Integrated taxonomy reveals new threatened freshwater mussels (Bivalvia: Hyriidae: Westralunio) from southwestern Australia

The freshwater mussel Westralunio carteri (Iredale, 1934) has long been considered the sole Westralunio species in Australia, limited to the Southwest and listed as vulnerable on the IUCN Red List and under Australian legislation. Here, we used species delimitation models based on COI mtDNA sequences to confirm existence of three evolutionarily significant units (ESUs) within this taxon and conducted morphometric analyses to investigate whether shell shape differed significantly among these ESUs. “W. carteri” I was found to be significantly larger and more elongated than “W. carteri” II and “W. carteri” II + III combined, but not different from “W. carteri” III alone. We recognise and redescribe “W. carteri” I as Westralunio carteri (Iredale, 1934) from western coastal drainages and describe “W. carteri” II and “W. carteri” III as Westralunio inbisi sp. nov. from southern and lower southwestern drainages. Two subspecies are further delineated: “W. carteri” II is formally described as Westralunio inbisi inbisi subsp. nov. from southern coastal drainages, and “W. carteri” III as Westralunio inbisi meridiemus subsp. nov. from the southwestern corner. Because this study profoundly compresses the range of Westralunio carteri northward and introduces additional southern and southwestern taxa with restricted distributions, new threatened species nominations are necessary.

. Phylogenetic trees obtained by maximum likelihood (left) and Bayesian inference (right) analysis of "Westralunio carteri" mtDNA COI sequences, including support values for the major genetic clades [ultrafast bootstrap values (left) and Bayesian posterior probabilities (right)]. Colour coded bars show support for the three major clades by the species delimitation methods (ASAP = dark shade; TCS = lighter shade). Green = WcI = "W. carteri" I; blue = WcIII = "W. carteri" III; red = WcII = "W. carteri" II. Results of bPTP analysis not shown (see supplementary data). Haplotype names correspond to Benson et al. 28 . Outgroup taxa are Velesunio ambiguus (Philippi, 1847) (Hyriidae: Velesunioninae) and Cucumerunio novaehollandiae (Gray, 1834) (Hyriidae: Hyriinae: Hyridellini). Table 1. Molecular diagnoses of "Westralunio carteri" Evolutionarily Significant Units (ESUs) from southwestern Australia (after Bolotov et al. 122 with reanalysis of data from Klunzinger et al. 27 and Benson et al. 28 ). New taxa include "W. carteri" II and III. Supplementary files 1 and 2 contain the alignments used to determine the single pure characters and p-distances. The position of each diagnostic character refers to its location within those alignments. SE, standard error. "W. carteri" I 60/40 5.02 (0.90) "W. carteri" II 2.95 (0.01) "W. carteri" II and beak length (BL)) and more elongated (i.e., had a lower maximum height index (MHI)) than shells of "W. carteri" II and "W. carteri" II + III combined (Table 2). However, there was no difference in size or shape metrics between "W. carteri" I and "W. carteri" III ( Table 2). The lack of significant differences in beak height index (BHI) and beak length index (BLI) among any of the taxa (Table 2) indicates that wing and anterior shell development was not discernibly different between any of the ESUs. This pattern was partly confirmed in the principal component analysis (PCA) of these three shell shape indices, where PC1, largely explained by variation in BLI ( Fig. 2A), did not differ between the two species (i.e., "W. carteri" I vs. "W. carteri" II + III) or among the three taxa ( Table 2). The PC2, largely explained by variation in MHI and BHI ( Fig. 2A), differed significantly between "W. carteri" I and "W. carteri" II (Table 2). Accordingly, 70% (70% jack-knifed) of specimens were assigned to the correct species in the corresponding discriminant analysis (DA), whilst this was true for only 55% (54%) at the MOTU-level.
