Enhancing knowledge of the ecology of a highly elusive species, the okapi (Okapia johnstoni), using non-invasive genetic techniques

Okapi (Okapia johnstoni) are an even-toed ungulate in the family Giraffidae, and are endemic to the Democratic Republic of Congo (DRC). Very little is known about okapi ecology in the wild. We used non-invasive genetic methods to examine the social structure, mating system and dispersal for a population of okapi in the Réserve de Faune à Okapis, DRC. Okapi individuals appear to be solitary, although there was some evidence of genetically similar individuals being associated at a very small spatial scale. There was no evidence for any close spatial association between groups of related or unrelated okapi but we did find evidence for male-biased dispersal. Okapi are genetically polygamous or promiscuous, and are also likely to be socially polygamous or promiscuous. An isolation by distance pattern of genetic similarity was present, but appears to be operating at just below the spatial scale of the area investigated in the present study. We therefore here provide new ecological information about a species that has recently been recognised by the IUCN as Endangered, and is a potentially important flagship species for Central Africa. INTRODUCTION The key to protecting and managing species of conservation concern is a good understanding of their ecology, including knowledge of their dispersal, sociality and mating system [1,2]. This information can have a considerable and very real impact on conservation (e.g. [3-5]). However, measuring these factors in wild animals by direct observation can often be very difficult, especially for elusive mammals, or those inhabiting difficult terrain [6,7]. There is a vast amount of variation in social structure, mating systems and dispersal strategy amongst mammals, even among those that are taxonomically and geographically similar [8-14]. This variation in social structure makes predictions of ecological and genetic ???(processes) difficult for any poorly studied mammal species. In terms of social structure, mammals that utilise densely forested habitats tend towards forming a smaller social unit, putatively because the coordination of a social group is difficult in a forest especially if the animal is large [15]. Also, animals at greater risk of predation are more likely to adopt a hiding strategy [15] and be predominantly solitary to reduce social interaction and therefore detection probability [16]. Mating systems are even more diverse (20) and difficult to predict. For example, the extent of polygamy can be affected by predation pressure [17], social group composition [18] and phylogeny [19]. Due to this complex interaction, mammals show a diverse array of mating systems, true for both males and females [20]. Dispersal (specifically natal dispersal [21]) also often varies between sexes, with some degree of sex-biased dispersal being virtually ubiquitous in mammals [22]. However, male-biased dispersal is the norm for mammals [23]. Due to this lack of predictive power of habitat and taxonomy, other methods are clearly needed to accurately elucidate the ecology of elusive, or otherwise difficult to observe animals. Non-invasive genetic methods are increasingly being used to investigate questions such as dispersal, mating systems and social structure in wild animals [24-26]. These methods may therefore provide a means of investigating the ecology of elusive animals without actually observing them. The okapi is a highly elusive even-toed ungulate, endemic to the Democratic Republic of Congo (DRC). Although widely distributed throughout the DRC, it occurs at low density across its range [27]. Okapi appear to only be present in dense forest, away from human presence [28,29]. Determining aspects of behavioural ecology using observations is therefore difficult for this species. Only two in situ ecological studies of okapi have been published [30,31]. However the studies are somewhat equivocal, are lacking in detail, and tell us nothing of okapi mating systems or dispersal. Noninvasive genetic methods therefore potentially provide a useful tool for the study of this species. We hypothesised that okapi are mostly solitary, due to their utilisation of dense rainforest, and the likelihood of them having a high predation pressure [30]). In captivity, okapi males are rotated among females and sire multiple offspring [32]. We hypothesised that this would also be true in the wild, with okapi showing evidence of genetic polygamy, or promiscuity. We also hypothesised that okapi would demonstrate male-biased dispersal, due to its higher incidence in mammals. The above hypotheses will be tested using dung samples from okapi in a population in the okapi faunal reserve (Réserve de Faune à Okapis, RFO), DRC. METHODS Study species and site Okapi are an even-toed ungulate in the family Giraffidae, separated from the giraffe by an estimated ~16 million years of independent evolution [33]. The limited number of long-term ecological studies that have been carried out on okapi have been based in the RFO [30,31] and this reserve was also chosen for the present study (Figure 1). Four teams sampled the park, between December 2010 and February 2011, and collected 208 putative okapi fecal samples. These samples were collected as part of a great ape and human monitoring survey [34],..Briefly, surveys comprised a total of 164X one km transects, and fecal samples were collected on and between transects [34]. Transect location was determined randomly using the program DISTANCE 6.0 [35]. Each transect was walked once. DNA extraction and amplification DNA was extracted from faecal samples (stored in 100% ethanol for 24 hrs and then silica) using a QIAamp DNA Stool Mini Kit (Qiagen). Thirteen microsatellite loci were amplified using the primers Oka-01–13 and PCR conditions from Stanton et al. [36]. Primers Oka-02, 10 & 11 were excluded from the analysis due to low PCR amplification success rate. From the 208 faecal samples, consensus genotypes were generated for 105. These 105 samples were confirmed to be okapi based on the following: 1) Correct species identification from this survey was 100% based on mitochondrial DNA analysis of a subset of samples (Stanton et al (submitted)). 