Convergent morphological and genetic patterns delimit Evolutionary Significant Units in endangered Aipysurus sea snakes

Abstract

Background

Identifying significant adaptive variation within species is a key focus of efforts to preserve evolutionary potential under environmental change. We analysed genome-wide SNPs and ecomorphological traits to define units for the conservation of Aipysurus sea snakes. This group contains nine fully-marine taxa that are primarily endemic to the northern Australian region, including two critically endangered and one endangered listed species.

Results

Results show convergent patterns of genetic and morphological divergence between coastal and offshore populations in four of five co-distributed Aipysurus lineages. In A. apraefrontalis, A. duboisii, A. foliosquama, and the A. fuscus-A. tenuis complex, individuals from the WA Coast and remote Timor Sea reefs form distinct genetic clusters with gene-tree based divergence estimates of five hundred to nine hundred and fifty thousand years. Coastal populations in all four of these lineages have longer bodies, higher vertebrate counts, and lighter colour patterns compared to offshore-reef populations. In A. laevis (the fifth co-distributed lineage), body length and colour pattern variation mirrored the coastal-offshore divisions seen in the other Aipysurus, but this species showed different patterns of geographic genetic structure that are incongruent with current species boundaries.

Conclusions

This striking genetic and morphological convergence implicates shared patterns of biogeographic isolation and suggests local adaptation related to locomotory performance and crypsis in open soft-bottomed coastal habitats versus structurally complex reefs. We recommend that coastal and offshore-reef populations of A. duboisii, A. apraefrontalis and A. foliosquama be recognised as evolutionarily significant units (ESUs), while the current taxonomy of A. fuscus, A. laevis, A. pooleorum and A. tenuis be retained.

Supplementary information

The online version contains supplementary material available at 10.1186/s12862-025-02481-w.

Introduction

Identifying genetically and ecomorphologically divergent populations within species provides valuable information for biodiversity conservation. Where such studies uncover common patterns of diversity among species, this is an important step in understanding how geographic and ecological factors shape speciation.

The nine species of Aipysurus sea snakes [1] are fully marine adapted and often thought to be primarily coral reef dwellers [2]. They are primarily Australian in distribution, with eight out of nine taxa found in waters surrounding the continent, five of which are endemic to Western Australia and the Timor Sea reefs of the Northwest Shelf. Two of these species (A. apraefrontalis and A. foliosquama) were only known from Ashmore and Cartier Reefs; however, unexplained and rapid declines led to their extinction from their entire known distribution by 2006 [3–5]. Examination of museum material and further collecting on the Western Australian (WA) Coast showed that these species have a significantly larger distribution and were not extinct as previously feared [6–8]. These species are classified as Data Deficient on the IUCN Red List and are listed as Critically Endangered under Australia’s EPBC Act 1999. The remaining Timor Sea endemic, Aipysurus fuscus, is more widespread on various reefs in the Timor Sea, but declines and evidence of contemporary hybridisation with the more common A. laevis [9] have led to its listing as endangered. The Shark Bay endemic A. laevis pooleorum was described as subspecies of A. laevis [8], although it was raised without justification to full species by Wells and Wellington [10] and is listed as Data Deficient on the IUCN red list. Analyses of mitochondrial and microsatelitte markers [11, 12] have shown significant population structure between WA Coast and Timor Sea lineages in all species of Aipysurus examined. Timor Sea populations were more closely related to those on the east coast of Australia than populations from the geographically closer WA Coast.

The north-western coast of Australia is a vast region of continental shelf between Exmouth and the Bonaparte Gulf. Within this region lies a recognised barrier for shallow water taxa, located between the Canning and Kimberley basins and meeting near Cape Leveque. There is no contemporary deep-water barrier between these basins to limit dispersal for shallow water taxa (reviewed in [13]). However, the extreme tidal regime in the region, and the outflow from King’s Sound, has been shown to restrict gene flow in fish and invertebrate taxa across this boundary [14, 15]. The strength of this barrier varies considerably among taxa, with life history traits associated with dispersal and site attachment likely influencing the degree of population structure [16, 17].

The geomorphology and therefore area of available habitats differ considerably between the Canning and Kimberley. The Canning basin is underlain by sedimentary rocks and gently slopes towards the continental shelf with relatively featureless flats of soft sediment [13, 18]. No large permanent rivers or estuaries drain into the Canning Basin meaning sediment outflow and turbidity is low. The Kimberley basin has a more complex geomorphology, with complex rias, and many offshore islands that represent high points in the landscape during lower sea level stands. Some of the largest tides in the world are present in this region, with large drainages that deposit great quantities of sediment into coastal habitats. Offshore of the Kimberley, tides and tidal outflow are significantly lower, but the underlying geomorphology remains complex. In deep regions, soft sediments predominate; along the edge of the shelf banks rise subaerially to within ~20 metres of the surface, and most of these are dominated by Halimeda. However, the largest of these banks are formed of coral and large reefs around low islands on the Ashmore platform [13, 19].

In fish, contrasting selection pressures in reef and open water habitats have driven differences in body shape. In coral reef fishes, a less elongate and (dorso-ventrally) deeper body is favoured as this improves the ability to manoeuvre and accelerate quickly in a cluttered environment. In contrast, fishes found in open habitats tend to be elongated with a thinner body to improve swimming efficiency over extended distances [20]. This trait, known as fineness, is expressed as the ratio of body length to body width [21]. Sea snakes share an ‘anguilliform’ mode of swimming with fish [22] and therefore are subjected to similar environmental and hydrodynamic pressures as fishes.

