Multi-gene insights into the taxonomy and conservation of Tasmania’s galaxiid fishes

Timeline of allozyme analyses

The original allozyme overview study was undertaken in 1985, to identify areas of taxonomic uncertainty, obvious phylogenetic and phylogeographic patterns and a suite of markers for follow-up assessments of population structure in selected species (L. sealii and Galaxias maculatus; Pavuk 1997). As such, it focused on small numbers of individuals per site and taxon, screened for as many allozyme markers as were established in the SA Museum’s allozyme laboratory at that time. Whole fish were collected and snap frozen, often with the assistance of staff from the Tasmanian Inland Fisheries Commission, before being shipped to the SA Museum for inclusion in their frozen-tissue collection (the Australian Biological Tissues Collection; ABTC). These collections were planned to include exemplars of all species of Galaxias (10 species), Paragalaxias (4 species) and Neochanna (Neochanna cleaveri at that time was included in the genus Galaxias; Waters and White 1997), plus a single population of the mainland galaxiid G. olidus sensu stricto (i.e. referable to correct taxon on the basis of the taxonomic revision of Raadik 2014). It should be noted that Galaxias pedderensis and Galaxias niger are morphologically similar to Galaxias brevipinnis, and specimens were largely assigned to these two species on the basis of collection locality. Furthermore, although many (but not all) researchers have regarded G. niger as likely synonymous with G. brevipinnis (McDowall and Fulton 1996; Hardie et al. 2006; Raadik 2014), we have conservatively applied the name to the original 1985 collection from its type locality.

A subsequent allozyme study was then conducted in 1988, using additional frozen material collected and dispatched earlier that year. The aim of this study was to explore population structure and species integrity in the Galaxias truttaceus species group, including G. truttaceus plus its two sister species G. tanycephalus and G. auratus. This study focused on the polymorphic markers identified for these three species in the initial overview study, plus a handful of additional enzymes that are generally likely to be polymorphic. Ultimately, this second screen of 20 polymorphic loci, along with companion mtDNA restriction fragment-length polymorphism (RFLP) data, were published as two discrete studies, one on the widespread G. truttaceus (Ovenden and White 1990), and the other on its two lake-restricted congeners (Ovenden et al. 1993). Of relevance here is that the raw genotypes for all individuals screened were generated in our laboratory, and so were able to be integrated with those obtained in the final study (see below).

We next conducted one final allozyme study in 2012, focused on assessing the genetic distinctiveness of the threatened Western Australian population of G. truttaceus (thought prior to the study to be a subspecies). By this time, the ABTC collection of frozen tissues had been enhanced to include additional Tasmanian and some mainland populations of G. truttaceus. As a consequence, this final allozyme screen surveyed all populations for which frozen tissues were available (12 coastal and two landlocked sites for G. truttaceus) and again included both G. tanycephalus and G. auratus. The resultant published study (Morgan et al. 2016) included a companion mtDNA tree, but excluded the genetic data for the eastern landlocked populations; this latter decision reflected our desire to focus on the broader Australian perspective and subsequently explore the distinctive genetic profiles of these landlocked populations in a future study.

Herein, we present two analyses of all existing allozyme data for the G. truttaceus species group, one for the full 2012 allozyme dataset and the other incorporating the allozyme profiles of all sites surveyed for the 20 polymorphic loci common across all three allozyme studies (excluding sites with n = 1). Sampling details and sample sizes for all allozyme studies and the ‘2022’ analysis (i.e. the integrated dataset of 20 loci for the 1985, 1988 and 2012 studies) are presented in Table 2 and the geographic arrangement of these sites are displayed in Fig. 2.

Fig. 2.

Composite map showing the location of all sites surveyed across the various molecular studies. Site numbers mirror those used in Table 1.

General allozyme methods

Allozyme electrophoresis on muscle homogenates was undertaken as described elsewhere (Richardson et al. 1986; Hammer et al. 2007). Homogenates were successfully screened for various combinations of the enzymes employed in our other studies on galaxiids (Adams et al. 2014; Morgan et al. 2016). Histochemical stain recipes plus enzyme, locus and allozyme nomenclature all follow Hammer et al. (2007).

