Young relicts and old relicts: a novel palaeoendemic vertebrate from the Australian Central Uplands

Climatic change, and in particular aridification, has played a dominant role in shaping Southern Hemisphere biotas since the mid-Neogene. In Australia, ancient and geologically stable ranges within the vast arid zone have functioned as refugia for populations of mesic taxa extirpated from surrounding areas, yet the extent to which relicts may be linked to major aridification events before or after the Pliocene has not been examined in detail. Here we use molecular phylogenetic and morphological data to show that isolated populations of saxicoline geckos in the genus Oedura from the Australian Central Uplands, formerly confounded as a single taxon, actually comprise two divergent species with contrasting histories of isolation. The recently resurrected Oedura cincta has close relatives occurring elsewhere in the Australian arid biomes with estimated divergence dates concentrated in the early Pliocene. A new taxon (described herein) diverged from all extant Oedura much earlier, well before the end of the Miocene. A review of data for Central Uplands endemic vertebrates shows that for most (including Oedura cincta), gene flow with other parts of Australia probably occurred until at least the very late Miocene or Pliocene. There are, however, a small number of palaeoendemic taxa—often ecologically specialized forms—that show evidence of having persisted since earlier intensification of aridity in the late Miocene.

Statistical analyses were performed in R [33]. All mensural variables were log transformed and we inspected all transformed variables for heteroscedasticity, normality and influential observations using boxplots and diagnostic plots in R. Body and head measurements were corrected for body size by regressing against principle component 1 from an initial principle components analysis (PCA) to account for possible non-allometric growth [34]. Analyses of tail morphology were restricted to specimens with original tails and tail measurements were standardized against snout-to-vent length (SVL).
We initially used PCA of the size-corrected body and head length data to examine how distinctive populations were in multivariate space. These analyses did not reduce the number of variables required to explain variation, with the first three components only explaining 38% of variation, and are therefore not considered further.
For subsequent univariate analyses, we used one-way ANOVA to test for overall differences in morphology between four genetically and geographically divergent lineages of Oedura in the AAZ (Oedura fimbria, two mtDNA lineages of Oedura cincta and Oedura sp.). For mensural variables where there was an overall difference (p ≤ 0.05), we ran pairwise t-test comparisons (Bonferroni corrected) between each population.

Biogeographic data
We synthesized distribution records (mainly from the Atlas of Living Australia [35]) and phylogenetic data for all apparently endemic or isolated vertebrate taxa and populations from all three Central Uplands Bioregions (MRB, CRB, MDRB). Where available, we noted the estimated timeframes of divergence between endemics and nearest extant relatives occurring elsewhere in Australia. Given the wide error bars associated with molecular dating, we chose to focus on broad eras: mid-Miocene (more than 10 Ma), late Miocene (approx. 10-5.3 Ma), Pliocene (5.3-2.5 Ma) and Pleistocene (less than 2.5 Ma). We also noted the broad ecological guild of endemic taxa or populations (aquatic (i.e. dependent on water to breed), arboreal, saxicoline (with adaptations for climbing in rocky habitats), scansorial, terrestrial or  Figure 1. Timeframe of diversification and distribution of major clades of Australian velvet geckos (Oedura). Divergence time estimates derived from Bayesian analysis of two nuclear genes and secondary calibrations. Deep nodes that were well supported are highlighted with an asterisk (Bayesian posterior > 95, maximum-likelihood > 70). The two divergent lineages occurring in the Central Upland are respectively highlighted in green (Oedura cincta) and red (Oedura luritja n.sp.). Distributional data downloaded from http://ozcam.ala. org.au/. Summary diagram of Australian environmental history reproduced from [5]. Different pulses of major aridification in the late Miocene and Pliocene are highlighted. fossorial), whether endemics are allopatric from known or putative sister lineages, the ecology of nearest relatives (as above), and the biome of nearest relatives (arid, temperate and monsoonal).

