Myeloproliferative neoplasms (MPN) constitute a group of clonal hematological disorders characterized by an expansion and accumulation of one or more mature cell types of the myeloid lineages. There are four main groups of MPNs that can be classified by the presence or absence of the Philadelphia (Ph) chromosome.1 Chronic myeloid leukemia is accompanied by the Ph chromosome, whereas polycythemia vera (PV), essential thrombocythemia (ET) and primary myelofibrosis (PMF) are not.1 In all three Ph-negative MPN subtypes (hereafter called MPN), a recurrent somatic mutation is frequently found in the pseudokinase region of Janus kinase 2 (JAK2V617F).2 The mutation is observed in about 80–95% of PV cases, 50% of ET cases and 60% of PMF cases.2 Owing to the activating nature of the mutation, it leads to constant stimulation of myeloid proliferation.2 Family studies have shown that germline variants may predispose to the disease as the risk of MPN among first-degree relatives of MPN patients in Sweden is five to sevenfold greater than that in the general population.2 Furthermore, common germline single-nucleotide polymorphisms (SNPs) at the JAK2 locus have been associated with MPN.3, 4, 5

The aim of the current study was to search for germline sequence variants that associate with risk of MPNs by performing genome-wide association analysis. For the association analysis, we used sequence variants identified by whole genome sequencing (WGS), to an average depth of 22 × , of DNA isolated from white blood cells of 2230 Icelanders (Supplementary Material). Using imputation assisted by long-range haplotype phasing, the genotype probabilities of all the 34.2 million sequence variants identified were determined for Icelanders genotyped with Illumina (San Diego, CA, USA) SNP chips. Furthermore, using Icelandic genealogical information, genotype probabilities were calculated for individuals who are close relatives of chip-typed individuals (familial imputation) (Supplementary Material). We then tested the identified variants for association with MPN, using 237 Icelanders diagnosed with MPN, including PV (n=98), ET (n=40), PMF (n=26) and 34 128 Icelandic controls with imputed genotypes (Table 1 and Supplementary Table S1). It is worth stressing that somatic MPN mutations do not affect the association analysis as the variants identified through WGS and subsequently imputed into MPN cases are almost exclusively from individuals not diagnosed with MPN (one MPN case, diagnosed 9 years after blood draw, among the 2230 sequenced individuals).

Table 1 Association of the TERT variant rs2736100_C and the JAK2 variant rs1034072_A with MPN in Icelandic samples
Full size table

Assuming a multiplicative model, we observed a genome-wide significant association between MPN in the Icelandic population and a common variant rs2736100_C located in the second intron of the TERT gene at the TERT-CLPTM1L locus (allele frequency (AF)=49.3%, odds ratio (OR)=2.09, P-value=6.39 × 10−10) (Figure 1 and Table 1). In addition, we replicate the association of rs10974944_G in JAK2 previously reported to associate with MPN5 (AF=28.7%, OR=1.78, P-value=1.90 × 10−6; Supplementary Table S2) and detect an even stronger association between MPN and a second variant, rs1034072_A, in JAK2 (AF=28.2%, OR=1.85, P-value=3.19 × 10−7; Table 1). This variant is highly correlated with rs10974944_G (r2=0.91; Supplementary Table S2). As reported for rs10974944_G, the association of the JAK2 variant rs1034072_A with PV (OR=2.76) is stronger than that with the other MPN subphenotypes (ET, OR=1.39 and PMF, OR=1.56)5 (Table 1). In contrast, the effect of the risk allele of rs2736100_C at the TERT-CLPTM1L locus is similar for the three MPN subphenotypes, with OR 2.32 (PV), 2.25 (ET) and 2.42 (PMF) (Table 1).

Figure 1
figure 1

Genome-wide association results for MPN case (N=237) and control (N=34 128) analysis in Iceland. The −log10 of the allelic P-values <0.05 for the 34.2 million sequence variants tested is presented. The chromosomal distribution is shown as a Manhattan plot.

Full size image

We tested all variants at the TERT-CLPTM1L locus that associate with MPN with a P-value<1 × 10−5 and those that have been associated with cancer and idiopathic pulmonary fibrosis,6, 7 conditioning on rs2736100. None of these SNPs remained significantly associated with MPN after conditioning the association on rs2736100 (Supplementary Table S3). This raises the possibility that rs2736100 is the causative variant at this locus as the association was based on WGS data and therefore, all simple variants at this locus were tested.

In a recent publication, rs2736100_C is one of the eight variants that were shown to associate with long telomeres in white blood cells,8 suggesting that this SNP acts on the TERT gene encoding the reverse transcriptase of the telomerase complex essential for maintaining the telomere length. To our knowledge, functional effect of rs2736100 has not been evaluated. However, for rs7705526, an SNP located 542 bp telomeric to rs2736100 and the variant at the TERT-CLPTM1L locus that shows the highest correlation with rs2736100 (r2=0.52; Supplementary Table S3), an increased enhancer activity has been described for the allele correlated with rs2736100_C.9 Therefore, rs2736100_C might increase transcription of TERT through increased enhancer activity and this enhanced expression could mediate the MPN risk.

