A genetic evaluation of conservation units and four decades of translocation history for Banksia cuneata, an iconic threatened species in a biodiversity hotspot

Abstract

Background and Aims

Translocation and genetic data are increasingly used in tandem for threatened species recovery. However, for species managed over many decades, records may be incomplete and have limited genetic data. Modern genetic analysis can assist with understanding the consequences of historical translocations and is key to informing ongoing management. Here, we assess genetic patterns in Banksia cuneata, an iconic species that has been under conservation management since 1980, to clarify conflicting signals from early genetic studies and evaluate the species’ translocation history.

Methods

We applied ddRADseq on current in situ and historical ex situ samples, including translocations and extinct populations, to assess spatial and temporal genetic variation in B. cuneata and relatives.

Key Results

Strong divergence confirmed B. cuneata to be a distinct species and separate conservation unit from a morphologically similar relative, B. oligantha, where previous allozyme data had indicated conspecificity or hybridization. Analysis of a known mixed-source translocation confirmed the expected increase in genetic diversity, while temporal changes in population structure revealed an additional, unexpected mixed-source translocation following historical documentation inconsistencies. We also confirmed strong population structure and moderate genetic diversity across the extant distribution of B. cuneata, clarifying contradictory signals in early allozyme and RAPD datasets. Finally, historical samples demonstrated sufficient diversity in ex situ seed collections and confirmed the genetic affinity of collections from now-extinct populations to enable their appropriate usage.

Conclusions

From much confusion and patchy records, our data provide clarity on appropriate conservation and management units, translocation history and the genetic composition of extant populations and germplasm collections to greatly improve the ongoing conservation management of B. cuneata. More broadly, our study provides vital knowledge concerning an active long-term translocation program, highlighting the importance of future-proofed record-keeping and the value of revisiting genetic data over the duration of any conservation program.

INTRODUCTION

As anthropogenic pressures on natural ecosystems and species accelerate globally, active conservation management of threatened plant species is increasing, and translocations have become a common approach to mitigating decline and promoting species recovery (Godefroid et al., 2011; Silcock et al., 2019; Bellis et al., 2024). Through the intentional transplantation of plants or regenerative material to new locations and to augment existing populations, translocation aims to reduce the risk of extinction by increasing population sizes and associated genetic diversity, increasing the number of populations and re-establishing connectivity among fragmented populations (Frankham et al., 2011; Weeks et al., 2011). The value of using genetic data to inform the planning and monitoring of translocations has been long known (e.g. Falk and Holsinger, 1991; Akeroyd and Jackson, 1995; Weeks et al., 2011), but in practice has been an underutilized tool in conservation management, particularly in non-vertebrate organisms (Laikre et al., 2010; Frankham et al., 2017; Holderegger et al., 2019). With growing affordability and accessibility to molecular workflows, as well as improved collaboration between academic and conservation agencies, genetic data and translocation are increasingly being used in tandem for effective conservation management (Ottewell et al., 2015; Frankham et al., 2017; Holderegger et al., 2020; Rossetto et al., 2023; Shaw et al., 2024).

From a planning perspective, quantifying the spatial distribution of genetic variation in wild populations can aid in the identification of conservation units, guide optimal germplasm collections and inform their use in recovery actions. Taxonomy is foundational to conservation in defining the units (i.e. species or subspecies) upon which conservation management is implemented (Mace, 2004; Frankham et al., 2012; Ely et al., 2017). Consequently, any undescribed taxa will be overlooked for conservation attention and oversplit taxa can misuse limited conservation resources and risk inappropriate management (Wilson et al., 2022; Cascini et al., 2025). To aid in this critical first step of identifying appropriate conservation units, genetic data are increasingly being used in the clarification of species boundaries, particularly when morphological evidence is limited or challenging (e.g. Rutherford et al., 2022; Anderson et al., 2023; Orel et al., 2023; Wu et al., 2023). Once conservation units are clearly defined, understanding the genetic context of wild populations is vital to planning appropriate recovery actions, identifying priority populations and guiding germplasm collections of genetically representative material to be used in translocation programs (Schoen and Brown, 2001; Li and Pritchard, 2009; Weeks et al., 2011; Basey et al., 2015). Consideration of factors such as within-population genetic diversity, among-population differentiation, clonality and inbreeding allow germplasm collections to be made and utilized in translocations with intention to optimize genetic diversity and adaptive potential, while also avoiding unwanted outcomes, such as outbreeding depression or hybrid swamping (Frankham et al., 2011, 2017; Weeks et al., 2011; Ottewell et al., 2015; McGlaughlin et al., 2015; Rutherford et al., 2019; Rossetto et al., 2023). The use of historical samples, which harbour genetic signals from previous time points, can also aid in planning processes. Genetic assessments of historical plant samples, such as herbarium specimens (e.g. Cozzolino et al., 2007; Rosche et al., 2022), ex situ seed collections (e.g. Coates et al., 2015) or the stratified soil seed bank (e.g. Morris et al., 2002; Summers et al., 2018), can reveal pre-decline levels of genetic diversity and gene flow from times when populations were larger or less fragmented. These assessments can be used to establish genetic baselines from which to develop restoration targets (Hofman et al., 2015).

Following the establishment of a translocation, genetic data are also valuable from a monitoring perspective to assess whether the intended goals are being achieved and to guide corrective actions, if needed. Assessments of genetic diversity and mating system parameters in founding and successive generations of translocations can inform whether reproductive functioning and adaptive potential have been adequately maintained or improved relative to the source populations (e.g. Monks et al., 2021; Dillon et al., 2023). For mixed-source translocations, this includes confirmation of genetic recombination among sources and assessment of the impacts of such admixture on fitness (e.g. Zavodna et al., 2015; Shemesh et al., 2018; St. Clair et al., 2020). Both short- and long-term genetic monitoring of plant translocations is important because even with carefully curated input, low-fitness individuals may be inadvertently introduced into translocations via unavoidable factors such as small founder sizes and subsequent inbreeding due to limited germplasm availability (e.g. Robichaux et al., 1997; Krauss et al., 2002), relaxed selection pressures on the founding individuals due to highly nurturing ex situ germination and growing conditions (St. Clair et al., 2020) or outbreeding depression in later generations (Frankham et al., 2011; Van Rossum and Hardy, 2022). Moreover, as translocation programs are typically iterative rather than singular, the genetic information gained from early monitoring can greatly inform adjustments for ongoing augmentation to meet long-term goals (Van Rossum and Hardy, 2022; Rossetto et al., 2023).

While the benefits of using genetic data to plan and monitor translocation programs are well established, translocation has been used as a conservation tool for much longer than genetic data have been readily available. Moreover, translocation programs tend to be less well documented, particularly in the scientific literature, such that detailed establishment or monitoring records of older translocations are often not available or become lost over time (Godefroid et al., 2011; Silcock et al., 2019; White et al., 2023). This is one of the reasons that long-term monitoring studies of plant translocations are rare (Van Rossum and Hardy, 2022), as they are limited, if not impossible, without detailed documentation. This is particularly problematic for scenarios in which undocumented translocations have modified the genetic composition of extant populations, germplasm or ex situ living collections because the lack of this information may lead to inappropriate decision-making with unintended consequences in future management actions (Latch and Applegate, 2006). In a similar way to that in which genetic data have been used to document other anthropogenic influences on species dispersal, such as the introduction histories of invasive species (e.g. Hawkins et al., 2007; Hopley et al., 2021) or ancient translocation practices (e.g. Rosvold et al., 2012; Rangan et al., 2015; Rossetto et al., 2017), genetic investigations using historical samples may also provide an opportunity to reconstruct conservation-based translocation histories for species that have been under long-term management.

