Domestic dog introgression in Australian dingoes: environmental drivers and evolutionary consequences

36 Introgressive hybridization between wild and domestic animals is a widespread phenomenon 37 with important implications for genetic diversity, local adaptation, and conservation 38 management. The causes and consequences of this process are poorly understood. In Australia, 39 hybridization between dingoes and domestic dogs presents a dual conservation challenge, 40 threatening the genetic integrity of dingoes while allowing potential adaptive introgression. To 41 investigate the environmental drivers of this process, we analyzed high-density SNP array data in 42 390 dingoes and 396 domestic dogs. Dingo populations showed regional genetic structure and 43 were clearly differentiated from domestic dogs. Using local ancestry inference and genome –44 environment association analyses, we found low levels of dog introgression in dingoes from 45 remote areas in Central and Western Australia, and moderate levels in Eastern and Southern 46 populations. Climatic variables (maximum temperature of the warmest month, mean temperature 47 of the driest quarter) and the Human Footprint Index (reflecting density of human populations 48 and environmental modifications) were significant predictors of introgression. We identified four 49 genomic regions with overrepresented dog ancestry, including a large introgressed block on 50 chromosome 27, which contained an olfactory receptor gene showing signatures of positive 51 selection, suggesting adaptive introgression. In addition, a chromosomal inversion previously 52 described in dogs and absent in dingoes was initially identified as an introgressed block. We also 53 detected eight genomic regions nearly free of dog ancestry, suggesting purifying selection 54 against maladaptive variants. Together, these results highlight the complex interplay between 55 introgression, human influence, and local adaptation in dingoes, offering valuable insights for 56 conserving the evolutionary potential of this apex predator in increasingly modified landscapes. 57 58

Canid hybridization, Dingo, Adaptive introgression, Genome -environment 59 associations, Landscape genomics, Local adaptation 60 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint 1. Introduction 61 Hybridization is a natural and widespread phenomenon that has long captivated evolutionary 62 biologists. Early conceptual frameworks provided key insights into its role in speciation, 63 challenging traditional views on how species boundaries are maintained (Dobzhansky, 1937; 64 Mayr, 1942; Stebbins, 1959; Grant and Grant, 1979). This process may involve gene flow 65 between genetically distinct lineages, leading to diverse outcomes such as the extinction or 66 displacement of parental taxa, the fusion of previously divergent taxa, or the formation of new 67 hybrid lineages that may eventually result in speciation (Ellstrand and Elam, 1993; Rieseberg 68 and Wendel, 1993; Mallet, 2007; Abbott et al., 2013; Thomas, 2015; Grant and Grant, 2017). 69 However, introgression, the integration of genetic variation into a recipient population through 70 hybridization followed by backcrossing, constitutes a key mechanism by which alleles move 71 across species boundaries (Aguillon et al., 2022). Advances in genomic technologies have 72 enabled the study of introgression at a genome -wide scale, highlighting its evolutionary 73 relevance and providing deeper insights into its role in shaping biodiversity across diverse 74 taxonomic groups, including fungi (Giraud et al., 2008; Kinnerberg et al., 2023), plants (Mallet, 75 2007; Stull et al., 2023), fish (Seehausen, 2004; Blumer et al., 2024; Kato et al., 2024), birds 76 (Grant and Grant, 2017; Singhal et al., 2021), and mammals (Leonard et al., 2013; Adavoudi and 77 Pilot, 2021; Tensen and Fisher, 2024). 78 Among mammals, introgression between canid species is a widespread phenomenon with 79 significant implications for genetic diversity, ecological dynamics, and evolutionary trajectories 80 (Leonard et al., 2013; Adavoudi and Pilot, 2021). This process may be influenced by 81 anthropogenic activities driving hybridization, especially in ecosystems where wild canids like 82 grey wolves (Canis lupus), coyotes (Canis latrans), and golden jackals (Canis aureus) overlap 83 with domestic dogs (Canis familiaris) (Wheeldon et al., 2013; Pilot et al., 2018; McFarlane and 84 Pemberton, 2019; Stefanović et al., 2024). In regions such as North America and Europe, where 85 free-ranging domestic dogs coexist with wild representatives of the genus Canis, hybridization is 86 well-documented and frequently results in the introgression of dog-specific genetic variation into 87 wild species (Leonard et al., 2013). A notable example is melanism in gray wolves in North 88 America, driven by the beta -defensin gene, a melanocortin pathway gene introduced via 89 historical hybridization with domestic dogs (Anderson et al., 2009). This trait has risen to high 90 frequencies under positive selection in forested habitats, exemplifying adaptive introgression 91 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint (Anderson et al., 2009). However, hybridization and subsequent introgression may pose risks to 92 the genetic and phenotypic distinctiveness of wild canids, challenging conservation efforts aimed 93 at preserving unique ecological roles and lineage integrity (Hohenlohe et al., 2021). 