Genetic mixing in conservation translocations increases diversity of a keystone threatened species, Bettongia lesueur

Abstract

Translocation programmes are increasingly being informed by genetic data to monitor and enhance conservation outcomes for both natural and established populations. These data provide a window into contemporary patterns of genetic diversity, structure and relatedness that can guide managers in how to best source animals for their translocation programmes. The inclusion of historical samples, where possible, strengthens monitoring by allowing assessment of changes in genetic diversity over time and by providing a benchmark for future improvements in diversity via management practices. Here, we used reduced representation sequencing (ddRADseq) data to report on the current genetic health of three remnant and seven translocated boodie (Bettongia lesueur) populations, now extinct on the Australian mainland. In addition, we used exon capture data from seven historical mainland specimens and a subset of contemporary samples to compare pre‐decline and current diversity. Both data sets showed the significant impact of population founder source (whether multiple or single) on the genetic diversity of translocated populations. Populations founded by animals from multiple sources showed significantly higher genetic diversity than the natural remnant and single‐source translocation populations, and we show that by mixing the most divergent populations, exon capture heterozygosity was restored to levels close to that observed in pre‐decline mainland samples. Relatedness estimates were surprisingly low across all contemporary populations and there was limited evidence of inbreeding. Our results show that a strategy of genetic mixing has led to successful conservation outcomes for the species in terms of increasing genetic diversity and provides strong rationale for mixing as a management strategy.

1. INTRODUCTION

Australia has an unenviable record of mammal extinction with 35% of the world's modern mammal extinctions (29 species) having occurred on the continent since European settlement (Short & Smith, 1994; Woinarski et al., 2015). Many species are continuing to decline, with 21% of terrestrial mammals considered threatened (Woinarski et al., 2015). The causes of these declines are often complex, but predation by introduced feral cats (Felis catus) and European red foxes (Vulpes vulpes), as well as changing fire regimes, are heavily implicated, with many extinctions occurring in sparsely populated, arid regions of the continent (Doherty et al., 2015, 2016; Short & Smith, 1994; Woinarski et al., 2015). Many formerly widespread Australian mammal species now only occur as natural populations on offshore islands (Woinarski et al., 2015), and conservation translocation programmes (IUCN/SSC, 2013) have increasingly been used to secure the survival of these threatened species. Conservation managers in Australia, and particularly Western Australia, have been proactive in undertaking conservation translocations (Short, 2009), with some schemes in operation for more than 30 years (Morris et al., 2015; Richards, 2012). Across Australia, predator‐free mainland enclosures and island safe havens contribute substantially to the conservation of a suite of native mammal species now otherwise extinct on the mainland—including boodies (Bettongia lesueur), greater stick nest rats (Leporillus conditor), mala (Lagorchestes hirsutus) and the Shark Bay (formerly ‘western barred’) bandicoot (Perameles bougainville) (Legge et al., 2018; Ringma et al., 2018). Translocated populations provide insurance against localised extinction, facilitate increases in population size and provide opportunities to manage gene flow among now permanently isolated populations via supplementation of managed animals (Ottewell et al., 2014; Weeks et al., 2011).

Genetic data are increasingly being used to monitor threatened mammal populations and to aid in decision‐making for effective translocations (Capel et al., 2022; Seaborn et al., 2021). Considered an essential biodiversity variable (EBV; Hoban et al., 2022), genetic diversity is a key metric in conservation monitoring, given its role in conferring species and ecosystems with resilience and adaptive potential (Gustafson et al., 2017; Hoban et al., 2022; Weeks et al., 2011; Willi et al., 2022).

Many translocated populations are established from single population sources, often remnant island populations and may also involve serial translocating (i.e. sourcing founders from populations established from translocations themselves; Lambert et al., 2005). This places these populations at risk of decreased genetic diversity and increased inbreeding depression relative to source populations as a result of founder effects and ongoing small population size and/or isolation (Cardoso et al., 2009; Frankham, 2007; Hedrick et al., 2001; Ottewell et al., 2014). For example, founder effects have been observed in the greater stick nest rat, where translocated populations founded by as few as six, but up to 153 animals, were less diverse and showed higher levels of pairwise relatedness than their source populations on Australian islands (White et al., 2020). In contrast, single‐source translocated populations of the Seychelles warbler (Acrocephalus sechellensis) showed a limited loss of genetic diversity in both microsatellites and the MHC class 1 region (Wright et al., 2014), and the saddleback bird (Philesturnus carunculatus rufusater) in New Zealand showed little loss in diversity over time despite sequential translocations (Taylor & Jamieson, 2008), although both species were noted to have low endogenous genetic diversity. An alternative to sourcing translocates from a single population is to source from multiple populations. Distinct from performing ‘genetic rescue’ of a species to alleviate inbreeding depression, multiple‐source translocations may result in ‘evolutionary rescue’ through increased genetic diversity relative to source populations (Hoffmann et al., 2021; White et al., 2018), whereby broader demographic effects of increased genetic diversity such as adaptive potential are realised (Gonzalez et al., 2013). This may lead to several beneficial outcomes including increased effective population size and enhanced fitness (Ralls et al., 2020; Zecherle et al., 2021). In some instances though, the use of mixing in translocation programmes may be avoided if there is a perceived risk of outbreeding depression or ‘genetic swamping’, whereby locally adapted alleles are lost via a rapid increase of an introduced genotype (Frankham, 2015; Hufford & Mazer, 2003). The decision of whether to keep separate or mix populations that are believed to display substantial phenotypic or genotypic differentiation remains a key consideration of any translocation strategy.

Long‐term genetic monitoring provides opportunities to track micro‐evolutionary changes in translocated populations over time and is particularly important given mammal species may take several years post‐translocation before showing declines in genetic diversity or an increase in relatedness and inbreeding (Duncan et al., 2020; Rick et al., 2019). In addition, in the absence of a limiting factor such as predators, many mammals experience irruptive population dynamics with rapid increases in population size—sometimes to the point where animals require removal from havens (Bannister et al., 2016; Treloar et al., 2021)—followed by dramatic declines, imposing demographic bottlenecks that can act to exacerbate the effects of genetic drift. These declines may occur outside of any initial monitoring period following translocation (Duncan et al., 2020). Understanding the baseline patterns of genetic diversity in both source and translocated populations and continued monitoring post‐translocation is therefore vital to revealing the genetic consequences of translocation and to inform the ongoing management of these species.

Although long‐term monitoring of populations is an important step towards improved genetic management, an understanding of historical diversity can establish genetic baselines, or targets for genetic restoration (Hofman et al., 2015). Accessing historical DNA, for example by using museum skin specimens, has provided insight into the genetic diversity sustained by larger populations prior to their decline and/or extinction (Bi et al., 2019; Byrne et al., 2021; Roycroft et al., 2021). A recent study showed no evidence of reduced genetic diversity prior to the extinction of a suite of Australian rodents, suggesting declines in population size were rapid (Roycroft et al., 2021). Where possible, inclusion of time‐stamped specimens into genetic studies provides a better understanding of historical levels of diversity present in the species, and provides a benchmark for where recovery programmes may target the genetic diversity of species into the future (Hofman et al., 2015).