Molecular diversity of Central Uplands velvet geckos
Genetic analyses of Oedura identified two highly divergent and distantly related lineages within Central Uplands: (i) the recognized Oedura cincta (  Analyses based on the combined nuclear and mitochondrial alignment (third codons removed) produced older date estimates for divergence events below the calibration prior (all taxa in the main Australian radiation of diplodactylid geckos, including both lineages of Central Uplands Oedura) (electronic supplementary material, table S2). We interpret this as most probably an artefact of combining mitochondrial data with deep calibration nodes, and focus on date estimates from the nuclear data only. However, we also emphasize that all analyses broadly support the contention that the divergence of isolated populations of Oedura cincta occurred through the Plio-Pleistocene, while in contrast Oedura n. sp. is a much more divergent relict dating back to the early, mid-or even late Miocene.
Dating analyses, using the preferred combination of nuclear data and the correlated lognormal model, estimated very late Miocene to early Pliocene divergence of central populations of Oedura cincta from relatives now living in inland eastern Australia (approx. 4.3 Ma (2.1-6.8)). By contrast, Oedura sp. has no strongly supported sister taxon (although it did tend to associate with an assemblage of taxa from eastern Australia in both mitochondrial and nuclear phylogenies) and is estimated to have diverged from all living congeners well into the Miocene (approx. 11.6 Ma (8.1-15.9)).
Tamura-Nei distances [36] estimated using the ND2 alignment within well-sampled and widespread clades within the MRB were relatively low (Oedura cincta mean 0.007 (0.000-0.017); Oedura n. sp. 0.019 (0.000-0.035)). However, three moderately divergent mitochondrial lineages of Oedura cincta were identified in the greater Central Uplands region (mean inter-clade divergences 0.051-0.061); one widespread throughout the eastern and central MRB, one from a single site in the mid-north of the MRB, and an apparently isolated lineage in the ranges at the southern edge of the Tanami Desert (outside the MRB) (figure 2).   Figure 3. Summary morphological data. (a-d) Boxplots summarizing proportional differences in head depth and tails shape between major lineages of velvet gecko (Oedura) in the Australian arid zone, (e) mid-dorsal scales of adult Oedura cincta, (f ) mid-dorsal scales of Oedura luritja n. sp. Scale bar, 5 mm.

Morphological analyses
Univariate analyses of morphology revealed consistent differences between Oedura n. sp. and other Oedura in the AAZ (figure 3a-d Morphological analyses also revealed additional diagnostic differences in the scalation and coloration of Oedura n. sp., in particular it is characterized by much smaller scales across the body (maximum diameter less than versus greater than 0.5 mm) (figure 3e,f ), and the absence of strong canthal or postorbital striping (figure 4).

Systematics
The synonymy of the Oedura marmorata complex has recently been considered elsewhere [22]. Two names have been applied to populations from the Central Uplands, but are not applicable to new taxon. Oedura greeri Wells & Wellington [37] (holotype: AMS R87677, Mt Doreen) was described without diagnosis and is regarded as a nomen nudum [38].   described based on a specimen from Jessie Gap close to Alice Springs and is currently a synonym of Oedura cincta. Oedura luritja n. sp. (Figures 3f and 4a

Paratypes
All from Northern Territory (n = 11

Etymology
Luritja is a collective name for people speaking several dialects of the Aboriginal Western Desert language. The western parts of the distribution of Oedura luritja (including Watarrka National Park) are in Luritja lands. Luritja is also believed to be derived from the Arrernte word 'Ulerenye' meaning foreigner or stranger, and is therefore further appropriate for such a deeply divergent lineage. Used as a noun in apposition.   greatly depending on body condition. Caudal scalation homogeneous. Fully regrown tails similar length (TL/SVL 0.58-0.78) and width (TW/TL 0.11-0.16) to original tails.

Description
In preservative, base coloration of dorsum dark purplish brown, generally with 5-6 distinct pale buff relatively straight and clearly defined transverse dorsal bands, or occasionally transverse series of blotches. Extensive further light buff flecking usually present between the bands, and elsewhere on the dorsal and lateral surfaces of the head, torso and limbs; light and brown pigmentation on head not forming clear lines or stripes. Two paratypes (AMS R52145 and AMS R52148) have strongly defined bands, lack extensive light flecking and have unusually large light blotches on the head and lateral edges of torso, and a single adult specimen (AMS R52143) lacks any clear dorsal bands. Venter plain light buff, with faint brownish tinge around the terminal lamellae and occasionally fine brown maculations around the infralabials. Original tail with 7-8 indistinct light bands on a brown background. Regrown tails brown with extensive and variable light flecking that does not form a distinct pattern.
In life, the basic pattern and coloration of individuals matches those of preserved specimens, however, the darker regions are purplish during the day, silvery grey at night, and light regions tend to be relatively bright yellow (figure 4a,b). Iris very dark brown (figure 4c).
Juveniles with simpler pattern of clear light transverse bands (and occasionally lateral blotches) on a plain dark-brown background, with light and dark bands becoming increasingly indistinct with size (figure 4d).

