Acquired copy-neutral loss of heterozygosity of chromosome 1p as a molecular event associated with marrow fibrosis in MPL-mutated myeloproliferative neoplasms

Philadelphia-negative myeloproliferative neoplasms (MPN), including polycythemia vera (PV), essential thrombocythemia (ET), and primary myelofibrosis (PMF), are characterized by clonal hematopoiesis and overproduction of mature myeloid cells.¹  The unique JAK2 (V617F) mutation is found in about 95% of patients with PV and in 60% to 70% of those with ET and PMF.^2-5  Several studies have been conducted to identify the somatic mutations underlying the remaining 30% to 40% of patients with JAK2 (V617F)–negative ET and PMF. The only consistent observation so far is that subsets of these patients carry activating somatic mutations of the MPL gene encoding the thrombopoietin receptor and that these mutations cluster in exon 10.^6-14  Since MPL-mutant clones are often small, their detection is dependent on the sensitivity of the molecular method used and the source of DNA.¹⁵  Studies using sensitive methods in large patient populations indicate that MPL-mutated ET and PMF represent about 10% and 15%, respectively, of the respective JAK2 (V617F)–negative conditions.^12-14 

MPL exon 10 mutations affecting the juxtamembrane (W515L/K/A/R) or the transmembrane (S505N) domains result in a ligand-independent receptor activation with constitutive activation of downstream JAK-STAT signaling.^6,16,17  Expression of MPL mutations in a murine bone marrow transplant model induced a myeloproliferative disorder characterized by leukocytosis, thrombocytosis, extramedullary hematopoiesis, bone marrow fibrosis, and splenomegaly.⁶  In the few studies that investigated the phenotypic associations of MPL mutations and their prognostic relevance, only minor relationships were found.^9,10,12 

We previously used high-resolution single neucleotide polymorphism (SNP) microarrays to detect chromosomal aberrations, such as loss of heterozygosity (LOH) and somatic copy number changes, in patients with MPN.¹⁸  Uniparental disomy (UPD) of chromosome 9p (9pUPD) involving JAK2 and UPD of chromosome 1p (1pUPD) involving MPL were found to be relatively common aberrations that are responsible for transition from heterozygosity to homozygosity for the JAK2 or MPL mutation, respectively.¹⁸  We also showed that the JAK2 (V617F)–mutant allele burden has a clinical effect in PV, representing in particular a risk factor for progression to myelofibrosis.^19-21 

In this study, we reasoned that the MPL-mutant allele burden, rather than the mere presence or absence of MPL mutations and their subtype, might have a clinical effect in MPL-mutated ET and PMF. Therefore, we did a mutation analysis of MPL exon 10 in patients with ET and PMF, assessed the mutant allele burden by using a deep sequencing approach, and then analyzed the clinical correlates of the MPL-mutant allele burden.

This study was approved by the Institutional Ethics Committee (Comitato di Bioetica, Fondazione Istituto di Ricovero e Cura a Carattere Scientifico (IRCCS) Policlinico San Matteo, Pavia, Italy). The procedures followed were in accordance with the Helsinki Declaration of 1975 (as revised in 2000), and samples were obtained after patients provided written informed consent.

The study design is schematically described in supplemental Figure 1. The main component was a prospective observational cohort study that included 892 patients with Ph− MPN other than PV diagnosed and followed at the Department of Hematology Oncology, Fondazione IRCCS Policlinico San Matteo and University of Pavia, Italy, between 2002 and 2012. Specifically, our cohort included 661 patients with ET, 187 patients with PMF, and 44 patients with post-ET myelofibrosis. The prospective observation started from the evaluation of MPL molecular status: 434 patients (49%) entered the study at diagnosis, and 458 (51%) were enrolled during follow-up.

