https://orcid.org/0000-0003-2371-6581
Abstract
Social monogamy is rare in mammals, but in those species in which it occurs, individuals frequently engage in extra-pair copulation (EPC). Numerous hypotheses have been proposed to explain EPC, but relatively few field studies have examined factors influencing EPC in socially monogamous mammals. Prairie voles (Microtus ochrogaster) are a socially monogamous rodent in which extra-pair paternity (EPP) is common. Using genetic data from a 3-year study of a natural population of prairie voles, we investigated whether the negative consequences of inbreeding and the structure of social units (male–female pairs vs. groups; adult sex ratio within groups) were factors affecting EPC. We found strong evidence that genetic similarity between males and females that produced offspring via EPC was less than that between social partners, as would be expected if avoiding inbreeding depression influenced the occurrence of EPC. Social unit structure was also a factor involved with extra-pair parentage. Contrary to our expectations, the greater the proportion of females in the group, the lower the chance of EPC resulting in offspring production by females, and, similarly, the chance of EPP by males declined as the proportion of males within the group increased. However, neither males nor females were more likely to produce offspring from EPC when they were living in pairs versus groups. One implication of these results is that EPC may be influenced more by female behaviors, such as mate choice, than male mate guarding. Overall, our results suggest that the proximate factors influencing EPC in prairie voles are complex but include the cost of inbreeding depression and the structure of social units.
Social monogamy is uncommon in mammals, occurring in only about 5% to 9% of mammalian species, but among those species in which it occurs, it is not generally characterized by mating exclusivity (Kleiman 1977; Lukas and Clutton-Brock 2013; Lambert et al. 2018). Cohas and Allainé (2009) reviewed field studies of 22 species of socially monogamous mammals in which genetic markers were used to assess parentage and found extra-pair paternity (EPP) in more than 75% (17/22) of the species they surveyed. The frequency of extra-pair offspring within a species was highly variable with 0% to 92% of offspring tested per species being from extra-pair copulations (EPCs). In addition, in more than half of the species surveyed (13/22), at least 25% of the offspring sampled were sired by a male other than the social partner of the female. Males likely engage in EPC because it can increase their reproductive success without substantially increasing paternal investment, but why females mate with males other than their social partner remains a challenging question (Trivers 1972).
One hypothesis that has received some empirical support from field studies of socially monogamous mammals is the inbreeding avoidance hypothesis, which posits that the negative consequences of inbreeding can favor the evolution of inbreeding avoidance behaviors such as EPC (Cohas et al. 2008; Driller et al. 2009; Reid et al. 2015). One assumption of the inbreeding avoidance hypothesis is that the choice of a current social partner had been constrained in some manner, e.g., infrequent encounters between potential social partners due to low population density (Kokko and Rankin 2006). This constraint in choice of a social partner would create the potential for a female to reduce the inbreeding level of her offspring by seeking EPC from less related males encountered after the social partner (Westneat and Stewart 2003; Duthie et al. 2016).
In socially monogamous species that reside, at least sometimes, in social units with more than 1 adult male and female, the occurrence of EPC may also be affected by the ratio of adult males to adult females within a social unit. In social units with more adult males than females, males can prevent or decrease the chances of females engaging in EPC by guarding the females and directing aggression toward conspecific males that are not members of the social unit. Because not all males in the social unit would necessarily be closely related to the female, the females would also have more mating options within the social unit, decreasing EPC among females (Cohas and Allainé 2009). However, males within these social units would have fewer opportunities for exclusive mating with the females in the social unit due to mate competition with other males in the social unit, so EPC for males could increase. In social units where the adult sex ratio is biased toward females, the occurrence of EPC for females may increase because male members of the social unit are less able to prevent access to these females by conspecific males that live outside the social unit (Isvaran and Clutton-Brock 2007). Males within the social unit would have more opportunities for mating with females within the social unit, so EPC for males may decrease.
Prairie voles (Microtus ochrogaster) are typically characterized as a socially monogamous species (Getz et al. 1981, 1990). The social organization of prairie voles primarily consists of male–female pairs that defend a shared territory, but some individuals live in social units containing more than 2 adults, which are referred to as groups (Getz and Hoffmann 1986; Getz et al. 1993). The frequency of each type of social unit varies temporally within populations and geographically among populations, but how these social units form is unknown (Getz et al. 1993; Streatfeild et al. 2011). The average lifespan of adult prairie voles in nature is 2 to 3 months, and it has been hypothesized that adult prairie voles often form pairs with the first available opposite-sex conspecific because individuals delaying breeding risk a decrease in lifetime reproductive success (Getz et al. 1997, 2004; Ozgul et al. 2004). Because there is no sex bias in natal dispersal rate or distance, some social pairs likely form between related individuals, creating the potential for inbred offspring in natural populations (McGuire et al. 1993). Inbreeding depression has been documented in laboratory and field populations of prairie voles (Bixler and Tang-Martinez 2006; Lucia and Keane 2015).
