The relationship of JAK2^V617F and acquired UPD at chromosome 9p in polycythemia vera

Letter to the Editor

JAK2^V617F mutation is the most common somatic event observed in patients with myeloproliferative neoplasms (MPN)^1-4. JAK2^V617F is found in over 95% of patients with polycythemia vera (PV), 55% of essential thrombocytosis and 65% of primary myelofibrosis. Homozygous JAK2^V617F is present in about half of PV patients whereas it is rarely seen in other MPNs^{3, 5, 6}. Patients bearing homozygous JAK2^V617F tend to have a longer duration of disease², higher hemoglobin levels, increased incidence of pruritis⁷ and are more likely to progress to post-PV myelofibrosis⁸. Homozygous JAK2^V617F is shown to result from mitotic recombination, leading to acquired uniparental disomy (aUPD) on chromosome 9p⁹.

It was reported that JAK2 46/1 (GGCC) haplotype may predispose carriers to JAK2 mutation^10-12, and JAK2^V617F facilitates the acquisition of homozygous JAK2^{9, 13}. Challenging this view was a single study reporting 9p aUPD in two PV subjects with wild-type JAK2¹⁴, suggesting that in these two individuals, 9p aUPD might have preceded JAK2^V617F. However, the relationship between JAK2^V617F and 9p aUPD, the frequency and stability of existing PV genotypes has yet to be systematically defined in a larger PV cohort. To address this, we performed whole-exome sequencing and SNP array in 31 PV patients, and further validated our findings in two additional cohorts totaling 59 patients (Supplementary Methods). In addition, we investigated the stability of PV genotypes using 36 longitudinal samples.

The allelic fraction of JAK2^V617F in granulocytes (GNC), measured by sequencing, varied from 0.01 to 1.0 (Figure 1). 9p aUPD spanning JAK2 locus was present in 48% of patients in the discovery set. The fraction of genomes harboring 9p aUPD in a given GNC sample, varied from 0.08 to 1.0 (Figure 1). Based on the quantitative relationship between JAK2^V617F allelic fraction and the fraction of genomes harboring 9p aUPD, we defined 4 PV subgroups and validated this finding in extended two cohorts (Figure 1). Across the three cohorts, 42% of patients harbored JAK2^V617F in a heterozygous state without detectable 9p aUPD (Subgroup I); 45% of patients had JAK2^V617F with an allelic fraction in direct proportion to the level of 9p aUPD (Subgroup II; homozygous JAK2^V617F); 10% of patients harbored 9p aUPD at approximately twice the level of the JAK2^V617F allelic burden (Subgroup III; aUPD with heterozygous JAK2^V617F). Although any single patient in Subgroup III, taken in isolation, could be explained as a mixture of cells from Subgroup I and II, the best explanation for the pattern of Subgroup III across these patients is that JAK2^V617F and 9p aUPD reside in the same cell with heterozygous JAK2^V617F in all such patients. Thus, Subgroup III reported in this study comprises a novel subtype that has not been previously described. Three patients (3%) exhibited trisomy of 9p, generating 2 copies of JAK2^V617F allele (Subgroup IV). The number of patients in Subgroup IV was too small and thus this group was not evaluated further.

Figure 1. The mutational pattern of JAK2^V617F and 9p aUPD.

The mutational pattern of JAK2^V617F and 9p aUPD in the discovery study (n=31) and two validation studies obtained from the archived PV DNA specimen from the same institution (n=40 and n=19, respectively) was shown. I, II, III indicates 3 subgroups defined by the mutational pattern. Two patients in the discovery study with triploid JAK2 was excluded from the analysis. Black dots and open circles, DNA copy number status determined by Affymetrix SNP6 array. Cross, DNA copy number status determined by Illumina610 SNP array. P values of Pearson's correlation test were shown. Line fit was done by JMP algorithm.

We then investigated the stability of each PV genotype using longitudinal samples. The JAK2^V617:9p aUPD ratio remained stable in, 11 of 17 patients, although the JAK2^V617 allele burden itself varied (Figure 2A). In contrast, 6 of 17 patients changed their JAK2^V617 Subgroup membership. Patients PV1, PV16, and PV27 progressed from prognostically more favorable heterozygous JAK2^V617F to less favorable homozygous JAK2^V617F clones (Figures 2B-C). For example, patient PV27 was a mixture of Subgroup-I and -II clones at the first timepoint but four years later only Subgroup-II clone was apparent. However, patients PV2, PV28 and PV31 transformed their JAK2 genotypes and gained new clones (Figures 2D-E). The Subgroup-II clone in PV2 was eliminated, possibly as a result of treatment, and a new Subgroup-III clone emerged. It was interesting that PV28 and PV31 transformed to myelofibrosis (MF) whereas none of the other 14 patients with stable or progressed JAK2 genotype showed phenotype transformation. It has been suggested that patients with homozygous JAK2^V617F were more likely to develop post-PV MF², a propensity we also observed in the Subgroup-II patients (Supplementary Figure 1). Among those 11 Subgroup-II patients that transformed to MF, we have longitudinal samples collected for two patients PV28 and PV31. We observed that, in both patients, there was an outgrowth of a new Subgroup-I clone (Figure 2D). However, we could not establish the sequential order of these two events.

Figure 2. The stability of JAK2 genotypes.