Boodies (Bettongia lesueur (Quoy & Gaimard, 1824), Potoroidae), or burrowing bettongs, have one of the longest translocation histories of any marsupial (Thomas et al., 2003). Once widespread and abundant, the species became extinct on the Australian mainland in the latter half of the 20th century, leaving only three remnant island populations off the western Australian coast. For nearly three decades, boodies have been managed via conservation translocations to both predator‐free offshore islands and fenced mainland havens (Figure 1), providing an outstanding opportunity to examine the genetic consequences of this management strategy. Specifically, we wanted to investigate whether translocated populations showed reduced genetic variation in comparison to their source/s; to determine whether using multiple sources increased genetic diversity in translocated populations, and via the inclusion of time‐stamped historical specimens, quantify the loss of diversity following extinction on the mainland. Using thousands of single nucleotide polymorphisms (SNPs) derived from double‐digest restriction‐associated DNA (ddRADseq) data, we assessed the current levels of genetic diversity, structure and relatedness present in all populations to determine the impact that founder source (whether single or multiple) has had on these patterns. We then investigated the patterns of exon sequence diversity and structure across seven contemporary populations (three natural and four translocated) as well as seven historical specimens from now‐extinct mainland populations (1896–1964). Mainland specimens were anticipated to show higher levels of exon diversity in line with theoretical expectations of lowered genetic diversity following isolation of populations on islands (Frankham, 1997). We use our findings to provide recommendations regarding the use of admixture approaches in future management of remnant and translocated mammal populations.

FIGURE 1.

Map of boodie populations analysed in this study including flow chart of translocation history and a heatmap of pairwise population boodie F _ST values based on 8076 ddRADseq SNPs from data set A (all populations). ALPH = Alpha Island, ARID = Arid Recovery, BARR = Barrow Island, BERN = Bernier Island, BOOD = Boodie Island, DORR = Dorre Island, DRYA = Dryandra Breeding Facility (now closed), FAUR = Faure Island, H.P. = Heirisson Prong (now closed), MATU = Matuwa Indigenous Protected Area, SCOT = Scotia Sanctuary, YOOK = Yookamurra Sanctuary. Map adapted from Rick et al. (2019). Photo credit: Talitha Moyle: B. lesueur at Scotia Reserve (AWC). Red dashed line denotes historical range, †denotes historical sample locations.

2. MATERIALS AND METHODS

2.1. Study species

Boodies are omnivorous potoroid marsupials that feed on fungi, vegetation and a range of insects, arthropods and carrion (Bice & Moseby, 2008; Chapman et al., 2015; Palmer et al., 2021). They breed throughout the year and females typically rear two young per year (Short & Turner, 1999). The average generation length is estimated as three years (Pacifici et al., 2013). Important ecosystem engineers, they create small diggings during their search for food (Burbidge et al., 2008), aid in seed dispersal (Palmer et al., 2021) and build warrens (Palmer et al., 2021; Sander et al., 1997). Boodies were once widespread across the Australian mainland, occurring west of the Great Dividing Range and south of the tropical savanna in the north of the continent (Figure 1). They were often noted to be the most abundant resident mammal (Burbidge et al., 2008). Decline in boodie numbers is primarily attributed to predation by feral cats and foxes (Short & Turner, 2000), and by the mid‐1900s, mainland populations were restricted to Western Australia before these too became extinct (Burbidge & Short, 2008). Only three natural populations now remain, all on Western Australian islands: Bernier and Dorre Islands of Shark Bay (‘Gatharragudu’ in the local Malgana language) and Barrow Island off the Pilbara coastline. Morphological differences between the Barrow Island and Shark Bay Island animals have prompted provisional classification into two subspecies, with Shark Bay Island boodies (B. lesueur subspecies lesueur) roughly double the weight of Barrow Island boodies (B. lesueur undescribed subspecies Barrow and Boodie Islands; hereafter subspecies Barrow Is). In 2010, both subspecies were introduced to a fenced enclosure at Matuwa and were found to successfully interbreed (Rick et al., 2019; Thavornkanlapachai et al., 2019). In addition to Matuwa, mixed‐source populations from subspecies lesueur (Bernier and Dorre Islands) have been established at Arid Recovery Reserve and Yookamurra Sanctuary in South Australia, as well as at Scotia Sanctuary in New South Wales (Table 1, Figure 1). In addition to the mixed‐source populations, single‐source translocations are located at Alpha and Boodie Islands (Barrow Island source), Dryandra (no longer extant) and Faure Island (Dorre Island source). Reintroduced populations were sampled 5–20 years after establishment and varied in founder number (Table 1). Boodies are currently listed as Near Threatened under IUCN criteria, Vulnerable under Australian Commonwealth legislation (EPBC Act) and as Conservation Dependent under Western Australian legislation.

TABLE 1.

Details and translocation history of the 11 Bettongia lesueur populations (natural and translocated (trans.)) sampled and year they were established, including total number of founders, census population size in 2019, number of individuals successfully genotyped using ddRADseq (sample size) and their inclusion in alternative data sets (Data set A, B or C) for genetic analyses.

Pop	Location	Type	Year sampled (established)	Subspecies	Source	No. founders	Census No. 2019	Sample size	Data set A	Data set B	Data set C
BARR	Barrow Island, W.A.	natural	2010	Barrow Is	–	–	3000	22	✓		✓
BERN	Bernier Island, W.A.	natural	2001/2016	lesueur	–	–	700	19	✓	✓
DORR	Dorre Island, W.A.	natural	2001/2016	lesueur	–	–	2500	24	✓	✓
ALPH	Alpha Island, W.A.	trans.	2016 (2011)	Barrow Is	Barrow Is.	40	114	20	✓		✓
ARID	Roxy Downs Arid Recovery, S.A.	trans.	2014 (1999–2000)	lesueur	Bernier/Heirisson Prong (Dorre Is.)	30	71 a	21	✓ b	✓ b
BOOD	Boodie Island, W.A.	trans.	2012 (1993)	Barrow Is	Barrow Is.	36	200	18	✓		✓
DRYA	Dryandra Reserve, W.A.	trans.	2010 (2003–2007)	lesueur	Heirisson Prong (Dorre Is.)	82	Closed	24	✓	✓
FAUR	Faure Island, W.A.	trans.	2018 (2002)	lesueur	Heirisson Prong (Dorre Is.)	17	15,000	20	✓	✓
MATU	Matuwa Indigenous Protected Area, W.A.	trans.	2015–2018 (2010–2012)	lesueur/ Barrow Is	Dryandra/Barrow Is.	201	250	22	✓
SCOT	Scotia Sanctuary, N.S.W.	trans.	2017–2018 (1997–2001)	lesueur	Yookamurra	19	477	19	✓	✓
YOOK	Yookamurra Sanctuary, S.A.	trans.	2017–2018 (1996)	lesueur	Bernier Is. via Earth Sanctuaries	100	160	18	✓	✓
								SNPs prefilt (ind)	54,584 (242)	41,587 (154)	27,583 (63)
								SNPs post (ind)	8076 (212)	6406 (131)	9682 (56)

^a At time of sampling ARID census size was approx. 6000. Note that the Dryandra (DRYA) haven was closed prior to 2019.

^b ARID population was not included in standardised genetic diversity analyses due to high degree of missing data.

2.2. Sampling and genomic DNA preparation

Ear notch tissue samples were opportunistically obtained during routine monitoring and scientific studies of boodie populations across Australia. A total of 289 samples were obtained, and the year of sampling varied across populations (Table S1). In addition to the 10 populations sampled here, we also included publicly available ddRADseq data on boodies sampled during 2014 from the Arid Recovery translocation programme in South Australia (ENA, PRJNA389944; White et al., 2018). For the exon capture study, we subsampled tissue from historical museum skin specimens collected from mainland Australia (n = 7). Historical specimens were collected between 1896 and 1928, with one naturally mummified specimen collected in 1964 likely deceased much earlier (Table S2).

For contemporary samples, genomic DNA was extracted from ear biopsy samples (1–2 mm) using a modified salting out method (Sunnucks & Hales, 1996) with the addition of 3 μL 10 mg/mL RNase to the TNES buffer to remove RNA contamination. For the historical museum specimens, DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen), with modifications based on Joseph et al. (2016) (see Appendix S1 for detail).

