Embracing biodiversity: multispecies population genomics of leafless Bossiaea species shows novel taxa, population dynamics and conservation strategies

Field sampling

We sampled the Bracteosa and Fragrans subgroups extensively, totalling 288 cladode tissue samples across 27 Bossiaea populations in south-eastern NSW, Vic. and Tasmania (Tas.) to investigate genetic variation and population dynamics. Sampling included all known populations of B. grayi, B. bombayensis, B. vombata, B. fragrans and B. milesiae, and several populations of the more widespread B. bracteosa, to capture representative diversity (Table 2, Fig. 1). In our sampling approach, we inadvertently collected two additional genetic groups, designated as genetically distinct population (GDP) 1 and GDP 2. Specimens of GDP 1 were obtained from two closely situated sites near the Cobberas Mountains, Vic. (Sites 17 and 18, ~4 km apart; see Fig. 1), and were initially identified as B. grayi and B. bracteosa respectively. Specimens of GDP 2 were collected as B. riparia from Oallen Ford (Site 2) and within Kosciuszko National Park (Site 11) in NSW. Another leafless Bossiaea from the Ensata subgroup, B. riparia, which is morphologically distinct (Thompson 2012) but within the distributional range of the Bracteosa and Fragrans subgroups, is included as an outgroup.

The collection strategy involved gathering 5–10 cm of tissue from each individual plant, ensuring spatial separation of sampled plants and recording the geographic coordinates (latitude and longitude) of each plant. To ensure the capture of genetic diversity across differently sized populations for the threatened species, sampling numbers were dependent on the number of individuals known and distribution patterns. After collection, the samples were stored at −80°C for a minimum of 12 h, then freeze-dried before storing in silica gel to ensure preservation until DNA extraction.

DNA extraction and DArTseq analysis

Genotyping was conducted using medium-density DArTseq, a reduced representation sequencing method implemented by Diversity Arrays Technology Australia Pty Ltd (Canberra, ACT, Australia). The DArTseq method involves a genome restriction digest followed by sequencing of the digested products by using an Illumina instrument, with SNPs called using proprietary analytical pipelines by DArT Pty Ltd (Jaccoud et al. 2001; Kilian et al. 2012).

Relationships among species and populations

Analyses of genetic relationships among the 288 Bossiaea samples were undertaken following methods from Rossetto et al. (2018), with some modifications. Our initial dataset had 81,158 loci. To enhance data quality, we eliminated loci with a DArT reproducibility score of 0.96 or lower, as well as monomorphic loci. Additionally, we selected a single SNP per locus. Subsequently, loci with more than 30% missing data across all genetic groups were excluded. We also removed samples with more than 30% missing loci to further refine the dataset. After implementing these quality-control measures, the resultant dataset utilised for subsequent analyses comprised 52,539 loci and 283 samples.

These SNP analyses included the estimation of pairwise kinship between individuals, principal-component analysis (PCA; Jombart 2008), estimation of fixation index (F_ST; Zheng et al. 2012), calculation of basic diversity statistics (Keenan et al. 2013), a SplitsTree phylogenetic network analysis (based on the Euclidean distance among genotype scores; Huson 1998) and SVDQuartets phylogenetic estimation (Chifman and Kubatko 2014). These analyses were performed in R for macOS (ver. 4.3.1, R Foundation for Statistical Computing, Vienna, Austria, see http://cran.r-project.org/bin/macosx/). We used in-house developed R code for the RRtools package, which is available on Github (see https://github.com/jasongbragg/RRtools) to filter and analyse the genotype data.

We estimated pairwise kinship between individuals using the PLINK identity by descent (IBD) method (Purcell et al. 2007; Chang et al. 2015) implemented in the snpgdsIBDMoM function of SNPRelate (ver. 1.36.0, see http://rdrr.io/bioc/SNPRelate/; Zheng et al. 2012). PLINK IBD estimation compares genotypes at specific SNP positions (shared loci with minor allele frequency (MAF) of ≥0.05) to determine the likelihood of shared alleles and estimate genetic relatedness between individuals in a dataset. Kinship was calculated within genetic groups (McMaster et al. 2024). We considered individuals to be genetically identical where the estimate of kinship exceeded a threshold value of 1 ÷ 2^(3÷2) (Manichaikul et al. 2010). Where pairs of individuals exceeded this threshold, we removed the sample with lower data quality from the data set for subsequent SVDQuartets and diversity analysis.

