b2. Frequency-dependent selection: are allele effects inconsistent
across lakes?
If stickleback and their parasites engage in a Red Queen race style
co-evolution, the efficacy of any single allele will shift through time.
For hosts and parasites with limited dispersal, physically disconnected
sites are unlikely to be in the same phase of the arms race. As a
result, a given MHC allele may have different effects on a given
parasite from one site to the next: effective against defense in some
places/times, ineffective or susceptible at others. Alternatively, if
the same allele has similar effects on the same parasite across
different lakes (without gene flow), such fluctuating
frequency-dependent selection is unlikely. To test these alternatives,
we first identified the moderately prevalent MHC-parasite combinations
that were present in more than one site. For each qualified combination,
we used a generalized linear model with negative binomial distribution
to examine if the infection intensity of the parasite is influenced by
the MHC allele, sampling site, and the interaction between site and MHC
allele (this differs from the GLMs described in b1, which were done
separately for each sample site). We corrected p values for
multiple-comparison with BH method. We used the anova function in
R to perform an analysis of deviance for each regression model. It
reported the reductions in the residual deviance as each term of the
formula was added in turn. We evaluated whether, across many
parasite-allele combinations, more variation was explained by the focal
allele’s main effect (implying consistent protection across
populations), or allele × population interactions (inconsistent
protection). Parasites transmitted by birds could spread in a larger
spatial scale, so they are less likely to be engaged in evolutionary
arms race. We did a Chi-squared test to test whether the parasite taxa,
which have birds as final hosts, were more likely to be found in the
models with significant main effect than in the models with significant
allele x population effect.