Genomic data show little geographical structure across the naturally fragmented range of the purple-gaped honeyeater

Keywords: purple-gaped honeyeater, honeyeater, phylogeography, Nullarbor Barrier, Eyrean Barrier, southern Australia, Lichenostomus.

Interest in the roles of biogeographical barriers in speciation and differentiation arises from several sources. These include whether present-day distributions of species and recognised subspecies correlate with and can be explained by biogeographical barriers. One approach to address these interests is to test for molecular phylogeographic structure within a species and link it to barriers. These approaches underpin the ever-burgeoning phylogeographic literature on the evolution of terrestrial and marine species. Examples from across southern Australia, this paper’s area of interest, are typical.

Two biogeographical barriers of interest in many southern Australian phylogeographic studies are the Nullarbor and Eyrean Barriers (Fig. 1). They have affected different species at different times (Crisp and Cook 2007) and according to different ecologies (e.g. diet: granivory, insectivory, nectarivory; habitat: mesic or xeric: Dolman and Joseph 2012). Some species with geographical ranges naturally fragmented by these two barriers have no subspecies and no discernible phylogeographic structure (e.g. tawny-crowned honeyeater, Gliciphila melanops: Dolman and Joseph 2012). Others are more continuously distributed across all of southern Australia. They show contrasting taxonomic and phylogeographic patterns, indicating a range of temporal and spatial effects of these two barriers (cf. Alpers et al. 2016 in wombats; Donnellan et al. 2009 in birds). More examples from the growing body of literature relevant to these questions in southern Australia now include work on reptiles (Edwards et al. 2015; Ansari et al. 2019), mammals (Cooper et al. 2000, 2018; Pestell et al. 2008; Neaves et al. 2012), fish (Donnellan et al. 2015), invertebrates (Pons et al. 2006; Rix et al. 2017) and plants (Crisp and Cook 2007); birds have been especially studied (Joseph and Wilke 2006; Dolman and Joseph 2015; McElroy et al. 2020; Norman et al. 2014).

Fig. 1.  Principle coordinates analyses of the SNP data from the nuclear genome of the two Lichenostomus species. Analyses with data from (a) L. cratitius (illustrated), the species of interest here, and L. melanops, and (b) L. cratitius only. EB and NB schematically indicate locations of the Eyrean and Nullarbor Barriers, respectively, and localities mentioned in the text are also shown (Eucla is highlighted by a star). Sample points circled in the top right quadrant of (a) show the slightest of separation between eastern and western samples on PCoA1 that was not supported in (b). See text for discussion.

Here we contribute a phylogeographic study of another southern Australian bird species, the purple-gaped honeyeater (Lichenostomus cratitius) (Fig. 1). The species occurs in semiarid habitats across southern Australia, specifically mallee woodlands, tall heath and associated low eucalypt woodlands (Menkhorst et al. 2017). Its range is naturally fragmented by the Nullarbor and Eyrean Barriers. The Kangaroo Island population (Fig. 1) is isolated by sea. The species is almost phenotypically invariant. Two weakly differentiated subspecies are currently recognised (Schodde and Mason 1999): L. c. cratitius on Kangaroo Island and L. c. occidentalis for all other populations. L. c. occidentalis is thus isolated into three groups of populations by the Nullarbor and Eyrean Barriers. Isolation by the Nullarbor Barrier may well have been partial. Pockets of suitable habitat do occur along otherwise unsuitable coastal cliffs where the Nullarbor Plain reaches the sea. Their connectivity has likely been augmented by anthropogenically provided habitats such as gardens around Eucla (Fig. 1). Here, we ask whether there is cryptic genetic differentiation in L. cratitius and, if so, can it be related to the Nullarbor and Eyrean Barriers, or indeed a sea barrier, and to currently recognised subspecies? To address these questions, we have gathered mitochondrial DNA (mtDNA) and single nucleotide polymorphism (SNP) data from the nuclear genome.

We wished to sample the purple-gaped honeyeater (Lichenostomus cratitius) evenly across its range. Fig. 1 and Table S1 (Supplementary Material) show locations and other details of the specimens studied. We lacked samples from what may be an isolated population at the species’ easternmost extremity well east of the Eyrean Barrier (Fig. 1) and east of the Bassian Volcanic Barrier (sensu Schodde and Mason 1999). That population aside, the largest gap from which no tissue samples were available before our study was ~1050 km, comprising 710 and 340 km west and east, respectively of Eucla, Western Australia (Fig. 1). We collected 16 specimens from the former region west of Eucla through field work in 2017. We were unable to obtain any further samples from east of Eucla. Nonetheless, we are confident that Fig. 1 and Table S1 show reasonably even sampling of the species across its range.