Distribution and habitat
Currently known only from the sandstone ranges of the southern MRB, extending from Rainbow Valley Conservation Reserve in the east to Watarrka National Park in the west. Apparently, suitable habitat is continuous between these localities and it presumably occurs throughout the intervening region.
All specimens with associated data have been collected from sandstone and it appears to be moderately common (on suitably warm nights they can be reliably spotlighted). They are generally found in association with deep but tight cracks under exfoliating sandstone, often near exposed vertical faces, and retreat into these if threatened ( [23]; P.J.M. March 2016, personal observation). Field observations indicate they are most active during the summer (daytime maxima above 35°C). In winter, they have been collected under flat exfoliating sandstone on the tops of ridges in Watarrka National Park. Gecko species observed in sympatry are Gehyra moritzi, Gehyra versicolor, Heteronotia binoei and Nephrurus amyae.

Comparisons
Similar in overall proportions to and has been confounded with Oedura cincta (both central and eastern populations) but can be distinguished by its shorter rostral crease (less than 50% divided versus fully divided). Further differs from both Oedura cincta and Oedura fimbria (Western Australia) in head and tail proportions (see Results and electronic supplementary material, table S3), in its smaller body scales (mid-dorsal scales on adults < 0.5 mm wide versus > 0.5 mm wide), in generally single cloacal spur (22 out of 23 specimens examined) (versus up to 3), in generally lacking obvious light canthal stripes, brown postorbital stripes and brown nuchal bands (versus present), and in generally retaining strong and distinctly edged dorsal bands into adulthood (versus much more indistinct or absent) (figure 4).
Differs from Oedura bella and members of the Oedura gemmata-marmorata species complex from northern Australia in possessing a longer tail (TL/SVL 0.65-0.87 versus 0.49-0.65 and 0.53-0.63, respectively) that is also narrower (always narrower than the head versus as wide or wider), and by generally having just one clocal spur (versus 2-3).
Differs from Oedura gracilis (Kimberley region) by its moderately long tail (versus very long (approaching length of body)) and flared lamellae series on fingers and toes 2-5 (versus tapering); and from Oedura filicipoda and Oedura murrumanu in having narrower proximal lamellae on fingers 3-4 (not wider than the apical lamellae versus distinctly wider), and further differs from the former species in having a narrow tail that is not wider than the head and near circular in cross-section (versus wider and very flattened).
Distinguished from the remaining Oedura in eastern Australia (here referred to as the tryoni group) by its dorsal colour pattern consisting of 5-6 distinct to indistinct narrow light bands with poorly defined light flecking (versus wide pale V-shaped transverse bands in Oedura castelnaui, distinct dark-edged ocelli or transverse bands of varying size in Oedura coggeri, Oedura monilis and Oedura tryoni, or two pale bands across the nape and base of tail in Oedura jowalbinna). It also has a less swollen tail than Oedura castelnaui, and is larger (SVL up 99 mm) than Oedura coggeri (70 mm) and Oedura jowalbinna (69 mm).

Biogeography of Central Uplands vertebrates
The ecology and divergence dates of endemic vertebrate taxa (n = 19) and apparently isolated populations (n = 10) in the Central Uplands are summarized in table 1. Most endemic taxa (afforded unique taxonomic status) and isolated populations (currently recognized as conspecific with taxa occurring elsewhere in Australia) are clearly allopatric from extant relatives. A majority of endemics have relatives that occur in arid or seasonally arid biomes and none are closely allied to taxa in mesic forest biomes of eastern or far southwestern Australia. Published median or mean estimates of divergence age for endemic taxa are concentrated around the very late Miocene and continue through the Plio-Pleistocene. Divergence timeframes for the majority of endemic populations are not available, but given putative conspecific status may be assumed to be young. The largest number of endemic taxa are saxicoline (n = 6), but there are also largely terrestrial forms that are closely associated with rocky habitats (4), and a suite of aquatic (3), fossorial (4), scansorial (1) and arboreal taxa (1) that appear to be dependent on isolated microhabitats with reliable water. Endemic populations tend be less outwardly specialized (usually terrestrial) taxa with relatives that occur in habitats often widely distributed across both arid and non-arid environments.

