Where are the beachmasters? Unexpectedly weak polygyny among southern elephant seals on a South Shetland Island

Intraspecific variation in animal mating systems can have important implications for ecological, evolutionary and demographic processes in wild populations. For example, patterns of mating can impact social structure, dispersal, effective population size and inbreeding. However, few species have been studied in sufficient detail to elucidate mating system plasticity and its dependence on ecological and demographic factors. Southern elephant seals (Mirounga leonina) have long been regarded as a textbook example of a polygynous mating system, with dominant ‘beachmaster’ males controlling harems of up to several hundred females. However, behavioural and genetic studies have uncovered appreciable geographic variation in the strength of polygyny among elephant seal populations. We therefore used molecular parentage analysis to investigate patterns of parentage in a small satellite colony of elephant seals at the South Shetland Islands. We hypothesised that dominant males would be able to successfully monopolise the relatively small numbers of females present in the colony, leading to relatively high levels of polygyny. A total of 424 individuals (comprising 33 adult males, 101 adult females and 290 pups) sampled over eight years were genotyped at 20 microsatellites and reproductive success was analysed by genetically assigning parents. Paternity could only be assigned to 31 out of 290 pups (10.7%), despite our panel of genetic markers being highly informative and the genotyping error rate being very low. The strength of inferred polygyny was weak in comparison to previous genetic studies of the same species, with the most successful male fathering only seven pups over the entire course of the study. Our results show that, even in a species long regarded as a model for extreme polygyny, male reproductive skew can vary substantially among populations.

Behavioural studies at South Georgia (Laws 1956, McCann 1980, Modig 1996, the Falkland Islands 86 harems of many tens to over a thousand breeding females on densely packed beaches. These dominant 89 males tend to be older and larger than subordinate males, who attempt to gain mating opportunities by 90 entering harems to secure 'sneaky' matings, intercepting females as they leave the harems to forage at 91 sea (McCann 1981), and potentially by mating with females at sea (De Bruyn et al. 2011 The most southerly breeding sites for southern elephant seals are located in the South Orkney Islands 111 and the South Shetland Islands (Laws 1956). These populations could potentially have different patterns 112 of male reproductive skew than previously studied populations for two main reasons. First, at Signy 113 Island in the South Orkneys, around 70 pups per year were recorded as having been born in 4-6 harems 114 (Laws 1956). These small harem sizes will limit the maximal reproductive success of harem holders 115 simply because these males will have access to fewer breeding females. Second, in particularly cold 116 years, breeding takes place on fast ice where space is unrestricted, leading to even greater female 117 dispersion and potentially lower male reproductive skew (Laws 1956). However, genetic studies have 118 not been conducted in these localities, so the realised degree of polygyny is unknown. Conversely, 119 smaller harems might in fact be easier for dominant males to monopolise, allowing beachmasters to 120 attain relatively high reproductive success at low population densities. This was previously shown for point for breeding females migrating northwards to larger colonies from southerly pelagic foraging 132 grounds closer to the Antarctic Front, with some females remaining there to breed (Krzemiński 1981, 133 Laws 1994. In warmer, more favourable years, many females breed at Cape Sherriff, whereas in 134 colder, less favourable years, the accumulation of sea ice and snow causes fewer animals to come 135 ashore. Consequently, the size of the breeding population varies appreciably from year to year. Similar 136 to the Valdés Peninsula, the beach at Half Moon Bay represents a large expanse of uninterrupted 137 breeding habitat that is sparsely occupied by southern elephant seals in the breeding season. In most 138 years, there is only one harem, but in years of high pup production there can be as many as five harems. hypothesised that the mating system of southern elephant seals at Cape Shirreff would be highly 147 polygynous, with a small proportion of males monopolising reproductive success. Both the degree of 148 reproductive skew between competing males and the size of the harem they compete over are likely to 149 contribute to dominant male reproductive success. We therefore also predicted that the maximum 150 annual reproductive success of harem holders may be lower at Cape Shirreff than at high density 151 breeding sites, due to the smaller sizes of the harems. 152