The difference in shell elongation between "W. carteri" I and "W. carteri" II was confirmed by Fourier shape analysis. As visualised by synthetic outlines in Fig. 2B, shell elongation is expressed along the PC1 (explaining 15% of total variation in Fourier coefficients). The PC1 as well as PC2 scores differed significantly between the two species (i.e., "W. carteri" I vs. "W. carteri" II + III) as well as between "W. carteri" I and "W. carteri" II, respectively ( Table 2). Combined with synthetic outlines, this indicated a tendency towards a more elongated, somewhat wedge-shaped shell in "W. carteri" I, whilst "W. carteri" II shells tended to be relatively high with a stout anterior margin (Fig. 2B). An analysis of similarities (ANOSIM) analysis on all Fourier coefficients revealed no significant difference between the two species (i.e., "W. carteri" I vs. "W. carteri" II + III; ANOSIM: R = − 0.018, p = 0.097), but did indicate a significant difference between the three ESUs (ANOSIM: R = 0.0625, p = 0.0051). Specifically, "W. carteri" I differed significantly from "W. carteri" II (Bonferroni-corrected p = 0.0009). Only Table 2. Shell size metrics [mm], shape indices [%] and scores for the first two principal components (PC) obtained by Principal Component Analysis of shape indices and 18 Fourier coefficients generated by Fourier Shape Analysis for each "Westralunio carteri" species and subspecies-level Evolutionarily Significant Units (ESUs): n, number of specimens measured; minimum (min) to maximum (max) and mean (± standard error (SE)). P-value (P), degrees of freedom (df) and F-ratio (F) were obtained from ANOVAs comparing "W. carteri" I, II and III (upper section of the Table) and "W. carteri" I and II + III (lower section of the Table), respectively. P-values < 0.05 in bold; different superscript letters indicate significant differences between groups in Tukey's pairwise posthoc-test. Shell size metrics: TL total length, distance from anterior to posterior apices of valve; MH maximum height, distance from ventral edge to apex of posterior ridge; BH beak height, distance from ventral edge to beak apex; BL beak length, distance from anterior apex of valve to the 90-degree vertex aligning with the beak apex. Shape indices determined as: MHI = MH/TL, BHI = BH/MH, and BLI = BL/TL.    (Iredale, 1934) (Table 2). Shell outline oblong-ovate to rounded; posterior end obliquely to squarely truncate, anterior end round; ventral edge slightly curved, nearly straight in larger specimens; hinge line curved, hinge strong. Umbos usually abraded in specimens > 20 mm in length; unabraded umbos with distinctive v-or w-shaped plicated sculpturing ( Fig. 3B and Zieritz et al. 46 ). Shell substance typically thick; shells of medium width with pronounced posterior ridge; periostracum smooth, dark brown to reddish, with fine growth lines. Pallial line less developed in smaller specimens and prominent only in large specimens (e.g., > 60 mm TL). Lateral teeth longer and blade-like, slightly serrated to smooth and singular in left valve, fitting into deep groove in right valve; pseudocardinal tooth in right valve coarsely serrated, thick, and erect, fitting into deeply grooved socket in left valve. Anterior muscle scars well impressed and anchored deeply in larger specimens; anterior retractor pedis and protractor pedis scars both small and fused with adductor muscle scar; posterior muscle scars lightly impressed; dorsal muscle scars usually with two or three deep pits anchored to internal umbo region.
