Computational protocol: Contrasted Patterns of Selection on MHC-Linked Microsatellites in Natural Populations of the Malagasy Plague Reservoir

Similar protocols

Protocol publication

[…] For non-duplicated MHC-linked and presumed neutral loci, we checked for Hardy-Weinberg equilibrium and genotypic linkage disequilibrum between pairs of loci with Genepop v. 4 . In both cases, we corrected for multiple testing by the false discovery rate (FDR) approach, with Q-value software (available from http://genomics.princeton.edu/storeylab/qvalue/); the tuning parameter λ was fixed at 0 . As some loci exhibited heterozygote deficiencies, we used Micro-checker 2.2.3 to evaluate whether it could be explained by the occurrence of null alleles. We then used the software Freena (INRA Montpellier website, available www.montpellier.inra.fr/URLB, accessed 2012 February 7) to verify that mean null allele frequencies were low at these loci, as high null alleles frequencies can impact genetic differentiation estimations .Under the hypothesis of overdominance (or heterozygote advantage, see introduction), an excess of heterozygotes would be expected at MHC loci. We thus tested for Hardy-Weinberg disequilibrium (H1: heterozygote excess) within each studied population for the two non duplicated MHC-linked loci using Genepop program. The FDR approach was also used to correct for multiple testing.For the “Madagascar” dataset, we expected the level of genetic differentiation to be higher for MHC-linked microsatellites than for neutral loci within population pairs. We evaluated the deviation from neutrality for all non-duplicated MHC-linked and presumed neutral microsatellites, using two approaches for the detection of selection, based on the FST values for each of the four population pairs. We first used the method of Beaumont & Nichols , as implemented in Fdist2 (available at http://www.rubic.rdg.ac.uk/~mab/software.html). This approach uses an island model and simulates the distribution of FST conditioned on heterozygosity, under the null hypothesis of drift and migration only. We carried out 50,000 simulations of two demes, assuming a stepwise mutation model. We then used the Vitalis et al. method, implemented in DetSel software . This method performs coalescent simulations under a divergence model in which an ancestral population splits into two isolated populations. The simulations were performed with a stepwise mutation model and a mean mutation rate μ = 0.0005. We used several different values for ancestral population size (Ne = 1,000, 10,000, 100,000), and the time since divergence was set at 1,200 generations (because R. rattus probably colonized the central highlands about 700 years ago ). For each pairwise analysis, we performed a total of 1,000,000 simulations, with each set of nuisance parameter values representing a third of the total simulations. We applied a minimum allele frequency of 0.01 to the observed and simulated data.We then used a method similar to the one described in Neff & Fraser to empirically compare the pairwise FST value for each MHC-linked (duplicated and non-duplicated) locus with the global FST for the neutral loci within the four population pairs. For each of the five MHC-linked markers, as well as for the global neutral dataset (13 loci), 1000 datasets were obtained by bootstrapping over individuals. For all the datasets, pairwise FST were computed within each population pair from allelic frequencies (see also , ) using the package Arlecore from Arlequin v 3.5 software . Note that allelic frequencies cannot be directly estimated for the duplicated microsatellites (data available only consist in allele presence/absence, genotypes may be unknown). Allelic frequencies were thus inferred from the number of individuals carrying a given allele divided by the total number of alleles observed in a population (see also , ). The proportion of bootstrap replicates in which the FST of MHC-linked markers were higher or lower than neutral FST was used as p-value for the null hypothesis that one of MHC-linked markers had higher or lower FST than the neutral markers (see also ).For the “Betafo” dataset, we expected the level of genetic differentiation to be lower for MHC-linked microsatellites than for presumed neutral loci, provided that the selection pressure exerted by plague was uniform over the whole area . We tested this hypothesis for non duplicated microsatellites, with Fdist2 (with the same parameters as for the analysis of the Madagascar dataset, see above). We did not use DetSel on the “Betafo” dataset: it does not conform to the model of a single ancestral population diverging into two isolated populations with no migration between them, as it corresponds to a small scale metapopulation sampling design.A significant isolation by distance (IBD) pattern was found for presumed neutral microsatellites within the “Betafo” dataset (as already reported ). For each MHC-linked locus, we performed partial Mantel tests , which evaluated the correlative link between the pairwise FST at the MHC locus and geographic distance while keeping constant differentiation at neutral microsatellites (excluding those identified by FDist2 as corresponding to selection signature in the “Betafo” dataset). Under the hypothesis of strong selection pressures acting on MHC loci, we would expect a lower isolation by distance pattern than under neutrality for these loci. Pairwise FST values were estimated for each locus with Genepop v. 4 for non duplicated loci, or from allelic frequencies for duplicated loci. Partial Mantel tests were performed with Fstat v. 2.9.3.2 , with 20,000 permutations. […]

Pipeline specifications

Software tools Genepop, Arlequin
Application Population genetic analysis
Organisms Rattus rattus
Diseases Epilepsy, Tonic-Clonic, Rodent Diseases, Yersinia Infections