2.3. ddRADseq library preparation and read processing

For each sample, DNA (approx. 50 ng) was sent to the Australian Genome Research Facility (AGRF, Melbourne) for ddRADseq library preparation and sequencing (total 289 samples and 22 replicates, Table S1). Samples were digested using the pstI and mspI restriction enzymes, purified and size selected to 280–375 bp using Blue Pippen (Sage Science). Libraries were amplified using indexed primers and sequenced (150 bp) on the AGRF Illumina NextSeq500. Raw sequence quality was screened using FastQC (Andrews, 2010) before demultiplexing using the process_radtags module in Stacks v. 2.4, removing reads with any uncalled bases, discarding reads with low quality scores and trimming to a final length of 125 bp using the ‐inline_index option for barcodes (Catchen et al., 2013). Demultiplexed reads were then aligned against a draft version of the closely related woylie (Bettongia penicillata) genome (10X assembly) (Peel et al., 2021) using BWA mem v. 0.7.17 (Li & Durbin, 2009). BAM files were filtered for secondary alignments and sorted using Samtools v. 1.6 (Li et al., 2009) and SNPs were called in the Stacks ref_map.pl pipeline using default settings. Filtering of SNPs within the Stacks Populations module was dependent upon the downstream analyses performed (see below). The gl.filter.sex‐linked function in dartR (Gruber et al., 2018) was used to confirm the absence of sex‐linked loci from the SNP data set. Due to the substantial genetic divergence between Barrow and Shark Bay Island populations that can impact SNP selection and diversity estimates (Schmidt et al., 2021), we partitioned our data into three data sets to enable comparative analyses across these highly divergent populations. Data set A (n = 227) included all 10 populations (both subspecies) sampled as part of this study as well as the Arid Recovery (ARID) population data. ARID data were generated using the same restriction enzymes as our samples, however, different library preparation and sequencing methods meant that only a proportion of the data overlapped with ours, and for this reason, ARID samples were only included in genetic structure analyses. For comparative analyses, the Matuwa (MATU) population was downsampled from n = 65 to n = 22 to avoid introducing bias associated with different sample sizes (Schmidt et al., 2021). Data set B (n = 145) included all seven populations comprised of subspecies lesueur individuals including single source and mixed Bernier and Dorre Island translocations but not including the ‘hybrid’ Matuwa population. Data set C (n = 60) consisted of the source population of the Barrow Island subspecies and two single‐source translocations from this (Table 1).

2.4. Comparative data analyses (ddRADseq)

2.4.1. Population structure

SNP filtering for ddRADseq data for population structure and admixture analyses required different filters to be applied in the Stacks Populations module to accommodate features of each data set, as detailed in Appendix S1, with consistent filters on final read depth (minDP = 8, maxDP = 200), reproducibility (>0.9) and minor allele frequency (MAF, 0.02) applied to all data sets. In all cases, repeat libraries were removed post‐filtering with the best‐performing replicate (least missing data) retained for analysis (Gosselin et al., 2020). Pairwise population F _ST and significance estimates were produced using the StAMMP package (Pembleton et al., 2013) in R version 4.0.5 (R Core Team, 2021) and visualised as a heatmap using pheatmap (Kolde, 2018). Genetic clustering was visualised with a principal coordinates analysis (PCoA) in the R package Adegenet (Jombart & Ahmed, 2011) using the dudi.pco function and plotting with ggplot2 (Wickham, 2016). We utilised the Bayesian clustering algorithm available in fastSTRUCTURE (Raj et al., 2014) to assess population genetic structure using the simple prior and performing 10 independent runs for each K value (K = 2–12 for data set (A), K = 2–8 for data set (B), K = 2–4 for data set (C)). The most likely number of K clusters that minimised the marginal likelihood was determined using the chooseK.py script available within the fastSTRUCTURE program.

2.4.2. Standardised population genetic diversity

To calculate genome‐wide diversity metrics, we ran the Stacks ref_map.pl across the three data sets and removed any sites with missing data (R = 1). We did not filter on short distance linkage disequilibrium (LD) or minor allele frequency (MAF). The VCF file was filtered on minimum mean depth (‐‐min‐meanDP = 8) and maximum mean depth (‐‐max‐meanDP = 300) in VCFtools (Danecek et al., 2011) to obtain a SNP blacklist which was then re‐passed through populations to exclude loci failing these criteria. The number of private alleles per population, percentage polymorphic loci, nucleotide diversity (Pi) and mean observed and expected heterozygosity were taken from the populations.sumstats_summary.tsv output file produced in Stacks. We chose to report a standardised genome‐wide heterozygosity rather than traditional SNP heterozygosity as the latter is heavily influenced by sample size, MAF filtering, missing data thresholds and population structure (Schmidt et al., 2021). In preliminary univariate analyses, we detected a significant correlation of founder group size and two genetic diversity parameters (observed heterozygosity and proportion polymorphic loci, Figure S1), though these were non‐significant when rerun in multivariate models testing founder size and source (multiple vs. single) population (Table S3). Furthermore, we found no effect of time since establishment on genetic diversity parameters. Subsequently, we tested for a significant difference in mean values of diversity metrics among single (including natural)‐ and multiple‐source populations using a one‐tailed t‐test after confirming normality with a Kolmogorov–Smirnov test.

2.5. Individual population‐based analyses

2.5.1. Population relatedness and kinship estimates

To assess relatedness within populations, we generated population‐specific data sets. SNPs were called for each population individually using the ref_map.pl pipeline and loci filtered on depth (min DP = 8, max DP = 200) with additional filters as detailed in Appendix S1. We calculated the internal relatedness (IR) statistic (Amos et al., 2001) in GENHET (Coulon, 2010) to represent a relative outbred–inbred continuum in which positive values are indicative of inbreeding and negative values of outbreeding. For BERN and DORR populations, where samples were taken from two distant time points, 2001 and 2016, we also calculated relatedness (RelW) and IR within temporal cohorts to remove any bias associated with non‐overlapping generations. We inferred familial relationships between pairs of individuals using kinship coefficients calculated with the KING programme (Manichaikul et al., 2010). The kinship coefficient represents the probability that the allele selected at random in two individuals is identical by descent (Manichaikul et al., 2010).

2.5.2. Effective population size (N _e)

The individual population data sets were also used to estimate contemporary N _e using the Linkage Disequilibrium method in the program NE Estimator V2 (MAF = 0.02 and 0.05) (Do et al., 2014). As with relatedness, we also calculated N _e on temporal cohorts within BERN and DORR. We also removed first‐order relatives as determined in the kinship analysis.

2.6. Exon capture

A custom sequence capture approach was used to target exons, using a set of marsupial probes that deliberately excluded Y chromosome exons (SeqCap EZ Developer Library; Roche NimbleGen) as described in Bragg et al. (2017). Some samples (n = 17) overlapped with the ddRADseq data set, five from YOOK, six from SCOT and six from MATU (Table S2). The remaining samples were sourced from the same cohorts as represented in the ddRADseq data set but from different individuals. Genomic library preparation followed the Meyer and Kircher (2010) method with modifications as described in Duchene et al. (2018). Libraries were dual indexed with a combination of one of 24 unique p5 index sequences and one of 96 unique p7 index sequences. Pooled exon capture libraries were sequenced on an Illumina NovaSeq SP 200 cycle paired end run at the ACRF Biomolecular Resource Facility, Australian National University. Raw sequence data were cleaned, assembled, mapped and aligned using the pipeline ECPP (https://github.com/Victaphanta/ECPP) with settings as described in Roycroft et al. (2020), and with a minimum coverage of 10X to call variants. Historical skins were mapped to the highest quality de novo assembly using the approach described in Roycroft et al. (2022). Final alignments for all data were filtered at a threshold of 3% maximum heterozygosity per locus and a minimum of 80% data completeness, and processed with BMGE (Criscuolo & Gribaldo, 2010) to remove poorly represented regions. Individual heterozygosity was calculated for each sample in the exon capture data set using the approach of Roycroft et al. (2021). To visualise the variance in genetic data, we selected one random SNP per exon to perform PCoA analyses (function gl.pcoa in the R package dartR; Gruber et al., 2018) on both extinct and living populations.

