Introduction

Despite the seminal discovery of JAK2 or MPL mutations in the majority of patients with BCR-ABL1-negative myeloproliferative neoplasms (MPN),1, 2, 3, 4 it is becoming increasingly evident that these mutations do not signify either disease-initiating or leukemia-promoting events.5, 6 It is therefore important to keep looking for additional molecular alterations to clarify the genetic underpinnings of both chronic- and blast-phase MPN. In the last 2 years, mutations involving TET2, ASXL1 and CBL have been described in some patients with BCR-ABL1-negative MPN, including polycythemia vera (PV), essential thrombocythemia (ET) and primary myelofibrosis (PMF).7 The precise pathogenetic contribution of these mutations and their clinical relevance are currently under investigation. The glioma-associated8 isocitrate dehydrogenase 1 (IDH1) and IDH2 mutations are the latest to be added to the ‘MPN mutations list’.9

IDH1, located on chromosome 2q33.3, and IDH2, located on chromosome 15q26.1, encode enzymes that catalyze oxidative decarboxylation of isocitrate to α-ketoglutarate. IDH1 (cytoplasm and peroxisome) and IDH2 (mitochondria) use NADP+ as a co-factor to generate NADPH, which is important in the production of intracellular glutathione. Intact IDH activity is therefore necessary for cellular protection from oxidative stress. Mutant IDH has decreased affinity to isocitrate, but displays neomorphic catalytic activity toward α-ketoglutarate, the net result being decreased supply of α-ketoglutarate and accumulation of 2-hydroxyglutarate.10, 11, 12, 13 It is currently believed that these intracellular changes facilitate oncogenic pathways including activation of HIF-1α.10

IDH1 and IDH2 mutations were first described in low-grade gliomas/secondary glioblastomas8 and subsequently in acute myeloid leukemia (AML),14 with respective mutational frequencies of ∼70 and 8%. We recently screened 200 patients with either chronic- or blast-phase MPN for IDH mutations, and identified 9 patients with either IDH1 (n=5) or IDH2 (n=4) mutations.9 Mutational frequencies were ∼21% for blast-phase MPN and ∼4% for PMF. In the current study, we expanded our study cohort to include 1473 patients recruited from three MPN centers of excellence, with the intent to accurately describe the prevalence of IDH mutations in chronic-, fibrotic- and blast-phase PV, ET and PMF. In addition, IDH-mutated patients were analyzed for their cytogenetic and molecular (that is, JAK2, MPL and TET2 mutation and JAK2 haplotype status) phenotype, as well as their prognostic relevance.

Materials and methods

This study was approved by the Mayo Clinic institutional review board. All patients provided authorization for use of their medical records for research purposes, and the research was carried out according to the principles of the Declaration of Helsinki. Patient samples were obtained from the Mayo Clinic, Harvard Medical Institute and University of Florence. Mutational analyses were performed on DNA derived from either bone marrow or peripheral blood granulocytes. JAK2 46/1 haplotype analysis on patient samples accrued from Harvard was performed on germline DNA. Diagnoses of MPN, post-PV/ET MF and AML, in patient samples accrued from the Mayo Clinic and the University of Florence, were according to the World Health Organization and International Working Group criteria.7, 15 Diagnoses in patients accrued from Harvard were self-reported during an internet-based collection of samples, as previously detailed.16

DNA from either bone marrow (Mayo Clinic samples) or granulocytes (samples from Harvard and the University of Florence) was extracted using conventional methods. MPL, JAK2 and TET2 mutation and JAK2 haplotype analyses were performed according to previously published methods.4, 17, 18, 19 With regard to IDH mutation analysis, Harvard patient samples were analyzed using the following primers for IDH1, which cover amino acid residues 41–138: sense, 5′-TGTGTTGAGATGGACGCCTA-3′ and anti-sense, 5′-GGTGTACTCAGAGCCTTCGC-3′. Sequencing of IDH2 used primers that covered amino acid residues 125–226: sense, 5′-CTGCCTCTTTGTGGCCTAAG-3′ and anti-sense, 5′-ATTCTGGTTGAAAGATGGCG-3′. Sequence analysis was performed using Mutation Surveyor (SoftGenetics, State College, PA, USA) and all mutations were validated by repeat PCR and sequencing on unamplified DNA from the archival sample.

