Reestablishing B. tropica in the southern part of its former range could greatly improve the species’ conservation status. Understanding source population genetics improves the chance of successful reintroduction and establishment of a genetically diverse population. We found four, existing, genetically differentiated populations, from which individuals could potentially be sourced for reintroduction. However, these populations had differing levels of genetic diversity, and there were differences in the degree to which populations were genetically isolated. One population, occurring at Mt Spurgeon, had a particularly small effective size and low genetic diversity. We use our findings to develop recommendations for reintroduction.
4.1 Genetic characteristics of extant B. tropica populations
Genetic diversity was similar across the three Lamb Range populations, but lower at Mt Spurgeon. Genetic diversity is important for fitness and adaption, and low genetic diversity is associated with inbreeding depression (Alho et al. 2009; Armstrong et al. 2015rénos et al. 2016; Frankham et al. 2017a; Weeks et al. 2017). The heterozygosity estimate for Mt Spurgeon is very low compared to most estimates for wild mammal populations found in the literature, with a few exceptions from endangered species such as arctic ringed seals, polar bears, black bears and western barred bandicoots (Kjeldsen et al. 2016; White et al. 2018). The Lamb Range populations have similar heterozygosity to other endangered marsupial species that have suffered major population declines such as natural (non-admixed) burrowing bettong (Bettongia lesueur) populations (White et al. 2018; Rick et al. 2019) and reintroduced or captive greater bilby (Macrotis lagotis) populations (Lott et al. 2020). The Lamb Range average heterozygosity is slightly lower than estimates for species with large or outbred populations (Kjeldsen et al. 2016). Nevertheless, cross-study comparisons of SNP heterozygosity should be viewed cautiously, and inference is limited by ascertainment and sample size biases (Schmidt et al. 2020).
All B. tropica populations had small effective sizes, suggesting that they may be vulnerable to further diversity loss through genetic drift. This has implications for the long-term survival of these populations and for management, including reintroduction. The Mt Spurgeon estimate fell below the ‘50’ minimum viable size threshold for preventing immediate inbreeding effects (Harmon and Braude 2010; Jamieson and Allendorf 2012). Populations below this threshold will theoretically lose heterozygosity at a rate of 1/(2*Ne) due to drift, thus and an Ne estimate of predicts a loss of 1.67% heterozygosity per generation, potentially explaining the low diversity observed at Mt Spurgeon. Although larger than Mt Spurgeon, effective population size estimates in the Lamb Range were well below the ‘500’ effective individuals theoretically required for long-term population viability. Moreover, Frankham et al. (2014) argue that the 50/500 rule is too low, and in fact more than 100 effective individuals are required to keep fitness decline due to inbreeding depression below 10%. Yet, despite the implications of having small effective sizes, populations in the Lamb Range appear to have been stable over the last 20 years (Whitehead 2018).
Removing migrants from the Emu Creek and Tinaroo samples reduced effective size estimates. Fewer immigrants may have had proportionally higher effect on Emu Creek effective size estimates compared to Tinaroo because of the smaller relative size of that population. Over-estimation of local effective size from linkage disequilibrium is expected when migration levels are high and migrant individuals are included in the sample, because the parent pool being sampled is actually the metapopulation, not just the local population (Waples and England 2011). Whilst using ‘local’ estimates (i.e., excluding migrants) is necessary to avoid double counting when inferring census size from effective size, we argue that for our objective of assessing drift risk, ‘metapopulation’ estimates (i.e., including migrants) are more suitable because they implicitly capture the mitigating effect of gene flow on drift. Indeed, we found that even though Emu Creek and Mt Spurgeon have similar local effective sizes, Emu has retained significantly higher genetic diversity and thus is less susceptible to inbreeding depression. Gene flow greatly increases population persistence (Mims et al. 2019), therefore connectivity likely plays a crucial role in maintaining genetic diversity and preventing decline of the Lamb Range populations, and explains the population stability.
