The conservation value of admixed phenotypes in a critically endangered species complex

Abstract

In today's environmental crisis, conservationists are increasingly confronted with terminally endangered species whose last few surviving populations may be affected by allelic introgression from closely related species. Yet there is a worrying lack of evidence-based recommendations and solutions for this emerging problem. We analyzed genome-wide DNA markers and plumage variability in a critically endangered insular songbird, the Black-winged Myna (BWM, Acridotheres melanopterus). This species is highly threatened by the illegal wildlife trade, with its wild population numbering in the low hundreds, and its continued survival urgently depending on ex-situ breeding. Its three subspecies occur along a geographic gradient of melanism and are variably interpreted as three species. However, our integrative approach revealed that melanism poorly reflects the pattern of limited genomic differentiation across BWM subspecies. We also uncovered allelic introgression into the most melanistic subspecies, tertius, from the all-black congeneric Javan Myna (A. javanicus), which is native to the same islands. Based on our results, we recommend the establishment of three separate breeding programs to maintain subspecific traits that may confer local adaptation, but with the option of occasional cross-breeding between insurance populations in order to boost genetic diversity and increase overall viability prospects of each breeding program. Our results underscore the importance of evidence-based integrative approaches when determining appropriate conservation units. Given the rapid increase of terminally endangered organisms in need of ex-situ conservation, this study provides an important blueprint for similar programs dealing with phenotypically variable species.

Figure 1

Map of Java and Bali showing the distribution of the three Black-winged Myna Acridotheres melanopterus subspecies (melanopterus, tricolor and tertius). Inset shows the location of Java and Bali within the Indonesian Archipelago. Myna drawings are by Yifan Pei. Map modified from https://maps-for-free.com/ (OpenStreetMap contributors) using Adobe Photoshop v.21.2 (https://www.adobe.com/products/photoshop.html).

Figure 2

Subspecific variation within Black-winged Mynas (BWM). Mitochondrial ND2 gene variation of a subset of samples representative of each taxon was visualized using (a) a median-joining haplotype network using a trimmed 902 bp alignment, and (b) a phylogenetic tree based on a maximum likelihood (ML) topology using an untrimmed alignment of 976 bp. Hatchmarks on the network correspond to single nucleotide changes. The four labelled tips on the ML tree are sequences obtained from GenBank. The remaining samples were sequenced for this study and are marked with a symbol connecting them to their position on the haplotype network and STRUCTURE plot (if applicable). The ML tree was rooted using a Common Hill Myna (Gracula religiosa). (c) Bayesian clustering analysis of 73 BWMs (filtered to exclude any first-order kin) in STRUCTURE with 7,229 genome-wide SNPs. (d) Upperparts coloration of four representative BWM individuals, including the two terminal subspecies (melanopterus and tertius) as well as two morphological hybrids (melanopterus-like hybrid and tricolor-like hybrid), relative to our plumage scoring scheme. (e) A color score of 0 corresponds to the lightest-backed individuals (melanopterus) and a score of 24 corresponds to the darkest individuals (tertius). Colored symbols represent four classes of morphological identity assigned to samples according to photos at the bottom. Samples without a morphological identity lacked photos, but can be assumed to be of the pure melanopterus phenotype (see “Results”).

Figure 3

Population subdivision of Black-winged Myna samples by principal component analysis (PCA) for dataset 2 (left; comprising the larger captive dataset with kin filtered out) and dataset 3 (right; comprising the smaller dataset of only captive founders based on pedigree data). The PCA for dataset 2 was based on 7,229 SNPs and explained 13.27% of the total variation observed, whilst the PCA for dataset 3 was based on 7,983 SNPs and explained 16.96% of the variation observed.

Figure 4

Four-taxon ABBA-BABA testing scheme for Acridotheres mynas. A four-taxon pectinate phylogeny shows two possible discordant gene tree patterns, ABBA and BABA, that typically occur in equal proportions under incomplete lineage sorting, with A and B denoting the ancestral and derived allele states respectively. Post-divergence gene flow (introgression) from lineage 3 to 2 generates additional instances of the ABBA pattern in this arrangement, leading to its preponderance over BABA arrangements. Branch tips have been labelled with tested taxa for clarity.