3. RESULTS

3.1. ddRADseq raw data

Sequencing generated a mean of 3.1 (SD = 1.7) million reads (approx. 150 bp) for each novel sample, of which a mean of 3 million reads (98.68% ± 0.37 SD) mapped to the Woylie genome. The ENA Arid Recovery samples (n = 28) generated a mean of 2.4 million reads of which on average 2 million reads (75 bp) per sample mapped to the Woylie genome (85.23% ± 8.32 SD) (Table S1). Poor quality samples, considered to be those with less than 400,000 reads (n = 7), were excluded from downstream analyses (Table S1).

3.2. Comparative data analyses

3.2.1. Population structure

Pairwise population differentiation was significant among all sampled populations using the overall data set of 8076 SNPs (data set A; global F _ST = 0.38, all comparisons p < .05) (Table S4). Each of the island source populations (BARR, BERN and DORR; acronyms in Table 1) displayed very high genetic differentiation, with the highest values observed between the most distantly located BARR and BERN/DORR populations (F _ST = 0.62–0.66). Despite BERN and DORR islands only being separated by a 500‐m wide channel, these populations also were highly significantly differentiated (F _ST = 0.54 in the overall data set and F _ST = 0.56 in the Shark Bay‐specific data set, Table S4). Island source populations formed distinct genetic clusters in PCoA and FASTSTRUCTURE analysis (K = 3, Figure 2).

FIGURE 2.

(a) Principal coordinates analysis (PCoA) of boodies based on 8076 ddRADseq SNPs from data set A (all populations). Axes show the amount of variance explained by each principal component. Remnant source populations are indicated as triangles and translocated populations with open circles. (b) FASTSTRUCTURE plot (K = 3) of boodies based on 8076 SNPs from data set A (all populations) showing the impact of admixture of different source populations during the translocation process. (c) FASTSTRUCTURE plot (K = 4) of boodies based on 6406 SNPs from data set B (subspecies lesueur populations). (d) FASTSTRUCTURE plot (K = 2) of boodies based on 9682 SNPs from data set C (subspecies Barrow Is). Population abbreviations include: ALPH = Alpha Island, ARID = Arid Recovery, BARR = Barrow Island, BERN  = Bernier Island, BOOD = Boodie Island, DORR = Dorre Island, DRYA = Dryandra Reserve, FAUR = Faure Island, MATU = Matuwa Indigenous Protected Area, SCOT = Scotia Sanctuary, YOOK = Yookamurra Sanctuary.

Single‐source translocated populations consistently showed lower genetic differentiation from their source populations in comparison to mixed source populations. For example, ALPH and BOOD cluster tightly with source, BARR (Barrow Island), in PCoA and FASTSTRUCTURE analyses and exhibit low differentiation with mean F _ST = 0.04. Similarly, FAUR and DRYA cluster tightly with the source, DORR (Dorre Island) (F _ST = 0.04–0.06) (Table S4). For mixed source translocations, F _ST values relative to source populations ranged from 0.157 to 0.314.

Genetic analysis also revealed very clear admixture patterns in the translocated populations derived from mixed sources. Within the overall data set A, PCoA clearly displays the intergrade of subspecies lesueur individuals with subspecies Barrow Is individuals in the admixed MATU population along axis1 and of BERN & DORR individuals in admixed ARID, SCOT and YOOK populations on axis2 (Figure 2), with similar admixture visualised in FASTSTRUCTURE analyses (Figure 2). Despite the difference in sequencing method used for the ARID population, admixture patterns were consistent with its translocation history; however, its placement in the PCoA relative to other populations sequenced in this study may have been impacted by a higher proportion of missing data (see Appendix S1 for detail).

Hierarchical analysis using subspecies‐specific SNP data sets revealed further detail of the relationships among translocated populations, particularly of the Shark Bay‐derived populations. In data set B with 6406 SNPs (subspecies lesueur), FASTSTRUCTURE analysis identified K = 4 genetic clusters (Figure 2), with older translocated populations DRYA and SCOT/YOOK (noting that SCOT was originally founded from YOOK animals, Figure 1) identified as distinct clusters in addition to source populations (Figure S2). Similarly, in data set C with 9682 SNPs (subspecies Barrow Is.), the older BOOD translocation was identified as distinct from its original source population (K = 2) (Figure 2, Figure S2). Such additional fine‐grained genetic structuring among translocated populations likely reflects their independent evolutionary trajectories, given all populations are completely isolated post‐translocation.

3.2.2. Genome‐wide diversity metrics

After filtering, data set A included 11,733 variable sites out of a total of 4,241,264 callable sites (Table 2). Measures of standardised genetic diversity (data set A) varied significantly among populations, with those founded from multiple sources consistently showing significantly higher estimates of percentage of polymorphic loci (t(8) = −3.43, p = .004; mean PPL = 0.15 vs. 0.10, respectively), nucleotide diversity (t(8) = −5.5, p = .0003; mean Pi = 0.18 vs. 0.12) and both observed (t(8) = −5.3, p = .0003; mean Ho = 0.0005 vs. 0.0003) and expected (t(8) = −5.3, p = .0004; mean He = 0.0005 vs. 0.0003) heterozygosity than single‐source and natural populations (Figure 3, Table 2). Across all populations, the single‐source translocated population at DRYA showed the lowest genetic diversity and the multiple‐source population of MATU (Matuwa‐mixed subspecies) the highest (Table 2). Out of the three natural island populations, BERN showed lowest genetic diversity at all metrics (Table 2, Figure S3). The number of private alleles was highest in BERN (n = 539) and lowest in YOOK (n = 12). Overall, private alleles were higher (but not significantly so) in the natural island populations (mean = 255, SE = 117.1), in comparison to the single‐source translocation (mean = 73.25, SE = 19.9) and multiple‐source translocation populations (mean = 30.3, SE = 12.6) (Table 2).

TABLE 2.

Standardised ddRADseq genome‐wide genetic diversity statistics for boodie populations in data set A based on 11,733 variant sites of 4,241,264 total sites (variant and invariant).

Pop	Source	n	Ap	% Poly Loci	Pi	Std. Obs Het	Std. Exp Het
BARR	Natural	22	78	0.1141	0.1279	0.00034 (SE 0.00001)	0.00035 (SE 0.00001)
BERN	Natural	19	539	0.0992	0.1242	0.00034 (0.00001)	0.00033 (0.00001)
DORR	Natural	24	148	0.1107	0.1251	0.00035 (0.00001)	0.00034 (0.00001)
Mean	Natural	21.6	255 (SE 117.1)	0.108 (SE 0.004)	0.126 (SE 0.001)	0.00034 (0.00000)	0.00034 (0.00000)
ALPH	Single	20	100	0.1104	0.1294	0.00036 (0.00001)	0.00035 (0.00001)
BOOD	Single	18	122	0.1077	0.1246	0.00035 (0.00001)	0.00034 (0.00001)
DRYA	Single	24	22	0.0928	0.1107	0.00031 (0.00001)	0.00030 (0.00001)
FAUR	Single	20	49	0.0974	0.1155	0.00032 (0.00001)	0.00031 (0.00001)
Mean	Single	20.5	73.25 (19.9)	0.1002 (0.004)	0.1201 (0.004)	0.00034 (0.00001)	0.00033 (0.00001)
Mean	Nat/single	21	151.1 (61.7)	0.1046 (0.003)	0.1225 (0.002)	0.00034 (0.00000)	0.00033 (0.00000)
MATU	Multiple	22	61	0.1878	0.2094	0.00060 (0.00001)	0.00057 (0.00001)
SCOT	Multiple	19	18	0.1344	0.1666	0.00046 (0.00001)	0.00045 (0.00001)
YOOK	Multiple	18	12	0.1236	0.1595	0.00045 (0.00001)	0.00043 (0.00001)
Mean	Multiple	19.7	30.3 (12.6)	0.1486 (0.016)	0.1785 (0.013)	0.00050 (0.00000)	0.00048 (0.00000)

Note: Mean values in bold and standard errors for all estimates in parentheses. Source, whether translocated population is sourced from single or multiple populations; n, number of individuals; A_p, number of private alleles; % Poly Loci, percentage polymorphic loci; Pi, nucleotide diversity; Std. Obs/Exp Het, standardised genome‐wide observed and expected heterozygosity.

