Extant and extinct bilby genomes combined with Indigenous knowledge improve conservation of a unique Australian marsupial

Bilbies (family Thylacomyidae) comprise the extant greater bilby (Ninu, Macrotis lagotis) and the extinct lesser bilby (Yallara, Macrotis leucura). They are culturally significant to Indigenous Australians, with many Indigenous names and associated knowledge systems. Historically, Ninu ranged widely across arid and temperate Australia, whereas Yallara occupied sandy deserts. Both declined following European settlement due to predation (cats, foxes), competition (rabbits), and altered fire regimes; Yallara is now extinct. Ninu persists in roughly 20% of its former range and is managed as a metapopulation across zoos, fenced sanctuaries, and islands. Management actions include captive breeding (since 1979) and translocations, but the genetic consequences and representativeness of metapopulation diversity relative to wild populations were unclear. Existing monitoring (tracks, scats, microsatellites) lacks resolution for sex ratios, relatedness, and gene flow. The study aims to: (1) generate a chromosome-level Ninu reference genome and sequence Yallara to explore evolutionary history and unique biology; (2) develop a cost-effective scat SNP genotyping tool to support Indigenous ranger-led monitoring; and (3) assess genetic diversity, inbreeding, and the impact of translocations across the Ninu metapopulation relative to wild populations to inform conservation.

Background research identifies key threats to bilbies, including invasive predators, competitors, and fire regime changes, leading to range contractions and extinction of the lesser bilby. Prior work describes bilby biology (low metabolic rate, arid adaptation, reproductive traits including chorioallantoic and choriovitelline placentas) and boom–bust demography with breeding linked to rainfall and food availability. Conservation practice has used metapopulation management with separate evolutionary units historically, later combined, and multiple translocations to sanctuaries and islands. Genetic monitoring has often relied on reduced-resolution markers (microsatellites) from scats, which inform presence/absence but not fine-scale demography. Advances in marsupial genomics (e.g., koala) demonstrate the utility of high-quality reference genomes for conservation genetics, immunogenomics, and understanding adaptation. The study builds on sequentially Markovian coalescent methods (PSMC/MSMC) to infer demographic history and on gene family evolution analyses to explore adaptation.

Genome sequencing and assembly: A female Ninu provided tissues for DNA/RNA. A hybrid assembly combined PacBio HiFi long reads (~10×), Dovetail Omni-C Hi-C, 10x Genomics linked reads (~57×), and Illumina polishing. Assembly steps included HiFi assembly (PacBio Improved Phased Assembler), purging duplicates (purge_dups), scaffolding with 10x linked reads (ARCS/LINKS), gap filling (PBJelly), polishing (Pilon), contamination and duplication filtering (Diploidocus), Hi-C scaffolding (HiRise), and manual curation using contact maps (pairtools, cooler, hicExplorer, HiGlass, PretextSuite). Assembly completeness assessed with BUSCO (mammalia_odb10), Merqury k-mer completeness, and repeat annotation (RepeatModeler/RepeatMasker). A Yallara genome was constructed from historical museum specimens (male skull, samples 1895–1931) using Illumina short-read sequencing; reads aligned to the Ninu genome (BWA aln) to build a Yallara assembly.

Resequencing and variant calling: 12 Ninu (6 temperate island, 6 semi-arid) resequenced on Illumina NovaSeq (PCR-free libraries; ~24–30× coverage). Four Yallara specimens sequenced (mean ~6×, range 0.73–12.82×). Reads aligned to the Ninu reference (v1.5/v1.9), variants called with Illumina DRAGEN Germline and Joint Genotyping; multi-allelic sites normalized (bcftools), annotated (ANNOVAR). Mitochondrial genomes were assembled from BAMs to infer phylogeny via maximum likelihood and Bayesian approaches; PCA on nuclear SNPs assessed population structure.

Demographic inference and diversity: MSMC2 and PSMC were used to infer effective population size through time for Ninu (selecting the 4 highest-coverage individuals per group) and Yallara (2 individuals), scaled using a Tasmanian devil mutation rate (1.17×10^-8 per site per generation) and 2-year generation time; 20 bootstrap replicates for species-level analyses. Runs of homozygosity (ROH) were identified with PLINK on high-density SNPs (post-filter: ~29.3M SNPs) using recommended parameters; F_ROH computed for thresholds >100 kb, >500 kb, >1 Mb. Observed heterozygosity estimated with VCFtools.

Metapopulation genomics (RRS): 363 Ninu (2011–2022) from 13 managed and wild sites genotyped by DArTseq (PstI–SphI). SNP calling and filtering followed established pipelines. Analyses grouped populations as source, translocated founders, and translocated offspring. Calculated observed/expected heterozygosity, allelic richness (R/hierfstat), inbreeding coefficients F_IS with bootstrapped CIs (diveRsity), mean kinship (COANCESTRY), pairwise F_ST (StAMPP), effective population size (NeEstimator), and structure (fastSTRUCTURE, choosek.py).

