Introduction

Although individual humans share some 99.6% of base pairs with others,[1] the size of the human genome dictates that there are nonetheless some 24 million base pairs that differ between individuals. This degree of variation offers substantial opportunity for the manifestation of heterozygote advantage, the phenomenon of biological advantage of mixed over pure breeds.

Multilocus heterozygosity may act in two ways. First, heterozygosity at individual loci may provide an advantage relative to homozygote wild-type or variant alleles, sometimes referred to as overdominance. The best known is variation in genes related to erythrocyte function, many of which confer resistance to malaria.[2] In this setting, homozygotes are disadvantaged by either susceptibility to malaria (for wild-type alleles) or clinical syndromes as sickle cell anemia, thallasemia, and glucose-6-phosphate dehydrogenase deficiency. There is also evidence that heterozygosity at specific loci may be associated with perceived attractiveness of humans to members of the opposite sex.[3,4] Second, heterozygosity as a manifestation of outbreeding reduces risk of homozygosity in disadvantageous alleles, leading in the most adverse situations to recessive diseases.

An important caveat in studies of human heterozygosity is the limited evidence that heterosis confers protection against chronic disease, which is less likely to exert selection pressure than do acute infectious diseases or attractiveness to mates, yet is the dominant cause of mortality in developed nations. Heterozygosity at selected loci has been associated with extreme longevity, but these have not been replicated and it is uncertain that they exemplify generalized heterosis.[5,6] In a genome-wide comparison of populations in the Dalmation Islands,[7] Campbell and colleagues found that overall heterozygosity across populations was associated with lower blood pressure and total cholesterol, but individual heterozygosity was not measured, raising the strong possibility that this reflects population stratification. Among individuals in the Framingham Heart Study, multilocus heterozygosity at 706 single nucleotide polymorphisms (SNPs) was associated with higher blood pressure, but not with several other cardiovascular traits measured simultaneously.[8] To our knowledge, no genome-wide assessment among individuals has evaluated whether heterozygosity confers an advantage on common chronic diseases among adults.

To elucidate whether heterozygote advantage extends to coronary heart disease (CHD), the most common cause of death in the United States,[9] we determined the association of variation in genome-wide single nucleotide polymorphisms (SNPs) with both risk of incident CHD and with several CHD risk factors in two well-characterized parallel cohorts of men and women.

Materials and Methods

Results

Following cleaning of genotyping data, we included 435 women with incident CHD and 931 control women and 435 men with incident CHD and 878 control men. Table 1 demonstrates the associations of genome-wide heterozygosity with risk of CHD among women and men. In both sexes, we observed no relationships of CHD with overall number of heterozygous loci, number of heterozygous loci within genes, or number of heterozygous nonsynonymous loci. These results were essentially identical when the proportions of heterozygous SNPs overall, within genes, or at nonsynonymous loci were used, or when we restricted to individuals below 65 years of age at blood draw.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Mean (±SD) numbers of heterozygous loci among CHD cases and controls and odds ratios (95% confidence interval) for CHD associated with heterozygous loci.

https://doi.org/10.1371/journal.pone.0124847.t001

Table 2 shows the relationship of heterozygosity with levels of selected CHD risk factors. There were only two modestly significant relationships in men (LDL among all SNPs and SNPs within genes and fibrinogen among nonsynonymous SNPs; p = 0.03–0.04) and none that were consistent in both sexes, regardless of which measure of heterozygosity was used. One possible exception was the association of overall heterozygosity and heterozygosity within genes with E-selectin (p = 0.03), which was only measured in women and hence could not be confirmed or refuted in men.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Association of CHD risk factors with overall heterozygous loci, loci within genes, and nonsynonymous SNPs.

https://doi.org/10.1371/journal.pone.0124847.t002

Discussion

In this analysis of two prospective case-control studies, we found no consistent evidence of heterozygosity advantage in risk for CHD. The number or proportion of heterozygous loci was unrelated to either incident CHD or to circulating CHD risk factors.