Effective conservation management is especially important in global biodiversity hotspots, where there are high levels of endemism, species diversity and major threatening processes (Myers et al., 2000). In one of these global hotspots, the Southwestern Australian Floristic Region (SWAFR), plant translocation has been utilized extensively as a management tool since the 1980s (Monks et al., 2023). In more recent decades, these translocations have been implemented under an experimental framework with careful documentation to allow monitoring (Monks et al., 2023); however, earlier translocation records can be limited. One species with over four decades of conservation management, including translocation, is Banksia cuneata, an iconic species that is known for its striking pink inflorescences that resemble matchsticks (Fig. 1A). The species is endemic to the SWAFR and has been actively managed since the species was listed as threatened in 1982, including repeated seed collections held in ex situ storage and the use of some of this seed for population augmentations through the 1990s (Department of Environment and Conservation, 2009). However, records of this augmentation history are patchy and thus the consequences of these recovery actions are largely unknown. This conservation program also represents a rare example of the early adoption of genetic data to inform the conservation management of threatened species, with several genetic studies of B. cuneata undertaken through the 1990s and early 2000s (Coates and Sokolowski, 1992; Maguire and Sedgley, 1997; Broadhurst and Coates, 2004). Interestingly, these studies reported conflicting patterns (summarized in Supplementary Data Fig. S1). Studies using allozymes reported low genetic diversity and strong population differentiation among wild populations (Coates and Sokolowski, 1992; Broadhurst and Coates, 2004), whilst a study using random amplified polymorphic DNA (RAPD) markers reported high genetic diversity and no population differentiation (Maguire and Sedgley, 1997). These conflicting signals have presented confusion in designating appropriate management units and conservation decision-making for B. cuneata. Moreover, genetic indications of admixture or hybridization between B. cuneata and a parapatric relative, B. oligantha (Broadhurst and Coates, 2004), also raised questions regarding species boundaries and appropriate conservation units. While these genetic studies utilized the best information and data available at the time, they were conducted on incomplete subsets of populations and applied early molecular technologies with limited markers. Ongoing conservation management of B. cuneata would benefit from clarity regarding the species’ genetic standing in terms of natural population dynamics and appropriate conservation units by revisiting these genetic studies with updated molecular technology. In addition, genetic data would clarify any consequences of previous translocation practice that might have impacted the genetic composition of germplasm collections or extant populations.

Fig. 1.

Images displaying key features of the three study species. (A) An inflorescence of Banksia cuneata, showing pink colouring along the full length of the perianth. (B) An inflorescence of B. oligantha showing reduced pink colouring and fewer flowers. (C) An inflorescence of B. ilicifolia with much reduced pink colouring and larger leaves. (D) The shrub-like growth habit of B. cuneata, which is similar to that of B. oligantha. This image also shows the setting of most B. cuneata populations on weedy roadsides and conservation staff undertaking seed collections. (E) The tree growth habit of B. ilicifolia. Image credits to authors A. Crawford (A, D), M. Edgley (B), N. Delnevo (C) and R. Binks (E).

This study reports on a comprehensive reassessment of genetic variation in B. cuneata and relatives to resolve the mixed signals from previous genetic studies, gain insight into historical management practices and inform ongoing recovery actions. Specifically, we generated high-resolution single-nucleotide polymorphism (SNP) data for all extant populations, the founding population of a recent, well-documented translocation, and historical ex situ seed collections that include some extinct populations. These data were used to address three primary aims: (1) confirm appropriate conservation units through clarifying the evolutionary relationships amongst B. cuneata and close relatives, B. ilicifolia and B. oligantha; (2) identify any indicators of previous translocation history for B. cuneata through use of historical samples to assess temporal variation in genetic diversity and structure; and (3) confirm management units through the clarification of spatial patterns of genetic diversity and population structure among extant populations of B. cuneata. Clarity regarding these long-unresolved uncertainties and previous translocation efforts will provide confidence in making informed decisions for the continued conservation of this iconic species.

MATERIALS AND METHODS

Study species and conservation history

Banksia cuneata is a long-lived woody shrub that grows in deep yellow sands in the semi-arid Wheatbelt region of the SWAFR. It sits in the small subgenus Isostylis with two morphologically similar species, B. ilicifolia and B. oligantha (Mast and Givnish, 2002; He et al., 2011) (Fig. 1). Banksia cuneata is clearly differentiated from the erect tree B. ilicifolia by its shrub-like growth habit, smooth bark, smaller leaves, flowers and fruit, and pink colouring along the full length of the perianth (largely yellow-cream in B. ilicifolia) (George, 1981) (Fig. 1A, C–E). Banksia cuneata is more similar to B. oligantha, with the latter differing in bearing consistently fewer flowers within inflorescences and reduced pink coloration in the perianth limbs, tending to be more green-cream (George, 1988) (Fig. 1A, B, D). Banksia ilicifolia is a common species with a wide distribution throughout the higher-rainfall regions of the SWAFR that follow the coastline, while B. oligantha has a narrow distribution with just four disjunct populations in the inland Wheatbelt region, some 50 km south and south-east of the southernmost B. cuneata populations (Fig. 2).

Fig. 2.

Geographic distribution and sampling sites for the three study species, Banksia cuneata (BC), B. ilicifolia (BI) and B. oligantha (BO) in south-western Australia. Distributional points are based on records from the Western Australian Herbarium. For the two threatened species (BC, BO), all known populations are displayed; population labels in black were sampled via in situ and/or ex situ collections, while greyed-out population labels were not included in this study, with X in the population labels indicating extinct populations. Grey lines on the broader map indicate IBRA (Interim Biogeographic Regionalisation for Australia) regions, while blue lines on the inset map indicate hydrological systems, including salt lakes. Photo inset: B. cuneata inflorescence (A. Crawford).

Banksia cuneata is monoecious and becomes reproductively mature from 5 years of age (Burgman and Lamont, 1992), with mass flowering occurring annually from September to December, attracting a wide range of nectar-feeding birds and insects. Honeyeaters are considered the primary pollinators, with at least six species known to feed on the nectar of B. cuneata flowers (McNee, 1997). Banksia cuneata has a mixed mating system that is largely outcrossing but the species is self-compatible, and reduced outcrossing rates have been detected in highly disturbed populations (Coates and Sokolowski, 1992; Coates et al., 2013). The species is serotinous, with fire being the primary trigger for the release of canopy-stored seed, which can accumulate in large volumes over successive years without fire (George, 1981, 1988; Burgman and Lamont, 1992). Consequently, while fire kills adult plants, it is important for recruitment in B. cuneata.

Twelve populations of B. cuneata have been recorded since the species was first documented in the 1980s, five of which are now locally extinct (Fig. 2). The seven extant populations have a patchy distribution along roadsides or in remnant bushland in a largely cleared agricultural region. Prior to land clearing, populations would have been larger and more numerous, but were likely still patchily distributed, coinciding with the patchy distribution of deep yellow sands amongst other soil types and saline lakes in this region (Coates and Sokolowski, 1992). The remaining extant populations have poor recruitment and are impacted by additional threats, including small population size, habitat degradation (e.g. weed invasion, salinity, drought), grazing and altered fire regimes (Department of Environment and Conservation, 2009). The species is currently ranked as Endangered under the Western Australian Biodiversity and Conservation Act 2016 and the Australian Commonwealth Environment Protection and Biodiversity Conservation Act 1999, both of which apply IUCN Red List criteria.

