Evolutionary history of Australian samphires (Salicornieae, Amaranthaceae)

Abstract

Background and Aims

Tecticornia is the most species-rich genus within the tribe Salicornieae. These halophytes are distributed across the Australian continent along coastlines and inland salt lake shores, playing a key ecological role in these hostile habitats. However, species delimitation within the genus remains controversial and little is known about infrageneric phylogenetic relationships. Therefore, the primary aim of this study was to infer the evolutionary history of Tecticornia and to genetically assess the reliability of current species concepts.

Methods

We sampled multiple accessions per taxon from nearly all currently recognized species in Australia. We used a target enrichment approach with two bait sets: Angiosperms353 and a custom Salicornieae bait set (Salibaits). Analyses were performed using HybPiper, and we addressed paralogy using a tree-based approach. In addition, we tested the potential influence of missing data and/or missing gene trees on the topology of the final phylogenetic tree.

Key Results

Despite extensive gene tree discordance and the presence of short branches, the customized Salibaits set consistently produced better-resolved trees than the Angiosperms353 bait set. Missing data were found to have a negligible effect on the final tree inference. These data highlight that there is genetic support for lineages in line with observed morphological variation, suggesting markedly more taxon diversity than is currently circumscribed. While we have shown there is genetic evidence to support the characterization of several new species awaiting formal description, it is clear that further molecular and morphological investigation is required to resolve continent-wide species aggregates, each comprising multiple novel taxa.

Conclusions

The target enrichment method effectively addressed the challenges of species delimitation in Tecticornia posed by reduced morphology and high ecological plasticity. We have shown that while there are several complexes constituting variants widely distributed across the Australian continent, some well-defined taxa have highly restricted distributions, which may represent conservation priorities.

INTRODUCTION

The tribe Salicornieae (Amaranthaceae) sensu Morales-Briones et al. (2021) comprises 12 genera and over 100 species that are often found growing on alkaline soils in saline mudflats, coastal salt marshes and along the shores of inland salt lakes worldwide (Piirainen et al., 2017). Fluctuating water levels, high salinity, waterlogging and extreme temperatures are just some of the features that make these environments uninhabitable to most plants (Pomeroy and Wiegert, 2012). In order to survive in such inhospitable habitats, Salicornieae have evolved a plethora of highly specialized anatomical and physiological adaptations, including significantly reduced leaf area, stem succulence, root aerenchyma, advanced osmoregulation and optimization of water use, and some species undertake carbon assimilation via C₄ photosynthesis (Kadereit et al., 2006; Voznesenskaya et al., 2008; Colmer et al., 2009). Alternate phyllotaxis and shrubiness are considered plesiomorphic in this group, and annuality, which has evolved several times independently, is believed to have allowed the successful occupation of ecological niches in cooler temperate zones (Kadereit et al., 2006; Cousins-Westerberg et al., 2023). While convergent morphological evolution in this lineage resulted in a highly uniform morphology characterized by succulent, articulated stems (lacking leaves) and inconspicuous flowers, the daily and seasonal changes in water and salinity levels in these habitats are reflected in the plants’ immense phenotypic plasticity (Ingrouille and Pearson, 1987; Kaligarič et al., 2008; Piirainen, 2015). There is also evidence of significant polyploidization (ranging from 2x to 8x) and hybridization within the tribe (Castro and Fontes, 1946; Contandriopoulos, 1968; Cristofolini and Chiapella, 1970; Nilsson and Lassen, 1971; Castroviejo and Coello, 1980; Castroviejo and Lago, 1992; Lago and Castroviejo, 1993), with taxa commonly exhibiting intraspecific polyploidy across populations (Shepherd and Yan, 2003). This combination of factors prevents us from fully understanding the complex evolutionary history of this ecologically highly specialized plant group and hinders the identification of known species, as well as the delimitation of new taxa (Kadereit et al., 2007, 2012b; Teege et al., 2011; Vanderpoorten et al., 2011; Slenzka et al., 2013).

Previous studies have shown that succulence and salinity tolerance were already present in the most recent common ancestor of Salicornieae (Kadereit et al., 2012a). These ancestral adaptive traits likely facilitated the successful colonization of inland continental lake systems surrounded by xeric steppe or desert habitats in the Late Eocene/Early Oligocene in continental Eurasia along the northern margins of the slowly retreating Tethys Sea (Kadereit et al., 2006). During this period, one of the most pronounced climatic cooling events occurred, which, with the retreat of the warm West Siberian Sea, the closure of the Turgai Strait and the uplift of the Tibetan Plateau, probably enhanced the aridification of the Asian interior (Hurka et al., 2019). The cooling and associated aridification continued throughout the second half of the Oligocene, Miocene and Pliocene, during which time the Salicornieae expanded around the globe through numerous long-distance dispersal events (Piirainen et al., 2017). One of these events during the late Miocene to Pliocene brought the most recent common ancestor to Australia, giving rise to the most species-rich genus in the tribe, Tecticornia (Kaderreit et al., 2005). Subsequently, the closely related genus Salicornia arrived sometime during the Pliocene (Kadereit, 2005; Steffen et al., 2015) and gave rise to three extant species, which collectively, along with Tecticornia, are referred to as Australian samphires.

Tecticornia (Figs 1 and 2) currently comprises around 68 hygrohalophytic taxa (Council of Heads of Australasian Herbaria, 2006), almost all of which are confined to arid Australia. Only T. australasica and T. indica subsp. indica, which occur along the tropical Australian coast, also extend their range to Java and New Guinea and subtropical coastal regions of southern Asia and eastern Africa, respectively. Tecticornia indica subsp. ciliolata is endemic to Java and the Lesser Sunda Islands (Shepherd et al., 2004). The genus is predominantly represented by stem-succulent subshrubs or shrubs, although there are a few annual or short-lived perennial species such as T. australasica and T. verrucosa (Fig. 1C). Samphires are widely distributed across the Australian continent (AVH, 2025; https://avh.ala.org.au/occurrences/search?taxa=Tecticornia#tab_mapView) along coastlines and around the margins of inland lake shorelines and saline river beds. The highest species diversity is recorded from the inland salt lakes of Western Australia, across the Murchison, Avon Wheatbelt, Coolgardie and Yalgoo bioregions (Western Australian Herbarium, 1998–; Australian Government Department of Agriculture, Water and the Environment, 2012), with a second centre of diversity in the Lake Eyre Basin, an endorheic river system, which spans 1.2 million km² across the borders of South Australia, Northern Territory, Queensland and New South Wales (AVH, 2025). Tecticornia forms the dominant vegetation in these saline habitats and establishes long-lasting stands that play an important role in soil stabilization and nutrient cycling, as well as providing shelter, nesting sites and a food source for local fauna (McComb et al., 1995; Saintilan, 2009).

Fig. 1.

Habit diversity in Tecticornia. (A) T. triandra (syn. Pachycornia triandra). (B) T. bulbosa. (C) T. verrucosa. (D) T. peltata. (E) T. moniliformis (syn. Sclerostegia moniliformis). (F) T. lylei. (G) T. bibenda. (H) T. uniflora (syn. Tegicornia uniflora). Vouchers: M. Alcorn (unvouchered) https://www.inaturalist.org/observations/53575011 (A); K.A. Shepherd & A. Žerdoner Čalasan KS 1896 (PERTH 09543147) (B); K.A. Shepherd & S.R. Willis KS 1797 (PERTH 09187766); K.A. Shepherd & S.R. Willis KS 1788 (PERTH 09187456); K.A. Shepherd & S.R. Willis KS 1787 (unvouchered); K.A. Shepherd & S.R. Willis KS 1789 (PERTH 09187421); K.A. Shepherd & C.F. Wilkins KS 836 (PERTH 07462786); K.A. Shepherd & C.F. Wilkins KS 1749 (PERTH 09187448). Images by M. Alcorn (A) and K.A. Shepherd (B–H).

Fig. 2.

Variation across the Tecticornia halocnemoides group. (A) T. halocnemoides subsp. caudata. (B) T. halocnemoides subsp. longispicata. (C) T. aff. halocnemoides subsp. caudata. (D) T. sp. Chinnocup (K.A. Shepherd KS 1191). (E) T. sp. Dennys Crossing (K.A. Shepherd & J. English KS 552). (F) T. halocnemoides. (G) T. halocnemoides. (H) T. halocnemoides subsp. tenuis. Vouchers: K.A. Shepherd KS 1183 (PERTH 08591180) (A); K.A. Shepherd & T.A. Edwards KS 1343 (PERTH 09664491) (B); K.A. Shepherd KS 1154 (PERTH 08591148) (C); K.A. Shepherd KS 1192 (PERTH 08631077, same population as PERTH 08016186) (D); K.A. Shepherd & S.R. Willis KS 1802 (PERTH 09187731) (E); K.A. Shepherd & L. Wai-Leng KS1321 (PERTH 09241124) (F); K.A. Shepherd & T.A. Edwards KS 1324 (PERTH 09662693) (G); K.A. Shepherd & T.A. Edwards KS 1367 (PERTH 09662545). Images by K.A. Shepherd.

