To the Editor:

The treatment of myelofibrosis (MF), a myeloproliferative neoplasm (MPN) driven by JAK-STAT pathway activating mutations, evolved with the advent of JAK inhibitors. The first-in-class agent, ruxolitinib (RUX), a JAK1/2 inhibitor, is now standard for treatment of splenomegaly and MF-associated symptoms [1]. However, MF is a biologically and clinically heterogeneous disease with certain difficult to treat patient subgroups. In particular, disease- or treatment-associated thrombocytopenia is associated with adverse outcomes [2, 3] and often requires RUX dose reductions or interruptions which may limit treatment efficacy. Thrombocytopenic patients who discontinue RUX have a median survival of less than 1 year [3].

Pacritinib (PAC), a JAK2/IRAK1/ACVR1 inhibitor that spares JAK1, has shown clinical benefit in thrombocytopenic MF in the PERSIST-1 and -2 trials [4, 5]. PAC203 was a randomized dose finding study of PAC in primary or secondary MF patients who were refractory or intolerant to RUX (RUX-ref/int), including patients with moderate and severe thrombocytopenia. Patients were randomized 1:1:1 (PAC 100 mg once daily [QD], 100 mg twice daily [BID] or 200 mg BID) stratified by baseline platelet count. This study established PAC 200 mg BID as the optimal efficacious and safe dose [6] and PAC is now FDA approved for the treatment of patients with MF who have thrombocytopenia.

Previous studies established the adverse prognostic implications of certain somatic gene mutations in MF; specifically mutations in epigenetic (ASXL1, EZH2) [7], splicing factor (SRSF2, U2AF1) [7] and IDH1/IDH2 genes [7] are associated with disease progression and shortened survival. Reduced likelihood of RUX response has been associated with ≥3 mutations [8] but not with mutation type [9, 10]. Shorter time to RUX failure has been reported in those with ASXL1/EZH2 mutations [9] and reduced time to RUX discontinuation in patients with ≥3 mutations [8]. Specific cytokine signatures have been correlated with RUX resistance [11] suggesting possible biologically relevant pathways (e.g. NFκB) mediating resistance. The mutation profiles of RUX-ref/int thrombocytopenic MF patients have not been well delineated. This represents a group with a major unmet need for effective management strategies, and a better understanding of their mutation profiles will assist the application of precision medicine in this challenging group.

We therefore performed mutational analysis on a subgroup the PAC203 cohort (110 patients at baseline and 42 patients at 24 weeks follow-up using a 32-gene TruSeq Custom Amplicon Panel (see Supplementary Methods). Furthermore, we interrogated cytokine profiles to understand the relationship between inflammatory signatures and clinico-genomic profiles in this cohort.

Characteristics of this group was representative of the overall PAC203 cohort [6]. Median follow-up time was 213 (95% confidence interval [CI]: 189–236) days. The median age was 68 (37–87) years, the median duration of prior exposure to RUX was 1.59 years (range 0–11 years) with 72.7% reporting prior exposure to non-RUX therapies (range 1–5 lines). Primary myelofibrosis (PMF) was the most prevalent disease category (56.4%, 62/110), followed by post-polycythemia vera MF (PPV-MF) in 29.1% (32/110) and post-essential thrombocythemia MF (PET-MF) in 14.5% (16/110). Thrombocytopenia was common: median baseline platelet count was 63 ×109/L, with 38.2% (42/110) <50 ×109/L. Baseline hemoglobin was <10 g/dL in 64.5% (71/110) of the cohort.