Mayo Clinic and University of Florence patient samples were screened for IDH1 and IDH2 mutations by direct sequencing and/or high-resolution melting assay. Direct sequencing for IDH1 exon 4 mutations was carried out using the following primer sequences: sense, 5′-CGGTCTTCAGAGAAGCCATT-3′ and anti-sense, 5′-CACATTATTGCCAACATGAC-3′.18 IDH2 exon 4 was amplified using sense, 5′-CCACTATTATCTCTGTCCTC-3′ and anti-sense, 5′-GCTAGGCGAGGAGCTCCAGT-3′.19 Both reactions were performed in 25 μl volume containing 100 ng of DNA, 0.25 U Taq polymerase, 0.3 mM each of dATP, dCTP, dGTP and dTTP, 5 μl of a 10 × PCR buffer (Roche Diagnostics, Indianapolis, IN, USA) and 0.2 μM each of sense and anti-sense primers. The reaction was denatured at 94 °C for 3 min followed by 35 cycles of denaturing at 94 °C for 30 s, annealing at 57 °C for 30 s and extension at 72 °C for 40 s. After a final extension at 72 °C for 2 min, the products were confirmed by running on 1.3% agarose gel and purified using Qiagen's PCR Quick Purification Kit. The product was sequenced using the ABI PRISM 3730xl analyzer (Applied Biosystems Inc, Foster City, CA, USA) to screen for the presence of mutations.

High-resolution melting was performed using the LightCycler 480 real-time PCR system (Roche Diagnostics), using the above-mentioned primers for IDH1 mutations (R130) and the following primers for IDH2 mutations (R140 and R172): R140 sense, 5′-GCTGAAGAAGATGTGGAA-3′ and anti-sense, 5′-TGATGGGCTCCCGGAAGA-3′; R172 sense, 5′-CCAAGCCCATCACCATTG-3′ and anti-sense, 5′-CCCAGGTCAGTGGATCCC-3′ (Figure 1).

Figure 1
figure 1

High-resolution melting (HRM) normalized and temperature-shifted difference plot for IDH1 (a) and IDH2 (b) and corresponding sequences (c and d).

Full size image

Conventional statistical procedures were used (SAS Institute, Cary, NC, USA). All statistically analyzed data were obtained at time of IDH mutation analysis. All P-values were two-tailed and statistical significance was set at the level of P<0.05. Categorical variables were described as count and relative frequency and compared by χ2 statistics. Comparison of continuous variables between categories was performed by the Mann–Whitney U-test. Survival analysis was performed by the Kaplan–Meier method taking the interval from the date of diagnosis, for chronic-phase disease, or from the date of leukemic transformation, for blast-phase disease, to death or last contact. The log-rank test was used to compare survival data. Cox regression model was used for multivariable analysis.

Results

Disease- and stage-specific IDH mutational frequencies

A total of 1473 patients with BCR–ABL1-negative MPN were recruited from the Mayo Clinic, Rochester, MN, USA (n=629), University of Florence, Florence, Italy (n=522) and Harvard Medical Institute, Boston, Massachusetts, USA (n=322). Specific diagnoses included ET (n=594), PV (n=421), PMF (n=312), post-PV MF (n=54), post-ET MF (n=41), post-PV AML (n=12), post-ET AML (n=7) and post-PMF AML (n=32). Table 1 provides clinical and laboratory details of the study population including age and sex distribution, specific diagnoses and JAK2, MPL and TET2 mutational and JAK2 46/1 haplotype status, stratified by center of patient recruitment. A total of 38 IDH mutations were documented (Table 2): 18 involved IDH1 (10 R132S, 7 R132C and 1 R132G) and 20 IDH2 (18 R140Q, 1 R140W and 1 R172G). IDH mutations were infrequent in chronic- or fibrotic-phase disease and significantly more prevalent in blast-phase disease (P<0.01; Table 3): 5 (0.8%) in ET, 8 (1.9%) in PV, 13 (4.1%) in PMF, 1 (1%) in post-ET/PV MF, none in blast-phase ET, 3 (25%) in blast-phase PV and 8 (25%) in blast-phase PMF.