Diagnosis of ET and PMF was in accordance with the World Health Organization (WHO) criteria.^22,23  Post-ET myelofibrosis was diagnosed according to the criteria of the International Working Group of Myelofibrosis Research and Treatment (IWG-MRT)²⁴ ; evolution into acute myeloid leukemia (AML) was defined according to the WHO criteria.²³  For the assessment of bone marrow fibrosis, paraffin sections were stained with Gomori’s silver impregnation technique, and fibrosis was assessed semiquantitatively following the European consensus guidelines.²⁵  Thrombotic events were defined as described in detail elsewhere.²⁶  Patients were fundamentally treated according to the recommendations of the European LeukemiaNet.²⁷ 

As described in supplemental Figure 1, in order to define the effect of MPL-mutant allele burden on clinical phenotype, we analyzed the Pavia cohort of 43 MPL-mutated patients together with a second cohort of 19 patients observed at the University of Florence, Italy, which was previously reported.¹⁰ 

Peripheral blood granulocytes and CD3⁺ T lymphocytes were isolated as previously described.²⁸  DNA extraction was performed by using the Puregene Blood DNA isolation kit (Qiagen, Hilden, Germany) according to the manufacturer’s procedures. The granulocyte JAK2 (V617F) mutation burden was assessed by using a quantitative polymerase chain reaction (qPCR) –based allelic discrimination assay on a Rotor-Gene 6000 real-time analyzer (Qiagen) and applying the standard curve method.¹⁹  The sensitivity of this discrimination assay is approximately 0.2% mutant alleles.²¹ 

MPL mutation scanning was performed on granulocyte and T-lymphocyte genomic DNA by using a high-resolution melt (HRM) assay, and samples found to be MPL mutated at HRM analysis were further characterized by using direct sequencing, as previously described.¹³ MPL mutation analysis was performed at diagnosis or within 6 months from diagnosis in 30 of 62 patients. The remaining 32 patients were evaluated during follow-up (median time from diagnosis, 4.3 years; range, 0.5 to 19.5 years).

Primers used for the HRM screening were adapted for the multiplexed 454 GS-FLX ultramassive sequencing of granulocyte genomic DNA from MPL-positive patients, as described in detail in supplemental Methods and supplemental Table 1.

MPL exon 10–amplified regions were purified by MinElute columns (Qiagen, Valencia, CA) to remove the fragments shorter than 70 bp and were quantitated by PicoGreen DNA Quantitation Kit (Life Technologies, Monza, Italy). Thirteen amplicons were pooled and then amplified by emulsion PCR as required by the manufacturer’s instructions (Roche Diagnostics, Monza, Italy). Libraries were recovered by isopropanol emulsion breaking and enriched for positive reaction beads. Each enriched pooling sample was separately loaded onto one-eighth of the PicoTiterPlate and was sequenced according to the 454 GS-FLX Titanium protocol (Roche Diagnostics). Raw reads from the GS-FLX sequencing were demultiplexed by using Roche’s proprietary “sfffile” and “sffinfo” utilities and were mapped against a reference sequence of exon 10 of the MPL gene (isoform ENST00000372470 from Ensembl release 69 assembly) by using the Amplicon Variant Analyzer v. 2.6 (Roche Diagnostics). For each sequencing pool, a quality check of the percentage of sequences correctly mapped and carrying the appropriate multiplexing tag was performed. Afterward, by using a script developed in-house, we estimated the mutation burden in all the nucleotides of the reference for each sample. Variations in the sequence were then converted to the corresponding amino acid substitutions according to the selected reference isoform. Read alignments of samples with a very low mutation burden (<1%) were further manually inspected to exclude sequencing and alignment errors.

Granulocyte and T-lymphocyte DNA were genotyped for the rs760567, rs2073025, rs2297634, and rs1801574 loci by using HRM, as described in detail in supplemental Methods and supplemental Table 2. The number of copies of the MPL gene on chromosome 1p was determined by using TaqMan qPCR and comparing MPL and the haploid SRY gene; the housekeeping diploid ALB gene was used to normalize the data, as described in detail in supplemental Methods and supplemental Table 3.