Fixation index (F_ST) was calculated between subpopulations of Bossiaea by using the (Buckleton et al. 2016) method implemented in the snpgdsFst function of SNPRelate (MAF ≥ 0.05). Subpopulations with three or fewer individuals after clone removal were excluded. The fixation index (F_ST) measures the degree of genetic differentiation between populations, with higher F_ST values indicating a greater divergence in allele frequencies owing to limited genetic drift or selection.

Diversity metrics were computed for individual genetic groups within each site and overall. This involved excluding clones, subsetting to the target genetic group, selecting loci with less than 30% missing data and a MAF of ≥5%. Subsequently, the basicStats function of diveRsity (ver. 1.9.90, see http://rdrr.io/cran/diveRsity/) was used to obtain allelic richness (Ar), observed heterozygosity (H_O), expected heterozygosity (H_E) and inbreeding coefficient (F_IS; Keenan et al. 2013).

We used the adegenet package (ver. 1.7-22, see http://rdrr.io/cran/adegenet/; Jombart 2008) to conduct PCA. The goal was to uncover and illustrate the underlying patterns of genetic diversity and population structure within our dataset. To achieve this, we visualised the first six principal components (PC1–PC6) in our analysis. We included more PCs than usual because of the involvement of seven different distinct genetic groups, allowing us to visualise how the observed distinctions contribute to the overall variance in the data.

The phylogenetic network was produced using the R package RSplitsTree (ver. 0.1.0, see http://rdrr.io/github/IVS-UZH/RSplitsTree/) and visualised in R using ggplot2 (ver. 3.4.2, see https://ggplot2.tidyverse.org; Wickham 2016) and tanggle (ver. 1.0.0, see http://rdrr.io/github/KlausVigo/tanggle/). Phylogenetic network analysis was used because it can capture and represent complex evolutionary relationships, such as reticulate evolution, hybridisation and horizontal gene transfer, which may not fit into a strictly tree-like structure (Huson 1998).

We estimated phylogenetic relationships of subpopulations using SVDQuartets in PAUP* (ver. 4.0a, see http://paup.phylosolutions.com/; Cummings 2004) and visualised the tree in R using ggtree (ver. 3.2.1, see https://rdrr.io/bioc/ggtree/; Yu et al. 2017). The SVDquartets method estimates a tree by exhaustively sampling combinations of four taxa from the data matrix, inferring a tree for each quartet and then using a quartet assembly algorithm to combine all sampled quartets into a species tree. SVDquartets is preferred for producing phylogenies from SNP data over other methods because of its ability to handle incomplete lineage sorting, its quartet-based approach for efficient analysis of large datasets, its direct inference of the species tree without relying on gene-tree estimation and its robustness to model violations commonly encountered in SNP data (Leaché and Oaks 2017). Input alignments were produced from the dart2svdquartets function in RRtools (ver. 0.1, see http://github.com/jasongbragg/RRtools/; Rossetto et al. 2018). Populations initially assigned during sampling were renamed into their putative subpopulation groups on the basis of genetic data (putative species and site). SVDQuartets was run under the multi-species coalescent model, standard bootstrap method with 1000 replicates, 100,000 random quartets and population taxonomic partitioning (Potter et al. 2021; Whitelaw et al. 2023). The final consensus tree was rooted using B. riparia (including GDP 2).

Distribution maps

We utilised the R package galah (ver. 2.0.2, see http://rdrr.io/github/AtlasOfLivingAustralia/galah/) to retrieve all records from the Atlas of Living Australia for Bossiaea fragrans, Bossiaea grayi, Bossiaea bombayensis, Bossiaea milesiae, Bossiaea riparia, Bossiaea bracteosa and Bossiaea vombata. Subsequently, we filtered these records to include only those records sourced from the Australian Virtual Herbarium, NSW BioNet Atlas and PlantBank Records.