Genetic analyses of parentage in field populations of prairie voles reveal that, depending on location and year, up to 60% of litters contain offspring sired by males outside the social unit, but factors shaping the occurrence of EPC have not been extensively examined in this species (Streatfeild et al. 2011; Lichter et al. 2020). Therefore, the objectives of this study were to examine whether extra-pair parentage in prairie voles was affected by the (i) genetic similarity between opposite-sex members of the same social unit (hereafter referred to as social partners) and (ii) structure of social units (male–female pairs vs. groups; adult sex ratio within groups). We predicted that the genetic similarity between males and females that produced extra-pair offspring would be less than the genetic similarity between social partners, as would be expected if EPC was a mechanism to reduce inbreeding depression within litters. In prairie voles, the number of offspring that females produce is not correlated with the number of adults residing in a social unit, but we do not know if EPC is affected by the type of social unit or the ratio of adult males to adult females within a group (Solomon and Keane 2018). Males and females living in pairs may engage in more EPC than those in groups because there are no opposite-sex conspecifics other than the social partner for them to mate with if the social partner is genetically very similar to them. But the frequency of EPC among individuals in groups could also depend on the composition of the group. We predicted that EPP among females would increase with a decreasing ratio of males:females in a group because the smaller number of males would not be as successful at preventing females from mating with conspecific males living outside the social unit (nonresident males), whereas male EPP would decrease because there would be more opportunities for mating with the females in the group. In contrast, we predicted a decrease in EPP by females with an increasing ratio of males:females in a group because males would be more successful at preventing access of females to males living outside the social unit and females would also have more potential mates within the group. We also predicted that the frequency of EPP among males would increase with an increasing ratio of males:females in a group because more males per female may reduce mating opportunities for a male within the group.
Materials and methods
Study area
Data for this study were collected from a natural population of prairie voles located at the Indiana University Bayles Road Preserve (Bloomington, Indiana, 39°13ʹ00″N, 86°32ʹ27″W). Our study site was situated within a large field (~6 ha), containing primarily grasses and forbs, that was mowed periodically to prevent the invasion of woody plants via ecological succession. Fieldwork began in mid-July in 2006 to 2008 and lasted for 4 consecutive weeks in each of the 3 years (for details, see Streatfeild et al. 2011; Chesh et al. 2012).
Field methods
Each year we conducted 4 consecutive weeks of livetrapping to monitor the prairie vole population. The size of the area that we livetrapped varied somewhat among years (1.5 ha in 2006 and 2008, 2.2 ha in 2007). We used a combination of grid and nest trapping using Ugglan multiple-capture traps (Grahnab, Hillerstorp, Sweden). Grid trapping allowed us to estimate population densities and capture adult females to track to their nests. From nest-trapping data, we could determine the voles residing at a specific nest and capture offspring for parentage analysis. For both types of livetrapping, we baited traps with cracked corn and covered each trap with an aluminum shield or wooden board overlain with vegetation to shield trapped animals from excessive heat or rain. Between trapping periods, traps were left in place but unset (for details, see Streatfeild et al. 2011; Chesh et al. 2012).
During either the first week (2006 to 2007) or first 2 weeks (2008) of each field season, we trapped on a grid with 10 m spacing between live traps. For grid trapping, we placed a single Ugglan multiple-capture live trap in a vole runway within 1 m of each grid marker. Traps were set in the late afternoon and checked later that evening and the following morning from Sunday afternoon through Friday morning for a total of 10 trap checks per week.
For every adult female captured during grid trapping weeks, we attempted to find the location of their nest using either radiotelemetry or fluorescent-powder tracking (for details, see Lucia et al. 2008). Once we located the nest of an adult female, we recorded the nest coordinates with a handheld global positioning unit (eTrex Legend, Garmin, Olathe, Kansas) and placed 4 Ugglan multiple-capture live traps within 30 cm of the entrance(s) of the nest. Immediately following the grid trapping period, we trapped only at nest sites for either 3 (2006 to 2007) or 2 (2008) consecutive weeks. During nest-trapping weeks, traps were set in the late afternoon and checked that evening and the following morning from Sunday afternoon through Tuesday evening and again from Wednesday afternoon until Friday evening, for a total of 10 trap checks per week.
Upon first capture, we gave each prairie vole a unique toe clip for identification and stored these tissue samples at −20 °C without preservative for subsequent genetic analysis of parentage and relatedness. For every capture, we recorded ID number, capture location, sex, body mass (g), age class, and reproductive condition of an individual. Body mass was determined to the nearest gram using a Pesola micro-line spring scale (Forestry Suppliers Inc., Jackson, Mississippi) and used to assign individuals to age classes. Individuals >29 g in mass were classified as adults, those 21 to 29 g were considered subadults, and those <21 g were classified as juveniles (Gaines et al. 1979; Getz et al. 1993). We never captured individuals marked in 1 year in succeeding years. The density of adult voles was 40/ha in 2006, 84/ha in 2007, and 90/ha in 2008 (Keane et al. 2015).
All the procedures involving the trapping, marking, tissue sampling, and handling of prairie voles in this study were approved by the animal care and use committees of Miami University and Indiana University and were consistent with the guidelines published by the American Society of Mammalogists for the use of wild animals in research (Sikes and the American Care and Use Committee of the American Society of Mammalogists 2016).