DNA was extracted from liver samples of L. cratitius (n = 34) and its sister species, the yellow-tufted honeyeater (L. melanops) (n = 3) (Nyári and Joseph 2011; Marki et al. 2017; specimens listed in Table S1) using the Qiagen Puregene^® Tissue Kit following the manufacturer’s protocols. DNA was provided to Diversity Arrays Technology (DArT) with a concentration of 30 ng µL^–1. The DArTseq™ genotype-by-sequencing approach uses a combination of DArT complexity reduction and high-throughput sequencing (Kilian et al. 2012; Courtois et al. 2013; Cruz et al. 2013) to simultaneously identify and genotype SNP markers in the absence of a reference genome. A combination of two enzymes were used for complexity reduction (PstI–HpaII), before addition of custom adapters (Kilian et al. 2012) to restriction site overhangs. Fragments were amplified using primers complementary to the adapters. These also incorporated molecular identifier barcode tags, to allow multiplexing of up to 96 samples per sequencing run. PCR conditions consisted of denaturation at 94°C for 1 min; 30 cycles of 94°C for 20 s, 58°C for 30 s and 72°C for 45 s; and a final extension period of 72C for 7 min. PCR products were pooled for sequencing on an Illumina HiSeq 2000 platform using 77 cycles of single end sequencing. Raw sequence reads were filtered and processed using a proprietary DArT analytical pipeline. Poor quality sequences were removed, with more stringent criteria being placed upon the barcode region than the rest of the sequence. Approximately 2 000 000 sequences per barcode were identified and used in marker calling, during which identical sequences were collapsed and filtered before screening to identify variable markers using DArT proprietary SNP and SilicoDArT algorithms (DArTsoft14).

The R package dartR (Gruber et al. 2018) was used for population genomic analyses of the DArTseq SNP data, including filtering data and principal coordinates analysis (PCoA). SNPs were filtered using the callrate function (99.5%) and the RepAvg function (1). No monomorphic loci and no duplicate loci per sequence tag were identified. The proportion of genetic variation that is attributable to between-population differentiation, F_ST, was calculated using the R package StAMPP (Pembleton et al. 2013).

We obtained mtDNA sequences for the ND2 gene of mitochondrial DNA from the 31 L. cratitius (including most of those analysed for SNP data) and three L. melanops samples. We used protocols described in Dolman and Joseph (2015) for DNA extraction and sequencing. Forward and reverse raw sequences were aligned in Geneious 10.2 (Biomatters) and subsequently checked and edited manually. Consensus sequences for all individuals including those of the outgroup L. melanops were aligned in MUSCLE (Edgar 2004). Gamma distribution, percentage of invariant sites and model of sequence evolution were chosen using AIC calculations in MEGA 7.0 (Kumar et al. 2018). Forward and reverse raw sequences were aligned in Geneious 10.2 (Biomatters) and subsequently checked and edited manually. The consensus sequence for all individuals were aligned in MUSCLE (Edgar 2004). The Tamura–Nei (TN93+G) (Tamura and Nei 1993) model was chosen using AIC calculations in MEGA X 10.0.5 (Kumar et al. 2018). Phylogenetic trees were prepared using Garli 2.01 (Zwickl 2006) and visualised in SumTrees (Sukumaran and Holder 2015).

Nodal support under maximum likelihood was assessed with 100 bootstrap pseudoreplicates. We used DnaSP 6.11.01 (Rozas et al. 2017) to calculate Da (net nucleotide divergence), which corrects for shared diversity among the individuals within groups being compared.

We generated 170 551 SNPs. After filtering to remove loci with missing data, we retained 40 437 SNPs for further analyses (data lodged at doi: 10.6084/m9.figshare.12798446). F_ST between L. cratitius and L. melanops were high, as expected (0.49–0.51). Within the focal species, L. cratitius, F_ST was low, at 0.01. A PCoA plot including both species (Fig. 1a) shows that 19.8% of the variation in the data was explained by the first two axes of variation, PCo1 and PCo2, most of this reflecting the separation, as expected, between the species L. cratitius and L. melanops. Still with L. melanops included, very narrow but sharp separation of L. cratitius samples east and west of the Nullarbor Barrier was suggested on PCo1 (Fig. 1a). This particular effect disappeared when the analysis was confined to L. cratitius (Fig. 1b) and a much smaller proportion of the variation within the species, 9.6%, was explained by PCo1 (5.2%) and PCo2 (4.4%). The only structure then apparent on PCo1 was that of all three Kangaroo Island samples being apart from all others. PCo2 similarly showed almost all samples clustered together and mostly with no geographical structure. Two pairs of samples, however – a pair from Eyre Peninsula and a pair from south of Madura at the western edge of the Nullarbor Plain – were each well separated on PCo2. Other samples from both localities fell in the main cluster. Further filtering the data to remove loci where rare alleles have frequencies less than 0.25 yielded essentially identical results (not shown).

