Apomixis goes a long way: Genetic evidence of persistence and long‐distance seed dispersal in an ancient landscape

Apomixis is a widespread trait in extreme environments worldwide, yet phylogeographical studies for species exhibiting these complex reproductive systems are still limited to temperate zones in the Northern Hemisphere. Through analyses of a combination of adult plants and seedlings, and nuclear and chloroplast DNA, we assessed the contemporary genetic outcomes of apomixis and phylogeographical patterns in an arid unglaciated landscape to understand the evolutionary trajectory of apomictic species.


| INTRODUC TI ON
The association between flowering plants and sexual reproduction is well established and yet despite the evolutionary advantages of sexual reproduction, most angiosperms are also capable of some form of asexual propagation.In the majority of cases, this remains an addition to sexual breeding but the balance between sexual and asexual reproduction varies widely across groups (Silvertown, 2008).Most asexual reproduction is via suckering or root propagation.However, in over 70 angiosperm families, facultative asexual reproduction via seed (apomixis) has become dominant through an (epi)genetically controlled modification of the ancestral sexual pathway, with an evolutionary transition away from obligate sexuality (Hojsgaard et al., 2014;Karbstein et al., 2021;Whitton et al., 2008).
Asexual propagation can provide an advantage in harsh environments, allowing well adapted genotypes to persist in refugial areas during adverse times while avoiding the negative genetic effects of sexuality, such as inbreeding, in retracted small populations.
Apomixis, defined as the asexual formation of a seed from the maternal tissues, combines the benefits of seed dispersal with those of asexual reproduction, allowing not only persistence but also the rapid expansion from refugial areas during favourable times (Mogie, 1992).
In some species, the parthenogenetic embryos can develop autonomously without pollination, although, most apomictic plants retain a facultative sexuality, and pollination and consequent fertilisation of the central nucleus may still be required to develop the endosperm (i.e.pseudogamy; Fei et al., 2019).Moreover, this life-history strategy is generally associated with polyembryony and polyploidy (Carman, 1997).Polyembryony allows the development of multiple embryos within a single seed (Mendes-Rodrigues & Oliveira, 2012; Nakano et al., 2012), while polyploidy allows the duplicated genes to contribute increased genetic variation for long-term diversification and evolutionary success under changing conditions (Hörandl & Hojsgaard, 2012).Indeed, having multiple genome copies might have manifold indirect positive effects on the establishment of apomictic individuals by altering many physiological features and providing greater phenotypic flexibility to respond to environmental stress, move into different niches, or rapidly colonise disturbed habitats (Hojsgaard & Hörandl, 2019;Karbstein et al., 2021;Molloy, 2019).
Despite these advantages, the complete loss of recombinant progeny is expected to facilitate Muller's rachet, the accumulation of recessive deleterious mutations in an irreversible manner over generations (Agrawal, 2006;Kimura & Maruyama, 1966).This is particularly problematic for polyploid organisms, where the mutation rate is multiplied by the ploidy level (Gerstein & Otto, 2009).
Although very large population sizes could theoretically be sufficient to sustain effective purifying selection in eukaryotic apomicts (Brandt et al., 2017;Hörandl et al., 2020), facultative sexuality with low frequencies of recombinants per generation is generally thought to be the mechanism that effectively exposes deleterious mutation to purging selection and may provide an escape from this seemingly irreversible process (Hodač et al., 2019;Hojsgaard & Hörandl, 2015).
Despite the deep interest in biogeographical and ecological aspects of apomixis that has emerged in the last 30 years, most of the early work tends to focus on cytological features of apomictic seed development and reproductive trials (e.g.Brock, 2004;Rambuda & Johnson, 2004).Much needed molecular studies at regional scales to understand the evolutionary and biogeographical history of apomictic species are emerging in recent years but are mainly limited to temperate zones in the Northern Hemisphere (Coughlan et al., 2017;Hörandl, 2023;Karbstein et al., 2021), while studies in more ancient, unglaciated landscapes are still lacking.
Molecular data can be useful for confirmation of apomixis, assessing the relative contributions of sexual and asexual reproduction, and understanding the genetic diversity and phylogeographical dynamics of apomictic species.Recent molecular investigations on polyploid apomictic plants have mainly concentrated on herbaceous plants (e.g.Cosendai et al., 2013;Paule et al., 2018;Voigt-Zielinski et al., 2012) and relatively few attempts have been made to evaluate the genetic structure and phylogeography of perennial woody plants (Kolarčik et al., 2023;Lo et al., 2009;Sochor et al., 2017), in which growth habit and longer generation times may influence the pattern of genetic variation.A predicted consequence of obligate asexuality is the formation of genetically uniform populations that are likely to become differentiated from each other by stochastic events (Dickinson & Campbell, 1991) and through natural selection over long timeframes.On the other hand, the residual sexuality in facultative apomictic species allows for backcrossing to progenitors and intercrossing of polyploids, resulting in large networks of hybridogenetic apomictic lineages (Karbstein et al., 2022), where intraindividual heterozygosity is also maintained at duplicated loci and can be as much as three times higher than in their diploid counterparts (Karbstein et al., 2021;Paun & Hörandl, 2006).Therefore, facultative apomixis, which combines the excess of heterozygotes and higher mutation rate of polyploidy with gene flow via seed dispersal, may counteract the theoretical loss of intrapopulation variation expected for obligate asexuals (Ellstrand & Roose, 1987;Hörandl & rather than an evolutionary 'dead-end', can be regarded as a mechanism facilitating evolutionary success of apomictic species in extreme and complex environments.