Samphire species exhibit variation in tolerance of salinity, waterlogging and prolonged periods of drought, which often results in clear species zonation across the shoreline profile of inland salt lakes. For example, T. pergranulata subsp. pergranulata and T. medusa can survive complete submergence for up to 10 months and continue photosynthesizing (English, 2004; Pedersen et al., 2006; Rich et al., 2008; Konnerup et al., 2015). These species are, therefore, often found close to the lake pans around the high flood waterline. In contrast, species such as T. indica subsp. bidens and T. mellarium are less tolerant of extreme salinity and waterlogging, and are found on the slightly elevated dunes surrounding flat lake salt pans and floodways (English and Colmer, 2013; Moir-Barnetson et al., 2016). Many populations of Tecticornia face growing threats due to changes in hydrology from the impacts of climate change (Monie et al., 2025), increased land clearing and infrastructure development from mining activity or pollution from agricultural run-offs. Despite their significance across the landscape, an understanding of the true diversity and drivers of speciation in Tecticornia are lacking. While several new species have been recognized in recent years (Shepherd, 2007a, b, 2018, 2020; Shepherd and Van Leeuwen, 2007, 2011; Shepherd and Lyons, 2009; Shepherd and Žerdoner Čalasan, 2024), several widespread species complexes that exhibit considerable variation in morphology remain unresolved. Furthermore, many salt lakes in remote inland desert regions are largely unexplored, and so potentially new species remain undiscovered. Therefore, it is likely that the current number of species is a significant underestimation of the true diversity of samphires present in Australia.

Tecticornia is distinguished from the rest of the members of the Salicornieae by having articulate stems without sclereids, opposite phyllotaxy, two or three perianth lobes (with the third one always abaxial), only one abaxial stamen, and seeds with abundant perisperm (Shepherd and Wilson, 2007). While most species have bisexual flowers arranged in a spike-like thyrse with flowers in opposite-decussate clusters of three, some species show varying degrees of deviation from this pattern, leading to their previous placement in separate genera (Wilson, 1980; Shepherd et al., 2005a). For example, T. uniflora (formerly Tegicornia) is a mat-like, dioecious species that has single male or female flowers on separate plants (Fig. 1H). Several species are also andromonoecious, where each triad of flowers within the inflorescence has a central bisexual flower flanked by a male flower on either side, as seen in T. triandra (formerly Pachycornia) (Fig. 1A) and T. arbuscula, T. disarticulata, T. medullosa, T. moniliformis (Fig. 1E) and T. tenuis (all formerly Sclerostegia; highlighted portion of Fig. 3A, denoted by the abbreviation AM). Early molecular studies (Shepherd et al., 2004, 2005b; Kadereit et al., 2006) confirmed these three genera, along with Halosarcia, were not monophyletic and were subsequently synonymized under Tecticornia (Shepherd and Wilson, 2007).

Fig. 3.

Coalescence-based phylogeny of group II (A) and group I (B) of the genus Tecticornia inferred using ASTRAL from 463 nuclear loci with stitched contigs, allowing up to 75 % missing data. Local posterior probabilities of 1.00 are indicated by ***, those between 0.95 and 0.99 by **, and those between 0.90 and 0.95 by *. Local posterior probabilities below 0.90 are omitted. A shaded area in the upper left corner indicates the specific region of the phylogeny shown. Orange boxes on the right indicate taxa with dorsoventrally flattened flowers, yellow boxes represent taxa with rectangular-shaped flowers, pink boxes denote taxa with firm infructescences, and blue boxes correspond to taxa with soft infructescences. Grey-shaded squares labelled ‘AM' represent andromonoecious taxa, while grey-shaded squares labelled ‘C₄’ represent taxa known to perform C₄ photosynthesis. ‘±❀⇓' denotes taxa with three or fewer flowers per cyme, ‘❀⇓' denotes taxa with fewer than three flowers per cyme and ‘❀⇑' denotes taxa with more than three flowers per cyme. ‘C₃–C₄?’ denotes taxa with a non-standard C₃ anatomy (according to Dakin, 2021).

Tecticornia is the only genus in the Salicornieae that has representatives known to perform C₄ photosynthesis (Kadereit et al., 2003; Sage 2016; highlighted portion of Fig. 3A, denoted by the abbreviation C₄). However, even within this genus, this syndrome occurs in only one clade (Shepherd et al., 2004) comprising T. indica, a widely distributed species complex consisting of four subspecies (including a fifth representative outside of Australia) and T. bibenda, which was previously known as Halosarcia sp. Yanneri Lake (S. van Leeuwen 3002) (Fig. 1G) and is a geographically highly restricted species known from only two small populations in the Little Sandy Desert of Western Australia (Western Australian Herbarium, 1998, see https://florabase.dbca.wa.gov.au/browse/profile/31677;  Shepherd and van Leeuwen, 2007). It has been argued for many C₄ lineages that the C₄ syndrome evolved as a response to decreasing atmospheric CO₂ concentration coupled with increased aridity and represents an evolutionary advantage of xerophytes over other mesic elements due to suppressed photorespiration (Sage et al., 2018). However, C₄  Tecticornia species do not seem to follow this pattern, as they occupy similar habitats as their C₃ relatives, pointing towards a different evolutionary history and ecology of the C₄ syndrome within the genus (Dakin, 2021; Berasategui et al., 2023).

Polyploidization is common in Tecticornia, and different cytotypes have been observed not only within a species but within populations and even between progeny of the same individual. In the case of T. leptoclada subsp. inclusa and T. pterygosperma subsp. pterygosperma, diploid and triploid individuals were found among siblings in the former, while triploids and hexaploids were recorded in seedlings germinated from seed collected from a single mother plant of the latter (Shepherd and Yan, 2003). This led the authors to hypothesize that certain Tecticornia species may exhibit facultative apomixis as a means of maintaining individuals with uneven ploidy numbers within populations that produce pure-breeding lineages, a proposition supported in part by observations of the possible cleistogamous and self-compatible nature of some species (Wilson, 1980).

Due to their highly reduced morphology, coupled with considerable ecological plasticity, polyploidization and potential apomixis, and the relatively young evolutionary age of the lineage, the biological history of Australian samphires is complicated and poorly known. Further factors, such as incomplete lineage sorting and hybridization, add further complexity to their biological history. While early molecular phylogenetic studies based on Sanger sequencing confirmed Tecticornia was monophyletic and placed as sister to Salicornia, resolution across the genus was poor with little backbone support (Shepherd et al., 2004, 2005a, b; Kadereit et al., 2006; Cousins-Westerberg et al., 2023).

The main aims of this study were (1) to resolve the backbone of Tecticornia, (2) to compare the resolution power of the universal Angiosperms353 bait kit and the customized Salibaits kit, (3) to genetically circumscribe taxon diversity across morphologically variable species aggregates, and (4) to identify potential drivers of diversification across the genus. To tackle this, we carried out a target enrichment-based approach with two sets of baits for the majority of Australian species of Tecticornia and Salicornia, including closely related genera. We wanted to test the ability of the increasingly popular Angiosperms353 bait kit (Johnson et al., 2019; Zuntini et al., 2024) to discriminate between samphire species by participating in phase 2 of the Genomics for Australian Plants initiative (GAP, https://www.genomicsforaustralianplants.com/tecticornia-salicornia/; Simpson et al., 2025 ) and compare these data with the results using the customized Salibaits kit. To further explore current species’ delimitations across various complexes in the genus, we expanded our sampling to include several accessions per taxon representing their contemporary distribution range as well as morphological variants that appeared to be distinct from the typical form of known species.

MATERIALS AND METHODS

Sampling

Initially, in the framework of the GAP initiative, DNA was extracted from 96 accessions from AD, MEL and PERTH herbaria (Thiers, 2024; Supplementary Data Table S1). However, to test for monophyly of widely distributed and morphologically variable species and potential species aggregates, an additional 134 accessions were subsequently sampled from geographically distant locations across the Australian continent (but with an emphasis on the centre of diversity in Western Australia, with non-Western Australia accessions annotated by the state in which the voucher was collected), bringing the total to 230 samples, which span several accessions of almost all currently accepted species and subspecies, as well as multiple accessions of potentially new taxa awaiting formal description. Some of these have already been designated by an informal phrase name with a reference specimen that is recognized by State institutions (Western Australian Herbarium, 1998; AVH, 2025), while others were given an informal tag name with a locality (e.g. sp. ‘Lake Throssell’) or an ‘aff.’ when thought to be allied to but morphologically distinct from currently known species. Morphological variants coalescing around known species (thought to represent widespread species aggregates, e.g. T. halocnemoides, T. pergranulata and T. undulata/T. pruinosa) were more densely sampled. Sampling was augmented with accessions collected during a field trip in Western Australia in December 2022, and the duplicates of these were deposited in PERTH and M. The outgroup samples comprised closely related Salicornieae obtained from AAU, MELU, MJG, OSBU and PERTH (sequenced here), as well as less closely related Caryophyllales published previously by Morales-Briones et al. (2021). Further information on the vouchers is available in the Supplementary Data Table S1. The initial 96 samples analysed within GAP were sequenced twice using Angiosperms353 and the Salibaits kit, whereas the additional 134 samples were analysed using only the Salibaits kit.