We employed a range of different approaches to analyse the allozyme data, depending on the dataset under consideration. For the initial overview study, we generated a neighbour-joining (NJ) tree among all individuals, constructed from a pairwise matrix of Roger’s genetic distances (Rogers’ R). This NJ tree was rooted using N. cleaveri, identified as the earliest-branching species among this group based on mtDNA and nDNA sequence data (Burridge et al. 2012). We also calculated the pairwise number of fixed differences and unbiased Nei’s distances (Nei Ds) among all species to assess their distinctiveness and for cross-referencing with the NJ tree among individuals. For the 2012 study we used principal co-ordinates analysis (PCoA) to assess the genetic affinities of individuals, independent of locality or taxon. Analysis of the ‘2022’ dataset involved generating an unrooted NJ network among sites on the basis of unbiased Nei Ds. All details for these methods have been published previously (Hammer et al. 2007; Adams et al. 2014).

mtDNA sequencing

As the majority of published mtDNA sequences for the Australian galaxiids involve the cytochrome b (cytb) gene, we constructed our mtDNA gene tree using a combination of newly generated cytb sequences plus Genbank-held exemplars for all taxa and most of the populations profiled in the allozyme studies. Importantly, we confirmed (Jonathan Waters, pers. comm.) that the G. pedderensis sequences published by Burridge et al. (2012) were obtained from a single individual, directly sourced from the population of G. pedderensis that was translocated into Lake Oberon prior to the species becoming extinct in the wild (Chilcott et al. 2013). Our final cytb dataset also included sequences for two individuals collected in 2014 by one of us (C. P. Burridge) from the type locality for G. niger.

Some of the new sequences were generated and edited using the methods and primers detailed in Morgan et al. (2016), whereas others were obtained using protocols that differed only in the choice of primers (those described in Nielsen et al. 1994), and edited using software Geneious Prime (ver. 2021.2, see https://www.geneious.com). All new sequences have been deposited in GenBank, with Accession numbers OQ738621–OQ738684 and OQ738809–OQ738811. The final gene tree was generated using RAxML (ver. 8.2.12, see https://github.com/stamatak/standard-RAxML; Stamatakis 2014) on the CIPRES cluster (see http://www.phylo.org/; Miller et al. 2010) by using the model GTRGAMMA and searching for the best-scoring maximum likelihood (ML) tree. Bootstrapping was set to finish on the basis of the autoMRE majority-rule criterion.

1985 allozyme overview study

The final dataset for the initial overview study comprised the allozyme profiles of 63 individuals at 43 putative loci. Excluded from this dataset was a single fish, allozymically identified as a G. truttaceus × G. brevipinnis F₁ hybrid (both species being present at that site; Table 2). Allozyme frequencies for each locus and species are summarised in Supplementary Table S1. Most of the 16 nominal species (15 Tasmanian species plus G. olidus) appeared as distinct clusters within the NJ tree (Fig. 3), the exceptions being one cluster comprising G. brevipinnis–G. pedderensis–G. niger and the other involving G. truttaceus–G. tanycephalus–G. auratus. This outcome is supported by the number of pairwise fixed differences among species (Table 3), with every species readily diagnosable by multiple fixed differences (range 6–34) except for the two clusters described above (range 0–3).

Fig. 3.

Neighbour-joining tree among the 63 individuals included in the 1985 allozyme overview study, rooted using N. cleaveri. Individuals are labelled by a unique species code (a one- or two-letter abbreviation) plus site; multiple individuals from the same site are labelled alphabetically. Highlighted individuals: supplied as G. pedderensis (blue); supplied as G. niger (green); G. tanycephalus (yellow).

Under the usually reliable assumption that phylogenetic relatedness is roughly correlated with overall genetic similarity, we infer from Fig. 3 the following results: (1) Galaxias johnstoni and G. fontanus appear to be sister species to the G. brevipinnis ‘complex’, (2) the four species of Paragalaxias comprise a monophyletic genus, (3) G. parvus and G. maculatus are not closely related to any other included species, (4) Tasmanian Galaxias species may not be monophyletic, and (5) the genus Galaxias is likely to be paraphyletic.

2012 allozyme study of G. truttaceus species group

The final allozyme dataset for this study comprised 91 individuals profiled at 57 putative loci (i.e. the dataset of Morgan et al. (2016) plus landlocked sites of Tr22 (n = 8) and Tr33 (n = 2)). A single fish identified by its allozyme profile as a G. truttaceus × Galaxias oliros F₁ hybrid (Site Tr33, Table 2) was excluded from this dataset. Allozyme frequencies by locus for each of the 16 sites are provided in Table S2. A PCoA of these 91 individuals displayed three primary clusters, corresponding to the three member species in this group. Conventional measures of among-species genetic divergence are similar for all pairwise combinations (Nei D range: 0.17–0.24); however, only G. auratus is fully diagnosable by fixed differences (Fig. 4). In contrast, G. tanycephalus and G. truttaceus are only marginally diagnosable by fixed differences, particularly when the latter is partitioned by habitat type (see next section).

