Social pairing of Seychelles warblers under reduced constraints: MHC, neutral heterozygosity, and age

Lay Summary We manipulated factors that normally constrain the social mate choice of female Seychelles warblers to test if, when less constrained, these females chose partners based on variation at specific immune genes that have been linked to mate choice in other species. We found that older and more generally genetically diverse males were more likely to end up paired with females, but contrary to our predictions, variation at the immune genes did not appear to influence social pairing.


INTRODUCTION
The prevalence and signi cance of precopulatory mate choice remains a keenly debated topic in sexual selection, mainly due to di culties in quantifying the evolutionary costs and bene ts of being "choosy" (Andersson 1994;Kokko et al. 2003). Mate choice can provide both direct bene ts, such as superior nest guarding, and indirect bene ts, such as increased genetic diversity in o spring (Hamilton and Zuk 1982;Andersson 1994). It can manifest in behavioral patterns, such as the choosing of a social mate, and eventually in genetic patterns, such as bias in o spring genotypes (Jennions and Petrie 2000;Consuegra and Garcia de Leaniz 2008). Genes of the major histocompatibility complex (MHC), which determine antigen recognition in the adaptive immune response of vertebrates (Klein 1986), have been a focus of mate choice research for decades (Yamazaki et al. 1976;Jordan and Bruford 1998;Milinski 2006;Kamiya et al. 2014). Di erent MHC genotypes confer di erential pathogen resistance and, therefore, tness to individuals (Briles et al. 1977;Ditchko et al. 2001;Wedekind et al. 2004). This makes the MHC an obvious candidate for genes that underpin the bene ts of mate choice, and several hypotheses have been proposed to explain how individuals may optimize the MHC genotypes that their o spring inherits (Penn and Potts 1999;Milinski 2006;Kamiya et al. 2014).
Various mechanisms have been invoked to explain apparent mating preferences. Individuals may choose mates based on "good genes": either particular bene cial alleles (a classical "good genes" scenario) or on heterozygosity at speci c loci (referred to as "good-genes-asheterozygosity"), or a combination of the two. When calculated over many duplicated loci this heterozygosity can translate into choice for overall diversity, thus the latter scenario can be referred to as a "diversity" mechanism (Kamiya et al. 2014). In an MHC-dependent scenario, choice under a "good genes" or a "diversity" mechanism may be achieved by assessing indicators of condition such as secondary sexual traits that are linked to MHC characteristics (Hamilton and Zuk 1982;Ditchko et al. 2001;Milinski 2006). By choosing a mate with a superior MHC genotype, individuals may obtain direct bene ts such as better provisioning for their o spring (Andersson 1994) as a result of the better condition (immunocompetence) of the mate or indirect bene ts by providing o spring with speci c advantageous alleles and/or increased MHC diversity (Hamilton and Zuk 1982;Reusch et al. 2001). Individuals may also choose mates based on their MHC similarity, in order to obtain an optimal level of MHC diversity in their o spring (Nowak et al. 1992). Mate choice under such a "compatibility" mechanism is based on indirect bene ts. What constitutes a "good match" depends on the complementarity of the maternal and paternal genotypes (Yamazaki et al. 1976;Milinski 2006), although maximizing dissimilarity with a mate may not necessarily be the best Skip to Main Content strategy if there could be negative consequences of too high a level of MHC diversity (e.g., Milinski 2006;Kalbe et al. 2009). Importantly, as well as the ability to assess the MHC characteristics of others (e.g., via olfaction), "compatibility" mechanisms require selfrecognition, so that individuals can gauge their compatibility with potential mates (Penn 2002;Milinski 2006). These mate choice models are normally viewed from the female perspective, though male choice is important in some systems (see Gillingham et al. 2009;Edward and Chapman 2011). Finally, MHC genes may act as markers of relatedness and be used to avoid close inbreeding, rather than to acquire speci c MHC characteristics per se (Brown and Eklund 1994;Penn and Potts 1998).
Numerous studies have investigated MHC-dependent pairing and fertilization patterns providing evidence for each of the di erent hypothesized mechanisms in di erent species (e.g., Penn and Potts 1999;Kokko et al. 2003;Andersson and Simmons 2006;Milinski 2006;Kotiaho and Puurtinen 2007;Griggio et al. 2011;Løvlie et al. 2013;Kamiya et al. 2014 Skip to Main Content new islands as part of the long-term conservation of this species (Richardson et al. 2006;Wright et al. 2014). Signi cantly more males than females were translocated to each island, where a large surfeit of optimal habitat meant that each male was able to establish a highquality territory compared with those inhabited on the original island. Consequently, females had the opportunity to pair up with any one of multiple males, all with high-quality territories. Thus, we provide a relatively unconstrained arena in which to test whether speci c social mating patterns occur and to assess whether such patterns may be driven by active female mate choice or other mechanisms. Given the o spring survival bene ts resulting from mating with a MHC diverse male, we predict MHC class I characteristics will play an important role in any social mate choice in these populations. Under a "diversity" mechanism (i.e., good-genes-as-heterozygosity), we expect females to prefer males with high MHC diversity.
Under a classical "good genes" mechanism, we would expect a link between male pair status and individual MHC class I alleles, if social mate choice is based on the presence of speci c alleles-such as Ase-ua4, which has previously been shown to in uence survival (Brouwer et al. 2010). Under a "compatibility" mechanism females would pair with maximally or optimally (Milinski 2006) MHC dissimilar males. Finally, we test whether stability of these pair bonds is MHC dependent. If "diversity" is important, males of low MHC diversity are more likely to su er the subsequent breaking of any established pair bonds (divorce) than higher diversity males. If "compatibility" is important, pairs that divorce are expected to be more MHC similar than pairs that remain stable.

