Influence of JAK2V617F allele burden on phenotype in essential thrombocythemia

Abstract

Background Fifty to sixty percent of patients with essential thrombocythemia harbor the JAK2^V617F mutation. The impact of this mutation on clinical phenotype is still debated. The aim of this study was to evaluate possible correlations between JAK2^V617F mutant allele burden and both clinical presentation and hematologic abnormalities in essential thrombocythemia patients.Design and Methods In this single-center retrospective study, JAK2^V617F allele load was measured by sensitive quantitative reverse transcriptase polymerase chain reaction (RT-PCR) in the granulocytes of 260 patients diagnosed as having essential thrombocythemia according to WHO criteria.Results Median V617F allele burden in patients with the mutation (n=165, 63.4%) was 24%, ranging from 1% to 87%; an allele burden greater than 51% was found in 5% of the patients. Older patients presented progressively higher percentages of the V617F allele. Signs of stimulated erythropoiesis and myelopoiesis, as well as higher PRV-1 levels, were found in patients with the mutation, but no linear correlation with load of mutant allele could be ascertained; on the other hand, the frequency of patients with erythropoietin-independent erythroid colonies progressively increased depending on mutant allele load. Splenomegaly and microvessel symptoms were significantly more represented among patients with greater than 50% and 25% JAK2^V617F allele burden, respectively. Increasing mutant allele load correlated with higher frequency of arterial thrombosis at diagnosis, as confirmed also in multivariate analysis; the relative risk was 3.0 (95% CI 1.3–6.8; p=0.01) in patients having a greater than 25% mutant allele burden.Conclusions The JAK2^V617F mutant allele burden contributes to determining the clinical phenotype in patients with essential thrombocythemia.

Introduction

An acquired mutation in exon 14 of the Janus tyrosine kinase 2 gene (JAK2) leading to a valine→phenylalanine change at position 617 (JAK2) in the JH2 pseudokinase domain,1–4 has been described in Philadelphia-negative chronic myeloproliferative disorders (MPD);5,6 it can be found in approximately 95% cases of polycythemia vera (PV) and in approximately 50–60% cases of of essential thrombocythemia (ET) or primary myelofibrosis (PMF).7 A role for the gain-of-function8 of mutant JAK2 in the pathogenesis of MPD is supported by in vitro studies, which showed growth factor independence and enhanced JAK-STAT signaling in cells transfected with the mutant allele, and by the results of in vivo animal models.2,9,10 The JAK2 mutation can be harbored in the heterozygous or homozygous state, the latter deriving from mitotic recombination;1–4 homozygosity for this mutation is found in about 30% of PV and PMF patients but it is rare (2–4%) in ET.

In patients with the different MPD, the JAK2^V617F mutation has been variably associated with higher indices of erythropoiesis,11,12 unchanged or decreased platelet counts,11,12 greater occurrence of thrombosis,4,13 bone marrow fibrosis or cytoreductive treatment,4 older patients’ age,3,4,14,15 longer disease duration or poorer survival.16 In particular, in the case of ET, it has been suggested that JAK2^V617F mutation imparts patients a PV-like phenotype;11 furthermore, the risk of thrombosis has been found to be increased in ET patients with the mutation,17 particularly in those harboring it in the homozygous state.13 It is still unclear, however, how the presence of the JAK2^V617F mutation can help the clinician in identifying categories of patients with unique clinical and prognostic characteristics. One of the reasons underlying current uncertainties is that genomic techniques currently employed for detection of the JAK2^V617F mutation do not produce quantitative information about the level of the mutant allele, which might represent an important variable influencing disease presentation. The levels of expression of genes such as PRV-1 and NFE-2, the number of CD34 cells in the peripheral blood, and the concentration of alkaline leukocyte phosphatase (ALP), have all been shown to be dependent on mutant allele burden.18–20 A prospective study in PV patients identified those with greater than 75% V617F allele as having more pronounced signs of myeloproliferation, a higher incidence of thrombosis and as being at greater risk of needing chemotherapy.21

In the current study we explored whether there was an association between burden of the JAK2 mutation and clinical presentation in a retrospective series of 260 patients with ET. We correlated the levels of JAK2^V617F allele load in granulocytes with clinical and hematologic characteristics and with some markers typical of MPD, which included X-chromosome inactivation pattern (XCIP), PRV-1 expression level and erythropoietin-independent erythroid colony (EEC) formation. Additionally, quantitative analysis of the JAK2 mutation was performed serially in 35 patients during follow-up to evaluate any fluctuations in mutant allele load.