It is well documented that MPN cells have shorter telomeres than normal cells10 as do cells in many types of neoplasms.11 Short telomeres have been proposed to promote tumor formation by inducing genomic instability. The association of TERT rs2736100_C with increased risk for MPN and with longer telomeres thus appears to be a paradox. However, the biology of the TERT gene and its link to cancer (including MPN) is complicated and could even be mediated through functions other than telomerase-mediated extensions of telomeres such as its effect on cell DNA repair, cycle regulators or on cell signaling.12 This hypothesis is further supported by a lack of association with MPN of rs10936599_C in the TERC gene (template-containing telomerase RNA that is part of the telomerase complex; Supplementary Table S4) that shows the strongest reported genome-wide association with increased telomere length.8

An association of TERT rs2736100_C was observed in Iceland with an increase in red blood cell counts similar to a previous report13 (effect (s.d.)=0.019, P-value=9.07 × 10−6, N=76 739 individuals; Supplementary Table S5). In addition, we observed an association with elevated platelet counts (effect (s.d.)=0.025, P-value=5.26 × 10−8, N=103441 individuals) and white blood cell counts (effect (s.d.)=0.016, P-value=4.10 × 10−4, N=126 853 individuals). The association with white blood cell counts is limited to cells of the myeloid linage (Supplementary Table S5) and the effect on blood cell counts remained unchanged after excluding MPN cases. In addition to the association of rs2736100_C with the three subphenotypes of MPN, these results for the blood cell counts indicate that the TERT variant asserts its effect on a common myeloid progenitor. As this effect on cell counts is seen in individuals without MPN it could mean that the risk of MPN conferred by the variant is mediated through an effect on cell numbers. In line with no effect on lymphocyte counts, rs2736100_C did not associate with lymphoproliferative malignancies (Supplementary Table S6). In contrast to the TERT variant, the germline JAK2 risk variant rs1034072_A is not associated with red blood cell count (P-value=0.91, effect (s.d.)=0.001, N=76 739 individuals), but associates with decrease in platelet count (P-value=2.50 × 10−4, OR=−0.019, N=103 441 individuals) (Supplementary Table S5). This might indicate that TERT and JAK2 predispose to MPN through different mechanisms.

The recurrent somatic mutation JAK2V617F is present in a large fraction of MPN cases.2 We analyzed available blood samples from Icelandic MPN cases (N=103) for the JAK2V617F somatic mutation using a real-time quantitative-PCR assay (Supplementary Material). We excluded blood samples drawn more than 2 years before MPN diagnosis (N=43). In the remaining samples (N=62) we observed positive JAK2V617F somatic mutation status in 69.4% of MPN cases and 80.6% of PV cases (N=31) (Supplementary Table S7), in line with previous studies.2 The germline MPN risk allele in JAK2, rs10974944_G, has been shown to associate with increased allelic burden of the JAK2V617F somatic mutation.14 We tested the counts of the MPN risk alleles rs1034072_A in JAK2 and rs2736100_C in TERT for correlation with JAK2V617F allele burden. We observed a significant correlation between the number of germline risk alleles and allelic burden of JAK2V617F at the JAK2 locus (P=0.02, effect=10.28% of allelic burden per germline risk allele) but not at the TERT locus (P=0.77) (Supplementary Table S8). However, owing to our small MPN sample set, a weak effect of the germline TERT variant on JAK2V617F allele burden cannot be excluded. Recurrent somatic mutations in the TERT promoter region, C228T and C250T, shown to increase TERT expression, have recently been reported in several cancer types.15 We screened for these mutations by Sanger sequencing in the MPN cases used for screening the JAK2V617F mutation (N=62) and observed only a single occurrence of the the C228T TERT mutation.

In conclusion, we report a common germline variant in TERT that associates with MPN. The risk mediated by the variant rs2736100_C is large (OR=2.09) and in our data even larger than the reported JAK2 germline signal rs1034072_A (OR=1.85). As the TERT variant is more common than the JAK2 variant (49.3% vs 28.2%) its population attributable risk is larger (57.7% vs 34.9%). Unlike the JAK2 variant, the TERT variant exerts similar risk on the three MPN subtypes PV, ET and PMF. Our data also suggest that the germline TERT variant does not associate with the recurrent somatic JAK2V617F mutation. Association of the TERT variant with increased counts of myeloid white blood cells, red blood cells and platelets but not lymphoid cells, in the general population, together with similar effects on all three subpopulations of MPN suggests that the variant exerts its effect on hematopoiesis by increasing proliferation of cells derived from common myeloid progenitor.