Isolation, marine transgression and translocation of the bare‐nosed wombat (Vombatus ursinus)

Abstract Island populations can represent genetically distinct and evolutionarily important lineages relative to mainland conspecifics. However, phenotypic divergence of island populations does not necessarily reflect genetic divergence, particularly for lineages inhabiting islands periodically connected during Pleistocene low sea stands. Marine barriers may also not be solely responsible for any divergence that is observed. Here, we investigated genetic divergence among and within the three phenotypically distinct subspecies of bare‐nosed wombats (Vombatus ursinus) in south‐east Australia that are presently—but were not historically—isolated by marine barriers. Using genome‐wide single nucleotide polymorphisms, we identified three genetically distinct groups (mainland Australia, Bass Strait island, and Tasmania) corresponding to the recognized subspecies. However, isolation by distance was observed in the Tasmanian population, indicating additional constraints on gene flow can contribute to divergence in the absence of marine barriers, and may also explain genetic structuring among fragmented mainland populations. We additionally confirm origins and quantify the genetic divergence of an island population 46 years after the introduction of 21 individuals from the Vulnerable Bass Strait subspecies. In the light of our findings, we make recommendations for the maintenance of genetic variation and fitness across the species range.

during glacial periods when sea levels are low (most recently in the Pleistocene; Burridge, 2012). Depending upon the timing, duration and frequency of these connections, and the nature of intervening habitats, gene flow may have been experienced between lineages occupying presently isolated regions. This raises questions regarding their conservation prioritization given uncertainty about their history of genetic isolation. Furthermore, phenotypic distinction of lineages on continental shelf islands may also be problematic to interpret if the peripheral geographic setting of these islands confers environmental differences (Mullen, Vignieri, Gore, & Hoekstra, 2009), in addition to potential influences of island size alone (e.g., dwarfism in island emus; Thomson et al., 2018). This is a question of broad conservation interest, as continental shelf islands are common and host high biodiversity, most notably in South-East Asia (e.g., the entire Malay Archipelago), but also Europe (e.g., England and many islands of the Mediterranean), North America (e.g., Newfoundland), South America (e.g., Falkland Islands) and Australia (e.g., Tasmania; Burridge, 2012).
Historical sea-level rise associated with the end of the last glacial maximum (LGM) potentially played a significant role in the biogeography of south-eastern Australia. This event isolated Tasmania and an array of islands from continental Australia during the flooding of Bass Strait, protecting some populations from causes of extinction that are present on the mainland (e.g., invasive predators; Kinnear, Sumner, & Onus, 2002), and shaping the population genetic structure of others (Firestone, Elphinstone, Sherwin, & Houlden, 1999;Toon, Mather, Baker, Durrant, & Hughes, 2007). These areas were connected by the Bassian land bridge during the LGM circa 25 kya (Lambeck & Chappell, 2001). As sea level rose, the mainland, Tasmania, and intervening islands remained connected through a western sill until around 17.5 kya and an eastern sill until around 14 kya (Lambeck & Chappell, 2001). Many species still occur across these now isolated regions, with Bass Strait and offshore Tasmanian islands exhibiting high species richness per unit area relative to other Australian islands (Burbidge, Williams, & Abbott, 1997), and supporting populations of mammals which are now extinct or declining on mainland Australia (Morris et al., 2018). These island populations may represent important genetic lineages and evolutionary legacies that are distinct from the mainland (e.g., platypus ;Furlan et al., 2012), or may be representative of the mainland genetic pool (e.g., white-bellied sea-eagles; Shephard, Hughes, Catterll, & Olsen, 2005).
Wombats are evolutionarily significant as the largest extant burrowing mammals (Johnson, 1998). The bare-nosed wombat (Vombatus ursinus) is a large (up to 50 kg), fossorial marsupial endemic and historically widespread in south-east Australia (mainland and islands, Figure 1; Triggs, 2009;IUCN, 2016). Within this range, there are three recognized allopatric subspecies: south-eastern mainland (Vombatus u. hirsutus; Perry 1810), Bass Strait islands (V. u. ursinus; Shaw 1800) and Tasmanian (Vombatus u. tasmaniensis;Spender and Kershaw, 1910) (Jackson, 2015). These subspecies are distinguished based on distribution and body size, with mainland individuals being the largest and Flinders Island being the smallest (Tate, 1951)-though these distinctions are in need of revisitation in an updated and comprehensive way. Despite being considered "common"-V. ursinus Least Concern on IUCN Red List (Taggart, Martin, & Menkhorst, 2016)-all three subspecies have experienced range retractions since settlement by Europeans (Figure 1), and may support F I G U R E 1 The bare-nosed wombat distribution across Australia. Sampling locations and sample size are indicated by the circles (Supporting information Data S1 for location coordinates). Spatial data for the current distribution accessed from the International Union for Conservation of Nature (IUCN, 2016) several genetically important, yet isolated populations. Specifically, the range of V. u. hirsutus has been fragmented and more than halved, and similar retraction has been observed in V. u. ursinus, which now exists only on Flinders Island, having gone extinct on King, Cape Barren, Deal and Clarke islands (Rounsevell, Taylor, & Hocking, 1991). The Tasmanian subspecies exists throughout Tasmania with seemingly stable populations across its range (Figure 1; DPIPWE, 2017). A growing population also exists on Maria Island (Figure 1; Ingram, 2015), which may represent the descendants of 21 individuals translocated from Flinders Island (Rounsevell, 1989), and hence potentially of conservation significance for V. u. ursinus. However, records are inconsistent as to whether V. ursinus existed on Maria Island prior to this translocation event (Plomley, Cornell, & Banks, 1990;Rounsevell et al., 1991).
Despite range retractions observed in V. ursinus, it is still distributed relatively continuously, but with areas of fragmentation in the western and northern edges of the mainland distribution (IUCN, 2016). Assessing genetic structure within subspecies could reveal important biological processes, such as dispersal limitations and barriers to gene flow, that are also relevant for conservation with respect to the maintenance of genetic diversity. Evidence for isolation by distance has been observed for V. u. hirsutus, with high levels of population differentiation at larger spatial scales (Banks, Skerratt, & Taylor, 2002). However, sampling in this study was spatially clumped, and patterns of genetic structure and isolation by distance should be addressed within a continuously sampled region (Bradburd, Coop, & Ralph, 2018;Rosenberg et al., 2005). Assessing genetic structure within regions (mainland and Tasmania) also provides a valuable contrast for genetic structuring that may be ascribed to isolation by historical sea-level rise.
Here, we utilize genome-wide single nucleotide polymorphisms (SNPs) to (a) quantify the population structure of bare-nosed wombats across their current range in the context of the presently recognized subspecies and their potentially dynamic history of connectivity, (b) document within region genetic variation to assess gene flow within a continuously distributed and sampled subspecies (V. u. tasmaniensis), and (c) assess the genetic provenance of the Maria Island population with respect to conservation genetic resources of V. u. ursinus. Discovery of genetically distinct populations across the wombat range will assist in determining spatial units that warrant independent management and support ongoing conservation planning for this Australian marsupial.

