Evolutionary processes in an undescribed eucalypt: implications for the translocation of a critically endangered species

Abstract

Background and Aims

Knowledge of the evolutionary processes responsible for the distribution of threatened and highly localized species is important for their conservation. Population genomics can provide insights into evolutionary processes to inform management practices, including the translocation of threatened plant species. In this study, we focus on a critically endangered eucalypt, Eucalyptus sp. Cattai, which is restricted to a 40-km² area of Sydney, Australia, and is threatened by increased urbanization. Eucalyptus sp. Cattai has yet to be formally described in part due to its suspected hybrid origin. Here, we examined evolutionary processes and species boundaries in E. sp. Cattai to determine whether translocation was warranted.

Methods

We used genome-wide scans to investigate the evolutionary relationships of E. sp. Cattai with related species, and to assess levels of genetic health and admixture. Morphological trait and genomic data were obtained from seedlings of E. sp. Cattai propagated in a common garden to assess their genetic provenance and hybrid status.

Key Results

All analyses revealed that E. sp. Cattai was strongly supported as a distinct species. Genetic diversity varied across populations, and clonality was unexpectedly high. Interspecific hybridization was detected, and was more prevalent in seedlings compared to in situ adult plants, indicating that post-zygotic barriers may restrict the establishment of hybrids.

Conclusions

Multiple evolutionary processes (e.g. hybridization and clonality) can operate within one rare and restricted species. Insights regarding evolutionary processes from our study were used to assist with the translocation of genetically ‘pure’ and healthy ex situ seedlings to nearby suitable habitat. Our findings demonstrate that it is vital to provide an understanding of evolutionary relationships and processes with an examination of population genomics in the design and implementation of an effective translocation strategy.

INTRODUCTION

Evolutionary processes drive lineage diversification and shape species diversity and distribution (Pagel, 1997). Processes such as inbreeding, hybridization and clonality greatly influence the persistence of populations and species, and can impact a species’ extinction risk (Pierson et al., 2015). For example, inbreeding depression in a rare species can significantly increase the probability of its extinction (Cook and Sgró, 2019). Alternatively, genetic swamping from a common species can prevent a sympatric rare species from replacing itself, resulting in hybrid swarms (Levin et al., 1996). It is now widely recognized that plant diversity worldwide is under considerable threat, with rapid and extensive anthropogenic changes over recent decades resulting in a dramatic increase in the number of threatened species (Wyse et al., 2018). Understanding evolutionary processes can help achieve better long-term outcomes when managing threatened plants for restoration efforts, including translocations (Rossetto et al., 2021).

Translocations (i.e. the movement of living organisms from one location to another) are increasingly being used to reduce the probability of species extinction through the establishment of resilient and self-sustaining populations (IUCN, 1987; Bragg et al., 2021). From a genetic perspective, translocations should be designed to reduce detrimental effects impacting small fragmented populations, such as inbreeding depression and reduced genetic variation (Weeks et al., 2011). Yet, while targeted threatened species management is a global priority for preventing further biodiversity loss (Howard et al., 2020), developing and implementing effective strategies to safeguard species remains a major challenge. Translocations can be especially challenging for species from taxonomically complex groups. For instance, species boundaries in recently radiated groups can be difficult to delineate due to secondary contact through interspecific hybridization (Suh et al., 2015). Until recently, in many cases we lacked the tools to distinguish whether priority species are distinct but blurred by secondary contact (Georges et al., 2018), or if they are part of an extensive morphologically variable continuum (Hundsdoerfer et al., 2019).

Genome-wide sequencing can provide useful insights into the processes driving species evolution and can thus be a powerful tool when investigating the evolutionary history of species (Rutherford et al., 2021). Genomic sequencing has enabled the resolution of species boundaries in many taxonomically challenging groups (e.g. Wagner et al., 2013), and can be used to assess patterns of gene flow and genetic variation within and between closely related species (Sansaloni et al., 2011). Genomic datasets also allow the calculation of population genetic parameters with higher statistical power, and are unlikely to be impacted by random or non-random processes affecting a single locus (Hoffman et al., 2014; Bragg et al., 2015). For example, the process of sequencing thousands of markers across the genome can more accurately detect interspecific hybridization, thereby providing a sensitive filter for the translocation of the purest individuals of an ex situ cohort (Allendorf et al., 2010).

In the current study, we focus on an undescribed eucalypt commonly known as Eucalyptus sp. Cattai (Myrtaceae), which is restricted to a 40-km² mostly urbanized area of Sydney, Australia (Fig. 1). In the 1990s, its distinctiveness compared to other eucalypt species was noticed (by S. Douglas and I. Brooker; S. Douglas, pers. comm.), which has been outlined in more recent identification tools (Klaphake, 2012; Royal Botanic Gardens and Domain Trust, 2021). Eucalyptus sp. Cattai is thought to be closely related to the red mahoganies (Royal Botanic Gardens and Domain Trust, 2021). The red mahoganies (subgenus Symphyomyrtus section Latoangulatae series Annulares, sensuBrooker, 2000) have trunks entirely covered by rough bark, fruit with prominent valves, ovules in 4–8 vertical rows and pyramidal seeds; they include commercially important trees (e.g. E. urophylla of Timor and other islands north of Australia), and widespread species in eastern Australia (e.g. E. resinifera and E. notabilis) (Fig. 1).

Fig. 1.

Morphology and distribution of Eucalyptus sp. Cattai, showing: (A) habit, buds, flowers and fruits, (B) distribution of red mahoganies in Australasia (subgenus Symphyomyrtus section Latoangulatae series Annulares, sensuBrooker, 2000), and (C) distribution of E. sp. Cattai and red mahoganies in the Sydney region. Photography by T. C. Wilson and S. Rutherford. Maps were created using the R package ‘leaflet’ and distribution data were obtained from the Australasian Virtual Herbarium (2021).

Understanding relationships among species in section Latoangulatae, including the red mahoganies, has been problematic (e.g. Le et al., 2009). Molecular dating suggests that many groups in subgenus Symphyomyrtus have undergone rapid and recent speciation, with most species diverging in the last 10 million years (Thornhill et al., 2019). Previous studies suggest extensive introgression in the subgenus with reticulate evolution playing a major role in the evolution of species (McKinnon et al., 2004). In a recent phylogenetic analysis, Jones et al. (2016) found the red mahoganies were not monophyletic, and that hybridization and incomplete lineage sorting had probably confounded the resolution of species boundaries. Based on these findings, Nicolle and Jones (2018) made taxonomic changes to section Latoangulatae. In this revised classification, although some of the red mahoganies (E. resinifera, E. notabilis and E. scias) remained in series Annulares, others (E. robusta, E. pellita and E. urophylla) were placed into series Robustae, while E. botryoides was put into series Transversae (Nicolle and Jones, 2018).