MPN driver mutation frequency was as expected for MF [10, 12]; JAK2V617F mutation was present in 77.3% (85/110), CALR-mutation in 12.7% (14/110; type 1: n = 11, type 2: n = 3), MPL-mutation in 8.2%, and “triple-negative” in 1.8% of cases, Fig. 1A. JAK2V617F variant allele frequency (VAF) was ≥50% in 68.2% (58/85) with VAF < 20% present in just 5.9% (n = 5) of patients. Non-MPN driver mutations (NDM) were present in 76.4% (n = 84) with ≥3 NDMs in 20.9% (23/110) of patients. Analogous to previous reports, the most prevalent NDMs were in ASXL1 and TET2 genes (in 29.1%, n = 32, and 26.4%, n = 29, of patients respectively) (Fig. 1A, Supplementary Table S1). Splicing factor (SF) gene mutations were mutually exclusive and detected in 34.5% (n = 38/110) of patients, which included SF3B1 [13.6%, n = 15], U2AF1 [12.7%, n = 14], SRSF2 [5.5%, n = 6], ZRSR2 [2.7%, n = 3]. Patients with SF mutations were more often categorized as PMF (76.3%) rather than PET-MF (15.8%) or PPV-MF (7.9%), P = 0.001 (Supplementary Table S2A). SF-mutated patients had lower baseline hemoglobin level (Hb <8 g/dL in 39.5% as compared with 16.9% in SF-wild type [WT], P = 0.009) and were more likely to be red cell transfusion dependent at trial entry (RCC-D) as compared with SF-WT patients (42.1% vs 22.2% respectively, P = 0.012; Supplementary Table S2A). SF3B1-mutated patients had higher trial entry platelet counts (platelet count >100 × 109/L) in 66.7% vs. 28.7% in SF3B1-WT patients, P = 0.004).

Fig. 1: The mutation and cytokine profiles in the PAC203 cohort.
figure 1

A Waterfall plot of mutation distribution in the PAC203 cohort. B Forest plot illustrating the results of logistic regression analyses of mutation statuses and baseline hemoglobin count associated with the likelihood of grade 3/4 anemia; TET2-mutated patients were more likely to experience grade 3/4 anemia independent of baseline hemoglobin level; odds ratio (OR) 4.5, 95% CI 1.4–13.9, P = 0.009 (upper panel). Forest plot illustrating the results of logistic regression analyses of mutation statuses and baseline platelet count associated with the likelihood of grade 3/4 thrombocytopenia; KRAS/NRAS-mutated patients were more likely to experience grade 3/4 anemia after adjustment for baseline platelet level; OR 3.65, 95% CI 1.2–11.3, P = 0.026. (lower panel). Univariate logistic regression was performed for each variable. Significant P values (<0.05) highlighted in red and OR denoted by (*) were adjusted. JAK2V617F JAK2 V617F-mutated; HMR high molecular risk mutation [IDH1/2, SRSF2, ASXL1, EZH2, U2AF1Q157]; SF splicing factor mutation [SF3B1, U2AF1, SRSF2, ZRSR2], ASXL1 ASXL1-mutated; TET2 TET2-mutated, RAS KRAS/NRAS-mutated, BL Plt <50 baseline platelet level <50 × 109/L. C Cluster dendrogram of cytokine levels in ruxolitinib refractory / intolerant study cohort with cluster scores 2 and 4 highlighted in the table for high molecular risk positive (HMR+) and negative (HMR−) patients.

Full size image

High molecular risk mutations (HMR; IDH1/2, SRSF2, ASXL1, EZH2, U2AF1Q157) [7] were present in 43.6% (48/110) and ≥2 HMR mutations were present in 15.4%, a prevalence similar to other high-risk enriched MF cohorts [12, 13]. No clinical parameters were associated with a HMR mutation (Supplementary Table S2B). Strikingly, RAS-pathway mutations, KRAS/NRAS/CBL (RAS/CBL-MT), were found at a higher frequency than previously described in MF cohorts [13, 14] in 20.9% of patients (n = 23; RAS n = 20, CBL n = 3; Fig. 1A, Table 1). These mutations were sub-clonal in the majority with a median VAF of 10% (range 1.9–95%). RAS mutations occurred in known mutation hotspots; the most prevalent was in codon G12 (n = 12/20) [13]. RAS/CBL-MT patients had a significantly higher frequency of NDMs (≥3 in 56.5% vs. 9.2% for RAS/CBL-WT patients, P = 0.0001, Table 1) and a co-mutated HMR mutation (65.2% vs 37.9% for RAS/CBL-WT patients, P = 0.02, Table 1). KRAS/NRAS/CBL and TP53 (n = 7 patients) mutations were mutually exclusive in this cohort.

Table 1 RAS/CBL-mutated patient baseline clinical and mutation characteristics.
Full size table

In patients with both molecular and 24-week clinical data, there were no significant correlations between driver or NDM mutation status (including specific analyses relating to HMR and RAS/CBL-MT status) and SVR or TSS response, although numbers of events for analysis were low. Grade 3/4 anemia occurred more often during the study period in TET2-mutated patients (odds ratio [OR] 4.2, 95% CI 1.4–13, P = 0.012), Fig. 1B. Grade 3/4 thrombocytopenia occurred more frequently in RAS/CBL-MT patients (OR 3.64, 95% CI 1.2–11.3, P = 0.026, Fig. 1B), including after adjusting for baseline platelet strata (< vs. ≥50 ×109/L). The presence of ≥3 NDMs was associated with an increased risk of infections (OR 7.59, 95% CI 2.45–23.4, P = 0.0001).