Since the 1980s, recovery actions by the Department of Biodiversity, Conservation and Attractions (DBCA) and its predecessors have included repeated seed collections that have been used for population augmentation, as well as maintaining a long-term ex situ germplasm collection that holds seed reserves from most populations, including now extinct populations. Records of the species’ management history are documented in the grey literature, largely in the form of internal reports, translocation proposals, letters of correspondence and recovery plans. These documents indicate that many of the 12 populations were augmented in the 1990s using ex situ seed collections (Department of Environment and Conservation, 2009). However, details of those augmentations are patchy, with limited information regarding plant numbers, source populations and individual survival over time, such that it is not currently possible to track which plants in the extant populations are wild versus augmented. Based on the available documentation, we understand that the standard practice of augmentation for any given population was for seed/seedlings to be sourced only from that population (i.e. single, local source). More recently in 2018 and 2019, plants were reintroduced into a part of population BC-2 where the species no longer occurred. This translocation is ∼1 km north of the extant portion of BC-2 and was planted with seedlings sourced from the five extant populations in the northern part of the species’ range (Monks et al., 2023). This population mixing followed the historical recognition of two management units corresponding to the disjunction of northern and southern populations within the species’ range (Fig. 2). All founder plants in this mixed-source translocation, which were randomly distributed across the site to maximize admixture among sources, were tagged and databased to allow long-term monitoring. To distinguish this well-documented translocation from the previous historical translocations, we hereafter refer to it as the ‘recent translocation’ (BC-T). A summary of the species’ documented populations and their translocation history to our best knowledge, is provided in Table 1.

Table 1.

Population and sampling summary for Banksia cuneata (BC), B. ilicifolia (BI) and B. oligantha (BO) in south-western Australia.

Species	Population ID	Status	Translocation history	Translocation source	Census (2021)	In situ collection year (n)	Ex situ collection year (n)
Banksia cuneata	BC-1	Extant	Augmented	BC-1*	21	2021 (20)	1996 (10), 2007 (10)
Banksia cuneata	BC-2	Extant	Wild		11	2021 (11)	1988 (12), 1996 (13), 2007 (5)
Banksia cuneata	BC-X3	Extinct	Wild		–	–	1988 (4)
Banksia cuneata	BC-4	Extant	Augmented	BC-4*	>50	2021 (20)	1988 (7), 1996 (11), 2007 (12)
Banksia cuneata	BC-X5	Extinct	Wild		–	–	1988 (8)
Banksia cuneata	BC-6	Extant	Wild		>100	2021 (20)	1988 (8), 2007 (8), 2017 (14)
Banksia cuneata	BC-7	Extant	Augmented	BC-7*	8	2022 (8)	1988 (2), 1996 (11)
Banksia cuneata	BC-8	Extant	Augmented	BC-8*	10	2021 (10)	1990 (10), 1997 (10), 2007 (10)
Banksia cuneata	BC-X9	Extinct	Wild		0	–	–
Banksia cuneata	BC-X10	Extinct	Augmented	BC-10*	0	–	–
Banksia cuneata	BC-X11	Extinct	Wild		–	–	–
Banksia cuneata	BC-12	Extant	Augmented	BC-12*	>100	2021 (20)	2007 (25)
Banksia cuneata	BC-T	Extant	Reintroduction	BC-1, 2, 4, 6, 7	213	2021 (24)	–
Banksia ilicifolia	BI-1	Extant	Wild		>100	2021 (10)	–
Banksia ilicifolia	BI-2	Extant	Wild		∼50	2021 (10)	–
Banksia oligantha	BO-2	Extant	Wild		>100	2021 (10)	–
Banksia oligantha	BO-4	Extant	Wild		>100	2021 (10)	–

In situ collection refers to leaf material collected from adult plants in extant populations, while ex situ collections refer to leaf material collected from seedlings that were germinated from separate maternal lines of historical seed collections held in ex situ storage.

Population numbers for the two threatened species (BC and BO) match their designated population numbers in the DBCA Threatened and Priority Flora database to allow ease of reference for conservation managers.

All augmentation was done in the 1990s and the recent translocation (BC-T) was established in 2018–2019 as a re-introduction to a locally extinct part of BC-2.

^*Assumed source population used in augmentation.

Sampling collections

Sampling of B. cuneata involved three different sources of genetic material: (1) leaf material collected from reproductively mature plants across all extant populations (hereafter referred to as the ‘in situ collection’); (2) leaf material collected from mature founding plants in the mixed-source ‘recent translocation’, and (3) leaf material collected from seedlings germinated from ex situ seed collections (hereafter referred to as the ‘ex situ collection’). The in situ sampling occurred in 2021, when leaf material was collected from 133 adult plants across the seven extant populations of B. cuneata (BC-1, 2, 4, 6, 7, 8, 12) (Fig. 2). For larger populations, sample sizes were capped at 20, spaced as widely as possible to cover population extent, while all mature plants were sampled in the smaller populations. For the recent translocation (BC-T), 24 plants were selected for sampling to ensure that all five source populations were represented in proportion to their original planting proportions. For the ex situ collection, 190 seed collected from the same seven extant populations, as well as two extinct populations (BC-X3, X5), were sourced from historical seed collections held in ex situ storage within DBCA’s Western Australian Seed Centre, Kensington (WASC-K). These collections were stored under standard gene bank conditions (stored at −20°C after drying to equilibrium at 15°C and 15% relative humidity) (Martyn Yenson et al., 2021). Only seed collections that were stored in separate maternal lines (i.e. the seed collected from each maternal plant was stored separately) were used to maximize the representativeness of the source populations. Collection years (1988–2017) and sample sizes varied depending on availability. Two to three seeds per maternal line were germinated by the WASC-K using methods that have been successfully applied to a range of Banksia species in the SWAFR (Crawford et al., 2011) and the resulting seedlings were grown under nursery conditions at the Kings Park and Botanic Garden nursery until they reached sufficient size to sample leaf material without impacting the growing seedlings (∼6 months). One seedling from each maternal line was randomly sampled for this study to maximize the genetic diversity represented in the collections. Finally, leaf material was also sampled from two populations of each B. ilicifolia (BI-1, 2) and B. oligantha (BO-2, 4), with each population consisting of leaf material from ten adult plants, again spaced out over the spatial extent of each population. A summary of all sampling undertaken is given in Table 1 and Fig. 2. The population numbers for both B. cuneata and B. oligantha are so labelled to retain the formal population identifiers for these conservation-listed taxa in DBCA’s internal Threatened and Priority Flora database to facilitate clear interpretation of the results generated in this study for ongoing conservation management purposes.

DNA extraction, library preparation and sequencing

Fresh leaf material was initially dried on silica during sampling and then freeze-dried upon transport to the laboratory. Genomic DNA was extracted using a 2% CTAB protocol (Doyle and Doyle, 1987), modified by the addition of 1% w/v polyvinylpyrrolidone and 0.1 m sodium sulfite to the extraction buffer (Byrne et al., 2001). A total of 383 samples, including 20 technical replicates, were sent to the Australian Genome Research Facility (AGRF, Melbourne, Australia) for library preparation and sequencing. Library preparation was based on a modified double digest restriction-site associated DNA (ddRAD) protocol (Peterson et al., 2012). This genotyping by sequencing approach reduces genome complexity to generate many thousands of anonymous SNPs, largely from the nuclear genome, that can be readily applied to non-model organisms. Initial screening of eight digestion enzyme combinations was followed by library generation for all 383 samples using the most optimal combination, PstI and MspI. Following enzyme digestion, adapter ligation, size selection (280–375 bp) and PCR amplification, the libraries were sequenced in a 150-bp single-end configuration on the Illumina NovaSeq 6000.