Interspecific hybridization has been shown to pose a considerable threat to many rare eucalypts through localized genetic swamping (e.g. Rutherford et al., 2019). Although E. sp. Cattai is considered a putative species by some authors, others have suggested that it may be an intergrade between E. notabilis and E. resinifera (K. Hill and L. Johnson; Royal Botanic Gardens and Domain Trust, 2021). Previous records indicate that E. sp. Cattai is sympatric to E. resinifera and E. notabilis (Australasian Virtual Herbarium, 2021), so if it is distinct then its small size and restricted distribution may make it particularly susceptible to genetic swamping by widespread related co-occurring species.

Conversely, genetic diversity and resilience to change of rare species may be highly affected by clonality (van der Merwe et al., 2010). Rare species often have small population sizes with lower sexual fecundity and reduced genetic variation (Kéry et al., 2000). As such, clonality is thought to be more common in localized or endangered species with older, long-established individuals (Silvertown, 2008), and has been shown in rare eucalypts (e.g. Smith et al., 2003). Given that E. sp. Cattai is rare and is often a mallee (long-lived multi-stemmed tree that can reproduce clonally via an underground lignotuber; Nicolle, 2006), it is expected that the propensity for producing clonal individuals may be relatively high (Rossetto et al., 1999).

Eucalyptus sp. Cattai is currently listed as critically endangered under both state (NSW Biodiversity Conservation Act 2016) and federal (Environment Protection Biodiversity Conservation Act 1999) legislation. With fewer than 700 individuals in variously sized fragmented populations, E. sp. Cattai is threatened by habitat loss through residential development, poor recruitment, and increasingly frequent and intense fires (NSW Department of Planning, Industry and Environment, 2020). To combat these threats, a targeted management strategy is being implemented to secure E. sp. Cattai from extinction, which includes translocating seedlings from threatened sites to nearby suitable habitat. However, the effectiveness and conservation value of this initiative has remained in jeopardy while the evolutionary history of the populations described as E. sp. Cattai was unknown.

In this study, we investigated species boundaries and evolutionary processes in E. sp. Cattai to assess whether its translocation was warranted. Using genome-wide scans, we explored evolutionary relationships among E. sp. Cattai and related species within section Latoangulatae. Additionally, we examined patterns of genetic differentiation and clonality within both adult plants and seedlings of E. sp. Cattai. We also assessed the risk of genetic swamping by studying hybridization between E. sp. Cattai and related and sympatric species. Here, we ask the following questions: (1) Is E. sp. Cattai a differentiated evolutionary lineage, and are the red mahoganies monophyletic? (2) Is there evidence of interspecific hybridization in E. sp. Cattai, and if so, do morphological traits vary between hybrid and genetically ‘pure’ seedlings? (3) If present, what is the extent of clonality in E. sp. Cattai?

MATERIALS AND METHODS

Field sampling

We sampled 104 individuals across the range of E. sp. Cattai (Fig. 1), including three sites in the north (northern Kenthurst and Annangrove), three sites in the central part of its distribution (North Kellyville and Kenthurst) and two sites in the south (Kellyville and Glenhaven, Table 1). Closely related species were also sampled, including E. resinifera and E. notabilis. Although both species were reported to co-occur with E. sp. Cattai, we found neither was sympatric to E. sp. Cattai during our sampling. Thus, a population of E. resinifera was sampled nearby (10 km from E. sp. Cattai), while a population of E. notabilis was sourced from Faulconbridge (50 km west of Kellyville). Another population of E. notabilis was sampled at Boonoo Boonoo (700 km north-west of Sydney). We sampled the other red mahoganies, E. robusta, E. botryoides and E. scias, which are not sympatric to E. sp. Cattai, but are found in the Sydney region. A putative hybrid (E. grandis × E. robusta) sympatric to E. sp. Cattai was also sampled, as was an herbarium specimen identified as E. notabilis × E. resinifera. We included species related to the red mahoganies from subgenus Symphyomyrtus, namely the blue gums, E. saligna, E. deanei and E. grandis; the grey gums, E. punctata and E. propinqua; and the red gum, E. parramattensis. The section and series for each of these species (based on the classification of Nicolle and Jones, 2018) are given in Fig. 2. Eucalyptus squamosa (subgenus Symphyomyrtus section Bisectae), which is often sympatric to E. sp. Cattai, was also sampled. Two species from subgenus Eucalyptus, E. sieberi and E. piperita, were used as outgroup taxa.

Table 1.

Eucalyptus sp. Cattai and other eucalypts sequenced for genomic DNA. For each species, individuals observed in the field are reported (if known). For E. sp. Cattai, the number of adults and seedlings, and the number of genets estimated from genomic analysis for each site are shown. All populations are from New South Wales, except for E. grandis Kahlpahlim which is from Queensland (see Fig. 1). Genetically pure unique seedlings do not include those that were found probably to be the result of self-pollination (Supplementary Data Table S7) or which were of hybrid origin (Table S3). Total genets include both unique adult genets (excluding hybrids) and genetically pure unique seedlings.

Species	Site	Code	Individuals observed	Adults sequenced	Seedlings sequenced	Genets with multiple ramets	Genetically pure unique adults	Genetically pure unique seedlings	Total genets
E. sp. Cattai	Clarke Way, Kenthurst	Clarke	292	25	4	3	18	3	21
	Shoplands Road, Annangrove	Shoplands	41	14	10	2	9	7	16
	Woodlands Road, Annangrove	Woodlands	38	6	0	2	3	0	3
	Saltwater Circle, North Kellyville	Saltwater	250	30	59	8	19	50	69
	Colbran Avenue, Kenthurst	Colbran	25	12	0	3	4	0	4
	Logie Road and Robson Road (Logie–Robson) intersection, Kenthurst	Logie	15	5	9	1	4	3	7
	Bannerman Road, Glenhaven	Bannerman	38	6	0	1	2	0	2
	Adjacent to Georgia Terrace and Cattai Creek Drive, Kellyville	Georgia	6	6	0	2	2	0	2
	Kellyville area	Gregson		0	1	0	0	1	1
E. notabilis × E. resinifera	Foxall Road, Kellyville	Foxall		1	0
E. grandis × E. robusta	Woodlands Road, Annangrove	Woodlands	1	1	0
E. resinifera	Caterson Drive, Castle Hill	Caterson	>20	6	0
E. notabilis	Casuarina Trail, Boonoo Boonoo National Park	Casuarina	>20	6	0
	Meeks Crescent, Faulconbridge	Meeks	6	6	0
E. scias	West Head Road, Ku-Ring-Gai Chase	West	>30	6	0
E. robusta	Cape Solander Drive, Kurnell	Cape	10	5	0
E. botryoides	South of Marie Byles Lookout, Bouddi	Marie	10	6	0
	Near Etta Road, Umina Beach	Etta	15	1	0
E. grandis	Trail to Kahlpahlim Rock	Kahlpahlim		1	0
	Measuring Hut Road, Cascade	Measuring		1	0
	Devos Trail, Bongil Bongil National Park	Devos		2	0
	Ferny Creek, Bungwahl	Ferny		2	0
E. saligna	Colongon Road, Boonoo Boonoo National Park	Colongon		2	0
	Cambridge Plateau	Cambridge		2	0
	Mount Royal	Royal		2	0
	Royal National Park	Roy		2	0
E. deanei	Boonoo Boonoo Falls Road, Boonoo Boonoo National Park	Boonoo		2	0
	London Bridge Trail, Moggs Swamp	London		1	0
	Putty Road, Howes Valley	Putty		2	0
E. punctata	Clarke Way, Kenthurst	Clarke		6	0
	Shoplands Road, Annangrove	Shoplands		5	0
	Saltwater Circle, North Kellyville	Saltwater		1	0
E. propinqua	Cowarra Access Road, Lake Innes	Cowarra	10	6	0
E. parramattensis	Ulan Road, Goulburn River National Park	Ulan		1	0
E. squamosa	Clarke Way, Kenthurst	Clarke	>20	6	0
	Saltwater Circle, North Kellyville	Saltwater	>20	6	0
	Heathcote Road, Lucas Heights	Heathcote	10	6	0
E. piperita	Clarke Way, Kenthurst	Clarke	20	5	0
E. sieberi	Mount Keira	Keira	>30	5	0

