The nuanced nature of mesic refugia in arid landscapes: a tale of two peas

Abstract

Background and Aims

Understanding how genetic diversity is distributed and maintained within species is a central tenet of evolutionary and conservation biology, yet is understudied in arid regions of the globe. In temperate, glaciated environments, high genetic diversity in plant species is frequently found in refugial areas, which are often associated with southern non-glaciated landscapes. In arid, unglaciated environments, landscape features providing mesic conditions are likely to be refugia, although our understanding needs more refinement in these biomes. We test whether refugia and nuclear diversity hotspots occur in high-elevation, topographically complex areas for co-distributed shrubs (Petalostylis labicheoides and Indigofera monophylla; Fabaceae) in the ancient, arid Pilbara bioregion of north-western Australia.

Methods

We conducted extensive sampling of the Pilbara (>1400 individuals from 62 widespread populations) to detect patterns in nuclear diversity and structure based on 13–16 microsatellite loci. Evidence of historical refugia was investigated based on patterns of diversity in three non-coding chloroplast (cp) sequence regions for approx. 240 individuals per species. Haplotype relationships were defined with median-joining networks and maximum likelihood phylogenetic trees.

Key Results

We found cpDNA evidence for a high-elevation refugium in P. labicheoides but not for I. monophylla that instead exhibited extraordinary haplotype diversity and evidence for persistence across a widespread area. Nuclear diversity hotspots occurred in, but were not exclusive to, high-elevation locations and extended to adjacent, low-elevation riparian areas in both species.

Conclusions

Phylogeographic refugia in arid environments may occur in high-elevation areas for some species but not all, and may be influenced by species-specific traits: a mesic montane refugium in P. labicheoides could be related to its preference for growth in water-gaining areas, while a lack of such evidence in I. monophylla could be related to maintenance of cpDNA diversity in a large soil seed bank and dynamic evolutionary history. Mesic environments created by the intersection of topographically complex landscapes with riparian zones can be contemporary reservoirs of genetic diversity in arid landscapes.

INTRODUCTION

Understanding the distribution and maintenance of genetic diversity within species is a central tenet of evolutionary, ecological and conservation biology (O’Brien et al., 2022). More specifically, phylogeographic and landscape genetic approaches have been utilized globally to understand both the refugial evolutionary history of plant species and important associations between genetic diversity and landscape features. This research has had a bias towards temperate biomes and glaciated environments across the Northern Hemisphere (Hewitt, 2004; Keppel et al., 2012; Manel and Holderegger, 2013). Arid and semi-arid landscapes are understudied in this context despite such environments harbouring high levels of biodiversity globally in the form of species richness, endemism and genetic diversity (Byrne et al., 2008; Durant et al., 2012; Maestre et al., 2012; Pepper and Keogh, 2021). Identifying locations of mesic refugia and genetic diversity hotspots in arid biomes will shed further light on how diversity has been distributed and maintained in these often-harsh landscapes, which may differ from expectations for temperate and tropical regions. This information will better inform conservation efforts for arid taxa, because refugia and/or centres of diversity represent areas for contemporary conservation priority (Keppel et al., 2012). Moreover, given that many environments are expected to become progressively more arid under ongoing climate change (Huang et al., 2015; Berdugo et al., 2020), it is increasingly relevant to improve our understanding of the distribution of genetic diversity in arid environments to predict future changes and delineate priority conservation areas and actions.

Research in unglaciated regions of the globe, such as parts of Australia, has frequently shown species-specific patterns of plant phylogeographic history with evidence of multiple, relatively localized range contractions and expansions (Byrne, 2008), in contrast to obvious major refugia with clear post-glacial expansion routes identified in the temperate Northern Hemisphere (Hewitt, 2000). Localized population persistence has also been observed in semi-arid parts of South Africa (Potts et al., 2013). Similarly, in unglaciated parts of South America, some plant lineages show evidence of localized contraction and persistence in multiple, patchy refugia (Bonatelli et al., 2014; Ossa et al., 2017), while others show larger scale contraction and post-glacial expansion (Ossa, 2019). The biogeographical history of large unglaciated regions can therefore be variable and not readily predicted, especially in arid and semi-arid regions that have been less well studied (Beheregaray, 2008). General theory suggests that landscape features such as high-elevation areas could have served as mesic refugia in times of heightened aridity during periods of climatic oscillations, due to their topographic complexity and the availability of diverse microclimates and edaphic conditions (Byrne et al., 2008; Pepper and Keogh, 2021). It has also been suggested that moisture availability could be a key component of refugia in arid biomes, in conjunction with elevation and topographical complexity, such that other features at lower elevation may also act as refugia (Byrne et al., 2017). Further evaluation of the characteristics of refugia in arid areas is warranted to explore these concepts. Analysis of phylogeographic signals in species that occur both on and off mountain ranges in arid and/or unglaciated biomes can reveal evidence of historical refugia in such areas. Landscape features associated with refugia may also maintain populations with high nuclear diversity, potentially representing contemporary genetic diversity hotspots and high priority zones for conservation (Selwood and Zimmer, 2020).

The Pilbara biogeographic region is an ancient, unglaciated, arid region of exceptionally high biodiversity and endemism in north-western Australia (Pepper et al., 2013; Mishler et al., 2014). The region is defined geologically by the Pilbara craton and characterizsed by topographic, edaphic and climatic heterogeneity (Pepper et al., 2013). Landscapes range from coastal sandplains and islands to high-elevation rocky ranges and gorges, while geological substrates and associated soil types also vary markedly. Extreme seasonal weather events such as cyclones and regular widespread fire also impact the region. These landscape, geological and climatic factors are expected to have had varying effects on the distribution of genetic diversity and species’ evolutionary history, but relatively little is known of plant evolutionary history in this region or the distribution of intraspecific genetic diversity. Consistent with theory, a phylogeographic study of a widespread tree in the Pilbara, Eucalyptus leucophloia, found evidence of refugia in the topographically complex high-elevation Hamersley Range, along with the lower elevation, less topographically complex Chichester Range (Byrne et al., 2017). Moisture availability, such as higher precipitation and proximity to permanent groundwater, was considered a factor contributing to refugia in the Chichester Range despite its lower elevation.

Interestingly, evidence for refugia has not been found in other phylogeographic studies of Pilbara plants including another widespread tree (Corymbia hamersleyana, Nistelberger et al., 2020) and several widespread shrubs and trees of the Acacia genus (A. pruinocarpa,Nistelberger et al., 2020; A. ancistrocarpa, Levy et al., 2016; A. hilliana and A. spondylophylla, Millar et al., 2022a). These inconsistent findings suggest that further refinement of our expectations for refugia in this arid landscape is required. For example, mesic refugia may not be detected in high-elevation areas in some species if there has been ongoing genetic connectivity with a general lack of vicariance or range contraction over long time scales and widespread gene flow (Nistelberger et al., 2020). Further, evidence for mesic refugia may be lacking if species that currently occupy the Pilbara have expanded within or into the region from desert-dwelling ancestors due to increasing aridification from the mid-Miocene into the Pleistocene, e.g. the hard spinifex grass, Triodia basedowii complex (Anderson et al., 2019), or some members of Acacia that probably radiated extensively during the expansion of the arid biome (Crisp and Cook, 2013).