Nest residency
Only the capture data from nest-trapping weeks were used to assign nest residency to adult male and female voles. We classified an adult prairie vole as a resident at a specific nest if it was captured at least once per week during each of the first 2 weeks of nest trapping, and ≥75% of all captures during these 2 weeks were at a single nest site (Cochran and Solomon 2000). Adults trapped at least once a week during the first 2 nest-trapping weeks but less than 75% of the time at any 1 nest or that were not caught during each of the first 2 nest-trapping weeks were not classified as a resident at any nest location. Unclassified individuals may have been wanderers (nonterritorial voles that were captured at more than 1 nest site), dispersers, or residents at nests located off the study grid (Getz et al. 1993). All adults classified as residing at the same nest were considered to be social partners in the same social unit.
Microsatellite analysis
All prairie voles captured were genotyped at 6 microsatellite loci previously used for analysis of parentage and genetic relatedness in this species (Keane et al. 2007, 2015; Solomon et al. 2009). We used either standard phenol/chloroform extraction techniques or DNeasy extraction kits (Qiagen, Valencia, California) to extract genomic DNA from tissue samples and conducted polymerase chain reactions (PCRs) to amplify microsatellite alleles at each locus (for details of PCR conditions, see Keane et al. 2007; Solomon et al. 2009). The resultant PCR products were diluted, combined with an internal size standard (LIZ GS500, Applied Biosystems, Foster City, California) and detected using an ABI 3730 DNA sequencer (Applied Biosystems, Foster City, California). Base pair lengths of the fluorescently labeled DNA fragments were determined with GeneMapper 3.7 software (Applied Biosystems), and microsatellite alleles were binned into discreet size classes using FlexiBin (Amos et al. 2006).d
We used Cervus 3.0 to test for departures from Hardy–Weinberg equilibrium at each locus in a particular year. With significance set at α < 0.05 (Kalinowski et al. 2007) and using a Bonferroni correction to reduce the probability of committing a type I error due to conducting multiple tests, we found only a single locus (MSMM-3) in the Indiana population in 2008 to exhibit significant departures from Hardy–Weinberg equilibrium, likely due to the presence of null alleles (Mabry et al. 2011). Because parentage assignment in Cervus 3.0 is robust to the presence of null alleles, and departures from Hardy–Weinberg equilibrium at a single locus were not likely to substantially bias exclusion probabilities for parentage analysis, we used the allelic data from all 6 loci in assigning parentage in the population during each year (Dakin and Avise 2004).
Parentage analyses
Parentage was assigned using the parent-pair analysis option in Cervus 3.0, which uses a simulation to calculate the statistical confidence of parentage assignments (Kalinowski et al. 2007). We conducted separate parentage analyses for each year, and all simulations were performed for 10,000 cycles with a genotyping error rate of 0.02. This error rate was based on empirical estimates of 2 potential sources of error: mutation and mis-scoring of alleles (Solomon et al. 2004). The remaining input parameters used in simulations were based on the actual data from the study population each year.
We assessed parentage using a multistage approach where we initially considered only those adults trapped within 20 m of the location of a juvenile’s (<21 g) first capture as candidate parents (for details, see Mabry et al. 2011). The 20 m criterion for initially selecting candidate parents was chosen because it was the mean home range diameter of adult prairie voles in the Indiana population (Streatfeild et al. 2011). We accepted parentage assignments whenever the confidence level among a male–female–juvenile trio was ≥95%. If parentage could not be assigned to a trio at a confidence level of at least 95% after initial analysis, we expanded the set of candidate parents to include all adults trapped within 40 m of the location of first capture of a juvenile and ran another parentage analysis. Finally, if we could only assign the female, but not the male parent at the 95% confidence level after these 2 analyses, we ran a parentage analysis using the “known mother” option and considered all adult males captured within 40 m of the location of first capture of the juvenile as candidate fathers. In every case where parentage was assigned using the known mother option, the putative female parent was also captured together with the juvenile in the same trap at least once, providing an independent corroboration of maternity. We used the genetic parentage data to determine which individuals bred with conspecifics that were not their social partner in order to estimate EPP. Because we were not able to directly observe copulations, our estimate of EPP should be an underestimate of EPCs for several reasons including (i) not all copulations result in fertilization, (ii) not all offspring survive to be captured, and (iii) not all surviving juveniles are captured.
Genetic similarity
The microsatellite allele data we used for analyzing parentage also were used to estimate genetic similarity among individuals. Pairwise similarity between voles for each year was estimated using RELATEDNESS 5.0, which uses the frequency of alleles in a population to calculate the probability that 2 individuals share alleles identical by descent (Queller and Goodnight 1989). Similarity values (r) between 2 individuals may range from −1 to +1, where a positive value indicates that 2 individuals share more alleles that are identical by descent than expected by chance (i.e., more related), whereas a negative value indicates they share fewer alleles identical by descent than expected by chance (i.e., less related). If a population is in Hardy–Weinberg equilibrium, first-degree relatives (e.g., parent–offspring or full-siblings) should have similarity values of 0.5, whereas pairs of unrelated individuals should have similarity values of 0.