Fig. 2.

Relationships among Eucalyptus sp. Cattai and other eucalypts derived from genome-wide scans in: (A) Splitstree and (B) SVDquartets phylogeny. The analysis in A is based on 13 054 SNPs and includes all samples (codes are given in Table 1). In A, ‘Putative hybrids A’ includes seedlings from the Logie–Robson Road intersection; while those labelled as ‘Putative hybrids B’ consist of E. notabilis × E. resinifera, seedlings and one adult from Clarke Way, and seedlings from Saltwater Circle and Shoplands Road. The phylogeny in B includes all species sampled, with E. piperita and E. sieberi (Eucalyptus subgenus Eucalyptus) used as outgroup species. Putative hybrids identified in A and all seedlings were excluded from the analysis in B. In B, taxonomic sections (sec.) and series (ser.) in subgenus Symphyomyrtus (sensuNicolle and Jones, 2018) are shown and numbers above the branches represent bootstrap values >50 %.

We collected four to six leaves from a minimum of six individuals per population of E. sp. Cattai (trees were sampled more than 10 m apart where possible). The location (latitude and longitude) of each individual was recorded. Population sizes of E. sp. Cattai varied from fewer than ten to almost 300 individuals and, as such, the larger populations (e.g. Clarke Way and Saltwater Circle) were sampled more extensively in number and distribution to more accurately capture the genetic variation across these populations, as well as the taxon. Leaf samples were kept cool after collection until freezing (minimum of 12 h at −80 °C), were freeze-dried and stored in silica gel until used for DNA extraction.

Genetic sampling and morphological measurements of ex situ seedlings

To examine the morphology and genetic identity of seedlings, we collected fruits from four populations of E. sp. Cattai (Clarke Way, Saltwater Circle, Shoplands Road, and the Logie Road and Robson Road intersection). Populations of E. sp. Cattai consisted mostly of older mallees (based on a comparison of trunk and lignotuber size), many of which did not appear to produce fruits. Over the last few years, we observed that less than 40 % of individuals at Shoplands Road and the Logie–Robson Road intersection showed signs of reproduction. Therefore, at each site, we collected capsules from any individual bearing fruit. Complicating the selection of individuals, E. sp. Cattai is usually a multi-stemmed tree, and therefore in certain cases it was difficult to distinguish between proximally located individuals. In such cases, fruits collected from different stems were labelled separately. In total 15 plants were sampled across the four populations (11 of which were genetically sequenced and included in the 104 adults sampled as described above). Fruits were air-dried in a cool, low-humidity environment to release seeds, germinated on Petri dishes, and the emergent seedlings sown into individual pots containing a Native potting mix in batches between October 2018 and February 2019. Plants were grown in a glasshouse at the Australian Botanic Garden (−34.07°S, 150.77°E) under natural light and a misting-bay.

One leaf was removed from each of 83 seedlings for genomic sequencing and processed according to the methods described in the field sampling strategy. Of these, 44 seedlings were measured for morphological traits. Thirteen morphological traits were measured: (1) plant height (cotyledon node to the base of the most recently produced leaf node), (2) internode length between the first and second juvenile leaf node, (3) stem diameter at node 2, (4) length of the smallest leaf (i.e. lamina length), (5) width of the smallest leaf (at the widest point), (6) length of the smallest leaf petiole, (7) width of the smallest leaf petiole, (8) length of the largest leaf (i.e. lamina length), (9) width of the largest leaf (at the widest point), (10) length of the largest leaf petiole, (11) width of the largest leaf petiole, (12) minimum leaf thickness and (13) maximum leaf thickness. These traits were measured to the nearest millimetre approximately 4 months after the seedlings were sown. Plant height and internode length were measured with a ruler, while stem diameter and all leaf measurements were recorded using digital Vernier calipers (Kincrome, Scoresby, Victoria, Australia).

DNA extraction and DArTseq analysis

Leaf samples were sent to Diversity Arrays Technology Pty Ltd, where total genomic DNA was extracted using the Plant DNA Extraction Protocol for DArT (https://www.diversityarrays.com/orderinstructions/plant-dna-extraction-protocol-for-dart ). Proprietary analytical pipelines (i.e. DArTseq analysis) were used to produce genotype data. DArTseq combines complexity reduction using restriction enzymes, implicit fragmentation size selection and next-generation sequencing (NGS) (Kilian et al., 2012), and is similar to double-digest restriction associated DNA sequencing (ddRAD; Peterson et al., 2012). DArTseq produces a set of co-dominant single nucleotide polymorphisms (SNPs) and associated metadata informing the quality of each SNP. The metadata includes a reproducibility score for each locus (i.e. the scoring consistency of genotype calls across replicates).

For this study, a dataset was generated for 293 samples comprising 68 355 SNPs (97.8 % had a reproducibility score ≥96 %). The proportion of missing data for the samples was between 5.8 and 84.4 %, with a mean of 27.2 %. The SNP dataset was checked for quality using the filtering scripts implemented by an in-house designed R package ‘RRtools’ v.1.0 (Rossetto et al., 2019). To ensure that high-quality markers were used for analyses, only SNPs with a reproducibility of ≥96 % and with <20 % missing data were included in the dataset. One SNP was randomly selected for markers that had more than a single SNP so that our analyses were not influenced by linkage. Finally, one sample with >60 % missing data after quality checking (E. sp. Cattai from Clarke Way) was excluded from the dataset, as was another sample that was found to be a misidentification (E. piperita from Clarke Way). This left 13 054 SNPs and 291 samples remaining for analyses.