Follow-up molecular analysis at week 24 was performed in 38.2% (n = 42/110). No significant driver or NDM molecular responses (≥50% reduction in VAF) were detected. At least one new NDM was acquired in 7.1% (3/42) including CBL [2], TET2 [1], TP53 [1], U2AF1 Q157 [1]. No associations were observed between follow-up mutation analyses and outcomes.

Unsupervised clustering identified 6 cytokine clusters at baseline, Fig. 1C. Elevated cluster 2 (P = 0.009) and 4 (P = 0.006) scores were associated with HMR mutations. Higher cluster 2 scores were also associated with driver mutation VAF ≥50%. The pro-inflammatory cytokines in cluster 2 linked to HMR mutations (HMR+) represented a cluster regulated by the NFκB pathway. The presence of a HMR mutation was particularly associated with significantly higher IL-8 levels (40.5 pg/ml) as compared with absence of an HMR mutation (24.5 pg/ml), P < 0.0001. Elevated tumor necrosis factor-alpha (TNF-α) was also associated with HMR mutations; TNF-α was 61 pg/ml in HMR+ vs. 48.5 pg/ml for HMR−, P = 0.009. Although RAS-pathway mutations were not associated with specific cluster scores, these patients did have higher levels of the NFκB-associated cytokine IL12P40 (1.1 ng/ml) as compared with RAS/CBL-WT patients (0.6 ng/ml), P = 0.001. There was no association between cytokine cluster scores and exposure to RUX.

We report the mutation landscape in RUX-ref/int cytopenic MF, showing enrichment for HMR mutations and, in particular, a higher frequency of RAS-pathway mutations (20.9%) than previously reported in MF cohorts (to date at a frequency of 6–8.1%) [13, 14]. RAS and HMR mutation co-occurrence has previously been described, which we also observed [14]. RAS-pathway mutations often showed low allele burden and correlated with the presence of multiple NDMs, consistent with presence of RAS-pathway mutations in patients undergoing genetic evolution. Although mutation data was not available prior to RUX treatment in this cohort, recent single cell genetic analyses in myelofibrosis show RAS-pathway mutations were one of most common emergent mutations after exposure to RUX [15]. Activating mutations of the RAS-pathway have also been reported to correlate with reduced likelihood of spleen and symptom responses in patients with myelofibrosis treated with dual JAK1/2 inhibitors [13]. RAS-pathway mutations in MF have also been associated with shorter survival and progression to leukemia [14]. The PAC203 cohort therefore represents a genetically high risk group of patients.

Importantly, we report for the first time a relationship between HMR and RAS mutations and a pro-inflammatory cytokine signature. This signature mirrors a previously described RUX resistant cytokine profile [11] involving NFκB signaling. A potential underlying mechanism may be that the inflammatory microenvironment creates a selective pressure promoting the evolution of subclones carrying HMR and RAS-pathway mutations. We speculate that this combination of cell-intrinsic genetic properties of the clone, and cell-extrinsic inflammatory microenvironment might collectively confer JAKi resistance. Therapeutic strategies, including dual blockade of JAK2 and NFκB, may prove beneficial for treatment of MF. As a JAK2/IRAK1 inhibitor, PAC targets both pathways, as IRAK1 is upstream of NFκB signaling suggesting a potential role in those with HMR and RAS-pathway mutations. Although no specific TSS or SVR responses were observed in these patients on PAC203, the numbers of patients available for analysis was low, and the follow-up period may not have been sufficient to capture responses in this subgroup. Other strategies, including combinations targeting JAK and MEK/ERK pathways together with inflammatory pathways, for example through bromodomain inhibition, could be an effective strategy to mitigate clonal evolution in high-risk patients.

In summary, the PAC203 cohort encompasses a molecularly high-risk group, with a high incidence of HMR and RAS pathway mutations that may be associated with JAK1/2 inhibitor resistance. Our findings will help inform the application of precision medicine for this group of patients with a major unmet need for new therapeutic strategies.