Thirty-two ND2 sequences newly acquired in this study have been lodged in GenBank (accession numbers MW310592–MW310623). Of these, 30 were from L. cratitius and three were from L. melanops. Net divergence, Da, between the two species L. cratitius and L. melanops was substantial at 3.31%. Within L. cratitius itself nucleotide diversity was low at 0.82%. So too, net divergence, Da, between eastern and western groups was low at 0.06%. Da between Kangaroo Island and all mainland samples was an order of magnitude higher but still low at 0.2%. Phylogenetic analysis showed two haplogroups, each comprising samples from across the range (Fig. 2). Only one of these haplogroups was strongly supported. It comprised five and two samples from the eastern and western parts of the species’ range, respectively. It was robustly supported by 99% bootstrap support. The other haplogroup was weakly supported at 79%. Only one specimen, ANWC B46696 from near Cummins, Eyre Peninsula, was in both this mtDNA clade and among weakly differentiated samples on PCo2 of the SNP analysis (Fig. 1b).

Fig. 2.  Phylogenetic analysis of ND2 sequences from mtDNA of Lichenostomus cratitius (and the L. melanops outgroup). Nodal bootstrap support values ≥79% are shown. Numbers 1 to 31 link to the listing of specimens in Table S1 (Supplementary Material). The bold horizontal line highlights the separation of the two haplogroups in L. cratitius. All specimens in this analysis are from the Australian National Wildlife Collection (ANWC).

We asked whether the Nullarbor and Eyrean Barriers and a sea barrier have generated genetic differentiation among populations of the purple-gaped honeyeater (Lichenostomus cratitius). We were interested in linking any structure to the species’ very low phenotypic diversity. Based on single nucleotide polymorphism (SNP) and mitochondrial DNA sequences, we conclude that the species has very low genetic diversity and appears almost panmictic. The most strongly supported genetic structure in the SNP data parallels the currently recognised subspecies, L. c. cratitius (Kangaroo Island) and L. c. occidentalis (mainland). Two aspects of our results warrant further comment.

First, closer inspection of the PCoA analysis of SNP data including both species L. melanops and L. cratitius (Fig. 1a) suggested a pattern arguably consistent with early stages of differentiation in the latter. Importantly, this pattern disappeared in the analysis using only data for L. cratitius, the species of interest (Fig. 1b). Still at this geographical scale, robust support for one mtDNA haplogroup is notable (Fig. 2). That haplogroup comprises seven samples (samples 25–31: Table S1), of which five (samples 27–31) are from east of the Eyrean Barrier. The other two samples (samples 25 and 26) are from west of the Nullarbor Barrier but they are among the easternmost such samples. If the Nullarbor Barrier initiated differentiation in mtDNA in this species, that differentiation has probably long since been arrested by gene flow. We note here that the species occurs in domestic gardens at Eucla (Fig. 1; authors’ obs.). Eucla is itself located in what is otherwise a gap in the species’ range (Atlas of Living Australia, http://www.ala.org.au, accessed 4 August 2020). This would at least facilitate recent bidirectional gene flow across the Nullarbor Barrier. Second, the differentiation of the Kangaroo Island population should be interpreted with care. That is, genetic drift in an island population may explain this as well as, if not better than, longer term biogeographical separation implied by treating the population as a separate subspecies.

We conclude that some differentiation may have occurred historically during more arid glacial cycles, which presumably fragmented the distribution across the Nullarbor Barrier. We infer that this has since been largely obliterated, however, by gene flow. Gene flow may also explain the diversity within L. cratitius at 0.82%. It is approximately equal to or potentially lower than what might be expected, given that L. cratitius and L. melanops diverged from a common ancestor ~2.5 million years ago (see Joseph et al. 2014; Marki et al. 2017). The resulting genetic patterns nonetheless correlate with currently recognised subspecies in L. cratitius. Overall, this is very similar to patterns seen in the western pygmy possum (Cercartetus concinnus), the distribution and habitat of which closely mirror those of L. cratitius (Pestell et al. 2008). No effect of the Eyrean Barrier is evident, unlike in many other bird species, some of which have even less of a contemporary geographical separation across that barrier than L. cratitius (examples cited in Introduction). If phenotypic variation recognised by the currently recognised subspecies is itself valid, then it may also include an effect of drift in the Kangaroo Island population. The species seems poised at an interesting stage in its evolutionary history.

The authors declare no conflicts of interest.