Relationships among species and populations

We used Splitstree v.4.14.6 (Huson et al., 2008) to generate a network from the reduced dataset (13 054 SNPs) to visualize genetic relationships among species and samples. Splitstree is useful for representing evolutionary histories, especially in groups with substantial reticulation arising from incomplete lineage sorting and hybridization (Huson and Bryant, 2006). We generated a network of all samples using the NeighborNet method in Splitstree. Additionally, we generated a principal component analysis (PCA) in adegenet v.2.1.1 (Jombart, 2008) to assess genetic similarity at the species, individual and population levels. Genetic structure analysis for all species was performed using LEA v.2.4.0 (Frichot and François, 2015), which estimates ancestry coefficients from large genotypic matrices and evaluates the number of ancestral populations. LEA implements the snmf function (sparse Non-Negative Matrix Factorization algorithms) to estimate individual admixture coefficients from the genotype matrix (François and Durand, 2010). A measure of fit (the entropy criterion) is evaluated between the statistical model and data, and is used to choose the best number of ancestral populations (K) that explain the data (Frichot and François, 2015). We examined the minimum cross-entropy for up to K = 8, wherein we selected the optimal K based on the post-stabilization of the steepest decline in cross-entropy values. In addition, we inspected all plots produced by LEA at all K values so that the most biologically sensible number of ancestral populations was selected. The use of mathematical algorithms to select the optimum value of K (implemented in software such as LEA) can be problematic as there is some degree of uncertainty as to what K value best fits the data (Janes et al., 2017). Therefore, we followed the recommendations of Janes et al. (2017), which involved inspecting all plots produced by each LEA analysis at different K values to select the optimum value of K. LEA was also used to examine the genetic structure and estimate the ancestral populations that best fit the data of samples of only E. sp. Cattai using the approach described above.

The coalescent-based phylogenetic tool SVDquartets v.1 (Chifman and Kubatko, 2014) implemented in the PAUP software v.4.0a (Swofford, 2002) was used to evaluate the position of E. sp. Cattai within the Eucalyptus phylogeny. This program is designed to accept SNPs and can accept a large number of specimens and data while still producing relatively robust phylogenetic results (Chou et al., 2015). Given that SVDquartets was used to examine relationships among species and to test the species descriptions based on morphology (e.g. Slee et al., 2006, Klaphake, 2012), samples with intermediate morphological characters (i.e. E. notabilis × E. resinifera and E. grandis × E. robusta) and putative hybrids identified using Splitstree were excluded from this analysis. Seedlings were also excluded as they could not be reliably identified using morphology (due to the absence of any description of seedling morphology). As such, this analysis was based on 13 073 SNPs (all with a reproducibility ≥96 % and <20 % missing data). The multispecies coalescent model was set up with the following parameters: 100 000 quartets and 1000 bootstrap replicates. We examined the results of all analyses using at least three independent runs for multi-species coalescent analysis by allocating samples within their respective populations. Eucalyptus piperita and E. sieberi were used as outgroup species (based on previous eucalypt phylogenies, e.g. Thornhill et al., 2019).

Hybridization and morphological variation across seedlings

To further investigate admixture across species, we used TreeMix v.1.13 (Pickrell and Pritchard, 2012). This software generates a maximum-likelihood (ML) tree (using allele frequency data and a Gaussian approximation for genetic drift among populations), and calculates admixture across the branches in a stepwise likelihood procedure, where the tree is searched for the optimal placement of migration events (Pickrell and Pritchard, 2012). Using the reduced dataset (13 054 SNPs), we generated an ML tree without any admixture. Subsequently, we inferred ML trees allowing for admixture. We estimated the optimal number of migration events using the OptM R package (Fitak, 2021). TreeMix was run with 0–10 migration edges in five iterations, and using a range of -k values (10, 50, 100, 200 and 500). The function optM was used to determine the optimum number of migration edges (using the approach in Fitak, 2021 and Senzcuk et al., 2021) from the output files from TreeMix, with both the linear modelling and SiZer methods.

To determine whether a focal population was admixed between two other source populations, we used the f3 population test of treeness (Reich et al., 2009) in the program Threepop provided within the TreeMix package. The f3 statistics and their corresponding Z-scores and standard errors were calculated in blocks of 100 SNPs (Von Wettberg et al., 2018). In any given analysis of three populations (i.e. X; A, B), a significantly negative f3 statistic (at P < 0.05) suggests admixture in the target population (X) from the source populations A and B (Schaefer et al., 2016).

To identify hybrids between E. sp. Cattai and other species, we used NewHybrids v.1.1 (Anderson and Thompson, 2002), a model-based Bayesian clustering analysis that assigns individuals to one of six genotypic classes: species A, species B, F₁ hybrid, F₂ hybrid, backcross to species A and backcross to species B. Due to the complex computation analysis implemented in NewHybrids, we used subsets of the dataset. These subsets were selected using the outputs from our clustering-based and phylogenetic analyses. We removed SNPs with >5 % missing data and reproducibility <98 %. From this reduced but high-quality dataset, 300 markers were randomly selected and imported into NewHybrids. In the final analysis, we used the uniform prior, a burn-in period of 100 000 runs (to allow the Markov chain Monte Carlo simulations to converge) and a run-time of 1 000 000 sweeps. In the output from NewHybrids, a specimen was assigned as a ‘pure’ species when its probability (for species A or species B) was P ≥ 0.90 (McIntosh et al., 2014). Individuals that had assignment values of P ≥ 0.90 to a hybrid category (F₁, F₂, backcross to species A, backcross to species B) were denoted as that category. Samples that could not be assigned to any genetic class (i.e. P < 0.90) were regarded as entirely admixed (McIntosh et al., 2014). To compare hybridization between seedling and adult plants, we calculated the proportion of seedlings and adults that were of hybrid origin. Given that many of the seedlings in our study were probably siblings or half siblings (as the seeds were sourced from 15 trees from four populations), it was necessary to select one seedling per adult plant when calculating the proportion of hybrid individuals (to account for non-independence). Where an individual plant produced both hybrids and pure seedlings, we randomly selected one seedling to represent that individual. We used Pearson’s chi-square test (Rosner, 2006) to test the significance of differences (at P < 0.05) between proportions of hybrids in the seedlings and adult plants using R v.3.6.1.