In this study, we use non-coding chloroplast sequence data and nuclear microsatellites combined with extensive sampling to detect evidence of phylogeographic refugia and to identify landscape-scale genetic diversity hotspots in two shrub species (Indigofera monophylla and Petalostylis labicheoides), to further refine our understanding of refugia and genetic diversity in arid, unglaciated landscapes. Both species are relatively short-lived peas (Fabaceae), and both are widespread, common and co-distributed across diverse environments within the Pilbara. Multiple populations of both species occur both on and off the high-elevation ranges, providing an ideal opportunity to comparatively test whether topographically complex landscape features elicit similar signals of refugia and elevated diversity in co-distributed shrubs in an arid biome (Edwards et al., 2022). Petalostylis labicheoides is a large, non-woody shrub that tends to grow in water-gaining sites lower in the landscape, while I. monophylla is a small, prostrate to erect woody sub-shrub that recruits seedlings from a soil seed bank following fire and/or ground disturbance. Both species occur in areas subject to extensive mining in parts of their Pilbara range and are key species in the rehabilitation of mined land. Information on the landscape-scale patterns of refugial evolution and contemporary diversity in these species, and those with similar functional traits, is essential to appropriately restore post-mined landscapes and conserve the significant biodiversity values of the bioregion.

Based on general expectations for environmental characteristics of refugia, both species may show signals of refugia and higher genetic diversity in high-elevation, topographically complex areas in contrast to lower elevation plains. Here, we test two hypotheses: (1) the high-elevation Hamersley Range, or parts thereof, have acted as evolutionary refugia for both species, represented by a concentration of chloroplast DNA (cpDNA) haplotype diversity on the Range with lower diversity off the Range on the lowlands and associated plains; and (2) nuclear diversity hotspots are detectable at the landscape scale and co-locate with cpDNA evidence of refugia. Evaluation of these hypotheses will aid refinement of expectations for refugia and diversity hotspots in arid, unglaciated environments.

MATERIALS AND METHODS

Study system

Petalostylis R.Br. (Fabaceae: Dialioideae) (Azani et al., 2017) is comprised of two species based on the most recent formal taxonomic revision: P. labicheoides and a morphologically variable P. cassioides, that are distinguished by leaf size, leaflet shape, number of leaflets and geographic distribution (Ross, 1986). Petalostylis labicheoides is a large, non-woody shrub with solitary yellow flowers and has a disjunct distribution within Australia (Ross, 1986). Within the Pilbara region, P. labicheoides is common and widespread, occurring on sandplains, stony ridges, skeletal soils, dry watercourses and creekbeds, frequently in water-gaining sites; while P. cassioides is known by the authors to currently occur in two discrete locations that were avoided during sample collection. Petalostylis labicheoides plants are disturbance opportunists and live for approx. 15–20 years (S. van Leeuwen, pers. obs.). Pollinators are unknown, but Petalostylis has a highly modified, incurved petal-like style that facilitates insect pollination, and the genus is expected to be self-compatible (Tucker, 1998) and diploid (Azani et al., 2017). Seeds have an elaiosome and, therefore, ant predation and dispersal are possible, but most dispersal is expected to occur by gravity. Plants are highly floriferous but not highly fecund (S. van Leeuwen, pers. obs.) and, for this reason, together with some expected seed predation, any potential soil seed bank is not expected to be large or long-lived.

Indigofera L. (Fabaceae: Papilionoideae) is a large genus with a global distribution (Schrire et al., 2003; Azani et al., 2017). Indigofera monophylla DC. is one of >50 endemic Australian species (Wilson and Rowe, 2015). This name was formerly used to describe a large morphologically variable species complex (Wilson and Rowe, 2004, 2015) but a recent revision has re-circumscribed the group (Wilson, 2021). The revised I. monophylla occurs predominantly within the Pilbara, growing on rocky limestone and ironstone hills, loamy plains and adjacent to watercourses. In contrast to P. labicheoides, plants do not show a preference for water-gaining areas. Plants typically grow as multi-stemmed, small prostrate and/or upright shrubs with red-maroon flowers, although there is some morphological variability (Wilson, 2021). The species is a disturbance opportunist and is shorter lived than P. labicheoides with an approx. 5–10 year life span (S. van Leeuwen, pers. obs.). Seeds do not have an elaiosome, therefore ant predation or dispersal is limited. Plants can re-sprout from mechanical disturbance, and fire also stimulates mass seedling recruitment even when the number of mature plants is low (S. van Leeuwen, pers. obs.), suggesting the presence of a large soil seed bank that is probably also long lived (e.g. seed longevity is relatively high; Erickson et al., 2017). Indigofera species are insect pollinated, with ‘tripping’ flowers that release a cloud of pollen when insects probe the base of the flower (Schrire et al., 2009), but mechanical tripping due to high wind, high temperature and heavy rain is also considered possible (Lakshminarayana and Solomon Raju, 2017). Indigofera monophylla is likely to be self-compatible (Wilson and Rowe, 2004) and expected to be diploid.

Sampling, DNA extraction and genotyping

Fresh leaf material was collected from 24 individuals per population from 31 populations of each species widespread across the Pilbara (Tables 1 and 2). Where possible, collections were spread widely across the geographic space of the population to avoid sampling potentially close relatives. DNA extraction, microsatellite development and PCR amplification followed the methods of Nistelberger et al. (2020). A final set of 16 (P. labicheoides) and 13 (I. monophylla) reliable polymorphic nuclear microsatellites (Supplementary data Tables S1 and S2) were used in the final analysis to genotype 739 individuals of P. labicheoides and 744 individuals of I. monophylla using GENEMAPPER™ v.3.7 (Applied Biosystems).

Table 1.

Nuclear genetic diversity parameters of 31 populations of Petalostylis labicheoides (Fabaceae) from the arid Pilbara bioregion of Western Australia based on 16 microsatellite loci