SNP calling and data filtering

Sequenced reads were cleaned to a phred quality score >30, demultiplexed and trimmed to 120 bp using the process_radtags pipeline in Stacks v2.62 (Catchen et al., 2013). Optimal running parameters for de novo assembly in Stacks were selected by applying the r80 optimization protocol (Paris et al., 2017) on a subset of 266 samples, which prioritized samples with the highest read depth in each population. The de novo pipeline was then applied to the full dataset using running parameters of n = 4, m = 3 and M = 4 and the populations module was run to output a variant call format (VCF) file with minimal filtering (P = 1, R = 0.2) to maximize output and allow flexibility in downstream filtering. This generated a large dataset of 41 054 SNP loci with 30.11% missing data across the 383 samples.

Custom scripts were used to calculate a measure of genotypic reproducibility based on the 20 pairs of technical replicates and to manipulate the data into a format suitable for filtering using the dartRverse v1.0.2 series of dartR packages (Gruber et al., 2018; Mijangos et al., 2022) in R v4.4.1 (R Core Development Team, 2023). All technical replicates were removed and two datasets were generated. Dataset 1 was intended to assess the variation amongst the three study species and consisted of the in situ collections from all extant populations of B. cuneata (excluding the recent translocation) and the representative populations of B. ilicifolia and B. oligantha (n = 149). Filtering removed all secondary loci and then retained only loci with read depth between 5 and 200x, reproducibility >0.95, call rate >0.80 and minor allele frequency >0.02. Any monomorphic loci resulting from this filtering process were removed. Individuals were subsequently filtered to remove any samples that still retained excessive missing data (call rate >0.50). This generated a dataset of 3756 SNPs with 6.18% missing data and average read depth of 32.12x across 145 samples (samples BC4-06, BC7-08, BI1-03 and BO2-08 were removed due to excessive missing data). Dataset 2 consisted only of B. cuneata and included all samples across the in situ, recent translocation and ex situ collections (n = 324). This dataset was filtered with the same parameters as above, except that loci were filtered with a higher call rate (>0.90) to minimize missing data. This resulted in a dataset of 2664 SNPs with 2.77% missing data and 37.46x average read depth across 322 samples. Again, samples BC4-06 and BC7-08 were removed due to excess missing data.

Data analysis

The evolutionary relationships among the three study species (B. cuneata, B. ilicifolia and B. oligantha) were assessed using both individual and population-level approaches on dataset 1. First, a phylogenetic network was generated using Splitstree v4.15.1 (Huson and Bryant, 2006). Given previous indications of close affinity between B. cuneata and B. oligantha (Broadhurst and Coates, 2004), a network approach was chosen instead of a bifurcating tree-based approach because networks allow for reticulation events and can present more complex evolutionary histories when assessing closely related species, where incomplete lineage sorting or hybridization are more likely. The network was generated based on the genomic proportion of alleles distance (GENPOFAD) metric, as calculated using the POFADINR v1.0 package (Joly et al., 2015) in R. In addition, principal component analysis (PCA) was undertaken to assess the genetic variation among individuals in multivariate space using the gl.pcoa function in dartR and the Bayesian analysis implemented in STRUCTURE v2.1.3 (Pritchard et al., 2000) was used to detect genetic clusters in the dataset (K). The latter analysis was run assuming correlated frequencies, with a 50 000 burn-in period followed by 100 000 Markov chain Monte Carlo replications to test K values ranging from 1 to 6, with 20 iterations for each value of K. This analysis was run without any a priori information regarding population, species identity or geographic location. The optimal K value was evaluated by identifying peaks in ΔK, the rate of change in the log probability of data between successive K values, which is known as the Evanno method (Evanno et al., 2005), and was implemented in Structure Selector (Li and Liu, 2018). Finally, levels of genetic differentiation (F_ST) were estimated between all pairwise population comparisons using the gl.fst.pop function in dartR, which is based on the Weir and Cockerham (1984) method.

To improve resolution of the spatial genetic structure detected within B. cuneata, the PCA and STRUCTURE analyses were repeated for the in situ collection, with and without the recent translocation, using dataset 2. All running parameters were the same, except that the upper K limit in the STRUCTURE analyses was set to the total number of populations plus 1. In addition, pairwise F_ST was recalculated and the correlation between genetic distance (F_ST/1 − F_ST) and geographic distance (log km) was assessed to test for a signal of isolation by distance. Finally, a hierarchical analysis of molecular variation (AMOVA) was undertaken using GenAlEx v6.503 (Peakall and Smouse, 2012) to partition the total genetic variation within and among the major genetic clusters detected by the PCA and STRUCTURE analyses. To assess any temporal variation in these population structuring patterns, the PCA and STRUCTURE analyses were repeated for the historical ex situ collection in a similar manner.

Finally, genetic diversity was estimated for the in situ, recent translocation and ex situ collections of B. cuneata. The percentage of polymorphic loci and allelic richness were calculated using dataset 2 with the gl.report.heterozygosity function in dartR and the allelic.richness function in hierfstat v0.5-11 (Goudet, 2005), respectively. Because these metrics are sensitive to uneven sample sizes, spatial estimates across the in situ collection were calculated by randomly standardizing all populations to the lowest sample size available (n = 7). To minimize the loss of smaller seed cohorts, the standardized sample size was dropped to five (n = 5) for temporal comparisons of diversity across the in situ and ex situ collections. This sample size excluded only the 1988 seed collection from BC-7. All F-statistics (i.e. observed and expected autosomal and SNP heterozygosity, inbreeding coefficient) were also calculated using the gl.report.heterozygosity function but with custom datasets filtered for each population individually from the raw output of Stacks. These filters aimed to retain more population-specific SNPs, remove all missing data and to standardize populations to the same sample sizes as above. These provisions were intended to minimize known biases in these metrics, particularly regarding missing data and large and/or uneven sample sizes, and to allow better comparison among studies (Schmidt et al., 2021; Sopniewski and Catullo, 2024).

RESULTS

Species relationships

The three study species showed a consistent pattern of strong divergence in dataset 1, with no indication of admixture or hybridization, across all analyses. All three species were clearly separated by long branch lengths in the Splitstree network, with no reticulation between species or population clusters (Fig. 3A). Population structure was evident within both B. cuneata and B. oligantha, with the samples of each species separating into distinct clusters. In comparison, little distinction was seen between the two B. ilicifolia populations. Similarly, the PCA showed three major clusters across the first two axes (Fig. 3B). PC1 explained 35.6% of the total variation to separate B. cuneata from both B. ilicifolia and B. oligantha. The latter two species clearly separated on PC2, which explained an additional 13.2% of the total variation. Population structure within B. cuneata was again evident, with samples separating into two closely associated clusters. The STRUCTURE analysis returned an optimal value of K = 3 (Supplementary Data Fig. S2) that corresponded to the three study species with very high assignment (100% for almost all individuals) (Fig. 3C). Pairwise F_ST values were very high in all comparisons between species (>0.7), while values among populations within species ranged from low (F_ST < 0.1) between B. ilicifolia populations to moderate–high (F_ST = 0.1–0.4) within each of B. cuneata and B. oligantha (Fig. 3D).

Fig. 3.