Scat SNP panel: Designed a MassARRAY SNP panel of 35 autosomal and 4 sex-linked markers (including Y-linked genes KDM5D and HCFC1 for sexing) from DArTseq-derived SNPs filtered for quality and MAF>0.30 across populations. Genotyped 195 scats collected by Kiwirrkurra Indigenous rangers at two sites (south and north-east) in 2021–2022. Included ~10% repeats for error estimation. Diversity (Ho, He, F_IS), relatedness, and structure assessed; comparisons made to tissue datasets restricted to the same 35 autosomal SNPs.

Comparative genomics and gene family analyses: Generated global transcriptomes (12 female tissues) and a testis transcriptome; aligned reads (HISAT2), assembled transcripts (StringTie, TAMA merge), predicted ORFs (TransDecoder), and annotated genes (GeMoMa using 10 mammalian references). Synteny analyses (GENESPACE) compared Ninu chromosomes with other marsupials and human. Orthogroups inferred (OrthoFinder); dated species tree (MCMCtree in PAML); gene family expansion/contraction evaluated (CAFE). Olfactory receptor gene counts compiled. Male meiosis and sex chromosome pairing examined using read depth analyses across X/PAR regions and immunofluorescence (DAPI, SYCP1, SYCP3) on primary spermatocytes. Immune genes annotated using BLAST/HMMER against known marsupial/vertebrate immune gene sets; MHC organization mapped and phylogenies inferred (MEGA11).

• High-quality Ninu reference genome: 3.66 Gb; 95.6% assigned to 9 nuclear chromosomes plus mitochondrial genome; scaffold N50 343.85 Mb; 0.34% gaps; BUSCO completeness 93.5%; repeat content ~47.9% (L1 LINEs ~20.9% dominant). Chromosome 1 is exceptionally large (934,426,298 bp). Global transcriptome: 39,106 genes and 303,420 isoforms; proteome BUSCO 96.0%.
• Yallara (lesser bilby) genome: 3.50 Gb (assembled from museum specimens), 6,329,012 contigs, 19.74% gaps, BUSCO 75.2%.
• Phylogeny and population structure: Mitochondrial and nuclear data support divergence between Ninu and Yallara; divergence time of greater and lesser bilbies ~3 million years ago. PCA differentiates Yallara vs Ninu (PC1 explains 61.6% variance) and separates temperate vs semi-arid Ninu (PC2 explains 11%).
• Demographic history (MSMC/PSMC): Both species show two major contractions coinciding with global cooling prior to the last glacial period (initial declines ~0.3–8 Myr bp). Ninu expanded ~100–500 kyr bp with possible recent decline (uncertain at SMC limits). No clear separation between semi-arid and temperate Ninu within inference timeframe. Yallara Ne likely underestimated due to low coverage (~6×).
• ROH and heterozygosity: Temperate island Ninu (Thistle Island) show lower heterozygosity and higher F_ROH than semi-arid Ninu, consistent with small founder size (N=21) and ~35–40 generations of isolation. Semi-arid individuals show variable ROH with fewer short ROHs (historic wild inbreeding) and some long ROHs (recent inbreeding). Overall, Ninu have relatively few long ROHs compared with other threatened mammals, potentially reflecting boom–bust demography, shorter generation time, and higher per-year recombination/substitution rates.
• Metapopulation management outcomes (RRS, n=363): Observed heterozygosity across source populations ranged 0.1005–0.1839; mixed translocated founder populations showed higher heterozygosity (0.1470–0.1916), which persisted in offspring at two new sites (0.1888; 0.1913). Significant inbreeding (F_IS) detected in several source populations (Pilbara, Scotia, Thistle Island, ZAA). Allelic richness high in translocated sites, reflecting admixture. Effective population sizes were low with wide CIs; small sample sizes reduce precision in some populations.
• Scat SNP panel and Indigenous-led monitoring: MassARRAY panel (35 autosomal + 4 sex-linked SNPs) successfully genotyped 195 scats from Kiwirrkurra, identifying more Ninu in the traditionally managed southern area (N=16) than in north-east (N=9). Genetic diversity (He) in Kiwirrkurra scats comparable to other wild populations (Pilbara He=0.34, Kimberley He=0.37; Kiwirrkurra combined 2021–2022 Ho≈0.338; He≈0.337). Evidence of connectivity between colonies ~70 km apart (half-sibling or higher relatedness; minimal structure in PCoA). Method enables low-cost, broad-scale monitoring.
• Translocations: Based on genetic recommendations, 225 individuals moved between 2016–2021 to found five new sanctuary populations; offspring data show benefits of genetic mixing. Recommendations include continued genetic management, sourcing from wild populations where appropriate, biobanking, and nationwide scat-based surveys.
• Adaptive and comparative genomics: GWAS between semi-arid and temperate Ninu identified 3,858 consistent SNPs across tests and 339 enriched genes, including those involved in anatomical development (e.g., SYNE1, FMR1), metabolic pathways (BBOX1, ACSBG1), and stress response (GRM7). Male reproductive genes with fixed/private alleles by population (SPEF2, TBC1D21, SYNE1, NME8) suggest differences in spermatogenesis-related loci, though functional implications remain speculative.
• Metabolic/olfactory adaptations: Ninu exhibit lowest standard metabolic rate among marsupials and largest olfactory bulbs; genomics shows expansions in metabolic and development-related gene families (CAFE) and the highest number of annotated olfactory receptor genes among ten species compared (total ORs=1377; OR1D2=980; OR1D5=397).
• Reproduction and immune system: All 115 conserved eutherian chorioallantoic placentation genes are expressed in Ninu uterus (TPM>2). The genome contains abundant LTR retroelements, providing candidates for additional syncytins; peramelemorphian placenta involves syncytial fusion. Immune repertoire annotated with 562 genes across major families (cytokines, TLRs, MHC I/II/III, NK receptors, Ig, TCR). Notably fewer MHC-I and Ig variable genes relative to some marsupials; MHC organization shows interspersed MHC-I/II and unique features (four MHC-I genes translocated to scaffold 1; high sequence similarity among some MHC-I paralogs).
• Chromosome evolution and sex chromosomes: Ninu karyotype shows rearrangements and an XY1Y2 system formed by fusion of X with an autosomal arm. Male read depth delineates ancestral X-specific region (half depth), new X-specific region, and a large PAR (full depth) pairing with Y2; microscopy confirms pairing patterns during meiosis.

