Different landscape effects on the genetic structure of two broadly distributed woody legumes, Acacia salicina and A. stenophylla (Fabaceae)

Abstract Restoring degraded landscapes has primarily focused on re‐establishing native plant communities. However, little is known with respect to the diversity and distribution of most key revegetation species or the environmental and anthropogenic factors that may affect their demography and genetic structure. In this study, we investigated the genetic structure of two widespread Australian legume species (Acacia salicina and Acacia stenophylla) in the Murray–Darling Basin (MDB), a large agriculturally utilized region in Australia, and assessed the impact of landscape structure on genetic differentiation. We used AFLP genetic data and sampled a total of 28 A. salicina and 30 A. stenophylla sampling locations across southeastern Australia. We specifically evaluated the importance of four landscape features: forest cover, land cover, water stream cover, and elevation. We found that both species had high genetic diversity (mean percentage of polymorphic loci, 55.1% for A. salicina versus. 64.3% for A. stenophylla) and differentiation among local sampling locations (A. salicina: ΦPT = 0.301, 30%; A. stenophylla: ΦPT = 0.235, 23%). Population structure analysis showed that both species had high levels of structure (6 clusters each) and admixture in some sampling locations, particularly A. stenophylla. Although both species have a similar geographic range, the drivers of genetic connectivity for each species were very different. Genetic variation in A. salicina seems to be mainly driven by geographic distance, while for A. stenophylla, land cover appears to be the most important factor. This suggests that for the latter species, gene flow among populations is affected by habitat fragmentation. We conclude that these largely co‐occurring species require different management actions to maintain population connectivity. We recommend active management of A. stenophylla in the MDB to improve gene flow in the adversity of increasing disturbances (e.g., droughts) driven by climate change and anthropogenic factors.


| INTRODUC TI ON
Environmental changes through space and time can produce genetic differentiation (Fenderson et al., 2020). However, determining the role of specific environmental factors that cause genetic differentiation is still challenging. Changes in the landscape produced by climate change and intensified land use can generate a severe decrease of genetic connectivity and population viability in many plants and animals (Frankham et al., 2010). Therefore, the pervasive effects of global changes, and particularly habitat fragmentation, increase extinction risk of native species in urban and agriculturally intensified areas even in apparently resilient plant species (Vranckx et al., 2012;Young et al., 1996). Shrubby legumes belonging to the genus Acacia are highly diverse and widespread across the Australian continent. Acacias form major components of many ecosystems across the continent including many arid ecosystems with poor soils (Bui et al., 2014;Maslin et al., 2003) and play an important role in ecosystem functioning including through the provision of resources and habitat to a broad range of insects and animals (Wandrag et al., 2015;Ward & Branstetter, 2017;Young et al., 2008). It also helps rapid colonization supporting ecosystem recovery following disturbance (Spooner, 2005). Consequently, acacias often play a critical role in the restoration of highly degraded areas (Jeddi & Chaieb, 2012).
Restoration using Acacia species primarily occurs within regions where fragmentation of native vegetation is extensive (Doi & Ranamukhaarachchi, 2013;Jeddi & Chaieb, 2012). Fragmentation of large and continuous vegetation results in smaller, more isolated populations, often with lower genetic diversity, an increased risk of further genetic loss through drift and elevated inbreeding (Aguilar et al., 2006(Aguilar et al., , 2008Hamrick, 2004;Young et al., 1996). Consequently, there are risks associated with using seed crops from small populations for restoration purposes. Our understanding, however, about how landscape fragmentation and other environmental factors (e.g., elevation) shape patterns of gene flow in Australian Acacia species remains unclear. A recent meta-analysis of patterns of genetic diversity highlighted that Australian species generally follow global expectations when factors including range size, form, and abundance are considered (Broadhurst et al., 2017). This study also found that genetic diversity is lower in Australian shrubs (primarily acacias) when compared to trees or herbs and that population genetic structure (F st /G st ) in shrubs and trees was estimated to be twice that observed in global studies. While observations such as these are useful for high-level comparisons, understanding the major drivers of among-species variation in genetic diversity and structure may be more important for guiding conservation decisions.
Here, we compare genetic diversity, population genetic structure, and landscape genetics using AFLP data in two functionally similar shrubby legumes (Acacia salicina and Acacia stenophylla) to improve our understanding of the main environmental factors shaping genetic connectivity in these two species. Acacia salicina and Acacia stenophylla are both broadly distributed across the Murray Darling Basin (MDB) ( Figure 1) in eastern Australia, one of Australia's most large river system that has been extensively used for agricultural production (Cai & Cowan, 2008). Importantly, these two species however have partially contrasting life-history and environmental requirements. A. salicina is a perennial woody shrub that mainly occurs in semi-arid habitats and it is a very successful colonizer of degraded areas with high tolerance of bare soil (Grigg & Mulligan, 1999). This species has been introduced successfully in different parts of the world to revegetate degraded areas and restore soil conditions (Jeddi & Chaieb, 2010), and it is invasive in some arid areas of Israel (Jeddi & Chaieb, 2012). Although not much is known about seed dispersal mechanisms of A. salicina, some evidence suggests that birds can disperse their seeds (O'Dowd & Gill, 1986). A. stenophylla is a small woody shrub that mainly occurs in riparian ecosystems of Australian river dryland areas. This species provides nesting habitat for many birds in floodplains of inland Australia, and its main seed dispersal mechanism is through hydrochory (Murray et al., 2019). Thus, this species might have specific patterns of genetic structure and diversity modulated by downstream unidirectional gene flow through the MDB river system (Ritland, 1989).
Given their ecological and biological differences and to uncover the effects of landscape fragmentation and other environmental factors on genetic connectivity, we have formulated the following hypotheses: We hypothesize that these two species potentially have different gene flow connectivity patterns through the landscape.
We expect A. stenophylla to be more sensitive to habitat fragmentation, historical changes in water fluctuations of the MDB (Cai & Cowan, 2008), and being affected by hydrological connectivity directly shaping its genetic structure and diversity, while A. salicina seems more resilient and can quickly (re)colonize degraded areas (Grigg & Mulligan, 1999;Jeddi & Chaieb, 2012) and the river system might act as geographic barriers for gene flow. To test these hypotheses, we were interested in determining: (a) Did levels of genetic diversity differ between the two species? (b) Did population genetic structure differ between the two species? And 3. if differences between the two species were evident, could this be explained by environmental factors, such as elevation and/or habitat fragmentation?