Population	Code	Latitude	Longitude	n	H O	H E	A R	P	Null	F IS
Pardoo	PAR	–20.287528	119.501389	24	0.40 (0.06)	0.40 (0.06)	3.08 (0.35)	1	0.02 (0.01)	–0.03 (0.05)
Shaw	SHA	–20.781472	119.316806	24	0.16 (0.05)	0.24 (0.06)	1.87 (0.22)		0.05 (0.03)	0.20 (0.10)*
Beabea	BBA	–21.787361	118.828139	24	0.19 (0.05)	0.25 (0.06)	1.86 (0.22)		0.05 (0.02)	0.20 (0.08)*
Ragged	RAG	–21.306194	121.142056	24	0.02 (0.02)	0.05 (0.03)	1.18 (0.10)		0.03 (0.01)	0.64 (0.19)*
Spear	SPR	–21.51875	119.4025	24	0.12 (0.04)	0.17 (0.05)	1.62 (0.18)		0.04 (0.02)	0.21 (0.12)*
Nullagine	NUL	–21.883472	120.094417	24	0.29 (0.07)	0.33 (0.06)	2.80 (0.32)		0.04 (0.02)	0.14 (0.09)*
Cajarina	CAJ	–20.391667	118.886944	24	0.52 (0.06)	0.51 (0.05)	3.90 (0.45)	5	0.02 (0.01)	–0.02 (0.05)
Nimingarra	NIM	–20.382278	120.026556	24	0.38 (0.05)	0.39 (0.05)	3.18 (0.25)		0.03 (0.01)	0.04 (0.06)
Gillam	GIL	–21.085528	118.744056	24	0.45 (0.05)	0.46 (0.05)	3.81 (0.33)	1	0.02 (0.01)	0.00 (0.03)
Coppin	COP	–20.921528	119.980500	24	0.40 (0.08)	0.39 (0.06)	3.18 (0.27)	1	0.03 (0.02)	0.00 (0.08)
Kulbee	KUL	–22.490444	119.998028	24	0.49 (0.07)	0.49 (0.07)	3.84 (0.45)	2	0.02 (0.01)	–0.01 (0.05)
Snappy	SNA	–21.580528	117.104750	24	0.48 (0.06)	0.47 (0.05)	4.02 (0.43)		0.02 (0.01)	–0.04 (0.05)
Tee Tree	TEE	–21.852278	117.617528	24	0.42 (0.07)	0.45 (0.07)	3.15 (0.34)		0.02 (0.01)	0.05 (0.05)
Corbay	COR	–22.031222	118.067278	24	0.43 (0.06)	0.45 (0.06)	3.26 (0.32)		0.03 (0.01)	0.01 (0.04)
Hesta	HES	–22.224639	118.969694	24	0.53 (0.05)	0.55 (0.05)	4.51 (0.35)	1	0.01 (0.01)	0.00 (0.03)
Whundo	WHU	–21.067194	116.972500	23	0.36 (0.07)	0.39 (0.08)	3.75 (0.52)	1	0.02 (0.02)	0.03 (0.05)
Deepdale	DEE	–21.667389	116.205333	23	0.38 (0.06)	0.39 (0.05)	2.91 (0.23)		0.03 (0.01)	0.00 (0.06)
Anderson	AND	–22.033583	117.577639	24	0.44 (0.07)	0.49 (0.07)	3.46 (0.41)	2	0.04 (0.02)	0.08 (0.06)*
Kalamina	KLM	–22.43625	118.385694	24	0.54 (0.05)	0.53 (0.05)	3.85 (0.30)	2	0.02 (0.01)	–0.04 (0.05)
East Munjina	EMU	–22.489583	118.736194	24	0.55 (0.04)	0.55 (0.04)	3.86 (0.39)	1	0.02 (0.01)	–0.03 (0.04)
Nameless	NAM	–22.719056	117.761028	24	0.52 (0.05)	0.52 (0.05)	3.88 (0.35)		0.02 (0.01)	0.00 (0.04)
Packsaddle	PAC	–22.934833	118.737056	24	0.50 (0.07)	0.52 (0.06)	4.38 (0.43)	1	0.03 (0.01)	0.03 (0.04)
Weeli Wolli	WEL	–22.925722	119.197139	24	0.54 (0.06)	0.51 (0.06)	4.07 (0.34)	3	0.00 (0.00)	–0.06 (0.04)*
Hilditch	HIL	–23.236556	118.761861	24	0.51 (0.05)	0.52 (0.05)	3.67 (0.30)	1	0.02 (0.01)	0.00 (0.04)
Kalgan	KAL	–23.229972	119.495167	24	0.51 (0.06)	0.51 (0.06)	3.92 (0.42)		0.02 (0.01)	–0.02 (0.05)
Spearhole	SPE	–23.362500	119.113500	24	0.53 (0.04)	0.56 (0.04)	4.21 (0.29)		0.03 (0.01)	0.04 (0.04)
Ophthalmia	OPH	–23.341083	119.833639	24	0.58 (0.04)	0.57 (0.04)	5.32 (0.43)	2	0.01 (0.00)	–0.03 (0.03)
Paraburdoo	PBO	–23.21125	117.646861	24	0.25 (0.05)	0.30 (0.06)	1.98 (0.24)		0.04 (0.01)	0.18 (0.05)*
Hardy	HAR	–22.953944	117.311000	24	0.30 (0.05)	0.35 (0.05)	2.51 (0.23)		0.04 (0.01)	0.14 (0.05)*
Metawandy	MET	–22.694861	116.610333	24	0.36 (0.06)	0.38 (0.06)	3.25 (0.28)	1	0.04 (0.01)	0.07 (0.05)*
Nanutarra	NAN	–22.561278	115.489000	23	0.27 (0.06)	0.35 (0.07)	2.93 (0.32)		0.06 (0.02)	0.18 (0.05)*
Mean					0.40 (0.03)	0.42 (0.02)	3.33 (0.17)		0.03 (0.00)	0.06 (0.02)*

H _O, mean observed heterozygosity (s.e.); H_E, mean expected heterozygosity (s.e.); A_R, mean allelic richness (s.e.); P, number of private alleles (if > 0); Null, mean proportion of null alleles (s.e.); F_IS (fixation index) (s.e.) (*indicates significantly greater than zero). Note that some populations have the same name as some I. monophylla populations (Table 2) because they were collected in a similar geographic location

Table 2.

Nuclear genetic diversity parameters of 31 populations of Indigofera monophylla (Fabaceae) from the arid Pilbara bioregion of Western Australia based on 13 microsatellite loci