Study populations
Each translocation was performed as outlined by Richardson et al. (2006) Skip to Main Content across the entire island by multiple catching teams, covering all areas simultaneously (Richardson et al. 2006;Wright et al. 2014

Social mate choice
Territories were mapped on each new island. The average territory size of Seychelles warblers on Cousin is 230 m (Brouwer et al. 2009). With at least 300000 m of suitable habitat on Denis (Richardson et al. 2006) and 390000 m on Frégate (Richardson and Hammers 2011), the spatial and habitat quality constraints that existed in the original population were greatly reduced (if not totally removed) for males released on the new islands. Individuals were located and followed for a minimum of 15min repeatedly (ca. weekly) during the eld period, during which interactions with other conspeci cs were recorded. After a pair bond and territory are established, Seychelles warblers display a speci c repertoire of behaviors that allows clear classi cation of pair status . Previous studies found that although a few pairs may form and nest within days of translocation, most stable pairings only formed after 2 months of rapid switching (Komdeur 1996). Hence, social mate choice was assessed after 3 months and by multiple observations where possible. Pairings were reassessed 1 year after the initial social pairing observations. Divorce was identi ed when both individuals were resighted, with at least 1 individual in a new pair. In cases where one of a pair was not resighted over consecutive eldwork periods, it was assumed that individual had died and the pairing was considered faithful until death. Any subsequent pairing of the surviving individual was not considered the result of divorce and was excluded from analysis, but we acknowledge this is a conservative measure.

