Targeted deep sequencing in polycythemia vera and essential thrombocythemia

Despite seminal descriptions of novel^1-6  and other less specific^7,8  mutations in the last 10 years, the molecular pathogenesis of myeloproliferative neoplasms (MPNs) remains incompletely understood; however, there has been increasing use of these mutations in diagnostics and disease prognostication.⁹  Mutations in MPNs are operationally classified into those that involve JAK2, calreticulin (CALR), and myeloproliferative leukemia virus oncogene (MPL) and those that involve other genes, such as additional sex combs like transcriptional regulator 1 (ASXL1).⁹  The former are believed to be “driver” mutations and are now formally integrated into the World Health Organization (WHO) diagnostic criteria for polycythemia vera (PV; 98% JAK2 mutational frequency), essential thrombocythemia (ET; 60% JAK2, 22% CALR, and 3% MPL mutational frequency) and primary myelofibrosis (PMF).¹⁰  In terms of disease prognostication, type 1/type 1–like CALR mutations have been associated with superior survival in PMF^11,12  and JAK2 mutations with thrombosis in ET.¹³ 

The prognostic contribution of mutations other than JAK2, CALR, or MPL has been demonstrated in PMF, where ASXL1, serine/arginine-rich splicing factor 2 (SRSF2), isocitrate dehydrogenase 1 or 2 (IDH1/2), and enhancer of zeste 2 polycomb repressive complex 2 subunit (EZH2) were identified as being unfavorable to survival.⁷  More recent studies have suggested the prognostic relevance of not only the specific type but also the number of mutations.^8,14  These observations are noteworthy considering the increasing application of next-generation sequencing (NGS) technology in clinical practice. In the current study, we used custom-designed targeted sequencing of 27 myeloid neoplasm–relevant genes that are also often included in most commercially available NGS gene panels used in myeloid cancer. Our main objective was to determine the prognostic and phenotypic relevance of sequence variants/mutations, other than JAK2, CALR and MPL, in PV and ET.

The current study was approved by the institutional review boards of the Mayo Clinic and the University of Florence, Florence, Italy. Diagnoses and treatment approaches were in accordance with what was considered standard of care at the time of initial diagnosis.¹⁵  Targeted capture assays for Mayo Clinic patients (the main study cohort) were carried out on bone marrow or whole-blood DNA for the following genes: TET2, DNMT3A, IDH1, IDH2, ASXL1, EZH2, SUZ12, SRSF2, SF3B1, ZRSR2, U2AF1, PTPN11, TP53, SH2B3, RUNX1, CBL, NRAS, JAK2, CSF3R, FLT3, KIT, CALR, MPL, NPM1, CEBPA, IKZF1, and SETBP1. Additional details regarding NGS analysis are outlined in the supplemental Methods, which also includes the methods used in the validation cohort from the University of Florence. Prognostic evaluation of sequence variants/mutations considered both the “number” of sequence variants/mutations and the specific genes affected. “Adverse sequence variants/mutations” were identified by age-adjusted multivariable analysis of their impact on overall survival or multivariable analysis of impact on leukemia-free or myelofibrosis-free survival.

All statistical analyses considered clinical and laboratory parameters obtained at time of diagnosis. Differences in the distribution of continuous variables between categories were analyzed by either the Mann-Whitney or the Kruskal-Wallis test. Patient groups with nominal variables were compared by χ² test. Survival analysis was considered from the date of diagnosis to date of death or last contact; because of the considerable effect of age on survival, univariate analysis for overall survival was adjusted for age. Leukemia-free and myelofibrosis-free survival calculations considered the transformation events as the uncensored variable. Survival curves were prepared by the Kaplan-Meier method and compared by the log-rank test. The Cox proportional hazard regression model was used for multivariable analysis. P values <.05 were considered significant. All analyses were conducted using the Stat View (SAS Institute, Cary, NC).