Following this, we reclassified certain records on the basis of our analysis; records of B. vombata with longitude less than 146 were reclassified as B. vombata subsp. orientalis, whereas those with longitude greater than 146 were reclassified as B. vombata subsp. vombata. Additionally, records of Bossiaea grayi with latitude less than −36 were also reclassified as B. vombata subsp. orientalis (see Taxonomy).

Identification of genetic groups

The genetic groupings in the data predominantly matched the expected classifications on the basis of their identification during collection. However, we also identified two additional genetic groups (referred to as GDP 1 and GDP 2) that did not neatly correspond to pre-existing taxonomic categories. The first lineage, GDP 1, was sampled from two sites near the Cobberas Mountains, Vic., that were proximate (Sites 17 and 18, ~4 km apart; Fig. 1) and were genetically similar. A previous collection from one site had been determined as B. grayi, and the other as B. bracteosa. GDP 1 is genetically similar to B. vombata, which, by current definition, occurs only in the Wombat State Forest in southern-central Vic., a considerable distance (~640 km) from the Cobberas (Ross 2008). The shared ancestry is indicated by the lineages clustering in the phylogenetic network and PCA (Fig. 2a–c) and clade forming in the SVDquartets phylogeny (Fig. 2e). Genetic separation correlated with the geographic distance is evident, and when coupled with observed morphological differences (longer calyx lobes, standard, claw and keel in B. vombata than in GDP 1 plants), this suggests that both lineages might be in the early stages of speciation. On the basis of the morphological resemblance of GDP 1 to B. vombata, we propose naming GDP 1 as Bossiaea vombata subsp. orientalis (described below in Taxonomy), thereby recognising the Wombat State Forest population as B. vombata subsp. vombata.

Genetically distinct Population 2 comprises individuals of B. riparia found near Oallen Ford (Site 2) and in Kosciuszko National Park (Site 11) in NSW, which are geographically distant from other included populations of B. riparia in Vic. (Site 16) and Tas. (Sites 26 and 27). The phylogenetic network and SVDquartet phylogeny shows that GDP 2 forms a cluster sister separate from the cluster of B. riparia from Vic. and Tas. (Fig. 2d, e). However, GDP 2 collections could not be separated morphologically from other collections of B. riparia (see Taxonomy below). This species is noted to have variable morphology (Thompson 2012), and our genetic findings support the concept that this widely distributed species contains genetic as well as morphological variability. Overall, the genomic data support the concept of B. riparia as a single species, despite the geographic separation and variation in habit and floral characteristics in its populations. However, the limited sampling of the Ensata subgroup (only one of five species included) restricts our ability to definitively identify the taxonomic status of GDP 2. Future studies should incorporate the other species from this subgroup to better delineate species boundaries and refine taxonomy within this group.

The genomic data also identified a genetic relationship inconsistent with previous taxonomic concepts, namely the previously described species B. milesiae from Brogo River and B. fragrans from Abercrombie Karst Conservation Area (~290 km apart; Fig. 1). Although these two species are morphologically different, with B. fragrans more commonly having multiple flowers per node, wider reproductive branches of cladodes and shorter stipes on the pods than does B. milesiae (McDougall 2009), the level of genetic separation between B. fragrans and B. milesiae is comparable to that between B. vombata and GDP 1, suggesting that they are better recognised as subspecies rather than species. This is exemplified by PCA outputs showing all individuals clustering together and separating only in further principal components (PC5 v. PC6; Fig. 1c) and the phylogenetic network showing a divergence between B. milesiae and B. fragrans similar to that of the two subspecies of B. vombata (Fig. 2d). To ensure consistent and reliable taxonomic assignment among the leafless Bossiaea species, we therefore propose changing the rank of B. milesiae to a subspecies of B. fragrans and supply the required combination below in the Taxonomy section.