Circulating CD34⁺ cells were enumerated by flow cytometry using a single-platform assay as previously described.²⁸ 

Numeric variables were summarized by their median and range. Categorical variables were described by count and relative frequency of each category. Hypothesis testing was carried out with a nonparametric approach. The Spearman coefficient was used to test for correlation between numerical variables. The Wilcoxon rank sum test was applied to the comparison of numerical variables between two groups, and the association between categorical variables (two-way tables) was investigated by means of Fisher’s exact test. All tests were two-tailed, and P values were considered significant when lower than .05. Microsoft Office Excel (Microsoft Corporation, Redmond, WA) and Stata 11.2 (StataCorp, Lakeway, TX) were used for data management and statistical analysis. The cumulative incidence of thrombotic complications and the cumulative incidence of leukemic evolution were estimated with a competing risk approach according to the Kalbfleisch-Prentice method.²⁹  Death in the absence of the event of interest (thrombosis or leukemia) was considered as a competing event. Associations between JAK2 or MPL mutations status and marrow fibrosis were investigated by multivariate logistic regression analysis.

The distribution of JAK2 and MPL mutations according to MPN subtype is reported in Table 1. Somatic MPL mutations, detected in granulocytes but not in T lymphocytes, were observed in 26 (4%) of 661 patients with ET, 10 (5%) of 187 with PMF, and 7 (16%) of 44 with post-ET myelofibrosis. The mutation status was significantly different among the three MPN subtypes (P = .021); in particular, Fisher’s exact test revealed significant differences in terms of MPL-mutated genotypes between ET and post-ET myelofibrosis (P = .004) and between PMF and post-ET myelofibrosis (P = .033), but no difference was observed between ET and PMF (P = .6).

Demographic and clinical characteristics at diagnosis of patients with ET, PMF, and post-ET myelofibrosis according to their genotype are reported in supplemental Tables 4, 5, and 6, respectively. In ET, the three subgroups (MPL-mutated, JAK2-mutated, and JAK2/MPL-unmutated patients) showed significant differences in terms of clinical phenotype. In particular, MPL-mutated patients had lower hemoglobin levels (P = .0001) and lower platelet counts (P = .0001), and JAK2-positive patients had higher hemoglobin levels (P = .0001), higher WBC counts (P = .0001), and higher incidence of thromboembolic complications (P = .0001).

The cumulative incidence of thrombosis was estimated with death as a competing risk. In ET, the 5-year cumulative incidence of thrombosis was 9.2% (95% CI, 1.5% to 25.6%) in MPL-mutated patients, 13.5% (95% CI, 9.6% to 18%) in JAK2-mutated patients, and 8.4% (95% CI, 4.7% to 13.4%) in JAK2/MPL-unmutated patients. JAK2-positive patients had a trend toward higher incidence of thrombotic events than JAK2/MPL-unmutated patients (P = .05).

In PMF, the 5-year cumulative incidence of thrombosis was 20% (95% CI, 0.9% to 57.3%) in MPL-mutated patients, 6.8% (95% CI, 2.7% to 13.4%) in JAK2-mutated patients, and 11.8% (95% CI, 4.8% to 22.2%) in wild-type patients. When comparing the three subgroups by pairs, we did not identify significant differences, possibly because of the low number of events.

In post-ET myelofibrosis, the 5-year cumulative incidence of thrombosis was 39.4% (95% CI, 12.3% to 66%) in JAK2-mutated patients, and no thrombotic event was observed in MPL-mutated and JAK2/MPL-unmutated patients.

During follow-up, 21 patients (2.3%) progressed to secondary AML, including 5 patients (0.7%) with ET, 12 (6.4%) with PMF, and 4 (9%) with post-ET myelofibrosis.