94 Wild canids increasingly occupy habitats highly modified by humans and with access to 95 human food waste (Kuijper et al., 2016). In such habitats, wild canids are likely to shift, at least 96 partially, from hunting wild prey to livestock depredation and/or scavenging anthropogenic food 97 (Newsome et al., 2015a). Because evolutionary divergence in canids may be strongly influenced 98 by differences in diet composition (Pilot et al., 2006), dietary shifts may trigger a contemporary 99 domestication process (Newsome et al., 2017). Hybridization with free -ranging dogs may 100 accelerate this process by enabling wild canids to rapidly acquire adaptations to the niche of 101 human commensal (Pilot et al., 2021). While dog -derived traits may be maladaptive in natural 102 habitats, they can be advantageous in human -dominated landscapes. Thus, introgression may 103 facilitate adaptation to anthropogenic habitats but simultaneously shift the ecological niche, with 104 potentially negative ecosystem-level consequences (vonHoldt et al., 2018). 105 The dingo presents a unique case within canid hybridization systems. Present in Australia 106 for at least 5,000 years (Fillios and Taçon, 2016), dingoes have assumed the role of apex 107 predators, with varying effects on both co-occurring predators and prey (Glen et al., 2007; Letnic 108 and Koch, 2010; Newsome et al., 2015b; Doherty et al., 2019; Castle et al., 2023). The dingo 109 genome displays patterns of natural selection distinct from domestic dogs, reflecting its 110 adaptation to the apex predator niche (Zhang et al., 2020). Dingoes exhibit social and behavioral 111 traits similar to other wild representatives of the genus Canis, including pack structures led by a 112 dominant pair, seasonal breeding and the cooperative hunting of large prey like kangaroos and 113 wallabies (Corbett, 1995; Glen et al., 2007; Pollock et al., 2022). Yet, this ecological function is 114 potentially threatened by extirpation, human -wildlife conflict, and hybridization with domestic 115 dogs (Newsome et al., 2015b). For the latter, the arrival of domestic dogs with European origin 116 created new opportunities for interbreeding, and concerns about the need to preserve the genetic 117 integrity of dingo populations are often raised (Stephens et al., 2015; Cairns et al., 2017, 2020). 118 Genetic studies have highlighted the distinct evolutionary history of dingoes relative to 119 domestic dogs, reflecting dingoes’ ancient origins and genetic uniqueness (Cairns et al., 2017; 120 Cairns et al., 2021; Souilmi et al., 2024). Numerous studies using microsatellites have 121 documented the genetic impact of dingoes interbreeding with dogs, suggesting it may erode 122 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint dingo genetic integrity in geographic regions characterized by high densities of dogs via genetic 123 swamping (Wilton, 2001; Elledge et al., 2006; Glen et al., 2007; Stephens et al., 2015, 2022; 124 Cairns et al., 2021). This situation has led to complex management challenges (Boronyak et al., 125 2023). However, recent SNP -based genomic studies suggest that the extent of contemporary 126 interbreeding between dingoes and dogs has been overstated (Cairns et al., 2023; Weeks et al., 127 2024). Whole genome sequence analysis with an ancient DNA baseline identified 9.7 to 22.5% 128 introgressed European dog ancestry persisting in dingoes from southeast Australia, while 129 minimal dog ancestry was detected in northwest dingoes (Scarsbrook et al., 2025). Consistently, 130 genetic surveys of free -living canines in Australia indicate that domestic dogs and first -131 generation hybrids are rare (<1%), and genome -wide admixture analyses reveal a bimodal 132 ancestry distribution dominated by either pure dingoes and dingo –dog backcrosses or pure 133 domestic dogs, with minimal numbers of first -generation hybrids (Cairns et al., 2023). A similar 134 pattern is observed in the case of wolf -dog hybridization across Eurasia (Pilot et al., 2021; Lobo 135 et al., 2025; Sarabia et al., 2025; Battilani et al., 2025). Additionally, studies on the skull shape 136 and the ecosystem impacts of dingoes in different parts of their range suggest that introgression 137 from domestic dogs have had a limited effect on the dingo’s ecological role (Parr et al., 2016; 138 Crowther et al., 2021). Yet, several key questions remain. First, discrepancies between previous 139 studies highlight uncertainty about the true extent of introgression, with concerns that estimates 140 may be biased by underlying population structure and detection of historical introgression. 141 Second, while human activities such as domestic dog ownership and lethal control have been 142 suggested as potential drivers of increased hybridization, the specific environmental factors 143 influencing admixture rates are not well understood. Finally, the evolutionary consequences of 144 introgression for dingoes, especially the possibility of adaptive introgression, remain largely 145 unexplored. 146 To address these knowledge gaps, we combined landscape genomics, local ancestry 147 inference, and introgression analyses, using DNA samples collected from dingoes and European 148 and Australian domestic dogs. We applied high -resolution local ancestry methods to improve 149 detection accuracy, explicitly accounted for population structure, and explored how 150 environmental factors shape spatial patterns of introgression between dingoes and dogs. We 151 further examined introgressed genomic regions for signatures of selection, providing insights 152 into the evolutionary impact of dog ancestry on dingoes and informing conservation efforts. 