Genetic diversity and drivers of dwarfism in extinct island emu populations

Abstract

Australia's iconic emu (Dromaius novaehollandiae novaehollandiae) is the only living representative of its genus, but fossil evidence and reports from early European explorers suggest that three island forms (at least two of which were dwarfs) became extinct during the nineteenth century. While one of these—the King Island emu—has been found to be conspecific with Australian mainland emus, little is known about how the other two forms—Kangaroo Island and Tasmanian emus—relate to the others, or even the size of Tasmanian emus. We present a comprehensive genetic and morphological analysis of Dromaius diversity, including data from one of the few definitively genuine Tasmanian emu specimens known. Our genetic analyses suggest that all the island populations represent sub-populations of mainland D. novaehollandiae. Further, the size of island emus and those on the mainland appears to scale linearly with island size but not time since isolation, suggesting that island size—and presumably concomitant limitations on resource availability—may be a more important driver of dwarfism in island emus, though its precise contribution to emu dwarfism remains to be confirmed.

1. Introduction

The fossil record in Australia contains the remains of at least nine species of ratite (large flightless palaeognathous birds), with most known from only single or a few bones [1]. Only two species of ratite occupy Australia today (electronic supplementary material, figure S1), the southern cassowary (Casuarius casuarius) and the emu (Dromaius novaehollandiae); however, reports from early European explorers suggest that three distinct ‘dwarf' island forms of the genus Dromaius existed until recently: the King Island emu (D. ater), Kangaroo Island emu (D. baudinianus) and Tasmanian emu (D. n. diemenensis) [2]. While these island forms only became extinct during the nineteenth century [3], their characteristics and evolutionary relationships remain poorly understood. Further, there are conflicting accounts of the Tasmanian emu's size [3–5], making its status as a dwarf and its distinctness from the mainland emu even more difficult to assess. Some accounts from early Europeans suggest that Tasmanian emus were the same size as mainland emus [5], but others report them to be slightly smaller [4]. Conversely, King Island emus and Kangaroo Island emus appear to be easily distinguishable from mainland emus based on their plumage coloration, osteological differences in limb bones and relatively short stature [6–8]. However, a molecular phylogenetic study of King Island emus has shown that they were simply an isolated population of mainland emus, rather than a distinct species, suggesting that their dwarfed phenotype was the result of phenotypic plasticity [9]. It remains unknown whether Kangaroo Island emus and Tasmanian emus were also dwarfed conspecifics of mainland emus or were in fact evolutionarily distinct lineages. Discriminating between these two competing hypotheses has important implications for understanding the speed and underlying biological mechanisms of extreme morphological changes like island dwarfism.

Island dwarfism in large animals is a common phenomenon, and it is generally accepted that reduced resource availability drives a decrease in body size, in contrast to increases in body size of small animal taxa resulting from predator release [10–14]. However, there is considerable uncertainty about adaptive mechanisms and environmental drivers that may have formed similar dwarfed emu phenotypes on distant islands (greater than 800 km apart). The islands in question range in size from 62 400 km² (Tasmania), to 4400 km² (Kangaroo Island) and 1100 km² (King Island), providing very different levels of resource availability and inter-specific competition. In addition, two large marsupial carnivores—the Tasmanian devil and Tasmanian tiger—are known to have inhabited Tasmania [15] (with the latter also on Kangaroo Island; G Prideaux 2012, personal communication), while King Island was free of large terrestrial predators [16]. Further, the time elapsed since the isolation of island populations has also been suggested to play a role in morphological size differences [17], since dwarfism likely takes considerable time to evolve. Rising sea levels isolated the islands at different times: Kangaroo Island (isolated 10 000 years ago (ya) from the mainland), King Island (isolated 12 000 ya from Tasmania) and Tasmania (isolated 14 000 ya from the mainland) [18].

To examine these issues, we generated 350 bp of the highly variable mitochondrial control region from current mainland emus as well as extinct populations of King Island emu, Kangaroo Island emu and Tasmanian emu. We also collected femur and tarsometatarsus measurements (as a proxy for body size) to examine morphological differences between populations and correlations between emu size and island characteristics (i.e. area and time since isolation).

2. Material and methods

(a). Samples

A total of 29 ancient emu bone, eggshell and feather samples were obtained from museum collections, along with 82 historical or modern mainland emu samples, for DNA analysis (electronic supplementary material, table S1 and figure S2). Based on radiocarbon dates, palaeontological context and molecular preservation, we assume that all of our ancient island emu specimens post-date the isolation of the islands (see electronic supplementary material, table S1). Our 111 new samples were combined with four published King Island emu and 16 mainland emu samples (Heupink et al. [9]), resulting in a full dataset of 131 samples.

(b). Ancient DNA extraction, amplification and sequencing

Samples were extracted, amplified and sequenced in specialist anciecnt DNA laboratories at the Australian Centre for Ancient DNA (ACAD) in Adelaide, South Australia, according to a range of strict protocols, including numerous controls [19] (see electronic supplementary material). The most variable 350 bp section of the mitochondrial control region (mtDNA CR) was amplified and sequenced using primers in two multiplex reactions (see electronic supplementary material).