Statistical analysis
We considered only animals from social units comprising male–female pairs or more than 2 adults as focal individuals for all analyses. For social units comprising more than a male–female pair, we considered all opposite-sex adults residing at the same nest to be social partners in the same social unit (i.e., have the same nest ID). Because we found no voles that lived more than 1 year, we pooled observations among years because in that respect, observations were independent of vole ID.
We measured effect sizes using Bayesian linear models. Effect sizes estimate biologically relevant quantities, and Bayesian methods estimate the probability of a hypothesis or parameter given the data, rather than a P-value (Wasserstein et al. 2019). All analyses were conducted using R v. 4.3.1 (R Core Team 2023), and R packages “rstanarm” (v. 2.21.4, Goodrich et al. 2023), “emmeans” (Lenth 2023), and “bayesplot” (Gabry and Mahr 2022). For each analysis, we created a set of related competing models, including those with different random effects structures with individual vole ID and nest ID. Each model was fit using Hamiltonian Markov chain Monte Carlo (MCMC) using 4 independent chains, and we assessed these MCMC samples visually, comparing the empirical cumulative probabilities of the samples to theoretical expectations (Säilynoja et al. 2022). We also checked standard numerical assessments using effective sample size and the Gelman–Rubin statistic (R-hat = 1.00) as calculated in “rstanarm.” In each analysis, we compared competing models using the expected log pointwise predictive density (ELPD), which was estimated using leave-one-out (LOO) validation using the “loo” R package (Vehtari et al. 2022). In our results, we present only the best-supported models on the basis of the greatest out-of-sample predictive power with a hypothetical new data set (i.e., largest ELPD). We used Pearson correlation to assess whether adult female body mass was correlated with the proportion of adult females in a social unit.
EPP versus sex and social unit type
To assess whether sex, type of social unit (male–female pair vs. group), or an interaction between sex and type of social unit explained the chance of EPP, we used Bayesian logistic regression. We relied on the default, weakly informative priors in “rstanarm” because so little evidence exists regarding genetic similarity of partner types in prairie voles.
Difference in genetic similarity between social and extra-pair partners
To assess our prediction that genetic similarity between an individual vole and the extra-pair partners was less than their genetic similarity to their opposite-sex social partner(s), we calculated the mean difference in genetic similarity between each individual’s social partner(s) and their extra-pair partner(s). We used a Bayesian linear model with Gaussian errors to test whether this difference was less than 0 (mean extra-pair similarity − social pair similarity). We also used nest ID as a random effect because this model yielded the ELPD. For this analysis, observations were averaged for each individual, and thus we did not have repeated observations of individual voles. In addition, we assessed this for voles that engaged in EPP, and also for all voles.
EPP versus genetic similarity and social unit sex ratio
To test our predictions that the probability of EPP was affected by genetic similarity and sex ratio within a social unit, we used Bayesian generalized linear mixed models with binomial errors and nest ID as a random effect. EPP was a binomial variable, where each vole pairing was scored as having produced offspring with a social partner (EPP = 0) or extra-pair partner (EPP = 1). Note that in these analyses, an observation is a mating event that produced offspring, whereas in our analyses of genetic similarity described previously, an observation is the mean degree of genetic similarity between 2 partner types of a single individual. We constructed competing models in which genetic similarity was a function of partner type (extra-pair or social partner), same-sex:opposite-sex ratio within the social unit, and their interaction. The same-sex:opposite-sex ratio depends on the sex of the focal individual used in the response. When the focal individual is female, the same-sex:opposite-sex ratio is females:males, whereas when the focal individual is male, the same-sex:opposite-sex ratio is males:females. Thus, same-sex:opposite-sex ratio increases with the relative frequency of whichever sex in the social unit matches the sex of the focal individual.
Results
We examined a total of 42 social units containing more than 1 adult, of which 25 were male–female pairs and 17 were groups containing more than a male–female pair. The number of adult voles in a group ranged from 3 to 7 individuals. Genetic similarity values among opposite-sex adults in the same social unit ranged from −0.37 to 1.00 with a mean (±SE) of 0.05 ± 0.02 (n = 94 male:female dyads).
Although it is possible for adult female prairie voles to produce a litter every 21 days during the breeding season (Nadeau 1985), we did not detect evidence that any females in our study produced more than 1 litter during the 4-week trapping period. We were able to assign both biological parents with 95% confidence to only about 62.4% of the juveniles trapped on our study site. Thirty-two percent (8/25) of females living in a male–female pair and 20% (9/44) of females living in a group engaged in EPP. Only 3 of these litters were sired by more than 1 male. For males, 32% (8/25) living in a pair and 25% (7/28) living in a group engaged in EPP. Given the relatively small sample sizes, the differences in these percentages were only suggestive. The best model of group type and sex included only group type and not sex, as well as no random effect of nest ID (Table 1)—its posterior distribution of the effect of groups versus pairs showed an only 87% probability that individuals in groups were less likely to engage in EPP than individuals in male–female pairs.