Colour pattern is important in sea snakes for predator avoidance through misdirection, crypsis and aposematism [23–25]. Most Aipysurus species exhibit cryptic colouration, likely shaped by colour matching to their environment, such as the shadowed crevices of coral reefs. However, industrial melanism has been suggested as a possible driver of melanic Emydocephalus in heavy metal rich waters in New Caledonia [26], even though melanic snakes are also present in relatively pristine offshore reefs in Australia.

Aipysurus sea snakes in Western Australia offer an excellent system to investigate evolutionary responses during habitat transitions due to the parallel population structure in five species groups between coral reef dwelling Timor Sea populations and soft bottom dwelling populations off the WA Coast. Here, we aimed to combine SNPs and morphology data to define evolutionarily significant units (ESUs) for conservation. We analysed several thousand SNPs from 116 individuals using phylogenetic and clustering methods. In addition, we measured a small number of ecologically relevant morphological traits from 201 specimens and used photographs of specimens and live individuals to characterise colour patterns.

Methods

Sampling

Tail and liver samples were obtained for 116 individuals of seven Aipysurus species (S1). Samples were gathered from museum collections (Museum and Art Gallery of the Northern Territory (MAGNT), South Australian Museum (SAMA) and Western Australian Museum (WAM)). 19 samples from Western Australia were obtained during fieldwork between Exmouth and Broome between 2012 and 2022. For population-level analyses, we targeted five species groups; three of these are endemic to Western Australia and the Timor Sea (A. apraefrontalis, A. foliosquama, and a combined data set of close sister taxa A. fuscus and A. tenuis), and two are widespread from Australia to New Caledonia (A. duboisii and A. laevis).

Genomic DNA was extracted using the Gentra Puregene tissue kit [27] following the manufacturers protocol with the following modifications: 1) tissue was ground directly in the cell lysis solution, 2) an in house lysis buffer solution (10MMTris/2% SDS/0.1 M EDTA) was used instead of the kit product, 3) Ammonium acetate was used instead of “protein precipitation solution”. Extracted DNA was then quantified using a Quantus fluorometer to obtain > 500ng of DNA at concentrations of 50ng/μl. Samples were concentrated using a Centrivap DNA concentrator if required and checked for quality using gel electrophoresis. Samples were sent to Diversity Arrays Technology Pty Ltd (DArTseq, Canberra, Australia) where SNP genotyping was conducted using a proprietary genome complexity reduction pipeline using a pair of restriction enzymes (PstI and HpaII) [28, 29]. After initial digestion/ligation reactions and amplification samples are sequenced on an Illumina Hiseq 2500.

We obtained raw demultiplexed reads from DArTseq which we checked for quality using FASTQC v.0.12.1 [30] and then filtered out adaptors and quality trimmed using default parameters in BBduk [31]. Common microbial and human contaminants were removed using kraken2 [32]. We called SNPs using iPYRAD v.0.9.85 [33] run on the University of Adelaide Phoenix HPC. Filtered and demultiplexed reads were assembled de novo, with the cluster threshold set to 0.85, mindepth (statistical and majrule) to 5, and maxdepth to 10,000. The minimum number of genotyped samples per locus was set to 90% in order to maximise SNPs called, as overly stringent filtering can cause allelic dropout at deeper phylogenetic scales [34]. Invariant Sites were removed and one SNP per locus was chosen for the subsequent phylogenetic analyses. SNPs found in only a single sample were excluded from final SNP matrices. SNPs were called separately for the phylogenetic dataset and for each of the five species groups (S3).

We reconstructed a phylogenetic tree using a concatenated dataset including all samples. Using a GTR+ASC Maximum likelihood tree reconstruction was implemented in IQTree using a GTR+ASC model to account for ascertainment bias in reconstructing trees from concatenated SNP data, node support was assessed using 1000 ultrafast bootstrap replicates. We initially rooted trees using Emydocephalus annulatus, as previous work has recovered this as the well supported sister to Aipysurus [35], but for final analyses we used Aipysurus mosaicus as this species was recovered as the well supported sister. Trees were visualised in FigTree v.1.4.4 [36]. Finally, we used Snapper [37] to estimate divergence times. Snapper is implemented in the BEAST2 [38] family of software and is an accurate implementation of the SNAPP multi-species coalescent species tree method, but is designed to handle greater numbers of samples in feasible computation times. We selected 1000 random SNPs to minimise computation time. We used default priors and a secondary calibration of 6.8 Mya for the root node (Aipysurus mosaicus vs. other Aipysurus) [39]. Four independent chains of 500,000 iterations were run with the first 125,000 discarded as burnin, with ESS values assessed using Tracer v.1.7.2 [40]. Individual runs were combined with LogCombiner and TreeAnnotator and trees were visualised in DensiTree and Figtree v.1.4.4 [36]. We performed a Principal Coordinate Analysis (PCA) using the SNP data for each species and species complex separately using the R package adegenet [41]. To further assess population structure and ancestry we used the likelihood-based program sNMF v.1.4 [42]. This program quickly and efficiently estimates optimal K values and ancestry coefficients for large SNP based datasets. For each species complex we ran 100 replicates from K 1 to 10 and used the K value which minimised cross entropy, with a tolerance of 0.0001.

Small sample sizes (n = 2 to 16 per population) were used for population genetic analyses of A. apraefrontalis, A. foliosquama and A. fuscus-A. tenuis. Simulation studies have shown that when large numbers of SNPs are used, population genetic estimates such as F_ST can be calculated from as few as two individuals [43, 44]. While we acknowledge that some degree of inaccuracy in calculating diversity metrics can be expected with smaller sample sizes, the remoteness and severe population declines of Timor Sea populations make obtaining greater sample numbers difficult. Nucleotide diversity (π) was calculated for each population using the program “pixy” [45]. To calculate heterozygosity, inbreeding coefficients, F_ST statistics and 95% confidence intervals we used the basic.stats, genet.dist and boot.ppst functions in the R package hierfstat [46].