Population	Code	Latitude	Longitude	n	H O	H E	A R	P	Null	F IS
Pardoo	PAR	–20.285861	119.501389	24	0.62 (0.07)	0.67 (0.06)	6.40 (0.54)	1	0.04 (0.02)	0.08 (0.05)*
Cajarina	CAJ	–20.391667	118.886944	24	0.57 (0.05)	0.64 (0.06)	6.48 (0.72)		0.05 (0.02)	0.08 (0.06)*
Shaw	SHA	–20.70975	119.318389	24	0.62 (0.05)	0.73 (0.04)	6.82 (0.75)		0.07 (0.02)	0.14 (0.06)*
Kangan	KAN	–21.035056	118.662583	24	0.57 (0.07)	0.68 (0.06)	6.71 (0.74)		0.06 (0.02)	0.15 (0.06)*
Whim Creek	WCK	–20.884944	117.754444	24	0.47 (0.08)	0.56 (0.07)	5.39 (0.65)	1	0.08 (0.03)	0.21 (0.11)*
Shay Gap	SGP	–20.508111	120.182278	24	0.46 (0.04)	0.65 (0.05)	5.60 (0.57)		0.11 (0.02)	0.26 (0.06)*
Ragged	RAG	–21.295861	121.217028	24	0.51 (0.07)	0.64 (0.08)	5.99 (0.81)		0.08 (0.03)	0.18 (0.07)*
Moolyella	MOO	–21.182389	119.947444	24	0.61 (0.06)	0.68 (0.06)	6.54 (0.77)	1	0.05 (0.02)	0.09 (0.06)*
Coolyia	COL	–21.390028	119.560694	24	0.48 (0.07)	0.57 (0.08)	5.61 (0.63)		0.05 (0.02)	0.12 (0.04)*
Beabea	BBA	–21.887111	118.825083	24	0.44 (0.07)	0.61 (0.08)	6.41 (0.81)	1	0.10 (0.03)	0.23 (0.07)*
Roy Hill	ROY	–22.460917	119.98275	24	0.51 (0.06)	0.60 (0.08)	6.13 (0.90)		0.04 (0.02)	0.10 (0.04)*
Karratha	KAR	–20.742944	116.872167	24	0.52 (0.08)	0.63 (0.08)	5.85 (0.79)		0.06 (0.02)	0.15 (0.06)*
Snappy	SNA	–21.580528	117.10475	24	0.59 (0.06)	0.69 (0.07)	7.65 (0.93)	1	0.05 (0.01)	0.13 (0.03)*
Harper	HAR	–21.871194	117.620972	24	0.49 (0.07)	0.61 (0.08)	6.30 (0.84)		0.06 (0.02)	0.14 (0.06)*
Hooley	HOO	–21.862639	117.984611	24	0.57 (0.07)	0.68 (0.07)	7.92 (0.94)	1	0.07 (0.02)	0.14 (0.05)*
Weelamurra	WEE	–22.074806	117.716472	24	0.56 (0.07)	0.66 (0.06)	6.79 (0.60)		0.06 (0.03)	0.14 (0.06)*
Sheila	SHE	–22.234278	117.607444	24	0.56 (0.07)	0.66 (0.07)	7.43 (0.93)		0.05 (0.02)	0.11 (0.06)*
Hamersley	HAM	–22.252917	117.673528	24	0.56 (0.06)	0.66 (0.07)	7.36 (0.98)		0.05 (0.02)	0.11 (0.05)*
Solomon	SOL	–22.216333	117.975194	24	0.55 (0.07)	0.66 (0.07)	7.28 (0.84)		0.07 (0.02)	0.15 (0.06)*
Wittenoom	WHI	–22.248417	118.359222	24	0.52 (0.07)	0.61 (0.08)	6.15 (0.88)	1	0.06 (0.02)	0.11 (0.05)*
East Munjina	EMU	–22.489583	118.736194	24	0.61 (0.07)	0.63 (0.07)	6.78 (1.00)	2	0.03 (0.01)	0.03 (0.04)
Dinner	DIN	–22.575639	118.300917	24	0.57 (0.06)	0.64 (0.07)	6.62 (0.77)	1	0.04 (0.02)	0.05 (0.05)
Lionel	LIO	–22.64125	117.640417	24	0.55 (0.06)	0.67 (0.08)	6.89 (0.93)		0.07 (0.02)	0.12 (0.07)*
Rhodes	RHO	–23.048778	119.243222	24	0.48 (0.07)	0.61 (0.07)	5.87 (0.77)		0.07 (0.02)	0.22 (0.05)*
Pebble Mouse	PEB	–23.096028	118.939111	24	0.53 (0.06)	0.63 (0.07)	6.25 (0.68)		0.05 (0.02)	0.12 (0.05)*
Giles	GIL	–23.245139	119.144361	24	0.52 (0.07)	0.60 (0.08)	6.15 (0.77)		0.04 (0.01)	0.09 (0.04)*
Ophthalmia	OPH	–23.341083	119.833639	24	0.53 (0.08)	0.59 (0.08)	5.71 (0.78)	1	0.04 (0.02)	0.07 (0.05)*
Paraburdoo	PBO	–23.242111	117.67375	24	0.54 (0.07)	0.67 (0.06)	6.29 (0.63)	1	0.08 (0.03)	0.17 (0.07)*
Beasley	BEA	–22.938694	116.958722	24	0.51 (0.09)	0.62 (0.08)	5.93 (0.76)		0.07 (0.03)	0.19 (0.08)*
Warney	WAR	–22.197444	115.581083	24	0.50 (0.08)	0.63 (0.08)	5.74 (0.89)		0.07 (0.03)	0.19 (0.06)*
Mesa A	MES	–21.653083	115.878694	24	0.65 (0.07)	0.66 (0.07)	5.98 (0.77)	1	0.03 (0.01)	0.00 (0.05)
Mean					0.54 (0.01)	0.64 (0.01)	6.42 (0.11)		0.06 (0.00)	0.14 (0.01)*

H _O, mean observed heterozygosity (s.e.); H_E, mean expected heterozygosity (s.e.); A_R, mean allelic richness (s.e.); P, number of private alleles (if > 0); Null, mean proportion of null alleles (s.e.); F_IS (fixation index) (s.e.) (*indicates significantly greater than zero). Note that some populations have the same name as some I. monophylla populations (Table 1) because they were collected in a similar geographic location

For cpDNA analysis, the methods of Byrne and Hankinson (2012) were used to sequence two D-loop introns (atpF and ndhA) and one intergenic spacer (rps16–trnQ) in P. labicheoides, and the intergenic spacers psbD–trnT, rpl32–trnL and trnG-trnS in I. monophylla, for eight individuals per population, per species. Sequence quality was checked in SEQUENCHER v5.0 (Genecodes Corp., USA) and sequences were aligned using CLUSTAL-W and trimmed and concatenated in MEGA-X v10.2.4 (Kumar et al., 2018). To obtain an outgroup for I. monophylla analysis, the relevant intergenic spacers were retrieved from a whole chloroplast genome of Indigofera tinctoria (GenBank NC_026680). For Petalostylis, there were no congeneric outgroup sequences available to use as an outgroup. Multiple insertion–deletions (indels) occurred in both species’ alignments. For P. labicheoides, indels were manually coded as single base substitutions except for one complex region formed by a large deletion of varying lengths, where each non-overlapping segment was coded as a distinct substitution. Indels occurring in mononucleotide A or T repeat regions were ignored. For I. monophylla, the preliminary inclusion of indels prevented confident sequence alignments due to the presence of multiple large, complex and overlapping gaps. For this reason, we excluded indels from I. monophylla alignments and analyses, as there was a very high level of chloroplast sequence polymorphism via substitutions alone.

Data analysis

Nuclear diversity and structure

FREENA (Chapuis and Estoup, 2007) was used to estimate null allele frequencies and to test their effect on global F_ST, which was calculated before and after null allele correction using the ENA method with 1000 bootstrapped replicates over loci to obtain 95 % confidence intervals. Deviation from Hardy–Weinberg equilibrium (HWE) and linkage disequilibrium (LD) among all pairs of loci in each population was tested using GENEPOP v4.7.5 (Raymond and Rousset, 1995; Rousset, 2008), applying the Bonferroni correction for multiple tests. Population-level genetic diversity parameters were estimated using MSA v4.05 (Dieringer and Schlotterer, 2003). Private alleles per population were identified using GENALEX v6.503 (Peakall and Smouse, 2006, 2012). To visualize the distribution of nuclear genetic diversity across the landscape and to identify genetic diversity hotspots, we used QGIS v2.18.16 (https://www.qgis.org/) to create genetic diversity heatmaps based on mean population allelic richness (A_R). The A_R values were mapped to population location then interpolated across the sampled landscape space using the inverse distance weighted (IDW) method.

To test for isolation by geographic distance (IBD), GENALEX was used to conduct a Mantel test for a correlation between linearized nuclear genetic distance [F_ST/(1 – F_ST)] and the natural log of geographic distance (kilometres), with significance tested by 999 permutations. Global F_ST was calculated using GENEPOP, and population-level differentiation was estimated by calculating pairwise F_ST between all populations, also with GENEPOP. Using GENALEX, genetic distance among individuals was visualized by a principal co-ordinates analysis (PCoA) ordination, which is a method free from any evolutionary assumptions. Finally, the Bayesian program TESS v2.3.1 (Chen et al., 2007) was used to define unique genetic clusters. TESS accounts for the presence of spatial autocorrelation in the data by incorporating geographical information in the analysis, and therefore minimizes the potentially confounding effect of IBD (Meirmans, 2012; Perez et al., 2018). The analysis was run using the CAR admixture model, 50 000 sweeps, a burn-in of 10 000, the default spatial interaction parameter (0.6) and 20 independent runs of each K value up to a K_max value of 15 or until the plot of the deviance information criterion (DIC) vs. K_max began to plateau.

Chloroplast diversity and structure

The number and identity of cpDNA haplotypes, haplotype diversity (H_D) and nucleotide diversity were identified using DNAsp v6 (Rozas et al., 2017). To identify the presence of chloroplast diversity hotspots representing potential evolutionary refugia, landscape-scale interpolated heatmaps of chloroplast diversity were generated based on H_D in QGIS.