Model comparisons for each analysis; models ranked by out-of-sample performance. Leave-one-out cross-validation assesses models for their error in predicting new observations. Better models have lower expected log pointwise predictive densities (ELPD). Difference refers to the difference between each model and the best model, and SE Diff is the standard error of the difference. Pairs of models with differences that are less than 1 SE apart have similar performance; however, the best models have the lowest uncertainty with effects least likely to include 0. The text reports only the best models. grp type = social unit type (pair or group); nest.unique = nest ID; r = genetic similarity; partner type = extra-pair or social partner; same-sex ratio = same-sex:opposite-sex ratio.
| Model | ELPD | SE ELPD | Difference | SE Difference |
|---|---|---|---|---|
| EPP vs. sex and social unit type | ||||
| EPP ~ grp type | −69.0 | 5.14 | 0 | 0 |
| EPP ~ sex | −69.6 | 5.08 | −0.6 | 1.23 |
| EPP ~ grp type + (1 | nest.unique) | −69.8 | 5.20 | −0.7 | 0.40 |
| EPP ~ sex + grp type | −70.1 | 5.23 | −1.0 | 0.33 |
| EPP ~ sex + (1 | nest.unique) | −70.4 | 5.15 | −1.4 | 1.23 |
| EPP ~ sex + grp type + (1 | nest.unique) | −70.9 | 5.30 | −1.9 | 0.52 |
| EPP ~ sex * grp type | −71.2 | 5.35 | −2.1 | 0.47 |
| EPP ~ sex * grp type + (1 | nest.unique) | −72.0 | 5.42 | −3.0 | 0.64 |
| Difference in genetic similarity between social and extra-pair partners (among individuals engaging in EPC) | ||||
| r ~ partner type + (1 | nest.unique) | −1.0 | 4.56 | 0 | 0 |
| r ~ partner type | −3.4 | 4.55 | −2.4 | 1.68 |
| Difference in genetic similarity between social and extra-pair partners (among all individuals) | ||||
| r ~ partner type + (1 | nest.unique) | 49.8 | 8.95 | 0 | 0 |
| r ~ partner type | 3.7 | 14.04 | −46.1 | 14.28 |
| EPP vs. genetic similarity and social unit sex ratio | ||||
| EPP ~ same-sex ratio + r | −72.8 | 6.7 | 0 | 0 |
| EPP ~ same-sex ratio * r | −73.1 | 6.8 | −0.3 | 0.77 |
| EPP ~ same-sex ratio + r + (1 | nest.unique) | −73.6 | 6.8 | −0.9 | 0.19 |
| EPP ~ same-sex ratio | −75.8 | 6.7 | −3.0 | 2.66 |
| Model | ELPD | SE ELPD | Difference | SE Difference |
|---|---|---|---|---|
| EPP vs. sex and social unit type | ||||
| EPP ~ grp type | −69.0 | 5.14 | 0 | 0 |
| EPP ~ sex | −69.6 | 5.08 | −0.6 | 1.23 |
| EPP ~ grp type + (1 | nest.unique) | −69.8 | 5.20 | −0.7 | 0.40 |
| EPP ~ sex + grp type | −70.1 | 5.23 | −1.0 | 0.33 |
| EPP ~ sex + (1 | nest.unique) | −70.4 | 5.15 | −1.4 | 1.23 |
| EPP ~ sex + grp type + (1 | nest.unique) | −70.9 | 5.30 | −1.9 | 0.52 |
| EPP ~ sex * grp type | −71.2 | 5.35 | −2.1 | 0.47 |
| EPP ~ sex * grp type + (1 | nest.unique) | −72.0 | 5.42 | −3.0 | 0.64 |
| Difference in genetic similarity between social and extra-pair partners (among individuals engaging in EPC) | ||||
| r ~ partner type + (1 | nest.unique) | −1.0 | 4.56 | 0 | 0 |
| r ~ partner type | −3.4 | 4.55 | −2.4 | 1.68 |
| Difference in genetic similarity between social and extra-pair partners (among all individuals) | ||||
| r ~ partner type + (1 | nest.unique) | 49.8 | 8.95 | 0 | 0 |
| r ~ partner type | 3.7 | 14.04 | −46.1 | 14.28 |
| EPP vs. genetic similarity and social unit sex ratio | ||||
| EPP ~ same-sex ratio + r | −72.8 | 6.7 | 0 | 0 |
| EPP ~ same-sex ratio * r | −73.1 | 6.8 | −0.3 | 0.77 |
| EPP ~ same-sex ratio + r + (1 | nest.unique) | −73.6 | 6.8 | −0.9 | 0.19 |
| EPP ~ same-sex ratio | −75.8 | 6.7 | −3.0 | 2.66 |
Model comparisons for each analysis; models ranked by out-of-sample performance. Leave-one-out cross-validation assesses models for their error in predicting new observations. Better models have lower expected log pointwise predictive densities (ELPD). Difference refers to the difference between each model and the best model, and SE Diff is the standard error of the difference. Pairs of models with differences that are less than 1 SE apart have similar performance; however, the best models have the lowest uncertainty with effects least likely to include 0. The text reports only the best models. grp type = social unit type (pair or group); nest.unique = nest ID; r = genetic similarity; partner type = extra-pair or social partner; same-sex ratio = same-sex:opposite-sex ratio.