K E Y W O R D S
apomixis, arid landscape, asexual reproduction, chloroplast DNA, clonal diversity, genetic diversity, microsatellite, phylogeography, polyploidy, Senna glutinosa Paun, 2007).These processes point to the evolutionary potential of apomictic species and yet phylogeographical studies to infer complex biogeographical scenarios, such as past fragmentation, recolonisation, or identification of refugia over historical timescales (e.g.Byrne et al., 2017), are limited in apomictic species.These gaps in knowledge present a great opportunity for evolutionary studies of apomicts in unglaciated environments with long evolutionary histories, such as arid regions in northwestern and central Australia.
Senna is one of the largest genera in Caesalpinioideae (Fabaceae), with approximately 300 species (Lewis et al., 2005).The genus has a pantropical distribution but is mostly represented in the arid zones of Australia (Randell, 1989) and offers an excellent system to investigate the phylogeography, genetic structure and rate of facultative sexuality for apomictic species in ancient landscapes.Many species of the genus occur throughout the Pilbara bioregion (Randell & Barlow, 1998), an arid and topographically complex landscape that comprises some of the oldest landmasses on Earth (McKenzie et al., 2009;Pepper et al., 2013).Karyological studies have found that taxa endemic to arid zones showed the highest frequency of polyploidy within the genus (triploid and tetraploid, with the latter being the most common), as well as high levels of hybridisation and polyembryony, with up to nine asexual embryos recorded in one ovule (Holman & Playford, 2000;Randell, 1970Randell, , 1989)).Apomixis has also been demonstrated for some species using histological, reproductive or molecular methods (e.g.Holman & Playford, 2000;Randell, 1989;Resende et al., 2014) but it remains unclear to what extent this complex reproductive system can influence genetic variation at the population level and whether gene flow via seed dispersal affects genetic structuring of populations across the landscape.
Recent molecular studies in the Pilbara bioregion on widespread, sexual, diploid trees and shrubs have found relatively low genetic structure across their distributions (Byrne et al., 2017;Nistelberger et al., 2020).This is remarkable given the vast distances and complex environments that characterise the region and has been attributed to extensive seed dispersal facilitated by flooding from cyclone and storm events and possibly also through large nomadic animals.The apomictic nature of Senna provides an opportunity to test the hypothesis of high seed dispersal in this arid region subject to extreme meteorological phenomena.
To better understand the influence of apomixis on genetic vari-

| Study area and species
The Pilbara region is a topologically complex landscape characterised by rocky ranges, incised gorges, granitic and alluvial plains in the northwest of the vast arid zone of the Australian continent (McKenzie et al., 2009;Pepper et al., 2013).The dominant natural feature is the Hamersley Range, running east to west and bounded by the Fortescue River on the northern side (Figure 1).It comprises a series of ridges and an elevated plateau, with the highest peak at 1256 m above sea level.In contrast to the complexity of the Hamersley Range, the Chichester Range, located north of the Fortescue Valley, comprises gentle, lower elevation hills, with a maximum elevation of ~510 m above sea level.
Senna glutinosa subsp.glutinosa is widely distributed throughout the Pilbara bioregion and its distribution extends eastwards to central Australia (Randell, 1989).Across this distribution, large populations of S. glutinosa subsp.glutinosa are dominant in arid shrublands on flat areas with red stony soil, as well as on rocky hillsides.The species grows as an erect shrub up to 3 m tall with glabrous leaves.
Senna glutinosa is primarily tetraploid, with triploids of sporadic occurrence, being the result of breeding with rare diploids which only occur in few taxa (Randell, 1970).Morphological identification is notoriously complicated within the genus due to the frequent occurrence of hybridisation with co-occurring congeneric species.Across the Pilbara, S. glutinosa subsp.glutinosa can be identified by its nonglaucous appearance and the occurrence of viscid secretions on the leaves and seed pods, as opposed to the pruinose and non-viscid features of the leaves of its cohabitant, S. glutinosa subsp.pruinosa.
Crossings between the two subspecies (i.e. S. glutinosa subsp.× luerssenii) are readily distinguished from the target subspecies by exhibiting the non-glaucous coloration of S. glutinosa subsp.glutinosa but without the viscid secretions (Randell, 1989).The flowering period extends March to August (Randell & Barlow, 1998), and similar to most species in the genus, the subspecies is pollinated by bees (Randell, 1989).Seeds are dull and produced in flat pods, with no obvious morphological structure to enhance dispersal.

