Balancing selection drives maintenance of genetic variation in Drosophila antimicrobial peptides

Genes involved in immune defense against pathogens provide some of the most well-known examples of both directional and balancing selection. Antimicrobial peptides (AMPs) are innate immune effector genes, playing a key role in pathogen clearance in many species, including Drosophila. Conflicting lines of evidence have suggested AMPs may be under directional, balancing or purifying selection. Here, we use a case-control gene approach to show that balancing selection is an important force shaping AMP diversity in two species of Drosophila. In D. melanogaster, this is most clearly observed in ancestral African populations. Furthermore, the signature of balancing selection is even clearer once background selection has been accounted for. Balancing selection also acts on AMPs in D. mauritiana, an isolated island endemic separated from D. melanogaster by about 4 million years of evolution. This suggests that balancing selection may be acting to maintain adaptive diversity in AMPs in insects as it does in other taxa.


15
Pathogens exert strong selective pressures on their hosts, both in terms of individual 16 fitness and the evolutionary trajectory of populations and species. Co-evolutionary 17 dynamics of hosts and pathogens results in continual selection for adaptive improve-18 ments in both players, often referred to as a co-evolutionary arms race (1, 2, 3). As 19 a consequence, genes involved in immune defense tend to undergo strong positive se-20 lection, such that they are among the fastest evolving genes in the genomes of many 21 hosts (4, 5, 6, 7, 8). 22 However, resistance mutations may not always become fixed. Balancing selec-23 tion is the process whereby polymorphism is adaptively maintained within genes over 24 extended timescales, sometimes described as trench-warfare dynamics (9). Several 25 processes are thought to contribute to balancing selection (reviewed in (10)). These 26 include heterozygote advantage, whereby individuals heterozygous at a given locus 27 have a fitness advantage over either homozygote; negative frequency dependent se-28 lection, whereby the benefit of an allele increases the rarer it is in a population; and 29 selection varying in a context-dependent manner, for example at different spatial or  (14); environmental variables (15); and 38 demographic factors such as gene flow and bottlenecks (16). 39 The dynamic selective pressures exerted by pathogens promote balanced poly-40 morphism of host immune genes in several cases. Perhaps the best documented ex-41 ample is the major histocompatibility complex (MHC) in vertebrates (reviewed in 42 (17,18,19,20)). Individuals tend to be heterozygous at MHC loci, and large numbers including Oas1b in mice (32), OAS1 in primates (33, 34) and TRIM5 in humans (35) 48 and primates (36). Balancing selection also appears to play a role in the evolution 49 of antimicrobial peptides (AMPs). AMPs are effectors of innate immunity that are 50 strongly induced upon infection (37,38). They tend to be membrane active (39, 40), 51 with a direct role in killing and/or impeding the growth of pathogens (41, 42). Bal-52 ancing selection has been implicated as a driver of AMP evolution in a diverse array of 53 species including birds (43,44), amphibians (45), fish (46), molluscs (47) and humans 54 (48,49 88 whereas FR and DGRP are derived populations. 89 We calculated three population genetic statistics: Watterson's θ (the sample size   Given the apparent differences in selection between AMPs and the genome averages  the DGRP (derived populations, Figure 1A, Table 1).

115
To identify if these signatures of balancing selection are unique to AMPs, or con-116 sistent across all immunity genes, we repeated all tests, this time for all non-AMP 117 immunity genes. We found very little evidence of balancing or directional selection 118 across the remaining immunity genes, with differences closer to zero (Supplementary

168
For other immunity genes, the differences from controls are primarily negative for π,

169
Watterson's θ and Tajima's D, suggesting directional selection may be acting on these temporally. There is some evidence for both seasonal (66) and spatial (67) variation 218 in selection pressure on AMPs. However, evidence for AMP specificity against par-219 ticular pathogens, especially different naturally occurring alleles of the same AMP, is 220 currently rare (but see e.g. (63, 86, 87, 88)).

221
Alternatively, AMP variation might be maintained because AMP alleles that are 222 more effective against pathogens also carry a higher autoimmune cost. This "autoim-223 mune hypothesis" states that more effective AMP alleles should be common during  For each analysis, (per population, including and excluding SD chromosomes) we 252 then resampled to find the average difference in scores between case and control genes.

253
Case genes were either a) AMPS, or b) immunity genes (using gene ontologies pre-254 viously described (58)). For each gene in these categories, we randomly sampled a 255 control gene within 100kbp upstream or downstream, that was no more than ten 256 times larger than this gene and not another gene in the given category (AMP or 257 immunity). We then found the average difference (∆) in each measure for the case 258 (AMP/immunity) group and the control group such that: where X Case represents the chosen gene, X Control represents the randomly sampled 260 control gene and n accounts for the number of genes in the group. We then repeated 261 this 10000 times to obtain an empirical distribution of the differences. 262 We employ this method to control for genomewide variation in recombination   for each population, also the mean for each statistic for all non-AMP immune genes.