Anatomy: Supra-anal opening absent, siphons of moderate size, not prominent but protrude beyond shell margin in actively filtering live specimens, pigmented dark brown with mottled lighter brown to orange splotches; ctenidial diaphragm relatively long and perforated. Outer lamellae of outer ctenidia completely fused to mantle, inner lamellae of inner ctenidia fused to visceral mass then united to form diaphragm; palps relatively small, usually semilunar in shape; marsupium well developed as a distinctive swollen interlamellar space in the middle third of the inner ctenidium of females. Outer ctenidia in both sexes thin, with numerous, short intrafilamentary junctions and few, irregular interlamellar junctions; in females similar, but marsupium has numerous, tightly packed, well-developed interlamellar junctions. Thus, brooding in females is endobranchous. Life history: Sexes are separate in W. carteri, and hermaphroditism appears to be rare [47][48][49] . Males and females both produce gametes year-round but brooding of glochidia appears to be seasonal and 'tachyticitc' (i.e., as defined by Bauer & Wächtler 19 , fertilisation and embryonic development occurring in late winter/early spring and glochidia release in early summer) 50 . Glochidia are released within vitelline membranes, embedded in mucus which extrude from exhalant siphons of females (i.e., 'amorphous mucus conglutinates') during spring/ summer. Glochidia attach to host fishes and live parasitically on fins, gills or body surfaces for 3-4 weeks while undergoing metamorphosis to the juvenile stage. Host fishes which have been shown to support glochidia metamorphosis to the juvenile stage in the laboratory include Afurcagobius suppositus ( 47 . Juveniles which have detached from host fishes have a characteristic ciliated foot and two distinct adductor muscles 47 . Probable age at maturity is 4-6 years old and estimated longevity is at least 36 to 52 years 59 . Inheritance of mitochondria is doubly uniparental 60 . Glochidium: Following release, glochidia hatch from vitelline membranes but remain tethered by a larval thread and characteristically 'wink'; valves with single adductor muscle; shells subtriangular and scalene in shape with smooth surface which lack surface spikes and dotted with pores, 305-310 μm long, 249-253 μm high and have a hinge length of 210-214 μm; apex of the ventral edge protrudes and is off-centre and closest to the www.nature.com/scientificreports/ posterior region of the glochidial shell, giving a sub-triangular scalene shape; larval teeth slightly curved towards adductor muscle with concave protuberance on base of the right valve tooth and convex protuberance on base of the left valve tooth; larval tooth of the right valve lanceolate, terminating with three sharp cusps; tooth of left valve blunt with two rounded cusps and groove at the midpoint to accommodate the middle cusp of the right valve; larval teeth lack microstylets ( Fig. 3C and Klunzinger et al. 48 ). Distribution: Found in freshwater catchments from Gingin Brook, north of Perth to westerly flowing drainages north and west of the Blackwood River, within 150 km of the coast 28,61 (Fig. 6).
Habitat: Found in freshwater streams, rivers and sometimes lakes or wetlands with permanent water, salinities less than about 3.0 mg/L, pH ranging from about 4.5 to 10 and more common in habitats not prone to nutrient pollution 61 .
Comments: McMichael & Hiscock 10 suggested that the species aligns with other Velesunioninae in having smooth umbos, later refuted by Zieritz et al. 46 , as illustrated in Fig. 3B. Additionally, Iredale 9 separated the genus Westralunio from Velesunio based on adult hinge tooth morphology, such that pseudocardinal hinge teeth are erect, serrated and strongly grooved in Westralunio as opposed to Velesunio which are suggested as not serrated and not strongly grooved. We contend that while W. carteri typically does have serrated pseudocardinal teeth that are usually erect/conspicuous and strongly grooved, so too are some Velesunio specimens (M. Klunzinger, unpublished data). In terms of distribution, there is one record of a specimen from the Gascoyne River collected ca. 1891 (BMNH-MP-110 listed as Diplodon ambiguus Parreyss in Philippi 26  Material examined: For W. inbisi inbisi (= "W. carteri" II), molecular data examined included 82 and 93 individual 16S rDNA and COI mtDNA sequences, respectively, for species delimitation. Additionally, Fourier shell shape outline analysis and traditional shell morphometric measurements were examined from 127 and 139 individuals, respectively. Complete details on all specimens examined are provided in Supplementary Table S2.
Diagnosis: Specimens of W. inbisi inbisi are distinguished from other Australian Westralunio taxa by having shell series that are significantly smaller and less elongated than W. carteri, but not W. inbisi meridiemus. The subspecies has three diagnostic nucleotides at COI (75 A, 87 T, 318 T) and none at 16S, which differentiate it from its sister taxa, W. carteri and W. inbisi meridiemus using ASAP and TCS species delimitation models.