To visualize relationships among haplotypes, median-joining haplotype networks were constructed from concatenated alignments using POPART v1.7 (Leigh and Bryant, 2015), and haplotypes were plotted on a geographic map to visualize their distribution across the landscape. We also constructed a maximum likelihood (ML) phylogenetic tree of all unique haplotypes for each species with IQTREE v1.6.12 (Nguyen et al., 2015), using ModelFinder (Kalyaanamoorthy et al., 2017) to find the best evolutionary model and the ultrafast bootstrap (UFBoot) method (Hoang et al., 2018) to assess statistical support. For P. labicheoides, the bases representing coded indels were treated as a separate partition (Chernomore et al., 2016) and the tree was rooted at the midpoint. For I. monophylla, the tree was rooted by the outgroup I. tinctoria.

RESULTS

Nuclear genetic diversity

All microsatellite loci in both species displayed low mean null allele frequencies (Supplementary data Tables S1 and S2). Global F_ST was largely unchanged before and after null allele correction (P. labicheoides 0.274 vs. 0.268; I. monophylla 0.115 vs. 0.113); we therefore report uncorrected values herein. Three out of 16 loci were out of HWE in up to six populations of P. labicheoides, and nine out of 13 loci in up to 18 populations of I. monophylla. Significant LD was detected in two out of 3720 comparisons in P. labicheoides, and in 63 out of 2418 comparisons in I. monophylla. In the latter, the locus pairs involved were not consistent across populations, therefore we consider all loci to be providing independent information.

Overall population genetic diversity was relatively low in P. labicheoides (Table 1) but fixation was not significantly different from zero in most populations, indicating a relative lack of inbreeding across the sampled range despite relatively low diversity. An exception was the Ragged population that had an extremely high fixation value, probably the result of a high number of identical multi-locus genotypes in this particular population (Supplementary data Table S3). Private alleles occurred in 15 out of 31 P. labicheoides populations, which were located relatively randomly throughout the sampled range (Table 1). Population-level diversity was generally higher in I. monophylla than in P. labicheoides (Table 2). In contrast to P. labicheoides and despite higher diversity, there was significant fixation in almost all populations of I. monophylla (Table 2), suggesting the presence of some inbreeding throughout the range. Private alleles occurred in 12 of 31 populations, which were located relatively randomly throughout the sampled range.

At the landscape scale, nuclear diversity hotspots were identifiable in both species (Fig. 1A, B) and these largely occurred within the Hamersley and Fortescue biogeographic sub-regions [i.e. Interim Biogeographic Regionalisation of Australia (IBRA) sub-regions; Thackway and Cresswell, 1995], but the specific locations of hotspots were not concordant among species. In P. labicheoides, a nuclear diversity hotspot occurred in the south-east Pilbara (Fig. 1A; Supplementary data Fig. S1), in populations on the central and south-eastern Hamersley Range, but also extending across the Fortescue River valley to lower elevation populations such as Hesta. However, there were populations of P. labicheoides with relatively high diversity that also occurred outside of this region, and standard errors of A_R values often overlapped (Table 1). In I. monophylla, a diversity hotspot appeared to occur in the central-west Pilbara (Fig. 1B), in populations occurring on the north-western Hamersley Range but also extending to adjacent lower elevation populations such as Snappy and Hooley. However, not all populations in the general hotspot region showed high diversity, and standard errors of A_R values often overlapped (Table 2).

Fig. 1.

Interpolated heatmaps of population-level nuclear allelic richness (A_R) (top; A, B) and chloroplast DNA haplotype diversity (H_D) (bottom; C, D) across the Pilbara landscape in north-western Australia based on 31 populations of Petalostylis labicheoides (left; A, C) and Indigofera monophylla (right; B, D). Note that populations of P. labicheoides denoted by grey stars formed a divergent lineage for both nuclear and chloroplast DNA and may represent a cryptic taxon (see the Discussion); comparable heatmaps excluding these populations are provided in Supplementary data Fig. S2. IBRA sub-regions are labelled and defined by grey lines. Note that the key and scale (e.g. minimum and maximum values and corresponding colours) are unique to each plot, since each interpolation is constrained by the range of observed values within each independent dataset.

Nuclear genetic structure

Significant IBD was detected in both species, and was initially stronger in I. monophylla (r² = 0.269; P = 0.001) than in P. labicheoides (r² = 0.131; P = 0.001). However, excluding the divergent lineage of P. labicheoides (see below), the r² value for P. labicheoides almost doubled (r² = 0.241; P = 0.001), and therefore became similar in magnitude to I. monophylla (Supplementary data Fig. S2).

Global F_ST was relatively high in P. labicheoides (F_ST = 0.27) but was lower when the divergent lineage (see below) was removed (F_ST = 0.20). Pairwise population F_ST values in P. labicheoides ranged from 0.05 to an exceedingly high 0.76 (Ragged vs. Paraburdoo) (Supplementary data Table S4), the latter of which was likely to be a result of the high fixation rate in Ragged. Excluding the divergent lineage, the maximum pairwise F_ST value was 0.46. Indigofera monophylla displayed a lower global F_ST value than P. labicheoides (F_ST = 0.10), suggesting higher levels of population connectivity. Pairwise population F_ST ranged from 0.01 to 0.24 (Supplementary data Table S5).

In the PCoA, many individuals overlapped on the ordination for both species (Figs 2B and 3B), indicating that genetic connectivity is maintained across a large geographic area with a general lack of major genetic barriers, and the degree of overlap was greater in I. monophylla. An exception to this general pattern occurred in P. labicheoides, where six populations from the north-east (Pardoo, Shaw, Spear, Beabea, Nullagine and Ragged) clearly clustered separately on co-ordinate 1, indicating a divergent lineage (Fig. 2B, C).

Fig. 2.

Nuclear genetic structure of 739 individuals of Petalostylis labicheoides from 31 populations in the arid Pilbara bioregion of north-western Australia, based on 16 microsatellites. (A) P. labicheoides flower and shrub at Nimingarra (NIM;(photos S. van Leeuwen). (B) PCoA of all individuals, with individuals coloured by the corresponding TESS cluster to which the population of origin had > 50 % Q membership. (C) Barchart of posterior estimates of individual admixture coefficients when K = 3 based on TESS analysis; C1 = Cluster One, and so on. See Supplementary data Fig. S2 for membership proportions at alternative K values. (D) Mean Q membership to each of K = 3 TESS clusters of each population plotted on a geographic map; black lines represent IBRA region boundaries, grey lines represent IBRA sub-region boundaries within the Pilbara (Chichester, Roebourne, Fortescue and Hamersley) and coloured contour lines represent elevation intervals within the Pilbara; the Fortescue River within the Fortescue IBRA sub-region is drawn in purple. Refer to Table 1 for full population names.

Fig. 3.

Nuclear genetic structure of 744 individuals of Indigofera monophylla from 31 populations in the arid Pilbara bioregion of north-western Australia, based on 13 microsatellites. (A) I. monophylla plant at PAR, flowers from East Munjina (EMU; photos S. van Leeuwen). (B) PCoA of all individuals, with individuals coloured by the corresponding TESS cluster to which the population of origin had >50 % Q membership. Note that the Shaw population did not belong to any cluster with > 50 % membership and is coloured grey. (C) Barchart of posterior estimates of individual admixture coefficients when K = 5 based on TESS analysis; C1 = Cluster One, and so on. See Supplementary data Fig. S3b for membership proportions at alternative K values. (D) Mean Q membership to each of K = 5 TESS clusters of each population, plotted on a geographic map. Black lines represent IBRA region boundaries. Grey lines represent IBRA sub-region boundaries within the Pilbara (Chichester, Roebourne, Fortescue and Hamersley). Coloured lines represent elevation contours within the Pilbara bioregion. The Fortescue River within the Fortescue IBRA sub-region is drawn in purple. Refer to Table 1 for full population names.