Our genetic results indicated that B. bracteosa, which is described as endemic to the Australian Alps of north-eastern Vic., mainly from Dargo to Mount Hotham area, is genetically distinct from other taxa in the Bracteosa subgroup. The results here support the current species concept of B. bracteosa, which was based on a re-circumscription of B. bracteosa sensu lato (also recognising B. fragrans, B. milesiae, B. bombayensis and B. grayi) by McDougall (2009).

Bossiaea riparia was included in this study as an outgroup to examine the relationships among B. bracteosa and its related species. However, our genetic results did not identify this species as highly distinct from other studied species. The outgroup was chosen on the basis of Thompson’s (2012) morphology-based classification of Bossiaea that placed B. riparia in the Ensata subgroup, whereas B. bracteosa, B. vombata, B. bombayensis and B. grayi belong to the Bracteosa subgroup, and B. fragrans and B. milesiae to the Fragrans subgroup. Here, our analyses showed that B. riparia is potentially a sister lineage to B. bracteosa or any of the study species, indicating that morphology-based categorisations will not always reflect genetic relationships in this group of species. Future studies should consider using multiple potential outgroup candidates to reduce the risk of encountering similar challenges.

The findings highlighted how genetic variation aids in resolving relationships, especially in cases of allopatric speciation within geographically isolated lineages, which may not be clearly discernible or consistent solely on the basis of morphology. Our study showed a shallower level of genetic differentiation between B. milesiae and B. fragrans, leading us to conclude that B. milesiae should be classified as a subspecies under B. fragrans (described below as B. fragrans subsp. milesiae). Furthermore, we identified two genetically distinct populations (GDPs) within samples initially labelled as B. bracteosa and B. grayi from sites in Cobberas, Vic. (Sites 17 and 18, designated GDP 1, described below as Bossiaea vombata subsp. orientalis, Fig. 2b), and B. riparia from near Oallen Ford and in Kosciuszko National Park (Sites 2 and 11, GDP 2; Fig. 2c). These findings underscore the importance of genetic data in refining our understanding of species relationships and taxonomy.

Our re-assessment of the B. vombata and B. fragrans subspecies aligns with cladistic subspecies concepts. The subspecies we identify exhibit slight, yet clearly distinguishable, taxonomically relevant biological traits (Mayr 1970), varying allele frequencies (Dobzhansky 1952) and genetic distinctions, associated with geographic isolation (Clausen 1941). The evolution of species occurs gradually, making the distinguishing between species and infraspecific classifications challenging (Reydon and Kunz 2021). However, in these instances, we believe the proposed classifications are the most precise, biologically significant and practical for managing these groups on the basis of our current knowledge. Future studies utilising marker genes such as the Angiosperms353 probe set have the potential to offer significantly enhanced resolution in understanding the evolutionary history and relationships among these species (Johnson et al. 2019; McDonnell et al. 2021), complementing the population-genomic approach of this study.

Diversity and reproduction

We conducted kinship analysis to investigate the modes of reproduction in various species and populations of Bossiaea species (Fig. 3). We found that high levels of pairwise kinship, denoting clonality, were prevalent. Clonality can be attributed to asexual reproduction through rhizomatous growth or apomixis, or it can result from inbreeding or selfing within a population if sexual reproduction is compromised. Bossiaea Group F is known for its extensive rhizomatous growth (Thompson 2012), leading us to assume that most clonal individuals arise from this reproductive method. However, apomixis and inbreeding might also contribute in some cases.

Fig. 3.

Boxplots of pairwise IBD kinship values on the x-axis among individuals grouped by species and site number (y axis) as indicated by the boxes. Each blue dot indicates a single pairwise relationship between two individuals at a site. The clonality threshold k > 1 ÷ 2^(3÷2) is indicated by the red range to the right of the plot. The other colour-coded thresholds represent theoretical values for relationships third-degree (blue), second-degree (pink) and first-degree relationships (orange, denoting parent–offspring and full siblings). These thresholds are accompanied by ranges that encompass potential values (Manichaikul et al. 2010; Purcell et al. 2007). This analysis used 52,539 loci and included all 283 samples, including clones.