| Site selection and collection of genetic material
Location data from herbarium specimen records of the Australian Virtual Herbarium were obtained to guide the selection of sites. A survey was then conducted to multiple agricultural areas of the MDB K E Y W O R D S Australia, connectivity, gene flow, habitat fragmentation, landscape genetics, population structure, resistance surfaces (New South Wales, Australia). We selected sites where the number of mature individuals of A. stenophylla and A. salicina exceeded 20-30 trees and distance between locations was greater than 30 km (Thrall et al., 2007). A total of 28 A. salicina and 30 A. stenophylla sampling locations were collected from across the MDB in southeastern Australia (Figure 1). 25 and 28 of those sampling locations of A. stenophylla and A. salicina; respectively, were located close to riverbanks or water streams. Phyllode material was collected from up to 30 trees in each sampling location, kept cool during transport to the laboratory, lyophilized for 2-5 days (Flexi-Dry MP FTS Systems, USA), and stored for DNA analysis.

| DNA extractions and AFLP genotyping
DNA was extracted from ~10 mg of dried tissue ground to a fine powder using 3-mm tungsten carbide beads in a Retsch MM300 mixer mill using the Qiagen 96-well DNEasy Extraction Kit (Qiagen, Melbourne) following the manufacturer's protocol. AFLP amplification largely followed that of (Vos et al., 1995) with the exceptions that 500ng of genomic DNA was digested for each sample using Amplicons were visualized on an ABI 3130XL sequencer using a LIZ 500-bp internal standard (Applied Biosystems) and scored using GeneMapper Version 4.0 software (Applied Biosystems). A binary matrix of present (1) and absent (0) bands was constructed for each species. Ninety-five samples from 3 to 4 populations from across the geographic range of both A. salicina and A. stenophylla were run twice for each of the proposed primer pairs to test for reproducibility across a range of 120 to 450 bp. Markers with an error rate of >5% were discarded with the error rate for markers selected for A. salicina ranging from 0-3.4 and for A. stenophylla being 1.5-4.3. A negative control was run with every set of samples in a 96-well block.