The cumulative incidence of leukemia was estimated with death as a competing risk. In ET, the 5-year cumulative incidence of leukemia was null in MPL-mutated and JAK2/MPL-unmutated patients and 1.2% (95% CI, 0.3% to 3.1%) in JAK2-mutated patients, without any significant difference among the 3 genotypic subgroups. In PMF, the 5-year cumulative incidence of leukemia was null in MPL-mutated and JAK2/MPL-unmutated patients and 9.5% (95% CI, 4.3% to 17%) in JAK2-mutated patients. JAK2-mutated patients showed a higher incidence of leukemic evolution in comparison with both MPL-mutated patients (P = .0019) and JAK2/MPL-unmutated patients (P = .013). In post-ET myelofibrosis, no leukemic evolution was observed in MPL-mutated patients; the 5-year cumulative incidence of AML was 12.7% (95% CI, 2% to 33.4%) in JAK2-mutated patients and 10% (95% CI, 0.5% to 35.8%) in JAK2/MPL-unmutated patients. JAK2-mutated patients showed a trend toward higher incidence of leukemic evolution in comparison with MPL-mutated patients (P = .051).

The whole cohort was observed for a median follow-up from diagnosis of 4.5 years (range, 0 to 21.1 years). Median overall survival (OS) was not reached in ET and post-ET myelofibrosis and was equal to 15.2 years in PMF. The molecular status (JAK2-mutated vs MPL-mutated vs JAK2/MPL-unmutated genotype) did not affect OS in patients with ET and post-ET myelofibrosis, but it did affect OS in patients with PMF (P = .011). In particular, in PMF, the median OS was 9.1 years in MPL-mutated patients, 12.9 years in JAK2-mutated patients, and not reached in JAK2/MPL-unmutated patients (supplemental Figure 2).

To assess the effect of MPL allele burden on clinical phenotype, we evaluated 62 MPL-mutated patients from Pavia (43 patients) and Florence (19 patients). Of these 62 patients, 40 patients were affected with ET, 14 with PMF, and 8 with post-ET myelofibrosis. Types of MPL mutations are reported in Table 2: MPL (W515L), MPL (W515K), and MPL (W515A) were the most common changes detected in 39, 11, and 7 patients, respectively. Four patients carried two MPL mutations, and one patient carried both MPL and JAK2 mutations.

The MPL mutant allele burden estimated by means of deep sequencing ranged from 0.8% to 94.7%, and the Kruskal-Wallis test showed a different distribution among different MPN subtypes (P = .0031), as shown in Figure 1. The median allele burden was significantly lower in ET compared with PMF (32.8% vs 59.4%; P = .0037) or post-ET myelofibrosis (32.8% vs 58.2%; P = .02), but there was no difference between PMF and post-ET myelofibrosis. In addition, the number of patients with a mutant allele burden greater than 50% was significantly different (P = .006) in the three MPN subtypes, being 8 (20%) of 40 patients in ET, 9 (64%) of 14 patients in PMF, and 4 (50%) of 8 patients in post-ET MF.

To establish whether the mutant allele burden was specifically related to subtypes of MPL mutation, we examined the most common MPL changes (Figure 2). The median allele burden was 32.9% (range, 4.7% to 94.4%) in the MPL (W515L) group, 56.2% (range, 2.0% to 94.7%) in the MPL (W515K) group, and 49.9% (range, 31.1% to 92.1%) in the MPL (W515A) group, with no significant difference among these median values.

We then analyzed the relationships between MPL-mutant allele burden and bone marrow fibrosis, circulating CD34⁺ cell count, or disease duration. The MPL-mutant allele burden was significantly higher (P < .0001) in patients with bone marrow fibrosis (median allele burden, 53.1%) than in those without fibrosis (median allele burden, 15.6%), as shown in Figure 3. The effect of mutation burden on fibrosis remained statistically significant (P = .0001), even when categorizing patients according to grade of fibrosis (Figure 3).²⁵ 

The Spearman rank test showed a correlation between MPL-mutant allele burden and circulating CD34⁺ cells (ρ = .518; P = .0001). More specifically, patients with a mutant allele burden greater than 50% had a significantly higher number of circulating CD34⁺ cells than those with an allele burden below or equal to 50% (median number, 46.9 vs 3.6 × 10⁶/L; P = .0028), as reported in Figure 4. A significant relationship was also found between MPL-mutant allele burden and disease duration (Spearman rank test ρ = .313; P < .013).