Description: This subspecies is of the ESU "W. carteri" II 27,28 . Shell morphology in juveniles and adults same as W. carteri as described above (see Fig. 7A-C). Total adult shell length generally < 80 mm but known to reach in excess of 90 mm 68,69 . MHI 55-74%; anterior portion of shell with moderate development, BLI 23-51%; larger shells with abraded umbos scarcely winged; wing development variable, generally decreasing with size, BHI 79-99% ( www.nature.com/scientificreports/ Distribution: Found in southerly to south-westerly flowing freshwater streams, rivers and sometimes lakes or wetlands with water salinities less than approximately 3.0 mg/L from Boodjidup Brook in the Capes region to the west of the Blackwood River catchment to Waychinicup River within 150 km of the coast, primarily along the South Coast of Western Australia [27][28][29]61,69 (Fig. 6).
Habitat: Similar to W. carteri although habitats can also include perched dune lakes. Waters that W. inbisi inbisi inhabit are often more tannin stained due to their occurrence in more heavily forested catchments with greater densities of native riparian vegetation than for W. carteri.
Material examined: For W. inbisi meridiemus (= "W. carteri" III), molecular data examined included 9 and 12 individual 16S rDNA and COI mtDNA sequences, respectively, for species delimitation. Additionally, Fourier shell shape outline analysis and traditional shell morphometric measurements were examined from 12 individuals each. Complete details on all specimens examined are provided in Supplementary Table S3.
Etymology: The subspecific epithet, meridiemus, is derived from Latin 'meridiem' , translating to 'southwest' in English in reference to the location of its type locality, which sits in the southwestern region of Western Australia.
Diagnosis: Specimens of W. inbisi meridiemus have five diagnostic nucleotides at COI (69 C, 123 C, 126 T, 483 A, 526 A) and none at 16S, which differentiate it from its sister taxa, W. carteri and W. inbisi inbisi using ASAP and TCS species delimitation models.
Description: This subspecies is of the ESU "W. carteri" III. Shell morphology in juveniles unknown, but adults same as W. inbisi inbisi and W. carteri as described above; total adult shell length generally < 80 mm. MHI 60-71%; anterior portion of shell with moderate development, BLI 28-35%; larger shells with abraded umbos scarcely winged; wing development variable, generally decreasing with size, BHI 84-93% (Table 2); anatomy same as W. carteri and W. inbisi inbisi, including siphon pigmentation and morphology (illustrated in Fig. 5B). Life history observations for this ESU cannot be derived from existing field observations as all known populations overlap the distribution of either W. carteri (in Margaret River) or W. inbisi inbisi (in the lower Blackwood Basin); however, mussels from those locations appear to have separate sexes based on macroscopic examinations of marsupia in both gravid and non-gravid females. Similarly, gravid female mussels have been observed from Margaret River during late spring to summer (November to December). Species of wild-caught fishes from Canebreak Pool in Margaret River that have been observed carrying glochidia include G. occidentalis and N. vittata 47 . Reproductive phenology, age and growth have not been elucidated in W. inbisi meridiemus.
Distribution: Found in the neighbouring catchments of Margaret River and the Blackwood River of Western Australia, where it is sympatric with W. carteri and W. inbisi inbisi, respectively 27,28 (Fig. 6).
Habitat: Similar to W. carteri and W. inbisi inbisi in either lotic or lentic freshwater rivers, streams, and pools with varying degrees of riparian vegetation.

Discussion
This study is the first to integrate molecular species delimitation and morphological analyses to describe new taxa of Australian freshwater mussels. In their review of the taxonomy, phylogeography and conservation of freshwater mussels in Australasia, Walker et al. 14 highlighted the need for such a taxonomic framework that uses both genetic and morphological data to gain a better understanding of species delimitation within this group. Overall, the study illustrates the value of using multiple data sources for species delimitation for cryptic taxa, provides an example when it is appropriate to recognise subspecies, and describes a case study of an IUCN listed species that would have to be re-assessed in terms of conservation status following the application of a robust taxonomic framework for recognising species boundaries.
In this study, we aimed to use multiple lines of evidence to investigate whether the three ESUs previously identified for the freshwater mussel Westralunio carteri 27,28 should be recognised as separate species. Using three species delimitation models run for 164 COI mtDNA sequences, a combination of both traditional indices of shell morphology and Fourier shell shape analyses and geographical distribution records, our results provided a clear case for the recognition of at least two separate species-Westralunio carteri (Iredale, 1934 9 ) which is found in rivers draining the western coast, and W. inbisi sp. nov. which occurs in rivers draining the southern and lower southwestern coast of southwestern Australia. The recognition of these two separate species was well supported, with congruent data sets confirming that they are both morphologically and genetically divergent.