FIGURE 3.

Box plots showing the standardised diversity metrics of B. lesueur calculated using data set A ddRADseq data (all populations). Metrics include the proportion of polymorphic loci, nucleotide diversity and standardised observed and expected heterozygosity on single‐sourced (including natural) populations versus multiple‐source translocated populations.

After filtering, data set B included 12,861 variable sites out of a total of 4,638,441 sites (Table S5). Data set B, encompassing all subspecies lesueur populations also showed that multiple‐source translocated populations were significantly higher than single and natural source populations in terms of percentage of polymorphic loci (t(4) = −3.8, p = .01), nucleotide diversity (t(4) = 8.83, p = .0004) and both observed (t(4) = −9.5, p = .0004) and expected heterozygosity (t(4) = −9.73, p = .0003). Again, DRYA showed the lowest levels of diversity across all metrics while the multiple‐source population of SCOT showed the highest genetic diversity. Of the natural island populations, BERN exhibited lower diversity than DORR in all metrics.

After filtering, data set C included 9016 variable sites out of a total 4,795,283 sites (Table S5). Given data set C (subspecies Barrow Is) was comprised of just three populations, we were unable to conduct statistical tests on variation of genetic diversity. The single‐source translocation population at ALPH showed the (marginally) highest estimates of percentage polymorphic loci, Pi and heterozygosity.

3.3. Individual population‐based analyses

SNP data sets were variable in size for individual populations, varying between 8941 SNPs (BARR) and 21,123 SNPs (MATU) post‐filtering (Table 3). Mean Internal Relatedness (IR) estimates of populations were all positive, but not significantly so (Figure S4), indicating low levels of inbreeding and ranged from 0.014 ± 0.013 in YOOK to 0.095 ± 0.022 in MATU. Values were comparable between multiple‐source (mean = 0.04 ± 0.004) and single‐source/natural population values (mean = 0.05 ± 0.019; Table 3). First‐ and/or second‐degree relationships were rare overall, and detected among sampled individuals in all populations, except ARID, FAUR and SCOT, with the highest relative number of related individuals detected in MATU and BARR Island boodies (Table 3, Table S6). N _e estimates were highest in DORR (N _e = 131.6) and lowest in MATU (N _e = 3.9) and DRYA (N _e = 25.2) (Table 3, Table S7). There was no significant difference between single‐source/natural populations (mean = 70.94 ± 14.8) and multiple‐source (mean = 36.3 ± 13.2) population estimates (Table 3, Table S7), even when MATU was removed from comparisons due to recent admixture (multiple source excl. MATU, mean = 52.5 ± 0.8).

TABLE 3.

The number of individuals and loci retained after filtering individual ddRADseq boodie population data sets.

Pop	Source	n	No. loci	No. loci	% Miss. data	N e	IR	1st degree	2nd degree
			Pre‐filter	Post‐filter				(%)	(%)
BARR	Natural	22	24,650	9203	4.5	28.1 (15–85)	0.051 (SE 0.016)	2.33	2.33
BERN	Natural	19	25,444	8941	2.6	51.4 (30–132)	0.035 (0.013)	0	1.73
BERN (2001)		14			4.9	37.6 (20–156)	0.023 (0.018)
BERN (2016)		5			1.8	Inf	−0.08 (0.018)
DORR	Natural	24	27,490	9387	2	131.6 (73–551)	0.026 (0.008)	0.98	0.49
DORR (2001)		13			2.3	102.5 (40–inf)	0.009 (0.011)
DORR (2016)		11			1.6	87.2 (36–inf)	−0.01 (0.01)
ALPH	Single	20	41,012	14,797	1.9	74 (50–137)	0.044 (0.009)	0.52	0.52
BOOD	Single	18	36,982	11,517	2.2	63.7 (33–401)	0.025 (0.010)	1.31	1.31
DRYA	Single	24	30,491	10,945	1.6	25.2 (12–121)	0.016 (0.011)	1.33	1.33
FAUR	Single	20	31,336	12,139	4.1	122.6 (71–403)	0.044 (0.013)	0	0
MATU	Multiple	61	63,341	21,123	3.8	3.9 (3–9) a	0.095 (0.022)	0.87	2.16
SCOT	Multiple	19	41,710	17,727	3.4	51.7 (37–83)	0.045 (0.018)	0	0
YOOK	Multiple	18	40,578	15,084	2.1	53.3 (34–111)	0.014 (0.013)	0	1.31

Note: n, number of individuals; number of loci pre‐filtering; number of loci post‐filtering; percentage missing data (%miss. data), Effective population size based on MAF = 0.05 (N _e) with first‐order relatives removed and confidence intervals in parentheses, Internal Relatedness (IR) (Amos et al., 2001) (standard errors in parentheses), the percentage of individuals identified as first‐ or second‐degree relatives using KING (Manichaikul et al., 2010). See Table S7 for more information on N _e estimates.

^a Given the MATU population involves recent admixture between two very divergent populations, it likely violates the conditions of the LDNE analysis due to its non‐equilibrium state.

3.4. Exon capture

An average of 2,102,157 collapsed de‐duplicated reads were obtained for the historical specimens with a mean of 41% of reads mapping to de novo assemblies (Table S2, Table 4). In comparison, an average of 2,799,648 collapsed de‐duplicated reads were obtained from contemporary specimens with a mean of 79% of reads mapping to target assemblies (Table S2, Table 4). After filtering, individual heterozygosity was calculated on SNPs with greater than 10X coverage (mean filtered SNPs = 1,144,520) (Table S2). Observed heterozygosity was highest in historical specimens (mean MAIN = 0.0011 ± 0.00) and lowest in the contemporary Bernier Island population (mean BERN = 0.0005 ± 0.00). Samples from the admixed Matuwa population showed levels of heterozygosity that were comparable to historical mainland specimens prior to extinction (mean MATU = 0.0010 ± 0.00) (Figure 4, Table 4). The exon capture data show, like the ddRADseq data, that mixing and successful breeding of animals from the different subspecies results in significantly higher levels of observed heterozygosity than mixing animals from different populations of the same subspecies (Figure 4). The PCoA based on 1580 SNPs separated the contemporary populations as per the ddRADseq data and the historical specimens lay central to the pattern of population distribution. Admixed Matuwa samples lay between the Barrow Is subspecies (Barrow Island and Alpha Island) and subspecies lesueur population at Dorre Island as anticipated (Figure 5).

TABLE 4.

Summary of boodie population samples used in exon capture analysis including extinct mainland specimens (MAIN).

