Results
A total of 342 individuals for which we had genetic, spatial and disease status data were included in our final dataset. This consisted of 155 males (of which one was a joey <1 year old caught with its mother) and 187 females (of which three were joeys <1 year old caught with their mothers). We were able to detect maternities for 195 individuals, who were the offspring of 106 unique mothers.
Out of the total 342 individuals, there were 60 that tested positive at first capture. Observed heterozygosity was estimated as 0.223, within population genetic diversity (also known as expected heterozygosity) was estimated as 0.278, and the population average inbreeding coefficient, Fis, was estimated as 0.196. The identity disequilibrium proxy (g 2) in our 5007 SNP dataset differed significantly from zero (g 2 = 0.013 ± 0.002 bootstrap confidence interval = 0.0101 – 0.0167, P(g 2 = 0) = 0.001), where permutations = 1000 and bootstraps = 1000). This indicated that our dataset met the requirements of variance between individuals in levels of inbreeding, which is required for detecting if there is inbreeding depression in a population. Accordingly, IR varied substantially between individuals, ranging from a minimum estimate of –0.39 (expected when the individuals’ parents are outbred) to 0.38 (expected when the individuals’ parents are related to one another) (mean IR = 0.05, standard deviation = 0.09, Fig 1a). The variance in pairwise relatedness values was 0.004, with approximately 232 pairs of first-degree relatives (i.e., parent-offspring pairs or full siblings) and 680 pairs of second-degree relatives (see Figure 1b).
There was no evidence for sex differences in the probability that a koala tested positive for C. pecorum (posterior modeβMALE = 0.037, Table 1). Koalas had a higher probability of testing positive for C. pecorum in the breeding season compared to the non-breeding season (posterior modeβBREEDING = 0.623, Table 1). The probability of testing positive for C. pecorum also increased with age (posterior mode β = 0.145, Table 1, Fig 2). There was no evidence that internal relatedness affected the probability that koalas tested positive for C. pecorum (posterior mode β = 1.071, Table 1, Figure 3). Furthermore, there was no evidence that the effect of internal relatedness changed with the age of the koala (IR*age interaction, posterior mode β = -0.038, Table 1). Together, the lack of an association between IR and disease suggests that there was no evidence for inbreeding depression in this population.
The probability of testing positive for C. pecorum was associated with moderate levels of additive genetic variance (VA) on both latent and observed data scales, whereby the posterior distribution was clearly different from zero on both scales (Table 1, see supplement for posteriors). More specifically, the VA for the probability of testing positive for C. pecorum was estimated at 1.35 on the latent scale (posterior mode, 95%CI = 0.23 – 2.93, Table 1), and 0.008 on the observed scale (posterior mode, 95%CI = 0.003 – 0.33, Table 2). The heritability (h2) for the probability of being diseased was estimated at 0.57 on the latent scale (posterior mode, 95%CI = 0.33 – 0.74), and 0.11 on the observed scale (posterior mode, 95%CI = 0.06 – 0.23, Table 1).
We found no evidence that the probability of testing positive forC. pecorum was associated with variance in shared environment effects (VS). The posterior distribution for VS (measured using the home range overlap matrix) was very low and bordered zero on both the latent and observed data scales (latent scale; posterior mode = <0.001, 95%CI= <0.001 – 0.071: observed data scale; posterior mode = <0.001, 95%CI = <0.001 – <0.001, Table 1). This represented <1% of variance in probability of being diseased on both latent and data scales. Furthermore, the inclusion of VS in the model did not improve the fit of the model (DIC VA + VS = 278.69, DIC VA= 277.88). Finally, VA estimates did not vary qualitatively between models with and without the home range overlap matrix, suggesting that its inclusion did not affect our reported estimates of h2(see supplement).
Using a subset of the data that included only individuals for which we knew the identity of their mothers (N = 195), we found only a small maternal effect variance (VME) in the probability of testing positive for C. pecorum . More specifically, the posterior mode of VME was estimated as 0.005 on the latent scale, and <0.001 on the observed data-scale, and the lower tail of the posterior distribution bordered zero on both scales (see Table 2 and supplement for full posteriors). This corresponded to a low proportion of phenotypic variance attributable to maternal effects (ICCME), with no indication that ICCMEwas statistically different from zero (Table 2). Given the low sample size, it may be that we lacked the statistical power to separate VA from VME within a single model. However, when running a model without estimating VA(i.e., without the relatedness matrix), we found that VME did not differ (qualitatively) to that estimated when fitting both simultaneously (see supplement). Moreover, the DIC was lowest for the model containing just the relatedness matrix, and highest for the model containing just the maternal effects (DIC: VA + VME = 277.66, VME = 314.20, VA = 140.64). Together, this suggests that maternal effects explained little to none of the phenotypic variance in the probability of individuals testing positive for C. pecorum . Finally, estimates for VA and for h2did not differ qualitatively between the model containing the maternal effect vs without, suggesting that our reported heritability estimates (Table 1) were not inflated by maternal effects that were unaccounted for.