| Model | ELPD | SE ELPD | Difference | SE Difference |
|---|---|---|---|---|
| EPP vs. sex and social unit type | ||||
| EPP ~ grp type | −69.0 | 5.14 | 0 | 0 |
| EPP ~ sex | −69.6 | 5.08 | −0.6 | 1.23 |
| EPP ~ grp type + (1 | nest.unique) | −69.8 | 5.20 | −0.7 | 0.40 |
| EPP ~ sex + grp type | −70.1 | 5.23 | −1.0 | 0.33 |
| EPP ~ sex + (1 | nest.unique) | −70.4 | 5.15 | −1.4 | 1.23 |
| EPP ~ sex + grp type + (1 | nest.unique) | −70.9 | 5.30 | −1.9 | 0.52 |
| EPP ~ sex * grp type | −71.2 | 5.35 | −2.1 | 0.47 |
| EPP ~ sex * grp type + (1 | nest.unique) | −72.0 | 5.42 | −3.0 | 0.64 |
| Difference in genetic similarity between social and extra-pair partners (among individuals engaging in EPC) | ||||
| r ~ partner type + (1 | nest.unique) | −1.0 | 4.56 | 0 | 0 |
| r ~ partner type | −3.4 | 4.55 | −2.4 | 1.68 |
| Difference in genetic similarity between social and extra-pair partners (among all individuals) | ||||
| r ~ partner type + (1 | nest.unique) | 49.8 | 8.95 | 0 | 0 |
| r ~ partner type | 3.7 | 14.04 | −46.1 | 14.28 |
| EPP vs. genetic similarity and social unit sex ratio | ||||
| EPP ~ same-sex ratio + r | −72.8 | 6.7 | 0 | 0 |
| EPP ~ same-sex ratio * r | −73.1 | 6.8 | −0.3 | 0.77 |
| EPP ~ same-sex ratio + r + (1 | nest.unique) | −73.6 | 6.8 | −0.9 | 0.19 |
| EPP ~ same-sex ratio | −75.8 | 6.7 | −3.0 | 2.66 |
| Model | ELPD | SE ELPD | Difference | SE Difference |
|---|---|---|---|---|
| EPP vs. sex and social unit type | ||||
| EPP ~ grp type | −69.0 | 5.14 | 0 | 0 |
| EPP ~ sex | −69.6 | 5.08 | −0.6 | 1.23 |
| EPP ~ grp type + (1 | nest.unique) | −69.8 | 5.20 | −0.7 | 0.40 |
| EPP ~ sex + grp type | −70.1 | 5.23 | −1.0 | 0.33 |
| EPP ~ sex + (1 | nest.unique) | −70.4 | 5.15 | −1.4 | 1.23 |
| EPP ~ sex + grp type + (1 | nest.unique) | −70.9 | 5.30 | −1.9 | 0.52 |
| EPP ~ sex * grp type | −71.2 | 5.35 | −2.1 | 0.47 |
| EPP ~ sex * grp type + (1 | nest.unique) | −72.0 | 5.42 | −3.0 | 0.64 |
| Difference in genetic similarity between social and extra-pair partners (among individuals engaging in EPC) | ||||
| r ~ partner type + (1 | nest.unique) | −1.0 | 4.56 | 0 | 0 |
| r ~ partner type | −3.4 | 4.55 | −2.4 | 1.68 |
| Difference in genetic similarity between social and extra-pair partners (among all individuals) | ||||
| r ~ partner type + (1 | nest.unique) | 49.8 | 8.95 | 0 | 0 |
| r ~ partner type | 3.7 | 14.04 | −46.1 | 14.28 |
| EPP vs. genetic similarity and social unit sex ratio | ||||
| EPP ~ same-sex ratio + r | −72.8 | 6.7 | 0 | 0 |
| EPP ~ same-sex ratio * r | −73.1 | 6.8 | −0.3 | 0.77 |
| EPP ~ same-sex ratio + r + (1 | nest.unique) | −73.6 | 6.8 | −0.9 | 0.19 |
| EPP ~ same-sex ratio | −75.8 | 6.7 | −3.0 | 2.66 |
Among voles that produced offspring with individuals outside the social unit, our best model included the random effect of nest ID (Table 1). We found that the genetic similarity between extra-pair partners was less than between social partners (Fig. 1). The expected difference in genetic similarity was −0.080 (90% Bayesian highest probability density interval: [−0.1613, −0.0005]), and the probability that the differences were less than 0 was P = 0.951. When we included all social partners, whether or not they produced offspring or engaged in EPP, the probability increased to P = 0.995.