(c). Morphometric analyses

A morphometric analysis of femora and tarsometatarsi was performed on Kangaroo Island, King Island and mainland emu bones. As there are few confirmed limb bones of the Tasmanian emu in existence, only one individual could be included in the morphological comparison (though due to missing measurements these limb bones could not be included in the principal components analysis (PCA)). The size and shape of the two leg bones as overall length, proximal width, proximal depth, distal width, distal depth, shaft width and shaft depth were measured.

3. Results and discussion

We successfully extracted, amplified and sequenced 350 bp of mtDNA CR for 111 emu samples (electronic supplementary material, figure S2 and table S1). When added to previous data [9] the 131 aligned sequences contained 12 variable sites defining 17 haplotypes. The network (figure 1) shows that haplotypes are largely shared between the mainland and the islands: mainland haplotype H3 is shared with Kangaroo Island, H6 is shared with King Island, H7 is shared with Tasmania, and H2 is shared with both King Island and Tasmania. In addition, unique haplotypes were observed on Kangaroo Island (H4) and King Island (H5). Shared haplotypes were also observed between contemporary and ancient mainland emus (H3 and H7 in figure 1). The haplotype network includes all four putative emu taxa and encompasses mainland emus from across mesic and semi-arid Australia (including samples from either side of the Nullarbor Plains, a common biogeographic barrier in southern Australia, and from samples as old as 17.6 kya), providing no evidence that the island emu populations were distinct (figure 1 and electronic supplementary material, S3). Instead, haplotypes from putative island taxa either are a subset of the sampled mainland diversity or only differ by a single substitution. Given this genetic homogeneity, we suggest that all recent emu taxa should be considered synonymous with D. novaehollandiae.

Figure 1.

Genetic haplotype network with the numbers on the branches referring to the relative nucleotide positions of mutations compared with reference sequence AF338711. Asterisks indicate ancient samples that are directly dated or derived from dated deposits.

The limb morphology ANOVA showed that there were significant differences between populations for each measurement (electronic supplementary material, table S2). The relative sizes of the mainland emu and the three island taxa were linear, with the mainland emu largest, the single Tasmanian emu and Kangaroo Island emu intermediate in size, and King Island emu smallest, with the relationship between the latter two forms being similar to Parker's results [8] (electronic supplementary material, table S3; figures S4–S7, S10 and S11). Our results, therefore, support Colonel Legge's [20] suggestion that Tasmanian emus were slightly smaller than mainland emus but larger than Kangaroo Island emus and ‘much in excess' of King Island emus. When the first principal component (PC) of femora and tarsometatarsi variation was plotted against natural log of island size (figure 2 and electronic supplementary material S12), and time since isolation (electronic supplementary material, figures S13 and S14), a linear relationship was observed for each comparison. By extrapolating the position of the Tasmanian emu in these plots, our estimate of the approximate overall body size of the mainland and island emus (i.e. where they fall on the first PC) could be used to examine the two major island characteristics likely involved in the degree of dwarfism (i.e. island size and time since isolation). The Tasmanian emu is expected to fall between the mainland emu and Kangaroo Island emu in body size by extrapolation from island size (figure 2 and electronic supplementary material, S12), which is consistent with available measurements (electronic supplementary material, table S3 and figures S4–S7) and recorded observations [4,5]. Conversely, a relationship between time since isolation and body size would predict that the Tasmanian emu should be smaller than the King Island emu, which is inconsistent with available measurements (electronic supplementary material, figures S13 and S14) and recorded observations [4,5]. Therefore, we suggest that island size is the stronger driver of island dwarfism in emus, with smaller islands perhaps limiting resource availability. Further testing this hypothesis would require comprehensively radiocarbon dating specimens from the dwarf island populations to confirm that they indeed post-date the isolation of the islands, as pre-isolation ages for the dwarf emus would suggest alternative drivers of dwarfism.

Figure 2.

Plot of PC1 of femur bone measurements from mainland emus, Kangaroo Island emus, and King Island emus versus natural log of island size. The regression line with formula is shown. Pearson's correlation coefficient and the p-value testing for a linear relationship are noted in the bottom left. Tasmania's position on the regression line is estimated from the actual island size, as the single Tasmanian emu sample had missing data and, therefore, could not be included in the PCA. The tarsometatarsi plot is qualitatively similar (electronic supplementary material, figure S12). (Online version in colour.)

Reduction in body size via changing growth rates and other life-history traits is a key mechanism for optimizing the reduced resources found on isolated islands [10,21–23]. The lack of genetic differentiation between island emus suggests that dwarfism evolved quickly and independently in each population, in the thousands of years since their geographical isolation. Consequently, these recently extinct island emu populations may represent an important natural laboratory for studying rapid and extreme adaptation to changing environments, as large flightless birds provide a unique contrast to island endemic mammals and reptiles.

Ethics

All samples were either sampled from museum collections or collected after death through other means (i.e. roadkill). All sample ID numbers are provided in the electronic supplementary information.

Data accessibility

Data are available in the electronic supporting information and on Dryad (http://dx.doi.org/10.5061/dryad.p7c6p) [24].

Authors' contributions

A.C. designed the study; V.A.T., K.J.M., A.C., J.D. and R.E. generated the data; V.A.T. analysed the data; all authors helped with interpretation of results and writing of the manuscript. In addition, all authors have given their approval for the publication of this manuscript, and have agreed to be accountable for the accuracy and integrity of the work.

Funding

This study was supported by Australian Research Council grants to A.C.

Competing interests

The authors have no competing interests.