| Genetic data analyses
The binary data matrix of each species was used to estimate the per-   , 2006). We also checked for sample size effects for both species (see Figure S12), and we did not find major effects on genetic diversity (H e ); except for sampling location 48 (N = 7) of A. stenophylla, which showed significant differences of genetic diversity (H e ) (Wilcoxon test: W = 322,230, p < .05) between all sample sizes.
GenAlEx was also used for an analysis of molecular variance (AMOVA (Excoffier et al., 1992) The popgraph R package (Dyer, 2009) was used to create population graphs that described the distribution of genetic variation among sampling locations for each species. This graph-theory approach simultaneously identifies genetic covariance structures among subpopulations, does not assume a priori hierarchical or bifurcating statistical models of population arrangement, and is independent of evolutionary assumptions that aim to minimize Hardy-Weinberg and linkage disequilibrium within populations (Dyer & Nason, 2004). Populations are represented as nodes with node diameter representing the level of within-site heterozygosity, lines connecting nodes show populations that are not significantly genetically differentiated with line length representing among-site genetic variation (Dyer & Nason, 2004). Paths connecting populations were also examined for "extended edges," designating long-distance dispersal, and "compressed edges," indicating topological or ecological sources of vicariance, both of which were identified by chi-square tests at α = 0.05.

| Resistance surface analysis
To assess the effect of landscape structure on genetic differentiation, we estimated four explanatory variables for each species:  to the extent of the study region for both Acacia species and masked them using a shape file of Australia. We also standardized by reprojecting (EPSG:4326) and aggregating all rasters to similar size grid cells. Spatial data were prepared using the R packages raster (Hijmans & van Etten, 2012) and rgdal (Bivand, Keitt, & Rowlingson, 2020).
The "forest cover" was aggregated by a factor of 3, using the mean function, "land cover" was similarly aggregate by a factor of 30 and reclassified as outlined below. The DEM was rescaled (min 0.001max 1) and also aggregated by a factor of 3 using the mean function.
We used circuit theory (McRae et al., 2008) to estimate the resistance to gene flow between sampling locations of both species for each of the explanatory variables (forest cover, land cover, water stream cover, elevation) using Circuitscape v4.0 (McRae, 2006) to estimate pairwise resistance distances. Because we hypothesized higher gene flow across intact forest remnants than between regions predominantly covered by agricultural areas, we created resistance surfaces where agricultural area pixels had higher resistance values. We created two separate resistance surfaces: one using land cover and one using forest cover maps. We used raw values of forest cover rasters ( Figure S5) and transformed the categorical values of land cover rasters to numerical resistance values ranging between 0 and 1 ( Figure S4). More specifically, we assigned a minimal resistance of 0.1 to all forested land cover classes, medium resistance values (0.4-0.5) to areas containing fragmented habitats, and a maximal resistance of 0.9 to all other classes (agricultural and permanent snow areas). We also tested two more the hypotheses: (a) that elevation influenced genetic connectivity with mountain ranges being a potential barrier to Acacia gene flow and (b) that water streams might be positively affecting gene flow between Acacia populations (particularly A. stenophylla). To do this, we created resistance surfaces with pixels at higher elevations having higher resistance values using the raw elevations from the DEMs as resistance values for each pixel ( Figure S6) and we created a "water stream cover" conductance surfaces only taking major perennial watercourses with conductance values of 1 ( Figure S7). Finally, to test for isolation by geographic distance (IBD), we created null-model rasters by replacing all values of the forest cover rasters with 0.5 and calculated resistance distances between sampling locations. Because Circuitscape does not accept zero resistance values, we replaced zero values in all rasters with 0.0001.

| Landscape genetic analysis
We used conditional genetic distance (Dyer & Nason, 2004) as our response variable and it was calculated using the R package gstudio (Dyer, 2009). This is an interindividual genetic distance, which considers genetic covariation among all studied sampling locations.
As explanatory variables, we used the various distance matrices described above: geographic distance, forest cover, land cover, water stream cover, and elevation. We tested correlations among genetic and environmental distances according to the different scenarios considered using: (a) Mantel (Mantel, 1967) and partial Mantel tests (i.e., to control for spatial autocorrelation using the geographic distance matrix); and (b) multiple regression on distance matrices (MRM, (Lichstein, 2007)). Mantel  We corrected p-values obtained for Mantel tests and MRM models for multiple testing using the Benjamini and Hochberg (1995) method as implemented in the stats package in R. A. stenophylla: Φ PT = 0.235, 23.5%, p = .010). Significant isolation by distance was also detected in both species (p < .001).