Design and Methods

Patients

This study involved 260 patients with a diagnosis of ET according to WHO criteria,22 115 of whom had been previously included in a multicenter GIMEMA study.13 The design of this study was that of a retrospective, single center study; the only criteria necessary for inclusion were a diagnosis of ET satisfying the WHO criteria, complete chart records, and availability of a stored or fresh granulocyte DNA sample for JAK2 mutational analysis. If patients had been first genotyped during follow-up, only those not receiving cytoreductive drugs were considered. All diagnoses were re-evaluated using the original clinical records and bone marrow biopsies by an ad hoc panel of four hematologists and two pathologists. The study was approved by the local Ethical Committee and informed consent was obtained from all patients included.

Clinical outcomes

The presence of splenomegaly was objectively determined on the basis of an ultrascan demonstrating an organ with a longitudinal diameter ≥12 cm. Major thrombotic events, whether arterial or venous (ischemic stroke, transient ischemic attack, myocardial infarction, angina pectoris, peripheral arterial thrombosis, deep vein thrombosis including cerebral and splanchnic venous thromboses, and pulmonary embolism) were recorded only if they happened after the diagnosis or in the preceding 2 years23 and were objectively documented.24 Only major bleeding events were considered, and included fatal hemorrhages, non-fatal intracranial bleeding or any other hemorrhages requiring surgery or causing a hemoglobin reduction of ≥2 g/dL and/or needing transfusion of two or more red blood cell units. The presence of symptoms due to microvessel disorder (headache, acral paresthesia, erythromelalgia, transient neurological and visual disturbances) was routinely investigated and recorded in the chart records by referring physicians, but for the purposes of this study they were considered only in the case that they had been described by the patient as non-occasional, of recent onset, and often ameliorated by aspirin.

Analysis of the JAK2V617F mutation

Peripheral blood granulocytes were separated by differential centrifugation over a Ficoll-Paque gradient, and contaminating red cells were removed by hypotonic lysis; DNA was purified using the QIAmp DNA blood kit (Qiagen, GmbH, Germany), and quantified with NanoDrop technology (Wilmington, DE, USA). All patients were routinely genotyped for the JAK2 mutation by an allele-specific (ASO) polymerase chain reaction (PCR), using 75 ng granulocyte DNA. To evaluate whether the mutation was carried in the homozygous or heterozygous state, PCR products were digested with BsaXI restriction enzyme (New England Biolabs, Hitchin, UK) as described by Baxter et al.1 The mutant allele burden was measured by a quantitative real time (QRT)-PCR assay, using 20 ng genomic DNA. PCR amplification and detection were performed on an ABI Prism 7300 analyzer (Applied Biosystems) using the following cycling conditions: 10 minutes at 95°C followed by 40 cycles of 15 s at 95°C and 60 s at 60°C. Primers flanking the mutant region (forward primer 5′-AAGCTTTCTCACAAGCATTTGGTTT-3′; reverse primer 5′-AGAAAGGCATTAGAAAGCCTG TAGTT-3′) were employed together with Taqman probes which were specific for either the wild type (VIC-5′-TCTC-CACAGACACATAC-3′MGB) or the mutant JAK2 allele (FAM-5′-TCCACAGAAACATAC-3′-MGB). All samples were analyzed in triplicate and the amount of JAK2 allele was calculated by comparison with serial dilutions of mutant DNA, obtained from a PV patient with 100% mutant alleles, and wild-type DNA from healthy subjects. The mean of triplicate ΔCT determinations (C_T^JAK2V617F−C_T^JAK2WT) was used to calculate the percentage of mutant alleles. Positive and negative controls were included in each assay; inter- and intra-assay variation was 3% and 5%, respectively.

Other biological markers

The level of expression of neutrophil PRV-1, and of GAPDH as the housekeeping gene, was quantified using the Taqman one-step procedure, as described previously.25 The PRV-1 cycle threshold (C_T) ratio was calculated from the mean value of triplicate PRV-1 C_T determinations divided by the mean of triplicate GAPDH C_T; the PRV-1 C_T ratio in healthy blood donors (n=30) was 1.37±0.09; PRV-1 over-expression was defined in the presence of a PRV-1/GAPDH C_T ratio ≤1.17.