Fig. 4.

Scatterplot of the relative scores in the first two dimensions for the principal coordinates analysis of 91 G. truttaceus group individuals, based on the 2012 allozyme study. Species are identified by a unique symbol corresponding to the cluster label. Symbols shown in green represent fish from landlocked or lake populations (blue circles for G. truttaceus being fish from ‘coastal’ sites that are presumed diadromous). A pairwise genetic-distance matrix among the major groups and subgroups is shown in the boxed inset; lower-left triangle shows number of fixed differences (numbers in parentheses allow a 10% tolerance for all shared alleles summed); upper right triangle shows unbiased Nei Ds.

As expected for a widespread species, G. truttaceus exhibited far greater genetic diversity than was present in its two lacustrine congeners. However, less predictably, a substantial proportion of this allozyme diversity (i.e. as shown in PCoA Dimension 2, Fig. 4) does not correlate with geographic distance (e.g. over 3500 km between WA and Tasmanian sites) but instead correlates well with whether the site is landlocked or ‘coastal’ (e.g. with easy access to the ocean). Indeed, putting aside species boundaries, a significant proportion of the genetic diversity in this complex clearly resides in landlocked habitats.

2022 combined allozyme analysis of G. truttaceus species group

The final 2022 allozyme dataset comprised 265 individuals at the 20 loci found to be polymorphic in the G. truttaceus species group (Ovenden and White 1990; Ovenden et al. 1993). Allele frequencies at all 21 sampling events across 19 sites are summarised in Table S3 (Sites Ta21 and A19 were independently sampled in each stand-alone allozyme screen). Four clusters corresponding to G. auratus, G. tanycephalus, coastal G. truttaceus and landlocked G. truttaceus are evident from the NJ network (Fig. 5) and PCoA (Fig. 4), with the mainland Tr 33 site again being the least divergent of the latter group. Importantly, this same pattern holds even after the inclusion of additional landlocked and coastal sites for G. truttaceus (Tr24 and Tr25 are lake populations; Tr10 and Tr12 are coastal sites), plus the addition of one extra site for both G. auratus (A18) and G. tanycephalus (Site Ta20; Site Ta21 also represented by a second sampling event).

Fig. 5.

Unrooted neighbour-joining network for the combined 2022 allozyme dataset (all populations of the G. truttaceus species group, profiled at 20 polymorphic loci). Populations represented in the 1985 study are in bold blue text; those in the 1988 study (Ovenden and White 1990; Ovenden et al. 1993) are in bold red text. Landlocked populations are highlighted in blue.

Matrilineal genealogy for the Tasmanian galaxiids

The final mtDNA dataset comprised 100 cytb sequences of length 1120 bp, representing all species and most populations. This included 68 novel sequences, 17 Genbank sequences directly linked to the 2012 allozyme study (Morgan et al. 2016), and 15 Genbank sequences from other sources. RAxML recovered a single tree and the rapid bootstrap search ended after 450 replicates (Fig. 6). This tree shares many features with the original allozyme overview study (Fig. 3). Both analyses support the monophyly of (1) G. johnstoni and G. fontanus with the G. brevipinnis ‘complex’, (2) the four species of Paragalaxias, and (3) the three species in the G. truttaceus species group. Likewise, both show that G. maculatus and G. parvus are distinctive early branching lineages, and both infer the genus Galaxias overall and Tasmanian Galaxias in particular as being unlikely to be monophyletic. A final point of similarity was the inability to distinguish G. niger from G. brevipinnis; indeed, the same cytb haplotype was present in both G. niger and the G. brevipinnis collected from Reservoir Lakes in 2014 (Site 5; Fig. 6).

Fig. 6.

Maximum-likelihood tree, based on 1120 bp of cytb and rooted using N. cleaveri. Nodes with bootstrap support of 90% or more are indicated by an asterisk. As for Fig. 3, individuals are labelled using their abbreviated species code (Ol = G. oliros) plus a site code (if known); multiple individuals at a site are labelled numerically (.1, .2, etc.). Genbank exemplars (red) are labelled with their accession number and, where traceable, their source site (Site 2# was the translocated population of G. pedderensis, originally sourced from Site 2); all other sequences were listed in Genbank only as from ‘Tasmania’. Highlighted sequences: G. truttaceus from Site 22 (Shannon Lagoon; yellow); F₁ hybrids (grey); G. pedderensis as included in the 1985 allozyme study (blue); G. niger (green).