Morphological analysis

We gathered morphological data from 191 sea snakes in the genus Aipysurus from seven nominal species from localities in the North-west of Australia from Shark Bay to the Timor Sea. Specimens were identified to species using the key in Rasmussen et al. [47], and they were held in the following museums: Australian Museum (AMS); Museum and Art Gallery of the Northern Territory (MAGNT); South Australian Museum (SAMA); Western Australian Museum (WAM). We collected data for seven ecologically relevant morphological characters: Ventral Scale counts (which show a 1:1 relationship with vertebral counts in Aipysurus); Snout to Vent Length (SVL); Tail Length (TailL); Forebody circumference at 4 head widths behind the head (Fore); Hindbody Circumference at ¾ of SVL (Hind); Head Length (HeadL); Head Width (HeadW) (S4) [23–25, 48].

We pooled sexes, as for some populations there were few available specimens, and scaled morphometric measurements by SVL to get relative measurements. We analysed morphological characters within each species to measure variation between populations. First, to assess whether number of vertebrae was correlated with body size we plotted SVL against ventral scale counts (which show a 1:1 relationship to vertebral counts). Second, to determine whether vertebral count was positively correlated with fineness (i.e. the ratio between body length and body width: [21]) we plotted Hind against vertebrae. We calculated Pearson correlation coefficients in R. We divided SVL by Hind to measure fineness, although we note as we measured the circumference of the body rather than the point-to-point width it is not directly comparable to fineness. We measured the circumference rather than width because formalin fixed sea snakes usually lose their laterally compressed shape, hence body width would give widely varying values. We tested for significance between lineages in fineness using a student’s T-tests. To visualise the datasets, we plotted them using ggplot2 [49]. Guided by the phylogenetic dataset, we analysed A. fuscus and A. tenuis together.

We used photographs from museum specimens and field-caught individuals to assess colour patterns. We used the same three broad colour pattern morphs used by Shine et al. [25], for Emydocephalus with a modified approach to scoring: melanic, intermediate and high contrast banded/light. To determine colour pattern morph we counted the ratio of light to dark scales across a single unit of bands at midbody. We used the midlateral scale row, for example in a snake with 19 midbody scale rows we would count scales at row 5. We counted scales mid laterally as most Aipysurus sea snakes are counter shaded with a dark dorsum and pale venter although there are exceptions. The ‘melanic’ morph was defined as those which were monochromatically dark with no obvious bands or pattern, ‘intermediate’ were those with a low contrast pattern or a ratio of dark:light scales > 0.5 and ‘light’ were defined by a ratio of dark:light scales < 0.5. The same observer (JHN) scored each specimen. Bar plots for colour pattern were produced in ggplot2 [49].

Results

Molecular analysis

The phylogenetic SNP dataset contained 6490 SNPs with 9.42% missing data. The phylogenetic tree topology had largely strong support with bootstrap values of 100% for most geographically concordant nodes, notably excluding internal nodes within Aipysurus laevis. (Fig. 1). At the species level the topology of our SNP tree was consistent with the mitochondrial tree previously recovered by Nitschke et al. [12]. We obtained conflicting results to those of Lukoschek [11], but this is due to misidentification rather than differing topology. In Lukoschek [11] A. tenuis samples are misidentified as A. fuscus, and WA Coast A. laevis are misidentified as A. tenuis (D’Anastasi, in prep.). Intraspecies relationships remained largely consistent with the mitochondrial dataset of [12], except for A. laevis detailed below, as well as the addition of New Caledonian samples.

Fig. 1.

Concatenated SNP tree of Aipysurus reconstructed from 6490 SNPs in IQTree. Rooted using Aipysurus mosaicus. Rooted using Aipysurus mosaicus. Yellow = WA Coast, Brown = Timor Sea, Red = Gulf of Carpentaria and Arafura Sea, Maroon = Eastern Australia, Blue = New Caledonia

After the root node of Aipysurus mosaicus, the remainder of Aipysurus was divided into two well supported clades (Bootstrap support = 100) (Fig. 1). First, the Aipysurus laevis clade consisted of A. laevis, A. fuscus and A. tenuis, with a close sister relationship between the allopatrically distributed A. fuscus and A. tenuis. Within A. laevis there was considerable conflict within early diverging nodes, particularly among Timor Sea samples. Broadly there was a well-supported split between WA Coast individuals, including two Aipysurus pooleorum samples from Shark Bay, and those from the Gulf of Carpentaria, East Coast and New Caledonia populations. Samples from Heywood Shoal grouped closely with WA Coast individuals, while samples from Ashmore and Scott Reefs were scattered among both clades. New Caledonian samples formed a well-supported lineage (Bootstrap Support = 100) within the broader Eastern clade.

The other major clade consisted of Aipysurus duboisii versus A. apraefrontalis plus A. foliosquama. The relationships among internal branches of A. duboisii were largely consistent with previous phylogenetic trees, with the WA Coast population as sister to the remaining populations. A single individual from the East Coast (Ad025) was recovered as sister lineage to other Eastern populations. The New Caledonian population was recovered within a broader East Coast clade.