Bait design

We designed a customized bait kit for chenopods I and II (Chenopodiaceae sensu Morales-Briones et al., 2021) that was tailored for Salicornieae (referred to as the ‘Salibaits’ kit). We identified putative single-copy nuclear (SCN) genes with MarkerMiner v.1.2 (Chamala et al., 2015) with default settings, except for the minimum transcript length, which was set to 500 bp. For this, we used the transcriptomes of 11 species distributed across chenopods I and II (Supplementary Data Table S2) and the genomes of Arabidopsis thaliana (Gan et al., 2011), Atriplex hortensis (Hunt et al., 2020), Chenopodium quinoa (Jarvis et al., 2017) and Spinacia oleracea (Xu et al., 2017) as references. SCN genes identified with MarkerMiner were further filtered and split into exons using GoldFinder (Vargas et al., 2019), filtering for loci at least 500 bp in length and spanning over at least six species, which resulted in 490 exons from 424 genes across chenopods I and II. For bait design, we finally filtered the 490 exon alignments to include only sequences from Salicornieae (Heterostachys olivascens (Speg.) Molfino and Salicornia pacifica Standl.) and Suaedeae (Suaeda maritima). This resulted in a final set of 488 exons from 424 genes. The custom set of 80 bp biotinylated RNA baits (myBaits) was manufactured by Daicel Arbor Biosciences (Ann Arbor, MI, USA) with a ×2 tiling density.

DNA extraction, library preparation and sequencing

The DNA extractions, and library preparation and sequencing were conducted at the Australian Genome Research Facility in Adelaide and Melbourne, respectively, following their standard protocols, which are available online: https://www.genomicsforaustralianplants.com/protocols/. Dried plant tissue (20–30 mg) was ground using a TissueLyser II (Qiagen) with tungsten carbide beads, and genomic DNA was extracted using the standard DNeasy^® Plant Mini Kit protocol (Qiagen) on a fully automated spin column-based nucleic acid extraction station, QIAcube Connect (Qiagen). DNA quality and quantity were checked using 1 % E-gel with Sybr Safe dye (Thermo Fisher) and the Quantifluor dsDNA assay (Promega), respectively. Library preparation was performed according to the standard protocol of the NEBNext Ultra II FS Library Prep Kit (New England Biolabs, Ipswich, MA, USA), with targeted inserts of ∼350 bp. After initial enzymatic fragmentation and end repair, adapter ligation, size selection (for samples with initial input >100 ng) and clean-up of adapter-ligated DNA fragments were performed. This was followed by a PCR enrichment and a clean-up step. Library quality was assessed using the Perkin Elmer GX Touch II with High Sensitivity DNA assay. Sixteen samples of similar quality and quantity were pooled and hybridized using the Angiosperms353 bait kit (Johnson et al., 2019) at 65 °C with the Arbor Biosciences MyBaits Expert Plant Angiosperms353 v.1 bait set with V5 chemistry (Cat. No. 308108.v5). This resulted in the generation of six pooled libraries (i.e. 96 samples). In parallel, the same 96 samples, supplemented with a further 134 new samples (see above), were pooled into 20 additional libraries and hybridized using the Salibaits kit at 64 °C using V5 chemistry (Cat. No. 300416.v5-ARB) according to the manufacturer’s instructions. Sequencing of both batches was performed on a NovaSeq 6000 (Illumina, San Diego, USA) using v.1.5 chemistry and 150 bp paired-end reads. We ended up with three datasets: a 96-taxon dataset using both the universal Angiosperms353 bait set as well as the Salibaits bait set, and a 230-taxon dataset using only the Salibaits set.

Data cleaning, assembly, extraction and alignment

Initially, adaptor trimming, as well as quality trimming and filtering, were performed using the option captus_assembly clean integrated into the CAPTUS pipeline (Ortiz et al., 2024) by removing read regions with phred quality scores <20 as well as complete reads with average minimum phred quality score <20 using the Falco algorithm (De Sena Brandine and Smith, 2021). Then, the raw reads were deduplicated using ParDRe v.2.2.5 (González-Domínguez and Schmidt, 2016) using flag –z 9 for maximum compression level. Further analyses were conducted in HybPiper v.2.1.8 (Johnson et al., 2016). Firstly, reads were assembled against the fixed target file using the DIAMOND algorithm (Buchfink et al., 2015) with a percentage identity threshold for retaining exonerate hits set to 75. We used two different target files: for the Angiosperms353 dataset we utilized a filtered mega353 target fasta file (McLay et al., 2021) using only sequences from Amaranthaceae (and former Chenopodiaceae), and for the Salicornieae dataset we used a custom target file with the sequences used for bait design. After generating the statistics, the results were visualized using a custom R script to assess the appropriate coverage cut-off for SPAdes (Bankevich et al., 2012) for each sample, and the assembly step was repeated based on these findings. Using HybPiper’s option recovery_heatmap, we identified loci and taxa with poor recovery, which were removed from further analyses.

Using paralog_retriever, we extracted all individual non-chimeric paralogue copies of the targeted loci in the 230-taxon dataset. These were then concatenated with a subset of the outgroup from whole-transcriptome data generated by Morales-Briones et al. (2021) using a modified bash script. For these samples, we used previously generated orthologous sequences from which we extracted loci using the function captus_assembly extract, retaining only hits with the 75 % minimum identity and 50 % minimum coverage to the reference protein. The sequences of the newly generated locus files were aligned using MACSE v.2.08 (Ranwez et al., 2018), an algorithm that accounts for frameshifts and stop codons. Shifted codons were replaced by gaps, and the maximum likelihood approach implemented in IQ-TREE v.2.1.2 (Minh et al., 2020) was employed to generate the tree, using extended model selection (Kalyaanamoorthy et al., 2017) followed by tree inference and ultrafast bootstrapping (Hoang et al., 2018).

Orthologue/paralogue detection and final tree inference

In the 230-taxon Salibaits dataset, the paralogue detection followed scripts from Yang and Smith (2014) and Morales-Briones et al. (2022). Firstly, we masked mono- and paraphyletic tips using a customized Python script and trimmed spurious tree tips with TreeShrink v.1.3.9 (Mai and Mirarab, 2018), excluding the outgroups. Secondly, we sequentially pruned the paralogues following a monophyletic outgroup (MO) approach to output one-to-one orthologues when no duplications were detected, and retained only those locus trees that contained information for at least 25 and 50 % of the taxa, respectively. From these trees, we wrote new locus alignment files using a customized Python script, realigned them using MACSE v.2.08 (Ranwez et al., 2011), trimmed the newly generated alignment by removing all columns with occupancy <30 % in pxclsq v.1.3 (Brown et al., 2017) and generated the trees employing IQ-TREE v.2.1.2 (Minh et al., 2020), under the same conditions as specified above. Final locus trees were then concatenated, rerooted accordingly, trimmed again with TreeShrink v.1.3.9 (Mai and Mirarab, 2018) and analysed using ASTRAL v.5.7.8 (Zhang et al., 2018).

In the two 96-taxon datasets, no proper paralogue detection was performed due to the lack of appropriate outgroups (see above). Here, instead of paralog_retriever we used the HybPiper option retrieve_sequences, which saves the longest individual locus copies with the highest sequence similarity to the target file. The generated locus files were aligned and processed as specified above. Due to the potential issues in using stitched contigs for phylogenomic inference as well as missing data, we generated four different 230-taxon data sets of trees during paralogy detection: (1) trees with stitched contigs and the maximum number of missing taxa per locus set to 75 %, (2) trees with stitched contigs and the maximum number of missing taxa per locus set to 50 %, (3) trees without stitched contigs and the maximum number of missing taxa per locus set to 75 % and (4) trees without stitched contigs and the maximum number of missing taxa per locus set to 50 %.

Inference of gene tree discordance and whole-genome duplications

The discordance among gene trees was investigated using PhyParts (Smith et al., 2015), using only individual gene trees with bootstrap support >70 %. The results were plotted in a form of pie charts using customized Python script phypartspiecharts_missing_uninformative.py (https://bitbucket.org/dfmoralesb/target_enrichment_orthology) to distinguish between uninformative and missing data. To assess the proportion of loci and sites that support the illustrated phylogeny, we furthermore carried out a concordance factor analysis implemented in IQ-TREE 2 (Minh et al., 2020). Finally, to infer potential whole-genome duplication events, we extracted and mapped the orthogroups onto the ASTRAL species tree using customized Python scripts extract_clades.py and map_dups_mrca.py from https://bitbucket.org/blackrim/clustering (Yang et al., 2018), setting the minimum input taxa value to 50.

RESULTS

Tecticornia backbone phylogeny

The Tecticornia phylogeny generated from the 230-taxon Salibaits dataset was for the most part well supported, with the majority of local posterior probability values being >0.95; however, some short branches and high discordance were found (Fig. 3, Supplementary Data Fig. S1). After paralogy detection, dataset 1 retained the highest number of gene trees, i.e. 463, whereas the other three datasets retained 454, 457 and 446 gene trees, respectively. Because all four datasets retrieved similar topologies (data not shown), we here present only the ASTRAL tree based on dataset 1, with gene trees generated from stitched contigs and the maximum number of missing taxa per locus set to 25 % (Fig. 3, Supplementary Data Fig. S2). The orthogroup mapping revealed no elevated levels of genome duplication in Tecticornia, with duplication levels spanning from 0.0024 to 0.0265 (Supplementary Data Fig. S3).