Notwithstanding the overall concordance of allozyme and mtDNA perspectives, there are two notable differences between them. First and foremost is the discrepancy between the cytb haplotypes displayed by the two G. pedderensis included in the 1985 allozyme study (Pe 2.1 and Pe 2.2, Fig. 6) when compared with the true-to-label cytb haplotype available on Genbank (accession number JN232623.1, from Burridge et al. 2012). Although both datasets confirm that G. pedderensis is closely related to G. brevipinnis, the mtDNA results suggest that the fish allozymed in 1985 may not have the same provenance as the separately sourced tissue sequenced by Burridge et al. (2012). Second, although both agree that the three members of the G. truttaceus species group are closely related, two modest discrepancies are evident; the nuclear markers suggest that G. tanycephalus and G. truttaceus are sister species (Fig. 3–5), whereas mtDNA places G. tanycephalus sister to G. auratus (Fig. 6). In addition, the mtDNA shows that the landlocked population near Shannon Lagoon (Site Tr22) is the only population of G. truttaceus that harbours cytb haplotypes from each of the two primary cytb subclades, which otherwise diagnose the G. truttaceus versus G. tanycephalus–G. auratus matrilineal dichotomy.

Species boundaries in Tasmanian galaxiids

Despite occupying less than 1% of Australia’s total landmass, Tasmania is home to 25% of its threatened freshwater fishes (10 of the 40 species listed nationally as ‘extinct in the wild’, ‘critically endangered’, ‘endangered’, or ‘vulnerable’; Lintermans et al. 2020; Department of Climate Change, Energy, the Environment and Water 2022). All the threatened Tasmanian species are galaxiids and most lack even the most basic multi-locus genetic assessment of their taxonomic integrity. Furthermore, their threatened status limits routine contemporary sampling of tissues, thus placing additional value on our historic tissue collections. In this light, the genetic appraisals presented herein provide an important baseline assessment of species boundaries plus an independent perspective on evolutionary relationships in this highly threatened group (Adams et al. 2014; Raadik 2014; Page et al. 2018; Lintermans et al. 2020).

Both allozyme and mtDNA analyses were concordant in unequivocally supporting the taxonomic validity of the following Tasmanian galaxiids: G. fontanus, G. johnstoni, G. maculatus, G. parvus, N. cleaveri, Paragalaxias dissimilis, Paragalaxias eleotroides, Paragalaxias mesotes and Paragalaxias julianus (all Tasmanian endemics except for N. cleaveri and all of conservation concern apart from G. maculatus). These nine species were readily diagnosable from each other at multiple genes and, apart from a single F₁ hybrid, displayed no evidence of hybridisation with co-occurring or sibling congeners. The remaining six species fell into one of two distinctive species groups, namely (1) a G. brevipinnis lineage (the diadromous G. brevipinnis plus the lacustrine endemics G. pedderensis and G. niger), or (2) the G. truttaceus complex (diadromous and landlocked G. truttaceus plus the lacustrine endemics G. auratus and G. tanycephalus).

Our G. pedderensis and G. niger exemplars, collected in 1985 and assigned to these taxa primarily on the basis of being taken from the type locality for each species, were genetically indistinguishable from G. brevipinnis for both allozymes and cytb. Such an outcome has two possible explanations, namely (1) the fish collected in 1985 were actually G. brevipinnis and thus the real endemics were not sampled, or (2) both endemic ‘species’ are simply lacustrine populations of G. brevipinnis that have undergone discernable phenotypic divergence since being isolated from coastal populations. Unfortunately, no voucher specimens were collected at that time to distinguish between these alternatives. In the case of G. pedderensis, we suggest that the first explanation (i.e. genuine G. pedderensis were not collected in 1985) is the most tenable, for two reasons. First, the G. pedderensis sourced by Burridge et al. (2012) displayed a cytb haplotype that was sister to but distinct from all G. brevipinnis haplotypes (n = 16; Fig. 6) plus was genetically most similar to G. johnstoni rather than to G. brevipinnis for two nuclear genes (S7 and RAG-1; Fig. 7), and thus seems genuine. Second, historical records indicate that G. brevipinnis was known to have colonised Lake Pedder by the early 1980s, although intensive surveys failed to detect G. pedderensis in the lake itself after 1987 (Chilcott et al. 2013). Thus, it is plausible that the fish collected from Lake Pedder in 1985 were G. brevipinnis rather than G. pedderensis.