Several species in this study, including B. vombata (Thompson 2012; Amor et al. 2020), B. grayi (McDougall 2009), B. bracteosa (Ross 2008; McDougall 2009; Thompson 2012), B. fragrans, B. milesiae and B. bombayensis (Thompson 2012), have previously been observed to exhibit rhizomatous growth patterns. Among these species, we observed widespread clonality in B. grayi (27.9% of samples were genetically unique), B. vombata (43.75% unique genets) and B. bracteosa (60% unique genets; Table 2). Remarkably, B. grayi demonstrated extensive asexual reproduction, with one genet spanning 9 km between the Murrays Corner sites (Sites 4 and 8). By contrast, the sampled B. bracteosa had more localised sucker production with the furthest distance between ramets of the same genetic individual being 30 m. Clonality was less prevalent but still noticeable in B. fragrans (77.42% unique genets) and B. milesiae (94.44% unique genets), and it was not detected in B. bombayensis (Table 2). We also identified intermediate levels of kinship among individuals in all species, varying from first- to third-degree relatives, indicating that sexual reproduction is occurring.

The diversity statistics for each species and subpopulation, with clones removed, showed that in most Bossiaea populations, there was a high level of observed heterozygosity and low inbreeding coefficient (Table 3). In exceptional cases, where the overall inbreeding coefficient was high for the species (GDP 2 and B. riparia), it mainly resulted from the Wahlund effect. This effect occurs when there’s an overestimation of expected heterozygosity owing to within-group structure, resulting in divergent allele frequencies rather than widespread inbreeding (Garnier-Géré and Chikhi 2013).

The maintenance of high heterozygosity across most subpopulation indicates a generalised outbreeding preference, with potential for strong self-incompatibility mechanisms (DeMauro 1993; Pickup and Young 2008). This may be why seed production is absent or rare (Fujii et al. 2016) in the case of B. vombata subsp. vombata (Ross 2008) and B. fragrans (McDougall 2009), where kinship analysis indicated that many individuals are clones or close relatives. In the case of B. vombata subsp. vombata, we identified 7 distinct genetic individuals among the 16 sampled individuals (Table 2), and all these genetic individuals were closely related. This aligns with the findings of Amor et al. (2020), who observed only five genetic individuals in the Wombat State Forest population and no evidence of viable seed (Fig. S1). Even though the pollen viability of this population ranges from 15 to 90% (E. James, unpubl. data), the species has limited seed production and recruitment (Thompson 2012). Given that heterozygosity remains high across B. vombata subsp. vombata (H_O = 0.315, F_IS = −0.051), it seems plausible that the observed infertility is due to incompatibly mechanisms (DeMauro 1993; Pickup and Young 2008). Similarly, although B. fragrans diversity remains high (H_O = 0.334, F_IS = 0.12), the species has many closely related individuals (Fig. 3), and there have been limited observations of fully developed seed (McDougall 2009). Excess heterozygosity caused by self-incompatibility also increases the mutation load, posing an additional risk to preferentially outbreeding species that become isolated or have population-size reduction (Navascués et al. 2010; Duthie and Reid 2016; Caballero et al. 2017), such as these Bossiaea species. Consequently, we do not consider heterozygosity alone to be a good indicator of risk in these species, but rather the combination of diversity information with kinship information and population history is more informative; i.e. if a population is small and has many relatives, even if heterozygosity is high, the population is at risk of inbreeding depression such as reduced fertility. Self-incompatibility and outbreeding preference may also be related to the widespread asexual reproduction observed, wherein sterile populations rely on vegetative or apomictic reproduction to avoid self-incompatibility in sexual reproduction (Navascués et al. 2010).

