Is MHC diversity a better marker for conservation than neutral genetic diversity? A case study of two contrasting dolphin populations

Abstract Genetic diversity is essential for populations to adapt to changing environments. Measures of genetic diversity are often based on selectively neutral markers, such as microsatellites. Genetic diversity to guide conservation management, however, is better reflected by adaptive markers, including genes of the major histocompatibility complex (MHC). Our aim was to assess MHC and neutral genetic diversity in two contrasting bottlenose dolphin (Tursiops aduncus) populations in Western Australia—one apparently viable population with high reproductive output (Shark Bay) and one with lower reproductive output that was forecast to decline (Bunbury). We assessed genetic variation in the two populations by sequencing the MHC class II DQB, which encompasses the functionally important peptide binding regions (PBR). Neutral genetic diversity was assessed by genotyping twenty‐three microsatellite loci. We confirmed that MHC is an adaptive marker in both populations. Overall, the Shark Bay population exhibited greater MHC diversity than the Bunbury population—for example, it displayed greater MHC nucleotide diversity. In contrast, the difference in microsatellite diversity between the two populations was comparatively low. Our findings are consistent with the hypothesis that viable populations typically display greater genetic diversity than less viable populations. The results also suggest that MHC variation is more closely associated with population viability than neutral genetic variation. Although the inferences from our findings are limited, because we only compared two populations, our results add to a growing number of studies that highlight the usefulness of MHC as a potentially suitable genetic marker for animal conservation. The Shark Bay population, which carries greater adaptive genetic diversity than the Bunbury population, is thus likely more robust to natural or human‐induced changes to the coastal ecosystem it inhabits.

