Conservation management strategy impacts inbreeding and mutation load in scimitar-horned oryx

Significance Conservation genetic management is becoming increasingly important for safeguarding and restoring wildlife populations. Understanding how the intensity of intervention influences genomic components of fitness is therefore essential for supporting species viability. We investigate the impact of contrasting management strategies on the genomic landscape of inbreeding and mutation load in captive populations of the scimitar-horned oryx. We reveal how several decades of management have prevented the formation of long runs of homozygosity and masked the expression of deleterious mutations. Our findings highlight the dynamics between inbreeding, mutation load, and population size and have direct implications for future management of threatened species.


Population origins
Scimitar-horned oryx have been kept in captivity since the 1800s yet the majority of known founders originated from a capture operation in Chad in the 1960s (1) Within the EEP, individuals undergo high-intensity genetic management to facilitate demographic stability and minimise inbreeding (2). Under this strategy, individual mean kinship and inbreeding coefficients derived from pedigree data are used to make breeding and transfer decisions. This approach should theoretically retain greater genetic diversity than under random mating (3). Typically, animals are kept in small herds and males are moved between collections to mimic natural dispersal patterns. No movement occurs with institutions outside the EEP. The current census size of the EEP population is around 619 individuals.
Within the USA there is a broader range of genetic management strategies in place. For example, some institutions within the SSP employ similar high-intensity genetic management practices as the EEP. However, this is not possible for all due to a lack of pedigree records and greater numbers of animals present in large enclosures (4). Under these scenarios, lowintensity genetic management takes place, where herds are assembled from multiple sources and bulls are regularly rotated among them. The current census size of the SSP population is around 223 individuals. In the USA, there are also substantial collections of scimitar-horned oryx held on privately-owned ranches (several thousand individuals). The management spectrum of ranch populations varies from low-intensity genetic management to collections of completely unmanaged breeding herds. Recently, increasing emphasis has been placed on metapopulation management within these private collections through the Source Population Alliance under the Conservation Centers for Species Survival initiative (5).
The Environment Agency -Abu Dhabi (EAD) also houses a population of several thousand scimitar-horned oryx in the United Arab Emirates. Most these animals were sourced from the late Sheikh Zayed bin Sultan Nahyan's private collection on Sir Bani Yas island (EAD A) which comprised thousands of individuals. An additional source came from a small population of around 70 animals that was identified in the UAE around 2014 (EAD B). Sparse historical records mean that founder numbers and origins for these source populations are unknown.
However, prior to moving under the EAD's management purview, both were completely unmanaged and therefore provide a unique comparison group against managed EEP and USA populations.

Sample selection
Samples from the EEP population (n = 8) originated from six European EAZA institutions employing high-intensity genetic management. Samples from the USA (n = 17) originated from both private ranches (USA Ranch: n = 10) and three AZA institutions employing a combination of high and low-intensity genetic management (USA AZA: n = 7). Samples were selected to be genetically representative of the founding lineages within the EEP and USA populations.
This was carried out on the basis of mtDNA control region haplotypes and was a necessary precaution as the genetically managed collections cannot be considered randomly mating populations. As we did not set out to carry out a systematic comparison of management strategies in the USA, we treated the USA Ranch and USA SSP samples as one population (USA: n = 17) on the basis of the following: The ranch populations in which the samples originate are unknown, however anecdotal evidence suggest they are comprised of excess SSP animals and that, at a minimum, low-intensity genetic management is employed through regular bull rotation.
(ii) Analysis of population structure clustered USA SSP and USA Ranch individuals together with respect to the other sampled populations ( Figure S1-3).
(iii) We identified one full-sibling pair in our dataset whereby one individual originated from a private ranch and the other from an AZA institution, corroborating the above.
(iv) We observe no differences in the main results of our manuscript between USA Ranch and USA SSP animals ( Figures S14-15). Systematic sampling across a broader range of ranch types would provide greater power to investigate this in further detail.
In the EAD, samples were selected using origin information i.e. whether they originated from EAD source A or EAD source B. Due to the absence of management in these populations, we considered them to be representative of panmictic populations and therefore sampled individuals opportunistically. As the two source populations had been moved together prior to when sampling took place, we evaluated individual admixture proportions to ensure our samples were truly representative of the source populations (see Main Text for Methods). Two individuals were identified with intermediate ancestry between EAD A and EAD B and were removed from subsequent analysis ( Figure S2).

Relatedness
To estimate relatedness among individuals, we first pruned our SNP dataset for linkage disequilibrium using the --indep function in PLINK, a sliding window of 50 SNPs, a step size of 5 and a variance inflation factor threshold of 2. We also removed SNPs that deviated significantly from HWE with a p-value threshold of 0.001 and with a minor allele frequency < 0.3. We then estimated KING, R0 and R1 coefficients (6) using NgsRelate v2 (7). These

Effective population size
In addition to estimating effective population size (Ne) based on FROH, we also employed a linkage disequilibrium (LD) based approach for comparison. This is because under intensive inbreeding in recent generations, a large fraction of the genome is made up of long ROH with short coalescent times. As a result, there is less genomic territory for reliably estimating Ne in deeper history, as demonstrated in Figure 2C. Furthermore, the performance and biases of IBD-based methods for estimating Ne are generally not well understood. LD-based approaches such as GONe (8), use the observed patterns of linkage disequilibrium across chromosomes to estimate Ne in the recent past. The approach can be applied to contemporary samples of fewer than 10 individuals making it particularly appropriate for our dataset.
The recommended number of SNPs per chromosome for analysis with GONe is between 50,000 and 100,000. We therefore randomly subsampled 2,000,000 filtered loci from across all 28 autosomes to achieve approximately ~70,000 SNPs per autosome. We then split the across much of recent history. These patterns are highly comparable to those observed using the IBD-based approach. However, as expected, Ne estimates in deeper history (>32 generations ago) were markedly larger using the LD-based method, at around 10-30K individuals. These values are likely to reflect more reliable estimates of historical Ne. than those inferred using patterns of IBD. Indeed, the long-term Ne estimate calculated for scimitarhorned oryx in (9) is around 22K individuals. Overall, these findings highlight the value of employing multiple methods for estimating Ne, particularly in populations with close inbreeding due to recent ancestry.  Table S2. Classifications used to assign missense, loss of function and intergenic categories to variants annotated by SNPeff and VEP.