| SNP discovery and filtering
High-molecular-weight DNA samples (n = 176), representative of the bare-nosed wombat distribution, were sent to Diversity Arrays Technology Pty Ltd (DArT), Canberra, Australia, for DArTseq analysis. DArTseq utilizes complexity reduction (restriction enzymes PstI and compliment, retained by DArT) and next-generation sequencing methodologies to produce genome-wide SNPs (Kilian et al., 2012;Sansaloni et al., 2011). A total of 28,081 SNPs were identified for V. ursinus. SNPs were filtered using the following exclusion criteria: reproducibility (<95%), missing data per locus (>20%), missing data per individual (>10%), secondaries (if multiple SNPs fall on the same sequence, removed the SNP with the lower read count average), minor allele frequencies (≤0.05), mean read depth per sample (<8) and heterozygosity (>0.5). Outlier SNPs identified according to both pcadapt (Luu, Bazin, & Blum, 2017) and sNMF (Frichot & François, 2015) were removed. Deviation from Hardy-Weinberg equilibrium (HWE) was assessed for three sampling regions in Genepop (Rousset, 2008): Tasmania, Maria and Flinders islands, and one mainland location (central Victoria). SNPs that were out of HWE in two or more of these sampling regions were removed from the data set (n = 372).
This approach was taken to reduce the risk of mis-identifying SNPs as out of HWE that are truly reflective of genetic structure (see Section 2.3 for comparative analyses performed including these SNPs). Filtering resulted in a total of 9,064 SNPs for 162 individuals (mainland, n = 76; Flinders, n = 6; Tasmania, n = 74; Maria Island, n = 6; Supporting information Data S1 and S2).