For the purpose of this analysis, patients with ET or post-ET myelofibrosis belonging to the Pavia cohort (Table 1) were considered as a phenotypic continuum in which the disease initially presents as ET and may progress to myelofibrosis over time.

A preliminary analysis comparing JAK2-mutated, MPL-mutated, and JAK2/MPL-unmutated patients (model 1 in Table 3) showed that MPL-mutated patients with ET had a higher risk of having marrow fibrosis than the JAK2 (V617F)–mutated ones (odds ratio, 3.22; P = .017). No significant difference was found between MPL-mutated and JAK2/MPL-unmutated ET patients.

To determine the association between mutation burden and marrow fibrosis, we then performed a multivariate logistic regression analysis on MPL-mutated and JAK2-mutated patients, with presence of fibrosis as outcome and age at diagnosis, type of mutation (MPL vs JAK2 mutation), and mutation burden as covariates. As shown in Table 3 (models 2 and 3), there was no significant effect of mutation subtype (MPL or JAK2) when adjusting by mutant allele burden, although a trend toward a significantly higher odds ratio for the MPL mutation was observed in model 3. The most striking finding was the strong association between elevated mutant allele burden (greater than 50%) and the presence of marrow fibrosis, irrespective of the mutation involved.

The MPL gene is located on chromosome 1p34. Direct sequencing allowed us to identify the intronic T→C rs839995 SNP located 70 bases before the beginning of MPL exon 10. Informative patients with MPL mutation burden greater than 50% had a homozygous TT or CC genotype in granulocytes and a heterozygous TC genotype in T lymphocytes (data not shown), suggesting 1pLOH.

To unequivocally demonstrate 1pLOH, we studied 8 patients with an MPL-mutant allele burden greater than 75% and adequate DNA supply. We developed HRM assays for the analysis of the two telomeric SNPs rs760567 and rs2073025 mapping on chromosomes 1p36.32 and 1p34.3, respectively, and the two centromeric loci rs2297634 and rs1801574 located at 1p22.1. In myeloproliferative neoplasms, granulocytes typically belong to the malignant clone and T lymphocytes normally do not.⁴  Thus, 1pLOH is defined as a condition in which a homozygous genotype is detected in granulocytes, and a heterozygous one is found in T lymphocytes. Detection of a heterozygous genotype in both granulocytes and T lymphocytes indicates the absence of LOH, but detection of a homozygous genotype in both cell types means that the patient is not informative. The results of our analysis of these polymorphic markers are shown in Figure 5: although one patient was not informative, various patterns of LOH of chromosome 1p34 and its telomeric sequences were observed in granulocytes from the remaining patients.

To evaluate whether 1pLOH resulted from deletion of the telomeric portion of one chromosome 1 or from mitotic recombination events, we used qPCR to determine the number of copies of MPL. Only one copy of DNA for the MPL locus would be expected in case of deletion, whereas two copies should be detected in case of mitotic recombination. To discriminate between these two possibilities, we determined the number of copies of MPL in granulocytes of patients with 1pLOH. Numbers of copies of MPL in healthy participants and in T cells from patients with 1pLOH were considered as two-copy DNA controls, whereas the number of copies of the haploid gene SRY on chromosome Y was used as a one-copy DNA control. As shown in Figure 5, no significant MPL copy number variation was found between granulocytes and T lymphocytes from patients with 1pLOH (P = .54), as well as between patients and healthy participants (P = .12), but significant differences were found with respect to the SRY gene (P < .001). Taken together, these results clearly indicate copy-neutral LOH (CN-LOH) of chromosome 1p in the patients studied.³⁰ 

