Ancient genomes reveal over two thousand years of dingo population structure

The study addresses long-standing uncertainties about the origins and early population history of Australian dingoes, including how many founding populations existed, their routes of introduction, and their relationship to New Guinea singing dogs (NGSD). Additional questions include when continent-wide dingo population structure arose and the extent to which modern dingoes carry post-Colonial domestic dog ancestry due to hybridization after European settlement. Modern morphological and genetic data suggest two main dingo populations (northwest and southeast), with southeastern dingoes showing affinities to NGSD, but timing and direction of gene flow are unclear. Prior analyses relied on contemporary dingoes, lacking a pre-Colonial genomic baseline, limiting power to detect recent admixture. This study uses ancient, pre-Colonial dingo genomes to clarify population structure timing, relationships with NGSD, and to provide a baseline for assessing modern dingo ‘purity.’

Prior work identified two major dingo lineages with geographically structured mtDNA and Y-chromosome haplotypes (northwest vs southeast), supported by morphology and genome-wide SNPs. Some authors infer multiple introductions to Australia, while others consider post-arrival processes and human impacts. Genetic similarities between southeastern dingoes and NGSD have been noted in mtDNA and nuclear SNPs, though Y haplotypes suggest complex relationships. The extent of post-Colonial hybridization with domestic breeds is debated: early microsatellite assays suggested extensive hybridization in southeastern Australia, while recent genome-wide SNP studies indicate hybrid ancestry is relatively rare. However, all prior studies lacked ancient pre-Colonial dingo genomes, complicating inference and raising risks of circularity. Advances in paleogenomics now enable recovery of ancient DNA from Australian contexts, motivating direct tests using ancient dingo material.

Samples and dating: 42 ancient dingo specimens were sampled from coastal Western Australia, the Nullarbor Plain, and coastal New South Wales (NSW), spanning >3,000 km. AMS radiocarbon dating was performed on 29 specimens (with 5 additional dates from prior work), calibrated in OxCal using SHCal20. Ages for key nuclear genomes ranged from ~400 to 2,746 years B.P. Sequencing and data generation: DNA extractions were conducted in dedicated ancient DNA facilities (Griffith University ARCHE and University of Adelaide ACAD). Libraries were prepared using Meyer & Kircher protocols with partial UDG treatment; some libraries underwent whole-genome hybridization capture using modern dog or K’gari dingo baits; others had mitochondrial capture; remaining libraries and modern K’gari dingoes were shotgun sequenced. Sequencing utilized Illumina platforms (HiSeq 4000, HiSeq X, NextSeq 500) with single- and paired-end reads. Data processing and authentication: Reads were adapter-trimmed and merged; mappings to canFam3.1 used BWA aln or MEM; duplicates removed; read termini soft-clipped to mitigate damage. Authentication included DNA damage profiling (DamageProfiler), contamination checks (ANGSD X/autosome heterozygosity; mitochondrial minor/major allele ratios via Calico), and QC (FastQC, fastp, Qualimap). Endogenous content, mt-to-nuclear ratios, and sex were assessed (endorS.py, MTNucRatioCalculator, Sex.DetERRmine). Reference datasets and SNP ascertainment: Ancient genomes were analyzed with modern dingoes, NGSD, village dogs, worldwide dogs, wolves, and ancient canids (Plassais 2019; Bergström 2020/2022; others). Variant ascertainment followed Bergström: retain transversion polymorphisms variable in coyotes to minimize biases; generate pseudohaploid genotypes via pileupCaller across ~19.1M biallelic SNPs. Analyses:

• Mitochondrial DNA: For samples with sufficient coverage, mtDNA consensus sequences were built and aligned with modern/ancient references. A median-joining haplotype network was constructed (PopART). Time-calibrated phylogeny inferred in BEAST2 with an informed substitution rate prior (from Zhang 2020), Coalescent Constant Size prior, and Optimised Relaxed Clock; three MCMC chains were combined after burn-in.
• Population structure: PCA (SmartPCA) on worldwide canids including ancient/modern dingoes and NGSD. Hierarchical clustering via qpWave P-value distances to define ancient dingo clusters.
• f-statistics: ADMIXTOOLS 2 to compute symmetry tests (f4), admixture tests (f3), and ancestry proportions (f-ratio). Tests compared affinities of modern and ancient canids to ancient Nullarbor vs ancient Curracurrang (coastal NSW) dingoes, and to NGSD.
• Graph modeling: qpGraph models scaffolded on prior best-fitting dog models; tested topologies with/without admixture among ancient SE/NW dingoes and NGSD; model fit compared by likelihood.
• Admixture dating: DATES used to estimate admixture pulse timing between SE ancient dingoes (Curracurrang) and NGSD, testing both directions, at population and individual levels. Data availability: Raw reads deposited in ENA (PRJEB75610), mtDNA consensuses on GenBank (PP812314–PP812321 & PP812323–PP812333), and genotype/supplementary data on Figshare (DOI: 10.25909/25885747).