2) Genetic structure and distance analysis of microsatellite data in the present study did not identify any unusually different genotypes within the 105 genotyped samples. The primer sequences SRY 1 (5’ CTTCATTGTGTGGTCTCGTG 3’) and SRY 2 (5’ CGGGTATTTGTCTCGGTGTA 3’; Wilson and White [1998]) were used to amplify a fragment in 5 blood samples from captive male okapi. Internal primers OJSEX-F (5’ CGTGAACGAAGACGAAAG 3’) and OJSEX-R (5’ TCAATATCTGTAAGCCTTTTCC 3’) were designed to amplify a shorter 101 bp fragment in non-invasive okapi samples. Sexing primers were multiplexed with an internal control, Oka-01 (forward: 5’ AAGAGAGACTGCACTGTGGACC 3’, reverse: 5’ GCTCTTGTGTCTGACATGTTCTC 3’, [36]). PCR was carried out in a 6.5 μl volume with 2.5 μl Multiplex Mix (Qiagen), 4 μg BSA, 2 nmol OJSEX primer, 0.8 nmol Oka01 and 2 μl DNA. The PCR was carried out twice for each of the samples that had been successfully genotyped, always with two negative controls. A sample was accepted as a female if both reactions showed the absence of a band from the sexing primers. Primers Mt 1 – 5 (Stanton et al. (submitted)) were used to amplify a fragment of the mitochondrial DNA control region (mtDNA CR), and cytochrome b, tRNA-Thr and tRNA-Pro genes in individuals with sexing information, using the conditions from (Stanton et al. (submitted)). A 325 bp fragment was amplified in 20 individuals (females n = 9, males n = 11), and a 543 bp fragment (that included the 325 bp fragment above) was amplified in a further 15 individuals. Data validation A preliminary genotyping error rate study was carried out using the programs PEDANT [38] and GEMINI [39] on 14 okapi faecal samples, comparing two genotyping repeats of each sample. GEMINI indicated that 2-3 repeats would be required to be able to accept a consensus genotype with >95% confidence, and PEDANT calculated an allelic dropout rate for each locus at between 0.0170 and 0.1645 (mean 0.0779), a false allele rate of between 0 and 0.0718 (mean 0.0170). The confidence converged on 100% with approximately three repeats. Therefore, for caution, at least four repeats (and up to eight) for each of the samples in the full study were carried out. Genotyping error rates were then recalculated on the full dataset. The allelic dropout rate for each locus was between 0 and 0.0429 (mean 0.0161) and false allele rate was between 0 and 0.0055 (mean 0.0010), demonstrating that the four repeats carried out were sufficient to give reliable consensus genotypes at the 95% level. Spatial autocorrelation To test the hypothesis of low social structure in okapi, the relationship between proximity of okapi dung samples and genetic distance was investigated. This was to determine if related individuals are spatially more closely associated than unrelated individuals, and was carried out using spatial autocorrelation analysis (SAA). Spatial autocorrelation measures the degree of dependency of observations, for example genetic distance, across space. Significant positive autocorrelation (in the example of genetic distance) indicates that genetically similar individuals are closer together than one would expect by chance, whereas significant negative autocorrelation indicates that individuals are arranged to maximize genetic distance between them [40]. We carried out spatial autocorrelation analysis using GenAlEx v6.4 [41,42], with significance assessed using 95% confidence interval and 9999 permutations. SAA was carried out on males (n = 27) and females (n = 29) separately, and on the combined dataset (n = 83) at distance intervals of (i) 2 km across 20 km, and (ii) 10 km across 120 km. The analysis was carried out on the combined dataset only (n = 83; there was insufficient data to analyse males and females separately) at distance intervals of 0.2 km across 2 km. Patterns of relatedness To further describe sociality of okapi in the study site, and to complement the spatial autocorrelation analysis, the association between spatial proximity and genetic relatedness was investigated. Pairwise relatedness was estimated using the program COANCESTRY v1.0.1.2 [43], which implements seven methods for estimating pairwise relatedness from individual multilocus genotypes. Duplicate genotypes were removed from the dataset and the spatial proximity of related dyads in the remaining individuals (n = 83) was described. This was done by investigating if there were significant differences between average spatial proximity of dyads with a relatedness greater than 0.5 verses less than 0.5, and greater than 0.25 verses less than 0.25, using t-test tests in R (R Development Core Team). This was carried out for all seven estimators. A rarefaction analysis was also carried out on the microsatellite genotypes using the program RERAT [44] to investigate the ability of the 10 markers used in the present study for inferring relatedness.