| DNA extractions and sequencing
For a broadscale assessment of spatial genetic variation using microsatellite markers, 20 populations were selected to cover the distributional range of S. glutinosa subsp.glutinosa across the Pilbara region.Initial sampling and genotyping was performed in 2016.Leaf material was collected from 24 adult plants per population, for a total of 480 samples.Collections targeted widely spaced plants that were sampled haphazardly across a 2500 m 2 area across each population to minimise the sampling of clones.To assess the extent of apomixis and polyembryony within populations, a second field expedition was undertaken at a separate time point to collect seed pods for a progeny study.Leaf material and five seed pods per mother plant were collected from five plants in a subset of four populations for microsatellite analysis.These populations were chosen based on having the lowest (EMU, FLA) and highest (GIL, ANG) genotypic diversity, as identified from the broadscale study (see Section 3).Up to six viable seeds per pod were germinated under controlled conditions for a total of 432 seeds, from which we obtained 516 seedlings.Cases of polyembryony were recorded on emergence and 96 seedlings per population were then retained for DNA extraction and genotyping.Finally, one representative for each multilocus genotype (MLG) identified in the broadscale study (See Section 3) were selected for chloroplast DNA analysis across all 20 populations.Genomic DNA was extracted from freeze-dried material using a modified 2% CTAB protocol, with 1% polyvinylpyrrolidone and 0.1% sodium sulphite added to the extraction buffer (Byrne et al., 2001).
Samples for the broadscale microsatellite study were genotyped using a set of 17 loci specifically developed for the taxon (see Appendix S1) following the PCR protocols detailed in Nistelberger et al. (2020) and scored using GeneMapper v5 (Applied Biosystems).
The use of these traditional genetic markers was dictated by the time period in which the data were collected (2016).Nonetheless, their codominant nature, high polymorphism, and ease of multi-allelic interpretation makes microsatellite markers well suited for genetic studies of polyploid taxa.Adult leaves and seedlings for the progeny study were genotyped using the 10 most informative of the 17 microsatellite markers (see Appendix S1).The 10 microsatellite loci were chosen based on two major factors: (1) being polymorphic and (2) displaying stutter patterns that were straightforward to interpret and score consistently.The latter point is very important when scoring polyploid data because the increased number of alleles can make even simple stutter patterns difficult to score consistently.
Finally, a range of chloroplast regions were trialled to identify those with informative intraspecific variation for the study species, as per Byrne and Hankinson (2012).The atpF, petD and rpl16 introns (these sequence data have been submitted to the GenBank database under accession numbers in Appendix S2) were subsequently selected for sequencing based on sequence quality and nucleotide diversity.Following amplification and thermocycling according to Byrne and Hankinson (2012), a Serapure method was used to purify PCR products (Faircloth & Glenn, 2012).Purified PCR products were then sequenced at the Australian Genome Research Facility (Perth, Australia).DNA sequence chromatograms were assessed in Sequencher v5.0 (Genecodes corp., MI, USA) and the resulting sequence data were aligned using AliView (Larsson, 2014).Indels arising from mononucleotide repeats were removed as these are often the result of sequencing error and are difficult to accurately score (Clarke et al., 2001).Subsequent analyses were conducted on data in which indels were manually encoded as binary characters.

