Chimpanzees breed with genetically dissimilar mates

Inbreeding adversely affects fitness, whereas heterozygosity often augments it. Therefore, mechanisms to avoid inbreeding and increase genetic distance between mates should be advantageous in species where adult relatives reside together. Here we investigate mate choice for genetic dissimilarity in chimpanzees, a species in which many females avoid inbreeding through dispersal, but where promiscuous mating and sexual coercion can limit choice when related adults reside together. We take advantage of incomplete female dispersal in Gombe National Park, Tanzania to compare mate choice for genetic dissimilarity among immigrant and natal females in two communities using pairwise relatedness measures in 135 genotyped chimpanzees. As expected, natal females were more related to adult males in their community than were immigrant females. However, among 62 breeding events, natal females were not more related to the sires of their offspring than immigrant females, despite four instances of close inbreeding. Moreover, females were generally less related to the sires of their offspring than to non-sires. These results demonstrate that chimpanzees may be capable of detecting relatedness and selecting mates on the basis of genetic distance.


Introduction
Inbreeding depression, a reduction of fitness in the offspring of closely related mates, is a widely documented phenomenon thought to result largely from the homozygous expression of recessive deleterious alleles [1][2][3]. Inbreeding can adversely affect fitness through numerous pathways, notably via early death or reproductive disadvantage [3]. In contrast, fitness is often positively correlated with increased heterozygosity, whether by male behaviour [47]. Daughters resisted the mating attempts of fathers at higher rates than those of unrelated males. There is some evidence for preferential interaction between males and their infant and juvenile offspring in this species [48,49], but the degree to which preferential interaction and/or association persists past juvenility is uncertain. Therefore, some form of phenotype matching may be involved in the identification of fathers in adulthood. In another study, dyadic relatedness calculated from microsatellite data was negatively related to mating frequency but mating did occur in most dyads [41]. Taken together, these patterns suggest that there is some recognition of relatedness which influences mating patterns in this species but that mating tactics may constrain choice for unrelated partners. As yet, the consequences of these patterns for the parentage of infants have not been examined. That is, the extent to which inbreeding is successfully avoided or genetic dissimilarity is favoured is not yet known.
Here we take advantage of differing demographic characteristics in two eastern chimpanzee communities in Gombe National Park, Tanzania to investigate patterns of mate selection and the factors influencing these by examining the parentage of 64 infants. The two communities (Kasekela and Mitumba) differ in size, number of adult males and the proportion of females that disperse. The larger Kasekela community is unusual in that about 50% of females breed in their natal community [50], whereas in Mitumba, 83% of females have dispersed, and the community currently contains only one natal female. First, we ask how effectively inbreeding is avoided by examining the incidence of moderate and close inbreeding based on known pedigree information. Second, we test our expectation based on sex-biased dispersal patterns that natal females are more closely related to resident males, and therefore at higher risk of inbreeding than immigrant females. Third, we examine the extent to which natal females are able to avoid inbreeding by testing whether they are more closely related than immigrant females are to the sires of their offspring. Last, we test the hypothesis that, because of the general advantages of heterozygosity, there will be overall mate choice for genetic dissimilarity whereby breeding dyads are less related than non-breeding dyads.