Despite some overlap on the PCoA ordination, TESS analysis revealed some structure within both species. For P. labicheoides, the DIC plot began to plateau at K_max = 3, but never truly plateaued (Supplementary data Fig. S3). At K_max = 3, there was one clearly distinct cluster (Cluster One) consisting of six north-eastern populations (Fig. 2C), corresponding to the distinct cluster in the PCoA. A second cluster (Cluster Two) included populations from the high-elevation central and south-eastern Hamersley Range, as well as some populations across the Fortescue River valley (Fig. 2C, D), similar to the location of the nuclear diversity hotspot (Fig. 1A). All remaining sampled populations belonged to a third cluster with a wide geographic range (Cluster Three). At higher K_max values (Supplementary data Fig. S3), the allocation of populations to Clusters One and Two remained consistent, whilst additional clusters were formed from within the original Cluster Three. For I. monophylla, the TESS DIC plot clearly began to plateau at K_max = 5 (Supplementary data Fig. S3), but there was substantial cluster admixture in multiple populations (Fig. 3C, D). At K_max = 5, two clusters formed the majority of the I. monophylla species range: Cluster One was formed by populations in the high elevation south-eastern Hamersley Range but also including Roy Hill across the Fortescue Valley (Fig. 3D). Cluster Two had a wider distribution, but the highest membership values occurred in central Pilbara populations, with increasing admixture with other clusters in populations towards the north-east (Cluster Three), west (Cluster Four) and north (Cluster Five) (Fig. 3D). Population structure was therefore not bound by IBRA sub-region or elevation in either species. Both species instead displayed a broad south-east to north-west pattern of division between major clusters, where cluster changeover occurred approximately in the central Hamersley Range. Both species also displayed a distinct cluster in the north-east Pilbara, although this was more pronounced in P. labicheoides.

Chloroplast haplotype diversity and structure

We detected 23 unique haplotypes in P. labicheoides (Supplementary data Table S6; Fig. 4), formed by 22 substitutions and 12 indels. Haplotype diversity was high (0.871 ± 0.013) although nucleotide diversity was low (0.003 ± 0.000). Almost double the number of haplotypes (45) were detected within I. monophylla (Supplementary data Table S7; Fig. 5), formed by 65 substitutions. Haplotype diversity was very high (0.942 ± 0.000) and nucleotide diversity was low (0.004 ± 0.000). The number of haplotypes per population ranged from one to three in P. labicheoides and from one to six in I. monophylla. In P. labicheoides, only 12 out of 31 populations (39 %) contained more than one haplotype, and 42 % of all populations contained at least one population-specific haplotype (Fig. 4). In I. monophylla, a much larger 23 out of 31 populations (74 %) contained more than one haplotype, and 68 % of all populations contained at least one population-specific haplotype (Fig. 5).

Fig. 4.

Chloroplast DNA haplotypes detected in 248 individuals of Petalostylis labicheoides in the arid Pilbara bioregion of north-western Australia based on three non-coding sequence regions. (A) Median joining haplotype network: circle size is proportional to the number of individuals, mutational steps are represented by dashes, white circles represent intermediate unsampled haplotypes, and haplotype labels inside the circle represent haplotypes that are shared by more than one population. (B) Maximum likelihood consensus phylogenetic tree. Only the branches with ≥95 % Ultrafast bootstrap support are labelled; see Table 1 for full population names. (C) Geographic distribution of haplotypes in each of 31 populations. Piecharts represent the proportion of sampled individuals in the population with each haplotype.

Fig. 5.

Chloroplast DNA haplotypes detected in 248 individuals of Indigofera monophylla in the arid Pilbara bioregion of north-western Australia based on three non-coding sequence regions. (A) Median joining haplotype network: circle size is proportional to number of individuals, mutational steps are represented by dashes, white circles represent intermediate unsampled haplotypes, and haplotype labels inside the circle represent haplotypes that are shared by more than one population. (B) Maximum likelihood consensus phylogenetic tree. Only the branches with ≥95 % Ultrafast bootstrap support are labelled; Group Two has 89 % support; see Table 1 for full population names. (C) Geographic distribution of haplotypes in each of 31 populations. Piecharts represent the proportion of sampled individuals in the population with each haplotype. Population names in italics only contain haplotypes from Group One; population names in normal font only contain haplotypes from Group Two; population names in bold contain haplotypes from both divergent Group.

Population-level haplotype diversity in P. labicheoides appeared to be concentrated in four populations in the central to south-eastern Hamersley Range (Figs 1C and 4C), concordant with the general wider location of the nuclear diversity hotspot in this species (Fig. 1A). For example, 45 % of all sampled haplotypes (excluding the divergent lineage, see below) occurred in these four populations, most of which were also unique to these populations (Fig. 4C). In contrast, I. monophylla showed relatively random distribution of haplotype diversity throughout the Pilbara (Fig. 1D).

The haplotype network for P. labicheoides (Fig. 4A) largely formed a relatively simple star-like structure around a rare central/ancestral haplotype (H12). A small degree of reticulation involved some common, widespread haplotypes (H01, H05 and H08) that occurred in approximately half of all sampled populations (Fig. 4C). Fifteen haplotypes (65 %) were specific to single populations, most (10) of which occurred in the Hamersley Range or Fortescue Valley (Fig. 4C). One highly divergent branch was present in the network, formed by three closely related haplotypes (H02, H17 and H20) from six populations in the north-east Pilbara, which collectively differed from the next closest sampled haplotype by at least nine mutations (Fig. 4A). These populations also formed a distinct cluster based on nuclear loci. This haplotype structure was similarly captured in the ML tree (Fig. 4B), which resolved the three divergent haplotypes as a strongly supported group that was sister to another strongly supported group containing all other haplotypes. Within the latter, H16 and H18 were relatively unresolved, while all remaining haplotypes formed a strongly supported group, but with weakly resolved topology within it.

The haplotype network for I. monophylla was complex and formed two distinct branches (Fig. 5A). Most haplotypes (76 %) were specific to single populations: 15 of these haplotypes occurred in the Hamersley Range or Fortescue Valley and 20 occurred elsewhere. One branch of the network formed a simple star-like structure with multiple haplotypes closely surrounding a common, central/ancestral haplotype (H10). This haplotype was shared by eight populations, all of which occurred in the Hamersley Range or Fortescue Valley (Fig. 5C). The other branch of the network had a more complex structure and contained more highly divergent haplotypes. This overall pattern was also captured by the ML tree (Fig. 5B), which resolved two major haplotype groups corresponding to the two branches of the network, one of which (Group One) was strongly supported while the other had moderate support. One haplotype found in the Ragged population of I. monophylla (H35) was the only haplotype that did not belong to either group (Fig. 5B) and branched in an unusual position near the outgroup in the network (Fig. 5A). Haplotypes from Group Two of the ML tree corresponded to the simpler, star-like branch of the network and were geographically widespread, but frequently occurred in the Hamersley Range and western Pilbara (Fig. 5C). Haplotypes from Group One were also geographically widespread, but frequently occurred in the north-east. Despite this trend, haplotypes from each divergent group also co-occurred together in seven single populations (Fig. 5C).