Carstens et al. 70 have argued strongly that species should be delimited based on the congruence of multiple data sets that could include genetic, morphological and distributional data, as was done in this study. This is particularly true for cryptic taxa. The use of an integrated approach for resolving taxonomic uncertainties for freshwater mussels has growing support. For example, Johnson et al. 4 used multiple lines of evidence to show that current taxonomy overestimated species diversity within the imperilled freshwater mussel genus Cyclonaias in North America. Similarly, Morrison et al. 71 tested species boundaries in the North American Pleurobema species complex using genetic and morphological data, finding that the most likely scenario was that the two named species they investigated were members of a single, widespread species. Despite this growing trend of using multiple sources of evidence for species delimitation, several freshwater mussel taxa have been named based on shell morphology alone in recent decades. For example, the hyriids Lortiella opertanea Ponder  www.nature.com/scientificreports/ described freshwater mussels. The propensity of shell shape to vary with the environment within freshwater mussel species can render it an often-unreliable character on which to base species taxonomy 12,14,73,74 . Alternatively, relying on genetic data without morphological support can also be problematic; however new species have been raised using this method 2,3 . While this may not be appropriate for practical purposes 75,76 , there is no doubt that recognising genetic differences between populations is an important aspect of describing biodiversity and indeed, has been used in several species' concepts 77,78 .
Our study also provides an example where subspecies have been described in recognition of the existence of phenotypically similar, but genetically distinct evolutionary lineages within the W. carteri species complex. In our case, shell morphology for W. inbisi inbisi and W. inbisi meridiemus was mostly similar, yet these taxa possessed COI character attributes that were unique to each, largely geographically separated lineage, suggesting a degree of reproductive isolation and an evolutionarily significant process. We did have a low replicate number of shells of W. inbisi meridiemus to examine which may have accounted for the lack of statistical differences between this subspecies and W. inbisi inbisi if indeed differences might exist. While there is additional material available from Margaret and Blackwood Rivers in museum collections, we were restricted to examining only shells from which genetic information is available given Benson et al. 28 found both taxa in the two river basins. The definition of a subspecies can vary but is widely accepted as an aggregate of phenotypically similar populations of a species inhabiting a geographic subdivision of the range of that species and differing taxonomically from other populations of that species 79 . In their study of land snails, Páll-Gergely et al. 80 suggest restricting subspecies to cryptic species delimitation, for example when molecular data support lineage separation but where no clear morphological differences are currently known. We agree with this definition but caution that molecular data can exhibit differentiation related to population structure and not speciation. Fixed molecular differences are critical to confirming reciprocal monophyly when molecular phylogenetic methods are employed. The use of subspecies in the Hyriidae is relatively common. Of the 261 names available for species or subspecies of Hyriidae, 57 (21.8%) have been used either exclusively as subspecies or as either species or subspecies 9,11,18,20,23,43,64,[81][82][83][84][85][86][87] . Here we recognise subspecies based on congruence in genetic and geographical differentiation but chose not to raise "W. carteri" III to species level given there was no significant difference in shell morphology between it and "W. carteri" II and because not all molecular species delimitation models revealed a separate taxon for "W. carteri" III.
