A Major Highway Acts to Genetically Structure a Sugar Glider (Petaurus Breviceps) Population


 Arboreal gliders are vulnerable to habitat fragmentation and to barriers that extend their glide distance threshold. Habitat fragmentation through deforestation can cause population isolation and genetic drift in gliding mammals which in turn can result in a loss of genetic diversity and population long-term persistence. This study utilised next generation sequencing technology to call 11, 292 genome-wide SNPs from 90 adult sugar gliders (Petaurus breviceps). Samples were collected from 12 locations in the Lake Macquarie Local Government Area (New South Wales), with two of these locations west of the Pacific Motorway, a potential major barrier to their dispersal. Overall, Lake Macquarie sugar gliders appeared to have high levels of gene flow and little genetic differentiation, however spatial least cost path analyses identified the Pacific Motorway as a barrier to their dispersal. This Motorway is still relatively new (< 40 years old), so man-made crossing structures should be erected as a management priority to mitigate any long-term effects of population isolation by assisting in the dispersal and gene flow of the species. This study provides further insight into the sugar glider after it was classed as three separate species in 2020 and could potentially be used as a model for its threatened congener in the area, the squirrel glider (Petaurus norfolcensis).


Introduction
Eucalypt forests have experienced the greatest rate of deforestation in Australia, with 80% of the remaining forests modi ed by human activity (Bradshaw 2012). This habitat loss and fragmentation is a known major threat to arboreal marsupials that are dependent on forest food resources and require tree hollows for nesting, shelter and predator avoidance (Braithwaite et al. 1988; Gibbons and Lindenmayer 2000;Lindenmayer 2002). Also, arboreal marsupials have adaptations for climbing and gliding between trees that leave them slow and vulnerable to predators and vehicles when forced to cross their habitat on the ground (Bassarova et al. 2009; Warburton et al. 2012; Rupert et al. 2014). Because of this, there is a real and ongoing risk that populations may become isolated due to habitat fragmentation, resulting in reduced gene ow and genetic diversity (Frankham et al. 2002). This presents a problem for the long-term viability and persistence of populations, as genetic diversity facilitates their ability to adapt to stochastic changes in their environment (Mahoney and Springer 2009).
Conservation genetic research has examined the effect of straight-line distance on genetic distance of populations, with a non-signi cant result indicative of features in the landscape that may present barriers or challenges to gene ow and dispersal (Whitlock and Mccauley 1999). Over time, researchers have incorporated spatial data into Mantel tests to examine the effect of least cost path on genetic distances of populations (Wang et al. 2009;Milanesi et al. 2016). This concept of "landscape genetics" has proved extremely useful for pinpointing barriers to gene ow such as roads (Keller and Largiadèr 2003) and biogeographic barriers (Pérez-Espona et al. 2008; Wei et al. 2012). Additionally, least cost path analyses have identi ed corridors for priority conservation when paired with genetic data (Wang et al. 2009; Etherington et al. 2014). When investigating the effect of habitat fragmentation on population genetics, it is essential to combine spatial and genetic data to gain a better understanding of genetic structure (Storfer et al. 2007).
While studies have used microsatellites to investigate the effect of Australian habitat fragmentation on gliding mammals in the past (Pavlova et al. 2010;Taylor et al. 2011;Goldingay et al. 2013; Malekian et al. 2015), researchers are yet to utilise the power of next generation sequencing and genome-wide single nucleotide polymorphisms (SNPs). SNPs are single base pair nucleotide changes along the genome that vary for individuals, representing the most common form of sequence variation (Brum eld et al. 2003). Thanks to advances in next generation sequencing, thousands of bi-allelic, co-dominant SNP markers can be produced (Kumar et al. 2012). SNPs have wider genome coverage than microsatellites, making them more precise and informative when answering questions regarding genetic diversity and structure in non-model organisms (Liu et al. 2005 Despite having similar life histories and being congeners, sugar gliders (Petaurus breviceps) are currently listed as common, while squirrel gliders (Petaurus norfolcensis) are listed as threatened in New South Wales (Threatened Species Conservation Act 1995). It is possible that these congeners may respond to habitat fragmentation differently, warranting further investigation into the effects of urbanisation on them. On the other hand, if similar trends are observed, then sugar glider populations could be used as a 'model' for its threatened congener. While they do not often share habitat, certain locations in their distribution present a unique opportunity to examine this. One such location is in central New South Wales (ie. Lake Macquarie Local Government Area (LGA)) where the two species occur in high densities with some overlap (Smith 2002;Smith and Murray 2003; Knipler unpublished raw data).
Although sugar gliders are listed as a common speacies, threats to populations from ongoing urbanisation and land clearing remains a concern for their ongoing survival. In addition to this, sugar gliders were recently divided into three separate species: the sugar glider (Petaurus breviceps, found in Lake Macquarie and east of the Great Dividing Range), krefft's glider (Petaurus notatus, found in eastern Australia, though west of the Great Dividing Range and in Tasmania where it is an introduced species) and savanna glider (Petaurusariel, found in northern Australia) (Cremona et al. 2020). Because of this division, the conservation status of the species' perhaps need revision since there has likely been an overestimation of its range and effective population size by past studies that considered these as a single species. Cremona et al. (2020) call for targeted research on the three seperate species to better understand their ecology and identify any areas of conservation concern. Here we use genome-wide SNP markers to investigate the effect of habitat fragmentation on sugar glider (P. breviceps) population genetics in the Lake Macquarie LGA. We report on their current genetic diversity and structure and thus provide the rst baseline data available on their population genetics. Finally, we investigate barriers to their historical and ongoing dispersal and in doing so hypothesised that biogeographical barriers and habitat fragmentation would reduce genetic diversity and in uence population genetic structure. Least cost path analyses were used to examine this in detail and subsequently direct conservation measures with the intended outcome of contributing to data driven conservation outcomes for sugar gliders.