The dated Snapper tree recovered the same topology as the concatenated SNP tree at the species level (Supplementary material: S2). Divergence times for currently recognised species pairs ranged from 1.8 Ma for A. apraefrontalis - A. foliosquama to 0.5 Ma for A. fuscus and A. tenuis. Within the sister species A. apraefrontalis and A. foliosquama divergence times between WA Coast and Timor Sea populations were near identical, 0.64 and 0.63 Ma respectively. Within A. duboisii three clades were well supported (posterior probability > 0.9), these were the WA Coast, Timor Sea and combined Gulf of Carpentaria + East Coast + New Caledonia clade with nodes dated as 0.95 Ma and 0.62 Ma. There was poor support for the relationship between the three Eastern clades of A. duboisii. Within A. laevis WA Coast + A. pooleorum and East Coast + Gulf of Carpentaria + New Caledonia clades were well supported (posterior probability > 0.9). Guided by the concatenated SNP tree we coded the Heywood Shoal and remaining Timor Sea (Scott and Ashmore Reef) samples separately. The sister relationship between the Heywood Shoal and Timor Sea lineages was poorly supported (posterior probability 0.72), and the sister relationship between all Timor Sea lineages and the WA Coast was equivocal (posterior probability 0.84). The divergence times between the two major but poorly supported (WA Coast + Timor versus East Coast + Gulf of Carpentaria + New Caledonia) clades was 0.71 Ma.

For the species level datasets we used the first SNP site from each RAD locus for downstream analysis. For A. apraefrontalis, we obtained 4,425 SNPs (4.22% missing); for A. duboisii we obtained 2,286 SNPs (3.93% missing); for A. foliosquama we obtained 2,790SNPs (1.79% missing); for A. fuscus/tenuis we obtained 3,101 SNPs (2.79% missing); and for A. laevis we obtained 5,661 SNPs (3.54% missing).

PCAs showed similar patterns of divergence in all three of the Timor Sea/Western Australian endemic species with a clear separation on PC1 (17.6% of variation explained in A. apraefrontalis; 34.3% in A. foliosquama; 14.1% in A. fuscus/tenuis). PC2 explained within-population variation, particularly in the most densely sampled groups (9.6% in A. apraefrontalis; 13.8% in A. foliosquama; 7.2% in A. fuscus/tenuis) (Fig. 2). In A. duboisii, PC1 explained 19.6% of the variation, and separated WA Coast, Timor Sea, and eastern Australian populations; PC2 explained 8.1% of variation and split the Timor Sea population from weakly separated clusters of eastern Australia and Gulf of Carpentaria versus New Caledonian individuals. In A. laevis PC1 explained 13.2% of the variation, separating New Caledonia at one extreme, the WA Coast and Timor Sea at the other, and the East Coast/Gulf of Carpentaria occupying an intermediate position.

Fig. 2.

PCAs (right) for five codistributed Aipysurus lineages. Colours follow map in Fig. 1

In the ancestral coefficient analyses, we found that the number of populations (K) minimising cross entropy for each species complex varied from 2 to 5 (Fig. 3). For A. apraefrontalis, A. foliosquama and A. fuscus/tenuis, K = 2 clearly separated the Timor Sea and WA Coast populations, although signals of admixture were particularly notable in A.fuscus/tenuis. For A. duboisii, we found K = 3 minimised cross entropy, with the WA Coast and Timor Sea populations clearly distinct with limited admixture. Samples from the Gulf of Carpentaria, eastern Australian coast and New Caledonia were interpreted as the same population. A. laevis had similar population structure to A. duboisii, but with K = 2 found to be the optimal number of populations. In contrast to A. duboisii, at K = 3 sNMF clustered the Timor Sea and WA Coast for A. laevis, while separating out New Caledonia and Eastern Australian populations. At K = 4 (Supplementary material: S2) sNMF largely clustered each species populations in the same manner (WA Coast, Timor Sea, Eastern Australia and New Caledonia), although showed potential admixture between the WA Coast and Timor Sea populations in A. laevis.

Fig. 3.

Ancestry coefficients calculated in sNMF of the five species of interest Aipysurus apraefrontalis (K = 2), Aipysurus duboisii (K = 3), Aipysurus foliosquama (K = 2), Aipysurus fuscus/tenuis (K = 2) and Aipysurus laevis (K = 3)

In both A. duboisii and A. laevis, the heterozygosity of the New Caledonian populations was substantially lower than in other populations, although inbreeding coefficients (Fis) were comparable to other populations (Table 1). In A. apraefrontalis and A. foliosquama the Timor Sea populations had notably higher heterozygosity and lower inbreeding coefficients, although both had sample sizes of only two. Outside of the aforementioned species, the nucleotide diversity was generally highest in the WA coast populations compared to others. F_ST between populations for the three NW Australia endemic Aipysurus ranged from 0.3209 in A. foliosquama, to 0.1497 in A. fuscus/tenuis. Among the more widely distributed species, pairwise F_ST was generally very low between East Coast and Gulf of Carpentaria populations (0.0238 and 0.0366 for A. duboisii and A. laevis respectively), while being relatively higher between New Caledonia and East Coast populations (0.1575 and 0.1603). A. duboisii and A. laevis differed in the patterns of F_ST between Timor Sea and other populations; in A. duboisii the F_ST was much higher between Timor Sea and WA Coast populations (0.1997) than in A. laevis (0.0206) (Table 2).

Table 1.