The study integrates high-quality genomes, population genomics, and Indigenous-led field data to address key conservation questions for the last surviving Thylacomyidae species. The chromosome-scale Ninu genome and Yallara assembly resolve evolutionary relationships, demographic histories, and reveal lineage-specific adaptations (olfactory expansion, reproductive genes), advancing biological understanding. Genomic analyses demonstrate that targeted admixture and translocations increase heterozygosity and allelic richness in new sanctuary populations, supporting ongoing metapopulation management to maintain adaptive potential. The low prevalence of long ROHs, despite small founding events and captivity, suggests that species-specific life history (boom–bust) and generation time may mitigate inbreeding impacts relative to larger mammals, informing expectations for genetic management. The validated MassARRAY scat panel empowers Indigenous rangers to monitor abundance, connectivity, sex, and diversity at low cost in remote regions, facilitating data-driven predator control and fire management. Evidence of connectivity between Kiwirrkurra colonies ~70 km apart informs landscape-scale planning. Comparative immune and placental gene findings generate hypotheses about peramelemorphian reproduction and immune evolution. Overall, the results directly inform conservation strategies (continued genetic mixing, sourcing wild founders, biobanking, nationwide scat surveillance) and exemplify how co-designed genomics with Indigenous communities can improve outcomes for culturally and ecologically important species.

This work delivers the first chromosome-length reference genome for the greater bilby (Ninu) and a genome for the extinct lesser bilby (Yallara), providing new insights into their evolutionary history, biology, and adaptation. Coupled with metapopulation genomics and a new scat SNP toolkit, the study demonstrates tangible conservation benefits: increased genetic diversity through informed translocations, effective non-invasive monitoring by Indigenous rangers, and an improved understanding of wild population connectivity. The research showcases the holistic value of reference genomes for applied conservation, cultural engagement, and fundamental biology. Future directions include: a nationwide scat-based survey to map distribution and gene flow; expanded sampling of wild populations to refine demographic inferences; functional studies of reproductive genes and mating systems; deeper investigation of the XY1Y2 system and meiotic sex chromosome inactivation; and characterization of placental tissues to test hypotheses regarding syncytins and immune gene repertoire in peramelemorphs.

• Some resequenced Yallara genomes had low coverage (mean ~6×), likely underestimating effective population size and precluding association analyses.
• MSMC/PSMC inferences at recent times are uncertain and sensitive to coverage and parameter limits; coverage differences between temperate and semi-arid Ninu may affect subtle differences observed.
• Several metapopulation estimates (e.g., Ne, F_IS) have wide confidence intervals and are sensitive to sample sizes; results for populations with small N should be interpreted cautiously.
• The lack of placental and pregnant uterine tissues constrains functional validation of hypotheses regarding syncytin usage and the relationship between placentation and immune gene repertoire.
• GWAS used a small number of individuals (n=12) and identified allele frequency differences that require functional validation; fixed/private alleles in testis-expressed genes may reflect drift rather than adaptation.