Table 1 Specific diagnoses, age/sex distribution, JAK2, MPL and TET2 mutational status and JAK2 non-46/1 haplotype frequency in 1473 patients with polycythemia vera (PV), essential thrombocythemia (ET), primary myelofibrosis (PMF), post-PV MF, post-ET MF, post-PV acute myeloid leukemia (post-PV AML), post-ET AML or post-PMF AML
Full size table
Table 2 Clinical, cytogenetic and molecular details, at time of mutation analysis, of 38 IDH-mutated patients with chronic- or advanced-phase polycythemia vera, essential thrombocythemia or primary myelofibrosis
Full size table
Table 3 IDH mutational frequencies in 1473 patients with polycythemia vera (PV), essential thrombocythemia (ET), primary myelofibrosis (PMF), post-PV MF, post-ET MF, post-PV acute myeloid leukemia (post-PV AML), post-ET AML or post-PMF AML
Full size table

Correlation of IDH mutations with other MPN-associated mutations and JAK2 46/1 haplotype

Considering the preponderance of informative cases with centrally confirmed diagnosis and availability of a more complete laboratory data, the current analysis was limited to patients from the Mayo Clinic cohort (n=629). IDH mutational frequencies were similar among JAK2- (3.6%), MPL- (4.3%) and TET2 (3.2%)-mutated patients and their respective mutation-negative counterparts (4.2, 5.3 and 6.3%; Table 3). In other words, mutant IDH was shown to co-occur with a JAK2, MPL or TET2 mutation, and mutational frequency did not appear to be influenced by either the type of the coexisting mutation (P=0.96) or the presence or absence of each specific mutation (Table 3). However, IDH-mutated cases were more likely to be nullizygous for JAK2 46/1 haplotype, especially when analysis was restricted to informative (that is, with JAK2 46/1 haplotype information) patients with chronic- (n=158) or blast (n=23)-phase PMF, analyzed together (P=0.007) or separately (P=0.04; Table 4).

Table 4 IDH mutational frequencies in 193 Mayo clinic patients with chronic-phase (n=166) or blast-phase (n=27) primary myelofibrosis (PMF) stratified by JAK2 mutational, JAK2 46/1 haplotype or cytogenetic status
Full size table

Clinical correlates and prognostic relevance

To avoid disease- or stage-specific confounding factors, as well as assure adequate sample size of informative cases, clinical correlative and prognostic studies were limited to PMF. In this patient cohort, detailed clinical information was available in 111 patients with chronic-phase PMF (including 7 IDH-mutated cases) and 27 patients with blast-phase PMF (including 8 IDH-mutated cases), both patient populations were accrued from the Mayo Clinic cohort. In both chronic- and blast-phase PMF, the presence of IDH mutations was not influenced by either age (P=0.51 and 0.70, respectively) or gender (P=0.09 and 0.3, respectively). In chronic-phase disease, comparison of prognostically relevant disease variables at diagnosis revealed that cytogenetic findings in IDH-mutated cases often belonged to a low- or intermediate-risk category,20 although the difference was not statistically significant (Table 4). Similarly, IDH-mutated blast-phase PMF was less likely to display complex karyotype (0 vs 64% in IDH-unmutated cases; P=0.001).

In addition to biological implications, the aforementioned associations of IDH mutations with favorable cytogenetic profile and JAK2 46/1 haplotype nullizygosity, both which have previously been shown to be prognostically relevant,19, 20 mandated their inclusion as covariates during multivariable survival analysis. In chronic-phase PMF, univariate analysis showed statistically significant adverse survival effect from JAK2 46/1 haplotype nullizygosity (P=0.0001; 34 nullizygous vs 74 not nullizygous), high-risk karyotype (P<0.0001; 13 high-risk vs 98 not high-risk) and higher International Prognostic Scoring System (IPSS; 27 high, 29 intermediate-2, 30 intermediate-1 and 25 low-risk patients)21 risk score (P<0.0001), but not from IDH mutational status (P=0.54; 7 mutated vs 104 unmutated; Figure 2). Multivariable analysis confirmed the independent prognostic value of JAK2 46/1 haplotype status (hazard ratio (HR) 2.2, 95% confidence interval (CI) 1.2–4.2), karyotype (HR 2.8, 95% CI 1.3–5.9) and IPSS risk score (HR 4.8, 95% CI 2.0–11.5).

