Concomitant analysis of EZH2 and ASXL1 mutations in myelofibrosis, chronic myelomonocytic leukemia and blast-phase myeloproliferative neoplasms

Somatic mutations in EZH2 and ASXL1 have recently been reported in myelofibrosis (MF) and chronic myelomonocytic leukemia (CMML).^1-4 The current study examines the frequency of these mutations in a larger cohort of patients with these diseases and their correlation with other myeloproliferative neoplasm (MPN)-associated mutations and disease phenotype. Given that many earlier studies of mutations in ASXL1 did not utilize paired normal tissue, and have largely reported mutations occurring in exon 12 (refs ^3,5-7), we also present previously unpublished/unannotated data regarding novel somatic missense mutations/single-nucleotide polymorphisms (SNPs) throughout the coding region of ASXL1.

Study samples from 143 patients were recruited from the Mayo Clinic, Rochester, MN, USA (n = 94; 46 primary MF, 22 post-polycythemia vera/essential thrombocythemia myelofibrosis (post-PV/ET MF), 11 blast-phase MPN and 15 CMML), Harvard Institutes (n = 25 primary myelofibrosis (PMF) with paired normal tissue) and MD Anderson Cancer Center (n = 24 CMML with paired normal tissue). High throughput DNA resequencing was used to screen bone marrow (Mayo and MD Anderson samples) or granulocyte (Harvard MPD cohort)-derived DNA coding regions of EZH2 (NM_015338), ASXL1 (NM_004456) and TET2 (NM_017628), as well as the regions of known mutations in IDH1, IDH2, JAK2 and MPL in all samples. CMML samples were also screened for known FLT3, K Ras and N Ras mutations. Nonsynonymous alterations not found in SNP databases (dbSNP) were annotated as somatic mutations or SNPs on the basis of sequence analysis of matched-germ line buccal swab DNA. Nonsynonymous alterations not in dbSNP nor determined to be somatic in paired samples or in recently reported data were censored.

Analysis of germ line DNA allowed us to distinguish between somatic missense mutations and previously unannotated SNPs in ASXL1 (Table 1); all unannotated SNPs were observed in matched normal tissue in at least two samples. All frameshift and nonsense mutations were not present in matched normal tissue. This strategy allowed us to identify nine novel somatic missense mutations in ASXL1 that were not previously reported. The majority of somatic ASXL1 mutations in CMML/MPN were reported to occur in the last exon of ASXL1 (Figure 1a);^3-7 however, full-length resequencing allowed us to identify an additional region of frequent somatic mutations in ASXL1 in exon 9, which encodes the ASXM domain (Figure 1a). The functional contribution of mutations in any region of ASXL1 has yet to be elucidated. We also identified six novel mutations in EZH2 in the PMF and CMML samples in this cohort. All of these mutations were heterozygous mutations, which resulted in a premature stop codon and were distributed along multiple exons of EZH2 (Figure 1b).

Among the 94 study samples from Mayo, ASXL1 mutations were identified in six (13%) PMF, five (23%) post-PV/ET MF, two (18%) blast phase MPN and three (20%) CMML cases. The corresponding TET2 mutational frequencies were 15, 14, 18 and 13%. EZH2 mutations were seen in three (7%) of 46 PMF cases, only. IDH1/2 mutations were restricted to blast-phase MPN (36%) and PMF (6%). Among 16 ASXL1 mutated cases from the Mayo series, one post-PV MF case had both TET2 and JAK2 mutations, one PMF mutant EZH2, one post-PMF AML mutant IDH and MPL, one post-PV mutant JAK2 and MPL and five mutant JAK2; among the 14 TET2 mutated cases, one PMF case carried mutant IDH, one mutant ASXL1, five mutant JAK2 and one mutant MPL. Of the three EZH2 mutated PMF cases, one case had a concomitant ASXL1 mutation and another a MPL mutation. Among the 25 Harvard PMF cases, 3 (12%) displayed ASXL1 and one (4%) IDH1/2 mutations; of note, a single case had concurrent ASXL1, IDH and JAK2 mutations. The MD Anderson CMML cases showed higher frequencies of ASXL1 (56%), TET2 (44%) and EZH2 (12%) mutations; Figure 1c shows the distribution of mutations in this cohort, including the co-occurrence of ASXL1 and TET2 mutations in seven cases and ASXL1 and EZH2 mutations in two cases.

Detailed clinical information was available for the Mayo cases. EZH2- and ASXL1-mutated PMF patients were similar to their unmutated counterparts in terms of age and sex distribution, were cytogenetically normal and none underwent leukemic transformation during follow-up. The three EZH2-mutated PMF patients died after 29, 48 and 67 months from the time of mutation analysis. In univariate analysis, the presence of mutant ASXL1 in PMF was associated with worse survival (P = 0.06), but the borderline significance was lost during multivariable analysis that included risk stratification according to DIPSS.⁸ The lone PMF patient which had both ASXL1 and TET2 mutations was diagnosed with PV at age 67 years after a 12-year history of unexplained leukocytosis. He was treated with phlebotomy and hydroxyurea and started to develop progressive splenomegaly by age 71, and was ultimately diagnosed with post-PV MF by age 73. At the time, he was enrolled in a thalidomide therapy-based clinical trial and developed pulmonary extramedullary hematopoiesis by August 2005 and ultimately died in October 2005, without progressing into AML. The three ASXL1 mutated CMML cases were alive after 40, 34 and 12 months from time of mutation analysis and none of them had progressed to acute leukemia; karyotype was normal in two of the patients and showed isolated trisomy 8 in one.

The current study illustrates a number of points. First, it provides incidence estimates for EZH2 and ASXL1 mutations in a relatively large cohort of CMML and MF cases. The results suggest that ASXL1 mutations are as frequent as TET2 mutations and the two were not necessarily mutually exclusive, although their co-occurrence was less likely in MF, as compared with CMML (Figures 1c and d). The biological consequence of this particular observation is currently not clear and will need to be investigated in future functional studies. Regardless, the current study underscores the profound lack of disease specificity and mutual exclusivity shown by recently described MPN-associated mutations and our observations do not support the suggestions from recent studies regarding the association of ASXL1 or EZH2 mutations with poor clinical outcome. However, a larger number of patients followed for a longer period of time is needed to validate these preliminary observations.