Most of MPL mutations identified in this study affected the amino acid W515 (W515L/K/A/R), which belongs to the RWQFP juxtamembrane domain of the MPL receptor (ie, to a domain that prevents spontaneous activation of the receptor).³¹  We also found a somatic mutation affecting the transmembrane domain (S505N); this mutation, originally identified in hereditary thrombocytosis and later observed in sporadic MPN, may promote or facilitate receptor dimerization.^9,16,17  Four patients had a second rare substitution (S505C, V501A, and Q516E), likely devoid of significant transforming activity.¹⁷ 

In agreement with previous observations,¹²  we found that compared with JAK2 (V617F), MPL mutation had little distinctive phenotypic and clinical effect. When comparing JAK2-mutated with MPL-mutated ET patients, the latter group was found to have lower hemoglobin levels, as expected.³²  With respect to clinical outcome, MPL-mutated patients had lower incidence of thrombotic complications and lower risk of leukemic evolution compared with JAK2-mutated patients.

The most remarkable findings of this study are that the MPL-mutant allele burden is highly variable in MPL-mutated myeloproliferative neoplasms and that higher mutation burdens originate from acquired CN-LOH of chromosome 1p and are associated with marrow fibrosis.

Acquired CN-LOH represents a common molecular mechanism of disease in myeloid malignancies.³⁰  A paradigmatic example is acquired CN-LOH of chromosome 9p, responsible for the transition from heterozygosity to homozygosity for the JAK2 (V617F) mutation and in turn for high JAK2 (V617F)–mutant allele burden.^4,33  A recent study has shown recurrent acquisition of JAK2 (V617F) homozygosity in both PV and ET patients, with different homozygous subclones being found in individual patients.³⁴  This is consistent with the notion that abnormal mitotic events occur frequently in hematopoietic cells from JAK2 (V617F)–mutated patients and give rise to multiple subtypes of acquired CN-LOH of chromosome 9p. The selective advantage of a dominant homozygous subclone, which might reflect additional genetic or epigenetic changes, would drive erythrocytosis and transition from ET to PV. Over time, a high mutation burden almost inevitably leads to secondary myelofibrosis, as we previously demonstrated.^19,21 

Acquired CN-LOH of chromosome 1p involving MPL was previously described in two patients with refractory anemia with ring sideroblasts associated with marked thrombocytosis carrying the MPL (W515L) mutation³⁵  and in a patient with PMF with the same mutation.³⁶  In addition, a few MPN patients with high MPL-mutant allele burden and/or evidence of homozygous MPL mutation have been described.^9,11  However, the clinical effects of acquired CN-LOH of chromosome 1p and high MPL-mutant allele burden have not been investigated systematically in patients with MPN.

In the 62 MPL-mutated patients in this study, the granulocyte mutant allele burden ranged from 1% to 95% and was significantly higher in patients with PMF or post-ET myelofibrosis than in those with ET, confirming previous findings by Jones et al.³⁷  Unlike Jones et al, however, we did not find any significant difference in mutation load between patients carrying MPL (W515K) and those with MPL (W515L). Taking into account statistical variability, a mutant allele burden greater than 50% would by definition indicate the existence of at least a subclone of cells that are homozygous for the mutation. In myeloid malignancies, this subclone is typically a result of mitotic events that generate acquired CN-LOH.³⁰  To demonstrate this, we studied 8 patients with a mutant allele burden greater than 75%, unambiguously indicating the existence of a dominant homozygous clone. In all the informative patients, we indeed detected acquired CN-LOH of chromosome 1p, consistently involving the location of MPL on chromosome 1p34 (Figure 5).