A total of 316 Mayo Clinic patients with PV (n = 133) or ET (n = 183) were evaluated; Tables 1 and 2 outline the presenting clinical and laboratory details. “Driver” mutation distribution was 98% JAK2 for PV and 52% JAK2, 26% CALR, and 4% MPL for ET. According to conventional prognostic models,^16,17  survival risk distribution at time of diagnosis was 45% high, 26% intermediate, and 29% low for PV, and 21% high, 39% intermediate, and 40% low for ET. Median follow-up was 10.4 years for PV and 10.1 years for ET. The respective numbers of documented deaths, leukemic transformations, and fibrotic progressions were, for PV, 62 (47%), 7 (5.3%), and 14 (11%) and for ET 61 (33%), 6 (3.3%), and 27 (15%).

Sequence variants/mutations other than JAK2, CALR, or MPL were seen in 70 patients with PV (52.6%) and 96 with ET (52.5%) (Figure 1); supplemental Table 1 provides additional information on specific variants/mutations. The respective percentages of patients with 1, 2, or ≥3 sequence variants/mutations were 30%, 20%, and 3% for PV and 41%, 8%, and 4% for ET. The most frequent sequence variants/mutations in both disorders were TET2 and ASXL1 (Figure 1). There were no significant associations between JAK2/CALR/MPL mutational status and number or type of other sequence variants/mutations. Among the most frequent sequence variants/mutations, ASXL1 in ET cosegregated with EZH2 (P < .001), IDH2 (P = .004), and RUNX1 (P = .002) and in PV with IDH2 (P = .003) and KIT (P = .02); TET2 in PV cosegregated with SH2B3 (P = .01).

Phenotypic correlations were examined for sequence variants/mutations with a recurrence rate of at least 5% (Figure 1). Parameters examined for possible association included age, sex, risk category,^16,17  karyotype, thrombosis history, leukocyte count, hemoglobin level, platelet count, palpable splenomegaly, pruritus, and microvascular symptoms. The most noteworthy association was between TET2 variants/mutations and thrombosis in ET (P = .01), which was independent of both age and JAK2/CALR/MPL mutational status; relative risk on age-adjusted multivariable analysis was 3.4 (95% confidence interval [CI], 1.4-8.4), compared with 4.8 (95% CI, 1.6-14.2) for JAK2 vs CALR. However, such association was not evident in either PV or an external cohort of 174 Italian patients with ET seen at the University of Florence (P = .58). In ET, TET2 and SF3B1 variants/mutations were associated with older age (P = .04 and 0.04, respectively), SF3B1 with higher platelet count (P = .02), and ASXL1 with palpable splenomegaly (P = .009). In PV, the only association noted was between SH2B3 variants/mutations and palpable splenomegaly (P = .045).

Median survival of the 133 Mayo Clinic patients with PV was 14.2 years. Age-adjusted univariate analysis identified SRSF2, ASXL1, and IDH2 variants/mutations as being significantly associated with inferior survival; ASXL1 (hazard ratio [HR], 2.2; 95% CI, 1.1-4.3) and SRSF2 (HR, 6.1; 95% CI, 2.0-19.2) remained significant on multivariable analysis. For leukemia-free and myelofibrosis-free survival, univariate analysis showed significant associations with SRSF2, IDH2, and RUNX1 for the former and with SRSF2 and RUNX1 for the latter; on multivariable analysis, SRSF2 (HR, 74.5, 95% CI, 4.4-1261.7) and IDH2 (HR, 55.5; 95% CI, 3.5-887.4) remained significant for leukemia-free survival and SRSF2 (HR, 27.2; 95% CI, 2.7-274.3) for myelofibrosis-free survival.

Based on their above-outlined independent impact on overall, leukemia-free, or myelofibrosis-free survival, ASXL1, SRSF2, and IDH2 variants/mutations were identified as being “adverse” in PV and at least 1 of the 3 was present in 20 of the 133 patients (15%). Patients with adverse variants/mutations displayed significantly worse survival when compared with both patients with other sequence variants/mutations (HR, 2.1; 95% CI, 1.1-4.1) and those without any sequence variant/mutation (HR, 2.2; 95% CI, 1.2-4.2; Figure 2A); there was no difference in survival between patients with other sequence variants/mutations and those without any sequence variant (P = .92).