| Broadscale microsatellite study
A high degree of asexual reproduction in S. glutinosa subsp.
glutinosa and its tetraploid status was apparent in scoring the electropherograms of the 480 adult plants sampled across the Pilbara.Because individuals generated by apomictic seeds are effectively clones of their mother plants, a certain number of markers is needed to provide enough power to discriminate all the MLGs present in the sample.Such number of markers depends on the genotypic resolution provided by the set of available loci and can be determined by Monte-Carlo procedures (Arnaud-Haond et al., 2007).Here, this genotype accumulation curve was calculated using the r package 'poppr' v2.9.3 (Kamvar et al., 2014).Genotypic variation in clonal species can reflect both variation among different sexually generated clones or variants within a single genetic lineage that have arisen through somatic mutation.The latter are expected to be differentiated from their clonal ancestor by only a few mutations and can be collapsed into the same multilocus lineage (MLL; Arnaud-Haond et al., 2007).To estimate the number of different MLLs present within each population, clone assignment was performed with the software genodive v3.06 (Meirmans, 2020) using the stepwise mutation model and a genetic distance of 6 as a threshold to allow for somaclonal mutation and genotyping errors.This threshold was determined by following the common approach of visual assessment of the bimodal genetic distance distribution (e.g.Roberts et al., 2017).In several analyses, the software genodive v3.06 allows for the correction of the unknown dosage of alleles (up to octoploid), which is a challenge for polyploid genetic data because it is often not possible to obtain the dosage from the marker phenotypes.Briefly, this correction uses a maximum likelihood method based on random mating within populations based on De Silva et al. (2005) in which the correction algorithm is run separately for every locus and every population until convergence is reached (Meirmans, 2020).Locus summary statistics, including allele frequencies and missing data, were calculated with genodive using this maximum likelihood method to correct for the unknown dosage of alleles.The same software was used to calculate clonal diversity indices, such as number of effective genotypes, Nei (1987) unbiased gene diversity, and evenness.Pairwise genetic fixation among populations was calculated using the ρ statistic (Ronfort et al., 1998), an F ST analogue that is independent of the ploidy level and the rate of double reduction, and is recommended for use in polyploid species (Meirmans et al., 2018).Pairwise differentiation between populations was calculated using Jost's D (Jost, 2008), which is also ploidy independent.Analysis of molecular variance AMOVA was calculated using the ploidy independent infinite allele model in genodive.
The distribution of alleles across the sampled region was assessed with a Mantel test (30,000 permutations) for isolation by distance (IBD), using transformed pairwise ρ values (ρ/(1ρ)) for genetic distance.To further clarify the role of landscape structure on the genetic variation of S. glutinosa subsp.glutinosa, we performed the Moran's eigenvector map analysis (MEM; Legendre et al., 2015) based on genotypes and spatial coordinates implemented in the R package 'memgene' v1.0.2 (Galpern et al., 2014).This analysis adopts a regression framework where the predictors are generated using Moran's eigenvectors maps, a multivariate technique developed for spatial ecological analyses and is recommended for genetic applications (Epperson et al., 2010;Jombart et al., 2009;Manel et al., 2012;Wagner & Fortin, 2013).Results from memgene were then superimposed on a raster map (http:// pid.geosc ience.gov.au/ datas et/ ga/ 72759 ) to identify spatial genetic neighbourhoods using the r package 'raster' v3.5 (Hijmans & Van Etten, 2021).
Clustering analysis was performed through discriminant analysis of principal components (DAPC) using the r package 'adegenet' v2.1.5(Jombart, 2008).This multivariate method does not focus only on the global diversity but maximises the variance between inferred groups while minimising within-group variation.Moreover, since it is based on a multi-allelic model of within-individual allele frequencies, it can be easily applied to polyploids (Meirmans et al., 2018).DAPC requires the a priori selection of both the number of retained principal components (PCs) and the number of discriminant functions.

| Broadscale chloroplast DNA study
The number of haplotypes and haplotype diversity were obtained using the R package 'pegas' v1.1 (Paradis et al., 2021).The same package was used to calculate Tajima's D (Tajima, 1989) and Ramos-Onsins and Rozas (2002) R 2 , while the software DnaSp v6 (Rozas et al., 2017) was used to calculate Fu's Fs (Fu, 1997).These tests detect departure from the neutral theory of molecular evolution and also predict population size changes under Wright Fisher assumptions of panmixia and constant population size.Significance was estimated using 10,000 replicates of coalescent simulations under the standard neutral model.Arlequin v3.5.2.2 (Excoffier & Lischer, 2010) was used for testing spatial expansion using Goodness of fit test (Harpending's raggedness index H rag ) and the sum of squared deviations (SSD) to determine significance.
The software SPADS v1.0 (Dellicour & Mardulyn, 2014) was then used to assess the presence of phylogeographical structure, where closely related haplotypes occur together in space.This tests whether the measure of genetic differentiation N ST , which considers not just differences in haplotype frequencies but also genetic distances, was significantly larger than G ST through a U test with 10,000 permutations (Pons & Petit, 1996).Lastly, a network of haplotype relationship using the median-joining method (Bandelt et al., 1999) was generated in NETWORK v 5.0.1.1 (fluxu s-engin eering.com).

| Progeny study
Assignment of seedlings as apomictic clones, selfed, or outcrossed progeny was performed with COLONY 2.0 (Jones & Wang, 2010) on the dataset without maximum likelihood impuation of allele dosage.
This software package can be used on polyploid species with a slight modification of the data: transformation of the polyploid codominant genotypes to pseudodiploid-dominant genotypes (Wang & Scribner, 2014).The input data included the genotypes of all sampled adult plants as candidate parents and the 384 seedlings.
COLONY was run allowing for male and female polygamy, assuming inbreeding and clones, and set for a long run and pair-likelihood score with default parameters.Because selfed progeny can potentially produce identical microsatellite patterns to clonal progeny, we opted for a conservative approach in which progeny with identical genotypes were considered clones, genotypes with missing alleles but otherwise identical to the mother plant were considered selfed, and genotypes with alleles not present in the mother plant were outcrossed.