| Genetic diversity and population genetic structure
The first two principal coordinates axes accounted for 56.9% of the total variation in A. salicina and 49.3% of the variation for A. stenophylla ( Figure S1). Divergent sampling locations were evident for both of these PCos, namely, A. salicina plants from sampling locations 13-16 were located in negative PCo1 space ( Figure S1a) and a group of A. stenophylla plants from sampling locations 37, 38, 42, 44, and 46 were in negative PCo1 and PCo2 space ( Figure S1c).
The analyses done by STRUCTURE showed that most likely there are six genetic clusters (K = 6) for both species (Figures 2 and   3) based on the statistic ∆K, although there was some evidence that a smaller number of clusters (i.e., 3) might also be present ( Figure S11a,b). There was little evidence of admixture in many of the A. salicina plants (Figure 2 and Figure S2). For example, some northern sampling locations (23, 24, 26, 27, and 29) were strongly associated with cluster 1 (dark blue); however, some plants in sampling location 24 had associations with cluster 4 (yellow) and 6 (red). In contrast, other A. salicina sampling locations (e.g., 2, 5, 6, 22, 25, and 28) showed evidence of admixture or potentially immigration from sampling locations belonging to other groups. The A.
salicina clusters were somewhat geographically partitioned across the study region (Figure 2 and Figure S9) with cluster 1 (dark blue) found to the northeast, cluster 2 (light blue) restricted to the most southerly edge while cluster 3 (purple) was broadly distributed.
Clusters 4 (yellow) and 5 (green) were distributed broadly in the southern half while the single cluster 6 (red) sampling location is found to the northeast. Two sampling locations (22 and 28) were a mixture of several clusters and could not be assigned to a single cluster at >70%. Results from K = 3, the second highest ∆K, show a clear geographic pattern of northern and southern clusters ( Figure S11a). Unlike A. salicina, many A. stenophylla sampling locations were dominated by cluster 1 (red, Figure 3d) and many sampling locations showed evidence of admixture or immigration ( Figure 3 and Figure S3). Many sampling locations could not also be assigned to a single cluster at >70%. These results are also clearly observed for K = 3 ( Figure S11b). Geographically the majority of sampling locations assigned to cluster 1 were located in headwaters with sampling locations not easily assigned or those belonging to other clusters located downstream.

| Landscape genetics analysis: Mantel tests and multiple regression on distance matrices (MRM)
We used resistance surfaces of each environmental variable calculated by Circuitscape for our landscape genetic analysis on each species. We did not find any significant relationship between genetic clusters and the different resistance surfaces (LC, FC, and DEM) for A. stenophylla and A. salicina, except for forest cover effects on genetic clusters of A. salicina (Kruskal-Wallis χ 2 = 12.657, p = .048).
More specifically, the admixed cluster (white) consisting of two sampling locations was mostly related to areas of high forest cover resistance ( Figure S7a).
Mantel tests showed that the genetic structure of A. salicina was affected by both geographic distance and land cover resistance (p < .05), while for A. stenophylla, only land cover resistance appeared to be important (Table 2). After correcting by spatial autocorrelation (partial Mantel tests), none of these factors had a significant effect on either of the Acacia species (Table 2). However, the MRM model considering only geographic distance was the best model explaining genetic variation between sampling locations in Acacia salicina whereas the MRM model including only land cover and elevation was the best model explaining sampling location genetic differentiation for Acacia stenophylla (Table 3, Table S1).

| D ISCUSS I ON
The MDB is Australia's longest river system (area of ~1.0 million square kilometers) and a vital economic resource for the agricultural industry (Cai & Cowan, 2008); however, it is also considered one of Australia's most impacted ecosystems (Cole et al., 2016).
Disturbed landscapes, such as the MDB, alter spatial structure and affect plant demography by decreasing population size and increasing population isolation due to geographical distance or barriers in the landscape (Kwak et al., 2009). While habitat fragmentation is expected to diminish gene flow between local populations, ultimately affecting patterns of genetic differentiation and viability (Ellstrand, 1992;Kwak et al., 2009;Young et al., 1996), some evidence suggests that a lack of intervening vegetation may However, STRUCTURE analysis might be overestimating genetic structure of A. salicina due to high levels of IBD (Frantz et al., 2009).
In contrast, A. stenophylla showed high levels of admixture in many sampling locations dominated by a single cluster (1, red) from six genetic clusters detected by STRUCTURE. This suggests that gene flow of A. stenophylla is relatively high across many northern and southern sampling locations within its geographical range. However, few sampling locations were composed of distinct genetic clusters [e.g., cluster 3 (green) only present in the south and cluster 4 (black) in the far southwest (see Figure 3). Our findings thus reveal that, contrary to A. salicina, A. stenophylla is able to maintain higher gene flow across large distances (at least 300 km). Although we did not