Molecular analyses
Samples were genotyped at 30 microsatellite loci following Spurgin et al. (2014). We tested for deviations from Hardy-Weinberg equilibrium and for linkage disequilibrium between loci using GENEPOP v. 4.1 (Raymond and Rousset 1995 Skip to Main Content analysis (RSCA) and the primer sets from Richardson and Westerdahl (2003), following the method of Worley et al. (2008). Each segregating RSCA variant corresponded to a unique 255 base pair amino acid coding sequence (of a total exon length of 274 base pairs, Richardson and Westerdahl 2003). A total of 10 MHC class I variants have been detected in the Seychelles warbler, with individuals possessing between 2 and 8 variants each, suggesting that at least 4 class I loci are ampli ed (Richardson and Westerdahl 2003). Although it is impossible to identify which locus each variant comes from, they are hereafter termed "alleles" for simplicity. This method does not provide a measure of locus-speci c heterozygosity, but an overall estimate of MHC class I exon 3 diversity, which will correlate with heterozygosity across the ampli ed loci. This heterozygosity measure has shown to be an important parameter linked to fertilization patterns and survival in the Seychelles warbler (Richardson et al. 2005;Brouwer et al. 2010). Previous work by Hutchings (2009) detected no variation at MHC class II loci in the Seychelles warbler and so these loci were not assessed in the current study.
MHC-dependent mate choice may be based on only the functional di erences between alleles.
To address this, codons comprising the PBR were superimposed onto the Seychelles warbler sequences (see Richardson and Westerdahl 2003). However, these PBR codons were identi ed in humans (Bjorkman et al. 1987) and while strongly conserved across taxa, may not be completely accurate in the Seychelles warbler. Therefore, another way to determine functional di erences between alleles is to compare between sites that have been identi ed as being  Skip to Main Content other male-female pair combinations possible in each analysis. All comparisons of pairs and dyads were performed using randomization tests (Manly 1997) in MSEXCEL plug-in POPTOOLS v. 3.2 (Hood 2010). This is a useful approach to testing whether observed data di er signi cantly from random expectation. In each instance detailed below, the di erence between observed pairs and all other possible male-female dyad combinations was tested with analysis of variance (Anova). The data were then resampled without replacement (shu ed) and retested 10 times. Estimates of signi cance were calculated as the proportion of repetitions in which the resampled Anova F value was equal to, or exceeded, the test Anova compatibility mechanism may be masked if the optimal similarity between pairs lies close to that observed in random dyads. In this case, observed pairs are predicted to show less variation around this mean similarity than random dyads. Therefore, we tested variance in MHC similarity of observed pairs versus simulated dyads using the same randomization approach. The association between the MHC similarity of a pair and the likelihood of divorce was tested using Mann-Whitney-Wilcoxon tests on divorced versus faithful pairings, using each measure of MHC similarity (S xy , PBR, and PSS). Finally, we tested for di erences in MHC diversity between divorced and stable-pair individuals of each sex using the randomization approach.

Social mate choice
Inbreeding avoidance

Good-genes-as-heterozygosity
The full generalized linear model (Table 1)   Skip to Main Content

Figure 1
View large Download slide Proportions of male Seychelles warblers paired (light gray) and unpaired (dark gray) in 2 newly established populations for each age class, 3-month postrelease: young (<2 years, n = 34) and old (>2 years, n = 33). Data are combined from both translocated populations.

Classical good genes
Fisher's Exact tests showed 1 allele, Ase-ua3, had a weak negative relationship with male pair status (i.e., the presence of Ase-ua3 meant lower probability of being paired) before correction for multiple testing (P = 0.03, odds ratio: 0.26), but there were no signi cant interactions between any of the individual MHC class I alleles and male pair status after Bonferroni correction for multiple testing (P crit = 0.005).

View large Download slide
Frequency histograms of individual standardized heterozygosity (upper) and MHC class I diversity (lower) in paired (n = 37) and unpaired (n = 30) male Seychelles warblers in the newly established populations 3-month postrelease.
Divorced males appeared to be less MHC diverse than those in stable pairings. This pattern is consistent with the previous work on MHC-dependent EPP (Richardson et al. 2005) (Richardson et al. 2005) suggests an interaction between MHC genes and fertilization patterns that is important in maintaining MHC diversity in this species. However, it may be that the historical constraints on, and costs associated with, social mate choice preferences (Kokko et al. 2003) have prevented the evolution of MHC-dependent social mate choice in the Seychelles warbler, with alternative mechanisms such as age-dependent pairing, promiscuity (driven by whatever mechanism), or postcopulatory selection (e.g., Løvlie et al. 2013) taking precedence. This study highlights that predicting the occurrence of a sexual selection mechanism, that is, MHC-dependent social mate choice, based on seemingly related observations, that is, the MHC-dependent EPP patterns and survival, is not straightforward and that the EPP pattern (Richardson et al. 2005) may re ect mechanisms other than active female choice. There are many potential sexual selection mechanisms that may evolve separately or in concert such as precopulatory or postcopulatory mate choice, coercion, or promiscuity (Andersson 1994;Andersson and Simmons 2006). Understanding how, when, and why particular mechanisms evolve, while others do not, or are observable, while others are not, requires an understanding of the constraints acting on any given species or population. We encourage future studies to focus on how patterns and mechanisms linked to sexual selection may be in uenced by the constraints acting within a system.