To the Editor

MPN-associated myelofibrosis is a clinical descriptor encompassing primary myelofibrosis (PMF) and post-polycythemia vera and post-essential thrombocythemia (ET) myelofibrosis (post-PV/ET MF) [1]. Most individuals with MPN-associated myelofibrosis have so-called driver mutations in JAK2, CALR, or MPL. Although the distribution of these mutations is similar in PMF and post-PV/ET MF, some data indicate that the impact of the type of driver mutation on prognosis differs [2]. Recently, non-driver mutations were reported to be associated with the initiation, progression, and prognosis of MPNs [3,4,5,6]. We analyzed whether the landscape of non-driver mutations differed among cohorts.

Forty-five subjects with PMF and 17 with post-PV/ET MF (13 post-PV MF and 4 post-ET MF) were studied. The subjects provided informed consent in compliance with the Declaration of Helsinki (The Ethics Committee of Institute of Hematology and Blood Diseases Hospital; No. KT2015010-EC-1). Diagnoses were based on the revised 2016 criteria of the World Health Organization (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, b. Twelve mutations in non-driver genes were detected in >5% of the subjects with PMF, including ASXL1 (N = 15), U2AF1 (N = 10), TET2 (N = 7), FAT1 (N = 7), SETBP1 (N = 6), SRSF2 (N = 4), CUX1 (N = 4), EP300 (N = 4), FAT2 (N = 3), NOTCH3 (N = 3), EZH2 (N = 3), and GATA3 (N = 3). In subjects with post-PV/ET MF, ASXL1 (N = 7) was the most frequent mutation, followed by TET2 (N = 5).

Fig. 1
figure 1

Mutational spectrum in primary myelofibrosis and post-PV/ET myelofibrosis. Frequency of non-driver mutations identified in the sequencing screen in 45 patients with PMF (a) and 17 patients with post-PV/ET MF (b). Frequency of gene mutations involved in common functional pathways (c)

Full size image

Changes in non-driver genes were grouped into several functional pathways. The most frequent mutation targets were genes associated with signal transduction (N = 31), followed by chromatin modification (N = 30), adhesion (N = 18), RNA splicing (N = 16), transcription (N = 15), DNA methylation (N = 15), cell cycle/apoptosis (N = 5), and DNA damage control (N = 4; 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). Figure 2a–c shows data from three subjects. Frequencies of ancestral clonal only vs. ancestral clone and sub-clone(s) architecture were significantly different between the subjects with PMF and those with post-PV/ET MF (ancestral clonal only: 49 vs. 13%; P = 0.013; 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).

Fig. 2
figure 2

Clonal architecture in PMF and post-PV/ET MF. a In a representative patient with PMF, mutant JAK2 co-occurs with non-driver mutations as ancestral mutations. b In a representative patient with PMF, mutant JAK2 was a sub-clonal mutation and other non-driver mutations precede mutant JAK2. c In a representative patient with post-PV MF, mutant JAK2 was the only ancestral mutation preceding other non-driver mutations. d More PMFs were classified as having an ancestral clonal only architecture than post-PV/ET MF

Full size image

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.