DISCUSSION

This study has revealed phylogeographic evidence of a localized, high-elevation refugium in the Hamersley Range for P. labicheoides within the arid, unglaciated Pilbara region of north-western Australia. This may be related to historically persistent moisture availability in a species showing a preference for growth in water-gaining sites. The restricted refugium is also co-located with evidence for a contemporary nuclear diversity hotspot in P. labicheoides in the wider area, and this hotspot extended beyond the Hamersley Range to the lower elevation, riparian Fortescue Valley. In I. monophylla, chloroplast haplotype diversity was exceptionally high, although, contrary to expectations, there was no evidence of this diversity being restricted to a high-elevation refugium. However, nuclear diversity was concentrated in the north-western Hamersley Range in I. monophylla and this diversity also extended to adjacent populations across the riparian Fortescue Valley. Patterns of refugia were therefore contrasting between the two study species, confirming that refugia are more nuanced in arid, unglaciated biomes and thus are less predictable in the context of individual species (Byrne et al., 2008; Pepper et al., 2013). However, repositories of nuclear diversity were similarly found in montane areas and their adjacent lowland riparian environments in both species. These data highlight that mesic environments, particularly those formed at the junction of high-elevation and lowland riparian areas, are potential reservoirs of contemporary diversity in arid biomes.

A high-elevation refugium for one shrub, but not the other

Consistent with our first hypothesis, we found evidence that the central to south-eastern Hamersley Range has acted as a historical refugium for P. labicheoides. This is represented by a concentration of high and unique cpDNA haplotype diversity, including a rare ancestral haplotype, within four nearby populations in this region, a scenario that is the expected signal of persistence in refugia (Hewitt, 2000; Gómez and Lunt, 2006; Byrne et al., 2008). In addition, the occurrence of two common, closely related haplotypes in geographically widespread populations mostly off the Hamersley Range that also tend to have lower haplotype diversity is consistent with a signature of range expansion. This refugial pattern was also reflected in the nuclear genome: higher nuclear diversity was concentrated in the central and south-eastern Hamersley Range, although this extended across the Fortescue Valley, indicating that nuclear diversity was not restricted to the high-elevation Range. Further, despite the signal of a refugium in the south-east Hamersley Range, populations with relatively high haplotype diversity were not exclusive to the Range (e.g. Snappy, Nanutarra, Cajarina and Coppin), and populations with low haplotype diversity also occurred on the Range. These patterns suggest finer scale patterns of refugial evolutionary history in P. labicheoides rather than a simple pattern of high diversity on the Hamersley Range and low diversity off the Range. Nevertheless, we interpret the overall pattern to reflect long-term persistence of populations in a localized part of the Hamersley Range, with concordant elevated nuclear diversity in the wider area, plus range expansion throughout a larger part of the rest of the Pilbara.

In contrast to P. labicheoides, our hypothesis of refugia in the Hamersley Range was rejected for I. monophylla. Instead, there was a relatively random pattern of higher vs. lower haplotype diversity throughout the Pilbara region, although some populations that formed a nuclear diversity hotspot in the central Pilbara also harboured high chloroplast diversity (e.g. Snappy and Hooley). This central region may therefore be considered as a contemporary reservoir of both nuclear and chloroplast genetic diversity in this species, together with the north-eastern part of the range (e.g. Cajarina, Shaw and Kangan) that also showed relatively high chloroplast diversity. The lack of any clear evidence of refugia for I. monophylla in the Hamersley Range is consistent with observations in the tree Corymbia hamersleyana and several Acacia species (Levy et al., 2016; Nistelberger et al., 2020; Millar et al., 2022b) in the Pilbara.

We therefore observe contrasting evidence for and against the existence of refugia in high-elevation areas in two common, similarly distributed Fabaceae shrubs in the arid Pilbara landscape. Our observed patterns of refugia might reflect nuance driven by species-specific traits. For example, P. labicheoides has a preference for growth in water-gaining areas and belongs to a subfamily (Dialioideae) that is otherwise strongly tropical in distribution (Azani et al., 2017). These traits may partially explain why the topographically complex, mesic environment of the Hamersley Range has facilitated persistence and acted as a refugium in this particular species. Similarly, E. leucophloia in the Pilbara has also shown evidence for refugia in the Hamersley and Chichester Ranges, which was similarly attributed in part to increased moisture availability (Byrne et al., 2017). In contrast, I. monophylla does not have a preference for growth in water-gaining areas and appears to have fire-stimulated seedling emergence, therefore the ranges and the more mesic microclimates they create may not necessarily represent optimal or refugial conditions in this species. Indeed, it is possible that the I. monophylla complex may have instead expanded its range within or into the Pilbara due to increasing aridity in the early Pleistocene. A phylogeographic study of the arid-tolerant Indigofera bungeana species complex in China similarly concluded that the complex underwent Pleistocene population expansion (Zhao et al., 2017), and previous authors have hypothesized an early Pleistocene radiation of Indigofera within Australia (Davidson and Davidson, 1993). Overall, our study suggests a requirement for greater understanding of the intersection of species traits and the attributes of topographically complex, high-elevation areas, or the environmental features that characterize them, that may facilitate their role as historical evolutionary refugia in arid landscapes.

The exceptional level of haplotype diversity we have observed in I. monophylla is the highest reported to our knowledge for a Pilbara plant species. Very high cpDNA diversity was also a feature of the I. bungeana complex (Zhao et al., 2017) and of other functionally similar legumes in mesic south-western Australia (Bradbury et al., 2016a, b). This pattern may be related to the obligate seeding life history where seeds are stored in a large and long-lived soil seed bank and seedlings are stimulated to emerge intermittently following disturbance such as fire. Long-lived seed banks could therefore act as multi-generational reservoirs of seed-based, maternally inherited diversity (Bradbury et al., 2016a, b), and this may provide a mechanism for maintenance of haplotype diversity in times of range contraction, thus obscuring any signal of historical refugia. An alternative explanation for widespread high haplotype diversity is that I. monophylla has not undergone any major range contractions within the Pilbara, instead maintaining a large, stable or expanding population. Similar patterns suggesting long-term persistence, connectivity or expansion appear to be characteristic of several other species in the Pilbara (Anderson et al., 2019; Levy et al., 2019; Nistelberger et al., 2020; Umbrello et al., 2020); and extremely high haplotype richness has been attributed to adaptive radiation in species from other diverse, arid and semi-arid global regions (Zhao et al., 2017; You et al., 2022).

The formation of two major haplotype groups in I. monophylla was unexpected and may have obscured the utility of these data for interpreting the refugial evolutionary history of the species. For example, the two clearly divergent groups did not show a strict association with geography, environment or morphology, given that seven populations harboured haplotypes from both groups in the same population. Such a pattern can be consistent with historical hybridization and introgression or can occur through retention of ancestral haplotypes via incomplete lineage sorting (ILS). These two explanations can be difficult to distinguish and are not necessarily mutually exclusive. Introgressive hybridization is possible for I. monophylla since at least 12 other Indigofera species occur within 5 km of our sampled populations. There are no known cases of contemporary I. monophylla hybrids, although one unusual specimen has been noted as showing morphology approaching I. rugosa (Wilson, 2021). Given that I. monophylla is part of a larger species complex with high haplotype diversity, and that other Indigofera species complexes have shown evidence of radiative diversification (Zhao et al., 2017), it is also plausible that the observed pattern in I. monophylla represents ILS due to recent/rapid evolution. Finally, the presence of two divergent but co-occurring haplotype groups could have arisen through a complex history of range shift dynamics within the species. For example, I. monophylla may have experienced historical contraction and divergence into two separate lineages within the Pilbara (e.g. north-east vs. south-west), with subsequent, possibly multiple, expansions or range shifts resulting in secondary contact and admixture. This would lead to the presence of divergent cpDNA haplotypes with an absence of divergence in the nuclear genome (i.e. cytonuclear discordance), as we have observed. A scenario of fragmentation, allopatric divergence and secondary contact has also been reported in cacti from arid, unglaciated areas of South America in rocky habitats of high diversity and endemism, similar to attributes of the Australian Pilbara (Bonatelli et al., 2014). Further, cytonuclear discordance promoted by the repeated isolation, divergence and secondary contact of lineages has also been reported in the Iberian peninsula, and has been proposed to be more frequent in such diverse, evolutionarily refugial parts of the world (Dufresnes et al., 2020).