F I G U R E 4 Visualization of A. salicina Popgraph compressed (A) (red dashed lines) and extended (B) (black dashed lines) edges for and
A. stenophylla compressed (C) (red dashed lines) and extended (D) (black dashed lines) edges. Geographical location of A. stenophylla and A. salicina sampling locations colored to match K = 6. White nodes indicate < 70% assignment to a single cluster Our spatial analysis suggests that distinct environmental factors influence genetic differentiation in the two studied species.  (Grigg & Mulligan, 1999), is fairly resilient to fragmentation because it has large seed banks, high growth rate and tolerance to bare soil (Jeddi & Chaieb, 2012).
In the case of A. stenophylla, habitat fragmentation (predicted by land cover) did have a significant effect on genetic structure. Thus, genetic connectivity among A. stenophylla sampling locations does not seem to be mainly driven by geographic distance (contrary to A. salicina) and MRM showed that elevation might also act as a potential barrier for gene flow between sampling locations. Interestingly, given the high levels of admixture in many sampling locations of A.
stenophylla, the effects of land cover and elevation are likely to explain the presence of distinct genetic clusters in southern sampling locations (green, yellow, and light blue clusters; see Figure 4c,d) where there is high habitat fragmentation produced by extensive areas of agricultural land ( Figure S4). Despite these factors, genetic structure and diversity analysis shows that this species is largely unconstrained across its range with long-distance seed movement possibly helping to maintain homogeneity among sampling locations, especially within rivers and tributaries as it has been shown in a previous study (Murray et al., 2019). Although we did not find a significant effect of water stream cover, occasional one-directional long-distance seed dispersal may partly explain high levels of admixture in many sampling locations located at the river margins ( Figures 1 and 3).
Although high levels of gene flow have been found in the north- European settlement and, in recent times, are known to suffer from lower and more even flow volumes (Adamson et al., 2009;Cai & Cowan, 2008). This suggests that the population connectivity of A.
stenophylla may now be partially affected by severe water flow fluctuations and management, particular in the Lower Murray River, of the MDB (Oliver & Merrick, 2006).
Several studies have pointed out the negative impacts of habitat fragmentation on plant population viability and genetic diversity (Millar et al., 2014;Young et al., 1996) show that (a) both species had relatively high levels of genetic diversity and differentiation; (b) both species also had high levels of genetic structure across the MDB, although A. stenophylla also showed high admixture levels in several sampling locations; and (c) habitat fragmentation and elevation do not equally affect the genetic connectivity of these two woody legumes supporting our hypothesis.
While it seems that A. salicina genetic differentiation and connectivity are mainly driven by geographic distance, anthropogenic disturbances in the MDB do have an important impact on gene flow in A. stenophylla and it is likely that it affects other less resilient plant species in the region (for example, wetland specialists (Colloff et al., 2014)). Previous studies show that severe impact it is already occurring in freshwater fauna in the MDB (Chessman, 2011;Cole et al., 2016) augmented by the increasing effects of climate change (Adamson et al., 2009;Balcombe et al., 2011). We also suggest that this work could serve as a reference for studies aiming to assess the importance of their associated legume symbionts (nitrogen-fixing rhizobial bacteria) (Thrall et al., 2005(Thrall et al., , 2007 to understand how their composition and genetic variation across large geographic scales might be associated with the survival and reproduction of Acacia species.

ACK N OWLED G M ENTS
This research was jointly funded by CSIRO and the New South Wales Environment Trust. The authors would like to thank Jo Slattery and Jacqui McKinnon for field assistance and for bacterial isolation and purification and Michelle Watt and Michael Grossman for constructive comments on the original manuscript.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Genetic data of both Acacia species will be deposited at CSIRO Data Access Portal.