Figure 2
figure 2

Survival curves of 111 patients with chronic-phase primary myelofibrosis stratified by their isocitrate dehydrogenase (IDH) mutational status (a), cytogenetic risk (b), JAK2 46/1 haplotype status (c) or International Prognostic Scoring System risk category (d).

Full size image

In blast-phase PMF, despite its association with noncomplex karyotype, the presence of mutant IDH predicted shortened survival, calculated from the time of disease transformation (P=0.04), and there was a similar trend for JAK2 non-46/1 haplotype (P=0.14; Figure 3). Significance was lost for both during multivariable analysis, probably because of small sample size. IDH mutation status also predicted worse survival when the analysis included all blast-phase MPN cases from the Mayo cohort (Figure 4; n=43; P=0.01). In this instance, significance was sustained during multivariable analysis that included JAK2 46/1 haplotype as a covariate.

Figure 3
figure 3

Survival curves of patients with blast-phase primary myelofibrosis stratified by their isocitrate dehydrogenase (IDH) mutational (a; n=27 including 8 mutated cases) or JAK2 46/1 haplotype (b; n=23 including 8 nullizygous cases) status.

Full size image
Figure 4
figure 4

Survival curves of 43 patients with blast-phase myeloproliferative neoplasm stratified by their isocitrate dehydrogenase (IDH) mutational status.

Full size image

Discussion

IDH1 point mutations involving exon 4 occur in the majority (60–90%) of patients with low-grade gliomas and secondary glioblastomas, and always affect the amino acid arginine at position 132 (∼93% R132H, 4% R132C, 2% R132S and <1% R132G, R132L or R132V).8, 22, 23 These mutations are relatively infrequent in primary glioblastoma (∼7%)22 and are usually not seen in other solid tumors.23, 24 A small fraction (∼4%) of glioma-associated IDH mutations involves IDH2, specifically the R132 analogous R172 residue on exon 4 (R172K, R172M, R172G, R172W).23, 25 IDH mutations in glioma are heterozygous, believed to constitute early genetic events and might be mutually exclusive of EGFR and PTEN, but not TP53 mutations. Clinical correlates of IDH mutations in glioma include younger age, longer survival and reduced risk of disease progression after conventional therapy.8, 22, 23, 26, 27

The first study on IDH mutations in AML included 188 patients with primary AML and reported IDH1, but not IDH2, mutations in 8.5% (n=16) of the cases and 16% of those with normal karyotype: R132C in 8 patients, R132H in 7 and R132S in 1.14 In a subsequent AML study of 493 patients,28 27 (5.5%) expressed IDH1 mutations (37% R132C, 26% R132H, 19% R132S, 15% R132G and 4% R132L). In both studies,14, 28 IDH1 mutations clustered with normal karyotype, NPM1 mutations and trisomy 8. IDH1 mutations are rare in pediatric AML.29 More recently, IDH2 mutations, affecting R172 (R172K)12, 13 or R140 (R140Q),13 were also shown to occur in primary AML.12, 13 In one of these studies, IDH1 or IDH2 mutations were seen in 18 (23%) of 78 AML cases and the majority of the mutations (12 of 18) involved IDH2, primarily R140Q.13 In general, survival in primary AML did not seem to be affected by the presence of IDH mutations.13, 14, 28, 29, 30 However, more recent studies suggest that specific IDH mutation variants might be prognostically relevant in certain molecular subsets of AML.31

The first reports of IDH mutations in MPN came from three independent groups.9, 32, 33 In one of these studies, IDH1 mutations were seen in ∼8% (5 of 63) of blast-phase MPN patients, mostly occurring in the absence of TET2 and ASXL1 mutations.32 The second study was focused on blast-phase MPN that arose from JAK2-mutated chronic-phase MPN.33 In this study, mutant IDH was seen in 5 (31%) of 16 blast-phase MPN (three cases with R132C and two with R140Q) and in none of the 180 PV or ET patients.33 The third study from the Mayo Clinic included 200 MPN patients and showed IDH mutational frequencies of ∼21% for blast-phase MPN, regardless of JAK2 mutational status, and ∼4% for PMF.9 The specific IDH1 mutations found in the particular study included R132C and R132S and the IDH2 mutations R140Q and R140W.