Relationships among in situ collections of the three study species, Banksia cuneata (BC), B. ilicifolia (BI) and B. oligantha (BO), based on various analytical approaches. (A) Splitstree network. (B) Principal coordinates ordination, displaying the first two axes. (C) STRUCTURE barplot, where vertical bars represent individuals that are coloured according to their ancestry assignment to K = 3 genetic clusters. (D) F_ST heat map of all pairwise population combinations within and between species. Population codes are detailed in Table 1.

Population structure within Banksia cuneata

The repeated clustering analyses that focused only on B. cuneata (dataset 2) provided greater resolution for the population structure that was evident for the species in dataset 1. Across the in situ collection, samples clustered tightly in the PCA to indicate hierarchical structuring of four genetic clusters across the first two axes, which together explained 34.9% of the total genetic variation (Fig. 4A). PC1 (25.5% variation) separated the southern populations (BC-8, 12) from the remaining populations further north. PC2 (9.4% variation) separated the northern populations into three clusters that grouped BC-1 and BC-2 together, and separate clusters for each of BC-4 and BC-6. All the samples of any given population clustered together with the exception of BC-7, where samples either clustered with BC-1 and BC-2 or with BC-4; this alternative placement of BC-7 samples could also be seen in the Splitstree network from dataset 1 (Fig. 3A).

Fig. 4.

Principal component analyses for sampled Banksia cuneata populations. (A) All in situ populations. (B) All in situ populations plus the recent translocation, BC-T (white), with the arrow indicating the admixed individual sourced from BC-7. (C) All historical ex situ samples including two extinct populations, BC-X3 (purple) and BC-X5 (dark green). Population codes are detailed in Table 1.

The addition of the recent translocation to the above PCA split the BC-T samples amongst the three northern clusters and were largely consistent with their documented source populations; the ten samples sourced from BC-1 and BC-2 clustered with these populations, as did the four samples from BC-4 and the six samples from BC-6 (Fig. 4B). However, similar to the pattern seen in the in situ samples collected directly from BC-7, the translocation samples that were reportedly sourced from BC-7 did not cluster together; three of the four BC-T samples clustered with BC-1 and BC-2, while the remaining sample fell midway in ordination space between that cluster and the BC-4 cluster. Finally, the PCA on the ex situ collection showed no temporal variation in clustering patterns across collection years (Supplementary Data Fig. S3); thus we present one ordination with all of these historical samples combined (Fig. 4C). This analysis showed near identical clustering patterns to the in situ collection, but with two key observations. First, that seed from the extinct populations, BC-X3 and BC-X5, clustered with BC-4, and second, that all seed sourced from BC-7 between 1988 and 1996 clustered together with seed sourced from BC-1 and BC-2, showing no affinity to BC-4.

The spatial and temporal clustering patterns in the STRUCTURE analyses were consistent with those from the PCAs. The hierarchical structure seen amongst the in situ collection in the PCA was reflected by two peaks in ΔK at K = 2 and K = 5 (Supplementary Data Fig. S4). At K = 2, samples from the southern populations, BC-8 and BC-12, clustered together, as separate from all remaining northern samples, all with near 100% assignment (Fig. 5A). At K = 5, the two southern populations split into two distinct clusters, each with high assignment, and the remaining three clusters matched the northern clusters seen in the PCA, but with low levels of mixed ancestry (Fig. 5B). The mixed signal from BC-7 in the PCA was also seen in the STRUCTURE results, with four samples clustering with BC-4 with >90 % assignment, while the remaining three samples clustered with BC-1 and BC-2 with the same signal of mixed ancestry seen in those populations. The geographic arrangement of these hierarchical structuring patterns across the in situ collection of extant populations is shown in Fig. 5C, D. The recent translocation returned high levels of mixed assignment to the above genetic clusters, consistent with the differing source populations (Fig. 5D). Finally, STRUCTURE assignment from the ex situ collection showed identical clustering patterns to the in situ collection (Supplementary Data Fig. S4) but with the two extinct populations clustering with BC-4 and all historical samples from BC-7 clustering with BC-1 and BC-2, again with the same low level of mixed ancestry seen in these populations (Fig. 5E).

Fig. 5.

Results of STRUCTURE and pairwise differentiation (F_ST) analysis for all sampled extant and extinct Banksia cuneata populations. (A, B) STRUCTURE bar plots for the in situ collection of extant populations, based on K = 2 and K = 5, respectively. (C, D) Spatial distribution of the above clusters at K = 2 and K = 5, including the bar plot for the recent translocation (BC-T). (E) Spatial distribution of STRUCTURE clusters for the historical ex situ collections (K = 5), with inset bar plots for extinct (BC-X3, BC-X5) and extant (BC-7) populations to highlight results that differed from the in situ collection. (F) Heat map of F_ST values between all pairwise combinations of populations from the in situ collection. Blue lines in (C) and (E) indicate hydrological systems, including salt lakes. Population codes are detailed in Table 1.

Pairwise F_ST values among extant populations in the in situ collection were moderate to high (Fig. 5F), consistent with those estimated using dataset 1 (Fig. 3D) and with the above clustering patterns. The highest values occurred between the northern and southern genetic clusters (F_ST range 0.38–0.46), with more moderate values between clusters within the northern region (F_ST range 0.14–0.29) and between BC-8 and BC-12 in the south (F_ST = 0.19). AMOVA partitioned 27.12% of the total genetic variation between the northern and southern clusters, 15.69% among populations within these clusters and 57.19% within populations (Supplementary Data Table S1). Finally, there was a significant positive relationship between genetic and geographic distance (P = 0.02) among the seven extant populations, but the structure detected among the northern populations produced outliers that deviated from this relationship at short- and mid-range geographic distances to weaken the strength of this isolation by distance signal (R² = 0.54; Supplementary Data Fig. S5).

Genetic diversity within Banksia cuneata

Genetic diversity was largely maintained among populations of B. cuneata across both space and time (Fig. 6). Spatially across the in situ collection, the percentage of polymorphic loci and allelic richness varied in a similar pattern, with the highest values exhibited in BC-1 and BC-7, with lower values for the remaining populations (Fig. 6A, B). This patterning had no correlation with population size. Autosomal heterozygosity showed some variation across populations but with overlapping standard errors (Fig. 6C). Autosomal estimates were highly correlated with SNP-based heterozygosity estimates (Supplementary Data Fig. S6), which ranged from 0.334 ± 0.005 to 0.389 ± 0.006. For most populations, inbreeding coefficients did not deviate significantly from zero, with the exception of a significantly positive value for BC-7 (Fig. 6D). Diversity estimates in the recent translocation were higher than in all seven in situ populations in terms of the percentage of polymorphic loci and allelic richness, but not heterozygosity. The recent translocation also exhibited a significantly positive inbreeding coefficient. Diversity levels across the in situ and ex situ collections showed minor variation over time, but overall did not show marked increases or decreases in diversity between 1988 and 2021 (Fig. 6E–H). The inbreeding coefficient did exhibit some temporal variation, with a high positive value recorded in the 1996 seed cohort of BC-1 that became non-significant in the collection years thereafter, including the 2021 in situ collection, and an increase in the inbreeding coefficient for BC-7 following the 1996 collection. The two extinct populations were only sampled from a single time point (1988) and exhibited similar diversity estimates to all other seed collections from that year (Fig. 6E–H).

Fig. 6.