Gene-flow limitations

Our data identified limited gene flow between populations separated by a minimum of 3 km, which is likely to exacerbate inbreeding and potentially hinders recovery of small populations. Connectivity, as indicated by low F_ST values (<0.1; Fig. 4), primarily occurs over very short distances (<1 km; Supplementary Fig. S2). Restricted gene flow between population and subsequent drift is likely to contribute to the distinct morphological and genetic divergence observed in the subspecific lineages GDP 1 (B. vombata subsp. orientalis) and B. milesiae (B. fragrans subsp. milesiae).

Fig. 4.

Heatmaps of pairwise geographic distance rounded to the nearest kilometre (lower-left triangle) and pairwise fixation index (F_ST; upper-right triangle) values for Bossiaea sites. Annotations on the side and bottom of the heatmaps indicate genetic group (colour) and numbers indicate site (numbers). This analysis used 52,539 loci and included 267 samples (excluding sites where n ≤ 3) as input, and further filtered so that loci with MAF of ≥5% and missingness of <30% were used for the pairwise comparisons.

Despite limited gene flow between species, we identified two outlier individuals within the leafless Bossiaea species (NSW1151220 within GDP 2 and NSW1151575 within B. grayi) showing genomic signatures suggestive of incomplete lineage sorting or hybridisation. Hybridisation has been suspected in Bossiaea species (Thompson 2012), although not among those included in this study. This potential for hybridisation among closely related species could be leveraged as a genetic rescue mechanism (Whiteley et al. 2015) to reintroduce diversity into genetically depleted populations, thereby enhancing fertility and overall adaptive potential for small isolated populations (Frankham 2015).

Conservation implications

The genomic outcomes of our study have enhanced our understanding of population dynamics and taxonomic relationships of leafless Bossiaea taxa to provide informative insights to guide species conservation. A population-genomic approach is ideal for addressing multispecies conservation and taxonomic problems because it provides a comprehensive understanding of genetic diversity and relationships among species, enabling more effective conservation strategies and taxonomic classification based on genetic data (Rossetto et al. 2021).

Through our population-genomics analysis, we have identified restricted gene flow between geographically separated populations, resulting in well-defined and genetically isolated groups. Concurrently, we have detected probable outcrossing preference across all studied Bossiaea species, consistent with indications of strong self-incompatibility mechanisms and a capacity for reproductive modes such as vegetative sprouting or apomixis. Although these mechanisms may aid survival in demanding environments, they also raise concerns about the reproductive potential of species that are characterised by small effective population sizes, and the potential accumulation of detrimental mutations over time.

We have also identified genetically similar lineages within B. vombata and B. fragrans, namely B. vombata subsp. orientalis and B. fragrans subsp. milesiae, which merit formal taxonomic recognition and warrant conservation attention because of ongoing threats. Our emphasis rests on channelling resources and actions towards the restoration and support of self-sustaining populations, given that these lineages currently contend with challenges resulting from inbreeding or clonality.

Rather than pursuing separate conservation strategies for each lineage, we advocate a coordinated approach to yield more informed conservation solutions. To mitigate the challenges associated with inbreeding depression and mutation load, and ensure the conservation of these threatened species, one potential solution involves introducing genetic diversity through controlled crossbreeding with closely related individuals. For example, in the case of B. vombata subsp. vombata, which has limited seed production and high relatedness within its population, exploring crossbreeding with the geographically separated subspecies B. vombata subsp. orientalis could be a valuable avenue to investigate whether reintroducing diversity improves fertility in either or both subspecies. Similarly, the feasibility of crossbreeding between B. fragrans subsp. fragrans and B. fragrans subsp. milesiae should be explored to determine whether this approach can enhance seed production and viability for B. fragrans subsp. fragrans, which also faces limitations in seed production.