observed across a variety of vertebrate species are commonly explained by balancing selection (Garrigan & Hedrick, 2003). Balancing selection maintains high levels of MHC diversity by two possible, not mutually exclusive, mechanisms: frequency-dependent selection (Borghans, Beltman, & Boer, 2004) and heterozygote advantage (Doherty & Zinkernagel, 1975). The frequency-dependent selection model suggests that MHC diversity is pathogen-mediated, because rare MHC variants are selected for by host-pathogen co-evolution.
In contrast, heterozygote advantage explains balancing selection due to heterozygotes having greater fitness than homozygotes.
Compared to terrestrial vertebrates, relatively little is known about MHC diversity in cetaceans, and the extent to which cetacean MHC diversity is associated with population viability remains uncertain. The vaquita (Phocoena sinus) population, endemic to the Gulf of California, showed low levels of MHC II variation (Munguia-Vega et al., 2007) and is now considered functionally extinct (Taylor et al., 2017). In contrast, the extinct baiji (Lipotes vexillifer) of the Yangtze River exhibited very high MHC diversity (Xu et al., 2012;Yang, Yan, Zhou, & Wei, 2005). Reduced MHC diversity may not necessarily adversely affect population viability (Radwan, Biedrzycka, & Babik, 2010). Caveats for many of these studies are that they had no baseline measure of genetic diversity in a conspecific viable population or no comparison of MHC and other types of genetic variation. No study to date has compared MHC and neutral genetic diversity of conspecific cetacean populations that differ with respect to population parameters and viability forecasts.
In this study, we used two genetic markers, MHC and neutral microsatellites, to assess genetic diversity of two contrasting bottlenose dolphin (Tursiops aduncus) populations-one in Shark Bay (SB) and another off Bunbury (BB), Western Australia ( Figure 1).
These two populations, more than 1,000 km apart (Figure 2), are not connected by dispersal. Each population exhibits limited genetic exchange with its neighboring populations Manlik et al., 2018). The two populations differ greatly with respect to population viability. A comparative population viability analysis showed that the SB population appeared stable, but the BB population was forecast to decline with a high probability of extinction, unless supported by immigration . The large difference in viability between the two populations was best explained by considerable differences in reproductive rates . Besides this difference in reproductive rates, the two populations also differ with respect to anthropogenic pressure . The SB population occurs in a remote UNESCO World Heritage area with markedly lower anthropogenic activity, whereas BB inhabits waters adjacent to an expanding regional city and port with comparatively high vessel traffic (Manlik, 2019; Our findings are consistent with the hypothesis that viable populations typically display greater genetic diversity than less viable populations. The results also suggest that MHC variation is more closely associated with population viability than neutral genetic variation. Although the inferences from our findings are limited, because we only compared two populations, our results add to a growing number of studies that highlight the usefulness of MHC as a potentially suitable genetic marker for animal conservation. The Shark Bay population, which carries greater adaptive genetic diversity than the Bunbury population, is thus likely more robust to natural or human-induced changes to the coastal ecosystem it inhabits.
SB and BB also differ with respect to reported population sizes.
SB population size was estimated by aerial surveys to be about 2,000-3,000 individuals (minimum estimates; Preen, Marsh, Lawler, Prince, & Shepherd, 1997) in a 14,900 km 2 area, but other studies investigating various sections of SB suggest that the population may be much larger (e.g. Nicholson et al., 2012). BB population size was assessed to be approximately 260 individuals for the 120 km 2 area . However, smaller seasonal abundance estimates have been reported for BB (Smith, Pollock, Waples, Bradley, & Bejder, 2013;Sprogis et al., 2016). Different methodologies to estimate population sizes, and the issue of connectivity, make comparison difficult, but all studies suggest that SB is substantially larger than BB.
The aim of this study was to compare MHC II DQB genetic diversity and microsatellite diversity between these two contrasting dolphin populations. Given that only few MHC studies have been conducted on populations with differing reproductive success or population forecasts, this provided a rare opportunity to compare MHC and neutral genetic diversity between two natural populations with considerable differences in viability. If MHC variation reflects differences in fitness, and given the large difference in reproductive output between the two populations , we would expect to observe a larger inter-population difference in MHC diversity than in microsatellite diversity. Additionally, to assess evolutionary and ecologically relevant genetic variation, we evaluated signals of selective pressure on MHC II DQB. We did this by assessing nonsynonymous versus synonymous nucleotide substitutions (Nei & Gojobori, 1986), whether substitutions occurred at codons expressing antigen-binding residues, and by performing a Tajima's D test (Tajima, 1989).
F I G U R E 1 Mother and calf bottlenose dolphin (Tursiops aduncus) in Shark Bay, a UNESCO World Heritage Site in Western Australia. Photograph: Ewa Krzyszczyk F I G U R E 2 (a) Shark Bay, a UNESCO World Heritage area, is about 13,000 km 2 in size and is divided by the Peron Peninsula, which bisects the bay into a western and an eastern gulf. Sampling sites included a 300 km 2 area (circled) north of Monkey Mia and an area of ca. 260 km 2 (circled) in the western gulf. (b) The inset shows the relative location of the study sites (Shark Bay & Bunbury). The coastal study area of Bunbury covers about 120 km 2 and extends approximately 1.5 km offshore with a linear distance of 50 km. The study site includes the coastal areas, embayment, Leschenault Inlet and outer harbors (5 km 2 ), estuary and river mouth (15 km 2 ). Transects of the outer-water Bunbury study site are shown. These figures are modified from Figure S1 of Manlik et al. (2016) Sampling in SB included two sites in western and eastern SB ( Figure 2), that are connected by extensive gene flow (Krützen, Sherwin, Berggren, & Gales, 2004) and appear to form one large continuous population. A total of 686 and 125 dolphins were biopsied in SB and off BB, respectively. Sex of individuals was determined by various methods, as described by Sprogis et al. (2016), including genetic sexing (Baker et al., 1998). We performed chisquare tests to assess whether the numbers of males and females in the samples were significantly different from those in the surveyed populations or different from an expected 50:50 male to female ratio. Sex ratios for surveyed individuals versus sampled individuals were not significantly different (SB = χ 2 = 0.42, p = 0.515; BB: χ 2 = 0.16, p = 0.693) nor were the ratios of sampled individuals significantly different from 50:50 (SB: χ 2 = 0.10, p = 0.757; BB: χ 2 = 1.7, p = 0.190). To assess whether it was justified to pool samples collected from eastern and western SB, we estimated subpopulation fixation index (F ST ) based on microsatellite data, using GenAlex 6.501 (Peakall & Smouse, 2006, and compared MHC and microsatellite diversity between the two sampling locations.

| Assessing signals of selection acting on MHC DQB
To assess signals of selection, we compared rates of nonsynonymous (d N ) and synonymous (d S ) substitutions within the 172-bp MHC DQB region. We used the Nei-Gojobori method (Nei & Gojobori, 1986) for a codon-based test of positive selection (two-sided z-test) implemented in MEGA version 7.08 (Kumar, Stecher, & Tamura, 2016) to test whether d N > d S for (a) all codons of the entire sequence; (b) codons of the putative peptide binding region (PBR), that is, variable codons that code for amino acids that have been reported to bind to antigens; and (c) putative nonpeptide binding regions (non-PBR).
Variance estimation for the z-test was based on 1,000 bootstrap replicates. Additionally, we used DNASP to perform Tajima's D test (Tajima, 1989), which detects departure from selective neutrality or historical changes in population size.
Sampling variances and standard deviations were calculated for nucleotide diversity and haplotype diversity according to Nei (1987) and for Watterson mutation estimator according to Tajima (1993).
We also calculated standard errors of the mean between the three SB conservative samples and across all subsamples (SB: 19 subsamples; BB: 5 subsamples). We used t tests to compare the mean π, Hd, Ө W , and Ө Eta values between the two populations across all subsamples.