Spatial and temporal variation in genetic diversity parameters (percentage of polymorphic loci, allelic richness, autosomal heterozygosity and the inbreeding coefficient, F_IS) for Banksia cuneata. (A–D) Spatial variation in genetic diversity across all extant populations (in situ collection; grey) and the recent translocation (pink), with populations in each plot ordered from lowest diversity to highest. (E–H) Temporal variation in genetic diversity across historical ex situ (1988–2017) and current in situ (2021) collections. Error bars have been omitted for visual clarity but populations with significant temporal changes are indicated with red lines. The two extinct populations (blue) were only available for a single time point (1988) but are included on all four plots for comparison. Population codes are detailed in Table 1.

DISCUSSION

This study reports on a comprehensive reassessment of genetic variation in B. cuneata and relatives that was necessary to resolve mixed signals from previous genetic studies and gain insight into historical management practices. Through the application of high-resolution SNP data across both in situ and ex situ collections that span 1988 to 2021, we provide clarity regarding three key aspects of conservation: delineation of appropriate conservation and management units, reconstruction of translocation history, and the utility of ex situ germplasm collections. The knowledge gained from this study will allow informed decision-making for the ongoing conservation management of B. cuneata. Moreover, this study presents a rare example of an active, long-term translocation program that provides key insights for the utility of genetic data in both current and historical conservation contexts.

Clarification of conservation units

The network, ordination and Bayesian analyses in our study presented a clear and consistent signal of three divergent lineages in the Banksia subgenus Isostylis that were consistent with B. cuneata, B. ilicifolia and B. oligantha under their current circumscriptions. This contrasts with a previous allozyme study that suggested a close genetic affinity between the southernmost B. cuneata populations and B. oligantha (Broadhurst and Coates, 2004), which, in conjunction with their parapatric geographic distribution and morphological similarities, had raised concerns regarding possible conspecificity or hybridization. Even with strong population structure within these two species, our data showed that levels of interspecific differentiation were two to three times greater than those within. Moreover, there was no evidence of admixture or reticulation in the clustering analyses that might have indicated interspecific hybridization occurring between them. Taking an integrative taxonomic approach to species delimitation (Dayrat, 2005), the congruence of our molecular data with the morphological data underlying the original taxonomy provides confidence that B. cuneata and B. oligantha are distinct species and that the southernmost populations of B. cuneata are correctly assigned to species. Thus, each of these threatened species should remain as separate conservation units under their current taxonomic circumscriptions, each with their own conservation ranking and management plans.

Translocation history of Banksia cuneata

The recent translocation presents a typical example of the genetic signals to be expected from the early stages of a mixed-source translocation. The varied genetic assignments of individuals in the clustering analyses confirmed the documented sources of most founding plants, and this mixed composition has resulted in an increase in allelic diversity (i.e. allelic richness and the percentage of polymorphic loci) relative to all source populations. This is promising in demonstrating that these founding plants from differentiated source populations have successfully introduced a new assemblage of alleles to the translocation, which is a key goal of mixed-source translocations to increase resilience and evolutionary potential (Weeks et al., 2011; Frankham et al., 2017; Rossetto et al., 2023). The lack of a similar increase in heterozygosity is not surprising, given that heterozygosity did not vary substantially among the source populations and the new assemblage of alleles has not had chance to recombine. The elevated inbreeding coefficient, which would usually be considered concerning for B. cuneata (Coates and Sokolowski, 1992) and long-lived, outcrossing Banksia species in general (Scott, 1980; Carthew et al., 1996; Heliyanto et al., 2005), is a common artefact in mixed-source founder populations due to the Wahlund effect, where the translocation is treated as a single population, despite comprising individuals from differentiated populations (Fant et al., 2013; Fotinos et al., 2015; Monks et al., 2021; Dillon et al., 2023). We expect that heterozygosity will increase and the inbreeding coefficient will decrease as this new assemblage of alleles recombine into new genotypic combinations with future admixed generations of the translocation.

Whilst the recent translocation was expected to be the only mixed-source translocation undertaken during the conservation management of B. cuneata, the marked spatial and temporal changes detected in the genetic composition of population BC-7 strongly suggest that this population has been augmented using seed or seedlings sourced from BC-4 at some point in time. The consistent genetic composition of BC-7 seen within our ex situ collection predates any human-mediated conservation intervention and thus demonstrates the historical genetic integrity of BC-7 as naturally affiliated with populations BC-1 and BC-2. This result is independently supported by the genetic analysis of this population in the early 1990s (Coates and Sokolowski, 1992). As such, the detection of BC-4-sourced plants in our in situ collection of BC-7 represents a relatively recent change in the composition of this population. Biological explanations for this mixed composition are inconsistent with pollen dispersal from BC-4 to BC-7 as this would have resulted in admixed individuals rather than individuals with alternate genetic signatures. Repeated, recent seed dispersal between these disjunct populations would result in the pattern seen but is highly unlikely given that B. cuneata has passive seed dispersal (i.e. by gravity) and seeds do not disperse far from the parent plant (Burgman and Lamont, 1992). Thus, we conclude that the introduction of BC-4 plants to this population was the result of augmentation during previous conservation management of B. cuneata.

Whilst records are patchy, many B. cuneata populations were augmented in the mid-1990s, including BC-4 and BC-7 (Department of Environment and Conservation, 2009). Over this lengthy time period, the populations of B. cuneata have been given various numerical and text labels across both the grey and white literature (Coates and Sokolowski, 1992; Broadhurst and Coates, 2004; Department of Environment and Conservation, 2009), including at least three documents where BC-7 is referred to as ‘population 4’ (Coates and Sokolowski, 1992; Maguire and Sedgley, 1997; McNee, 1997) (Supplementary Data Fig. S1). When all other documentation indicates that the augmentation practices of B. cuneata were undertaken using single, local sources, it is reasonable to conclude that this mixed planting at BC-7 was likely an unintentional consequence of inconsistent population labelling rather than an intentional mixed-source translocation. The alternate genetic assignments of the individuals in BC-7 demonstrate that this mixed planting has not yet resulted in any admixed mature plants, which is not surprising given that the species has poor natural recruitment in the absence of fire (Burgman and Lamont, 1992) and there are no records of fire occurring at BC-7 following the augmentation. However, the one admixed BC-7 seedling detected in the recent translocation provides reasonable evidence that the two sources of plants at BC-7 have interbred to produce admixed seed. While unintended, this demonstrated mixing of genetically differentiated sources is encouraging for the prospects of the mixed-source recent translocation. It also indicates that in addition to the mixed genetic composition of BC-7, the genetic composition of any seed collected from BC-7 since the late 1990s will also be mixed with that of BC-4.

Aside from BC-7, the remaining populations of B. cuneata showed a consistent signal of population structure over time, confirming that all other historical augmentations were undertaken using locally sourced seed. For these populations, without documented tracking of planted individuals it is difficult to know what impact such augmentations may have had. However, our data showed that one population, BC-1, exhibited an elevated inbreeding coefficient in the 1996 seed cohort that was not detected in the seed cohorts of later years. This is notable because the species is largely outcrossing (Coates and Sokolowski, 1992) and no indication of inbreeding was detected in the other seed collections or the adult plants of any non-admixed extant populations. This apparent temporal variation in inbreeding is consistent with mating system studies that showed decreased outcrossing, as well as elevated inbreeding coefficients in seed versus adults, from this same roadside population around the same 1990s period (Coates and Sokolowski, 1992), followed by a reduction of inbreeding in that populations’ seed cohorts in subsequent years (Coates et al., 2013). This may simply represent a natural variability in outcrossing in highly disturbed populations of the species (Coates and Sokolowski, 1992; Coates et al., 2013) but may be linked to the repeated augmentation of this population in the mid-1990s (Department of Environment and Conservation, 2009) that may have sufficiently improved the reproductive pool to alleviate the trajectory of inbreeding in this population.

