXL1G646Wfs*12 was the most frequent variant (24.4% of total). Of the 144 patients who died (53.1%), 43.7% harbored ASXL1 mutations, and the proportion of those with ASXL1G646Wfs*12 was 31.2% (P < .0001 vs ASXL1 others). The median (range) survival of patients harboring ASXL1G646Wfs*12 was 3.8 years (0.5-5.5 years), compared with 5.9 years (4.0-7.7 years) (P = .04) for other ASXL1 mutations and 8.4 years (7.0-9.7 years) for wild-type genotype (P < .0001) (supplemental Figure 2).

Thirty-five patients transformed to leukemia (10.5%), of whom 27.3%, 14.0%, and 6.5% had ASXL1G646Wfs*12, other ASXL1 mutations, and wild-type genotype, respectively (P < .0001). The LFS was significantly shorter in patients with ASXL1G646Wfs*12 (6.7 years; range, 4.6-8.7 years) compared with ASXL1 others (not reached; P = .03) and wild-type patients (not reached; P < .0001) (Figure 1E). The HR (95% CI) vs wild-type patients was 6.0 (95% CI, 2.6-14.0) and 2.7 (95% CI, 1.2-5.7) for ASXL1G646Wfs*12 and other ASXL1 mutations, respectively. Due to the low number of events in the validation cohort, attempts to validate these findings were not performed.

In summary, data reported herein suggest that care should be taken to accurately detect and report the ASXL1G646Wfs*12 variant in patients with PMF because of its strong adverse impact on OS and LFS. A recent report provided extensive bioinformatic advice to correctly call this variant from background noise in NGS traces, concluding that ASXL1G646Wfs*12 variant is a bona fide mutation¹⁴ ; another study indicated that this variant may represent an early mutational event in a dominant clonal population of hematopoietic cells.¹⁵  Finally, it seems reasonable, but remains to be shown, that ASXL1 G646Wfs*12 has similar adverse prognostic relevance in myelodysplastic syndromes and acute leukemia.