Study site
Gombe National Park, in western Tanzania, has an area of 35 km 2 and contains three chimpanzee communities. The chimpanzees of the Kasekela and Mitumba communities have been habituated and observed almost daily since 1966 and 1995, respectively [51], and the dispersal patterns of all females born into or entering these communities are known. Individuals in the Kalande community are known genetically from faecal sampling [52]. All individuals surviving beyond 2 years of age and alive between 1995 and 2012 in Kasekela and2004 and2012 in Mitumba have been genotyped. Within these periods, the Kasekela community included 12-25 adult females (age 12+ years) and 10-12 adolescent and adult males (age 12+ years), and the Mitumba community included five to nine adult females and two to five adolescent and adult males.
Since the 1990s, the park has been isolated from other forested areas and is primarily surrounded by anthropogenic landscapes, potentially limiting dispersal opportunities [53]. However, healthy females occasionally disappear from the park shortly after sexual maturity and females of unknown provenance sometimes immigrate into communities in the park. In the last 5 years, two immigrant females have been determined to originate from outside the park; they were never observed in Mitumba or Kasekela and their genotypes did not match any known chimpanzee, including those genetically sampled in the Kalande community. A chimpanzee community of at least 20 individuals is known to reside approximately 12 km north of the park (D. Mjungu 2016, personal communication). Although this community is likely to be a potential source and destination for dispersing females, it is highly unlikely that offspring born within the park are sired by males outside the park, because extra-group paternities are rare in chimpanzees [38,[54][55][56][57] and, in this case, would require significant travel across risky landscapes by within-park females. Paternities for 37 offspring born into the Kasekela community had previously been reported (including one set of non-identical twins (GLI, GLD) whose paternity was tested individually) [38,52,56,58]. For this study, we repeated the analysis for these 37 paternities and we also report paternities for 28 new individuals born into the Kasekela and Mitumba communities. This provides a total sample size of 65 offspring with known paternities: 52 in the Kasekela community and 13 in the Mitumba community (electronic supplementary material, figure S2). All infants had known mothers that were also genotyped, and we had good sampling of candidate males, with genotypes for a mean of 97.2% of candidate males from within the community, and 91.6% of males between the two habituated communities (KK and MT; electronic supplementary material, figure S2). Ten is the youngest age documented for a chimpanzee sire in the wild [59], whereas 11 is the youngest age of any sire at Gombe (2015, unpublished data). All males at least 9 years of age at the time of conception that were resident in the mother's community (either Kasekela or Mitumba) were genotyped for 57 conceptions (87.7%), and all males were sampled from both communities for 41 (63.1%) conceptions (electronic supplementary material, figure S2). Fathers were first identified using the exclusion principle and confirmed with likelihood methods using the program CERVUS [60]. Detailed information on the paternity assignment procedure is included in the electronic supplementary material.

Pedigrees and inbreeding
We classified offspring as inbred if the father was related to the mother at the level of half-siblings (R = 0.25) or above based on known pedigrees. Using known pedigrees combined with genetic paternity information, we can detect any inbreeding between parent-offspring dyads, full siblings and maternal half-siblings for all genotyped offspring but can detect inbreeding between paternal siblings in only 53 of 65 offspring. Because extra-group paternity is rare in chimpanzees [38,[54][55][56][57], we assumed that sires of offspring born to immigrant mothers were unrelated to the mother. Among 32 offspring born to natal mothers, 10 were born to parents where the father of both the mother and the sire was unknown, preventing us from determining if they were paternal siblings. Two mothers in Mitumba were present as adults at the time of habituation and their provenance is unknown (i.e. natal or immigrant); to be conservative, the offspring born to these mothers were also excluded from the determination of rates of inbreeding among paternal siblings (n = 2). A lack of generational depth in the pedigree prevented a systematic analysis of moderate inbreeding between other relatives related at the 0.25 level (i.e. grandmothers and grandsons, aunts and nephews, etc.) [61].

Pairwise relatedness measures
We used Co-ancestry to calculate pairwise relatedness [62]. In this programme, allele frequencies were calculated once from the same set of microsatellite genotypes (8-11 loci) of 135 individuals from the Kasekela and Mitumba communities (electronic supplementary material, figure S1c). Bias can result from using the same markers to determine paternity and calculate relatedness [63][64][65]; however, our microsatellite data could not be split into two subsets as recommended, because the full set is necessary to determine paternity. Nonetheless, the high degree of polymorphism and heterozygosity of our markers reduces the potential for bias [63] (electronic supplementary material, figure S1a,c) and we found that when we calculated R-values based on different allele frequency subsets as recommended by Wang [65], values were highly correlated with the full dataset (see electronic supplementary material).
To facilitate comparison with previous work on genetic relatedness in chimpanzees [54,55,66], we used the Queller and Goodnight pairwise relatedness estimator [67] to compare relatedness between classes of dyads. We found that estimates of pairwise relatedness from this method correlated highly with kinship coefficients calculated between dyads of known parentage confirmed through paternity analysis and pedigree information (n = 70) using the KINSHIP2 package in R (see electronic supplementary material, table S1 and figures S3,S4) [68]. These estimates were also correlated with six other pairwise estimators calculated in Co-ancestry [62] (see electronic supplementary material, table S1) including the more recent triadic likelihood (TL) method [69], which allows for inbreeding, weights loci based on their informativeness and accounts for typing errors. We repeated our analysis using the TL method and found similar results (see electronic supplementary material, figures S5-S7).

Definitions and inclusion criteria 2.3.1. Females and offspring
All females present in each community, aged 12 years or over and encountered in more than 10% of male focal follows (i.e. resident rather than peripheral community members [70]) were considered adults for relatedness analyses and were categorized as either natal (n Kasekela = 15, n Mitumba = 1) or immigrant (n Kasekela = 20, n Mitumba = 9) depending on their community of birth. Two females in Mitumba were of unknown provenance (they were present in the community when observations began) and were excluded from analyses in which females of different residence status were compared. Females that did not give birth (n Kasekela = 8, n Mitumba = 3) or have offspring genotyped (n Kasekela = 3, n Mitumba = 1) were excluded from analyses that required the use of breeding females. For the purposes of this study, the conception of twins was treated as a single event because they were sired by the same male (electronic supplementary material, figure S2). This leads to a total of 64 breeding events (51 in Kasekela and 13 in Mitumba).