Salicornia was retrieved as the sister clade to Tecticornia, within a monophyletic Salicornieae (Fig. 3, Supplementary Data Figs S2 and S4), which also included representatives of the genera Microcnemum, Arthroceras, Arthrocaulon, Halostachys, Kalidium, Heterostachys and Allenrolfea. Tecticornia was broadly sampled with 222 accessions representing 57 described taxa, 9 phrase-named taxa, and several morphologically variable potentially new taxa, and was shown to be monophyletic (Fig. 3, Supplementary Data Figs S2 and S4). The genus largely resolved into two well-supported groups, Group I (= grade I) (Fig. 3B) and Group II (= clade II; Fig. 3A), with T. lylei (Fig. 1F) placed as sister to them (Fig. 3B, Supplementary Data Figs S2 and S4).

Comparison of Angiosperms353 and customized bait sets

In the absence of suitable outgroups, the two 96-taxon phylogenies were rooted with Arthrocaulon franzii and A. macrostachyum, mirroring their placement in the 230-taxon dataset phylogeny (Supplementary Data Fig. S2). In both phylogenies, Salicornia was placed as sister to Tecticornia, and both genera were rendered monophyletic with maximum statistical support (Supplementary Data Fig. S5). All three species aggregates recognized in the expanded analyses (Fig. 3) were genetically well characterized in the 96-taxon datasets and were also supported as monophyletic with either maximum or near-maximum support (Supplementary Data Fig. S5).

In both datasets the support for morphologically well-determined Group II was inferred with markedly different statistical support (Supplementary Data Fig. S5). The most profound difference was in the placement of the species closely related to Group II. In the Angiosperms353 dataset, T. uniflora and T. lylei formed a poorly supported sister species relationship, with T. loriae being most closely related to Group II, whereas in the Salibaits dataset T. lylei was placed as a sister species to the whole genus, with T. loriae resolved as sister species to Group II, and T. uniflora as a sister species to the Group II and T. loriae together (Supplementary Data Fig. S5).

Within Group II, the C₄ clade was retrieved as monophyletic in both datasets; however, further topological differences were observed in the relationships of the closest C₃ relatives, comprising T. peltata and former Tecticornia s.s. (sensu Wilson, 1972), albeit in both datasets with low statistical support (Supplementary Data Fig. S5). Furthermore, both datasets rendered T. indica polyphyletic. While the Angiosperms353 dataset supported apparent morphological similarities in T. moniliformis, placing it as a sister to T. arbuscula, this was not the case in the Salibaits dataset, rendering T. moniliformis polyphyletic. While discordance levels were very high in both datasets (Supplementary Data Fig. S6), the Angiosperms353 dataset showed markedly higher levels of main discordant alternatives and missing data. Moreover, the shallower nodes in the phylogenetic tree of the custom bait dataset had, on average, higher local posterior probability values (Supplementary Data Fig. S5).

In the 96-taxon dataset analysed with the Angiosperms353 bait set, the number of reads per sample spanned from 1 257 971 to 6 387 851. Only an average of 10.38 % (ranging between 0.7 and 21 %) of all reads were on target, and out of 353 loci targeted the average number of loci that had extracted coding sequences was 313, spanning between 175 and 335. The retrieved locus length averaged 448 bp, ranging between 75 and 1491 bp. A concatenated dataset was 189 408 bp long and comprised 22 673 parsimony-informative positions, 166 735 parsimony-uninformative positions and 178 151 variable positions.

The same 96-taxon dataset yielded markedly better results when analysed with a Salibaits set. Here, the number of reads per sample spanned from 974 330 to 111 69 208, of which between 48.9 and 79.4 % were on target, yielding an average value of 74.85 %. Out of 488 loci targeted with the Salibaits set, the average number of loci that had extracted coding sequences was 483, and it did not drop below 479. The retrieved locus length averaged 986 bp, ranging between 202 and 3797 bp. Here, the concatenated dataset was markedly longer, 484 515 bp, and comprised 172 855 parsimony-informative positions, 311 660 parsimony-uninformative positions and 257 791 variable positions. In neither of the datasets was a comprehensive amount of chloroplast or mitochondrial loci retrieved (data not shown).

Tecticornia clade-level phylogeny

The two major groups recovered in the Tecticornia phylogeny broadly represent distinct morphological groups. Species in Group I (Fig. 3B) have generally more rectangular flowers and infructescences that become soft and pithy or chartaceous (Fig. 4A–C), largely disintegrating at maturity to release seeds ornamented with small bumps and finger-like projections (Shepherd et al., 2005a). However, there were some exceptions within Group I, such as T. crotalus, T. fimbriata and T. willisii, which have the typical soft, chartaceous fruits, but also dorsoventrally flattened flowers (see Fig. 20F in Wilson, 1980). In contrast, the majority of taxa in Group II (Fig. 3A, Supplementary Data Figs S2 and S4) have flowers that are dorsoventrally flattened and generally have fruits and infructescences that become firmly pithy, corky or woody at maturity (Fig. 4D–F). Fruits are usually retained on the mother plant for an extended period, and their seeds tend to be smooth and unornamented. However, there are exceptions, including representatives of the former Tecticornia s.s. sensu Wilson (1972) (including T. verrucosa, T. australasica and T. arborea; highlighted portion of Fig. 3A) and T. annelida, which have dorsoventrally flattened flowers but they become soft and papery in fruit, and easily disintegrate and fall at maturity. Furthermore, nested within Group II, T. peltata and T. leptoclada subsp. inclusa, as well as T. loriae, which was resolved as sister species to the core Group II (Fig. 3A, Supplementary Data Figs S2 and S4), are morphologically more similar to species in Group I in both flower and fruit morphology. Tecticornia lylei exhibits a morphology that resembled both groups, having rectangular flowers (representing Group I) but firm fruits (representing Group II).

Fig. 4.

Tecticornia flower and fruit morphology comparing Group I with square-shaped flowers with soft fruits (A–C) and the dorsoventrally flattened flowers with firm fruits (D–F) of group II (with exceptions). (A) Inflorescence with each opposite-decussate triad of square-shaped flowers exposed above the subtending bract. (B) Side view of a transparent, square-shaped flower showing a withered brown stigma and an ovate seed within. (C) Mature inflorescence with soft, pithy fruits. (D) Inflorescence with each opposite-decussate triad of flowers completely covered by subtending bracts (with only the bilobed stigmas exposed). (E) Side view of a triad of dorsoventrally flattened flowers. (F) Maturing inflorescence showing fruits becoming hardened. Vouchers: Tecticornia lepidosperma (A–C) T. syncarpa (D–F). Images: K. A. Shepherd.

Group I included two large and morphologically complex taxonomic groups. In the first group, the T. pergranulata species aggregate consisted of five subspecies of T. pergranulata, namely monophyletic T. pergranulata subsp. elongata and T. pergranulata subsp. queenslandica, and non-monophyletic T. pergranulata subsp. divaricata and T. pergranulata subsp. pergranulata (Fig. 3B, Supplementary Data Figs S2 and S4). Furthermore, this aggregate comprised morphologically well-defined species T. cupuliformis, T. doliiformis and T. fontinalis as well as phrase-named T. sp. Coorow (P.G.Wilson 12750) WA Herbarium and T. sp. Connewarre (A.C.Beauglehole 70251) K.A.Sheph. (Fig. 3B, Supplementary Data Figs S2 and S4).

The second group, T. halocnemoides species aggregate, included three lineages. The first lineage comprised T. crotalus and T. indefessa, whereas the second one was made up of morphologically well-defined T. fimbriata, T. medusa and phrase-named T. sp. Lake Way (P.Armstrong 05/961) WA Herbarium, among numerous T. halocnemoides accessions, including non-monophyletic T. halocnemoides subsp. caudata, T. halocnemoides subsp. catenulata and T. halocnemoides subsp. longispicata (Fig. 3B, Supplementary Data Figs S2 and S4). The third lineage within this species aggregate comprised accessions of morphological variants of T. halocnemoides mostly not determined to the subspecies level, as well as T. enodis, T. globulifera, T. willisii and phrase-named taxa T. sp. Lake Chinocup (K.A.Shepherd KS 1191) WA Herbarium, T. sp. Dennys Crossing (K.A.Shepherd & J.English KS 552) WA Herbarium and T. sp. Lake Wallambin (K.A.Shepherd KS 1157) WA Herbarium (Fig. 3B, Supplementary Data Figs S2 and S4).

While the sister species to Group II remained unresolved due to poor resolution, both candidate species, T. uniflora (formerly the monotypic Tegicornia uniflora) and T. loriae, were monophyletic, reinforcing the validity of their species concepts (Fig. 3A, Supplementary Data Figs S2 and S4). Within Group II, the topology started with a polytomy of four well-supported branches representing T. peltata, former Tecticornia s.s. sensu Wilson (1972), C₄  Tecticornia clade and the remaining taxa. Within the remaining taxa, the former genera Sclerostegia (including two accessions of T. arbuscula and all accessions of T. disarticulata, T. medullosa, T. moniliformis and T. tenuis) and Pachycornia (T. triandra), both synonymized into Tecticornia in Shepherd and Wilson (2007), formed a clade, which also included T. annelida (Fig. 3A, Supplementary Data Figs S2 and S4). These were in a polytomy with a clade consisting of T. sparagosa and T. entrichoma and a clade with two accessions of T. leptoclada subsp. inclusa.