Discussion
The Central Uplands of Australia are considered a hotspot of localized endemic relicts stemming from widespread extinction or range contraction in surrounding regions [18]. In support of this contention, most endemic vertebrates in the Australian Central Uplands have allopatric sister taxa or conspecific populations occurring elsewhere in arid or semi-arid Australia (table 1). The overarching role of refugial dynamics in generating endemism is also emphasized by the lack of evidence for ecological speciation [54] in the Central Uplands (although a gecko occurring in ranges just to the north provides one possible exception [42]). A further emerging theme from recent genetic research on plants, as well as the available data we have compiled here (table 1), is that the majority endemic species and populations are not particularly old, with many showing evidence of genetic exchange with relatives occurring in more peripheral parts of Australia during the Pliocene, and in some cases even more recently [15,21].
By striking contrast, our estimates of the divergence time for Oedura luritja suggest much earlier diversification around the mid-to late Miocene. Even more unusually for Central Uplands endemics, this taxon also has no strongly supported sister lineage, implying both long-term persistence and widespread extinction [55]. This lineage diverged from extant relatives well before the major expansion of Australia's vast sandy deserts through the Plio-Pleistocene [2,5,56], but on a timeframe that is consistent with isolation by an earlier period of intensifying aridity in the late Miocene [16]. While the pattern shown by this taxon is currently unique, phylogenetic and distributional data suggest that a number of other relatively restricted and specialized saxicoline lizard taxa in the Central Uplands also show Miocene divergences and/or a lack of close relatives [13,42]. Thus, while a majority of Central Uplands relicts probably do postdate the Pliocene wet phase, as dated molecular phylogenies accumulate, it seems likely that a smaller number will be shown to have longer histories of isolation (especially in specialized taxa, those with low mobility and/or a high capacity to persist in localized microrefugia, e.g. [13,57]).
Based on our review of the endemic vertebrate fauna in the Central Uplands, the presence of two morphologically and ecologically similar, congeneric vertebrate relicts occurring in the same bioregion is also unusual. Oedura luritja has smaller scales and a generally thinner tail, characteristics respectively linked to higher rates of evaporative water loss [58,59] and a reduced capacity to store resources in arid climates [60,61]. Conversely, this species also has a relatively flat head, a trait often observed in specialized crevice dwelling lizards [62], and its known distribution is entirely restricted to the rocks of a single geological landform (Mereenie sandstones). By contrast, Oedura cincta has a deeper body and larger tail, a comparatively wide distribution across eastern and central Australia and includes populations that use both rocks (predominantly granite, quartzite and limestone) and trees, with genetic data suggesting gene flow across now uninhabited regions during the Plio-Pleistocene (presumably on trees along watercourses where eastern populations still exist). The regional coexistance of these two taxa   could be linked to this variation in specialization and climatic tolerance, coupled with local microhabitat differences in different Central Uplands Ranges. Further physiological and ecological data are required to develop and test this hypothesis further. In the AAZ, it is clear that upland ranges, particularly in the Pilbara region, are characterized by deeper genetic diversity and higher genetic structuring than surrounding regions [63,64]. However, most studies have focused on comparative patterns of genetic diversity in rocky regions, rather than the timing of isolation. Thus the relative contribution of periods of aridification before and after the early Pliocene to shaping patterns of relictual endemism in the Central Uplands, other arid zone refugia in Australia, and elsewhere in the Southern Hemisphere, remains to be systematically addressed. However, molecular phylogenetic analyses of other Australia arid zone taxa, including other diplodactylid gecko genera with Gondwanan affinities [13,65], subterranean diving beetles [66] and salt lake beetles [7] are increasingly suggesting that the initial isolation of at least some arid zone relicts predates the Pliocene, and was potentially concurrent with widespread aridification during the Miocene. Many taxa showing long persistence tend to be ecologically specialized (e.g. fossorial or saxicoline), or otherwise show traits linked with vulnerability to environmental change and extremes (i.e. very small size [13]), but which may also predispose them to localized persistence on areas of stable habitat within an otherwise dynamic biome.

Conclusion
Previous taxonomy of Australian velvet geckos (Oedura) masked the existence of two highly divergent relicts that may have been isolated by different aridification events before and after the Pliocene. Available evidence indicates that the isolation of the majority of endemic vertebrate taxa in the Central Ranges of Australia probably postdates the onset of the Pliocene, although a small number of generally specialized lizard taxa (including the new species we describe herein) show evidence of deeper Miocene divergences. Similar deeply divergent and biogeographically significant, but overlooked, relict lineages continue to be discovered in arid and seasonally arid biomes across several major landmasses [11,12,[67][68][69][70]. In many of these taxa a close association with geological landforms that provide stable and protected microhabitats appears to be one key factor underpinning their persistence at very local scales, over long timescales and through major climatic changes.