| Broadscale microsatellite study
All the investigated loci were polymorphic, ranging from two to 17 alleles per locus with an average of 8.353 (±1.216) alleles.The average number of alleles per locus varied among populations, ranging from 2.059 in the KAR population to 5.353 in ANG, and 70% of populations had at least one private allele (Table 1).Nei's (1987) unbiased gene diversity, more commonly referred to as expected heterozygosity (H S ), ranged from 0.361 to 0.520 and was generally lower than the observed heterozygosity (H O ), which ranged between 0.380 and 0.582 (Table 1).This excess of heterozygotes was common across the study area, with 13 populations showing G is values that were significantly less than zero.
We found a total of 76 MLGs that condensed into 72 MLLs across the 480 samples, of which 25 were sampled only once.The genotype accumulation curve showed that values approaching this plateau would have been reached using nine markers from our set of 17 (see Appendix S3).The number of genets in a population ranged between one (KAR) and seven (ANG and HAM), with five MLLs shared among different populations (Table 1; Figure 1).KAR was the only monoclonal population, resulting in Simpson's diversity and evenness indices of 0.000 and 1.000, respectively.Across the remaining populations, the Simpson's index (i.e.estimation of the probability that two randomly selected genotypes are different) was the highest in HAM (0.819) and the lowest in OXE (0.083); and the most even distribution of genets occurred in EMU (0.852), while WRA had the lowest evenness index of 0.313 (see Appendix S4).
At the MLL level, estimates of pairwise population genetic fixation were moderate across populations, with an average ρ value of 0.160 (±0.016), while the differentiation index D was considerably lower, averaging 0.078 (±0.024).The monoclonal KAR was the most differentiated population, having the highest average ρ and D values of 0.349 (±0.019) and 0.456 (±0.037), respectively (Table 1).Nuclear genetic variation was partitioned more among (0.588%) than within (0.412%) populations.
There was no signal of isolation by distance with an R 2 value of 0.010 and a non-significant Mantel test (Spearman's r = 0.102; p = 0.159).However, a small but important proportion (28%) of the total genetic variation in S. glutinosa subsp.glutinosa was attributed to landscape structure (MEMGENE R 2 adj = 0.280) and explained greater genetic patterns than straight-line distances.The first MEMGENE variable, explaining 43.5% of the variation, showed a spatial genetic neighbourhood mainly centred on the Hamersley Range (Figure 2a).
The second MEMGENE variable (explaining 19.0% of the variation) showed an additional neighbourhood among the eastern-most populations (Figure 2b).
Consistent with the spatial neighbourhood analysis, populations with shared MLLs largely occurred in the central region of the sampling distribution (Figure 1).Nonetheless, we found MLLs shared between populations as far as ~250 km apart (i.e.populations MET and HES).Distinct MLLs from the same population did not consistently group together and no geographically structured genetic clustering was detected in the DAPC analysis of the 72 MLLs (Figure 3).The first axis mostly separated population KAR from the rest of the populations, but also populations WAB and SNA to a lesser extent.The second and third axes showed additional variation among populations but with no geographical pattern.The DAPC retained 55 PCs and three discriminant functions following cross-validation.

| Broadscale chloroplast study
The three cpDNA regions were concatenated to form a total sequence length of 945 bp, consisting of 11 transversions, six transitions, and three varying length indels.Thirteen haplotypes were identified in total, with the most common haplotype (60%, H2) present in all populations except KAR, and two other haplotypes (H1: 10% and H5: 12.5%) present across several populations.However, most haplotypes were unique to specific populations in low frequency (1.25%, H3, 7, 9-13) (Figure 4).All but four populations consisted of multiple haplotypes.Of these four populations showing no haplotype variability, FLA, OXE, and SNA displayed the common H2, whereas the peripheral population KAR presented the highly divergent haplotype H9.The majority of genetic variation was maintained within populations (78.39%) rather than among populations (21.61%).There was no evidence of departure from neutrality from TA B L E 1 Genetic diversity parameters based on microsatellite data from adult Senna glutinosa subsp.glutinosa for 20 populations sampled across the Pilbara bioregion.Tajima's D and Fu's Fs (Table 2).Ramos-Onsins and Rozas' ( 2002) R 2 statistics also showed no support for a model of population growth (Table 2).Nonetheless, the model of spatial expansion could not be rejected based on both the SSD between the observed and the expected mismatches, and the Goodness of fit test based on Harpending's raggedness index (Table 2).

| Progeny study
Overall, multiple seedlings (up to three) emerged from 81 out of the 432 seeds.The average across populations was 18.69% (±3.56), with ANG having the highest rate of polyembryony per plant (26.23% ± 6.70) and GIL the lowest (13.23% ± 2.06) (Table 3).Of these polyembryonic seedlings, the ratio of apomictic versus sexuals emerging from a single seed was heavily skewed towards apomixis in all the tested populations (Table 3).
Of the 20 mothers across the seedling study, 19 exhibited MLLs that were identified in the broadscale study, but one mother in the GIL population exhibited a new MLG not previously identified.Within populations, those with the lowest genotypic diversity (EMU and FLA) also presented the lowest number of genets across the five mothers (1 and 2 genets, respectively).Mothers sampled from the more diverse populations (ANG and GIL) had three and four genets represented among the five mothers, respectively.Apomictic progeny represented 84.56% (±3.50) of the genotyped seedlings and outcrossed offspring accounted for just 3.22% (±1.39), with the remaining being the outcome of selffertilisation within or between ramets.At the population level, GIL produced the most outcrossed progeny (9.54% ± 4.43) while the other three populations averaged very low levels of outcrossing (1.12% ± 0.61) (Table 3).