In vitro assays for EEC was performed by plating 2.5 × 10 peripheral blood mononuclear cells in Methocult H4531 medium (StemCell, Vancouver, Canada) in the absence of added erythropoietin. Concurrent plates containing optimal amounts of erythropoietin (2 U/mL) were prepared as control cultures. Dishes were incubated in a humidified atmosphere of 5% CO₂ at 37°C, and hemoglobinized colonies were enumerated on day 14 by standard criteria.

XCIP evaluation was performed at the HUMARA locus in female patients aged less than 60 years old using a PCR assay and enzymatic digestion with RsaI and HpaII, as described elsewhere.26 Immunomagnetically purified CD3 T-lymphocytes were used as control cells.

Statistical analysis

Data were processed using the GraphPad InStat (GraphPad Software, San Diego, CA, USA) and the SPSS ( Statsoft Inc, Tulsa, OK, USA) software. We used the χ or Fisher’s exact test (two by two table) or χ test for trend (larger contingency table), as appropriate, to compare categorical variables among the groups which had been categorized according to mutation load. The analysis of continuous variables among the groups was performed using the Mann-Whitney U test or Kruskal-Wallis test with Dunn’s method in the case of multiple comparisons. Spearman’s rank non-parametric correlation test was performed to analyze correlations between JAK2 allele burden and hematologic parameters. Unconditional logistic regression models with backward stepwise analysis were used to identify factors that were associated with an increased probability of having thrombotic events. A p value of less than 0.05 was considered to indicate statistical significance; all tests were two-tailed.

Results

Characteristics of the patients

We studied 260 subjects in whom a diagnosis of ET was made based on WHO criteria; their clinical characteristics and hematologic parameters at diagnosis, as well as major clinical events associated with disease, are reported in Table 1. Their median age was 52 years (range, 16–93). Of these 260 subjects, 195 (75%) were females, making the proportion of females higher than that in most published series of ET; this was due to trivial, uncontrolled reasons, and was not the result of a selection. At diagnosis, 72 patients (28%) had splenomegaly and in eight patients (3%) the spleen diameter was greater than 15 cm. Twenty-three patients (9%) reported aquagenic pruritus and 19 (7%) presented with systemic symptoms. One hundred and sixty-one patients (62%) required cytoreductive treatment, either because they were high-risk patients, according to current criteria,27 or because signs and/or symptoms of myeloproliferation needed to be controlled. Fifty-six patients (21%) had major thromboses, 26 of which were detected at diagnosis or had occurred within the preceding 2 years, and 25 during follow-up; five patients had a second thrombotic event during follow-up. A detailed description of the thrombotic events is presented in Table 1. Microvessel symptoms were reported by 73 patients (28%). Hemorrhages affected 15 patients (6%), and in all cases were gastrointestinal bleeds; seven of these occurred at the time of diagnosis. The median follow-up was 36 months (range, 1–120); during this period four patients died of causes unrelated to their hematologic disease, one patient developed myelofibrosis and two still-alive patients developed solid tumors.

JAK2V617F mutational burden

The median time from diagnosis to blood sampling for mutational analysis was 2.0 years (range, 0–3). DNA samples for mutational analysis were collected at diagnosis in 102 patients (39%), within 1 year from diagnosis in 80 (31%), between 1 and 2 years from diagnosis in 50 (19%), while the remaining 28 patients (11%) were genotyped from 2 to 3 years after diagnosis. According to ASO PCR, 157 patients (60%) carried JAK2^V617F mutations, of whom 151 (96%) were considered heterozygous and six (4%) homozygous, accounting for 58% and 2% of the entire patient population, respectively. However, by using QRT-PCR, the frequency of patients with the mutation rose to 63% (n=165) and the percentage of those displaying >51% mutant alleles was 5%. The median JAK2 mutant allele ratio in these 165 patients with the mutation was 24% (range, 1–87%), which is significantly lower than the level we previously found in PV patients (n=173, 52% range 1–100%)21 or in patients with PMF (n=55; 8%: range, 1–100%) or secondary forms of myelofibrosis (n=20; 61%: range, 1–100%) (p<0.001 for all) (Figure 1). JAK2^V617F levels ranged from 1% to 25% in 101 patients (39%), from 26% to 50% in 51 patients (20%), from 51% to 75% in 10 patients (4%) while only three patients (1%) had JAK2^V617F levels greater than 75% (Figure 1).