Males
All males 12 years or older and present in the community during the study period were considered adults for relatedness analyses (n Kasekela = 24, n Mitumba = 9). All males were born in the community in which they resided as adults except one (BE) that immigrated into Kasekela with a female assumed to be his sister at approximately 4 years of age [36]. The dominance rank of each male on the estimated date of conception of each infant [36,71] was given by his Elo score, calculated from the occurrence of submissive pant grunts [72]. In Mitumba, rank was only known for the alpha male, preventing a systematic analysis of rank and relatedness in this community.

Analyses of pairwise relatedness 2.4.1. Female-male dyadic relatedness
To test for differences in the degree of genetic relatedness between natal and immigrant females with resident males, we calculated and compared average relatedness between all immigrant female-male dyads (n = 388) with average relatedness between all natal female-male dyads (n = 279) that overlapped as adults in the Kasekela community. This analysis was restricted to the Kasekela community, beause Mitumba had only one natal female (with no infant). Differences between natal female-male dyads and immigrant female-male dyads were tested for significance using a permutation procedure that compares the normalized difference between means from 10 000 random pairs of replicates from the pooled dataset using JMP v. 11 [73].

Relatedness of natal and immigrant breeding dyads
Relatedness between immigrant mothers and sires of 30 offspring (19 in Kasekela and 11 in Mitumba) was compared with relatedness between natal mothers and sires of 32 offspring (all in Kasekela) in two ways: (i) including both communities and (ii) restricting analysis to just Kasekela since, as above, Mitumba had only one natal female (with no infant). Differences between groups were tested using the same permutation procedure explained above.

Relatedness of breeding and non-breeding dyads
Relatedness between breeding and non-breeding dyads was examined by comparing the relatedness between the mother and sire of an infant (breeding dyads) with the relatedness between the mother and each other adult male present in the community at the conception of her infant (non-breeding dyads) [36,71]. We tested each community separately, using 50 offspring in the Kasekela community and 11 offspring in the Mitumba community. One offspring born to a transient mother in Kasekela (KAR) and two offspring born in Mitumba prior to 2004 (APL and LAM) were excluded from this analysis because not all candidate sires were genotyped. Differences between groups were first compared using the same permutation procedure explained above. To determine if any observed difference was due to female residence status, we also compared breeding and non-breeding dyads among natal and immigrant females separately. Male rank and age influences the probability of paternity [56]. Therefore, to be confident that selection of breeding partners was influenced by genetic distance, in addition to other known factors, we also ran a linear mixed model using the lme4 package [74] in R [75] controlling for male rank and age. Relatedness was the outcome variable and dyad type (breeding or non-breeding), male rank and male age were included as fixed effects, and male and female ID were both included as random effects. To investigate if breeding patterns varied by community, we ran a second linear mixed model where relatedness was the outcome variable, and dyad type, community and the interaction term community × dyad type were included as fixed effects. Male ID and Female ID were also included as random effects.

Paternity and inbreeding
The fathers identified for all 37 offspring reported previously were confirmed here, with one exception (DIA). Typing at additional microsatellite loci resolved paternity between two maternal half-brothers for DIA, which we update here (electronic supplementary material, figure S2). Detailed information on paternity likelihood is included in the electronic supplementary materials. All 65 offspring, including 28 for which paternity is newly reported here, were sired by within-community males except for one (KAR), whose mother (TT) resided primarily in the southern Kalande community but conceived during a visit to Kasekela (table 1). KAR was also determined to be inbred, a product of breeding between maternal siblings. TT emigrated before her maternal brother was born and the conception occurred when she revisited her natal community after a sudden decrease in the adult male population in her transfer community (see the electronic supplementary material) [52].
Of the 28 new infants, two were the products of inbreeding (R ≥ 0.25 as determined from pedigree data) and strong evidence points to a third case (table 1). KAR was the offspring of maternal halfsiblings (as discussed above); GGL and probably GIZ (see the electronic supplementary material) were the offspring of paternal half-siblings.
Offspring conceived by close relatives (R = 0.5; n = 1) accounted for 1.5% of all genotyped infants and 3% of those born to natal females. Offspring conceived by half-siblings (R = 0.25; n = 3) account for 5% of all genotyped infants and 12% of those born to natal females. We currently lack depth of pedigree to identify most offspring born to other relative classes at the 0.25 level; however, in the instances where such relationships were known, no inbreeding was detected.
Three of the four inbred infants died prior to attaining sexual maturity; two died around or prior to 2 years (FI and KAR) and the third (GIZ) died at 6 years. Importantly, none of these infants, nor their mothers, were positive for simian immunodeficiency virus (SIVcpz) infection, which is known to increase infant mortality [76].

