Non-driver mutations in myeloproliferative neoplasm-associated myelofibrosis

We studied non-driver mutations in 62 subjects with myeloproliferative neoplasm (MPN)-associated myelofibrosis upon diagnosis, including 45 subjects with primary myelofibrosis (PMF) and 17 with post-polycythemia vera or post-essential thrombocythemia myelofibrosis (post-PV/ET MF). Fifty-eight subjects had ≥1 non-driver mutation upon diagnosis. Mutations in mRNA splicing genes, especially in U2AF1, were significantly more frequent in PMF than in post-PV/ET MF (33 vs. 6%; P = 0.015). There were also striking differences in clonal architecture. These data indicate different genomic spectrums between PMF and post-PV/ET MF. Electronic supplementary material The online version of this article (doi:10.1186/s13045-017-0472-5) contains supplementary material, which is available to authorized users.

(WHO) [7]. The median interval from diagnosis of PV or ET to post-PV/ET MF was 12 years (range, 5-30 years). Prognosis was evaluated with the Dynamic International Prognostic Scoring System (DIPSS) [8]. The cytogenetic findings were classified as favorable or unfavorable using the DIPSS-plus classification [9]. Subject-, disease-, and therapy-related variables are displayed in Additional file 1: Table S1.
Targeted gene sequencing was carried out upon diagnosis of myelofibrosis. Non-driver mutations were compared against expected patterns and categorized into oncogenic mutations, possibly oncogenic variants or unknown significance. Germline control DNA was used to ascertain the somatic mutations. To account for the absence of a matched control for some subjects, a bespoke variant selection pipeline was used. Details of the methods are described in the Additional file 2 and Additional file 1: Table S2-S3. Forty-two subjects with PMF (93%) had ≥1 non-driver mutation. Two of the 17 subjects with no driver mutation in JAK2 V617F , JAK2 exon 12, CALR, or MPL W515 had mutations in SH2B3 or PIAS3 of the JAK-STAT signaling pathway, 13 had mutations in other genes, and 2 had no mutations. Non-driver mutations were detected in 16 subjects with post-PV/ET MF (94%). There was no difference in the median numbers of non-driver mutations in the subjects with PMFs vs. those with post-PV/ET MF (3; [range, 0-5] vs. 3; [range, 0-7]; P = 0.885). There was also no difference in the median numbers of non-driver mutations (3; [range, 0-5] vs. 3 [range, 0-5]; P = 0.668) in the subjects with PMF with and without driver mutations.
The distribution of non-driver mutations in PMF vs. post-PV/ET MF is shown in Fig. 1a Changes in non-driver genes were grouped into several functional pathways. The most frequent mutation targets were genes associated with signal transduction (N = 31),  Fig. 1c). The mutation frequencies of the mRNA splicing genes were significantly different between the subjects with PMF and those with post-PV/ET (33 vs. 6%; P = 0.015). U2AF1 mutations were significantly more common in the subjects with PMF than those with post-PV/ET MF (22 vs. 0; P = 0.008).
We used copy number-adjusted variant allelic frequencies (VAF) in each subject with a non-driver mutation to determine whether a mutation was in the ancestral clone or whether it arose in a sub-clone. Using statistically significant differences in VAFs, the subjects were classified as representing two different clonal architectures: (1) ancestral clonal only; or (2) ancestral clone and sub-clone(s).  Fig. 2d). In the subjects with PMF, the driver and non-driver mutations were ancestral in 14 subjects (e.g., subject 2015-R02413 in Fig. 2a) and sub-clonal in 9 subjects (e.g., subject 2015-R02406 in Fig. 2b). Eleven subjects with post-PV/ET MFs had JAK2 V617F as the only ancestral mutation (e.g., subject G121517R00701 in Fig. 2c).
In conclusion, we found differences in the non-driver mutation profile and the clonal architecture among the subjects with PMF compared with those with post-PV/ET MF. There was a significant difference in the frequency of the mRNA splicing genes among cohorts, similar to data reported by others [2,10]. In individuals with PMF, approximately 30% of driver mutations co-occur with non-driver mutations as ancestral events, whereas approximately 20% of driver mutations were in sub-clones. In contrast, driver mutations occurred only in the ancestral clone in most subjects with post-PV/ET MF.

Additional files
Additional file 1: Table S1. Clinical and laboratory features of 45 patients with PMF and 17 patients with post-PV/ET MF. Table S2. Gene list of the 190-gene NGS panel. Table S3. Two hundred and twenty-nine