The current study is an extension of the above-mentioned Mayo Clinic study and involves a large number of patients (n=1473) recruited from three major MPN centers of excellence. The results of the study clarify a number of issues regarding IDH mutations in MPN. First, the study provides robust incidence figures for IDH1 and IDH2 mutations across different disease stages of specific MPN variants. Accordingly, we now show that both IDH1 and IDH2 mutations can occur in chronic-phase ET, PV or PMF, although infrequently. Mutational frequency was equally low in post-PV/ET MF and this fact combined with the significantly higher mutation incidence observed in blast-phase disease suggests a pathogenetic contribution to leukemic but not fibrotic disease transformation. Two additional observations support this contention (i) complex karyotype was infrequently encountered in IDH-mutated blast-phase MPN, which suggests an independent pathogenetic contribution that might be tied to distinct molecular alterations, such as, for example, overexpression of the APP (amyloid â (A4) precursor protein) gene, which has previously been shown in AML to be associated with either complex karyotype or IDHR172 mutation31 and (ii) the absence of mutual exclusivity between IDH and other MPN-associated mutations (for example, TET2, MPL), which is consistent with the suggestion that the former are later-arising cooperating mutations that are more involved in disease progression rather than disease initiation.

The types of IDH mutations seen in our patients with MPN (mostly IDH2R140Q and IDH1R132S/C) are distinctly different than those seen in gliomas (mostly IDH1R132H) and more similar to those seen in AML, although IDH1R132H was significantly more prevalent in AML. Within the context of MPN, IDH2R140Q was over represented in chronic-phase ET and PV, whereas IDH1 mutations were more prevalent in PMF and blast-phase MPN. More studies are needed to confirm this apparent trend. Regardless, there is currently no good explanation for the observed diversity in IDH mutation variants among gliomas and myeloid malignancies and current information suggests similar biological consequences.13 Whether or not different IDH mutations carry different prognostic relevance in MPN is currently not known and we did not attempt to address the particular issue because of our relatively small number of informative cases. Of note, in a recent study of primary AML with normal karyotype, different types of IDH mutations appeared to variably influence disease-free survival and complete remission rates.31

One particularly interesting observation from the current study was the significant association between mutant IDH and JAK2 non-46/1 haplotype. The latter phenomenon is further evidence for the JAK2 mutation specificity of the previously described association between the JAK2 46/1 haplotype and MPN.19, 34, 35 In other words, whereas JAK2 exon 1419, 35 or exon 1236 mutations have been shown to be associated with JAK2 46/1 haplotype, we did not see the same effect involving MPL mutations34 (although others have shown otherwise),37 and now show an association with JAK2 non-46/1 haplotype for IDH mutations. This latter observation is also consistent with our previous report on the prognostically detrimental effect of JAK2 non-46/1 haplotype in PMF;19 it is possible that patients with PMF who are nullizygous for JAK2 46/1 haplotype are susceptible to additional adverse molecular events, such as IDH mutations, which might lead to biologically more aggressive disease. Consistent with this possible scenario, in the current study, the negative prognostic impact of mutant IDH was accounted for by the JAK2 46/1 genotype in PMF but not in blast-phase MPN, in which risk factors other than JAK2 non-46/1 haplotype might have promoted the development of IDH mutations.

It is becoming increasingly evident that there are many more mutations than JAK2 and MPL mutations in BCR–ABL1-negative MPN including those that involve TET2,38, 39 ASXL1,40 IDH1,32, 33 IDH2,9, 33 CBL,41 IKZF142 and LNK.43 Some of these mutations might be later-arising and more prevalent in blast-phase disease. What is currently lacking is a composite evaluation (that is, concurrent analysis of all relevant mutations), which includes paired chronic- and blast-phase samples of a large number of patients with blast-phase MPN. Such an approach is essential for clarifying the individual pathogenetic or prognostic contribution of the aforementioned mutations and their chronological order of appearance. It is very likely that additional mutations in MPN will be described soon, but practical relevance in terms of either disease prognostication or value as drug targets has so far been limited.