Chelodina

The raw data obtained from the overview study comprised the allozyme genotypes from 102 specimens of

Chelodina

representing 45 populations overall. These data are summarized in Table 2 View Table 2 as allele frequencies for each diagnosable taxon (including the morphologically diagnosable C. reimanni plus the various geographical/putative forms of the C. rugosa group). Table 3 View Table 3 gives details of the subgroups that were subjected to further screening in Stage 2 of the analysis. The genotypic data for both the overview and various Stage 2 analyses are not presented due to considerations of space, but are available from the authors upon request.

It was clear from the outset that four individuals in the overview study and one in the Stage 2 analyses had hybrid origins. These specimens were omitted from the species boundary analyses. The analysis of their hybrid origins is presented separately, later in the results.

A matrix of fixed differences between the 45 populations was constructed using the raw allozyme data from the overview study and reduced to a matrix of differences between diagnosable taxa only by the procedure outlined in the Material and Methods. The following taxa were strongly supported as distinctive by this overview analysis: Chelodina expansa (6–20 fixed differences from other taxa, N = 8), C. oblonga (14– 22 fd, N = 7), C. parkeri (6–24 fd, N = 6) and C. steindachneri (4–20 fd, N = 6). Additional specimens of C. expansa and C. steindachneri , the two taxa which were sampled from multiple localities across their entire range, were examined for a suite of diagnostic and polymorphic loci in Stage 2 ( Table 3 View Table 3 ). The genetic profiles of these additional animals confirmed both their provisional identification and the utility of the diagnostic loci employed. Of the two taxa, only C. expansa showed any genetic substructuring across its range. Populations from the Murray–Darling drainage to the west of the Great Dividing Range showed some differentiation from those in coastal eastern Queensland ( Fig. 2 View Figure 2 ). However in the absence of any fixed differences, these two groupings could not be regarded as diagnosable taxa.

Among the remaining taxa, forms currently assigned C. rugosa and C. siebenrocki split into three interim diagnosable taxa that did not correspond with existing species boundaries. These three interim taxa were combined as one subgroup for further detailed analysis using all available specimens ( Table 3 View Table 3 ). Similarly, forms currently assigned to C. novaeguineae , C. reimanni , C. mccordi , C. pritchardi and C. longicollis were combined as a separate subgroup ( Table 3 View Table 3 ).

A summary of the number of fixed differences between taxa that were diagnosable at one or more loci is provided in Table 4 View Table 4 . This table was constructed by integrating the overview analysis and the various subgroup analyses on the assumption that the monomorphic loci from the overview analysis that were not further analysed would have remained largely monomorphic (they certainly would have failed to become in any way diagnostic) had they been subject to greater scrutiny in the subgroup analyses.

Principal co-ordinates analysis (PCoA) applied to the C. rugosa / seibenrocki forms revealed three major groupings ( Fig. 3 View Figure 3 ). Specimens from the Kimberley and Arnhem Land formed one group, corresponding to C. burrungandjii (= Chelodina sp. aff. rugosa (Mann] of Georges & Adams (1992)). A second group comprised lowland forms of C. rugosa from the Northern Territory east to the Macarthur River in the Gulf of Carpentaria. The third group comprised lowland forms of C. rugosa from Cape York (the Type locality), rivers of the eastern Gulf of Carpentaria in Queensland, and C. siebenrocki of New Guinea. These forms are very closely related, differing by only one fixed difference in each instance which, in cases of allopatry, is insufficient to establish a diagnostic taxon by our rule of thumb on sample size and statistical significance. Under this rule, there is insufficient evidence to regard C. rugosa from the Northern Territory , C. rugosa from Queensland, and C. siebenrocki from New Guinea as more than a single diagnosable taxon. The Kimberley and Arnhem Land forms are a single diagnosable taxon which is in sympatry with C. rugosa in the Victoria, Daly and South Alligator rivers of the Northern Territory (only the latter two drainages are included in our sampling of C. rugosa ). One fixed difference and the level of divergence illustrated in Fig. 3 View Figure 3 (equivalent to Rogers D = 0.16, reflecting the additional presence of several near-fixed differences) in sympatry is regarded as sufficient evidence of two discrete taxa, particularly as the two taxa are also morphologically diagnosable ( Thomson et al., 2000) .

PCoA applied to the second subunit of taxa ( C. novaeguineae and related species or forms) revealed three distinctive groupings of populations that corresponded to the recognized taxa C. mccordi , C. pritchardi and C. longicollis ( Fig. 4 View Figure 4 ). Each differed from the other by 6–8 fixed differences and so represent clear diagnosable taxa. Two additional groupings corresponded to the Australian C. novaeguineae vs. a cluster comprising C. reimanni plus New Guinea C. novaeguineae . These groupings differed from C. mccordi , C. pritchardi and C. longicollis by three fixed differences in each instance, but from each other by only one fixed difference. Again, this is insufficient evidence in allopatry to establish separate diagnosable taxa using the allozyme data alone.