Study site and sample collection 154
This study was conducted at a small breeding colony at Half Moon Beach, Cape Shirreff, South Fieldwork for this project was conducted during the Austral spring (October to December inclusive) of were lost in transit from Cape Shirreff back to the USA. In most seasons, adults were sampled on the samples were immediately placed in 95% ETOH and kept frozen (-20°C) until analysis. Sampled adults 181 were marked using hair dye whenever the bulk of sampling efforts spanned more than a day (sampled 182 individuals were otherwise identified by their biopsy punch mark). Marks persisted throughout the 183 season but were lost at the moult so were not informative from one season to the next. Most individuals 184 were sampled within a few days of the first visit to the colony but the site was checked for unsampled 185 individuals thereafter during broad pinniped censuses at Cape Shirreff and / or to retrieve 186 instrumentation. 187

Genetic data analysis 207
We removed 33 samples that were genotyped at <15 loci and then checked the dataset for duplicate 208 genotypes (representing resampling events) using the R package "poppr" (Kamvar, Tabima & from Hardy-Weinberg equilibrium (HWE) based on 10.000 Monte Carlo permutations using the R 214 package "pegas" (Paradis 2010). The resulting p-values were corrected using the Bonferroni correction 215 in the p.adjust function of the "stats" R-package. Finally, we tested for population substructure by 216 implementing a principal component analysis (PCA) of the dataset using "adegenet". Because PCA can 217 be sensitive to missing data, any missing genotypes were imputed and the allele frequencies were 218 transformed by centring and scaling the data. The probability of identity and the exclusion probability 219 were calculated using "GenAIEx" version 6.5 (Peakall & Smouse 2005). 220

Parentage analysis 221
Parentage analysis was conducted for the 290 pups within "COLONY" 2.0.6.6 (Jones & Wang 2010). 222 All sampled adult males (33 individuals) and females (101 individuals) were included as potential 223 parents and we did not specify any known parents or sibships. We set weak priors that the true mother 224 and father were in the candidate lists as 0.5 and 0.2 respectively and specified a polygynous mating 225 system in a diploid species. We used a medium run size, which provides high confidence parentage 226 assignments without exceeding practicable run-times. Parentage assignments were accepted with ≥ 227 0.95 probability. An advantage of COLONY is that, instead of using pairwise comparisons to assign 228 parentage, it uses a full-pedigree maximum likelihood approach, which considers the likelihood of the 229 entire pedigree structure and allows the simultaneous inference of both parentage and sibship. This 230 allows the program to assign genetically unsampled individuals as parents, which provides additional 231 insights into mating patterns despite the incomplete sampling of parents. Colony has been shown to be 232 highly accurate and is the only program available that can use genetic information to assign offspring 15-20 microsatellites (for details, see Table 1 and Table S1). The loci carried on average 9.3 alleles 239 and none deviated significantly from Hardy-Weinberg equilibrium after table-wide Bonferroni 240 correction (Table S1). The genotyping error rate, calculated by independently repeat-genotyping 96 241 samples, was low at 0.003 (0.3%) per allele or 0.006 (0.6%) per genotype. The probability of identity of paternity. First, we focused on the small number of genotyped fathers assigned to 31 pups, and 247 second, we analysed the best configuration fathers inferred by COLONY. Best configuration fathers 248 include genotyped fathers together with "hypothetical genetic fathers" inferred using maximum 249 likelihood to explain the offspring genotypes. Using the offspring genotype and the genotype of the 250 assigned mother together with data on allele frequencies, COLONY is able to assign unsampled fathers 251 to pups, including assigning the same unsampled father to full or paternal half siblings. Best 252 configuration parentage assignments therefore allow us to investigate likely patterns of paternity, even 253 when many fathers remain unsampled. When analysing best configuration fathers, we conservatively 254 restricted our analysis to the 108 pups for which maternity could be assigned, thereby excluding pups 255 with neither parent sampled. 256 At least one pup was assigned to 75% of genotyped females (n = 76 mothers) and 36% of genotyped  reproductive success in individuals that consistently bred at the study site, we identified a subset of 276 'core' genotyped individuals that were associated with the beach across multiple years based on genetic 277 recaptures and parentage assignments (Figure 4). These comprised 11 males (33% of sampled males), offspring produced by core individuals (Table S2); the most successful male in each year fathered 281 between 17% and 60% (mean = 32%) of these pups, equivalent to a mean of 2.6 pups per year (range 282 = 1-5). Furthermore, PCA did not identify any obvious genetic differences between core and transient 283 individuals (Figure 5), or between pups that were and were not assigned a father (Figure 6), implying 284 that they most likely originate from a single panmictic population. 285 Finally, although core breeders returned to the colony across multiple years, they were not always 286 present every year. Consequently, many comparisons between candidate males and offspring will 287 involve males who were not present in the colony during the year a given pup was conceived, while 288 some pups may also have been conceived elsewhere if their mother was absent from the colony in the 289 conception year. We therefore restricted our analysis to include only comparisons involving mothers 290 who were known to be present in the colony during the year of conception. Although our sample size 291 was reduced to only 20 pups, we did not find evidence for strong polygyny (Table S3) variation in the environmental potential for polygyny. We therefore investigated patterns of parentage 301 at a southern elephant seal breeding colony in the South Shetlands, where low densities and high rates 302 of turnover among animals may lead to different outcomes in terms of male reproductive success. We 303 found that polygyny was relatively weak at Cape Shirreff, where paternity could only be assigned to 304 around ten percent of pups and reproductive skew was only slightly stronger in males than in females. 305 We discuss these findings in the context of reproductive skew in harem-holding species and mating 306 system flexibility, as well as in light of specific features of the focal population. 307 308