Genetic diversity metrics and sample sizes across populations in five Aipysurus sea snake species

	N	Ho	He	Fis	π
Aipysurus apraefrontalis
WA Coast	13	0.138	0.153	0.051	0.0654
Timor Sea	2	0.177	0.148	−0.229	0.1580
Aipysurus duboisii
WA Coast	6	0.126	0.136	0.039	0.0444
Timor Sea	7	0.102	0.107	0.034	0.0067
Gulf of Carpentaria	10	0.082	0.094	0.077	0.0059
East Coast	9	0.082	0.087	0.037	0.0217
New Caledonia	4	0.062	0.094	−0.024	0.0162
Aipysurus foliosquama
WA Coast	8	0.173	0.194	0.199	0.1606
Timor Sea	2	0.190	0.258	0.061	0.1912
Aipysurus fuscus/tenuis
Aipysurus tenuis	3	0.136	0.151	0.040	0.0317
Aipysurus fuscus	16	0.111	0.125	0.069	0.0544
Aipysurus laevis
WA Coast	12	0.099	0.109	0.054	0.0499
Timor Sea	8	0.099	0.109	0.052	0.0710
Gulf of Carpentaria	4	0.067	0.102	0.109	0.0388
East Coast	3	0.102	0.112	0.028	0.0428
New Caledonia	6	0.067	0.070	0.028	0.0133

N = sample size, Ho = Observed heterozygosity, He = Expected heterozygosity, Fis = Inbreeding coefficient and π = Nucleotide diversity

Table 2.

Pairwise F_ST values for codistributed Aipysurus populations in Australia and New Caledonia are on the lower diagonal

Aipysurus apraefrontalis	WA Coast	Timor Sea
WA Coast		0.1664
Timor Sea	0.3047
Aipysurus duboisii	WA Coast	Timor Sea	Gulf of Carpentaria	East Coast	New Caledonia
WA Coast		0.075	0.0804	0.0832	0.117
Timor Sea	0.1997		0.0399	0.0451	0.0597
Gulf of Carpentaria	0.2787	0.1733		0.0327	0.0475
East Coast	0.297	0.187	0.0238		0.0401
New Caledonia	0.3634	0.3045	0.1563	0.1575
Aipysurus foliosquama	WA Coast	Timor Sea
WA Coast		0.3051
Timor Sea	0.3209
Aipysurus fuscus/tenuis	Aipysurus tenuis	Aipysurus fuscus
Aipysurus tenuis		0.1112
Aipysurus fuscus	0.1497
Aipysurus laevis	WA Coast	Timor Sea	Gulf of Carpentaria	East Coast	New Caledonia
WA Coast		0.0779	0.0823	0.0723	0.0619
Timor Sea	0.0206		0.0996	0.0873	0.0804
Gulf of Carpentaria	0.0885	0.0785		0.0771	0.0703
East Coast	0.1071	0.0961	0.0369		0.0508
New Caledonia	0.2125	0.2063	0.1491	0.1603

All F_ST values were found to be significantly greater than 0 at a 95% confidence interval. Dxy values are presented on the upper diagonal

Morphological analysis

We sampled 25 Aipysurus apraefrontalis for the morphological analyses, with 11 from the Timor Sea and 14 from the WA Coast. A. apraefrontalis from the Timor Sea had a lower SVL (345–810) than those from the WA Coast (670–1050) and slightly lower but overlapping vertebral counts (139–152 vs. 147–157). The WA Coast population of A. apraefrontalis had a marginally higher fineness value than the Timor Sea (9.65 vs. 10.45) (Table 3), but this was not significant. In A. apraefrontalis, individuals from the WA Coast had greater SVL, accompanied by a roughly proportional increase in vertebral number and hindbody circumference compared to those from the Timor Sea (Figs. 4 and 5). Notably, the WA Coast A. apraefrontalis population also has a relatively smaller head (both HeadW and HeadL) than the Timor Sea population (Supplementary material: S4).

Table 3.

Mean fineness in Aipysurus species comparing the Timor sea to WA coast populations, calculated by dividing SVL by hindbody circumference

Population	N	Mean	Standard Deviation	Max	Min	T-Test
Aipysurus apraefrontalis
Timor Sea	11	9.64732065	1.65856914	12.1212121	7.66666667	0.19464652
WA Coast	14	10.4455529	1.18427573	11.8421053	8.38383838
Aipysurus duboisii
Timor Sea	23	10.4351797	1.57777076	13.0357143	8	0.04918824
WA Coast	10	12.413673	1.96064635	16.3461538	10.25
Aipysurus foliosquama
Timor Sea	41	9.3176247	1.17573889	11.6891892	7.6	0.57592804
WA Coast	5	9.81321648	1.79080083	13	8.70588235
Aipysurus fuscus/tenuis
A. fuscus	40	9.12866406	0.74800648	10.8333333	7.58241758	0.0147334
A. tenuis	5	12.0422843	1.87300126	14.0697674	10.1041667
Aipysurus laevis
Timor Sea	14	7.46502859	1.12255631	9.80769231	5.58673469	0.58116366
WA Coast	28	7.26741371	0.98630742	9.56	5.69306931

Fig. 4.

Scatterplots of ventral scales (y axis) versus SVL (x axis) in five codistributed Aipysurus lineages. Pearson correlation coefficients and p values are listed next to the respective lines of best fit. Yellow = WA Coast, brown = Timor Sea

Fig. 5.

Scatterplots of hindbody circumference (y axis) versus SVL (x axis) in five codistributed Aipysurus lineages. Pearson correlation coefficients and p values are listed next to the respective lines of best fit. Yellow = WA Coast, brown = Timor Sea

We sampled 33 Aipysurus duboisii for morphometric and vertebral data, with 23 from the Timor Sea and 10 from the WA Coast. A. duboisii from the Timor Sea had a lower SVL (420–1020) than those from the WA Coast (850–1340) as well as lower and non-overlapping vertebral counts (149–167 vs. 171–188) (Fig. 4). The WA Coast population of A. duboisii had significantly higher fineness than the Timor Sea population, which was the highest of any species measured (12.41 vs. 10.44) (Fig. 5, Table 3). In Aipysurus duboisii SVL is strongly correlated with both vertebral number and fineness.