| Assessment of microsatellite diversity
All sampled BB individuals were previously genotyped for 25 polymorphic microsatellite loci (Manlik et al., 2018). We followed the same procedure and checks for genotyping individuals of the SB population as described in Manlik et al. (2018): We used previously tested primers for polymorphic microsatellite loci (Hoelzel, Potter, & Best, 1998;Kopps et al., 2014;Krützen, Valsecchi, Connor, & Sherwin, 2001;Nater, Kopps, & Krützen, 2009;Shinohara, Domingo-Roura, & Takenaka, 1997). All primer sequences used in this study are listed in Dryad/Table S1. Microsatellite amplification was performed using the Qiagen Multiplex Kit TM in three multiplex PCR reactions as described in Manlik et al. (2018). Fragment analysis of PCR amplicons was performed on a 3730XL DNA Analyzer (Applied Biosystems), employing a Genescan-500 LIZ TM size standard. Alleles were scored using GenemAPPer 4.0 (Applied Biosystems) and the microsatellite plugin for Geneious 6.0 (Drummond et al., 2010 (Chybicki & Burczyk, 2009) was used to estimate the frequency of null alleles at microsatellite loci in each population.
Microsatellite diversity was summarized by measuring observed heterozygosity (H o ), expected heterozygosity relative to HWE (H e ), the number of effective alleles (A e ), and Shannon's Index ( 1 H) (Brown & Weir, 1983;Sherwin, Chao, Jost, & Smouse, 2017), using GenAlex 6.501. We used paired t tests to compare the mean values of these measures between the two populations across the microsatellite loci.

| Sampling for comparison of inter-population genetic diversity
Due to the sample-size difference between SB and BB, we used three sampling approaches to compare genetic diversity between the two populations:  (209) subsamples for SB and 5 × 11 (55) subsamples for BB.
Other methods, such as rarefaction, are often used to investigate the effect of sample size, but we believe our three sampling approaches address this more thoroughly.

| RE SULTS
Pooling the eastern and western SB datasets was justified because the subpopulation fixation index (F ST ) comparing the two sampling sites in SB showed very little differentiation (F ST = 0.006; Dryad/

| Sequence variants of MHC DQB
Forward and reverse MHC DQB sequences of a total 341 individuals (SB: 276; BB: 65) were analyzed. Totals of 186 and 43 MHC DQB sequence variants were inferred by haplotype reconstruction for SB and BB, respectively. We did not detect any patterns in the sequences that indicated multiple allelism (i.e. having more than two alleles or sequence variants per amplicon/individual), gene duplications, stop codons, or frameshifts. Comparing MHC DQB sequences of seven mother-father-offspring trios (Kopps, 2007) did not reveal any patterns that were inconsistent with single-locus Mendelian inheritance. Nonetheless, we refrain from classifying these inferred sequence variants as novel MHC alleles, which are commonly confirmed by sequencing clones (Marsh et al., 2010) or by re-genotyping all individuals with rare haplotypes (Ahmad et al., 2002). Thirty-nine individuals were homozygous for all 172 bps for one of four unique sequences (Dryad/ Figure S1). All sequences showed high similarity (98%-100%) to published MHC DQB alleles in dolphins (Dryad /Table S4).

| Signals of selection acting on MHC
We detected signals of selection acting on MHC DQB. In both populations, nonsynonymous (d N ) substitution rates were significantly greater than the synonymous (d S ) substitutions rates in the entire 172-bp region and in the putative PBR, but not in the non-PBR (Table 1). About 82% of the variable nucleotide sites (18 out of 22) are within codons that have been associated with the PBR (Dryad/ Figure S1). Notably, the large majority of variable nucleotide sites were detected within the PBR (Dryad/ Figure S1). Tajima's D was near zero for MHC DQB in both populations (Table 2)

| Microsatellite diversity
Neither population showed evidence for scoring errors due to stuttering or large-allele dropouts for any of the microsatellite loci.
We also did not detect evidence for null alleles for any of the loci.  Tables 4 and 5). There was also no significant difference between the SB subsamples with respect to the microsatellite measures of genetic diversity (Dryad/ Table S5).
Results of paired t tests for all sampling approaches are shown in Table 5.
TA B L E 1 The estimated rates of nonsynonymous (d N ) and synonymous (d S ) substitutions (±standard errors of the mean) for putative peptide binding regions (PBR) and nonpeptide binding regions (non-PBR) and their ratios for DQB exon 2 in the Shark Bay (SB) and Bunbury (BB) dolphin population