Correlation of JAK2V617F burden with hematologic parameters

Patients with the JAK2^V617F mutant allele had significantly higher leukocyte counts (p=0.02), hemoglobin concentration (p<0.0001), hematocrit (p<0.0001), alkaline phosphate level (p=0.04) and lower platelet count (p=0.03), as compared to wild-type patients (Table 1). However, there was no clear dose-dependent correlation between these hematologic parameters, considered either at diagnosis or at the time of blood sampling, and the burden of JAK2 mutant allele (not shown in detail).

EEC formation was detected in 52% of evaluated patients (n=113), accounting for 40% and 60% of JAK2 wild-type and mutant patients, respectively (p=0.05). We observed a significant correlation between EEC and load of JAK2 mutant allele (p=0.0019) (Table 2); indeed, all 13 patients with more than 50% mutant alleles displayed EEC formation as compared to half of those with 1–25% V617F alleles. XCIP analysis was performed in 122 patients, and was informative in 104 (85%); 11 females were excluded because of constitutional skewing and seven because of homozygosity at the HUMARA locus. Sixty-five per cent of informative females had clonal hematopoiesis. The frequency of patients with clonal hematopoiesis was similar among those with wild-type (68%) or mutant JAK2 (63%); furthermore, there was no difference in JAK2 mutant allele load between patients with clonal or polyclonal hematopoiesis (13% in both groups). Finally, we found no significant correlation between PRV-1 expression level and JAK2 allele burden, although the level of PRV-1 was significantly lower in ET patients with the wild-type genotype than in the ET patients with the mutation (1.23±0.15 and 1.14±0.12, respectively; p=0.03) (Table 2). In 74 patients, the three tests (EEC, PRV-1 and XCIP) were concurrently evaluated; overall concordance was found in only 25% of the cases, which comprised nine and ten patients who had all of the tests positive or negative, respectively.

Figure 1.JAK2^V617F allele burden in the 165 patients with the JAK2 mutation included in the study. For comparison, results from patients with PV (n=135)44, primary myelofibrosis (PMF) (n=55) or secondary forms of myelofibrosis (Post-MF) (n=20) are presented. The level found in ET patients was significantly lower than the levels in all other MPD categories (p<0.001). Boxes represent the interquartile range that contains 50% of the subjects, the small square inside the mean value, and the horizontal line inside marks the median; the bars show the upper and lower range of values.

Table 1.Hematologic and clinical features at diagnosis in the 260 patients with ET considered in the study.

Correlation of JAK2V617F burden with clinical characteristics

Patients with the JAK2 mutation were older than those with wild-type alleles (p=0.005), and the mutant allele load progressively increased in older patients (Figure 2) (p<0.0001; 95% CI 0.16–0.44, Spearman’s rank test). In contrast, the median disease duration was inversely related to patients’ age: it was 42, 60, 27 and 18 months in patients less than 50 years old, 51–60 and 61–70 years old, or older than 70 years, respectively (p<0.001).

Splenomegaly was equally present among patients with or without the JAK2 mutation (29% and 24%, respectively; p=0.8); however, median allele burden was significantly higher (27% versus 20% p=0.04) in patients who had splenomegaly as well as in those with a larger spleen (30% versus 20%, p=0.03). Furthermore, there was a correlation between greater JAK2 allele burden and the frequency of patients who had either splenomegaly (p=0.03, χ test for trend; Figure 3) or a larger spleen (p=0.02, χ test for trend). The relative risk (RR) of having splenomegaly was 2.2 (95% CI 1.2–4.1, p=0.03) in the presence of greater than 50% V617F allele using patients with wild-type alleles as the reference population.

There was no difference between patients with mutations or wild-type alleles as regards the frequency of total (arterial plus venous) thrombotic events, which affected 24% and 17% of the patients, respectively (p=0.3), either considering thrombosis at diagnosis (14% and 8%) or in the follow-up (10% and 8%, respectively). The mutant allele burden was similar in patients who did or did not have thrombosis (21% and 22%, respectively). However, when arterial and venous events were considered separately, it was found that allele burden was significantly higher in patients with arterial thrombosis at diagnosis (33% versus 20%; p=0.005), and that the frequency of patients with thrombosis progressively increased with greater allele burden (p=0.003, χ test for trend). The RR for arterial thrombosis at diagnosis in patients with greater than 25% mutant allele was 3.0 (95% CI, 1.3–6.8; p=0.01). The impact of V617F mutant allele burden on arterial thrombosis at diagnosis was mantained in multivariate analysis, which included age, leukocyte count, hemoglobin and platelet count as co-variates (p=0.04). On the other hand, there was no difference between patients with mutant and wild-type alleles as regards the rate with venous events (Table 1) nor there was any meaningful effect of allele burden (not shown in detail).