Among only prairie voles (Microtus ochrogaster) that engaged in EPP, genetic similarity between extra-pair partners (EPR) was lower than between social partners (SPR). “EPR − SPR” shows the expected difference (extra-pair minus social partner) in genetic similarity between different types of mating partners, based on the posterior probability given the data and model. “SPR” shows the expected genetic similarity of social partners and is centered close to 0. Circles indicate the mean genetic similarity values, heavy black lines are 50% credible intervals, and thin black lines are 90% credible intervals.
Our best model for extra-pair parentage included only main effect of genetic similarity and sex ratio and no random effect of nest ID (Table 1). The model showed strong evidence that the probability that an offspring was produced by EPP was higher among voles with lower genetic similarity (90% credible interval: −4.9 to −1.3) and that lived in groups with relatively more opposite-sex social partners (90% credible interval: −1.2 to −0.1; Fig. 2). We found no evidence that adult female body mass differed with the proportion of adult females in the social unit (r = −0.098; P = 0.439).
Among all prairie voles (Microtus ochrogaster), the expected posterior probability of EPP for each sex declines with increasing genetic similarity (r), seen from left panel where r = −0.5 to the right panel where r = 0.5 (difference due to genetic similarity < 0, P > 0.99). Expected EPP also declines as the proportion of same-sex nest partners increases (x axis; slope < 0, P = 0.980). Vertical gray bars are 90% credible intervals.
Discussion
The negative consequences of inbreeding have been hypothesized to be an important driver of EPC in socially monogamous species (Reid et al. 2015). We found strong evidence that genetic similarity between males and females that produced offspring via EPC was less than that between social partners, supporting the hypothesis that EPP can be a mechanism to avoid inbreeding. The greater genetic similarity of social partners compared to extra-pair partners was true whether we considered only those voles that produced offspring with individuals outside the social unit or when we included all social pairs regardless of whether they produced offspring with individuals within or outside the social unit. This result indicates that extra-pair parentage could reduce inbreeding depression within litters.
Voles that never produced offspring by EPC had higher mean genetic similarity than voles that produced offspring with their social partner and via EPC. This result appears to be inconsistent with the conclusion that extra-pair parentage could reduce inbreeding depression. However, when we looked more closely at the data, we found that mean genetic similarity among voles that never produced offspring by EPC was greatly influenced by 1 male and 2 females within the same group that had extremely high genetic similarities (i.e., r > 0.88), whereas all other pairs of social partners or extra-pair partners had r < 0.50 (Supplementary Data SD1). Therefore, this result might not be inconsistent with the hypothesis that extra-pair parentage reduces inbreeding depression after all. Additionally, it is likely that the factors influencing EPC in prairie voles are complex and avoiding inbreeding is unlikely to be the only factor influencing EPC.
We also hypothesized that social unit structure would influence extra-pair parentage. Contrary to our expectations, when there was a greater proportion of females in the group, there was a lower chance of females producing offspring via EPC. Similarly, the chance of EPP by males declined as the proportion of males within the group increased. Finally, we found that the type of social unit did not influence offspring production by EPC. Males and females, whether in pairs or groups, were equally likely to produce offspring through EPC.
Our study provides support for the hypothesis that EPC plays a role in reducing the cost of inbreeding depression. Support for this hypothesis also comes from a few other studies of socially monogamous mammals. In alpine marmots (Marmota marmota; Cohas et al. 2008) and Lariang tarsiers (Tarsius lariang; Driller et al. 2009), females that produced offspring with males from outside their social unit were less genetically similar to these males than to their male social partner(s), which should reduce the cost of inbreeding. Ethiopian wolves (Canis simensis) show very low rates of successful dispersal, which creates the potential for incestuous matings (Sillero-Zubiri et al. 1996). Among observed copulations, 70% were between dominant females from one pack and an adult male from another pack, despite the presence of opposite-sex social partners in the natal group of the female, suggesting that Ethiopian wolves also avoid inbreeding by high rates of EPC.
The process by which prairie voles establish social pairs in nature is unknown, but pair formation may be opportunistic, with adults likely pairing with the first available opposite-sex conspecific that they encounter (Getz et al. 2004). A field experiment with prairie voles maintained in seminatural enclosures found that social pair formation was not influenced by genetic relatedness (Lucia and Keane 2012). The apparent lack of kin-based discrimination during social pair formation, coupled with the absence of sex differences in natal dispersal rates or distances, makes it likely that social pairs sometimes form between kin in natural populations (McGuire et al. 1993). Furthermore, studies of laboratory and field populations of prairie voles demonstrate that there are substantial costs to inbreeding (Bixler and Tang-Martinez 2006; Lucia and Keane 2015). If the probability of social partners being related is non-negligible and inbreeding depression is substantial, this could favor the evolution of inbreeding avoidance mechanisms, including individuals actively seeking copulations with less related mates outside the social unit (Charlesworth and Charlesworth 1987).