We used a multivariate analysis of variance (MANOVA) to examine differences in morphological traits of seedlings. Given that we had measured multiple traits, many of which are likely to be highly correlated, a MANOVA (which allows all variables to be considered simultaneously; Quinn and Keough, 2002) was considered to be appropriate for investigating between-group differences. Individual traits were further analysed using one-way analysis of variance (ANOVA), focusing on differences between ‘pure’ and hybrid seedlings among sites, as well as between all pure and all hybrid seedlings. Significant differences (at P < 0.05) between pairwise mean values within each ANOVA were assessed using Tukey’s honest significant difference (HSD) test. Prior to each analysis, homogeneity of variances among groups was examined with Levene’s test. Traits were log₁₀-transformed where necessary to meet the variance assumptions of ANOVA. For traits where log₁₀ transformation did not stabilize the variances, we performed a type III ANOVA followed by Tukey’s HSD test (using the HSD.test function in the R package ‘agricolae’). A PCA biplot based on all morphological traits was used to visualize differences between hybrid and pure seedlings (using the fviz_pca_biplot function from the R package ‘factoextra’). All analyses were performed in R v.3.6.1.

Clonality and genetic diversity

To assess the degree of clonality and parentage in E. sp. Cattai, genetic similarity between individuals located at the same site and corresponding seedlings was estimated. In the case of this endangered taxon, it was possible that there would be substantial inbreeding and high levels of homozygosity, which could make it challenging to identify levels of genetic similarity that distinguish clonal ramets and close relatives. This is because closely related individuals can have highly similar genotypes, and clonal individuals also typically do not have identical genotypes in large SNP datasets (due to non-zero rates of genotyping error). Therefore, we used kinship values since this was a flexible approach that would enable us to contemplate relatedness among all pairs of individuals, and to exploit the known parent–offspring relationships to identify useful thresholds for classifying related pairs and clonal pairs. An unweighted pair group method with arithmetic mean (UPGMA) hierarchical clustering method was implemented in the R package ‘phanghorn’ v.2.4.0 (Schliep, 2011). Pairwise kinship (i.e. relatedness) coefficients were estimated based on the reduced dataset (6496 SNPs) using an identity-by-descent (IBD) analysis in the R package ‘SNPrelate’ v.1.17.1 (Zheng et al., 2012). Distance matrices of pairwise kinship were generated for each population after observing the preliminary results from principal component and network analyses that clonality occurs within each site. The matrices were combined to generate a supermatrix that was drawn using the heatmap function from the R package ‘Phytools’ v.0.6-60 (Revell, 2012). Kinship values were used to examine the degree of clonality across the in situ plants, and to identify genets and ramets of E. sp. Cattai in cultivation as detailed in Bragg et al. (2021). A kinship value for any pair of seedlings and in situ plants between 0.25 and 0.4 was used to indicate genetic similarity and whether an offspring–parent relationship is likely.

To estimate population-level measures of genetic diversity of E. sp. Cattai, we used the R package ‘diveRsity’ (Keenan et al., 2013). Individuals revealed to be hybrids or clones were excluded from this analysis (so only one pure genet per individual was used). Expected (H_E) and observed (H_O) heterozygosity, and inbreeding coefficient (F_IS) for each population were calculated from a reduced dataset (4,913 SNPs). Genetic diversity parameters were calculated for seedlings from Clarke Way, Saltwater Circle, Shoplands Road and the Logie–Robson Road intersection to compare with the values obtained for the adult plants. Pairwise F_ST was calculated across all populations (i.e. location combinations) using ‘SNPrelate’ v.1.17.1.

RESULTS

Relationships among populations and species

Most samples of E. sp. Cattai formed a group distinct from all other species in the Splitstree and PCA (Fig. 2A; Supplementary Data Fig. S1). Eucalyptus resinifera, E. notabilis and E. scias clustered together in the PCA and Splitstree (Fig. 2A; Fig. S1). All populations of E. squamosa grouped together, as did the populations of E. robusta and all samples of E. punctata (Fig. 2A; Fig. S1). Eucalyptus botryoides was grouped with E. saligna and E. grandis, while E. propinqua clustered with E. parramattensis (Fig. 2). In the SVDquartets phylogeny, E. sp. Cattai was sister to a clade comprising E. resinifera, E. notabilis and E. scias (with 100 % bootstrap support, Fig. 2B).

In the Splitstree network, some seedlings of E. sp. Cattai from the Logie–Robson Road intersection (hereafter Logie) were intermediate between the majority of E. sp. Cattai and a group consisting of E. resinifera, E. notabilis and E. scias (‘Putative hybrids A’, Fig. 2A). Other E. sp. Cattai samples (mostly seedlings) and the specimen identified as E. notabilis × E. resinifera occupied an intermediate position between the main E. sp. Cattai group and other species (‘Putative hybrids B’, Fig. 2A). The putative E. grandis × E. robusta from Woodlands was recovered separately from all other species in the Splitstree network (Fig. 2A). LEA revealed that most E. sp. Cattai samples were assigned to one group (Q > 0.80; Supplementary Data Fig. S2, Table S1). Eucalyptus grandis × E. robusta, E. notabilis × E. resinifera and samples identified as putative hybrids in the Splitstree network and PCA were found to be admixed in LEA (Fig. S2). At the population level, E. sp. Cattai displayed low levels of structure in the PCA, with only some Saltwater and Shoplands individuals (mostly seedlings) being differentiated from the other samples (Fig. 3). This was supported by LEA, which revealed limited population genetic structure within E. sp. Cattai (see Fig. S2 for plots at K = 2–8; Table S2).

Fig. 3.

Principal component analysis (PCA) showing groupings of Eucalyptus sp. Cattai populations based on genome-wide scans (5080 SNPs). Seedling and adult samples are indicated. Codes are given in Table 1.

Hybridization and morphological variation across seedlings

In an investigation of gene flow across lineages using TreeMix, the log likelihood (LL) increased by 97.45 after the addition of the first migration event, followed by an increase of 97.31 after the second migration event, and an increase of <26 with subsequent migration events. Additionally, the outputs from the optM function (at all -k values) indicated two migration events was the most reliable number for our dataset (Supplementary Data Fig. S3). As such, we discuss the results from the second migration event here. The ML tree shared a similar topology to our earlier analyses; it revealed moderate gene flow (16 %) between E. sp. Cattai Bannerman and a clade comprising E. notabilis, E. resinifera and E. scias (Fig. 4). Moderate gene flow (29 %) was also detected from a clade of E. grandis to E. sp. Cattai seedlings from Logie (Fig. 4). TreeMix also revealed that seedlings from the two most northern populations of E. sp. Cattai, Clarke and Shoplands, were sister to each other and this is probably due to pollen immigration between the two locations (since Clarke is closer to Shoplands than all the other populations, Fig. 4). The residual matrix revealed that E. notabilis × E. resinifera Foxall was a likely candidate for admixture with E. parramattensis (Fig. S4).