The next clade comprised the large and complex T. undulata/T.pruinosa species aggregate (Fig. 3A, Supplementary Data Figs S2 and S4). This aggregate included several taxa such as T. auriculata, which was retrieved as sister to several accessions with morphological affinities to T. pruinosa, and an undescribed taxon from Lake Throssell east of Laverton in Western Australia. This clade was in a polytomy with an isolated accession of T. aff. moniliformis collected from Lake Hammersley, three accessions of T. arbuscula and the core clade of the T. undulata/T. pruinosa species aggregate (Fig. 3A, Supplementary Data Figs S2 and S4). This core clade was made up of morphologically and genetically well-defined species such as T. bulbosa, T. calyptrata, T. chartacea, T. laevigata, T. mellarium, T. syncarpa and T. undulata, as well as phrase-named taxa T. sp. Burnerbinmah (D.J.Edinger Nats 101) WA Herbarium, T. sp. Christmas Creek (K.A.Shepherd & T.Colmer et al., KS 1063) WA Herbarium, and T. sp. Yoothapina Station (A.A.Mitchell 883) WA Herbarium. Furthermore, this aggregate included a number of taxa morphologically closely resembling either of the aforementioned species, as well as a number of T. pruinosa accessions, scattered across the aggregate (Fig. 3A, Supplementary Data Figs S2 and S4).

Putative hybrids were also nested within this clade. For instance, T. bulbosa × pruinosa accessions grouped either with T. bulbosa or with T. chartacea and T. syncarpa. Several accessions allied to T. pruinosa formed a grade close to T. undulata and its variants, while one accession from the Gibson Desert represented a distinct lineage (Fig. 3A; Supplementary Data Figs S2 and S4).

DISCUSSION

Overall topology

Our target enrichment approach supports previously known topologies but markedly improves our understanding of the evolutionary history at shallower nodes. While the monophyly of most other Salicornieae genera has been supported in previous studies based on morphology and a low number of loci (Shepherd et al., 2004 (ITS); Kadereit et al., 2006 (ITS, atpB-rbcL); Steffen et al., 2015 (ETS, atpB-rbcL, rpL32–trnL); Piirainen et al., 2017 (ETS, ITS, atpB-rbcL, matK-trnK); Dakin, 2021 (ETS); Cousins-Westerberg et al., 2023 (ETS, ITS, atpB-rbcL, matK-trnK)), we here for the first time illustrate the clear monophyly of the genus Tecticornia, supported not only by the maximum local posterior probability, but also by high levels of gene and site concordance (Fig. 3, Supplementary Data Figs S1, S2 and S4). However, the sister group relationship between Salicornia and Tecticornia is not well supported and shows high levels of discordance (Supplementary Data Fig. S1). Coupled with short branches, these are typical signs of incomplete lineage sorting. While no hybrids are known from these two genera that would indicate recent gene flow, both genera are wind-pollinated, often occur sympatrically and share highly similar morphologies and ecologies. Consequently, we speculate that the last common ancestor of the samphires either had large effective population sizes, with a pre-existing population structure prior to speciation, or the two lineages underwent protracted speciation. Looking at the morphology and ecology of living relatives, both scenarios are possible, as samphires often occur in vast monospecific stands (Pennings and Callaway, 1992; Pellegrini et al., 2022) and show high levels of inbreeding (Teege et al., 2011; Vanderpoorten et al., 2011; Piirainen, 2015; Slenzka et al., 2013). While the placement of closely related genera Microcnemum and Arthroceras remains problematic, as observed in previous studies, we can conclude with higher certainty that this lack of resolution is not a result of limited data, but a true biological signal.

Fidelity of the Angiosperms353 bait set

Regarded as a universal nuclear probe set, the Angiosperms353 bait set has proven useful in providing data that are phylogenetically informative for many plant groups (A. E. Thomas et al., 2021; S. K. Thomas et al., 2021; Thureborn et al., 2022; Bratzel et al., 2023; Joyce et al., 2023; Maurin et al., 2023), even on a population level (Slimp et al., 2021). However, its effectiveness is far from consistent, as numerous studies also reported a lack of resolution on different taxonomic levels (Lee et al., 2021; A. E. Thomas et al., 2021; S. K. Thomas et al., 2021; Yan et al., 2024). Comparative approaches also consistently show advantages of custom bait sets over Angiosperms353 (Shah et al., 2021; Siniscalchi et al., 2021; Yardeni et al., 2022; Hay et al., 2023), as is the case in our study. When comparing the same 96 samples of Tecticornia, the Angiosperms353 bait set only resulted in 10.38–21 % reads being on target compared with 48.9–79.4 % for our customized bait set. While these numbers may appear low, they are influenced by mapping tools, with BLASTx typically mapping fewer reads on average than BWA, but retrieving more genes (https://github.com/mossmatters/HybPiper/wiki/Troubleshooting,-common-issues,-and-recommendations). This is also evident in our data, with accessions with the lowest percentage of on-target reads still yielding proportionally high numbers of recovered loci.

While both datasets analysed here supported Tecticornia as monophyletic, the more conservative Angiosperms353 bait set only moderately helped to resolve shallower nodes. In contrast, while branch lengths remained relatively short, the topology of the custom bait set showed markedly better statistical support values across the phylogeny. Despite proportionally high levels of discordance in both bait sets (Supplementary Data Figs S5 and S6), the number of uninformative trees and the amount of missing data were higher in the Angiosperms353 dataset, possibly due to substantially shorter loci with lower numbers of informative sites, which further points to the potentially limited value of this generalized bait set.

Drivers of diversification

Tecticornia resolved into two well-supported lineages (Fig. 3, Supplementary Data Fig. S2), broadly representing morphologically contrasting groups that either have dorsoventrally flattened flowers, which develop hard fruits enclosing usually smooth seeds (Group II; Fig. 3A) or truncate-flowered species with soft fruits and seeds with different degrees of ornamentation (Group I; Fig. 3B). Exceptions are evident; for example, T. australasica and allied species in Group II have soft fruits and ornamented seeds, while T. fimbriata in Group I has dorsi-ventrally flattened flowers (still with soft fruits and ornamented seeds) while T. enodis, also from Group I, has smooth seeds. Species numbers are similar across both morphogroups and neither of the morphogroups shows markedly different distributions or ecologies. Soft-fruited species in Group II readily drop seeds as the fruits mature and begin to disintegrate, and while it is hypothesized that the sculpturing on the ornamented seeds may facilitate dispersal, no studies have been conducted to test this hypothesis. In contrast, the hard-fruited species in Group I tend to hold fruits and, therefore, seeds for one season or even longer. Some species, such as T. triandra and T. medullosa, may only release seeds when branches fall to the ground and eventually decay (Wilson, 1980). Therefore, it could be assumed that species of Group I may be more narrowly distributed overall due to the potential limits to seed dispersal. This is not the case, as hard-fruited species such as T. indica, T. moniliformis and T. lylei have a widespread distribution across the Australian continent (AVH, 2025). In contrast, 14 soft-fruited species have a much narrower range, as they are only known from one to very few populations (Shepherd and Žerdoner Čalasan, 2024), suggesting other factors may be at play. While samphires are wind-pollinated and have small seeds, usually <2.5 mm long (Shepherd et al., 2005a , b), features that are normally associated with outcrossing and long-distance dispersal, many samphire species in the genus Salicornia are known for their high levels of inbreeding and even cleistogamy (Kadereit et al., 2007).

At first glance, samphires appear ecologically highly restricted, with their distribution closely tied to salt lakes and clay pans. Salt lakes are a familiar feature of arid Australian landscapes, and while samphires form the dominant vegetation fringing these habitats, significant variation is evident in the composition of species within and between sites. Species diversification may be influenced by microhabitat complexity, as while salt lakes are common in the arid zone, they are fragmented across the landscape and show significant differences in geology, soil chemistry and hydrology. For example, the chemical composition of Australian salt lakes is highly variable in levels of potash, boron, lithium and uranium, as well as in their pH values, which can differ significantly even within the same lake system (Bowen and Benison, 2009; Mernagh et al., 2016).