Fig. 7.

Neighbour-joining trees constructed using p-distances for G. brevipinnis and G. truttaceus complex species, on the basis of the published sequences for two nuclear genes of S7 and RAG-1 (Burridge et al. 2012). Sequences are labelled with their Genbank accession numbers. Nodes with bootstrap support of 90% or more are indicated by an asterisk.

In contrast to G. pedderensis, all available genetic data suggest that G. niger is likely to represent a darker, conspecific form of G. brevipinnis (see also Fulton 1990; McDowall and Fulton 1996; Hardie et al. 2006; Raadik 2014). Nevertheless, these data cannot exclude the possibility that, as for G. pedderensis, there was once an endemic galaxiid in Reservoir Lakes when first collected, but which subsequently became rare or was extirpated following colonisation by G. brevipinnis. This possibility could be confirmed or refuted in either of two ways. First, an explicit morphological comparison of the two species could re-assess the original assertion by Andrews (1985) that G. niger is readily diagnosable from G. brevipinnis on the basis of the latter’s large, paired fins and canine teeth (the two species were not otherwise compared). Second, it is now possible to generate genomic data from formalin-preserved voucher specimens, an approach that would help resolve this and many other taxonomic grey areas (Card et al. 2021), including any uncertainty surrounding G. pedderensis.

As well as confirming previously observed low levels of genetic divergence between G. truttaceus and its sister species G. auratus and G. tanycephalus (Ovenden et al. 1993; Morgan et al. 2016), our study further explored the relationships among these species by including allozyme and mtDNA assessments of additional landlocked G. truttaceus (Sites Tr22, Tr24, Tr25 in Tasmania, and Tr 33 from the Murray–Darling Basin; Fig. 2, 4–6). Together these analyses showed a consistent pattern whereby landlocked G. truttaceus display levels of nuclear genetic divergence from their diadromous coastal counterparts that mirror those between the three species themselves (Fig. 5). Importantly, whereas most landlocked G. truttaceus individuals exhibited the cytb subclade unique to this species, three of the five fish from Shannon Lagoon (Site Tr24) possessed a haplotype belonging to the subclade representing the two landlocked species (Fig. 6). Given how similar both subclades are to one another (average cytb sequence divergence = 2% and minimal divergence at S7 and RAG-1; Fig. 7), such an outcome either reflects lineage sorting among recently diverged taxa or post-isolation gene flow, both further indicators of the close affinities among these three species.

Evolutionary relationships

Our nuclear and matrilineal datasets complement and support the comprehensive, family-wide phylogeny of Burridge et al. (2012). Both studies found that the Tasmanian galaxiids are a heterogeneous assembly of evolutionarily distinct species (herein G. maculatus, G. parvus and N. cleaveri, but also including G. pusilla and L. sealii), plus three distinctive and monophyletic clades, namely, Paragalaxias, the G. brevipinnis complex (herein G. brevipinnis, G. ‘niger’, G. pedderensis, G. johnstoni and G. fontanus, plus many other galaxiids from New Zealand and New Caledonia), and the G. truttaceus complex (G. truttaceus, G. auratus and G. tanycephalus). Both studies also demonstrated that the genus Galaxias is not monophyletic, particularly with respect to Paragalaxias, which, along with G. parvus, appears to have closer evolutionary ties to members of the G. olidus complex in south-eastern Australia (Fig. 3; Burridge et al. 2012).

Isolation and landlocking as drivers of speciation

As Burridge et al. (2012) have already published a thorough exploration of evolutionary patterns, biogeographic affinities and life-history characteristics for the Galaxiidae, further discussion here will be limited to one key evolutionary theme applicable to this group, namely landlocking. Detailed studies of galaxiids, both in Australia and elsewhere, have demonstrated that the establishment of a non-migratory landlocked population from a diadromous, vagile ancestor has often driven rapid morphological, ecological and genetic divergence in the newly founded population which, if left untainted by subsequent gene flow, may ultimately result in speciation (Humphries 1990; Waters and Wallis 2001; Zattara and Premoli 2005; Waters et al. 2010). This phenomenon is most notable in the extended clade of species containing the diadromous G. brevipinnis and its many congeners (especially in New Zealand and, as further demonstrated herein, for Tasmania; Burridge et al. 2012). Moreover, it may also explain speciation events in two of Australia’s other galaxiid species complexes, one involving the diadromous G. maculatus and its non-migratory sister species G. rostratus and G. occidentalis, and the other being the G. truttaceus complex in Tasmania.