These results confirm that current taxonomy underestimates species diversity of freshwater mussels in southwestern Australia and that this has implications for the listing of W. carteri as a threatened species. Freshwater mussels are amongst the most threatened aquatic species worldwide, and many authors have expressed concern about the global decline of this group [88][89][90] . It is not surprising therefore that a recent review by Benson et al. 91 revealed that 44% (18 of 41 species) of described freshwater mussel species known to occur in Mediterraneanclimate regions have been listed globally as either Critically Endangered, Endangered or Vulnerable on the IUCN Red List. A further six species (15%) have been classified as Near Threatened. Freshwater mussels are also notorious in having shells which are morphologically plastic within the same taxa 12 and taxa which have morphologically similar shells but are molecularly different, leading to 'cryptic speciation' 92 . Given the increasing interest in using multiple data sets to confirm species boundaries in the group, amendments to the Red List conservation status of these listed species can be anticipated as species delimitation based originally on morphological characters is further clarified with this integrative taxonomy approach. In the case of W. carteri, the species is currently listed as 'Vulnerable' internationally (IUCN Red List), nationally under the Australian Environmental Protection and Biodiversity Conservation Act 1999 (EPBC Act) and at the state level under the Western Australia Biodiversity Conservation Act 2016. More recently, in a preliminary analysis based on past distribution data 61 , Klunzinger et al. 27 suggested that an estimated reduction of 72% in extent of occurrence (EOO) of the W. carteri lineage identified here might qualify this species as 'Endangered' under criterion A2c of the IUCN Red List. More robust analyses which include recent distribution records would be needed to confirm the conservation status of this species as well as the new species described in this study. Our evidence suggests that once formal re-diagnoses and descriptions of the Westralunio taxa are published, fresh nominations for listing as threatened would be required. This would entail submission of a nomination to the Threatened Species Scientific Committee to amend the conservation status of W. carteri, and the preparation of new nominations for one or both new subspecies if deemed necessary. As more studies use an integrative approach for delineating freshwater mussel species, implications for conservation are inevitable, and this is likely to take place on a global scale. For example, the suggestion that the freshwater mussel species Pleurobema clava (listed as Endangered under the Endangered Species Act in the USA) and P. oviforme (a species being considered for listing) are members of a single, widespread species will have management implications for these species 71 .
Several research gaps in our knowledge of Westralunio will benefit from future investigation. The position of W. albertisi and W. flyensis from West Papua and Papua New Guinea within the genus Westralunio is unclear 10,14 and employing molecular analyses will undoubtedly resolve this biogeographic and taxonomic conundrum. At a higher classification level, given that both species of juvenile Australian Westralunio have distinctive v-or w-shaped shell sculpturing on their umbos 46 , in combination with strongly grooved and serrated pseudocardinal hinge teeth in adult shells, calls into question their placement within the Velesunioninae defined by McMichael & Hiscock 10 . This is corroborated by phylogenetic data presented by Graf et al. 21 and Santos-Neto et al. 22 who showed Westralunio separate to the other Velesunioninae. However, until complete data are available for other Westralunio and other velesunionine species, we are reluctant to make any changes to the current arrangement of the subfamily. Also of value would be an examination of glochidia morphology and morphometry. Glochidia of W. carteri were described by Klunzinger et al. 48 , but the morphology and morphometry of glochidia from W. inbisi inbisi and W. inbisi meridiemus are entirely unknown and worthy of investigation. Indeed, glochidia morphology and morphometry has been shown to have taxonomic value for other species 19 . For example, Jones et al. 93 showed divergence in larval tooth arrangement and shell size in the south-east Australian Hyridella australis ( www.nature.com/scientificreports/ Pimpão et al. 95 were able to distinguish several species of Brazilian Hyriidae using glochidial characters and made taxonomic changes to some genera and subgenera on this basis. More recently, Melchior et al. 96 presented contrasting glochidia release strategies and glochidia size in between two sympatric species of Echyridella from New Zealand, further strengthening the taxonomic division between Echyridella menziesii (Gray, 1843 97 ) and Echyridella aucklandica (Gray, 1843 97 ) recognised by Marshall et al. 20 . This study has highlighted the utilisation of morphology and phylogeography to make sound decisions on drawing taxonomic boundaries between ESUs. Formalising the taxonomy of ESUs identified by Klunzinger et al. 27 and Benson et al. 28 for Westralunio taxa will be beneficial for conservation management of the species and subspecies identified in this study. We suggest that a similar approach of taxonomic division may be applied to other freshwater fauna of the region (and elsewhere), such as for the multiple lineages of pygmy perches or galaxiid fishes alluded to by Morgan et al. 98 and Buckley et al. 99 . Furthermore, the taxonomic approach illustrated in our study should be applied to other as yet undescribed 'cryptic species' of Australian freshwater mussels 92,100,101 as a way forward in resolving taxonomic uncertainty within the group. Gross anatomy. Siphon characters, which are not easily examined in preserved specimens (for example, due to tissue contraction and discolouration), were observed in live specimens of W. carteri in the Canning and Margaret Rivers with mask and snorkel and photographed in their natural state with a waterproof digital camera. Tissue anatomy of freshly dead and preserved specimens was examined employing well-established dissection methods 10 on specimens collected for data published by Klunzinger et al. 27 . For shell sculpture examination, we drew on data published by Zieritz et al. 46 . Individual specimens of dry shells lacking soft anatomy, also examined for shell shape and measured morphometrically (see below) were examined for the arrangement of adductor and retractor muscle scars following McMichael & Hiscock 10 .