HYBRIDIZATION

Two forms of Chelodina , initially thought to be distinct species ( Cann, 1998), appear to be hybrids based on their allozyme profiles. The first is a morphologically distinctive form from the Fitzroy–Dawson drainage on the boundary of the distributions of C. longicollis and C. novaeguineae . It is a large animal, with shell attributes most closely resembling C. novaeguineae and external attributes of the head and jaws most closely resembling C. longicollis (see Cann, 1998, 98–99, for photographs and an historical account of its discovery). The two individuals we obtained of this form appear to be hybrids between C. longicollis and C. novaeguineae ( Table 5 View Table 5 ). Both animals displayed the heterozygous genotypes expected for a C. longicollis X C. novaeguineae F 1 cross at the three loci ( Ada, Est and Hb ) that were unequivocally diagnostic for these two species. In addition, a comparison of their allozyme profiles at the other 42 loci found only two loci displaying any allele not detected in one or both parents, and in both cases (Me-2 a and PepA f) these were unique alleles and present only in the heterozygous state. Independent support for their F 1 hybrid status can be seen from the PCoA of this subgroup, which places them intermediate between the C. longicollis and Australian C. novaeguineae clusters ( Fig. 4 View Figure 4 ).

The second instance of hybridization involves two species that are quite distant phylogenetically, C. novaeguineae and C. rugosa (see Fig. 7 View Figure 7 ), and may one day be placed in separate genera ( Legler, 1981). It involves two specimens collected from the Gilbert River, near Georgetown in Queensland. The first specimen is morphologically distinctive, possessing an admixture of attributes from each of C. novaeguineae and C. rugosa , but it too was initially regarded as a separate species (see Cann, 1998, 96 for photographs). The allozyme profile obtained for this individual clearly demonstrates its status as an F 1 hybrid between C. novaeguineae and C. rugosa , as it displays the heterozygous genotypes expected for an F 1 hybrid at all of the 14 loci that display absolute or effective fixed differences between the parental species, and no genotypic inconsistencies at any other locus ( Table 5 View Table 5 ).

Unlike the previous case of hybridization between C. longicollis X C. novaeguineae , the large number of diagnostic markers available to distinguish C. novaeguineae from C. rugosa allow us to distinguish the allozyme profiles expected for an F 1 hybrid vs. those which would characterize a second generation hybrid (e.g. F 2, F 1 backcrossed with C. novaeguineae , F 1 backcrossed with C. rugosa , etc.). Thus an F 1 hybrid would be heterozygous at all 14 diagnostic markers, whereas a backcross animal will on average be heterozygous for only half of the diagnostic loci, displaying a genotype expected for one of the parental species at the remaining diagnostic loci. A second specimen from the Gilbert River, initially assigned to C. novaeguineae in the field, provides strong evidence of backcrossing between hybrid individuals and the parent species C. novaeguineae . Of the 14 key diagnostic loci, this individual displayed an expected F 1 genotype at seven loci and an expected C. novaeguineae genotype at the remaining seven loci ( Table 5 View Table 5 ). The genotypes at the other 31 loci were also in line with the hypothesis that this individual has resulted from a genetic cross between an F 1 ( C. novaeguineae X C. rugosa ) and C. novaeguineae . The allozyme data therefore provide direct evidence for the partial-fertility of F 1 hybrids, opening up the possibility of introgression between these distantly related species.

The third instance of hybridization involves C. rugosa and C. burrungandjii , occasionally sympatric in the Northern Territory. These two species are genetically similar, displaying a single fixed difference plus major differences in allele frequency at several other loci. PCoA for the C. rugosa subgroup ( Fig. 3 View Figure 3 ) revealed a single individual from the Katherine River which was genetically intermediate between C. rugosa and C. burrungandjii , and the allozyme profile of this individual is consistent with it being an F 1 hybrid between the two species ( Table 5 View Table 5 ). The individual concerned displayed the heterozygous genotype expected for an F 1 at the diagnostic locus Ca1 and the nearly diagnostic loci Acon1 and G6pd, plus no genotypic inconsistencies at any other locus. Nevertheless, given the lack of multiple fixed differences between the parental species, our data cannot rule out alternative scenarios involving second-generation hybridization events for this animal, nor can they determine whether introgression is occurring between these two species. Such an analysis would require additional genetic markers and detailed survey of populations in sympatry, parapatry and allopatry across several drainages.

PHYLOGENETIC RECONSTRUCTION

Figure 5 View Figure 5 shows a neighbour-joining tree applied to the matrix of Roger’s D distances ( Rogers, 1972) of Table 4 View Table 4 . It did not differ topologically from the maximum likelihood Fitch–Margoliash or in any substantial way from trees generated using percentage fixed differences as the distance measure. The tree is shown as an unrooted network because without an outgroup for this study, it was not possible to root the tree using data presented here. However, a number of previous molecular studies that have included Chelodina have been unanimous in placing the root on the branch between C. oblonga and C. expansa ( Georges & Adams, 1992; Seddon et al., 1997; Georges et al., 1998).