| DISCUSS ION
Our study shows the influence of extensive apomixis on the genetic diversity and structure of a woody perennial in an arid landscape.
Genetic and reproductive evidence for apomixis in S. glutinosa subsp.glutinosa showed frequent polyembryony and a high proportion of asexual offspring, resulting in low within-population genotypic diversity and shared MLLs among populations via seed dispersal across the landscape.Polyploidy and the high level of apomixis can explain the significant excess of heterozygotes found in most populations, as well as the moderate level of genetic fixation found in the broadscale nuclear genetic analysis.On the other hand, gene flow via seed counteracted fixation, as shown by the low level of population differentiation.A small proportion of the total genetic variation was linked to landscape features rather than straightline geographical distance but there was no evidence of genetic structuring.The species showed patterns of haplotype diversity that are indicative of long-term persistence, with widespread distribution of haplotypes indicating extensive seed dispersal throughout the Pilbara bioregion.

| Reproduction
The results of this study indicate a strongly apomictic reproductive strategy for S. glutinosa subsp.glutinosa.Apomictic seed production in Senna requires pollination and subsequent fertilisation of the central nucleus (i.e.pseudogamy).Indeed, the development of apomictic embryos starts about the time of fertilisation, so that they develop in phase with the endosperm but before the sexual embryo (Randell, 1970).The high proportion of apomictic to sexual embryos identified in our study is consistent with the suggestion that early initiation of asexual embryo  (Kearney, 2003).This strategy, coupled with polyploidy, is beneficial for persistence and/or establishment in challenging environments because it conserves well-adapted, highly heterozygous genotypes and can provide important benefits such as heterotic effects and buffering from recessive deleterious mutations (Hojsgaard & Hörandl, 2019;Hörandl & Paun, 2007;Karbstein et al., 2021).
Nonetheless, the facultative nature of the study species' apomictic pathway ensures that sex is not entirely lost; the 15.44% production of recombinant progeny recorded in our study likely provides the capacity to escape the theoretical irreversible accumulation of mutations over time (Hörandl, 2009;Karbstein et al., 2021;Kolarčik et al., 2023).Hodač et al. (2019) showed that approximately 6% of recombinants per generation sufficed to prevent the accumulation of deleterious mutations in a hexaploid apomictic species.Indeed, the purging selection of deleterious mutations in polyploid pseudogamous apomictic plants acts efficiently at both the gametophytic and sporophytic level (Hojsgaard & Hörandl, 2015).Even with the low levels of outcrossing detected in our study, our results are consistent with the theory that deleterious mutations can also be efficiently exposed to selection after 'cross'-fertilisation between clone-mates (Hodač et al., 2019), providing an effective way to escape Muller's ratchet for S. glutinosa subsp.glutinosa.Additionally, there is the possibility that some of the seeds were of hybrid origin.Hybridisation of S. glutinosa subsp.glutinosa with co-occurring species (Randell, 1989) also likely provides similar benefits in purging/gaining genetic diversity via guaranteed outcrossing and this may be the selective drive behind the extensive hybridisation networks that characterise apomictic Senna species more generally (Hojsgaard & Hörandl, 2019;Randell, 1970).

| Genetic outcomes of apomixis
Apomictic populations generally harbour genetic diversity at two main levels: high heterozygosity within individuals, and multilocus genotypic diversity (i.e.presence of distinct individual genotypes within and among populations).The high observed heterozygosity found in this study points toward the allopolyploid origin of this taxon (Hörandl & Paun, 2007), which is feasible given its known propensity for hybridisation (Randell, 1989).Indeed, elevated heterozygosity in apomicts is traditionally attributed to allopolyploidy that can result in 'fixed' heterozygosity at duplicated loci.
Moreover, the level of heterozygosity may be increased by the higher mutation rate of polyploids, especially at hypervariable microsatellite loci (Paun et al., 2006;Paun & Hörandl, 2006).This also explains the negative inbreeding coefficients recorded in most of the studied populations.
The occasional sexual events recorded in S. glutinosa subsp.
glutinosa provide a means whereby this allelic diversity may be released to create a broad array of new genotypes in the offspring over time.The overall moderate to low clonal evenness found within the majority of the studied populations is consistent with the typical pattern of one predominant clone and several, less frequent genotypes that has been found in other apomictic species (Hörandl & Paun, 2007).Our findings also corroborate the general pattern in which the genotypic diversity of apomictic plants is partitioned more among than within populations.While apomictic species generally have almost all genotypes restricted to a single site (Houliston & Chapman, 2004;Paun et al., 2006;Robertson et al., 2004), our results showed the presence of wide-ranging MLLs shared across several populations, suggesting extensive seed dispersal in the Pilbara.Accordingly, our DAPC analysis and the low levels of differentiation among populations highlighted a lack of spatial genetic clustering, suggesting high levels of admixture among most populations across this ancient landscape.This demonstrates the importance of the local context in predicting genetic outcomes from apomixis in long-lived woody perennials.b Percentage of polyembryony per plant within population.
c Ratio of apomictic versus sexual seedlings per population when multiple seedlings emerged from a single seed.
d No sexual seedlings recorded.