Study area
Sugar gliders require tree hollows for sleeping and nesting and primarily rely on the sap, gum and nectar of eucalypt, acacia, and banksia species (Smith 1982;Lindenmayer 2002). Together the Lake Macquarie LGA and neighbouring Newcastle LGA contains 47, 100 ha of native forest (55% of total area), the majority of which is comprised of medium open eucalypt forest that is suitable for glider species (Department of Agriculture Water and Environment 2021). Due to the eucalypt, acacia and banksia habitat spread across the landscape, Lake Macquarie LGA holds the most abundant squirrel glider population in New South Wales (NSW) (Smith 2002) and similarly holds a large population of sugar gliders (Smith and Murray 2003). This location is recognized as being the most genetically diverse for squirrel gliders in Australia, and is located 130 km north of Sydney, NSW (Pavlova et al. 2010) (Fig. 1). Additionally, Lake Macquarie LGA contains the largest coastal saltwater lake in Australia, (120 km 2 ) a potential biogeographical barrier to glider gene ow.

Live trapping and DNA collection
Live trapping was conducted at 32 sites from 2017 to 2020, with 113 sugar glider individuals caught across 12 of the sites ( Fig. 1). A combination of Mawbey traps, Elliot B traps, cage traps and Winning and King pipe traps were used to live trap gliders (Mawbey 1989;Quin 1995;Winning and King 2008). The Winning and King pipe traps were secured one metre from the base of the tree and the bottom was lled with leaves for bedding and a bait ball made from a mixture of peanut butter, honey and oats (Winning and King 2008). The other traps were instead secured to wooden planks that were drilled into tree trunks two metres above the ground. Each of these trap types contained a mix of leaves for bedding and a bait ball. A 1:4 ratio of honey water was then sprayed up and down the tree to a height of six metres as well as around the entrance to each trap as an olfactory attractant (Sharpe and Goldingay 2007). Each site was subject to at least one week of live trapping using 12 traps per site (two rows of six with each trap spaced 50 meters apart), and further focus was given to locations that required a larger sample size for intended genetic analyses. Trapping ceased when no new individuals were detected after one week or when a previously marked individual was recaught three times within a week.
Traps were checked each morning at sunrise. When a sugar glider was caught, body measurements were recorded. These measurements included weight, tail length, right hind foot length, head width, head length and sex to monitor the body condition of any recaptured animals. Individuals were given a unique identifying number in the form of a metal ear tag or a unique ear marking combination. Before being released, their DNA was collected in the form of an ear biopsy. A 2 mm metal ear punch (Able Scienti c, Australia) was sprayed with 70% ethanol and amed for sterilization. Once cooled, a small clipping was taken from the outside edge of the ear and stored in sterilized vials containing 95% ethanol. These were kept at -20°C prior to DNA extraction. The processing of each sugar glider was limited to ve-minutes to avoid unnecessary disturbance and following approved animal ethics protocol (AE15/11 and AE19/02). When the processing of an individual was complete, the glider was safely released onto