The difference in survival between patients with and without adverse variants/mutations was independent of age, the IWG prognostic model for PV,¹⁷  and karyotype, with respective HRs (95% CIs) of 2.7 (1.5-4.9), 2.8 (1.5-5.1), and 2.4 (1.2-4.6). Additional multivariable analysis that included the individual variables used in the IWG prognostic model¹⁷  confirmed the independent prognostic contribution of adverse variants/mutations (HR, 3.0; 95% CI, 1.6-5.5), along with age (HR, 5.6; 95% CI, 3.1-10.0 for age ≥65 years) and leukocytosis (HR, 2.0; 95% CI, 1.2-3.5 for leukocyte count ≥15 × 10⁹/L); venous thrombosis was no longer significant in the particular multivariable model (P = .32). The presence of adverse variants/mutations was also associated with shorter leukemia-free (Figure 3A; P = .02; HR, 5.6; 95% CI, 1.3-25.4) and myelofibrosis-free (Figure 4A; P = .11; HR, 2.5; 95% CI, 0.8-8.1) survivals, although the latter did not reach significance.

Patients with 3 sequence variants/mutations (n = 4) had significantly worse survival, compared with those with 2 (n = 26; P = .04), 1 (n = 40; P = .003), or no (n = 63; P = .003) sequence variant/mutation, whereas there was no difference in survival among the latter 3 groups (P = .29), especially after adjusting for age (P = .72). Furthermore, the shorter survival seen in the 4 patients with 3 sequence variants/mutations was fully accounted for by the presence of adverse variants/mutations in all of them.

The observations from the Mayo Clinic patients were validated in an external cohort of 215 Italian patients with PV (Tables 1 and 2), followed for a median of 11.9 years and with 84 deaths (39%), 55 fibrotic progressions (26%), and 6 leukemic transformations (3%); the particular cohort was enriched for patients who experienced fibrotic progression, in order to increase the power for molecular prediction for fibrotic transformation. In multivariable analysis, the unfavorable survival effect of adverse variants/mutations on survival (Figure 2B; HR, 2.6; 95% CI, 1.5-4.6) was independent of the IWG prognostic model for PV¹⁷ ; HRs (95% CIs) were 2.3 (1.3-4.1) for adverse variants/mutations and 3.6 (1.7-7.8) for high- or intermediate-risk category. Adverse variants/mutations also affected myelofibrosis-free survival (Figure 4B), but the effect on leukemia-free survival was not significant (Figure 3B). Multivariable analysis confirmed the individual prognostic contribution of ASXL1 (HR, 2.41; 95% CI, 1.3-4.5) and SRSF2 (HR, 3.84; 95% CI, 1.2-12.6) variants/mutations to overall survival and ASXL1 to myelofibrosis-free survival (HR, 1.9; 95% CI, 1.2-3.6).

Median survival of the 183 Mayo Clinic patients with ET was 19.9 years. Age-adjusted univariate analysis identified IDH2, EZH2, and SH2B3 variants/mutations as significant risk factors for survival; SH2B3 (HR, 3.0; 95% CI, 1.03-8.2) and IDH2 (HR, 22.1; 95% CI, 2.8-176.9) remained significant on multivariable analysis. For leukemia-free and myelofibrosis-free survival, univariate analysis showed significant associations with TP53, EZH2, SRSF2, and IDH2 variants/mutations for the former and with SF3B1 and U2AF1 for the latter; on multivariable analysis, TP53 (HR, 82.8; 95% CI, 7.5-916) and EZH2 (HR, 146.8; 95% CI, 11.1-1935.6) remained significant for leukemia-free survival and SF3B1 (HR, 8.1; 95% CI, 2.5-25.8) and U2AF1 (HR, 30.3; 95% CI, 3.4-271.0) for myelofibrosis-free survival.