We sampled 46 Aipysurus foliosquama for morphometric and vertebral data, with 41 from the Timor Sea and five from the WA Coast. A. foliosquama from the Timor Sea and WA Coast broadly overlapped in all measurements taken although there were some trends. Generally, the Timor Sea population had a lower SVL, vertebral count and fineness. WA Coast A. foliosquama reach greater sizes but this is less pronounced than in the preceding species.

We sampled 41 Aipysurus fuscus and 5 A. tenuis. Aipysurus fuscus had a significantly lower SVL, with no overlap among sampled specimens and a lower and non-overlapping number of vertebrae. A. tenuis had a higher mean fineness than A. fuscus (12.41 vs. 9.25) although this was not significant, likely due to small sample sizes of A. tenuis. This species complex followed the same trends as A. duboisii and had the greatest difference in mean fineness between any of the codistributed species measured.

We sampled 14 Aipysurus laevis from the Timor Sea and 28 from the WA Coast. There was a slight negative correlation between SVL number of vertebrae and SVL in was no significant correlation between number of vertebrae and SVL in any A. laevis populations, however the WA Coast population reached a greater SVL than the other populations sampled. Mean fineness was nearly the same between Timor Sea and WA Coast populations of A. laevis.

All Aipysurus species sampled showed a trend towards lighter colour patterns in WA Coast populations compared with those from the Timor Sea (Figs. 6–7). Of the eleven Aipysurus apraefrontalis examined from the Timor Sea three (27%) were dark brown with no discernible pattern, while the remaining eight (73%) were of the intermediate morph. The intermediate morph had a simple banded colour pattern with a dark brown base colour and approximately equally sized mid brown bands, typically specimens of this morph were smaller individuals. Ten of the eleven (91%) specimens examined from the WA Coast belonged to the light morph, with the ground colour being bright yellow. Dark brown bands that tapered to a point were formed by dark triangular centres of scales, giving the snakes of this morph a reticulated appearance. In live photos the pale edges of scales were orange, but this has likely faded from preserved specimens.

Fig. 6.

Barplots showing colour pattern (defined in methods) distribution in five codistributed Aipysurus lineages

Fig. 7.

Representatives of light (left) and dark (right) morphs for each species: A) Aipysurus apraefrontalis, Exmouth Gulf (Ruchira Somaweera); B) Aipysurus apraefrontalis, Ashmore Reef [50]; C) Aipysurus duboisii, Pilbara (James Nankivell; D) Aipysurus duboisii, Ashmore Reef [50]; E) Aipysurus foliosquama, Pilbara (James Nankivell); F) Aipysurus foliosquama, Ashmore Reef; G) Aipysurus tenuis, Pilbara (James Nankivell); H) Aipysurus fuscus, Heywood Shoals (Vinay Udyawer); I) Aipysurus laevis, Pilbara (RV Naturaliste); J) Aipysurus laevis, Gulf of Carpentaria (Mahree Dee-White)

A total of 7 from 25 (28%) Aipysurus duboisii examined from the Timor Sea were melanistic, which were typically larger adult individuals. In preservative, these specimens were dark brown but photos from live individuals show these were likely jet black in life, with some indistinct pale scale margins in some cases. The intermediate morph made up the majority of the remaining Timor Sea specimens (17, 68%) and typically had a dark brown ground colour with thin white bands on the lower half of the flank before tapering to a point. Most of these specimens were juveniles and smaller individuals. WA Coast Aipysurus duboisii were largely of the light morph (7 of 9, 78%), which consisted of a cream ground colour with black triangles centred on individual scales forming bands that tapered finer down the flanks. The two intermediate individuals had a similar colour pattern to the light morph however the black bands were more extensive. In contrast to the Timor Sea populations, the larger individuals tended to be lighter and juveniles darker.

A total of 42 A. foliosquama specimens were examined for colour pattern, of which 36 were from the Timor Sea and only 6 were from the WA Coast population. In the Timor Sea population, 23 were melanistic and had no discernible pattern, with a dark brown ground colour in preservative. The remainder belonged to the intermediate morph, which had a simple banded pattern with a dark brown ground colour and brown bands, nearly identical to the co-occuring A. apraefrontalis morph. The WA Coast population of A. foliosquama was evenly split between the intermediate and light morph; the light morph was banded like the intermediate morph but with a light red or orange ground colour. No specimens from Shark Bay were available, but low-resolution photos from commercial trawl fishers suggest that they largely belong to the intermediate morph (Nankivell, pers obs).

A total of 39 specimens of A. fuscus and A. tenuis were examined for colour pattern, of which 29 were A. fuscus from the Timor Sea and 10 were A. tenuis from the WA Coast. All A. fuscus specimens examined belonged to the melanistic morph and were uniform dark brown with no discernable pattern. However, some photos of live specimens show a faint pattern, but this was not visible in preserved specimens examined, most of which were from Ashmore Reef. Most A. tenuis were light morphs, with a cream ground colour, a brown head, and a varying degree of black scales predominantly on the dorsum. In some A. tenuis, the black scales formed tapering bands and in two individuals the body was predominantly black with thick cream bands.

A total of 29 A. laevis specimens were examined from Timor Sea (8) and the WA Coast (21). All the preserved specimens examined from the Timor Sea were melanistic with no discernible pattern; however, live photos of this population show that it is typically counter shaded in life and it is possible that the lack of pattern may be an artefact of preservation. The majority of A. laevis from the WA Coast were assigned to the light morph (20 of 21), with large individuals having immaculate cream body colour and striking dark brown heads. In smaller individuals, some dorsal scales were dark brown and in juveniles these formed stripes along the body.