| Diversity estimates and population structure
Heterozygosity, allelic richness and F ST were estimated using the R packages diveRsity (Keenan, McGinnity, Cross, Crozier, & Prodöhl, 2013) and strataG (Archer, Adams, & Schneiders, 2017). Population structure was explored using a combination of multivariate and Bayesian methodologies. We focused on understanding structure at two different geographic scales: (a) among the three bare-nosed wombat subspecies and (b) within the Tasmanian subspecies only, to reveal fine-scale structure across a continuous sampling range.
PCA and fastSTRUCTURE were also performed including SNPs that violated our HWE filtering criterion for comparative purposes (Supporting information Data S3).
No additional structure analyses (beyond PCA and fastSTRUC-TURE) were performed for the mainland region given the discrete spatial sample distribution and potential for false inference of genetic breaks if isolation by distance operates (Bradburd et al., 2018;Serre & Pääbo, 2004). However, further estimates of genetic diversity and differentiation were performed for the discrete populations located across the mainland (Supporting information Data S4). In Tasmania, where sampling was more continuous, a spatial principal component analysis (sPCA, package adespatial, Dray et al., 2018) was performed. sPCA incorporates both genetic variation and spatial autocorrelation (spatial weighting matrices) to explain observed patterns (Jombart, Devillard, Dufour, & Pontier, 2008). A Gabriel's graph was employed as the connection network, and sPCA scores were visually represented using the R package ade4 (Dray & Dufour, 2007).
To complement the sPCA, we investigated isolation by distance in Tasmania by employing a redundancy analysis (RDA) following the methodology of Meirmans (2015). The RDA was performed as an individual-rather than population-based analysis, whereby the dependent variable was the allele count per locus per individual, and the independent variable was a set of spatial polynomials derived from geographic coordinates. It is important to note that potential landscape inhibitors to movement (e.g., lakes and rivers) are not considered by this approach. The RDA was performed in R using the package VEGAN (Oksanen et al., 2018).

| Diversity estimates
Diversity estimates are described in Table 1

| Population structure
Pairwise fixation indices estimated among regions (pooled locations: mainland, Flinders and Maria islands, and Tasmania) ranged from 0.24 to 0.33 (Table 2), and all were significant (p ≤ 0.01) after correction for false discovery rates. The mainland was less differ-

| Genetic differentiation among V. ursinus subspecies
Designations of V. ursinus subspecies originated in the mid-1800s and early 1900s when differences in body and skull size were observed in the geographically separated groups. The mainland subspecies was described as the largest, and Flinders Island V. ursinus were the smallest (Tate, 1951). While body size is often a distinguishable feature between island populations and their mainland conspecifics (Lomolino, 1985), observed differences between groups do not necessarily denote genetic divergence (Thomson et al., 2018  an additional ~5-7 kyr (Coller, 2007;Lambeck & Chappell, 2001).
However, mainland and Flinders Island subspecies exhibit less genetic distinction from each other than when compared to Tasmania.
Two plausible explanations exist for these patterns. First, it is possible that gene flow across the Bassian Plain was influenced by factors other than sea level, and that despite being physically connected, geneflow was not achieved between Tasmania and Flinders Island following their isolation from the mainland. Second, F ST is influenced by both population size and gene flow (Meirmans & Hedrick, 2011), and thus, a combination of our sample sizes and the population sizes may have influenced the genetic divergence observed. Therefore, estimates of divergence time are required to assess whether marine barriers initiated or reinforced the isolation of these populations (e.g., Burridge et al., 2013), and should be pursued in future analyses.

