Common Variation at 6q25.3 (TULP4) Influences Risk for Arterial Thrombosis in Myeloproliferative Neoplasms

Inherent tendency for thrombosis is a major complication in myeloproliferative neoplasms (MPN). The molecular basis of thrombosis in MPN is not well understood, however, genetic factors have been proposed to play a role. To agnostically investigate the role of common germline variation in MPN thrombophilia, we performed genome-wide association studies (GWAS) in MPN patient cohorts characterized for arterial thrombosis (AT) events. In our discovery cohort (n=383) from Vienna, Austria, 18% of patients have suffered from AT after MPN diagnosis, and 33% of patients showed records for AT events at any time. Of this discovery cohort, a subset of patients (n=302) selected independently from thrombotic status was genotyped on the Affymetrix Genome-Wide SNP 6.0 array platform. After assessment of case-control setup and genotyping quality as standardly implemented in GWAS, confidently genotyped single nucleotide polymorphisms (SNPs) were tested for allelic association with occurrence of AT after MPN diagnosis. We observed an association signal beyond genome-wide statistical significance (P<5x10^-8) at chromosome 6q25.3, tagged by rs6455579 and six additional closely related tag-SNPs. The 6q25.3 risk haplotype is common with a minor allele frequency (MAF) of ~8% in the Vienna MPN cohort. The minor allele rs6455579_C correlated with increased risk for AT. We next genotyped the full discovery cohort from Vienna (n=383) and a replication cohort from Pavia, Italy (n=505) using qPCR-based SNP genotyping for rs6455579 and computed odds ratios (ORs) to estimate the effect sizes of the association. In the Vienna cohort, rs6455579 genotype was significantly associated with both “AT after diagnosis” (P=1.90x10^-8, OR_het=6.93) and “AT at any time” (P=6.96x10^-4, OR_het=3.07). In the Pavia cohort, an association with “AT after diagnosis” could not be observed (P=0.14), however, the association of rs6455579 with “AT at any time” could be reproduced at formal statistical significance (P=7.92x10^-3, OR_het=2.55). Moreover, Kaplan-Meier statistics on the Pavia cohort revealed a significant difference in AT-free survival after diagnosis upon rs6455579 genotype (log-rank-test P=0.009; rs6455579 homozygous major vs. heterozygous), underpinning the relevance of disease duration for the impact of the haplotype on MPN-related AT. To gain insight into the physiological mechanisms behind the 6q25.3 risk haplotype, we next evaluated the discovery cohort for correlations of the haplotype with clinical parameters other than AT.  We could observe a significant trend for increased white blood cell (WBC) count in CALR mutated (n=90, P=2.50x10^-3) but not in JAK2 mutated (n=182, P=0.52) ET and PMF patients carrying the risk allele. CALR mutated patients have been previously reported to exhibit significantly prolonged thrombosis-free survival as compared to JAK2 mutated patients. Thus, our observation indicates that the 6q25.3 germline risk haplotype might impact on WBC count most strongly in a subgroup of patients considered low risk for thrombosis based on the somatic mutational status. The 6q25.3 core haplotype (R²>0.3) spans ~300 kilobases, covering intergenic sequence as well as promoter and 5’ exons of the TULP4 gene. To test for the possibility of a rare coding variant in TULP4 or other more distant genes underlying the association through long-range linkage disequilibrium, we used genotype imputation in conjunction with the 1000 genomes reference panel to infer genotypes on all untyped variants (MAF>1%)  in a 5 megabase region centered on the core haplotype. We did not detect any coding variants reflecting the association, and the tag-SNPs from the initial GWAS remained the most strongly associated variants. Causative non-coding genetic variation identified in GWAS is thought to exert its function through differential regulation of specific target genes. Therefore, we evaluated a potential influence of the 6q25.3 risk haplotype on TULP4 gene expression in peripheral blood (The Cancer Genome Atlas (TCGA) LAML dataset, RNA-Seq on 173 acute myeloid leukemia patients). Indeed we detected significantly decreased TULP4 expression in risk haplotype carriers (P=0.029), providing indirect evidence for reduced TULP4 transcript levels impacting on elevated risk for AT in MPN. Further studies will be required to functionally assess the potential role of TULP4 in MPN-related AT.