Discussion

Our study reveals common patterns of biogeographic and adaptive evolution in four of five co-distributed Aipysurus lineages. In the discussion below, we use these results to delimit units for conservation management and identify the geographic and ecological mechanisms by which these distinct populations may have arisen.

Identifying phylogenetically and ecologically significant units for conservation

Aipysurus apraefrontalis, A. duboisii, A. foliosquama, and the A. fuscus-A. tenuis complex all show convergent morphological variation in genetically differentiated population pairs from coastal versus offshore habitats. The SNP phylogeny was time-scaled using a secondary calibration; hence, our splitting times are very approximate. Nonetheless, minimum divergence estimates of ~500,000 years, together with the presence of distinct genetic clusters in Bayesian and ordination analyses, indicate restricted gene flow between disjunct coastal and offshore populations. These genetic divisions are mirrored by quantitative differences in morphology. Coastal populations in all four lineages have longer bodies, higher vertebral counts, and lighter colour patterns compared to offshore-reef populations. The adaptive significance of these trait changes is discussed below; however, all are ecologically relevant and their convergence in four population and species pairs is consistent with divergent local adaptation in coastal versus offshore-reef Aipysurus.

The presence of ecologically and genetically distinct populations of sea snakes has important implications for their conservation management. Existing occurrence records show that offshore populations of all four species and species complexes occupy smaller and much more fragmented distributions compared to conspecific coastal populations. Many of the remote offshore reefs and shoals occupied by Aipysurus are within marine protected areas. However, unexplained population declines at some of these protected reefs have led to EPBC Act 1999-listing of A. fuscus as endangered, and A. apraefrontalis and A. foliosquama as critically endangered. Surveys at Ashmore prior to 2013 indicated population declines of Aipysurus apraefrontalis and A. foliosquama of at least 90% over 15 years (three generation lengths), and dedicated searches of both reefs failed to sight either species in 2005, 2007, 2012 and 2013 [3–5, 51]). Speed et al. [50] reported a single individual A. apraefrontalis at 60-100 m at Ashmore Reef, indicating that the species might be retained in (or have migrated to) deeper habitats than those surveyed during previous expeditions. Samples from the Timor Sea were obtained before these drastic declines were known, Sampling from after this collapse may provide new perspectives on the drivers of this event. Further fieldwork is required in the Timor Sea to better understand the current vulnerability of the Timor Sea Aipysurus populations. Coastal Aipysurus populations may be larger and more continuous than their offshore counterparts given that they occupy extensive habitats along Western Australia’s shallow continental shelf. However, Aipysurus are a significant component of the bycatch of multiple coastal trawl fisheries [52, 53] and may also be vulnerable to habitat loss as a result of major infrastructure developments in this region. Fine-scale genetic and habitat analyses within coastal and offshore populations are needed to quantify connectivity, and assess the scale of threats to Aipysurus in these regions.

Our results suggest that coastal and offshore populations of Aipysurus apraefrontalis, A. duboisii, A. foliosquama, and A. fuscus–A. tenuis may have undergone divergent adaptive evolution during a period of historical isolation. A. fuscus and A. tenuis are already recognised as separate species based on non-overlapping patterns of variation in vertebral counts [8, 54]; and supported here). We suggest retaining the current taxonomy for A. duboisii, A. apraefrontalis and A. foliosquama, but recommend that the coastal and offshore-reef populations of these species be recognised as Evolutionarily Significant Units (sensu [55; 56]). Separate management of these functionally discrete units is likely to be the surest strategy for maximising their evolutionary potential to respond to future change. These sea snake ESUs should also be considered in the planning of regional protected areas networks for the preservation of functionally distinct communities.

The fifth co-distributed lineage, Aipysurus laevis, presents substantially different genetic and morphological boundaries compared to the patterns shared by its congeners. Populations of this species from the WA Coast and offshore-reefs form a single cluster in the SNP analyses, which contrast with mitochondrial markers, which show clear separation between WA Coast and Timor Sea populations. Little evidence was found supporting the specific status of A. pooleorum, but this was not the primary focus of the study, and we suggest further molecular and morphological sampling before making taxonomic changes. This lack of nuclear population structure might be explained by contemporary migration, the male-biased dispersal characteristic of this species [57], or perhaps the influence of historical introgression between A. laevis and A. fuscus in the Timor Sea [9]. Morphological variation in A. laevis does not follow geographic boundaries as closely as the other codistributed Aipysurus species, although individuals on the WA Coast tend to be larger and paler, but without corresponding increases in vertebral counts or fineness.

Ecological and geographic mechanisms of lineage formation

Similar levels of divergence between coastal and offshore lineages are indicative of shared responses to geological and climatic history in four co-distributed Aipysurus species. Our approximate date estimates, from 950k years ago in Aipysurus duboisii to 500k years ago in A. fuscus-tenuis, coincide with a period of numerous cycles of sea level change in the region. During peak glacial periods much of the North West shelf was exposed and the shallow water habitats used by sea snakes would have been less extensive. The barrier between the Canning and Kimberley bioregions likely coincided with a narrow constriction in the shelf where there was little shallow water habitat before the drop off from the continental slope (reviewed in [13]). The Fitzroy River formed an estuary at this narrow point, and it may have been a combination of extensive freshwater outflows, large tidal regime and a lack of shallow water habitats that maintained this biogeographic barrier. The continuing existence of this barrier through high sea level interglacials in four of the five species warrants further investigation based on more complete geographic sampling across this putative boundary.