Relatedness of natal and immigrant breeding dyads
Despite the fact that natal females were more closely related to resident males, natal females were not more closely related to the sires of their offspring (n = 32 dyads, mean R = −0.052) than were immigrant females (n = 30 dyads, mean R = −0.063; p = 0.81), including data from both communities. There was also no significant difference when analysis was restricted to dyads from the Kasekela community (natal breeding dyads; n = 32 dyads, mean R = −0.052; immigrant breeding dyads: n = 19 dyads, mean R = −0.101; p = 0.35).

Relatedness of breeding and non-breeding dyads
In the Kasekela community, breeding dyads including both natal and immigrant females (n = 50 dyads, mean R = −0.072) were significantly more distantly related than non-breeding dyads (n = 486 dyads, mean R = 0.000, p = 0.01; figure 2). This difference remained significant in a model that included the

Discussion
Our results suggest that, as expected for a slowly reproducing species in which females invest heavily in each offspring and sometimes reside with close relatives, several mechanisms promote inbreeding avoidance and mate choice in chimpanzees. First, sex-biased dispersal reduced the relatedness between breeding adults in the community: immigrant females were more distantly related to the males in the community than were females that did not disperse. Second, although they often resided with close male relatives, natal females were not more closely related to the sires of their offspring than were immigrant females. Third, females in the Kasekela community were more distantly related to the sires of their offspring than to the other resident males. These latter two patterns provide evidence for the existence of mechanisms to detect genetic similarity among individuals and promote outbreeding. Relatedness between breeding and non-breeding dyads in Kasekela by residence status. Among immigrant females, breeding dyads are less related than non-breeding dyads (p < 0.01) and among natal females, there was a trend in the same direction (p = 0.10).
Nonetheless, these systems are not perfect and we identified four cases of inbreeding; one between closely related parents and three between moderately related parents. Three of these occurred between relatives residing together. One resulted from a rare forced son-mother mating [38], whereas two occurred between females that did not disperse from their natal Kasekela community and their male paternal siblings. The fourth case occurred between maternal siblings under unusual circumstances in which the parents never resided together. We may have underestimated the prevalence of inbreeding in the population, because most infants that die prior to age two are not genotyped (19 of 23 offspring in Kasekela and 2 of 2 offspring in Mitumba); however, 15 of these infants were born to immigrant mothers and were unlikely to be inbred. Furthermore, we were only able to measure the extent of inbreeding between half-siblings and parents/offspring because, owing to the long lifespan of chimpanzees, our pedigree was not deep enough to accurately assess rates of other forms of inbreeding [61]. In the cases where we could identify grandparents/grandchildren and aunts/uncles/nieces/nephews, we found no evidence of inbreeding. Although the sample size is small, the outcome of the four cases of inbreeding is suggestive of inbreeding depression. Fifty per cent of inbred offspring died in the first 2 years and 75% died by age six, compared with an overall average infant mortality of 29% in the first 2 years and 40% in the first 6 years [77].
In support of selection for genetically dissimilar mates, we observed that breeding pairs were more distantly related than non-breeding pairs. Importantly, this result was not driven by natal females avoiding close relatives as breeding partners. Rather, we found a general bias among all females, including immigrant females, for unrelated mates. These patterns suggest that chimpanzees are able to detect degrees of genetic relatedness and breed with more distantly related individuals and do not simply avoid relatives with whom they have a history of association. While we found a strong effect of choice for genetic dissimilarity in Kasekela, the effect was diminished in the smaller Mitumba dataset, though when modelled together, no community difference was found. Nevertheless, it is possible that demographic differences between communities may influence patterns of mate choice. In small groups, more effective mate guarding by the alpha male may constrain female choice, whereas larger groups may provide more scope for alternative mating strategies such as consortships [56,78] and future work, with a larger sample, should clarify these effects. Mate choice for genetic dissimilarity is likely to confer fitness benefits on the resultant offspring and infant survival is, thus, expected to vary by the degree of relatedness between the parents. Here, we document high infant mortality in moderate and close inbreeding events, but future work should examine fitness effects across all degrees of relatedness.