the tree where it was caught. New geographic distances were calculated using the three friction matrices (cell factor = 15, function = mean) and the least cost paths between spatial coordinates of individuals. The new pairwise geographic distances were then tested for correlation with genetic distances and number of neighbours = 8. The results from the four spatial analyses (IBD and three least cost paths) were compared to nd the best t, as recommended by Milanesi et al. (2016 Next, PCoA analyses were conducted to visualise genetic differentiation between individuals and between sampling locations. Population differentiation was evident (yet minimal) in the PCoA plots, with populations separating out on different axes. The cumulative percentage of genetic variation explained by the rst eight axes was 20.4%, with axis 1 accounting for only 3.7% of the variation (Fig. 2). The clearest patterns noted in the PCoA plots included the separation of sampling location CP on the rst axis, OR on the second axis, AD on the fourth axis, and BEC on the sixth.
The AMOVA examined population structure and population differentiation and showed that a high proportion of the genetic variation was contained within sugar glider samples (89.2%) while only 5.4% of the genetic variation was observed between populations and 5.4% of the genetic variation was observed between samples within populations ( Table 2).  (Table 3).

STRUCTURE
The Structure Harvester results for individual-based structure and admixture showed that ΔK had the highest value at K = 5 (Fig. 3). Therefore, it was understood that the sugar glider samples collected from within the Lake Macquarie LGA were derived from ve ancestral genetic clusters. When K = 5, unique, pure clusters appeared in populations CP and OR in the form of cluster three and cluster ve respectively (Fig. 4).
Cluster admixture was present in all sampling locations except for population AD (Fig. 4). In population AD, 100% of the individuals were considered pure to cluster 1. In BEC, 75% of the individuals were pure to cluster 2. In CP, 70% of the individuals were pure to cluster 3. In FDW, 27% of the individuals were pure to cluster 4. In OR, 11% of the individuals were pure to cluster 1, 22% were pure to cluster 4 and 22% were pure to cluster 5. In WR, 27% of the individuals were pure to cluster 2. And nally, 89% and 55% of individuals were pure to cluster 1 in WYB and WYC consecutively. The rest of the individuals (including all of those in populations BR, HD, SP and WRW) were admixed (Fig. 4).

Results of least cost path analyses
The IBD analysis produced a signi cant result, with a correlation between genetic distance and euclidian geographic distance (IBD Mantel's r = 0.389, p = 0.011). The proportion of genetic variation explained by geographic distance did not increase when the lake was included as a biogeographical barrier (Mantel test "B", Mantel's r = 0.334, p = 0.098) , however it did signi cantly increase once highways were included alongside the biogeographical barriers (Mantel test "BH", Mantel's r = 0.619, p = 0.001) (Fig. 5a). When habitat was incoroporated, the Mantel r statistic only slightly increased (Mantel test "BHH", Mantel's r = 0.647, p = 0.001) (Fig. 5b), indicating that the highways explained the most genetic variation out of the three factors examined: biogeographical barriers, highways and habitat (Table 4).