• Ancient data generated: 16 new ancient dingo mtDNA sequences and nuclear genomic data from nine ancient dingo individuals (≥0.01×), including high-coverage genomes: Nullarbor (16.6× and 2.83×) and Curracurrang, NSW (2.55× and 0.75×). Three modern K’gari dingoes were newly sequenced.
• Early emergence of structure: Ancient mtDNA haplotypes fall within the two known modern dingo lineages: ancient Nullarbor/WA within the modern northwest (NW) lineage; ancient Curracurrang within the southeast (SE) lineage. BEAST TMRCAs: SE cluster (cluster B) 3,424–8,244 B.P. (mean ~6,018); NW cluster (cluster A) 2,856–7,944 B.P. (mean ~5,480).
• Nuclear affinities mirror mtDNA: PCA places ancient dingoes with modern dingoes and NGSD, distinct from worldwide breed dogs. f4 tests show modern K’gari and northern QLD dingoes share excess drift with ancient coastal NSW dingoes (Curracurrang), while Kimberley and Gibson Desert dingoes share excess drift with ancient Nullarbor dingoes. NT and Simpson Desert dingoes are more symmetric.
• No detected mixture of ancient east/west sources in moderns: Admixture f3(Test; Curracurrang 2k, Nullarbor) values were positive for all tested modern dingoes, indicating no clear evidence of modern populations being simple admixtures of the two ancient clusters (caveats include drift, unequal contributions, and short separation times potentially yielding false negatives).
• Stronger affinity between SE ancient dingoes and NGSD: Ancient coastal NSW dingoes (and modern alpine dingoes) show excess allele sharing with NGSD compared to ancient/modern western groups. Ternary plots highlight Curracurrang ancient samples with highest SE affinity and Nullarbor ancient with highest NW affinity.
• Demographic modeling supports admixture: qpGraph models allowing admixture between SE ancient dingoes and NGSD fit better than no-admixture models; direction of gene flow cannot be resolved conclusively.
• Admixture timing: DATES estimates a mean admixture time of ~147.7 ± 57.0 generations, corresponding to 2,456 ± 171 years B.P. (assuming 3 years/generation), consistent across tests and aligning with mtDNA clade coalescence and X-chromosome cross-coalescence estimates.
• Limited post-Colonial dog introgression in modern dingoes: Using ancient genomes as an unadmixed baseline, most of 14 modern dingoes (13 previously microsatellite-defined as ‘pure’) showed no evidence of recent breed dog ancestry. Exceptions: one Gibson Desert and one alpine dingo with detectable but minimal post-Colonial ancestry. Results are consistent whether using Curracurrang 2k or Nullarbor ancient references for f-ratio estimates.
• Phenotypic concordance: The Curracurrang ancient dingo with highest coverage (2019 bp) exhibits reduced stature and dental proportions, paralleling traits of NGSD and southeastern modern dingoes.

The findings demonstrate that the major continent-wide dingo population structure (NW vs SE) was already in place by at least ~2,000 years B.P., indicating long-term local continuity from ancient to modern populations. Nuclear genomic patterns broadly corroborate mtDNA lineage structure and regional affinities, supporting the persistence of differentiated populations despite later human impacts. Excess allele sharing between ancient southeastern dingoes and New Guinea singing dogs indicates a complex relationship involving secondary contact after the initial split between dingo and NGSD ancestors. Graph modeling favors models with admixture between SE dingoes and NGSD, and dating analyses place this event around 2,450 years B.P., though directionality remains unresolved. These results refine interpretations of prior genetic signals connecting SE dingoes and NGSD and suggest limited, likely Holocene human-mediated movements across Sahul’s water gaps. Using ancient pre-Colonial dingoes as a genomic baseline provides a robust test of recent hybridization in modern dingoes: most tested modern individuals show minimal or no detectable post-Colonial domestic dog introgression, challenging earlier reports of pervasive hybridization based on microsatellites alone. This underscores the importance of incorporating ancient genomes into conservation genomics to calibrate ‘purity’ assays and avoid biases due to population structure and marker selection. Overall, the study addresses core questions regarding dingo origins and evolution, clarifying timing and persistence of population structure, the nature of relationships with NGSD, and the genuine extent of modern hybridization, all of which are central to evidence-based conservation and management strategies.

Modern dingoes predominantly descend from ancient dingo lineages that were already regionally structured by ~2,000 years B.P. Ancient genomes establish a pre-Colonial baseline that validates low levels of recent domestic dog introgression across most modern dingoes tested and will aid the refinement of ancestry assays used in conservation. The relationship between southeastern dingoes and New Guinea singing dogs involves secondary contact dated to roughly 2,456 ± 171 B.P., though the direction of gene flow is unresolved. Future research should target ancient canid genomes from northern Australia, Island Southeast Asia (e.g., Indonesia, Borneo, the Philippines), and New Guinea to test alternative migratory and admixture models, resolve directionality of gene flow, and better constrain the number and timing of introductions to Sahul.

• Ascertainment and capture biases: Some ancient coastal NSW genomes were enriched using K’gari dingo baits, potentially biasing affinities toward SE groups, though multiple controls argue against this as the sole cause of observed signals.
• Power and f-statistic caveats: Admixture f3 can yield false negatives under strong drift, unequal source contributions near boundaries, and short intervals between splits and admixture—conditions plausible for dingoes. Directionality of admixture with NGSD cannot be resolved with current data and qpGraph fits are non-unique.
• Coverage and sample distribution: Several ancient genomes are low coverage, and geographic/temporal sampling is sparse (notably a lack of ancient dogs from northern Australia, ISEA, and New Guinea), limiting resolution of routes and number of introductions.
• Environmental constraints: DNA preservation in hot, humid regions is poor, restricting recovery of critical ancient genomes needed to refine demographic models.
• Potential metadata uncertainties in legacy datasets and the possibility of minor lab or field biases, despite negative controls and authentication steps, remain general considerations in paleogenomics.