The palaeochannel hydrology of these lakes and associated rivers is complex and not fully understood (Beard, 1999, 2002). As Australia became increasingly arid from the mid-Miocene (15 mya) onwards, the existing inland drainage systems became more endorheic and saline (Martin, 2006). It was postulated that the last common ancestor of Tecticornia arrived on Australian shores ∼8 mya (Kadereit, 2005) and rapidly spread across the country as the continent aridified and inland lakes salinized (Martin, 1998, 2006). Assuming ecological niche conservatism, Tecticornia may have thus greatly benefited from increased salinization, as their primary habitats became more prominent during this period (Habeck-Fardy and Nanson, 2014). However, despite this ongoing trend, there were brief intervals during the Late Neogene and Quaternary when increased precipitation significantly altered the size and connectivity of existing water bodies (Cohen et al., 2011; Fitzsimmons et al., 2013; Sniderman et al., 2016). This can still be observed today, as the direction of water flow in these systems is highly variable, with adjacent drainages sometimes flowing in opposite directions. This is evident in the case of Lake Mason, Lake Noondie and Lake Barlee in Western Australia, which lie north to south along the same longitude and span less than 200 km. In each of these lakes, the water flow generally moves eastward (Mernagh et al., 2016). However, in neighbouring lakes, Lake Annean, Lake Austin and Lake Moore, which are situated in a similar north–south distribution ∼100–150 km to the west, water flow generally moves westward (Mernagh et al., 2016). While the habitat differences provide variable niches for samphires to occupy, species also exhibit physiological and ecological differences in their ability to tolerate salinity, drought and waterlogging (Short and Colmer, 1999; English, 2004; Colmer et al., 2009), which can also influence the localized distribution of species. Species zonation can be particularly evident in the fringing vegetation around inland salt lakes, with some species, such as T. crotalus and T. dactylifera, confined to the very narrow margins of the high-water line (Shepherd and Žerdoner Čalasan, 2024), while others, such as T. indica, always occur further away from the lake bed in the surrounding low-lying sand dunes where there is less waterlogging during wet periods. However, not all species can tolerate extremes of salinity; T. verrucosa (Fig. 1C), T. arborea and T. chartacea show a preference for growing in less saline clay pan habitats, often forming a monoculture across the bed of the clay pan. In contrast, T. disarticulata can often be found some distance from the margins of salt lakes growing on stony hillsides or over limestone, for example, along the southern margin of the Nullarbor Plain (Wilson, 1980).

Due to highly specialized physiology, halophytes are, on average, poor competitors in non-saline environments due to the reduced advantage of costly adaptations (Rozema and Schat, 2013; Flowers and Colmer, 2015). For example, Salicornia has been shown to be rapidly outcompeted for light by adjacent vegetation in less-saline habitats (Ellison, 1987; Callaway and Zedler, 1997). This appears to be the case in Australian vegetation communities adjacent to areas dominated by Tecticornia vegetation as well. Tecticornia species often build extensive samphire flats that comprise several species that coexist with other salt-tolerant species, such as low-growing procumbent or mat-like halophytes of the genera Gunniopsis and Frankenia or scattered mangroves. However, further from the saline water-line, less salt-tolerant Tecticornia are rapidly replaced with stands of low heath to open shrublands of Atriplex, Eremophila, Maireana and Melaleuca (Lyons et al., 2004; Barrett, 2006). This prostrate and/or mat-forming growth form arose multiple times in Tecticornia (e.g. T. crotalus, T. indefessa, T. sparagosa and T. uniflora; Fig. 1H), which points towards an ecological parallelism, similar to that previously observed in Salicornia (Steffen et al., 2015). This growth form is often found in other non-related lineages (see above) and has been interpreted as advantageous in frequently flooded saline environments (Youngman and Heckathorn, 1992; Voronkova et al., 2008; Steffen et al., 2015; Itoh, 2021). However, some shrub-like species, such as T. pergranulata and T. medusa, have been observed growing close to the water-line in large salt lake systems, and have been observed surviving prolonged periods of inundation (English, 2004; Konnerup et al., 2015). This suggests the mat-like growth form, at least in samphires, may not solely be an adaptation to flooding or frost (e.g. in Salicornia pacifica), but rather a side effect of selective pressures such as substrate stability.

By increasing genetic variation, polyploidization is often considered a key driver of diversification, as it may facilitate adaptation to new ecological niches, promote reproductive isolation, and enhance morphological or physiological variability (Soltis and Soltis, 2009). However, while polyploidy is present in both Salicornia and Tecticornia (Shepherd and Yan, 2003; Kadereit et al., 2007 , 2012b), it appears to be restricted to individual polyploidization events at the infraspecific level, as there are no observed whole-genome duplication events present at deeper nodes (Supplementary Data Fig. S3). Although direct evidence is lacking, suggestions of apomixis in Tecticornia, i.e. observations of seed-producing triploid and hexaploid plants of T. pterygosperma subsp. pterygosperma (Wilson, 1980; Shepherd and Yan, 2003) and high levels of inbreeding in the closely related Salicornia (Kadereit et al., 2007), may suggest autopolyploid rather than allopolyploid origins for polyploid taxa in the group. This is further supported by the low number of paralogues retrieved in our study.

Evolution of flower sexuality

Most Salicornieae exhibit a triad of bisexual flowers per bract (Piirainen et al., 2017). However, there are some notable exceptions to this rule. Arthrocaulon and Microcnemum occasionally produce lateral flowers that bear only a single stamen without a pistil (Piirainen et al., 2017). Individual populations of Salicornia quinqueflora from New Zealand are gynodioecious, where stigma exertion from female flowers coincides with anther exertion from hermaphrodite flowers, thus functionally preventing selfing (Connor, 1984). Further gynodioecious species reported from the literature include five other Salicornia species, whose identity, however, has to be investigated further (Soriano, 1947; Tölken, 1967; Connor, 1984). Among the Australian samphires, floral sexuality was previously thought to be a key diagnostic character, as the former genera Pachycornia (Bentham and Hooker, 1880) and Sclerostegia (Wilson, 1980) were characterized as distinct in part due to being andromonoecious, where each flower triad comprised a bisexual central flower flanked by two male flowers. In our phylogeny, the former Sclerostegia species T. arbuscula, T. disarticulata, T. medullosa, T. moniliformis and T. tenuis, along with T. triandra (syn. Pachycornia triandra), form a well-supported clade (highlighted portion of Fig. 3A, denoted by the abbreviation AM) that includes the bisexual T. annelida, and is also placed sister to the bisexual species T. entrichoma and T. sparagosa. Detailed anatomical studies postulated that the presence of andromonoecy in this clade is not homologous (Shepherd and Wilson, 2007) as the lateral male flowers of T. disarticulata and the morphologically similar T. papillata (previously known as Sclerostegia sp. Blue Hill (D.J. Edinger Nats 61) and not sequenced in this study), form protuberances of thickened material that are fused to each side of the ovary wall of the mature central fruit, likely representing aborted gynoecial material that is not present in other species of the former genus Sclerostegia (Shepherd et al., 2005a, b). However, analyses show that while T. disarticulata forms a well-supported clade with the bisexual T. annelida, it is nested within the andromonoecious clade, suggesting this is a homologous state change across this group, with T. annelida reverting to the plesiomorphic bisexual state.

While the majority of Australian samphires have axillary three-flowered cymes, flower numbers vary in some species scattered across the phylogenomic tree. The androdioecious T. pluriflora produces up to seven flowers per cymule, with the outermost lateral flowers being smaller in size, and sometimes with all flowers functionally male (as they still produce small pistillate stigmas). In contrast, other species, such as T. bibenda and the recently described species T. dactylifera, have seven and five bisexual flowers per cyme, respectively, indicating that space constraints may not always impact floral sexuality in the lateral flowers of each cyme (Fig. 1B; Shepherd and Žerdoner Čalasan, 2024). Alternatively, flower numbers may also be reduced. Tecticornia willisii, for example, has single bisexual flowers. Dioecious T. uniflora (syn. Tegicornia uniflora) plants have individual male or female flowers, although both still bear vestiges of the male (i.e. staminode) and female (i.e. pistillode stigma) organs (Fig. 1H; Wilson, 1980). Furthermore, floral numbers may not be fixed even within species, as glasshouse experiments show that T. uniflora can produce flower triads under certain environmental conditions (Shepherd and Wilson, 2007) and populations of the mat-like species T. indefessa may have between one and three flowers per cyme (Shepherd, 2007a), further supporting the view that floral characters are not always fixed in Tecticornia. Despite T. uniflora being the only dioecious Tecticornia species, dioecy and/or polygamodioecy are common sexual systems in other Amaranthaceae (e.g. several genera of Chenopodioideae; Kadereit et al., 2010; Dufay et al., 2014).

Evolution of C₄ photosynthesis

C₄ photosynthesis evolved only once in Tecticornia, and thus far only two species have the Kranz-tecticornoid anatomy that facilitates this type of carbon concentration mechanism: T. bibenda and T. indica (which includes T. indica subsp. bidens, T. indica subsp. indica, T. indica subsp. julacea and T. indica subsp. leiostachya) (Carolin et al., 1982; Kadereit et al., 2003; Shepherd and Van Leeuwen, 2007; Dakin, 2021) (highlighted portion of Fig. 3A, denoted by the abbreviation C₄). The C₄ carbon concentration mechanism is a complex trait and it evolved as a result of an interaction of different environmental conditions under low CO₂ concentration, such as increased temperatures and decreased nutrient availability (Ghannoum et al., 2008; Sage et al., 2018; Losemann, 2024). The evolution of C₄ often leads to either an expansion or specialization of an ecological niche (Lundgren et al., 2015; Aagesen et al., 2016; Young et al., 2022); however, neither seems to apply in Tecticornia as both species frequently co-occur with other C₃ samphires. Tecticornia bibenda (Fig. 1G) is restricted to the flood zones and fringing heath of gypseous playas and salt lake systems in the Little Sandy Desert of Western Australia, sharing its habitat with other xerophytic C₃ shrubs and forbs and C₄ hummock grasses, while C₄  T. indica subsp. indica and T. indica subsp. julacea occur only on the mudflats of the subtropical northern coastlines of the continent, alongside C₃ mangroves and C₄ mangrove-associated graminoids. Meanwhile, the other two subspecies, T. indica subsp. bidens and T. indica subsp. leiostachya, are distributed across the continent. While it remains unclear what was the driving force behind the evolution of the C₄ syndrome in Tecticornia and how this newly acquired physiological feature manifested, it seems that the newly acquired C₄ carbon concentration mechanism posed no distinct disadvantage from the energetic point of view when colonizing new habitats, as long as these are not in shade (Sonawane et al., 2018). Despite the overall well-resolved phylogeny, it remains unclear which species are the closest relatives to the C₄ clade.

