Not dead yet: GWAS insights into a common bleeding disorder Free

In this issue of Blood, Thibord et al present results from the largest genome-wide meta-analysis of heavy menstrual bleeding where they identify over 30 common genetic loci associated with the risk of heavy menstrual bleeding.¹ 

In the era of whole-genome sequencing, many have predicted the demise of the big genome-wide association study (GWAS) meta-analysis. However, these GWASs continue to provide interesting insights into the genetic determinants of complex human disease traits. For the readers of who may yawn when they see the term “single nucleotide polymorphism” (SNP) or long for the days of a simple hypothesis-based study, I will attempt to briefly describe why GWAS remains relevant and useful to our field.

First, we learn new things when we conduct a good GWAS. The success of a GWAS is determined by the precision of the phenotyping, the number of individual genomes tested, the quality and depth of genotyping, and the analytic pipeline. A well-powered GWAS will identify areas of the genome where common DNA variants (found in >1% of the population) are linked by association to complex genetic traits, which are the majority of the disease traits physicians treat. These genetic loci point to altered gene expression or altered gene function that plays a role in disease risk. Visually we can see the areas of the genome with common variant signals with the Manhattan plot. Here, the skyscrapers are the clusters of SNPs linked by ancestral inheritance blocks, and their very low P values represent a statistically significant association. Below them lies the mass of insignificantly associated SNPs lost in the millions of other SNPs and pointing to false-positives from previous single-gene association studies (like MTHFR and venous thromboembolism). The multiple loci detected in GWASs allow for a ranked list of gene variants with different effects on a trait. GWAS meta-analyses add to the bottom of that list as they gain power to detect weaker associations. Conditional analyses can further define the number of independently inherited signals at each locus and detect gene-gene interactions.

Second, GWASs can uncover new biology. This has been evident since 2005 when a GWAS for macular degeneration made a connection with complement pathway biology.² Since then, thousands of GWASs have identified signals at loci not previously connected to the phenotype. Due to several factors, including the sheer number of novel results, it remains a challenge to understand how variances at these unpredicted loci influence the phenotypes.

Third, GWASs generate useful data for downstream studies. Results of GWASs can be used to assemble and test a genetic risk score. Thibord et al show that individuals with genetic risk scores in the lower 25 percentile had a lower risk (odds ratio 0.77) for heavy menstrual bleeding and those in the upper 25th percentile had a higher risk (odds ratio 1.47) compared with people in the middle stratum. These genetic risk scores may be useful for clinical decision-making when they are comprehensive enough to reliably predict disease risk.³ Catalogs of hundreds of different GWASs can be used to link shared genetic etiologies across different traits, shedding light on new genetic networks. GWASs generate data that can be used in subsequent Mendelian randomization studies. These studies can be a tool to go from association to causation.⁴ Specific SNPs from GWASs can be used as proxies for a phenotype in much larger genotyped but not phenotyped populations. For example, high von Willebrand factor (VWF) levels are associated with risk for venous thromboembolic disease (VTE). But are VWF levels causing the increased risk for the disease? When a set of SNPs from VWF GWASs were used as instrumental variables in a Mendelian randomization study, authors demonstrated that higher VWF levels were causal for VTE.⁵ A common error in Mendelian randomization studies occurs when investigators include variants that have unknown pleiotropic associations. Thousands of GWASs have demonstrated that certain areas of the genome contain common variants that are highly pleiotropic, having significant associations with a variety of traits. In hematology, the ABO locus serves as an excellent example. The ABO gene encodes a glycosyltransferase responsible for the posttranslational modification of hundreds of different plasma proteins. The common “O” allele is a nonsense variant that has no transferase function. Therefore, O/O genotypes have different glycosylation patterns on secreted proteins and altered clearance rates compared with type A or type B genotype. This is a major reason why blood-type O/O individuals have lower VWF levels, are protected from VTE (strong signals in VTE GWAS),⁶ and are more likely to have type 1 von Willebrand disease.⁷ This pleiotropy makes ABO SNPs biologically interesting but a rotten choice as instruments in Mendelian randomization studies because they are associated with altered plasma levels of many other proteins and add confounders.

In the present study, several interesting biological connections are made. Given the increased power of this study compared with a previous GWAS, the investigators uncovered multiple new signals. Strong signals were generated at sites associated with the risk for uterine fibroids, a known risk factor for heavy menstrual bleeding. Other top signals included polymorphisms at loci encoding chorionic gonadotropin, luteinizing hormone, and follicle-stimulating hormone loci, shedding light on how alterations at these loci influence menstrual bleeding.

Perhaps most interesting to the readers of Blood is the apparent protective effect of the factor V (FV) Leiden variant against heavy menstrual bleeding.⁸ Here, FV Leiden (rs6025) reduced the risk for heavy menstrual bleeding with an odds ratio of 0.757 (P = 6.77 × 10⁻³³), the strongest signal in the European population in this study. Although there does not appear to be statistical evidence of balancing selection, it is not hard to understand why the procoagulant FV Leiden might reduce the severity of a bleeding disorder. To further investigate association with other common variants connected to VTE risk, the authors looked for colocalization with previous GWASs performed for VTE risk or plasma VWF/FVIII, fibrinogen, and several other hemostatic factors. Here, they found shared association signals at ABO, STAB2, MPHOSPH9, and DCST2 and CALCRL. Like FV Leiden (rs6025), the STAB2 signal encoded a missense variant associated with increased VTE risk (odds ratio 1.68), but protection against heavy menstrual bleeding (odds ratio 0.85).

GWASs are not without limitations. These studies cannot demonstrate the role of rare genetic variants in a population, and they have well-documented problems detecting the specific “functional” variants or genetic mechanisms responsible for a trait. Those type of functional but rare variants require whole-exome or whole-genome studies. GWASs may not explain a large proportion of the heritability of a disease if it is difficult to accurately phenotype or has a very small common genetic component. But a GWAS like the current study in Blood, which is well powered, phenotyped, and analyzed, still has much to give.

Conflict-of-interest disclosure: K.C.D. declares no competing financial interests.