Fig. 4.

TreeMix analysis of Eucalyptus sp. Cattai and other eucalypt species based on genome-wide SNP data showing the maximum-likelihood (ML) tree inferred under two migration events. Gene flow is indicated by arrows coloured yellow to red according to their weight (0–50 %). The corresponding residual matrix is presented in Supplementary Data Fig. S4. Codes are given in Table 1.

NewHybrids was used to test for hybrids between E. sp. Cattai and species that were closely related (i.e. E. resinifera, E. notabilis and E. scias), sympatric (e.g. E. squamosa), occurred nearby (e.g. E. saligna) or were from the Sydney region (e.g. E. botryoides; Fig. 1C). Since NewHybrids assigns samples to specific classes and will only allow two taxa and putative hybrids to be analysed at any one time (Anderson and Thompson, 2002; Rutherford et al., 2019), it was necessary to perform multiple analyses, each with subsets of the data that could meet the assumptions of the software (a full description of the analyses performed is given in Supplementary Data Table S3). Although E. grandis does not naturally occur in the Sydney region, it was included in our analysis since it has been widely planted and the TreeMix analysis detected admixture between this species and E. sp. Cattai. We found the 13 seedlings identified as putative E. sp. Cattai hybrids in the Splitstree network (Fig. 2) were of hybrid origin (Table S3). Six adult plants sampled as E. sp. Cattai were also revealed to be hybrids (two each from Bannerman and Georgia, and one each from Clarke and Colbran). These adults were either backcrosses to E. sp. Cattai or entirely admixed, and were morphologically similar to ‘pure’ E. sp. Cattai. Eucalyptus notabilis × E. resinifera and E. grandis × E. robusta were confirmed to be hybrids. No hybrids were detected between E. sp. Cattai and either E. squamosa or E. robusta (Table S3). Overall, we found that the proportion of seedlings of hybrid origin was 29 % (after excluding probable siblings and half siblings), which significantly differed (P < 0.001) from the proportion of adult plants which were hybrids (6 %, Pearson’s chi-square test, d.f. = 1, Fig. S5 and Table S4). A number of adult plants were found to be clonal individuals (see results from the kinship analysis below), and we found that even after excluding clones (to account for non-independence), the proportions of seedlings and adults of hybrid origin were still significantly different (P < 0.01, Table S4). The higher degree of admixture within E. sp. Cattai seedlings was supported by f3 statistics, which revealed that seedlings of E. sp. Cattai from Logie, Saltwater and Shoplands were admixed between E. sp. Cattai and other eucalypt species (they all had significantly negative Z-scores, Table S5).

Morphological traits varied significantly between genetically ‘pure’ and hybrid seedlings (P < 0.0001, Supplementary Data Table S6). Overall, hybrids had significantly greater height, leaf width (for both the largest and smallest leaves) and petiole length (for largest leaf only) compared to the pure seedlings (Fig. S6). However, in a PCA biplot of all the morphological traits, while many hybrids were separate from pure seedlings (by having either more negative values on PC1 or more positive values on PC2), some hybrid and pure individuals clustered together (Fig. 5).

Fig. 5.

Seedlings of Eucalyptus sp. Cattai showing: (A) a genetically ‘pure’ individual from Shoplands Road, (B) a hybrid from Saltwater Circle and (C) a hybrid from Shoplands Road; (D) principal component analysis (PCA) biplot based on 13 morphological traits (Fig. S6). The hybrid in B is morphologically similar to the genetically pure individual of E. sp. Cattai shown in A, while leaves of the hybrid in C are visibly larger than in the pure individual. Genet type and sample codes of each seedling in D are provided in Supplementary Data Table S7. Photography by T. C. Wilson.

Clonality and genetic diversity

Our kinship analysis of E. sp. Cattai detected clonality at all locations, although the extent of clonality varied between sites (Fig. 6). The number of genets with multiple ramets (clones) ranged from one at Logie (out of seven total genets) to eight at Saltwater (out of 69 total genets). The maternal parent of the seedlings was identified, except for those where the parent was not genotyped (Supplementary Data Table S7). A few cases of near clonal relationship between seedlings and adults were found at Logie and Saltwater (Fig. 6). Given their pairwise kinship values were >0.40, these individuals were suspected to be the result of self-pollination (this included five seedlings from Saltwater, and one from Logie, Table S7).

Fig. 6.

Pairwise kinship analysis of Eucalyptus sp. Cattai showing samples from (A) the Logie–Robson Road intersection (Log), Georgia Terrace (Geo), Bannerman Road (Ba), Colbran Avenue (Col), Clarke Way (Cla), Woodlands Road (Woo) and Shoplands Road (Sho); and (B) Saltwater Circle (Sa) only. The heatmap displays kinship estimates for each pair of samples for each site, with red corresponding to the highest kinship (≥0.40 = clone), orange-yellow corresponding to medium pairwise kinship coefficients (>0.25 and <0.40 = sibling), and white corresponding to the lowest kinship (0). The descending red diagonal is the result of an individual matched with itself. Coloured boxes highlight samples that belong to genets of multiple ramets, and the labels adjacent to each box indicate genet type and name (given in Supplementary Data Table S7). In B individuals that belong to genets of multiple ramets (Sa1 to Sa8) are labelled in the kinship heatmap and the locations of the adults are indicated as triangles in the map. Genets made up of a single ramet are shown as white squares in the map.

The population at Saltwater (both adults and seedlings included) had the highest H_E and F_IS (0.31 and 0.213 respectively), and plants from Bannerman had the lowest values (H_E = 0.224 and F_IS = −0.25, Table 2). Overall H_O was highest for Bannerman (0.277) and lowest for Shoplands (0.226). Both H_O and H_E were higher for adult plants than for seedlings in all populations. Seedlings at Saltwater had the highest H_O, H_E and F_IS (0.232, 0.307 and 0.239), while those at Logie had the lowest values (0.159, 0.126 and −0.251 respectively). Pairwise F_ST values ranged from 0.071 to 0.202 (Table 3). The lowest pairwise F_ST with other populations were found for Clarke, while the highest values with other populations were for Bannerman, Colbran and Woodlands. Overall, Shoplands, Colbran, Woodlands and Bannerman had relatively high pairwise F_ST values with either all or most other locations (>0.100, Table 3).

Table 2.

Genetic diversity parameters of Eucalyptus sp. Cattai showing observed heterozygosity (H_O), expected heterozygosity (H_E) and inbreeding (F_IS) of seedlings, adult plants and overall values for each site (codes given in Table 1).