The former Tecticornia s.s. (sensu Wilson, 1972, comprising T. arborea, T. australasica and T. verrucosa; highlighted portion of Fig. 3A), is one of the potential C₃ sister clades to the C₄ clade (Dakin, 2021). These species are annual or short-lived perennials with free rather than the more typically fused opposite bracts that subtend each paired triad of bisexual flowers. In contrast to the majority of other samphires, T. arborea and T. verrucosa inhabit periodically inundated non-saline or only moderately saline claypans (Wilson, 1972). Another potential C₃ sister taxon to the C₄ clade is T. peltata, which also has free opposite bracts, a feature that supports this species’ close relationship to Tecticornia s.s., and flower triads that are fused to each other and to the upper bract (Wilson, 1980). A recent anatomical study found that despite being currently recognized as C₃ species, T. australasica and T. verrucosa (but not closely related T. peltata), possess the following unusual anatomical characters: a high number of vascular bundles positioned adjacent to the chlorenchyma, a clear anatomical distinction between chlorenchyma and water storage tissue, and clear differentiation of chlorenchyma cells and high vein density, all features indicating a potential C₃–C₄ intermediate nature (Dakin, 2021, Fig. 3). This clade’s high levels of discordance (Supplementary Data Figs S1 and S4) might point towards its more complicated evolutionary history, potentially involving hybridization and incomplete lineage sorting, which may have resulted in their somewhat peculiar ecology and physiology. Interestingly, with the exception of clear differentiation of chlorenchyma cells, similar morphological tendencies towards C₄ photosynthesis were observed in both Group II species, such as T. auriculata and the closely related T. pruinosa and T. tenuis, as well as distantly related species in Group I such as T. lepidosperma and T. medusa (Dakin, 2021 ; Fig. 3). Due to its young age (Kadereit, 2005), we postulate that the C₄ syndrome has only recently been acquired in Tecticornia, which thus might be a good model group for investigating the initial steps in C₄ photosynthesis evolution (Kadereit et al., 2017; Morales-Briones and Kadereit, 2023; Tefarikis et al., 2022).

Species complexes

It is evident that while many currently recognized species or potentially new phrase-named taxa (e.g. T. sp. Christmas Creek (K.A. Shepherd & T. Colmer et al., KS 1063) or T. sp. Yoothapina Station (A.A. Mitchell 883)) are well supported in all the conducted analyses (Fig. 3, Supplementary Data Figs S1, S2 and S4), several broad complexes with unclear affinities are apparent across both Groups I and II. Seed coat morphology is a critical diagnostic feature for many taxa in Group I. Seed size, shape and degree of bumps and projections are reasonably consistent across the T. pergranulata species aggregate, which consists of four subspecies, as well as T. fontinalis, T. cupuliformis, T. doliiformis and informal phrase-named taxa T. sp. Corrow (P.G. Wilson 12750) and T. sp. Coonewarre (A.C. Beauglehole 70251). With the exception of T. fontinalis, the taxa in this group all bear similar dark reddish brown to black seeds covered with concentric porcate ribs, which supports their close relationship (Wilson, 1984). While the seeds of T. fontinalis are similar in colour and there are porcate ribs present over the top edge of each seed, the sides appear smooth. This species also possesses oblong-shaped flowers with a prominent external perianth abaxial lobe, thus placing it morphologically within the T. pergranulata species aggregate (Fig. 3A).

Tecticornia pergranulata subsp. pergranulata occurs along the coastal swamps and inland salt lake margins of south-western Australia and southern Australia, exhibiting a typical disjunct distribution, a phenomenon observed in many other plant species as well (Crisp and Cook, 2007; Ladiges et al., 2012). This taxon is morphologically variable, and regional variation is observed along its distribution, with the external adaxial lobe becoming smaller in plants found in the eastern parts of the continent (Wilson, 1980). In this particular case, denser sampling of this relatively widely distributed species is needed to infer whether this variation can be explained due to genetic divergence through breeding isolation of disparate populations, with ecotypes appearing under the different environmental conditions in which this species grows or if more than one valid taxon is present. Tecticornia pergranulata subsp. divaricata is less well understood and was not supported as monophyletic in our analyses. In contrast, T. pergranulata subsp. queenslandica, a decumbent plant with narrow inflorescences confined to the coastal mudflats in the north-eastern part of the continent, and T. pergranulata subsp. elongata, a widespread coastal and inland subspecies with distinctly long and often glaucous inflorescences, are both well characterized (Wilson, 1980). High levels of topological concordance and gCF and sCF values indicate that the latter subspecies is also the only genetically well-isolated taxon within T. pergranulata (Supplementary Data Figs S1 and S4). The close phylogenetic position of T. pergranulata species aggregate and T. pluriflora is somewhat expected based on the morphology. However, while the seed and flower characteristics of this species are highly similar to T. fontinalis from the T. pergranulata species aggregate, its vegetative characters are much closer to T. halocnemoides (Wilson, 1980). High topological concordance and relatively high gCF and sCF values point towards a limited genetic exchange with either of these two taxa and further studies are needed to speculate on the biological reasons behind this morphological pattern. Given that several genetically and morphologically well-supported species are recognized across this clade, further investigations and the clarification of the morphological features that consistently distinguish T. pergranulata s.s. from allied taxa may well lead its current subspecies to be eventually recognized as distinct species.

The most taxonomically problematic and biologically poorly understood group is the T. halocnemoides species aggregate (Figs 2 and 3B). Its seeds are variable in size, shape and degree of bumps and projections. Furthermore, this group is characterized by having the smallest vegetative articles and flowers within the genus. This not only makes them extremely difficult to identify accurately but results in the group becoming a catch-all for several soft-fruited entities with diminutive morphology. Wilson (1980) acknowledged that the classification he proposed for this complex (treated at that time as a broadly circumscribed species with five subspecies) would ‘prove unsatisfactory … nor has it seemed possible to provide critical characters for their differentiation’. He did note that seeds provide the most useful characters; however, a large proportion of specimens available for examination did not have mature seeds.

Wilson’s (1980) taxonomic concepts are not easily applied, evidenced by the fact that 51 % of specimens lodged in Australian herbaria remain unallocated to subspecies (644 of 1270 specimens; AVH, 2025). Furthermore, several newly recognized and morphologically well-characterized species, such as T. medusa (Shepherd and van Leeuwen, 2011) and T. willisii (Shepherd, 2018), potentially new taxa (e.g. T. sp. Chinocup (K.A. Shepherd KS 1191), T. sp. Dennys Crossing (K.A. Shepherd & J. English KS 552), T. sp. Lake Wallambin (K.A. Shepherd KS 1157) and T. sp. Lake Way (P. Armstrong 05/961)), or morphologically variable forms (‘aff.’) are present in the T. halocnemoides clade, the latter represented by several accessions that do not come together within a single subclade, not even those collected within the region of the type locality along the Swan River near Fremantle in the city of Perth. Extremely high levels of topological discordance coupled with low gCF and sCF values that persist along the backbone of this species complex (Supplementary Data Figs S1 and S4) further illustrate that, with the exception of some well-defined species (morphologically as well as genetically), most of the taxa in this complex probably still undergo either extensive gene flow or more possibly incomplete lineage sorting. Difficulties in confirming the diagnostic characters for these taxa are compounded by concerns about observed morphological variation being due to plants growing under highly selective but somewhat different environmental conditions, the presence of possible hybrids, or variation being due to the presence of polyploid races (e.g. impacting the size of seeds).

A third complex in the genus includes several morphologically variable taxa allied to T. undulata and T. pruinosa across the most derived subclade in Group II (Fig. 3A). Similarly to the T. halocnemoides complex, this species complex shows extensive topological discordance with low gCF and sCF values, which may indicate extensive incomplete lineage sorting (Supplementary Data Figs S1 and S4). Species within this broad group have dorsoventrally flattened flowers (that often have two rather than three perianth lobes), which are more or less completely covered by their subtending bracts. The whole infructescence remains intact and becomes firm or crustaceous at maturity, except in T. chartacea, where the soft fruits disintegrate into individual plate-like rings to release the seed (Shepherd, 2007b).

In his diagnosis, Wilson noted that at least two different forms fall under T. undulata (Wilson, 1980, page 59), one with acuminate articles and bracts and one that had ‘black carbonate pericarps’ and seeds with a papillate testa, listing several specimens as examples. Some of these specimens were subsequently found to represent a distinct species, later named T. laevigata, while others remain identified as ‘T. aff. undulata’. Wilson postulated that differences observed in the field were possibly due to differences in growth under different seasonal conditions rather than reflecting genetic diversity; however, this study does confirm there is genetic diversity across the aggregate.