Discussion
Habitat fragmentation can inhibit gene ow of arboreal gliders, particularly when gaps exceed their glide distance threshold and contain dangers to their survival in the form of high-speed vehicles or large bodies of water. In the current study, genome-wide SNPs proved useful in identifying ne-scale genetic structure and population differentiation of the sugar glider in the Lake Macquarie LGA.
Locations FDW, HAD, WR and WRW had little to no differentiation from each other and had expansive remnant bushland between them. On the other hand, location CP appeared moderately differentiated from other locations in the pairwise F ST and the STRUCTURE analysis results. One plausible explanation is biogeographic isolation since CP is on a peninsula, however the lake did not signi cantly explain genetic variation in the least cost path analysis. Overall, subtle genetic variation was detected in the PCoA and the AMOVA. This combined with the initial signi cant isolation-by-distance result suggests moderately high connectivity and random mating, and little structure in the overall sugar glider population. This In Lake Macquarie LGA the Paci c Motorway has been a dispersal barrier for the last 30 -35 years, impeding gene ow for approximately ten generations of sugar gliders. This is consistent with microsatellite research by Goldingay et al. (2013) who found genetic differentiation in populations of squirrel gliders within 30 years of landscape change in Queensland. It is important to catch this effect early on (as this study has done) to mitigate any long term effects on the genetic structure and viability of populations. The installation of rope bridges and glider poles would act as an asset to conservation management of sugar and squirrel gliders, as well as conserving any trees in the median strip to allow them to reach an optimal height for sugar gliders to successfully utilise them as stepping stones. Taylor and Rohweder (2013) observed sugar gliders using mature trees to cross a median strip when they were tall enough to support a successful glide (tall enough to support average glide angle of 29.69°, Jackson 2000). Additionally, rope bridges and glider poles have successfully encouraged squirrel gliders and sugar gliders to cross highways across the east coast of Australia. These species were observed using the man-made crossing structures through radio-tracking, hair-trapping and camera trapping While this study has been particularly insightful into the sugar glider population of Lake Macquarie LGA, future study should examine the threatened squirrel glider population in the LGA to see if they show similar responses to fragmentation. As a common species, sugar gliders may occur in higher abundance and therefore have a greater effective population size.
Genetic structure is not as pronounced in larger populations, but can have strong effects in small populations (commonly the case with threatened species) (Coleman et al. 2018).
In conclusion, this research was particularly insightful due to the recent division of the once sugar gliders into three glider species (sugar glider, krefft's glider and savanna glider) (Cremona et al. 2020). The results of this study suggests that the overall sugar glider population in Lake Macquarie LGA has appropriate levels of random mating with high levels of gene ow, however there is evidence that the relatively recent installation of the Paci c Motorway (30 -35 years ago) is starting to impose genetic structure on either side of the road. This can be mitigated with the installation of man-made crossing structures such as glider poles. Overall, the Lake Macquarie LGA sugar glider population has high levels of gene ow with potential for recovery in the area of concern (the Paci c Motorway).

Declarations
Funding: University of Wollongong Centre for Sustainable Ecosystem Solutions, Research Student Support Grant (2020).
Con icts of interest/Competing interests: The authors declare no competing interest.
Availability of data and material: DArTseq data is available from the authors upon request.  Figure 1 Sugar glider trapping locations from 2017 -2020 in the Lake Macquarie Local Government Area (LGA). See Table 1 for number of individuals and location information. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries.
This map has been provided by the authors.  had the largest Delta K value and is displayed in red.

Figure 4
Page 18/18 STRUCTURE admixture results for 90 sugar glider individuals and K = 5 clusters. Putative populations are divided by the white dashed lines and labelled below the plot. Each cluster is coloured as shown.

Figure 5
Relationship between genetic distances (FST/(1 -FST)) and logarithm of geographic distances (km) for 12 locations of sugar gliders in Lake Macquarie, using the results of the least cost path analysis (a) "Mantel test BH" and (b) "Mantel test BHH". Least cost path distances were calculated for each friction matrix as seen in top left corner of the plots. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.