| Genetic connectivity
Generally in angiosperms, nuclear DNA variation is the result of both pollen and seed dispersal, whereas cpDNA, when maternally inherited, reflects only seed movements (McCauley, 1995).However, in apomicts the influence of pollen dispersal on variation in the nuclear genome is minimal because the majority of seeds are effectively clones of their mother plants.Furthermore, the slowly evolving chloroplast genome is reflective of historical patterns of seed dispersal, while rapidly mutating microsatellite loci inform more recent patterns of seed movement.Such a reproductive system in our study species, combined with the use of multiple markers, has allowed unique insight into patterns of seed dispersal across the Pilbara bioregion.Although the presence of divergent haplotypes suggests that some populations of this species have been isolated at some point in time (see section below), the general lack of phylogeographical structure observed in our study suggests subsequent dispersal of diverged lineages.
The presence of common and widespread haplotypes, as well as the sharing of divergent haplotypes in populations at the opposite ends of the study area, indicates long-distance seed dispersal and suggests that populations were predominantly not established by single individuals, but rather by several ones from different genetic lineages.
Widespread seed dispersal is also supported by our broadscale microsatellite analyses, suggesting a high degree of recent connectivity among populations of S. glutinosa subsp.glutinosa.Seed dispersal has been generally found to be highly localised in shrubs and small stature trees with no adaptations for dispersal (e.g.Delnevo et al., 2021;Millar et al., 2022).Moreover, mountainous ranges are often identified as barriers to seed movement (Schurr et al., 2009).Yet, despite a genetic neighbourhood being identified in the Hamersley Range, our findings showed MLLs shared between populations as far as ~250 km apart, supporting the extensive seed dispersal inferred across the region for other widespread species (Byrne et al., 2017;Levy et al., 2016;Nistelberger et al., 2020).Extensive flooding and high wind from intense cyclonic activity during the wet season (Nathan et al., 2008) and the presence of large nomadic animals, such as emus, which are known to ingest Senna seeds (Calviño-Cancela et al., 2006), can explain these findings that contrast with general expectations of low seed dispersal.
Comparable long-distance dispersal may apply to other environments with similar meteorological or biological elements.
It is interesting to note that the most common haplotype was widespread across all populations (except KAR), which contrasts somewhat with the most common MLLs that were dispersed among distant populations but were generally shared at a more local scale.This suggests that the process to establish new lineages in new populations may take a long time and be hindered by the apomictic nature of the species (Lo et al., 2009) or be under frequency-dependent selection for the most fitting genotypes (Van Dijk et al., 2003).Indeed, this may reflect the time taken to widely establish a new genotype in an area when competing with established apomictic lineages, or simply that the new genotypes may be less fit for the new micro-habitat conditions.The actual frequency of recruitment of novel genotypes in apomictic species and the dynamics of genetic diversity in a temporal or spatial context are still not well understood.