This aggregate also encompasses T. pruinosa (with several morphologically divergent variants labelled ‘aff. pruinosa’), and includes representatives with a disjunct distribution across coastal and inland regions of Western and South Australia. Wilson also noted in his treatment that T. pruinosa was morphologically variable and some collections seem to be functionally exclusively female or male (Wilson, 1980, page 55), which should theoretically exclude inbreeding. Tecticornia undulata and T. pruinosa are both widespread species with distinct fruiting spikes, which act as a ‘spotting feature’ for these species; however, it is evident in both cases that there are more undescribed taxa being included under these names, which are genetically and morphologically distinct.

Phylogenomic analyses also show evidence for potential recognition of several species allied to T. calyptrata, known for its unusual thick cap-shaped woody pericarp, and formal recognition of the phrase-named species T. sp. Yoothapina Station (A.A. Mitchell 883), a taxon with distinctly divaricate branches and narrow inflorescences that occur in lateral pairs that are positioned almost at right angles to the central axis. Wilson (1980) noted several accessions in his treatment that he thought may represent hybrids due to exhibiting intermediate morphologies. Two accessions currently thought to be possible hybrids within Group II designated as T. bulbosa × pruinosa (PERTH 08010870 and PERTH 0863088) are genetically distinct from each other, with the former placed sister to T. bulbosa s.s. while the latter was more closely allied to T. syncarpa. Their hybrid origin is not supported by our results, as the majority of internal nodes within this complex receive maximal statistical support (Fig. 3A). As we have an extensive taxon sample and no way of knowing exactly which could be the two potential parental species to test, we refrain from doing any further computationally demanding hybridization analyses here.

Many of these complexes include variants that are widely distributed across the Australian continent, and considerable fieldwork in combination with dense molecular sequencing will be required to begin to untangle many of these seemingly intractable taxonomic problems. However, some taxa are also morphologically well defined and have highly restricted distributions, which consequently may represent taxa of conservation concern. Taxonomic resolution of such taxa should be prioritized to facilitate their appropriate conservation management.

Species concepts in samphires

Species concepts are difficult to apply in samphires. Convergent evolution, polyploidy and ecological plasticity likely exacerbate identification issues already caused by the limited number of easily observable morphological characters on which species were originally described, resulting in the group being dubbed a taxonomic nightmare (Kadereit et al., 2007). Nevertheless, there are numerous cases of morphologically well-circumscribed and narrowly endemic species dispersed throughout the phylogenomic tree. Examples include, but are not restricted to, the easily recognizable, hard-fruited C₃  T. bulbosa (Fig. 1B) and C₄  T. bibenda (Fig. 1G), both characterized by very large vegetative articles, as well as the more dainty, mat-like T. crotalus and T. indefessa. There are also morphologically uniform, easily recognizable and ecologically specialized but widely distributed species such as T. australasica, which is restricted to the wetter northern and north-eastern coasts of Australia and New Guinea, and the elusive T. verrucosa, a clay-loving, non-halophytic species with a peculiar vegetative architecture (Fig. 1C), whose short-lived populations tend to disappear after one season and reappear years later. There are also cases of morphologically and genetically closely related species with markedly different distributions. For example, T. enodis, T. globulifera, T. sp. Lake Chinocup, T. sp. Dennys Crossing, T. sp. Lake Wallambin, and T. willisii are all morphologically and genetically distinct, yet geographically isolated, while still embedded within the T. halocnemoides species aggregate.

While such cases may suggest extreme ecotype selection, well-studied lake systems such as Lake Way, Lake Sunshine, Lake Ten Mile and Hannan Lake, show numerous genetically and morphologically well-differentiated taxa occurring sympatrically or in close proximity (English, 2004; Datson and Coleman, 2018; Pick, 2019). In addition, cases of allopatric speciation can be observed, such as in T. flabelliformis and T. pterygosperma, or between T. pluriflora and T. dactylifera, where these sister species exhibit a clear east–west geographic separation. Nevertheless, ecological parallelisms are common and often render current species concepts obsolete. For example, T. arbuscula and T. halocnemoides subsp. longispicata, once considered to be taxa with a clear east–west disjunct distribution, have been shown to represent polyphyletic groups in need of taxonomic reconsideration. These geographic groupings are particularly evident in the more densely sampled T. halocnemoides species aggregate, which includes at least two genetically well-circumscribed clades occurring in Perth and its vicinity, and at least two coastal clades from north-western Australia.

Overall, Australian samphires represent an assemblage of diverse taxa, ranging from well-characterized taxa with wide to narrow ranges to genetically and morphologically highly problematic taxon swarms, in urgent need of clarification. Thus, Australian samphires represent a highly complex yet exciting group for investigating ecological influences on potentially recent or ongoing speciation processes.

Conclusions and research outlook

It is evident that the Salibaits customized bait kit has proven to be a powerful tool that has helped resolve the systematics and taxonomy of this recalcitrant genus, which is known for its paucity of obvious morphological characters that are further confounded by ecological plasticity. Here, we present markedly improved phylogeny of Tecticornia, which supports the genus as monophyletic and highlights two well-supported sublineages designated as Groups I and II. The short branches and pronounced gene tree discordance that persist throughout most of the phylogenetic tree illustrate that the evolutionary history of this genus has been influenced by extensive incomplete lineage sorting, probably due to rapid radiation as the Australian continent salinized. This is further supported by the highly restricted distributions of many species and the small temporal scale (Piirainen et al., 2017; Cousins-Westerberg et al., 2023). It remains unclear, however, why so many species exhibit such restricted distributions despite being presumably anemophilous and both anemo- and hydrochoric, and occurring in such open habitats. An important factor that should be taken into consideration is the evolutionary history of salt lakes, which Tecticornia is closely associated with, and ploidy variation within and between populations, as these may be significant drivers of diversification across the genus. Both aspects remain understudied and should be addressed in greater detail in the future. Our study further highlights the need to re-circumscribe certain species, such as T. halocnemoides and T. pruinosa, which clearly represent several, albeit morphologically highly similar, species. Moreover, we illustrate the need for unified taxonomic concepts for taxa distributed across Australian state boundaries to facilitate consistent identification and appropriate conservation management, in the face of escalating environmental and anthropogenic threats. Finally, this study lays the groundwork for genome-level research that will provide valuable insights into the evolution of C₄ (and potential CAM) photosynthesis and salinity tolerance in Salicornieae, traits of considerable agricultural relevance.

SUPPLEMENTARY DATA

Supplementary data are available at Annals of Botany online and consist of the following. Table S1: voucher list of the taxa analysed in this study. Table S2: list of the transcriptomes used for the customized Salibaits with corresponding references. Figure S1: concordance/discordance topology analysis of Tecticornia inferred using PhyParts, plotted on the coalescent-based species tree from an ASTRAL-III analysis of 463 nuclear loci (with stitched contigs, allowing up to 75 % missing data). Figure S2: coalescence-based phylogeny of Tecticornia inferred using ASTRAL-III from 463 nuclear loci (with stitched contigs, allowing up to 75 % missing data). Figure S3: orthogroup mapping results showing potential gene duplications based on the subclade orthogroup tree topology method, plotted on the coalescent-based species tree from an ASTRAL-III analysis of 463 nuclear loci (with stitched contigs, allowing up to 75 % missing data). Figure S4: gene (left) and site (right) concordance/discordance values using IQ-TREE 2, plotted on the coalescent-based species tree from an ASTRAL-III analysis of 463 nuclear loci (with stitched contigs, allowing up to 75 % missing data). Figure S5: gene (middle) and site (right) concordance/discordance values using IQ-TREE 2, plotted on the coalescent-based species tree from 353 loci (Angiosperms353 bait set, left) and 485 loci (Salibaits bait set, right) with local posterior probability values on the left. Figure S6: concordance/discordance topology analysis of Tecticornia inferred using PhyParts, plotted on the coalescent-based species tree from 353 loci (Angiosperms353 bait set, left) and 485 loci (Salibaits bait set, right).

FUNDING

This work was supported by the German Research Foundation grant to G.K. (1816/11-2) within the priority programme 1991 ‘Taxon-Omics: New approaches for discovering and naming biodiversity’. We would also like to acknowledge the contribution of the Genomics for Australian Plants Framework Initiative consortium (https://www.genomicsforaustralianplants.com/consortium/) in the generation of data used in this publication. The initiative is supported by funding from Bioplatforms Australia (enabled by NCRIS), the Ian Potter Foundation, Royal Botanic Gardens Foundation (Victoria), Royal Botanic Gardens Victoria, the Royal Botanic Gardens and Domain Trust, the Council of Heads of Australasian Herbaria, CSIRO, Centre for Australian National Biodiversity Research and the Department of Biodiversity, Conservation and Attractions (Western Australia).

AUTHOR CONTRIBUTIONS

A.Z.C. and K.A.S. carried out the fieldwork and wrote the original draft of the manuscript, with A.Z.C. carrying out all analyses. D.F.M.B. designed the baits and K.A.S. and G.K. provided the funding. All authors discussed the results, revised the draft and contributed to the final version of the manuscript.

DATA AVAILABILITY

Part of the raw data can be accessed through the Bioplatforms Australia data portal under the bioplatforms sample id 376800-376895. The remaining target enrichment data generated for this study can be found in the NCBI BioProject PRJNA1224975.