Management units in Banksia cuneata

With clarity of both conservation units and human-mediated changes in genetic composition following historical management practices, previous confusion regarding the spatial distribution of genetic variation in B. cuneata can also be resolved. Our data detected strong, hierarchical structure across the range of B. cuneata, which contrasts with a previous RAPD study that did not detect any population structure in the species (Maguire and Sedgley, 1997). The clear distinction between B. cuneata and B. oligantha in our study also rules out hybridization as an explanation for the divergence between the major northern and southern clusters in B. cuneata, as previous allozyme data had implied (Broadhurst and Coates, 2004). However, the differentiation we detected within the northern cluster is entirely consistent with patterns previously documented using allozymes. Whilst the southern populations were not known at the time, Coates and Sokolowski (1992) described ‘eastern’ and ‘western’ groups among the northern populations of B. cuneata, with further differentiation of populations within the western group (Supplementary Data Fig. S1). This exactly matches the structure patterns in our data between the western BC-4 and BC-6 populations that are distinct from the more homogeneous eastern cluster consisting of BC-1, BC-2 and BC-7. The southern populations were known by the time of the RAPD study so the inability of these markers to detect even the major north–south differentiation within B. cuneata is striking. This may be attributable to the dominant (e.g. RAPDs, AFLPs) versus codominant (e.g. allozymes, microsatellites, SNPs) nature of the markers involved. Dominant markers only allow the estimation of genotypic but not allelic frequencies (Ouborg et al., 1999), such that analyses relying on the latter, for example the AMOVA undertaken by Maguire and Sedgley (1997), may be flawed. In addition, the low levels of diversity detected by the allozyme markers (Coates and Sokolowski, 1992) may explain their lack of discriminatory power to clearly delineate B. cuneata and B. oligantha. The application of high-resolution SNP markers in the current study provides much-needed confirmation of strong population structure across the extant range of B. cuneata, highlighting that all populations hold unique genetic variation for the species and are critical from a conservation perspective. Moreover, the high level of differentiation detected between the northern and southern groups also supports the historical decision to recognize them as separate management units.

In general, the strong population structure detected across the range of B. cuneata is consistent with studies of other bird-pollinated species in the SWAFR (Sampson et al., 2014; Tapper et al., 2014; Bradbury et al., 2019). The high differentiation between the northern and southern populations can readily be explained by dispersal limitations between these disjunct geographic regions that are separated by >40 km, greatly exceeding demonstrated pollen dispersal distances by honeyeaters in the region (Byrne et al., 2007). However, the additional substructure detected within the northern cluster is not geographically intuitive. This is best demonstrated by the two geographically closest populations, BC-1 and BC-2, which clustered together and are nearest to BC-6 (<10 km in an east to west direction) but were genetically closer to pre-management BC-7, which is >20 km away to the south-east (Fig. 2). Moreover, the differentiation of populations BC-4 and BC-6 appears inconsistent given that they are not much further away from each other than populations BC-1 and BC-2 are from BC-7. This indicates the presence of barriers to gene flow in addition to geographic distance, and with a clearer understanding of the species’ translocation history we can rule out any human-mediated explanation for these patterns. Coates and Sokolowski (1992) noted that these seemingly odd genetic breaks across the northern distribution of B. cuneata coincide with the extensive network of salt lakes in the region (Figs 2 and 5) and suggested that these saline water bodies present unfavourable habitat for birds, limiting their movements across the landscape and resulting in ecological barriers to pollen dispersal across the range of the species, even over relatively short distances. Whilst the disparity in population structure among the early genetic studies raised uncertainty in this unusual patterning across the northern populations of B. cuneata, the congruence with our high-resolution data provides confidence in this genetic result and strengthens support for this ecological explanation.

Based on this interpretation of geographic distance and salt lakes driving patterns of gene flow in B. cuneata, the genetic affinity of extinct populations can also be inferred. The two extinct populations we were able to sample using historical seed collections (BC-X3 and BC-X5) clustered with BC-4, which is consistent with their close geographic proximity. Also due to very close geographic proximity, we expect that BC-X9 would be genetically affiliated with BC-8 and similarly that BC-X11 would cluster with BC-X3, BC-4 and BC-X5. Finally, BC-X10, being geographically close and on the same side of the nearby salt lake system, would likely be genetically affiliated with BC-6. These assumptions would indicate that, at least since the species came under conservation management, there has not been a loss of any major genetic subpopulation of B. cuneata and that all subpopulations are still represented in extant and/or ex situ seed collections.

Whilst our SNP-based patterns of population structure largely agreed with previous allozyme data, our estimates of genetic diversity did not. Previous allozyme studies have remarked that B. cuneata exhibited low genetic diversity for an outcrossing plant species (Coates and Sokolowski, 1992; Broadhurst and Coates, 2004; Coates et al., 2013), whilst RAPD markers indicated that diversity was high (Maguire and Sedgley, 1997). With the relatively new measures of autosomal diversity currently being generated using SNP data, it is challenging to interpret our values in the context of comparison with other marker types until more SNP studies utilize these new methods (Schmidt et al., 2021; Black et al., 2023; Sopniewski and Catullo, 2024). However, given that our autosomal estimates were correlated with the more widely applied SNP-based estimates, our data do not indicate that B. cuneata is genetically depauperate. Moreover, the relative temporal stability of genetic diversity across our ex situ and in situ collections indicates that genetic erosion is not an urgent conservation concern for B. cuneata. Given the naturally patchy distribution of B. cuneata and the presence of ancient salt lakes in this landscape (George et al., 2008), the strong spatial genetic structure in B. cuneata is likely driven by historical processes of isolation and drift, although these ecological limitations may have been exacerbated by relatively recent land clearing that has led to increased fragmentation and salination across the SWAFR. Similarly, our diversity estimates showed no correlation with population size, further indicating that genetic variation is largely driven by historical factors, rather than any factors that have driven recent decline. Such retention of genetic diversity in B. cuneata, despite significant population decline, is likely attributable to the species being long-lived and the storage of multiple generations of seed in the tree canopy providing some buffer for genetic drift and/or inbreeding to significantly impact diversity (Broadhurst and Coates, 2004). This is further evidenced during our study by the absence of a signal of low diversity or increased inbreeding in 1988 cohorts of BC-X3 and BC-X5 prior to their subsequent local extinction. This is encouraging from the perspective of having sufficiently diverse germplasm collections to conserve the species’ evolutionary potential through ongoing conservation management (Frankham et al., 2017). However, it also provides a clear warning that genetic signals of population decline, such as genetic erosion or increased inbreeding, may not present in B. cuneata, or other long-lived outcrossing species, prior to rapid local extinction.

Ex situ seed collections retain valuable stores of genetic diversity

The ex situ seed collections in our study were primarily utilized to track any changes in genetic diversity within B. cuneata populations over time, with the assumption that they represented the genetic composition of each population at various time points. The relative consistency between our in situ and ex situ collections, as well as with previous genetic studies using historical adult and seed samples (e.g. Coates et al., 2013), validates this assumption. Thus, these seed collections also provide insight into the benefits of best-practice seed collecting in the context of ex situ conservation. Four common concerns regarding long-term ex situ storage of seed collections are that (1) older seed collections, prior to the establishment of standardized germplasm guidelines, may not adequately capture the genetic diversity of natural populations (Martyn Yenson et al., 2021; van der Merwe et al., 2023), (2) genetic diversity may decline over time during long-term ex situ storage (Richards et al., 2010), (3) utilization can be compromised when information about a seed collection is lacking, such as the number of maternal lineages (van der Merwe et al., 2023), and (4) seed collections and subsequent controlled growing conditions may evade natural selection processes, risking the use of inbred or unfit seed in translocations (Basey et al., 2015).