This proactive intervention carries a low level of risk, particularly given that these species are already experiencing declining populations with limited recruitment, potentially owing to self-incompatibility mechanisms. Moreover, considering the evolutionary history of these species, especially in species with small, disjunct populations, it becomes evident that landscape evolution and historical factors may have contributed to their current genetic state. In some cases, the disjunct populations may not represent genuine evolutionary divergence, but, rather, recent loss of populations owing to factors such as habitat clearing since European settlement. Therefore, interventions such as genetic rescue could offer significant conservation benefits by restoring genetic diversity and mitigating the impacts of recent population declines. Prior instances of genetic rescue have demonstrated significant improvements in the viability of inbred plant species (Reinartz and Les 1994; Fischer et al. 2003; Tallmon et al. 2004), making it a strategy worth considering in the conservation efforts for these species.

From a conservation perspective, these discoveries underscore the significance of understanding the genetic dynamics of isolated lineages to devise effective management strategies. Measures such as targeted conservation initiatives, safeguarding habitats and enhancing genetic diversity by assisted gene flow or population relocations could prove pivotal in securing the prolonged endurance and adaptability of these distinct and genetically diverse Bossiaea populations. By using a multispecies approach to address the genetic ramifications of isolation and inbreeding, it is possible to make informed choices that will safeguard the biodiversity and evolutionary potential of this remarkable genus, despite environmental challenges and human-induced impacts.

Bossiaea vombata subsp. orientalis K.L.McDougall & N.G.Walsh, subsp. nov.

Type: Victoria. ~130 m W of Cowombat Flat Track, 3.3 km direct NNE from crossing of Limestone Road with Buchan River (Native Dog Flat), 13 Dec 2021, N.G. Walsh 9246, K.L. McDougall & E.A. James (Holo: MEL 2500750!; iso: NSW).

[Bossiaea grayi auct. non K.L.McDougall: I.R.Thompson, A revision of eastern Australian Bossiaea. Muelleria 30, 106–174 (2012)]

Ascending to erect clonal shrub up to 0.5 m high, rarely up to 1.5 m, young growth with scattered, fine appressed hairs up to ~0.5 mm long, older cladodes shortly pubescent in axils, otherwise glabrous, cladodes flattened, with ultimate branches (3–)3.5–9 mm wide, somewhat incised at the nodes, dull green to greenish-grey. Leaves absent, nodes bearing dark ovate-triangular scales 1.0–1.75 mm long, dark brown to blackish, glabrous, persistent. Flowers usually solitary at the nodes, occasionally twinned, the pedicels 1.0–2.0 mm long, glabrous; bracts imbricate, increasing in size from the outermost basal bract to the innermost, broadly ovate to oblate, the largest persistent bract 2.5–3 mm long, coriaceous, brown, glabrous except for the shortly fimbriate margin, persistent; bracteoles resembling bracts, but slightly larger, thinner-textured and paler, ~3.5 mm long and 2.5 mm wide, glabrous except for shortly fimbriate margin, caducous. Calyx, 3.5–5 mm long, green-flushed with purple towards apex, the tubular part glabrous externally and internally, lobes triangular, ±equal, 1–1.5 mm long and wide (or dorsal pair up to 2 mm wide), puberulous on inner surfaces and margins shortly fimbriate. Standard 8.0–9.0 mm long including a claw 1.5–2.0 mm long, 7.0–8.0 mm wide, abaxially yellow with a reddish to purplish basal flare, externally purplish at midline fading to yellow at margins; wings 7.0–8.0 mm long including a claw ~2 mm long, yellow-tinged with red at base; keel 6.0–7.5 mm long, including a claw ~2 mm long, uniformly crimson. Stamen-filaments 2.5–3 mm long with a sheath ~4 mm long, red or yellow. Ovary at anthesis ~4 mm long, glabrous; style ~4 mm long, glabrous. Fruits and seeds unknown.

Bossiaea fragrans subsp. milesiae (K.L.McDougall) K.L.McDougall & N.G.Walsh, comb. nov., stat. nov.

Bossiaea milesiae K.L.McDougall Telopea 12(3): 358–359 (2009). Type: New South Wales: South Coast. Brogo River, ~25 km NNW of Bega (~1 km downstream from Brogo Dam), K.L.McDougall 1193, J.Miles & P.Jeuch, 12 Sep. 2006 (holo: NSW 785654; iso: CANB 766111, MEL 2318264).