Nuclear diversity hotspots occur in high-elevation and adjacent lowland mesic environments

Our second hypothesis that nuclear diversity hotspots are detectable at the landscape scale and co-locate with evidence for refugia was supported by P. labicheoides but partially rejected by I. monophylla. Evidence for an evolutionary refugium of P. labicheoides in the Hamersley Range overlapped with the general location of a wider nuclear genetic diversity hotspot. This would be expected if refugia have facilitated persistence and maintained diversity over long time periods and are therefore also associated with elevated contemporary diversity, representing high conservation priority areas (Keppel et al., 2012; Selwood and Zimmer, 2020). Unlike the refugium, the nuclear diversity hotspot was not exclusive to the high-elevation Ranges but extended to adjacent lower elevation populations in the Fortescue Valley, and population genetic structure also tracked this pattern. Partial support for the hypothesis was found in I. monophylla, where a landscape-scale nuclear diversity hotspot was detectable in the central-west Pilbara, and, similarly, this hotspot was not exclusive to the Ranges but extended across the Fortescue Valley to the lower elevation Snappy and Hooley populations. However, in contrast to our hypothesis, this hotspot did not co-locate with any evidence for historical refugia. A similar scenario was observed in two other small shrubs (Acacia hilliana and A. spondylophylla) that showed higher nuclear diversity in the Hamersley Range but lacked chloroplast evidence of refugia in the same location (Millar et al., 2022a). Therefore, montane regions (or parts thereof) might act as reservoirs for contemporary genetic diversity but not necessarily as historical refugia or to the exclusion of other areas that may also be reservoirs of diversity in heterogeneous arid environments.

High nuclear diversity in riparian areas adjacent to Hamersley Range sites in both P. labicheoides and I. monophylla (although in nuanced locations) is consistent with similar observations in E. leucophloia of higher diversity and unexpected refugia in areas of permanent water availability in the lower elevation Chichester Range, including sites associated with the Millstream aquifer (Byrne et al., 2017) that is a significant wetland in the region (McKenzie et al., 2009). The Snappy populations of both P. labicheoides and I. monophylla also occur in the Millstream area and both exhibited elevated diversity in either the nuclear or chloroplast genome. Therefore, riparian environments that maintain moisture availability and access to permanent groundwater sources might equally be considered as candidate reservoirs of diversity in concert with rocky/montane areas in arid environments (Selwood and Zimmer, 2020). Further, both species exhibited continuous genetic structure across the Fortescue Valley from the Range, indicating that the river and associated lowland elevational gradient was not a barrier to gene flow. Instead, the Fortescue Valley with its associated riparian, grassland and saltmarsh ecosystems may even be facilitating connectivity with the Range (Fremier et al., 2015) by maintaining population size, persistence and diversity in a mesic environment in times of heightened aridity. This result highlights the role that major waterbodies, floodplains or riparian areas may play in maintaining centres of intraspecific genetic diversity in arid biomes, further confirming the conservation value of riparian areas that are known globally (Sabo et al., 2005; Riis et al., 2020). In arid and dryland areas, this may be due to riparian areas enhancing connectivity by minimizing the formation of strong genetic structure across elevational gradients, and preventing diversity loss due to extensive dispersal facilitated by the movement of water (e.g. Murray et al., 2019a, b; Nistelberger et al., 2020). Based on our data, this role may be amplified where riparian areas co-occur with topographically complex landscapes, since we did not observe comparable nuclear diversity hotspots in the topographically subdued north-east Pilbara where major non-perennial river systems also occur.

Despite an overall lack of strong genetic structure, we observed a broad south-east to north-west division of nuclear clusters in both species. A similar pattern has also been documented in the small shrubs Acacia hilliana (Millar et al., 2022a) and Mirbelia viminalis (Millar et al., 2022b) in the Pilbara. This warrants attention as evidence of a potential wider biogeographic pattern in plant species of similar functional type across the region that may be relevant to conservation strategies. A divergent nuclear cluster in the north-eastern Pilbara also occurred for both species, which in P. labicheoides was also repeated in the chloroplast genome. We consider that this unique, strongly divergent lineage may represent a cryptic Petalostylis taxon and warrants further taxonomic investigation. An east–west partitioning of genetic diversity in the northern Pilbara has also been observed in geckos, prompting calls for the north-east to be considered a distinct bioregion relative to the north-west (Pepper et al., 2013), and our data support that proposal.

Conclusions

Phylogeographic and landscape genetic studies in understudied arid regions of the globe are required to refine our understanding of evolutionary refugia and the distribution of genetic diversity in these biomes. Our study demonstrates that high-elevation landscapes in arid biomes may be phylogeographic refugia for some species but not all, such that patterns of refugia are more likely to be nuanced and species specific in comparison with those of more temperate biomes. Interestingly, nuclear diversity hotspots were identifiable in both species, and their locations suggested that mesic environments formed by the intersection of topographic complexity with riparian and floodplain zones can be reservoirs of contemporary genetic diversity in arid landscapes.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Table S1: characteristics of 16 microsatellite loci based on 739 individuals from 31 populations of Petalostylis labicheoides in the Pilbara bioregion of Western Australia. Table S2: characteristics of 13 microsatellite loci based on 744 individuals from 31 populations of Indigofera monophylla in the Pilbara bioregion of Western Australia. Table S3: sets of identical multilocus genotypes detected in each species based on 13 microsatellite loci in Indigofera monophylla and 16 loci in Petalostylis labicheoides. Table S4: pairwise F_ST values between all sampled populations of Petalostylis labicheoides from the Pilbara bioregion, Western Australia, based on 16 nuclear microsatellite markers. Table S5: pairwise F_ST values between all sampled populations of Indigofera monophylla from the Pilbara bioregion, Western Australia, based on 13 nuclear microsatellite markers. Table S6: GenBank accession numbers for sequences of each non-coding chloroplast region that formed 23 unique haplotypes of Petalostylis labicheoides based on 248 individuals in the Pilbara bioregion. Table S7: GenBank accession numbers for sequences of each non-coding chloroplast region that formed 45 unique haplotypes in Indigofera monophylla based on 248 individuals in the Pilbara bioregion. Figure S1: interpolated heatmap of population-level nuclear allelic richness and chloroplast haplotype diversity in 25 populations of Petalostylis labicheoides within the Pilbara region, excluding six populations belonging to a divergent lineage that potentially represent a cryptic taxon. Figure S2: isolation by geographic distance represented by a significant positive correlation of geographic distance with genetic distance in the Pilbara bioregion. Figure S3: full TESS results for genetic clustering analyses of Petalostylis labicheoides and Indigofera monophylla in the Pilbara bioregion based on nuclear microsatellites.

FUNDING

This work was supported by Rio Tinto and BHP.