Pop.	Subspecies	Source	Sample type	Sample size	Collected	# Raw reads	# Dedup reads	% Dedup target reads	Mean coverage	% Loci captured	Prop. heterozygous
MAIN	graii	Natural a	Historical skin	7	1896–1964	10,183,489	2,102,157	41.0	43.0	92.6	0.0011 (0.00)
BARR	Barrow Is	Natural	Ear clip	12	2010	5,530,522	2,618,846	78.8	102.9	94.4	0.0006 (0.00)
ALPH	Barrow Is	Single	Ear clip	12	2016	5,871,612	2,725,295	78.7	109.2	94.5	0.0006 (0.00)
BERN	lesueur	Natural	Ear clip	11	2001–2018	5,754,047	2,885,777	81.0	120.9	94.5	0.0005 (0.00)
YOOK	lesueur	Multiple	Ear clip	5	2017–2018	5,436,174	3,005,125	74.2	113.6	94.5	0.0007 (0.00)
DORR	lesueur	Natural	Ear clip	3	2017–2018	4,564,257	2,323,617	75.2	90.5	94.4	0.0006 (0.00)
SCOT	lesueur	Multiple	Ear clip	13	2009–2018	7,527,436	4,205,082	79.2	165.0	94.5	0.0007 (0.00)
MATU	lesueur/Barrow Is	Multiple	Ear clip	19	2011–2017	3,670,692	1,970,412	81.8	84.1	94.3	0.0010 (0.00)

Note: Source, whether natural, single‐source translocated or multiple‐source translocated; sample type used in DNA extraction, museum skin specimens or ear clip from extant populations; sample size, the number of samples included in analysis; collected, year the animal was sampled; # raw reads, the mean number of raw reads post sequencing; #dedup reads, the mean number of de‐duplicated reads; % dedup target reads, the mean percentage of deduplicated reads on target; mean coverage across all population samples, % loci capture, the mean percentage of exon capture loci captured; prop. heterozygous, the mean proportion of heterozygous loci identified with standard errors in parentheses.

^a Extinct.

FIGURE 4.

Exon capture (EC) individual heterozygosity grouped according to whether boodie populations were natural or single‐sourced translocation populations (Single source), translocated populations from mixed sources of the same subspecies, B. lesueur ssp. lesueur (Mix Within Ssp), translocated populations from mixed sources of the two different subspecies, B. lesueur ssp. lesueur and B. lesueur ssp. Barrow Island (Mix between Ssp) or historical mainland (extinct) specimens (Mainland).

FIGURE 5.

Principal coordinates analysis (PCoA) of boodie natural and translocated populations based on 1580 SNPs from exon capture data, including historical mainland specimens (MAIN). Remnant source populations are indicated as closed triangles and translocated populations with open circles. The extinct mainland population is represented as closed squares. Axes show the amount of variance explained by each principal component. Population abbreviations include: ALPH = Alpha Island, ARID = Arid Recovery, BARR = Barrow Island, BERN  = Bernier Island, BOOD = Boodie Island, DORR = Dorre Island, DRYA = Dryandra Reserve, FAUR = Faure Island, MATU = Matuwa Indigenous Protected Area, SCOT = Scotia Sanctuary, YOOK = Yookamurra Sanctuary.

4. DISCUSSION

Genetic data can provide insight into the population history and health of threatened species and inform their ongoing conservation. When combined with historical data, there is a unique opportunity to glimpse pre‐decline levels of a species' diversity, that not only illustrate what has been lost in recent population fragmentation and decline, but what we might hope to return to with the aid of breeding and management programmes. Here, we combined ddRADseq and exon capture data to investigate the genetic status of both natural and translocated boodie populations across Australia, a species subject to active conservation measures for the last 30 years. We identified a dramatic reduction in genetic diversity in remnant island populations and observed that mixing divergent contemporary populations, currently assigned to different subspecies, restored heterozygosity to levels approaching that in pre‐decline, historical specimens. The ddRADseq data confirmed a significant increase in a range of genetic diversity parameters in translocated populations where multiple sources had been used. Our results have broad implications for translocation programmes and suggest consideration of alternative approaches to management of taxonomic units, such as subspecies, when working with genetically depauperate mammal populations (Zecherle et al., 2021).

4.1. Genetic diversity and structure of Bernier, Dorre and Barrow Islands

Understanding genetic patterns present in the three remaining natural island populations provides essential context for genetic outcomes observed in translocated populations. The three natural populations were highly differentiated in both data sets, with similar levels of genetic diversity detected. Genetic differentiation (ddRADseq data) between neighbouring Bernier and Dorre Islands was remarkably strong, and consistent with previous mtDNA and microsatellite analysis (Thavornkanlapachai et al., 2019), despite the islands being isolated only relatively recently (3–6 Kya). Island populations typically experience cyclical population bottlenecks (Mills et al., 2004). Shark Bay boodie numbers have been shown to closely track rainfall, albeit with a 2‐year time lag, with large population crashes noted when rainfall falls below 200 mm/year (Chapman et al., 2015; Short et al., 1998). The combination of small island size (approx. 45 km²) and regular population crashes in response to declining climatic conditions are likely to cause strong genetic drift, driving rapid genetic differentiation as observed in our analyses of genetic structure (Frankham, 1997; Kimura, 1983; Mills et al., 2004). In the exon capture PCoA, the central position of the historical mainland animals relative to island populations suggests that the former carry ancestral variation now differentially fixed in the latter. This shows how rapidly populations of mammals can differentiate due to drift processes once isolated and serves as a caution that current taxonomic processes identifying subspecies on changes in allele frequencies risk over‐splitting taxa which may ultimately hinder conservation efforts, particularly if these are kept separate on the basis of preserving genetic differentiation (Weeks et al., 2016).

The historical samples also highlight the loss of genetic diversity in these island populations that now exhibit approximately half of the heterozygosity observed in the seven, pre‐decline mainland specimens. Loss of genetic diversity is expected among island populations but the stark contrast in diversity estimates between historical and contemporary samples emphasises the need, where still possible, to focus conservation efforts on populations of threatened species that remain on the mainland and are likely to be large reservoirs of genetic diversity. Despite being exposed to recent fragmentation effects, Roycroft et al. (2021) determined mainland populations of multiple rodent species had not experienced prolonged genetic drift prior to recent and rapid population crashes sustained following European occupation of Australia. This makes remaining mainland populations of vulnerable mammal species a valuable resource for conservation programmes.

Of the three natural island populations, Bernier Island consistently displayed the lowest levels of genetic variation. Previous studies have detected lower genetic diversity in other mammal populations present on Bernier Island in comparison to Dorre Island, including the banded hare‐wallaby (White et al., 2020), rufous hare‐wallaby (Eldridge et al., 2004, 2019) and the Shark Bay bandicoot (Smith & Hughes, 2008). Bernier Island is almost 20% smaller than Dorre Island, and commonly receives lower rainfall (C. Simms—unpublished data) despite close proximity. In addition, the vegetation on Bernier island has previously been described as more unstable because of grazing pressure from goats that were introduced to the island in 1899 (eradicated 1984) which may have reduced carrying capacity for species reliant on vegetation for their diet or shelter, as those above. Mammal abundance has been noted to be lower on Bernier Island (Chapman et al., 2015), and it is likely that drought impacts are more pronounced on that Island. Rainfall in the region is predicted to decline under future climate change (Watterson et al., 2015) and these challenges pose a significant risk to the future of these island boodie populations.

All three natural populations retained private alleles not represented in the derived, translocated populations suggesting that additional unsampled diversity in these populations has not been captured in the translocation programme or that low frequency alleles might have been lost subsequent to translocation. Despite its low diversity, private alleles were highest in Bernier Island, which concords with the over‐representation of Dorre and Barrow Island sources in translocations. However, given the low to moderate sample sizes in this study (n = 18–24 per population), these observed patterns should be interpreted with caution due to stochastic sampling effects, with comprehensive evaluation required to assess variation and/or loss of site‐specific genetic diversity within the translocation programme.