Population	Overall				Adults				Seedlings
	N	H O	H E	F IS	N	H O	H E	F IS	N	H O	H E	F IS
Clarke	21	0.248	0.303	0.169	18	0.263	0.306	0.312	3	0.205	0.175	−0.175
Shoplands	16	0.226	0.268	0.151	9	0.28	0.285	0.018	7	0.173	0.188	0.138
Saltwater	69	0.243	0.31	0.213	19	0.284	0.313	0.083	50	0.232	0.307	0.239
Logie	7	0.233	0.237	0.017	4	0.294	0.26	−0.128	3	0.159	0.126	−0.251
Woodlands	3	0.272	0.237	−0.153	3	0.272	0.237	−0.153
Colbran	4	0.274	0.237	−0.153	4	0.274	0.237	−0.153
Georgia	2	0.27	0.234	−0.184	2	0.27	0.234	−0.184
Bannerman	2	0.277	0.224	−0.25	2	0.277	0.224	−0.25
Overall mean for all populations	124	0.255	0.256		61	0.277	0.262		63	0.192	0.199

Table 3.

Pairwise F_ST of unique adult genets across all populations (i.e. sites) of Eucalyptus sp. Cattai. Codes for sample sites correspond to those given in Table 1. Higher values (i.e. F_ST > 0.100) are highlighted in bold.

	Clarke	Shoplands	Woodlands	Saltwater	Colbran	Logie	Bannerman	Georgia
Clarke	0
Shoplands	0.090	0
Woodlands	0.087	0.121	0
Saltwater	0.071	0.106	0.105	0
Colbran	0.139	0.196	0.189	0.156	0
Logie	0.099	0.143	0.148	0.098	0.187	0
Bannerman	0.103	0.138	0.157	0.106	0.202	0.121	0
Georgia	0.078	0.116	0.126	0.088	0.164	0.128	0.103	0

DISCUSSION

This study produced a large, high-quality SNP dataset that, compared to previous research, provided a more robust and resolved understanding of evolutionary relationships in the red mahoganies and related eucalypts, and supports E. sp. Cattai as a distinct species. The dataset enabled the identification of genetically ‘pure’, hybrid and clonal plants, and revealed a range of historical and contemporary evolutionary mechanisms (e.g. hybridization, post-zygotic barriers and clonality) that have shaped the genetic structure of E. sp. Cattai, thereby improving our capacity to manage genetic diversity for both existing populations and those that need to be translocated to new locations. With the addition of ex situ seedling propagation, this is the first study to collate morphological measurements of E. sp. Cattai seedlings, improving our ability to filter out hybrids in conservation management strategies.

Genomics confirms a rare and yet to be formally described eucalypt species

Our findings strongly support E. sp. Cattai as a distinct species in subgenus Symphyomyrtus section Latoangulatae. In all analyses performed, most samples of E. sp. Cattai grouped together and were separate from all other species (Figs 2 and 4; Supplementary Data Figs S1 and S2). Although E. sp. Cattai has some morphological traits that are distinctive from the species that we found were its closest relatives (E. notabilis, E. resinifera and E. scias), including a crooked trunk covered with thick furrowed bark that is loose to scaly from the base to the top of the tree (Royal Botanic Gardens and Domain Trust, 2021), other traits are shared across species. For example, E. sp. Cattai, E. resinifera and E. notabilis all have adult leaves that are dark green and discolorous, with overlapping leaf length and width measurements, as well as fruits with overlapping size dimensions, hemispherical shape, flat to raised discs and exserted valves (Royal Botanic Gardens and Domain Trust, 2021). We also found that most plants previously identified as hybrids between E. resinifera and E. notabilis were E. sp. Cattai. This includes all samples from Colbran, where E. sp. Cattai was previously considered absent, although an undisclosed number of E. notabilis × E. resinifera hybrids were reported (Klaphake, 2014). In contrast to previous reports (Australasian Virtual Herbarium, 2021), we could not find any populations of E. resinifera or E. notabilis that were sympatric to E. sp. Cattai during our sampling (E. resinifera was the closest, located 10 km away). Thus, our results are consistent with previous work demonstrating that eucalypts can be difficult to identify on the basis of morphological characters alone (e.g. Rutherford et al., 2018, 2019), and confirmed that genetic sequencing is sometimes the only way to determine species identity and hybrid parentage in taxonomically challenging groups.

All analyses in this study indicate that the red mahoganies (section Latoangulatae series Annulares) are not monophyletic, which was consistent with the most recent molecular phylogeny of Jones et al. (2016), but disagreed with classifications based primarily on morphological traits (Brooker, 2000; Slee et al., 2006). As traditionally thought, E. resinifera, E. notabilis and E. scias are closely related, but the other red mahoganies, E. botryoides and E. robusta are not genetically close to this group. Instead, E. botryoides is more closely related to the blue gums, E. grandis and E. saligna, which agrees with Jones et al. (2016) and the most recent classification of Nicolle and Jones (2018). We found that E. robusta is more distantly related to E. resinifera, E. notabilis and E. scias than E. botryoides, and not grouped with any member of section Latoangulatae. The classification of Nicolle and Jones (2018) recognizes E. robusta is not related to the other red mahoganies through its transfer to a different series (ser. Robustae), however, still places this species in section Latoangulatae. Our results therefore indicate a more comprehensive examination of E. robusta is warranted. Overall, our findings are consistent with previous genomic sequencing studies on eucalypts showing how evolutionary relationships can be different to those hypothesized based on morphological evidence (e.g. Jones et al. 2016), and highlights the importance of population genomic work (together with a detailed morphometric analysis) for navigating the complexities in evolutionary mechanics for an improved understanding of relationships within Eucalyptus. A morphometric analysis and formal species description of E. sp. Cattai is currently the focus of ongoing research (T. C. Wilson et al., unpubl. res).

Hybridization in seedlings versus adult plants

Gene flow was detected between E. sp. Cattai and related species (Fig. 4), with our analyses revealing that there were a greater proportion of seedlings than adult plants of hybrid origin (Supplementary Data Table S4). Unfortunately, high levels of reticulation among species prevented the identification of the parent species of these hybrids (NewHybrids indicated multiple potential parent species). This finding is not unexpected given that posterior probabilities calculated in NewHybrids can be difficult to interpret in groups of closely related species or sub-populations (Anderson, 2008). While it is possible that our finding of hybrids potentially having more than two parent species may be due to few samples being included as species B in the NewHybrids analysis, the f3 statistics suggested that seedlings from Logie were admixed between E. sp. Cattai and multiple species (Table S5). The findings here also agree with other studies of Eucalyptus showing that individuals may be the result of hybridization or admixture between more than one species (e.g. Rutherford et al., 2019).