Genetic distinctness was correlated with population genetic diversity, suggesting drift is the primary driver of genetic structure in this species. In fact, almost the entire variance in population divergence, measured by population-specific FST, was explained by genetic drift. Mt Spurgeon is the most geographically isolated population, and as expected, was also the most genetically distinct. However, the genetic uniqueness of Mt Spurgeon was predominantly characterized by an absence of genetic diversity found in the Lamb Range. In other words, many loci heterozygous in other populations were homozygous in all individuals sampled at Mt Spurgeon. This suggests strong drift has led to a loss of ancestral diversity and fixation at Mt Spurgeon since the time it diverged from its last common ancestor with the Lamb Range populations. Given the small effective size, it is unsurprising that little novel, potentially adaptive, variation has accrued at Mt Spurgeon. This adds to the growing body of evidence that maladaptive drift is a dominant force in the genetic differentiation of threatened species (Coleman et al. 2013; Weeks et al. 2016; Casey et al. 2018).
Genetic and geographic distances were much shorter between the Lamb Range populations, and we found genetic evidence of recent migration among populations. The three Lamb Range populations occur in continuous habitat and individuals are known to be capable of dispersing at least several kilometers (Chris Pocknee, personal communication, 2021). However we also found an isolation-by-distance effect occurring over short distances, confirming the findings of Pope et al. (2000). Populations were panmictic at 1.5km (Lower - Upper Davies Creek), but structure appeared at 7km (Emu Creek – Tinaroo), and then jumped to a much greater level of isolation at 11km (Davies Creek – Emu Creek). Gene flow is insufficient for panmixia across the Lamb Range.
The observed patterns of B. tropica genetic structure may, in part, reflect their historic biogeography. Pope et al. (2000) showed there was a deep split in B. tropica mitochondrial haplotypes within the Lamb Range, dividing northern B. tropica populations (i.e. Davies Creek, Mt Spurgeon and the now extinct Mt Windsor population) from southern populations (Tinaroo, and the now extinct Coane Range population), with admixture at Emu Creek (see Fig. 1). Pope et al. hypothesized this divergence was associated with periods of contraction of Wet Tropics rainforest during the Pleistocene, with secondary contact occurring relatively recently around 5000–7000 years ago when rainforest expanded (Moritz et al. 2009; Graham et al. 2010). It’s possible that some of this relic structure remains in the nuclear genome, if migration has been extremely limited over millennia. Conversely, if migration has occurred consistently at the rate we observed it would have presumably dissolved historic structure within this timeframe. We found that gene flow favored dispersal from Emu Creek south to Tinaroo, however this is opposite to the northward migration bias that was recorded in the late 1990s (Pope et al. 2000), suggesting that dispersal patterns are in flux and migration may have occurred in pulses rather than at a steady equilibrium rate. Alternatively, structure could be evolving dynamically, if very strong drift (due to the small effective size) is consistently causing diversity loss at a rate faster than is replenished by gene flow. Population history of B. tropica is likely complex and is yet unresolved.
4.2 Recommendations for reintroduction
Based on our genetic findings and existing information, we argue that the best strategy would be a ‘mixed source’ approach to reintroduction (Houde et al. 2015), whereby individuals are sourced from all three Lamb Range populations, rather than sourcing all individuals from a single population. This recommendation is supported by four arguments.
First, there are no closely related, or environmentally matched, populations available that are likely to confer pre-existing adaptation to the proposed reintroduction site. Until 2003, a natural B. tropica population occurred at the site, however, since the species’ regional extinction from the Coane Range, the nearest extant population is 250 km north, at Tinaroo in the Lamb Range. Pope et al. (2000), who sampled 10 individuals from the Coane Range population prior to its extinction, did find mitochondrial sequences more closely related to Tinaroo than the other populations, and Houde et al. (2015) recommended that if population ancestry could be identified, then the closest ancestral match should be used as the source to reestablish an extinct population. However, this is based on the assumption of genetic similarity, and genomic patterns are often discordant with mitochondrial patterns due to complex population histories of isolation and admixture (Singhal and Moritz 2013), as appears to be the case for B. tropica (Pope et al. 2000). Moreover, the environmental differences, and therefore divergent selection pressures, are far greater between the Coane Range and Lamb Range than they are within the Lamb Range. For example, the Coane Range is drier and more seasonally variable than the Lamb Range (Bateman et al. 2012). Thus, it is unlikely that the shared ancestry between Coane Range and Tinaroo (Lamb Range) populations would endow Tinaroo animals with fitness advantages over the other Lamb Range populations. In the absence of a source population with conferrable local adaptations, maximising diversity is the recommended approach for selecting sources (Houde et al. 2015).