Our genetic data for the G. truttaceus complex showed a range of populations at differing stages along the path from landlocking to speciation. Mirroring an earlier morphological study (Humphries 1990), lacustrine populations of G. truttaceus displayed levels of genetic divergence from their diadromous counterparts that approached those among the three described species within the complex. Given that the budding of isolated populations from a vagile ancestor seems to be a common life-history characteristic of galaxiids, isolated populations of G. truttaceus and other species ought to be priority candidates for any genomic search for cryptic species or to identify high-value populations for conservation. A landlocked population of G. truttaceus in the Murray–Darling Basin (Site Tr33), variously regarded as either a bait bucket introduction or long-lost native population (Humphries 2009), was genetically intermediate between lacustrine and diadromous, which implies that it may either be native to the region or display a founder effect after translocation. This prediction requires mtDNA combined with SNPs or similar genomic data to confirm which scenario is in play.

Conservation perspectives

Freshwater fishes are well known for their tendency to hybridise with congeners (Verspoor and Hammart 1991; Hammer et al. 2013b), particularly after experiencing anthropogenic habitat modification (McFarlane and Pemberton 2019). We detected two F₁ hybrids in this study, both involving sympatric species from quite different evolutionary lineages (G. brevipinnis and G. truttaceus, Site 24; G. truttaceus and G. oliros, Site 33) and neither identified a priori on the basis of external appearance. Unacknowledged introgression from widespread common species can seriously affect conservation efforts for rarer and narrowly distributed species (Guildea et al. 2015; Frankham et al. 2017), particularly if managers unwittingly choose only admixed populations for conservation management or as source populations for future translocations.

In recognition of these concerns, Adams et al. (2011) advocated that congeneric species should initially be included in any ‘best practice’ conservation genetic or genomic study of an imperilled freshwater fish if at least one of the following criteria was satisfied for the target species and its congener:

currently sympatric or parapatric, currently allopatric but with a reasonable likelihood of prior sympatry or parapatry, allopatric but co-occurring in a natural corridor (e.g. a river) capable of facilitating gene flow, or conspecific prior to a taxonomic revision [Adams et al. 2011, p. 768].

With conservation concerns over so many Tasmanian galaxiids, our study effectively satisfies this recommendation by demonstrating that all vulnerable species were, in the 1980s, readily diagnosable at multiple unlinked loci except for those involved in the G. brevipinnis–pedderensis–‘niger’ and G. truttaceus–tanycephalus–auratus complexes. Nevertheless, as anthropogenic-directed catchment changes, accidental and deliberate translocations, and climate change have seen G. brevipinnis colonise new catchments in Tasmania and elsewhere (e.g. the Murray–Darling Basin; Waters et al. 2002), conservation managers should always be mindful of the potential for hybridisation and cryptic introgression between congeneric species.

Future molecular research priorities

In the current age of genomics, it is likely that all future molecular genetic explorations of species boundaries, population structure, and wildlife conservation will aspire to large-scale SNP datasets (Georges et al. 2018). One such study is already being undertaken on taxonomic integrity and population structure in the Swan River galaxias G. fontanus (Bruce Deagle, CSIRO, pers. comm.) and the fine-scale assessments provided for the few remaining natural populations and their translocated counterparts will be invaluable in guiding future management efforts. We advocate comparable studies for all of Tasmania’s imperilled galaxiids, using the evolutionary and taxonomic affinities presented herein to guide which studies can be stand-alone and which require the inclusion of a few samples of the relevant congeners (e.g. G. brevipinnis ought to be included in any conservation genomic study of G. pedderensis). Such studies can be non-invasive (Le Vin et al. 2011) and need not involve large sample sizes to be highly informative, provided they include all known populations and, if possible, historic exemplars (Baverstock and Moritz 1996; Card et al. 2021).

A second molecular research priority ought to be the contemporary genomic sampling of all landlocked populations of apparently common species, to assess the possibility they represent unacknowledged ‘cryptic’ species (whether truly cryptic or simply not identifiable a priori morphologically). Even if shown to be conspecific, the levels of genetic diversity contained in landlocked populations are often disproportionally high compared with those in other, more connected populations. Such populations merit recognition as having high value conservation and evolutionary potential should the species suffer future declines in abundance.