Genetic methods. Sequence alignment construction. We assembled mitochondrial Cytochrome c Oxidase
Subunit I (COI) and 16S rDNA (16S) gene sequences originally published by Klunzinger et al. 27 and Benson et al. 28 . Available nuclear gene sequences (18S and 28S rDNA) were not used in our analysis because they have insufficient variation to be informative 28 . Sequences were retrieved from GenBank, along with sequences for two other Hyriid species for use as outgroups (Velesunioninae: Velesunio ambiguous (Philippi, 1847 26 ) and Hyriinae: Hyridellini: Cucumerunio novaehollandiae (Gray, 1834 94 )) (see Table 3). We limited the outgroup species to two hyriids in our tree because of the extreme divergence of W. carteri, even from other Hyriidae (see Graf et al. 21 ). Individual alignments for both genes were built using the ClustalW accessory application in Bioedit 7.2.5 104 , before inspecting and trimming to equal length in MEGA X version 10.1.8 105 .
Phylogenetic analyses and species delimitation. The COI alignment, including outgroups, was reduced to unique haplotypes using DnaSP v6 106 . Individual haplotype names correspond to those used in Benson et al. 28 . The alignment was partitioned by codon position in Mesquite version 3.61 107 , and the best fitting substitution model for each partition was identified by jModelTest 108 . This alignment was then used to construct Maximum likelihood (ML) and Bayesian inference (BI) trees in W-IQ-TREE 109 and MrBayes version 3.2.7a 110 respectively. The ML analysis was performed with 10,000 ultrafast bootstrap replicates using the standard settings of the selected substitution models. The BI analysis was performed with two independent runs of 1 × 10 7 generations sampling every 500 generations to achieve an average standard deviation of the split frequencies that was consistently < 0.01, and ESS values > 200. Convergence of the MCMC chains was also confirmed in Tracer version 1.7.1 111 . In order to determine the number of distinct taxa within W. carteri, three methods of species delimitation were applied to the COI dataset, excluding outgroups. Firstly, the distance-based method, Assemble Species by Automatic Partitioning (ASAP) 112 , was implemented using the Jukes-Cantor (JC69) option on the online webserver (https:// bioin fo. mnhn. fr/ abi/ public/ asap/). Secondly, a statistical parsimony method was run in TCS 1.21 113 using a 95% connection limit. Finally, the BI tree was assessed using Bayesian implementation of the Poisson Tree Process (bPTP) 114 on the bPTP webserver (https:// speci es.h-its. org/ ptp/) with the maximum allowable number of MCMC generations (5 × 10 5 ) and 20% burn-in. The ML and BI trees were edited in FigTree v1.4.4 (http:// tree. bio. ed. ac. uk/ softw are/ figtr ee/) and Inkscape v1 (https:// inksc ape. org).

Molecular diagnosis.
For the molecular diagnosis of taxa within W. carteri we considered fixed nucleotide differences found within the alignments of the full dataset for each gene (i.e., character attributes that were present in all individuals of one taxon while being absent in all individuals of the other two taxa [115][116][117] . For each gene, these characters were identified using the 'toggle conserved sites' option in MEGA X 10.1.8 105 . The uncorrected p-distance (mean and SE) to the nearest neighbour of each taxon was determined in the same software.