4.2. Genetic consequences of translocation

4.2.1. Multiple‐source translocation populations have significantly higher observed heterozygosity

As expected, mixing boodies from divergent multiple sources had a clear, positive impact on population‐wide genetic diversity. Populations that had multiple founder sources had significantly higher genetic diversity than those from a single‐source origin, as well as the three remaining natural island populations. Diversity estimates have been made previously for Matuwa (Rick et al., 2019; Thavornkanlapachai et al., 2019) and Arid Recovery populations (White et al., 2018) and these noted an increase in diversity relative to their source populations. Here, we comprehensively evaluated this pattern using ddRADseq data across all extant contemporary populations of boodies, using standardised heterozygosity estimates to remove biases in diversity estimates associated with genetic structure and variation in sample size among populations (Schmidt et al., 2021). We found this pattern, of higher diversity in multiple‐source populations relative to their source populations, to also hold in Scotia and Yookamurra Sanctuary populations, over and above the impacts that founder group size, or time since establishment may have had on genetic diversity. Furthermore, by standardising our diversity estimates, we have been able to quantify the relative change in diversity that is afforded by mixing between subspecies as opposed to mixing within subspecies. Observed heterozygosity estimates were ~30% higher in Matuwa (mixed between subspecies) than the Scotia and Yookamurra Sanctuaries (mixed within subspecies), and ~45% higher than the natural island populations. These patterns were also observed in estimates of heterozygosity using exon capture data, suggesting that gains in diversity are occurring at both neutral and functional genetic loci. These genome‐wide increases in diversity are promising given that founder effects often result in lower genetic diversity in translocated populations (Frankham et al., 2010) and that populations held in fenced reserves or on small islands are subject to genetic drift with time. With no significant decreases in genetic diversity observed even in single‐source translocation populations, founder effects appear to be limited within the boodie translocation programme, perhaps owing to the ability for populations to experience rapid population growth (Nei et al., 1975).

Establishing translocated populations with increased genetic diversity is desirable from a management perspective to increase adaptive potential and heterosis (Frankham et al., 2010; Hoffmann et al., 2021), or to undertake genetic rescue, where practitioners deliberately seek to increase fitness by reducing the impacts of inbreeding depression (Hoffmann et al., 2021). While these goals are frequently cited as the motivation for mixing populations, it is currently laborious and often impractical to quantify the changes in adaptive diversity or population fitness to empirically validate this approach (Holderegger et al., 2006; Teixeira & Huber, 2021). While our exon capture data suggest that mixing populations has resulted in an increase in diversity at functional loci, we cannot determine whether this variation will confer adaptive advantages and translate to improved survivorship and/or fecundity, but it does represent an increase in adaptive potential through standing genetic variation for natural selection to act upon. At present, there is no indication that mixing the two subspecies in the Matuwa population has resulted in outbreeding depression in either the F1 or F2 hybrids (Rick et al., 2019), and the positive gains in diversity warrant consideration of further mixing of the two subspecies.

4.2.2. Limited inbreeding detected in boodie populations

In addition to increasing genetic diversity, minimising relatedness among individuals within a population is a key goal of translocations to help limit inbreeding in subsequent generations and retain genetic diversity (Ralls et al., 2018). We had expected to see reduced relatedness in populations with multiple sources, given these populations are founded by individuals with different genealogical histories, over single sources. There was no strong indication in the genetic data that extensive inbreeding was occurring in any population with few highly related individuals detected in kinship analyses, although a slight trend towards inbreeding rather than outbreeding was observed in internal relatedness (IR) estimates. Several life‐history traits may be enabling avoidance of inbreeding in boodies, as detected in other macropodoids (Farquharson et al., 2021; Potter et al., 2012; Spencer et al., 1997). Male boodies have been noted to disperse a mean distance of 4.6 km from mothers upon reaching sexual maturity and female boodies are known to disperse up to 2 km from paternal burrows (Parsons et al., 2002). In addition, the potential for a high growth rate (Hone et al., 2010) and rapid increase in population size during the irruptive ‘boom’ dynamics of managed populations may explain the low levels of relatedness observed, even in populations that were founded with few animals (i.e. Faure Island and Scotia Sanctuary) (Mills & Smouse, 1994; Nei et al., 1975). These characteristics may help to explain the high success rates of the boodie translocation programme in comparison to other mammal species (Short, 2009). Given these results, our estimates of effective population size were generally low in all populations (mean = 47.6) and indicate populations may potentially be susceptible to inbreeding depression over time (i.e. N _e < 100; Frankham et al., 2014). We stress that cautious interpretation of these results is required, given the bias that exists in N _e estimation when there have been recent population fluctuations as experienced by boodies (Russell & Fewster, 2009), and when there may be genetic substructuring present (Kopatz et al., 2017). It is important to note that fluctuations in population size have occurred in many of our study populations since sampling was undertaken and more recent evaluation may be required.

4.3. Implications for management of translocated mammal populations

4.3.1. General recommendations

Genetic data allow conservation practitioners to make strategic decisions when sourcing animals for use in a translocation programme and for evaluating past conservation management actions (Weeks et al., 2011; White et al., 2018). Here, we have demonstrated positive genetic outcomes, in terms of increased genetic diversity, as a result of strategic mixing in conservation translocations of a threatened marsupial species. Although this outcome might seem intuitive, the fate of translocated populations is often uncertain. Translocation success depends on many factors, including the number, sex ratio and genetic diversity of founders as well as the potential for both inbreeding and, in mixed‐source translocations, outbreeding depression to occur (Mock et al., 2004; Poirier et al., 2019). The speed of population growth following establishment also influences success, with slower growth contributing to a loss of genetic diversity (Puckett et al., 2014). For example, a rapidly growing translocation population of the black‐footed ferret (Mustela nigripes) was found to maintain levels of diversity comparable to the source population, whereas populations that did not grow rapidly or receive augmentation with additional individuals were observed to decline dramatically after 5–10 years (Wisely et al., 2008). In the Alpine ibex (Capra ibex ibex), even 100 years post‐translocation, re‐introduction history was still the greatest driver of contemporary genetic patterns in translocated populations, all of which showed lower diversity than the ancestral source (Bieback & Keller, 2009). Even with apparent positive outcomes following translocation, these studies highlight the importance of long‐term population studies to monitor success.

The results presented here add weight to the use of deliberate admixture in conservation management and provide a further empirical example of how this can be applied to conservation of isolated relictual populations. Outbreeding depression is often cited as a reason to avoid mixing genetically divergent populations, but this perceived risk may be overestimated (Frankham, 2015), and a lack of action may ultimately be to the detriment of the species' survival (Lott et al., 2022; Ralls et al., 2018), particularly when genetic erosion and/or inbreeding depression is a real and proximal risk.