| Genetic structure within subspecies
Within Tasmania, where sampling was more continuous, there was evidence for isolation by distance. While bare-nosed wombats are capable of dispersal across varied landscapes (as their distribution suggests), they exhibit relatively small home ranges (on average 17.7 ha; Evans, 2008). Furthermore, dispersal is female-biased in all wombat species (Banks et al., 2002;Johnson & Crossman, 1991;Walker, Taylor, & Sunnucks, 2008), and though the extent of these movements is not well understood, there is molecular-and tracking-based evidence that suggests they are of short distances (100-3,000 m). These short-distance dispersal behaviours may provide some explanation for the isolation by distance observed within Tasmania. The exception to this pattern was observed in east-west comparisons in the north-central region of Tasmania, where geographically close individuals were genetically dissimilar, in a manner akin to a "ring species" (Irwin, Irwin, & Price, 2001). This likely reflects long-term barriers to gene flow present in this region, such as the Tamar River, with more recent (and likely weaker) impact from urbanization (the city of Launceston, the second largest city in Tasmania) and degraded landscapes (agricultural lands). Future research should investigate landscape features at finer scales to disentangle the potential contributors to this genetic break.
While most Tasmanian individuals were assigned to the same population cluster (n = 66, >90% ancestry assigned to the same cluster), it is worth noting that eight individuals were assigned (>99% ancestry) to a separate population cluster. Seven of these individuals were from the Tamar Valley region (north-central Tasmania), specifically Narawntapu National Park and Greens Beach area. These Though our mainland sampling was more spatially discrete, which places constraints on the interpretation of genetic structuring (Bradburd et al., 2018), we found high genetic differentiation within V. u. hirsutus specifically against the South Australian samples (SA).
This longitudinal pattern of genetic differentiation is consistent with previous studies of V. u. hirsutus, using microsatellite loci (Banks et al., 2002). This may be reflective of the recent fragmentation across the western range of V. u. hirsutus (IUCN, 2016), as the eastern mainland is less differentiated over comparable spatial scales. Further, the SA population is likely smaller and thus more susceptible to genetic drift (Frankham, 1996). These patterns may also be observed in the fragmented range in southern Queensland and northern New South Wales; however, samples from these regions were absent from our analyses. Finer spatial sampling across the mainland is required to determine factors responsible for genetic structuring in this region.

| V. u. ursinus on Maria Island
Bare-nosed wombats have been subjected to considerable human interference across the Bass Strait islands, becoming extinct on King, Cape Barren, Flinders, Deal and Clarke islands. Given this history, F I G U R E 3 fastSTRUCTURE (a) and spatial principal component (sPCA) (b and c) results for Tasmanian individuals. Spatial mapping of the principal components 1 (b) and 2 (c) of the sPCA visually represents genetic differentiation proportional to difference in square size and shade. Arrows designate where the eight individuals assigned to the separate cluster (>99% ancestry proportion; a) are geographically located  (Rounsevell et al., 1991). However, the present population is prolific and has experienced growth over the last decade (Ingram, 2015 (Frankham, 1997), but may require management action if low fitness is observed in the future (i.e., genetic rescue; Frankham, 2015;Whiteley et al., 2015).

| Applied evolutionary management
There is ongoing debate regarding the genetic identification of intraspecific units warranting independent conservation (Coates, Byrne, & Moritz, 2018 (Hedrick & Fredrickson, 2010;Ralls et al., 2018;Weeks et al., 2016). Research on bare-nosed wombats to assess fitness and adaptive distinction has been insufficient, although dramatic population declines have been observed in some areas (e.g., in response to novel pathogens; Martin et al., 2018).
As additional resources become available (i.e., the annotation of the wombat genome), questions regarding adaptive distinction can also be investigated more thoroughly (Pardo-Diaz, Salazar, & Jiggins, 2015). Regardless, if fitness is low, there are potential benefits through the incorporation of genetic variation from other populations ("genetic rescue"; Frankham, 2015;Ralls et al., 2018).
However, controlled crosses need first be conducted to assess potential fitness benefits, and the risk of outbreeding depression (although these appear overstated, generally; Frankham et al., 2011).
The Bass Strait islands previously harbouring V. u. ursinus provide an ideal opportunity to both establish additional insurance populations of pure V. u. ursinus, and also to test the potential fitness benefits of crosses within and between subspecies, if indeed natural populations are ascertained to be threatened by low fitness.

ACK N OWLED G EM ENTS
We thank our funding bodies, the Wildlife Disease Association Australasia sector and the Holsworth Wildlife Endowment.
We also thank everyone involved in tissue collection-Regina Magierowkski, Georgia Roberts, Rowena Hamer, Riana Gardiner, Sebastien Comte, Tamieka Lunn, Hanh Nguyen, Georgeanna Story, Julie Old, Janine Deakin, Tim Collins-among others. Last, but certainly not least, we would like to thank Adam Smolenski, whose invaluable expertise in molecular techniques and methodologies made this project possible.

CO N FLI C T O F I NTE R E S T
The authors declare no competing interests. and Australian National University (A2016/08).

DATA AVA I L A B I L I T Y
Reference sequences and SNP genotypes are available at the Dryad