Shell measurements. Shells were measured to the nearest 0.1 mm using manual callipers or, for photographed specimens, to the nearest 0.5 mm using a ruler as a scale bar appearing in each photograph. Shells Morphometric analysis. We conducted traditional morphometric analysis and outline (Fourier shape) analysis to assess whether shell shapes are different for each of the Westralunio ESUs identified 27,28 . Details of specimens used in these two analyses are provided in the 'material examined' sections below. Geographic distribution of these samples is illustrated in Fig. 6. A total of 446 and 411 specimens were included in traditional and Fourier shape analysis, respectively: 294 and 273 "W. carteri" I, respectively; 140 and 126 "W. carteri" II, respectively; and 12 and 12 "W. carteri" III, respectively. Like Ponder & Bayer 11 and Sheldon 101 , we tested for differences in shell measurements and sagittal shell shape between the two species and among the three taxa, respectively. 'Traditional' shell shape indices 10 were calculated from shell measurements for maximum height index (MHI), beak height index (BHI), and beak length index (BLI), such that: MHI = MH L ; BHI = BH MH ; BLI = BL L . Firstly, significant differences in TL, MH, BH, BL, MHI, BHI and BLI, respectively, were tested for using ANOVAs followed by pairwise Tukey's posthoc comparisons. Secondly, a Principal Component Analysis (PCA) on these variables was performed, following ANOVAs and Tukey's posthoc comparisons on the first two PC-axes. Thirdly, we employed Discriminant Analysis (DA) to assess the proportion of specimens that would be assigned to the correct taxa based on these shell indices.
Overall shell shape was analysed using Fourier shape analysis 118 . This method breaks information on sagittal shell outlines of specimens into a set number of harmonics, each of which is explained by two Fourier coefficients, which are then analysed statistically. Specimens held in the Western Australian Museum collections were photographed using a Canon EOS 3D digital camera. A photographic stand was set up to hold the camera at the  27 WAM S112805 28 WAM S112806 28 WAM S112807 28 WAM S112808 28 WAM S112809 28 www.nature.com/scientificreports/ same distance, angle and focus for each photographed specimen. Black felt was used as a background medium to minimise shadows in the photos and a silver ruler was included in each photo in the same position as a scale bar. Additional digital photos of shell specimens from other museum collections were obtained, with permission, from the Mussel Project website 63 . Shell outlines of specimens were digitised into xy-coordinates using the program IMAGEJ 119 . The digitised outlines were then subjected to Fast Fourier transformation using the program HANGLE, applying a smoothing normalisation of 20 to eliminate high frequency pixel noise. Preliminary analysis indicated that the first 10 harmonics described the outlines with sufficiently high precision. Discarding of the first harmonic, which did not contain any shape information, resulted in a set of 18 Fourier coefficients per individual. After rotating outlines to maximum overlap with program HTREE, a PCA was performed on the 18 Fourier coefficients using program PAST 120 . Synthetic outlines of extreme shell forms were drawn using program HCURVE 118 .
To test for statistical differences in overall sagittal shell shape between the two species (W. carteri and W. inbisi), we conducted t-tests on the first two PC-axes and carried out an analysis of similarities (ANOSIM; 9,999 permutations, Euclidean distance) and DA on the set of 18 Fourier coefficients. The degree of shell inflation (relative shell width) was not considered because, in contrast to sagittal shell shape, this morphological character is strongly influenced by ontogenetic growth in Unionida 14,73 . Statistical analyses were conducted in PAST 3.22 121 (PCAs, DAs, ANOSIM), and R version 3.6.3 (ANOVAs and Tukey's post hoc comparisons).
Nomenclatural acts. The electronic edition of this article conforms to the requirements of the amended ICZN 44 .

Data availability
Spatial data for mapping are provided in the Methods section. Data for specimen records we examined for this study (Tables 3, S1, (Tables 3,  S1, S2 and S3) are available from GenBank (https:// www. ncbi. nlm. nih. gov/ nucco re/? term= Westr alunio+ carte ri). The dataset generated in this study is also available from the corresponding author on reasonable request. www.nature.com/scientificreports/