We found that genetic diversity was consistent across collection years, indicating that even the oldest (1988) seed collections were undertaken effectively so as to capture the genetic diversity of each population, and that there has been no major loss of genetic diversity in these collections over nearly four decades in storage. This is a positive reflection of the best-practice methods that have long been used to collect and store seed from threatened species in Western Australia, including the separation of maternal lines and regular viability testing (Crawford et al., 2007). Moreover, with the exception of the 1996 BC-1 seed cohort, we did not detect evidence of inbreeding in the seed collections, which is consistent with the high levels of outcrossing determined for the species (Coates and Sokolowski, 1992; Coates et al., 2013) and demonstrates that the translocation management of B. cuneata poses a low risk of introducing excessive inbred individuals to existing or new populations. These results are encouraging for the continued use of these ex situ collections for future recovery actions in B. cuneata that will ensure populations harbour sufficient genetic variation to avoid inbreeding and facilitate evolutionary potential. Beyond B. cuneata, these results also promote the benefits of using standardized germplasm collection guidelines in a global context and support the continued collection and storage of ex situ seed collections for the threatened flora of Western Australia.

Conservation implications and research priorities

Clarity regarding appropriate conservation and management units, as well as improved understanding of the genetic composition of all extant populations and seed collections of B. cuneata, provides confidence in decision-making for the species’ ongoing conservation management. Most of the extant populations are declining with little to no natural recruitment and thus ongoing management should focus on increasing population sizes and the number of populations. This may occur via continued augmentation or careful fire management to stimulate natural recruitment processes within existing populations, as well as increasing the number of populations via additional translocations. There has been much debate over the risks of inbreeding versus outbreeding depression in the use of germplasm for translocation (Frankham et al., 2011; Weeks et al., 2011; Ralls et al., 2020; Kardos et al., 2021). Our data indicate that genetic erosion is not a current concern for B. cuneata and that the existing ex situ collections are both sufficiently diverse and abundant to enable continued independent management of the northern and southern management units without risking inbreeding depression. Mixing these units, on the other hand, may pose a risk of outbreeding depression given the timeframe for separation without gene flow between populations is likely to be greater than 500 years (Frankham et al., 2011). Thus, at this time we do not recommend mixing between the disjunct northern and southern management units and advise that any population augmentations continue to be done using single-sourced, local seed, while any introduction translocations are mixed-source within each of the northern and southern management units. With known and inferred knowledge of the genetic affinity of all extinct populations, such translocations can confidently include ex situ seed from these extinct sources to maximize diversity. However, if concerns arise surrounding germplasm availability or adaptive capacity, these management units could be used as sources of genetic rescue for one another with careful monitoring for any indications of reduced fitness. Finally, population BC-7 presents an intriguing scenario because the BC-4-sourced plants at this site are reproductively mature and make up half the population. As such, we recommend that all plants are retained, and the population is treated as a mixed-source translocation with monitoring of any new recruits for interbreeding between the two sources. Moreover, any augmentation of BC-7 can make use of the admixed ex situ seed collections made after the late 1990s but any new translocations wanting to source the unique genetic variation of pre-management BC-7 should utilize the ex situ seed collections made prior to the late-1990s.

Our study has also indicated further avenues for future conservation-focused research for each of the three study species. For B. cuneata, our genetic assessment of the recent translocation showed a promising founding population, but its success will be contingent on the trajectory of subsequent generations. Being established in 2018, the founding plants are now reaching reproductive maturity, and with detailed documentation of all individuals, the reproductive, ecological and genetic consequences of this mixed-source translocation can be monitored in future studies. For B. oligantha, the two populations sampled in our study are highly differentiated (F_ST = 0.4), which is consistent with their substantive geographic disjunction (∼90 km apart). The remaining populations occur within this geographic span, but with the presence of extensive salt lake systems throughout the region, we predict that these populations will be differentiated in a similar manner to that seen within the northern region of B. cuneata. A complete genetic study of all populations and any available ex situ seed collections would aid in decision-making for establishing single or mixed-source translocations for B. oligantha. Finally, while B. ilicifolia is not a threatened species, it is a dominant component of the Banksia woodlands that are listed as a Threatened Ecological Community (TEC) in southwestern Australia. These woodlands occur along a broad climate gradient in the SWAFR, which is predicted to experience increased daytime temperatures and reduced rainfall in future projections (Andrys et al., 2017). The apparent greater connectivity among populations seen in our limited sampling of B. ilicifolia indicate that it may be an excellent candidate for genetic research on climate adaptation that would aid climate-informed management of this TEC and other Banksia species in the SWAFR.

Conclusions

Our comprehensive genetic assessment of the natural population dynamics and historical management of B. cuneata provides clarity on previous confusion, as well as new knowledge, to inform the ongoing conservation of this iconic threatened species. More broadly, our study provides a rare examination of a long-term translocation program and highlights the value of genetic data to improving conservation outcomes. While it is now well recognized that genetic data provide critical knowledge for undertaking conservation management, we show that it can also provide insights into historical management actions, either to confirm available records or to reveal key information where records have become disordered or lost. Our clarification of historical actions through genetic data also highlights the importance of detailed and consistent record-keeping for tracking the metadata of populations, germplasm collections and individual plants that is future-proofed for long-term storage and utilization. Finally, our study also demonstrates that molecular methods inevitably change and improve over time. As such, the application of genetic data to any given threatened species or translocation program should not be considered a one-time-only input, but rather an ongoing requirement to revisit older methods and provide updated, best-practice data over the lifetime of any conservation program.

SUPPLEMENTARY DATA

Supplementary data are available at Annals of Botany online and consist of the following. Table S1: summary of AMOVA for population groupings of Banksia cuneata. Figure S1: stylized summary of population genetic studies undertaken on Banksia cuneata to highlight major similarities and differences in sampling, molecular markers and results. Figure S2: ΔK plot for STRUCTURE analysis on the three study species. Figure S3: principal component analysis for ex situ historical seed collections of Banksia cuneata, with populations indicated by colour and collection year indicated by shape. Figure S4: full suite of STRUCTURE analyses undertaken on various Banksia cuneata collections. Figure S5: isolation by distance plot for the correlation between linearized geographic distance (km) and genetic distance (F_ST) between all population pairs of Banksia cuneata. Figure S6: correlation between SNP and autosomal heterozygosity.

FUNDING

This project was supported by the Threatened Species Initiative (TSI; https://threatenedspeciesinitiative.com/), the Department of Biodiversity, Conservation and Attractions (DBCA), and resources provided by the Pawsey Supercomputing Centre.

AUTHOR CONTRIBUTIONS

R.M.B., L.M., A.D.C., M.E. and M.B. conceived and designed the study. Field collections were undertaken by R.M.B., L.M., A.D.C. and M.E. A.D.C. and L.M. coordinated the seed germination and seedling rearing, while R.M.B. coordinated the molecular laboratory work, with N.D. performing DNA extractions. Data analysis was undertaken by R.M.B. and N.D. R.M.B. wrote the draft manuscript, and all authors contributed to editing the manuscript.

DATA AVAILABILITY

Demultiplexed sequencing data are available at the NCBI Sequence Read Archive under Project PRJNA1261784.