| D ISCUSS I ON
Compared to selectively neutral genetic variation, variation of adaptive genes, such as those of the MHC, is a better proxy for genetic diversity relevant to population viability (Oliver & Piertney, 2012;Sommer, 2005;Ujvari & Belov, 2011). A loss of adaptive genetic diversity reduces reproductive success and survival in the short-term and ultimately diminishes the evolutionary potential of populations to adapt to environmental changes (Frankham, 2005;Frankham et al., 2010). Our results show that the more stable SB population, which displayed greater reproductive success , harbors greater MHC diversity compared with the BB population that was forecast to decline. It is important to note that this was the case, regardless of the sampling approach. Our finding that microsatellites do not show any significant differences between SB and BB suggests that the higher MHC diversity in SB is unlikely due to differences in  1-3). Tabulated are n = the sample size; π = nucleotide diversity; Hd = haplotype diversity; Ө W = Watterson mutation estimator; Ө Eta = the mutation parameter theta based on number of mutations, Eta. Standard deviations for π, Hd, and Ө W are shown in parentheses.
population size, because the resultant genetic drift is expected to affect MHC and microsatellites equally (although see Eimes et al., 2011). Therefore, it seems likely that other interactions, such as differential fitness or parasite pressure, are driving the observed MHC pattern.
The number of sequence variants we detected in both populations is unusually high, but a high number of single-locus MHC class II variants have been detected in other cetacean populations as well (e.g. Xu et al., 2012). As mentioned in the Methods, we refrain from classifying the sequence variants as novel MHC alleles, but having  followed the same methodology of inferring sequence variants for both populations allowed us to compare MHC sequence variation between the two populations. Further confirmation of alleles could be achieved by sequencing clones (Marsh et al., 2010) or by re-genotyping all individuals with rare haplotypes (Ahmad et al., 2002).

| Potential factors contributing to the interpopulation differences in MHC diversity
Differences in MHC diversity between the two populations might be related to fitness. Adult females in SB displayed higher reproductive success than BB females , and preliminary data suggest that SB females with greater reproductive success also exhibit greater MHC DQB diversity than females with low reproductive success (Manlik, 2016). Another selective effect associated with the inter-population difference in MHC diversity is differences in pathogen communities. High levels of MHC diversity can be maintained by balancing selection due to MHC's function in binding to pathogen-derived antigens (Eizaguirre, Lenz, Kalbe, & Milinski, 2012;Takahata & Nei, 1990;Wegner, Reusch, & Kalbe, 2003). The signal that we detected by the d N /d S analyses relates to long periods, with time for mutations to accumulate very slowly, at a rate of about 10 −9 per generation per nucleotide site. These patterns were originally proposed for differentiation between species, but the same patterns are expected for variation within a single population, though weaker (Kryazhimskiy & Plotkin, 2008). The higher ratios of nonsynonymous to synonymous substitutions that we observed in the MHC DQB region of both populations are consistent with balancing selection (Kimura, 1977;Yang & Bielawski, 2000). There are studies on numerous vertebrate taxa that show an association between pathogen load, infectivity, and MHC diversity (e.g., Paterson, Wilson, & Pemberton,1998;Sepil, Lachish, Hink, et al., 2013;Wegner et al., 2008). Vassilakos et al. (2009) proposed that differential pathogen pressure across the range of cetacean populations could explain geographic variation in MHC diversity.
Although the d N /d S analyses can detect balancing selection over long periods, on shorter time scales, there might be other influences, such as bottlenecks, or directional selection due to a recent change in pathogen load; these can be detected by Tajima's D, with the proviso that because it is sensitive to demographic and selective effects, they could cancel each other out. Bottlenose dolphin mortalities due to pathogens, such as the cetacean morbillivirus, have been reported in Western Australia (Stephens et al., 2014), and outbreaks are associated with high mortality (van Bressem et al., 2014;Di Guardo & Mazzariol, 2014). If this mortality is selective, then it could give a signal with Tajima's D, unless counteracted by some demographic effect. However, little is known about pathogen communities across geographic locations, including the two sites of this study. Other factors are unlikely to explain the differences in MHC diversity between SB and BB: Age and sex are unlikely because our sample sizes had equal numbers of each age class and sex; effects of mate choice (Kamiya, O'Dwyer, Westerdahl, Senior, & Nakagawa, 2014;Yamazaki & Beauchamp, 2007) are possible but unlikely because both SB and BB exhibit a promiscuous mating system (Connor, Richards, Smolker, & Mann, 1996;Smith et al., 2016).
Regardless, the difference in MHC diversity between the two populations likely also confers a differential potential to respond to pathogen pressure.
The diverse function and variability of MHC genes reflect evolutionary adaptive processes and thus make them suitable candidates to evaluate genetic diversity relevant to conservation. In this study, we compared MHC genetic diversity and microsatellite diversity of two contrasting bottlenose dolphin populations. We revealed signals of selective processes acting on the MHC DQB in both populations. In comparison with the BB population, the more stable SB population exhibited larger MHC diversity. This is congruent with our hypothesis that the difference in reproductive output and viability between the two populations    (Figure 1).

CO N FLI C T O F I NTE R E S T S
None declared.