Heterozygosity–fitness correlations and associative overdominance: new detection method and proof of principle in the Iberian wild boar



Aurelio F. Malo and Tim Coulson, Fax: +44(0)20 75942237; E-mails: a.malo@imperial.ac.uk and t.coulson@imperial.ac.uk


Heterozygosity-fitness correlations (HFC) may result from a genome-wide process — inbreeding — or local effects within the genome. The majority of empirical studies reporting HFCs have attributed correlations to inbreeding depression. However, HFCs are unlikely to be caused by inbreeding depression because heterozygosity measured at a small number of neutral markers is unlikely to accurately capture a genome-wide pattern. Testing the strengths of localized effects caused by associative overdominance has proven challenging. In their current paper, Amos and Acevedo-Whitehouse present a novel test for local HFCs. Using stochastic simulations, they determine the conditions under which single-locus HFCs arise, before testing the strength of the correlation between the neutral marker and a linked gene under selection in their simulations. They used insights gained from simulation to statistically investigate the likely cause of correlations between heterozygosity and disease status using data on bovine tuberculosis infections in a wild boar population. They discover that a single microsatellite marker is an excellent predictor of tuberculosis progression in infected individuals. The results are relevant for wild boar management but, more generally, they demonstrate how single-locus HFCs could be used to identify coding loci under selection in free-living populations.

It is over 30 years since the first correlations between heterozygosity at allozyme markers and fitness-related traits were detected in natural populations (Mitton & Grant 1984). These results contributed significantly to the neutralist–selectionist debate and helped identify the genetic causes of inbreeding depression and heterosis. The discovery of correlations between heterozygosity at neutral markers, including microsatellites, led to a renewed interest in the cause of heterozygosity-fitness correlations (HFC). Various possible causes of HFCs were identified: (i) inbreeding depression, (ii) overdominance, and (iii) associative overdominance. Given the markers are putatively neutral, overdominance could be ruled out. Because fitness was negatively correlated with inbreeding coefficients in livestock, the cause of HFCs at neutral markers was attributed to inbreeding depression. This assumes that neutral marker heterozygosity is a good predictor of genome-wide homozygosity and consequently the coefficient of in breeding (f).

Some studies have cast doubt on the ability of heterozygosity measured at a small panel of markers to capture inbreeding (Balloux et al. 2004; Slate et al. 2004), leading to the hypotheses that published HFCs predominantly represent type I statistical errors or publication bias. However, Hansson et al. (2001) provided convincing empirical evidence for associative overdominance, which has led to a renewed interest in the cause of HFCs. What we currently do not understand is how likely any individual neutral marker is to carry the signature of over- dominance at a linked locus.

Amos and Acevedo-Whitehouse's (AA, 2009) current paper was stimulated by previous results they had found in a population of wild boar from Southern Spain (Acevedo-Whitehouse et al. 2005). This work, based on 32 microsatellite markers, suggested that both genome-wide inbreeding and single-locus effects influenced the risk of tuberculosis infection, but that only single-locus effects affected the progression of the disease. In wild boar, individuals become infected with bovine tuberculosis (bTB), which can be detected by the presence of caseocalcareous tubercules and miliary lesions, but not all individuals progress to develop contaminated organs. In their new paper, AA conduct further analyses of the wild boar data after having conducted simulations to investigate the conditions under which single-locus HFCs are likely to occur and to be detected under a range of mutation and recombination rates and strengths of selection (Fig. 1).

Figure 1.

 Representation of how HFCs are detected in the real world (left) and how the strength of the association between the marker and the locus under balancing selection has been tested in the simulations performed by the Amos and Acevedo-Whitehouse (right). Single headed arrows indicate causation effects and two-headed arrows stand for correlations. Black straight lines represent chromosomes. Note that in the simulated data, locus genotype is identified with individual fitness, not allowing stochastic variation in the form of environmental effects to mediate the relationship between the genotype and fitness. Waa and Wab stand for the fitness of the homozygote and heterozygote, respectively. The most influential factors and parameters building up linkage in the simulations are marked in bold.

AA first used their simulations to demonstrate three points: first, that intermediate mutation rates produce the highest linkage disequilibrium (LD) between a locus under selection and a neutral marker. Second, as expected, low recombination rates generated the highest LD. Third, that different strengths of selection did not impact the marker-gene association. AA then used data from their simulation to examine how best to statistically detect associative overdominance at a neutral locus and developed a novel method to do just this. This method involves estimating the maximum possible strength of association from the data and comparing this quantity with a distribution of the same chi-squared statistic estimated from randomized sets of the data. If the chi-squared value describing the maximum possible strength of association obtained from the data is beyond the 95 percentiles of the distribution obtained from randomization, then the observed HFC is considered unlikely to have arisen by chance. AA found that their test performed well at identifying HFCs at individual loci from their simulation, so they next used their method to examine HFCs in their wild boar population. They find that neutral marker heterozygosity is negatively associated with the probability of bTB infection suggesting that inbred animals are more susceptible to bTB than those that are outbred. AA did not identify evidence for localized effects associated with bTB infection, but they did find effects for bTB progression. Specifically, progression of the disease was predicted by the genotype at the microsatellite marker SW787. Interestingly, the number of copies of allele 156 determines susceptibility: individuals bearing one single copy of this allele have fourfold higher probability of resisting disease progression (P = 0.2) than those boars bearing two copies of allele 156 (P = 0.045). The authors go even further and identify a candidate gene associated with inflammatory responses to bacterial infections which is located 8 cm from the microsatellite locus.

AA achieve three things. First, they have contributed to the debate on the cause of HFCs by providing a method to help distinguish local effects from genome-wide processes. Second, they have demonstrated how this method can be used to identify genes associated with disease using their method. Third, they have provided useful insights on the dynamics of disease progression of an economically important disease in an ecologically important species. Given concerns about the spread of wild boar into many parts of Europe, any improved understanding on their genetics and disease ecology is welcome.

As with any novel research, AA's paper raises questions. Their simulations were kept deliberately simple, would the inclusion of additional realism influence results? Similarly, if further information on sex, age, habitat or boar density were to be included in the statistical models, would additional insight on the prevalence of individual locus effects and the modus operandi of HFC be unearthed? These questions will be addressed in time, and there is no a priori reason to expect conclusions to be altered when the work is done. So in conclusion, AA have provided a novel and stimulating paper on an important topic and their results will undoubtedly stimulate further research and reignite the debate on the cause of HFCs in the wild.

Aurelio F. Malo is a biologist interested in the determinants of fitness in wild and captive mammal populations. Tim Coulson is a population biologist (http://www.bio-demography.org/).