Furthermore, our results emphasise that in threatened species, the common practice of prioritising conservation of genetic uniqueness and maintenance of taxonomic integrity over other conservation actions is potentially problematic in the modern landscape (Ralls et al., 2018). This approach, often implemented in the past in order to preserve species' adaptive diversity, may be misguided when genetic ‘uniqueness’ is measured through commonly used metrics of population genetic differentiation such as the fixation index, F _ST (Weeks et al., 2016). F _ST responds rapidly to changes in allele frequencies that occur as a result of population isolation and genetic drift, and as such, in fragmented populations, is most likely to be reflecting random drift processes, rather than directed evolution through adaptation and natural selection. Taxonomic inflation through formal or informal naming of allopatric, differentiated (albeit drifted) population segments can therefore be a barrier to effective conservation action for threatened species. This problem is particularly pervasive in Australian mammals, with many threatened species now persisting in very small, isolated mainland remnants or on offshore islands (Lomolino & Channell, 1995) where many have been named as different subspecies. We have shown here that gene pool mixing has been an effective conservation strategy to improve the adaptive potential of reintroduced populations, restoring genetic diversity to similar levels observed in historical mainland specimens, with no evidence of outbreeding depression (under present conditions) currently identified. This implies that reconsideration of subspecies concepts in taxonomy may be required to ensure that they are fit for purpose in contemporary conservation contexts and/or that conservation managers more pragmatically assess the risks and benefits of gene pool mixing in threatened species. Currently, this is not consistently done (Liddell et al., 2021).

Finally, we advocate for the use of historical samples such as museum specimens as invaluable sources of genetic data that provide context for understanding the loss of genetic variation in contemporary populations as well as providing a benchmark for managers to aim for in terms of restoring genetic diversity to pre‐decline levels.

4.3.2. Management of Boodie populations

Boodies provide an excellent case study for investigating conservation of isolated, relictual populations and their management via the establishment of translocation populations. Boodie populations located on Shark Bay islands (subspecies lesueur) have been isolated from the mainland approximately 8–10 kya which puts the coalescence of subspecies lesueur and subspecies Barrow Is lineages at least greater than 10,000 years before present (Sander et al., 1997), with substantial genetic divergence (measured as F _ST) also indicated between populations within subspecies (Bernier and Dorre Island), isolated from each other for only a comparatively short timeframe (3–6 kya) (Ride et al., 1962). Each of the remnant populations experiences a paucity of genetic diversity, relative to the previously extant mainland animals, implicating genetic drift as a major source of differentiation among island populations and a source of ongoing risk to these populations.

Here, through retrospective analysis, we demonstrated that by using multiple sources in translocations, we have increased the genetic diversity of boodie populations, with the greatest gains achieved by mixing between the two subspecies. Deliberate mixing between divergent populations or lineages has, for a long time, been controversial due to concerns relating to the risk of outbreeding depression (Frankham, 2015; Frankham et al., 2011). However, empirical evidence for outbreeding depression is rare and a meta‐analysis concluded that outcrossing differentiated lineages resulted in beneficial effects in nearly 93% of all cases studied (Frankham, 2015). Indeed, a similar positive increase in diversity and fitness was documented in a recent study that examined the impact of mixing subspecies of Equus hemionus for the purposes of reintroduction (Zecherle et al., 2021). Given the long period of isolation between the Barrow Island and Shark Bay island boodie populations and the heritable phenotypic differences between them (Rick et al., 2019; Thavornkanlapachai et al., 2019), outbreeding depression in later generations was considered a risk (Frankham, 2015). Despite these differences, and the apparent initial asymmetrical introgression observed at Matuwa (Thavornkanlapachai et al., 2019), there has been no evidence of outbreeding depression in hybrid (up to F2) survivorship or recruitment occurring in over three generations of admixed Matuwa boodies within the fenced haven (Rick et al., 2019). Fitness assessment of hybrid individuals is still required under more challenging environmental conditions. Experimental evidence suggests that boodies with larger body size (measured as hind foot length) may have higher survivorship in the presence of feral cats due to their size (Bannister et al., 2021) and enhanced escape response (Tay et al., 2021). This implies that dominance of Barrow Island genotypes (smaller body type) as observed at Matuwa might be maladaptive for establishing populations outside fences in the presence of feral predators. Ultimately, the goal for boodie conservation is to return populations to the wild, which in the short term, involves surviving alongside feral predators. Managing selection and evolutionary processes under semi‐wild conditions in conservation safe havens where feral predators are absent is an emerging challenge to species conservation in Australia.

Future translocations would benefit from sourcing animals from all three natural populations of boodies, or their representative single‐source translocated population equivalents, that is, Faure Island, Alpha Island, to maximise genetic input to founder populations. Considering that translocations to mainland Australia will be to habitats that have been modified compared to pre‐decline conditions, and to climatic zones different from those on source islands, maximising genetic diversity in founder populations should facilitate adaptation to novel conditions. We note that, in our analyses, Bernier Island and Barrow Island boodies are under‐represented in mainland translocations. Efforts to improve representation of these lineages are recommended.

The genetic data did not identify any translocation populations of immediate conservation concern. Nevertheless, a lack of gene flow is likely contributing to ongoing genetic erosion as the majority of populations are small and below the expected threshold (N _e > 500) required to prevent genetic diversity loss (Franklin, 1980; Jamieson & Allendorf, 2012). To maintain genetic diversity at the metapopulation level, a programme of periodic supplementation through managed transfer of animals across sanctuaries would be beneficial (Ottewell et al., 2014). Likewise, consideration might also be given to genetic exchange between the adjacent Bernier and Dorre Island populations, which we know to be susceptible to cyclical population crashes and ongoing genetic erosion. Exchange of low numbers of animals between these islands during low periods in the population cycle could improve genetic diversity without overwhelming patterns of local adaptation (although adaptation is shown to persist in some circumstances even under high gene flow; Fitzpatrick et al., 2015). Animal transfer should be balanced against the risk of introducing novel pathogens or parasites that may disrupt local populations (Dalziel et al., 2017). In addition, since sampling of the Arid Recovery population in 2014 and Matuwa in 2018, both populations have experienced significant declines as a result of severe drought conditions, and re‐evaluation of the genetic diversity of these populations is recommended.

AUTHOR CONTRIBUTIONS

KO and HMN were responsible for conceptualisation of the project and analyses, HMN was responsible for formal analysis and preparing the original draft; ER and KO were involved in formal analysis; ER, AJM, SM, PGSG, CS, SC, LW, JP, KM and KO in data curation; CM, MDBE, MB and KO acquired funding; AJM, MDBE and KO administered the project; CS, SC and LW provided resources; all authors reviewed and edited the manuscript.

CONFLICT OF INTEREST STATEMENT

There are no conflicts of interest to report.

BENEFIT‐SHARING STATEMENT

Our research findings are used to actively inform the boodie translocation programme and apply more broadly to the conservation of endangered mammals both in Australia and across the globe. This research was a multi‐institutional collaboration and these researchers form co‐authors on this publication.

Supporting information

Appendix S1

Nistelberger, H. M. , Roycroft, E. , Macdonald, A. J. , McArthur, S. , White, L. C. , Grady, P. G. S. , Pierson, J. , Sims, C. , Cowen, S. , Moseby, K. , Tuft, K. , Moritz, C. , Eldridge, M. D. B. , Byrne, M. , & Ottewell, K. (2025). Genetic mixing in conservation translocations increases diversity of a keystone threatened species, Bettongia lesueur . Molecular Ecology, 34, e17119. 10.1111/mec.17119

DATA AVAILABILITY STATEMENT

All raw sequencing data for ddRADseq and exon capture data are available online from the Oz Mammals Genomics Initiative data portal (https://data.bioplatforms.com/dataset) (data set IDs 102.100.100/52618 and 102.100.100/52588 (ddRAD) and 102.100.100/52583, 102.100.100/52584, 102.100.100/52629, 102.100.100/52636, 102.100.100/52637, 102.100.100/52638 and 102.100.100/52639 (exon capture)). These data are made available openly under a Creative Commons Attribution licence. ddRADseq de‐multiplexed fastq files and metadata are available on the NCBI Short Read Archive (PRJNA988247; SUB13568044). Tissue sample collection from Matuwa was undertaken with approval from the Animal Ethics committee of the Department of Biodiversity, Conservation and Attractions (permit number AEC: 2016‐10 and 2018‐02).