The hypothesis of overdominance associated with heterozygosity at individual loci is attractive and has clear support in a few well-described instances. Certainly, the best recognized of these is the selection pressure exerted by P. falciparum, which infects erythrocytes. The widespread prevalence of falciparum malaria appears to have selected for hemoglobin variants that, while disadvantageous on their own, appear to mitigate the consequences of plasmodial infection.[2] Another proposed manifestation of overdominance may exist in perceived attractiveness, even in non-human primate species.[18] For example, in one study, women rated the faces of men heterozygous at three major histocompatibility complex loci as more attractive than the faces of homozygotes, regardless of the level of genetic similarity between female raters and male subjects.[19]

At the same time, evidence that generalized heterozygosity provides an advantage for risk of chronic disease, and particularly for CHD, is sparse. The Framingham Heart Study reported a positive association of heterozygosity at 706 loci with blood pressure and two related echocardiographic parameters among approximately 1000 participants, but no association with any circulating risk factors or with BMI; no association with risk of CHD per se was tested.[8] Because we did not have standardized blood pressure or echocardiographic measurement in our cohorts, we cannot directly replicate their findings, but our results do confirm the lack of association with a large number of putative risk factors and extend these results to clinical CHD endpoints. Why blood pressure might differ from other risk factors in its association with heterosis is necessarily speculative, but it is striking that the observed association in Framingham was positive—that is, greater heterozygosity conferred a phenotype of higher blood pressure that might well be protective in the setting of acute primordial stressors like trauma or dehydration, but could be harmful for conditions like CHD that manifest in late adulthood. Thus, we find no clear evidence that multilocus heterozygosity is associated with a lower risk of CHD, the most common cause of mortality in developed nations. Its associations with cancer, dementia, and other chronic diseases remain to be tested, although recent work suggests it may influence schizophrenia.[20]

We found few significant relationships—about the number expected from chance alone—and unsurprisingly these generally did not appear consistent across cohorts. The exception of E-selectin is interesting, as its genetic variation is strongly related to ABO blood group genotype[21] and hence could also be subjected to plasmodial selection pressure.[22] However, there appears to be little ancestral selection pressure in erythrocyte traits among populations of European ancestry and hence even this possible association clearly requires confirmation in other cohorts.

Two differences of our study with more common GWAS designs warrant mention. First, although we performed standard chip-based genotyping as in other GWAS designs, we used a nested case-control design for efficiency within the very large parent cohorts. This has the advantage of preserving power to detect associations with CHD with efficiency and produces unbiased and reasonably precise estimates of relative risk comparable to what would be achieved with a GWAS of the full cohorts of NHS or HPFS (because we genotyped all of the cases that would have been present in the full cohort). While power to examine associations on biomarkers was not as efficiently maintained (because the case-control design explicitly tests a single case definition or endpoint), even those associations had ample precision in our analyses because of their continuous distribution.

Second, we did not conduct an unbiased effort to identify single, specific loci that influence risk of CHD or its risk factors, as occurs in most GWAS. Rather, we tested a specific, global hypothesis about the benefit of widespread variation across the genome. Had our hypothesis been proven correct, the natural next step would be to search for specific loci at which that hypothesis was borne out, more similar to standard GWAS efforts. As a consequence of our single underlying hypothesis, however, our statistical analyses use conventional levels of statistical significance and not the extreme thresholds more familiar in GWAS.

Our study has strengths, but important limitations as well. In both studies, participants who provided blood were middle aged or older, and hence we did not study cases of CHD that occurred early in life, which may be most closely associated with genetic factors, although our findings in those below 65 years of age were also null. We also studied a geographically diverse sample but occupationally homogeneous sample, in whom ascertainment of CHD is reliable but who may not generalize readily to the full US population. We performed genome-wide scans, and hence had an unbiased and comprehensive assessment of variation in SNPs, but we cannot exclude the possibility that heterozygosity at specific loci (e.g., major histocompatibility complex) or in other forms of genetic variation (e.g., copy number variants) favorably influence CHD or its risk factors. Likewise, we studied CHD and its risk factors but cannot rule out associations of heterozygosity with other chronic diseases nor with overall survival.

Power is an important concern in studies of heterozygosity. Number or proportion of heterozygous SNPs are an imperfect estimate of inbreeding,[23] potentially requiring large samples to overcome. Further, we were limited to case-control samples of moderate size with narrow ranges in heterozygosity, illustrated by the small standard deviation in number of heterozygous loci relative to the number genotyped. Nonetheless, our point estimates demonstrated no association whatsoever with CHD risk, and our study compares favorably in size to the Framingham Heart Study, which showed a significant association with blood pressure. Studies in more inbred populations are needed to rule out associations with lower levels of heterozygosity; at the extreme, consanguinity, which maximally suppresses heterozygosity, has been associated with CHD.[24] Equally, studies in populations in other ethnicities are warranted, as populations of African descent in particular are apt to exhibit variability at substantially larger numbers of loci.

In conclusion, in two cohort studies, genome-wide measures of heterozygosity were not associated with risk of incident CHD nor with CHD risk factors. If generalized heterozygosity provides any fitness advantage in humans, it seems unlikely to do so for chronic disease phenotypes like CHD at the levels of outbreeding seen in populations like these.