Based on their above-outlined independent impact on overall, leukemia-free, or myelofibrosis-free survival, SH2B3, IDH2, U2AF1, SF3B1, EZH2, and TP53 variants/mutations were identified as being “adverse” and at least 1 of them was present in 27 (14.8%) of the 183 patients with ET. Patients with adverse variants/mutations displayed significantly worse overall survival, when compared with both patients with other sequence variants/mutations (HR, 4.3; 95% CI, 2.2-8.5) and those without any sequence variant/mutation (HR, 7.0; 95% CI, 3.5-14.1; Figure 2C). In univariate analysis, patients with other sequence variants/mutations also showed significantly worse survival, when compared with patients without any sequence variant/mutation (HR, 2.0; 95% CI, 1.1-3.7); however, in multivariable analysis that included age, the difference in survival remained significant for patients with adverse variants/mutations (HR, 3.4; 95% CI, 1.6-7.0) but not for those with other sequence variants/mutations (HR, 1.5; 95% CI, 0.8-2.8). The presence of adverse variants/mutations was also associated with significantly shorter leukemia-free (Figure 3C; HR, 36.2; 95% CI, 3.8-347.9) and myelofibrosis-free (Figure 4C; HR, 6.6; 95% CI, 2.6-16.4) survivals.

Patients without any sequence variant/mutation (n = 87) had significantly longer survival when compared with those with 3 or 4 (n = 7; P < .0001) or 1 (n = 75; P = .0009) sequence variant but not to those with 2 sequence variants (n = 14; P = .28). However, when analysis was repeated after adjusting for age and excluding patients with adverse variants/mutations (n = 27), the survival impact of the number of mutations was no longer evident (P = .21); 6 of the 7 patients with 3 or more sequence variants also had adverse variants. In other words, in both ET and PV, patients with multiple sequence variants were older and more likely to harbor adverse variants, thus giving the appearance of unfavorable prognostic impact during unadjusted survival analysis.

The difference in survival between patients with and without adverse variants/mutations was independent of age (HR, 2.6; 95% CI, 1.4-4.8), IWG prognostic model for ET¹⁶  (HR, 2.6; 95% CI, 1.4-4.8), and karyotype (HR, 4.7; 95% CI, 2.4-9.2). Additional multivariable analysis that included the individual variables used in the IWG prognostic model¹⁶  confirmed the independent prognostic contribution of adverse variants/mutations (HR, 2.4; 95% CI, 1.2-4.7), along with age (HR, 3.2; 95% CI, 1.6-6.2 for age ≥60 years) and leukocytosis (HR, 2.1; 95% CI, 1.2-3.7 for leukocyte count ≥11 × 10⁹/L); thrombosis was no longer significant in the particular multivariable model (P = .68).

The observations from the Mayo Clinic patients were validated in an external cohort of 174 Italian patients with ET (Tables 1 and 2) followed for a median of 13 years and with 62 deaths (36%), 58 fibrotic progressions (33%), and 8 leukemic transformations (5%); the particular cohort was enriched for patients who experienced fibrotic progression, in order to increase the power for molecular prediction for myelofibrosis-free survival. In multivariable analysis, the unfavorable survival effect of adverse variants/mutations (Figure 2D; HR, 2.5; 95% CI, 1.5-4.4) was independent of the IWG prognostic model for ET¹⁶ : HRs (95% CIs) were 2.0 (1.1-3.6) for adverse variants, 3.5 (1.6-7.7) for the intermediate-risk category, and 8.7 (3.6-21.0) for the high-risk category. Multivariable analysis confirmed the individual prognostic contribution of U2AF1 to overall (HR, 2.9; 95% CI, 1.1-8.7) and myelofibrosis-free (HR, 3.2; 95% CI, 1.1-9.0) survival and that of TP53 to leukemia-free survival (HR, 7.3; 95% CI, 1.5-35.9). Adverse variants/mutations also affected leukemia-free (Figure 3D; HR, 6.7; 95% CI, 1.6-28.1) and myelofibrosis-free survival (Figure 4D; HR, 2.6; 95% CI, 1.4-4.5).

The revelation, in the current study, of DNA sequence variants/mutations, other than JAK2/CALR/MPL, in the majority of patients with PV or ET is consistent with the results from a smaller previous study that showed the presence of proven mutations in 44% and 29% of the respective diseases.⁸  The current study identifies prognostically important specific sequence variants/mutations and demonstrates the subsidiary role, in this regard, of the “number” of mutations.