| Historical persistence
Our analysis of chloroplast variation in S. glutinosa subsp.glutinosa revealed numerous haplotypes and high haplotype diversity.
Theory predicts that historical persistence of populations of relatively equal size that are approaching mutation-drift equilibrium will result in the accumulation of genetic variation (Nei, 1987).
Here, our results suggest historical persistence in localised populations throughout the Pilbara bioregion.This is in line with broad patterns in flora from other ancient landscapes such as the South Western Australian floristic region and the Greater Cape region (Byrne, 2008;Cowling & Lombard, 2002), and the Californian floristic province (Sork et al., 2016), where haplotype diversity of widespread plants typically far exceed that encountered in European or eastern American species.This is likely due to the topographic complexity and/or lack of glaciation events within these ancient landscapes, which have favoured low extinction rates (Lancaster & Kay, 2013) and, in turn, have supported population persistence and preserved genetic diversity in situ over long time scales.
Similar levels of haplotype diversity and divergence were found in the uplands and in the areas surrounding the ranges, despite the contrast in topographic complexity between the lowlands and the Hamersley and Chichester ranges.This further suggests that S. glutinosa subsp.glutinosa has persisted across these various habitats during periods of climatic change rather than contracting to refugial areas during the repeated cycles of aridification that characterised the Plio/Pleistocene mid latitude continental regions, where the ranges have provided some refugial buffer for other taxa (Byrne et al., 2017;Pepper et al., 2011).
The ability of this species to persist in different niches during unfavourable times may be the result of its allopolyploid status, as allopolyploidy could indirectly help to establish highly heterozygous apomictic individuals in a novel ecological niche by increasing adaptive potential (Hodač et al., 2023;Hojsgaard & Hörandl, 2019).Our findings support the notion that biotic responses to changing climatic conditions in old unglaciated landscapes may be largely idiosyncratic and species specific (e.g.Byrne et al., 2017;Levy et al., 2016;Millar et al., 2022;Nistelberger et al., 2020).Despite the general indications of broad seed dispersal, the high divergence observed among some haplotypes is indicative of historical isolation of some populations that has allowed drift of genetic lineages.This is best exemplified by the coastal population KAR, which appeared to be a monoclonal population that is highly divergent in both the nuclear and chloroplast genomes.
ation and species persistence in arid environments, we assessed nuclear microsatellite markers and chloroplast DNA sequences from adult plants and seedlings of Senna glutinosa (DC) Randell subsp.glutinosa in the ancient Pilbara bioregion of Western Australia.Specifically, we aimed to: (a) assess contemporary genotypic diversity using microsatellite markers to determine the broad extent of asexuality both within and among populations across the Pilbara region; (b) undertake a progeny study in a subset of populations to assess polyembryony and estimate the relative proportion of recombinant versus clonal progeny among seeds produced; and (c) use chloroplast sequence data to identify the phylogeographical patterns and contribution of historical seed dispersal to distribution of apomictic lineages across the landscape.

F
Spatial distribution of multilocus microsatellite genotypes within adult Senna glutinosa subsp.glutinosa across the sampled Pilbara range.The number in parentheses indicates the number of multilocus lineages within each population (out of 24 sampled plants) and the pie graphs indicate the frequency of each genotype.Coloured genotypes indicate those that are shared among populations, whilst uncoloured genotypes are restricted to single populations.The grey background represents the topography.
073 [−0.142 to 0.035] Note: 95% confidence intervals for H O , H S , and G is are given within square brackets.Bold values of G is indicate <1% significant deviation from Hardy-Weinberg equilibrium values, in which positive and negative values indicate deficit and excess of heterozygotes, respectively.N: number of sampled plants; N G : number of multilocus lineages; Na: average number of alleles; PA: number of private alleles; H O : observed heterozygosity; H S : gene diversity; ρ: mean pairwise Rho value (±standard error); D: mean pairwise Jost's D value (±standard error); G is : inbreeding coefficient.

F
I G U R E 2 MEMGENE analysis of 480 adult samples of Senna glutinosa subsp.glutinosa collected across the Pilbara region.Interpretation of the graph is similar to that of a spatial Principal Component Analysis (sPCA), where smaller circles represent data closer to the origin of the PC axes and larger circles data further away from the origin.Here, circles of a similar size and colour indicate clusters of individuals with similar scores, where large black and large white circles describe opposite extremes on the MEMGENE axes.Insets represent the values of the MEMGENE axes.(a) MEMGENE variable 1, explaining 43.5% of the variation.(b) MEMGENE variable 2, explaining 19.0% of the variation.Colour scale on the background represents elevation in metres.F I G U R E 3 Discriminant analysis of principal components of microsatellite data from 20 populations of Senna glutinosa subsp.glutinosa.Fifty-five PCs and three discriminant functions were retained following cross-validation.Insets show the relative importance of each component.Labels have been included only for the three most diverged populations.development provides apomictic embryos an advantage in the competition for space and endosperm against the embryo developed from the zygote.Such extensive asexual reproduction is also consistent with the trend of 'loss of sex' recorded across plants and animals in the Australian arid zone

F
I G U R E 4 (a) Network of cpDNA haplotypes identified across the 81 MLGs of Senna glutinosa subsp.glutinosa with frequency indicated by circle size and black dots representing mutational steps.(b) A map of the 20 populations sampled across the Pilbara, with pie charts indicating frequency of haplotypes at each population.The number in parentheses indicates the sample size for each population.The grey background represents the topography.
TA B L E 2 Chloroplast DNA diversity statistics, tests for neutrality and spatial expansion based on three sequenced regions (atpF, petD and rpl16 introns) from 20 populations of Senna glutinosa subsp.glutinosa sampled across the Pilbara bioregion.Hd, haplotype diversity; π, nucleotide diversity; R 2 , Ramos-Onsins & Rozas neutrality index; ns, not significant.Record of polyembryony and percentage of seeds produced via apomixis, selfing or outcrossing in four populations of Senna glutinosa subsp.glutinosa in the Pilbara bioregion.
a Number of polyembryonic seeds per population.