Restricted gene flow likely facilitated the morphological divergence of coastal versus offshore-reef populations. Convergent shifts in body length, vertebral count and colour pattern characters in four co-distributed lineages provide compelling evidence of local adaptation. Selection pressures relating to diet and habitat are dominant drivers of body shape variation in snakes. Dietary specialisation in particular has been linked to body length, girth, and vertebral counts in Hydrophis sea snakes [48, 58]. Hydrophis species with thicker, shorter bodies and low-moderate vertebral counts typically have generalist diets or specialise on toxic-spinous prey. Those using mostly elongate (especially eel) prey have longer bodies and higher vertebral counts (but to greater extremes than seen in coastal Aipysurus; e.g. ~280 vertebrae in H. kingii versus ~190 in A. tenuis). It is possible that differences in the diets of coastal and offshore Aipsyurus have contributed to their body shape differences to some degree. However, almost all available diet records for Aipysurus indicate a generalist diet for most species with a wide variety of fish taken (mostly fusiform and goby-like), suggesting a minor role for diet in driving body shape changes in most of these species. One notable exception might be the coastal ESU of A. apraefrontalis, which has a significantly smaller head than the offshore-reef ESU and for which all three available diet records are of burrowing eels (Nankivell pers. obs.). Hydrodynamic adaptations to open coastal versus structured offshore-reef habitats might also explain the morphological differences between these lineages. Sea snakes are anguilliform (eel-like) swimmers that propel themselves through water using lateral deformation waves that move backwards along the full length of their bodies [22]. The kinematics of anguilliform locomotion has been studied extensively in teleosts. In elongate, finless species, there are strong causal relationships between body size and shape and swimming performance. Body length is a major determinant of critical swimming capacity [59, 60], and fineness ratio (body length relative to cross-sectional width) is related to endurance capacity, with greater fineness providing higher sustained swimming efficiency via reduced drag [61]. Comparative studies in fish and cetaceans have linked optimised fineness ratios with the locomotory requirements of different habitat types; species that occupy open habitats or rely on patchily dispersed resources have higher fineness ratios than those in structurally complex reef or coastal habitats [62]. The coastal habitats of Aipysurus are relatively simple and open, consisting largely of unstructured mud and sand flats. In contrast, their offshore-reef habitats in the Timor Sea are structurally complex, with thick coral growth. Coastal populations of all species had longer bodies compared to offshore populations, but in some species (A. duboisii and A. fuscus/tenuis) this was accompanied by higher fineness, while in others (A. apraefrontalis, foliosquama and A. laevis) body length increased proportionally with hindbody circumference. It is reasonable to expect that coastal sea snakes experience selection for increased endurance-swimming performance, or critical-swimming performance, or both. Telemetry studies are beginning to shed light on individual movements in sea snakes and these may provide greater insight into their activity patterns and fine-scale habitat use [63–65]. While there must be many functional demands on snake body shape, we suggest that selection on locomotory efficiency is a major driver of change in Aipysurus. This hypothesis is consistent with the parallel patterns of diet, habitat and body shape evolution reported for Emydocephalus, which is the sister lineage to Aipysurus [66]. The two Emydocephalus share adaptations to a highly specialist diet of fish eggs, but the coastal species, E. orarius, is longer bodied and has higher ventral (and vertebral) counts than its offshore sister species, E. annulatus. The deep divergence (~ two million years) separating E. orarius and E. annulatus might indicate that coastal and offshore Aipysurus lineages are also on independent evolutionary trajectories and perhaps represent incipient species.

Finally, colour pattern is an important and often highly labile trait that is subject to strong evolutionary pressure in nearly all photic environments. Several sea snake species are regarded as aposematic, although it is also possible that bands provide effective crypsis through flicker fusion in some environments [23]. In our study, coastal populations of Aipysurus were universally paler than those from the Timor Sea reefs, and frequently had cream bodies with some darker markings and bands, as well as dark brown heads. In contrast, Timor Sea Aipysurus were generally dark brown or black in head and body colouration, frequently with no discernible bands, and when present, the bands were often low in contrast to the ground colour. We consider it highly unlikely that this is driven by industrial melanism [26] in this case because the Timor Sea reefs are not currently impacted by elevated trace-element concentrations. We suggest that the darker colour patterns are more likely to provide camouflage to snakes sheltering within reef crevices. Notably all the species in our study are either confirmed or suspected to have phototactic tails, which are thought to facilitate effective sheltering [67]. Coastal Aipysurus are less likely to have access to dark crevices and may instead shelter on the seafloor, where pale sandy colours could offer some degree of crypsis.

Conclusions

Identifying genetically and ecologically unique lineages within species is of key conservation concern. In a threatened genus with codistributed species of Australian sea snakes, we show that common biogeographical and ecological factors have led to the formation of parallel lineages in four species which we designate as Evolutionarily Significant Units (ESUs). Given their differing habitat use and possible responses to threats, treating these ESUs as separate management units is likely to improve conservation outcomes. Further investigation of potential contact zones between these ESUs north of the Dampier Peninsula, as well as the inshore Kimberley, where sea snake collections have been limited, will provide additional insights into their evolutionary history and taxonomic status.

Electronic supplementary material

Below is the link to the electronic supplementary material.

Author contributions

JHN and KLS conceived the research. JHN performed quantitative analyses. Both JHN and KLS interpreted results and drafted the manuscript.

Funding

This project received Australian Government funding through the Our Marine Parks Grant Round 3, 4-H0MJUFC.

Data availability

Data used in this study is available in the supplementary material. Raw sequence data is available via SRA (NCBI): PRJNA1338855.

Declarations

Ethics approval and consent to participate

All animals used in this research were covered in University of Adelaide Animal ethics committee number S-2021-017. Fieldwork was conducted under Western Australian fauna taking licence numbers FO25000393-2 and authorisation to take or disturb threatened species TFA-2425-0240.

No permits are required for specimen-based research at museums.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.