Models predict that time constraints on reproduction can result in individuals tolerating some degree of inbreeding depression if delaying reproduction to find a less related mate leads to a decrease in lifetime reproductive success (Batova et al. 2021). The very short average breeding lifespan of prairie voles (2 to 3 months) may explain why individuals do not appear to discriminate among social partners on the basis of kinship during pair formation and therefore risk some degree of inbreeding depression (Getz et al. 2004; Lucia and Keane 2012). A potential decrease in lifetime reproductive success caused by delaying pair formation to find a less related social partner may have led to the evolution of mating with multiple opposite-sex conspecifics through EPC in prairie voles as a genetic bet-hedging strategy to reduce the production of inbred offspring even in the absence of any kin discrimination capabilities (Watson 1991; Stockley et al. 1993). Yellow ground squirrels (Spermophilus fulvus) are solitary rodents in which females are sexually receptive for only 1 day per year (Batova et al. 2021). Time constraints on the opportunity for breeding likely limit mate choice and female yellow ground squirrels do not mate selectively with unrelated males. On average, a female Yellow Ground Squirrel mates with 2 males, which appears to dilute the costs of producing inbred offspring enough to maintain an overall low level of inbreeding.
Social unit composition also could be an important factor influencing the occurrence of EPC because it can affect mating opportunities with residents of a social unit as well as the ability or motivation to mate with individuals that are not members of the social unit (Isvaran and Clutton-Brock 2007). Contrary to our expectations, we found that the likelihood of females producing extra-pair offspring decreased when the proportion of females within social units increased and there were fewer males to mate guard females. This suggests that male mate guarding may not be as important a deterrent to EPC in prairie voles as had been thought or that the mate guarding of females does not increase when there are more males in a group. This idea is supported by data showing that, in seminatural populations, the frequency of multiple paternity in litters of female prairie voles living with a male social partner did not differ from that of females living singly (Lichter et al. 2020). Female behavior, such as mate choice, might be a more important determinant of which males they mate with than male mate guarding. If the males in social units with greater female-to-male sex ratios were high-quality males, then females may prefer to mate with them than with nonresident males. These higher-quality males may also attract more females to these social units.
We expected that competition for access to females in the social unit from males living in the same social unit should increase the propensity for EPC by males. Instead, we found that males sired more pups through EPC as the proportion of males within social units decreased, which should have decreased mate competition within a group. One possible explanation for this result is that the females in these social groups may have been closely related to the males (e.g., due to natal philopatry) and so males were mating outside the group for that reason. Another possibility is that females in these social units with fewer males were of lesser quality or not as fecund as nonresident females. One way this could occur is if females were living with male siblings, which can suppress growth of females (Batzli et al. 1977). In a laboratory study, female body mass did not influence litter size but did influence pup growth prior to weaning when litter size was held constant (Solomon 1994). If pups gained weight more quickly, especially when they were very young, perhaps they may be better at escaping from predators or acquiring higher-quality territories when dispersing. In the current study, resident females in groups with less male-biased sex ratios did not have a lower body mass than females in other social units, so female body mass would not have influenced male reproductive success if pup growth influenced pup survival. Furthermore, Keane et al. (2007) found that breeding among males and females was random with respect to body condition (the ratio of observed body mass to expected body mass based on body length). A review of 24 mammalian species found no differences in male EPP in socially monogamous versus socially polygynous species (Clutton-Brock and Isvaran 2006). Overall, these results suggest that the potential for increased sexual competition among multiple males living in a social unit does not result in more EPC by males, although why this is the case is not clear.
Engaging in EPC is likely a costly endeavor, and numerous benefits have been proposed as to why females, in particular, in socially monogamous species of mammals engage in EPC (Berteaux et al. 1999). Whereas the hypothesized benefits of EPC are not mutually exclusive, our study joins a small but increasing number of studies of socially monogamous mammals, showing that EPC may function to reduce inbreeding depression within a litter in the population we studied. However, rates of EPP in prairie voles vary geographically and are correlated with vegetation characteristics and population density (Streatfeild et al. 2011). This indicates that the proximate factors influencing EPC in prairie voles are complex and future field studies examining the range of potential benefits of EPC and ecological factors (e.g., social setting, population density) will be important for understanding the evolution of EPC in socially monogamous mammalian species.
Supplementary data
Supplementary data are available at Journal of Mammalogy online.
Supplementary Data SD1. Plot of mean genetic similarity between (i) prairie vole pairs that produced offspring through extra-pair copulation (EPC), (ii) each adult and all opposite-sex social partners in the same social unit, and (iii) each adult and all opposite-sex social partners in the same social unit for voles that did not engage in EPC.
Acknowledgments
We thank K. Clay for logistical support at the Indiana University field site. Thanks to A. Kiss and C. Wood for their help at the Miami University Center for Bioinformatics and Functional Genomics. We also thank K. Mabry, C. Streatfeild, and numerous undergraduate students who assisted with the field and laboratory research.
Author contributions
BK: conceptualization, investigation, data curation, methodology, writing—original draft, funding acquisition. MHHS: formal analysis, writing—reviewing and editing. NGS: conceptualization, investigation, data curation, methodology, writing—reviewing and editing, funding acquisition.
Funding
Funding was provided by the National Institutes of Health (NIGMS GM 06409-01 to NGS) and the National Science Foundation (IOS-0614015 to BK and NGS).
Conflict of interest
None declared.
Data availability
Data are available upon request.