Second, while we found the Lamb Range populations each have higher genetic diversity and larger effective size than Mt Spurgeon, they are still small and have each lost a (different) portion of the species’ diversity through genetic drift. The reintroduced population would most likely benefit from sourcing individuals from each of the populations so that more of the species’ diversity is captured, making it better equipped to adapt (Weeks et al. 2011). Offspring of mixed parents may also have increased fitness due to genetic rescue, helping the new population establish and increase quickly to a self-sustaining size (Frankham 2015; Weeks et al. 2017; MacFarlane 2019; Fitzpatrick et al. 2020). Rapid increase in size after founding from a few individuals is important to avert major genetic drift and allow purging of deleterious recessive alleles (Hoffmann et al. 2020). Other examples of bettong genetic rescue attest the benefit of mixing, across subspecies (Rick et al. 2019).
Third, the risk of outbreeding depression from mixing sources is low. We have demonstrated that the population differences within B. tropica are primarily attributable to genetic drift, and drift rarely generates outbreeding depression unless it has led to fixed chromosome differences (Frankham et al. 2017b). While it is unknown if there is any karyotype variation within B. tropica populations, it seems improbable, given chromosomes across species within this genus are conserved (Sharman et al. 1980). Outbreeding depression is most likely to occur when mixing allopatric populations adapted to different environments (e.g. Byrne and Silla 2020), whereas the Lamb Range populations inhabit similar environments, and all occur within 17 km of each other. Since outbreeding depression can occasionally evolve between isolated populations in similar environments, and this risk increases with increasing time since isolation, Frankham et al. (2011) recommended a 500-year gene flow threshold for mixing populations. Despite evidence of historic isolation of northern and southern Lamb Range B. tropica populations for tens or potentially hundreds of thousands of years in the mitochondrial genome (Pope et al. 2000; Haouchar et al. 2016), we found evidence of recent gene-flow. The range of admixture coefficients present in the three Lamb Range populations indicates not only migration, but survival and presumed fitness, of population hybrids over multiple generations. Furthermore, a 500-year threshold may be overly conservative for bettongs as mixing divergent Bettongia lesueur populations which are estimated to have been separated for more than 8000 years did not result in reduced fitness (Rick et al. 2019). Perhaps this is because most bettongs, including B. tropica and B. lesueur, have small, fragmented populations and therefore much larger genetic rescue effects obscure outbreeding depression upon mixing. Alternatively, mitochondrial divergence may simply not be a good predictor of reproductive isolation (Singhal and Moritz 2013). Even if outbreeding depression were to occur, fitness can be restored by selection within a few generations if the population size is sufficiently large (Frankham et al. 2011, 2017b).
Fourth, reintroductions should not be done at the expense of risking localised extinction by removing too many individuals from natural populations (IUCN 2013). Yet, reintroductions require a minimum number of founding individuals to be successful (Tracy et al. 2011). A total of 50 founding individuals have been proposed for this translocation (Australian Wildlife Conservancy, personal communication, 2020). We found that effective population sizes of all B. tropica populations were small, all less than 100, and therefore removal of 50 individuals from any one population may increase its extinction risk. A mixed source approach allows the impact of removals to be spread across the three populations, thereby reducing extinction risk to any particular population. Given that the Lamb Range populations all have similar levels of diversity, demographic or effective population size and genetic distance should be used as the key determining factors. Selecting more divergent individuals will capture greater diversity, and more individuals should come from the larger population, to ensure that existing populations are not harmed.
We do not, however, recommend sourcing any animals from Mt Spurgeon at this time. This population has low genetic diversity, and therefore would contribute little additional variation, and may even reduce fitness of the founding population by introducing deleterious genetic load (Wilder et al. 2020). Moreover, its very small effective population size means that removing any individuals would be overly risky. Persistence of the Mt Spurgeon population is of utmost conservation importance, and further management interventions should be considered to prevent its extinction, including genetic rescue.