The association between high MPL-mutant allele burden and marrow fibrosis was supported by different univariate and multivariate analyses. In particular, when we did a multivariate logistic regression analysis of patients with ET or post-ET myelofibrosis considered as a phenotypic continuum, the MPL-mutant allele burden was found to be the most significant risk factor for marrow fibrosis (Table 3). Our observation that MPL-mutated ET patients are more likely to directly progress to post-ET myelofibrosis than JAK2 (V617F)–mutated ET patients has a coherent pathophysiological explanation. In fact, acquired CN-LOH of chromosome 9p in a JAK2 (V617F)–mutated ET patient is expected to first determine a transition from ET to PV³⁴  and later on, progression to post-PV myelofibrosis.²¹ 

Although the relationship between high MPL-mutant allele burden and marrow fibrosis is strong, the underlying causal mechanisms remain to be defined. The thrombopoietin receptor is expressed not only in megakaryocytes but also in hematopoietic stem cells and plays a crucial role in maintaining the quiescence of the latter.³⁸  In PMF, MPL (W515L/K) mutations occur in multipotent hematopoietic stem cells and induce spontaneous megakaryocyte differentiation.³⁹  In a mouse model, the expression of the MPL (W515A) mutation induced a myeloproliferative phenotype with severe marrow fibrosis.⁴⁰  A number of studies have suggested that the abnormal megakaryocytes of patients with MPN release profibrotic growth factors in the bone marrow,⁴¹  and more recently, abnormal proplatelet formation has been described in MPN patients with marrow fibrosis.⁴²  Thus, if somatic mutations of MPL involving the W515 residue are relevant to the abnormal proliferation and differentiation of megakaryocytes, these abnormalities are expected to be more severe in clonal cells that are homozygous compared with those that are heterozygous for the mutation.

In conclusion, acquired CN-LOH of chromosome 1p, involving location of MPL, likely represents a mechanism of disease progression in MPL-mutated MPN, since it generates a subclone that is homozygous for the MPL mutation and expands, leading to a high mutant allele burden. This mechanism may specifically play a role in the progression from MPL-mutated ET to post-ET myelofibrosis. With increasing adoption of deep sequencing in diagnostic laboratories, assessing the MPL-mutant allele burden may become part of the diagnostic work-up of MPN, allowing clinicians to identify ET patients at high risk of fibrotic transformation.

This study was supported by a grant from Associazione Italiana per la Ricerca sul Cancro (Milan, Italy), Special Program Molecular Clinical Oncology 5x1000, to Associazione Italiana per la Ricerca sul Cancro-Gruppo Italiano Malattie Mieloproliferative, project number 1005; by grant number RBAP11CZLK from “Fondo per gli investimenti della ricerca di base” (A.V. and M.C.); by grant number 2010NYKNS7 from “Programmi di Ricerca Scientifica di Rilevante Interesse Nazionale” (A.V. and M.C.); by grant number GR‐2010‐2312855 from the Italian Ministry of Health and a grant award from the Italian Society of Experimental Hematology (E.R.); and by a grant from Regione Toscana, “Programma per la ricerca regionale in materia di salute 2009” (P.G.). A detailed description of the “Associazione Italiana per la Ricerca sul Cancro-Gruppo Italiano Malattie Mieloproliferative” project is available at http://www.progettoagimm.it.

Contribution: E.R., D.P., and M.C. conceived this study, collected and analyzed data, and wrote the manuscript; P.G., I.C., E.S., C.A., F.P., and A.V. collected clinical data; R.B., M.S., A.Pi., and G.D.B. performed deep sequencing investigations; D.P. and C.M. performed HRM, LOH, and MPL copy number analyses; A.Pa. and G.R. did PCR analyses; V.F., E.F., and C.P. did statistical analyses; E.B. studied bone marrow fibrosis.

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

Correspondence: Elisa Rumi, Department of Hematology Oncology, Fondazione IRCCS Policlinico San Matteo, Viale Golgi 19, 27100 Pavia, Italy; e-mail: elisarumi@hotmail.com; and Mario Cazzola, Department of Hematology Oncology, Fondazione IRCCS Policlinico San Matteo, Viale Golgi 19, 27100 Pavia, Italy; e-mail: mario.cazzola@unipv.it.