Longitudinal samples were collected from 17 patients during a 6-year period from 2007 to 2013. (A) Patients with stable JAK2 genotypes. (B, C) Patients with progressed JAK2 genotypes. Patients were progressed from prognostically more favorite to less favorite JAK2 genotypes. (D, E) Patients with transformed JAK2 genotypes. Patient transformed their JAK2 genotypes and gained new clones. (F) The pathway of acquisition JAK2^V617F and 9p aUPD in PV. G and T indicate the wild-type allele and the mutant allele of the codon 617 of JAK2 gene, respectively. For translated protein, G corresponds to valine and T corresponds to phenylalanine. The width of the arrows is proportional to the relative percentage of patients which carrying specific mutational pattern calculated by pooled patients from all three cohorts (n=90). The arrows are unidirectional. Therefore patients that transform from one end state to another must represent the outgrowth of a new clone.

We propose four pathways leading to PV (Figure 2F) to explain these results. PV stem cells or their self-renewing progenitors, in the majority of patients, acquire JAK2^V617F first. Approximately half of them remain stable in this configuration (Subgroup I).

Others in this pathway duplicate JAK2 by one of two different mechanisms: a minor fraction duplicated JAK2^V617F-bearing chromosome to produce trisomy 9p (Subgroup IV). Almost half of the patients in Subgroup I duplicate JAK2^V617F allele via mitotic recombination to produce 9p aUPD (Subgroup II). We observed progression from Subgroup I to II in two patients, but the factors controlling the progression could not be ascertained by our analysis. 10% of patients acquire 9p aUPD first followed by JAK2^V617F mutation, yielding patients in Subgroup III. Although the 9p aUPD cell with wild-type JAK2 has been reported in one case study¹⁴, we did not observe 9p aUPD alone in our unselected 3 cohorts, suggesting it might be asymptomatic until acquisition of JAK2^V617F. We found that, in a single female patient which was not included in these 3 cohorts, her granulocytes showed almost complete 9p aUPD with a low JAK2^V617F allelic burden (0.24), indicating the majority of PV clone was composed of 9p aUPD (Supplementary Figure 2). Further, this female was heterozygous for X-chromosome FHL1 gene and her granulocytes were clonal using a transactional clonality assay¹⁵. This patient is probably in a transient state from 9p aUPD with wild-type JAK2 to Subgroup III. Since we observed more patients in Subgroup III, it suggests that acquiring the JAK2^V617F mutation would provide a proliferative advantage. It was reported that JAK2 46/1 (GGCC) haplotype may predispose carriers to JAK2 mutation^10-12, but we didn't observe any significant difference in the frequency of this haplotype across 3 subgroups (Supplementary Figure 3).

The 9p aUPD region, in any given patient, includes many genes (Supplementary Figure 4). In the mitotic recombination event(s) leading to generation of aUPD, both common and rare alleles may be lost. We reasoned that genes with recurrent loss of wild-type alleles within aUPD regions underwent selection for the PV phenotype. Forty-eight genes lost their wild-type alleles in at least 3 patients (Supplementary Figure 5). Among them, 9 genes were related to cell division, 7 in transcriptional regulation, 4 in epigenetic regulation and 3 genes were potential tumor suppressors. KDM4C and SMARCA2, whichare involved in histone modification and chromatin remodeling, are among them. Moreover, 6 of the top 10 genes (Supplementary Table 1) play a functional role in regulation of cell division. Among them, DOCK8, a gene involved in intracellular signaling networks, also had a somatic loss-of-function mutation and a novel germline variant. Recent evaluation of recurrent loss of heterozygosity in 3131 tumors in the absence of clear driver mutations led Elledge and coworkers to hypothesize the existence of cancer gene islands¹⁶. These are regions harboring multiple cancer genes that act in a coordinated fashion within the cancer cell to produce a strong tumorigenic effect. Thus, our observations may suggest that genes in 9p aUPD region other than JAK2 could also contribute to PV pathophysiology.

Heretofore, the zygosity of the JAK2^V617F has been estimated by the allele burden with an arbitrary threshold of >0.5 for homozygous and <0.5 for heterozygotes. This approach assumed a pure tumor cell population, which often is not the case. In this study, we account for tumor cell purity by comparing JAK2^V617F to the level of 9p aUPD in the cell, rather than the absolute allele burden, leading to more accurate determination of mutant JAK2 zygosity. With this more accurate measure of JAK2 mutant zygosity, we observe that most patients can be classified in to one of 3 subgroups (Figure 1), setting the stage for more precise analysis of prognostic value of JAK2 zygosity, as well as improved resolution of the clonal status of a patient's tumor.

For the first time, we investigated the clonal stability of these PV genotypes. Two thirds of the patients were apparently stable in their PV clone for at least two years. Patients of all subgroups were represented among the stable patients. One third of the patients exhibited clonal changes, with emergence of both prognostically more or less favorable JAK2^V617F patterns (Figure 2B-E). However, our samples size is too small to investigate the factors controlling the stability of clones but the role of treatment regimen in clonal stability is clearly an important question. Further study of the clinical properties of a larger cohort of Subgroup III patients, compared to Subgroup I and II, is required to elucidate the contribution of 9p and the cancer genes therein in PV pathogenesis. This novel perspective on the molecular basis of the evolution of PV should lead to a better understanding of the roles of JAK2^V617F and 9p aUPD in this disease.