The detrimental nature of some of the adverse variants identified in the current study has also been recognized in other chronic myeloid malignancies, including PMF (ASXL1, SRSF2, EZH2, IDH1/2, U2AF1),^7,18  chronic myelomonocytic leukemia (ASXL1),¹⁹  and myelodysplastic syndromes (ASXL1, EZH2, TP53, SRSF2).^20,21  Similarly, TP53, IDH2, SRSF2, and SH2B3 mutations were previously found to be overrepresented in blast-phase MPNs.^8,22,23  In this regard, the possible leukemogenic cooperation of mutant JAK2 with IDH1/2 mutations has previously been clinically suggested²⁴  and that of TP53 loss with mutant JAK2 demonstrated in vivo.²³  Consistent with the latter observation, all 5 TP53-mutated patients with ET, in the current study, were also JAK2-mutated, and leukemic transformation in another study⁸  was temporally linked to the consequent loss of the wild-type TP53 allele in TP53-mutated patients.

The observations from the current study are practically relevant and timely considering the increasing use of targeted NGS in routine myeloid cancer practice. We show that the occurrence of sequence variants/mutations other than JAK2/CALR/MPL, in patients with PV or ET, is not infrequent and neither the type nor number of genes involved is necessarily detrimental to disease outcome, unless it involves specific genes with prognostic relevance. On the other hand, patients with adverse variants/mutations, regardless of their conventional risk category, might require closer monitoring. In this regard, the current study suggests the possibility of refining clinically derived prognostic models by the incorporation of molecular information.

Additional studies are warranted in order to refine and possibly expand the pool of clinically relevant DNA sequence variants/mutations that could be targeted by NGS. Such a strategy would ideally also include sequence variants/mutations that identify patients with increased risk of thrombosis. In this regard, the discrepant results between the Mayo Clinic and Florence patient cohorts, in regards to the possible association between TET variants/mutations and thrombosis in ET, remains unexplained and deserves further investigation.

This work was supported in part by The Mayo Clinic Harvey-Yulman Charitable Foundation for Myelofibrosis Tissue Bank and Clinical Database of Molecular and Biological Abnormalities, The Henry J. Predolin Foundation for Research in Leukemia, Mayo Clinic, Rochester, MN. The work in Florence was supported by Associazione Italiana per la Ricerca sul Cancro (AIRC) “Special Program Molecular Clinical Oncology 5x1000” to the AIRC-Gruppo Italiano Malattie Mieloproliferative (AGIMM) group (project #1005; a detailed description of the AGIMM project is available at http://www.progettoagimm.it). P.G. was supported by GR-2011-02352109 Bando Ministero della Salute Giovani Ricercatori–Ricerca Finalizzata 2011-2012 and by AIRC IG-15967. G.R. was the recipient of a fellowship from Amici di Beat Leukemia Onlus.

Contribution: A.T. designed and sponsored the study, contributed patients, collected clinical data and patient samples, reviewed the molecular data, performed statistical analysis, and wrote the paper; T.L.L. performed and analyzed the molecular analysis; P.G. contributed patients, collected clinical data and patient samples, performed and analyzed the molecular data, performed statistical analysis, and helped with manuscript writing; C.M.F. performed the molecular analysis; G.R., A. Pacilli, A. Pancrazzi, C.M., and T.F. performed sequencing and analysis of data; Y.E. and D.B. collected clinical data; C.A.H. performed pathology review; R.P.K. performed cytogenetics review; A. Pardanani contributed patients and reviewed the molecular data; N.G. contributed patients and collected clinical data; A.M.V. designed the study, contributed patients, collected clinical data and patient samples, reviewed the molecular data, and helped with data analysis and manuscript writing; and all authors reviewed and approved the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Ayalew Tefferi, Department of Internal Medicine, Division of Hematology, Mayo Clinic College of Medicine, 200 First St SW, Rochester, MN 55905; e-mail: tefferi.ayalew@mayo.edu.