Although many of the study species are very closely related, the hybridization we observed in the adult plants of E. sp. Cattai was not as extensive as found in other groups in subgenus Symphyomyrtus (Flores-Rentería et al., 2017). For example, in a previous study a high proportion (~40 %) of individuals sampled of a rare eucalypt (E. tetrapleura) were hybrids (Rutherford et al., 2019). In our study, despite the greater vigour displayed by the E. sp. Cattai hybrid seedlings compared to the pure seedlings (greater height and leaf width), the prevalence of hybrids detected among the adults was low (6 % including clones, 10 % excluding clones, Supplementary Data Table S4) indicating that post-zygotic factors may be limiting hybrid establishment. These results are consistent with other studies demonstrating that pure eucalypt seedlings outperform hybrids over the longer term (Lopez et al., 2000). For example, hybrids of E. globulus and E. nitens had a lower survival and reproductive fitness over a 20-year period compared to pure individuals (Larcombe et al., 2016). It should also be noted that the adult individuals of E. sp. Cattai are long-lived mallees (older mallees are hundreds of years old, with some individuals estimated to be almost 2000 years old, Kennington and James, 1997). Therefore, the relatively low number of adult hybrids probably reflects pre-urban conditions. In contrast, the greater proportion of hybrid seedlings may be due to increased urbanization in the area and population fragmentation. Regardless, hybrid and pure seedlings of E. sp. Cattai should be monitored over the long term to assess whether intrinsic post-zygotic barriers account for the smaller number of adult hybrids found in the field. Our observations correspond with the findings of Lee (2017) who reported an absence of in situ E. sp. Cattai seedlings, which may be a consequence of genetic swamping resulting in a short-lived nature of hybrids.

In our previous study, we found hybrids of E. tetrapleura could morphologically resemble pure trees, and that genomic sequencing was sometimes the only way to distinguish hybrid from non-hybrid individuals (Rutherford et al., 2019). This finding was supported in the current study, where it was revealed that while hybrids as a group could be statistically differentiated from non-hybrids (Supplementary Data Fig. S6), morphology alone does not allow for the identification of hybrid versus non-hybrid individuals at the seedling stage (Fig. 5; Tables S3 and S5).

Clonality, self-compatibility and genetic health

We found that clonality was present at all sampled sites, and clonal samples in some cases were located as much as 30 m apart (Fig. 6), suggesting extensive rhizomatous growth and long-term survival can occur in E. sp. Cattai. All populations of E. sp. Cattai had similarly high levels of heterozygosity, but half the populations showed positive F_IS values, suggesting some biparental inbreeding (Table 2). Additionally, we found a few cases of near clonal relationships between ex situ seedlings and in situ adults at Saltwater and Logie, indicating that E. sp. Cattai is capable of selfing (Supplementary Data Table S7). Although many Eucalyptus species are highly self-incompatible (Ellis and Sedgley, 1992), self-pollination has been detected in a number of species, such as E. grandis and E. urophylla (Jones et al., 2008; Wu et al., 2015). Moreover, in some species, such as the rare mallees E. rhodantha and E. caesia, self-pollination can be high (Sampson et al., 1989; Bezemer et al., 2016). However, selfing in many eucalypt species results in lower seed yields, reduced seed viability and suppressed seedling growth (Horsley and Johnson, 2007). Thus, self-compatibility in many eucalypts probably involves late pre- and post-zygotic components, and selection may remove most of the selfed progeny from the population (James and Kennington, 1993). The levels of heterozygosity in the E. sp. Cattai adults were higher than in the seedlings, suggesting that although low-level inbreeding and selfing might take place, post-zygotic mechanisms are likely to favour outcrossed individuals.

Use of genomic data to guide the translocation of a critically endangered species

Translocations should involve maximizing the persistence and resilience of the target species to respond to environmental changes (Weeks et al., 2011). Seedlings representing the maximum genetic diversity of E. sp. Cattai were chosen for translocation to suitable nearby habitat. A total of 60 seedlings were planted at two locations. These seedlings were mainly sourced from Saltwater (48 out of 60) because this site was partially cleared for development. Additionally, three seedlings were selected from Clarke, two from Logie and seven from Shoplands, based on their capacity to contribute diversity (Bragg et al., 2021). At each location, seedlings were planted in five clumps (seedlings in clumps at least 8 m apart and clumps at least 20 m apart) with each clump comprising six seedlings. This arrangement was to mimic the natural spacing observed at the source populations and to increase the likelihood of cross-pollination, seed production and recruitment (McCallum et al., 2018). Further analyses were performed to ensure that closely related seedlings were not planted close together (Bragg et al., 2021). At the first location, no plants with kinship >0 were located less than 35 m apart, while at the second location, some seedlings with kinship >0.2 were planted less than 20 m apart (this could not be avoided because seven plants at this site were closely related). Future monitoring is required to assess the success of the translocation of E. sp. Cattai and to ensure that translocated populations are genetically viable (van Rossum and Hardy, 2022). Specifically, genetic analysis of any in situ seedlings from planted stands is needed to ensure that genetic purity and health of individuals are maintained. Follow-up monitoring examining the response of populations to environmental change is also highly important for the future conservation of the species.

CONCLUSIONS

Our findings revealed many evolutionary processes (e.g. hybridization and clonality) operating within the one rare and restricted species. Such evolutionary processes have been observed operating together in other rare eucalypts (e.g. Bradbury et al., 2016) and these should be taken into account in conservation management solutions. Overall, our study demonstrates the ability of genomics to concomitantly resolve species boundaries within a taxonomically problematic group, while improving our understanding of evolutionary drivers to assist with ensuring long-term survival of translocations for threatened and highly localized species.

SUPPLEMENTARY DATA

Supplementary data are available online at https://academic.oup.com/aob and consist of the following. Fig. S1: PCA of all species based on genomic data. Fig. S2: LEA structure plot of E. sp. Cattai and related eucalypts. Fig. S3. Outputs from threshold models fit using the optM function. Fig S4: Residual matrix generated in TreeMix under a scenario of two migration events. Fig. S5: Frequencies of pure and hybrid seedlings and adult plants of E. sp. Cattai. Fig. S6: Morphological traits of E. sp. Cattai seedlings. Table S1: Q values for structure analysis implemented in LEA of E. sp. Cattai and related eucalypts. Table S2: Q values for structure analysis of E. sp. Cattai only. Table S3: NewHybrids analysis. Table S4: Proportion of seedlings and adults of hybrid origin. Table S5: f3 statistics testing for admixture. Table S6: MANOVA of seedling morphological traits. Table S7: Kinship analysis of E. sp. Cattai.

CONFLICT OF INTEREST

The authors declare no conflict of interest.

FUNDING

This work was supported by NSW Government’s Saving Our Species conservation programme. Susan Rutherford was supported by the Jiangsu University Research Foundation (20JDG055) and National Natural Science Foundation of China (32001087).