Symptoms due to microvessel involvement affected patients with and without JAK2^V617F mutations similarly (26% and 20%, respectively; p=0.27); however, when patients were categorized according to their V617F allele burden, the RR of suffering symptoms from microvessel disease increased progressively (p=0.0008, χ test for trend; Figure 3), being 2.2 (95%CI 1.3–3.8; p=0.03) and 2.4 (95% CI 1.2–5.0; p=0.03) in patients with a mutant allele burden greater than 25% and 50%, respectively.

Table 2.Relationships between JAK2^V617F allele burden and percentage of patients with abnormal results of laboratory assays.

Figure 2.JAK2^V617F allele burden in different age groups. The 165 patients with JAK2 mutation were divided in four age classes, as indicated. The age classes comprised 67, 29, 33, and 36 patients. The JAK2 mutant allele burden in the two oldest age classes was significantly different from that measured in patients less than 60 years old (p=0.01 for the 61–70 year old class, and p=0.0006 for the >70 year old age class). Boxes represent the interquartile range that contains 50% of the subjects, the horizontal line inside marks the median, the small square inside the mean value, and bars show the upper and lower range of values.

Finally, the mutant allele burden had no impact on pruritus, systemic symptoms, need for cytoreductive treatment, duration of disease or occurrence of hemorragic events (not shown in detail).

Kinetics of JAK2V617F allele burden

Changes in JAK2^V617F allele level measured at diagnosis were evaluated in 35 patients for whom at least one additional sample, taken during follow-up over a median time of 18 months (range, 6–36 months), was available. None of the 20 patients who had JAK2 wild-type alleles at diagnosis became positive for the mutation. Twelve out of the 15 patients with the mutation maintained stable V617F allele levels during follow-up, the median difference among samples being 5% (range, 1–12%). On the other hand, three patients showed overt changes in their mutant allele burden; in two of them V617F allele levels decreased from 22% and 29% to 4% and 3%, respectively, during treatment with hydroxyurea. However, an additional four patients receiving hydroxyurea did not show appreciable changes of allele burden over time. The third patient presented an increase from 82% to 100% of V617F allele in spite of cytoreductive treatment with hydroxyurea and a purine analog, which was instituted because of progressive leukocytosis with erythroblastosis, accompanied by the onset of transfusion-dependent anemia, suggestive of transformation to post-thrombocythemic myelofibrosis (the patient refused to undergo a bone marrow biopsy).

Discussion

Among the chronic myeloproliferative disorders, ET is characterized by greatest heterogeneity both as regards its clinical profile and at the cellular and molecular levels, since at least one third of females with ET have apparently polyclonal hematopoiesis,28–32 although there may be pitfalls in the interpretation of clonality assays;33 in addition, only 50 to 60% of patients have the JAK2 mutation, while a minority (1%) harbor MPL mutations.34 Current data about the prognostic relevance of the JAK2 mutation in patients with ET are still partially inconclusive. The presence of the mutation has been associated with venous thrombosis occurring in the years before diagnosis in a large series,11 while in other studies no such correlation was found.12,35 Furthermore, in a comparative analysis of patients with ET and PV, ET patients with mutations had significantly more thrombotic events than their wild-type counterpart, and did not differ under this respect from PV patients.17 Intriguingly, the JAK2 mutation was recently reported as being an independent risk factor for fetal loss in ET females.36 Finally, in a large multicenter study, ET patients harboring homozygous JAK2 mutation had a 3.9-fold higher risk of major cardiovascular events as compared to both their heterozygous and wild-type counterparts.13

In this study, we employed a quantitative RT-PCR assay to measure JAK2 alleles in 260 ET patients with the aim of evaluating whether the mutant allele load affected hematologic and clinical phenotype, possibly overcoming limitations of the qualitative assays that had been employed in previous studies. We were prompted to do this by our recent study in PV patients which showed that the extent of hematologic abnormalities and of some major clinical manifestations, in particular thrombosis and chemotherapy need, were actually dependent on the burden of V617F allele measured in granulocytes collected at diagnosis.21 However, another study in PV patients from the Mayo Clinic found no significant correlation between allele burden and clinical manifestations;37 a different source of DNA (bone marrow instead of granulocytes) and the fact that patients in this study were examined variably during follow-up rather than at diagnosis might underlie these discrepancies.