Paternity assignment rate and strength of inferred polygyny 309
We were only able to assign paternity to 31 out of 290 pups (10.7%). This low rate of paternity Shirreff. Specifically, the annual sampling and marking of adults could not usually be initiated before 349 the last week of October. We know from studies of other elephant seal colonies that female haulout 350 activity usually peaks sometime between the 2 nd and the 25 th of October, and that the males usually 351 arrive around a week earlier (Galimberti and Boitani 1999). Any males that were exclusively present 352 in the earlier part of the season will therefore not have been sampled. However, our analysis of the best 353 configuration fathers revealed a similar overall pattern to the parentage assignments and did not provide 354 any indications of the presence of a small number of disproportionately successful males. It is therefore 355 unlikely that our results can be explained by a failure to sample a small number of highly successful 356 males. If anything, the magnitude of polygyny inferred from the best configuration fathers appeared to 357 be slightly lower than the magnitude of polygyny inferred from paternity assignments. However, 358 parentage analysis was performed on the full dataset while the best configuration fathers were only 359 inferred for a subset of pups with known maternity. This high rate of turnover of breeding females may impact our results because many pups are likely to 380 have been conceived at other breeding colonies, thus diluting the perceived reproductive success of the 381 harem holders at Half Moon beach. To investigate how this might impact our results, we restricted our 382 analysis to pups born to females that were known to be present in the colony during the year of 383 conception. Although our sample size was substantially reduced, the overall pattern of paternity 384 assignment was qualitatively similar and again we did not find any evidence of high male reproductive 385 skew. Additionally, pups fathered at different colonies might be expected to carry different genetic 386 signatures given that the four main global populations of southern elephant seals show pronounced 387 genetic differentiation (Slade et al. 1998, Hoelzel, Campagna & Arnbom 2001. However, we could 388 not find any obvious genetic differences between pups with known fathers and pups that were not 389 assigned paternity. This implies that the majority of pups were probably not conceived at distant 390 localities, although we cannot discount the possibility that females may mate at closer, less genetically 391 differentiated sites such as elsewhere in the South Shetland Islands or at South Georgia. to access an area of high topographic complexity (rocky outcrops) that may provide sufficient cover for 400 sneaky copulations. High complexity breeding sites with gullies and dunes appear to diminish the 401 reproductive success of northern elephant seal (M. angustirostris) harem holding males because, even 402 though these features concentrate females who might otherwise sparsely occupy a beach, they also 403 provide cover for more agile peripheral males attempting to infiltrate harems to copulate (Hoelzel et al. 404 1999). to an elevated rocky area that could potentially provide cover for males pursuing alternative mating well as from the best configuration fathers inferred by COLONY (panels c,f). Note that the latter was 665 conservatively inferred from pups that were assigned maternity to a genotyped female, thereby excluding pups 666 with neither parent sampled. The number of offspring assigned to genotyped males was, in some cases, larger 667 than the number of offspring assigned to best configuration fathers because ~ 40% of paternities were assigned 668 to pups that were not assigned to a genotyped mother. Panels