Figure 3.Effect of JAK2^V617F allele burden on clinical outcomes in ET patients. The analysis was performed to calculate the relative risk of a patient presenting with splenomegaly (upper panel), developing major thrombotic events (middle panel) or having microvessel disease (lower panel) according to the amount of JAK2 mutant allele. Squares indicate the relative risk along with the 95% confidence interval (vertical bars). The reference category was patients who were wild-type (WT) for the mutation.

Consistent with previous data, ET patients with JAK2^V617F mutation had higher erythrocyte and leukocyte counts, and reduced platelet levels; however, the extent of these abnormalities was not significantly dependent on the burden of mutant alleles. Similar findings were reported in a recent paper by Tefferi’s group, in which 96 ET patients with mutations were analyzed using a quantitative assay.38 Intriguingly, these results differ from those we reported for PV patients,21 in whom abnormalities in hematologic parameters were dose-dependently correlated with V61F allele burden. We hypothesize that this different behavior might be accounted for by the significantly lower burden of mutant alleles actually present in the ET patients (Figure 1), which prevented a direct cause-effect relationship from being revealed. This statement is indirectly supported by the finding that the percentage of ET patients displaying EEC formation (a phenomenon which is a very sensitive indicator of autonomous activation of the JAK-STAT pathway)39 progressively increased depending on the amount of mutant allele. Also, no correlation was found between either the presence or burden of V617F allele with clonal hematopoiesis by XCIP analysis40 or PRV-1 expression level.41 Finally, patients with the JAK2 mutation were older than those with wild-type alleles,11,13,15 and the load of mutant allele progressively increased with age.

Confirming previous observations,42 we found that none of patients who had wild-type alleles at diagnosis later became positive, while we report some decrease of allele burden in three out of seven patients who were receiving hydroxyurea treatment. Given the low number of patients evaluated, we cannot draw any firm conclusions from these findings, which might have been non-specific unlike those reported in PV patients receiving interferon-α.43

The categorization of subjects according to the load of V617F allele measured by QRT-PCR in their granulocytes allowed us to reveal some effects that had been otherwise masked by the analysis of mutated versus wild-type patients considered as a whole. In fact, we observed that splenomegaly was significantly more frequent in patients having greater than 50% of the mutant allele, while symptoms due to microvessel disease were over-represented when the mutant allele level was just over 25%. Furthermore, we found that there was a 3-fold greater risk of arterial thrombosis at diagnosis in patients with more than 25% V617F allele, even though the difference between patients with the mutation and wild-type patients all together had not attained statistical significance. At variance with other studies, we found no association between the presence of mutant alleles and venous events,11 but this is likely because of the low number of subjects who had venous thrombosis in our series.

Overall, these observations in ET add to the growing idea that hematologic abnormalities, as well as defined aspects of clinical phenotype, in MPD patients are at least partially related to the relative representation of the V617F allele in the context of normal hematopoiesis; this would also help to explain why JAK2-associated abnormalities are seen more clearly in patients with PV21,45 or primary myelofibrosis,16 whose median allele burden is substantially greater than that of ET patients. However, whether JAK2 mutant status and/or allele burden might be fruitfully employed as novel criteria for risk stratification in patients with ET requires controlled prospective trials.

Footnotes

• Authorship and Disclosures EA and PG performed the research, evaluated the data and collaborated in writing the manuscript; GP, CB, AP, VP, LT, LP performed the molecular assays; VS, GL, AB and AMV contributed patients and evaluated clinical records; AMV designed the research, analyzed the data and wrote the manuscript. The authors reported no potential conflicts of interest.
• Funding; this study was supported by Associazione Italiana per la Ricerca sul Cancro, Milano; Ente Cassa di Risparmio di Firenze; and MIUR (COFIN 2006067001_003) grants to AMV. AP was the recipient of a fellowship from the Associazione Italiana per le Leucemie, Firenze.
• The project described was supported by grant number PO1CA108671 from the National Cancer Istitute. The content is solely the responsability of the authors and does not necessarily represent the official views of the National Cancer Institute or the National Institute of health.

• Received April 26, 2007.
• Accepted September 5, 2007.