153 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint 2. Material and Methods 154 2.1 Sample collection and SNP genotyping 155 Tissue samples were collected from wild dingoes across Australia, primarily opportunistically 156 following lethal management activities or retrieved as roadkill by private citizens. We also 157 collected blood or buccal swab samples from Australian domestic pet dogs, selecting commonly 158 occurring working, herding, and mixed breeds to capture the diversity most likely to interact with 159 regional dingo populations. DNA was extracted from blood, tissue, or buccal samples using 160 Qiagen DNeasy Blood and Tissue kits (Qiagen Sciences, Germantown, USA). Extracted DNA 161 was genotyped at the Ramaciotti Centre for Genomics (University of New South Wales, 162 Randwick, Australia) on the Axiom Canine HD Genotyping array (Thermo Fisher Scientific Inc., 163 Waltham, USA). In addition, we incorporated previously published genotype data from Cairns et 164 al. (2023), generated on the same Axiom Canine HD array, to augment our representation of wild 165 dingo diversity. This research complies with applicable laws on sampling from natural 166 populations and animal experimentation, including the ARRIVE guidelines (Du Sert et al., 167 2020). 168 The combined dataset analyzed in this study comprises genotypic data from 300,761 169 SNPs following rigorous QC and 170,465 SNPs following further LD -pruning in PLINK v1.9 170 (Purcell et al., 2007). Filtering steps included removing individuals with more than 10% of 171 missing data (option --mind 0.1) and excluding markers based on missingness ( --geno 0.1), 172 minor‐allele frequency ( --maf 0.01), and LD ( --indep-pairwise 50 10 0.5). Finally, individuals 173 with relatedness up to second -degree were removed using --king-cutoff 0.0885 option in PLINK 174 v2.0. To ensure balanced representation, we included equal numbers of dingoes and European 175 dogs (both purebred and free-ranging). Our dataset contained 390 dingoes sampled from multiple 176 regions across Australia (Big Desert, Central, East, North, South, West, and captive populations), 177 along with 54 domestic dogs representing Australian breeds (including 33 purebred dogs and 21 178 mixed-breed dogs), 116 European free‐ranging dogs, 226 European purebred dogs (i.e. dog 179 breeds of European origin sampled in the United States), and two F1 dog –dingo hybrids (Table 180 S1). Our sampling strategy for dingoes was population -focused, with multiple individuals 181 collected from specific localities to facilitate analyses of the influence of local environmental 182 factors and human footprint on introgression dynamics. Australian breeds are breeds developed 183 in Australia based on domestic dogs brought from Europe. Because we sampled owned dogs that 184 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint do not range freely and represent breeds whose reproduction is managed by humans, these 185 individuals are unlikely to represent populations that routinely hybridise with dingoes. However, 186 occasional reports exist of owned farm dogs breeding with wild dingoes. Our sampled 187 individuals therefore primarily represent managed lineages of domestic dogs introduced from 188 Europe, rather than free -ranging or feral dog populations directly involved in contemporary 189 hybridisation events, while still capturing the ancestral genetic background shared with many 190 rural Australian dogs. European dogs represent the parental population for the Australian dogs 191 and were included so that the broader European gene pool is represented, to account for potential 192 unsampled dog lineages that were introduced to Australia and interbred with dingoes. European 193 pure-breeds were drawn from Morrill et al. (2022), excluding any breeds of non -European origin 194 (e.g. Afghan Hound, Chow Chow). European free -ranging dogs were sampled from across 195 Eastern Europe and include samples previously used in Spatola et al. (2023). 196 2.2 Ancestry Analysis 197 To investigate how introgression patterns vary across Australia and to identify genomic regions 198 associated with introgression, we employed a combination of global ancestry analyses (which 199 provide an overall estimate of each individual’s ancestral composition across the entire genome ) 200 and local ancestry analyses (which detect chromosome‐level ancestry variation). Local ancestry 201 inference can reveal signatures of older admixture events that have since become pervasive 202 across the genome, enabling the detection of historical introgression that might be overlooked by 203 global methods (Sankararaman et al., 2014 ). By integrating both approaches, we captur ed broad 204 admixture patterns as well as discrete regions of introgression. 205 First, we assessed the genome -wide genetic structure using principal component analysis 206 (PCA) (Patterson et al., 2006) and ADMIXTURE (Alexander and Lange, 2011) to identify 207 distinct populations of dingoes across Australia that may differ in introgression patterns. These 208 analyses were carried out for the entire dataset and for the dingo dataset alone (see 209 Supplementary Material for details). 210 To examine local ancestry and introgression signals at the chromosome -level resolution, 211 we employed two complementary methods: LAMP -LD software v2.4 (Sankararaman et al., 212 2008) and ELAI software (Guan, 2014), both performed on a dataset without LD -pruning 213 (300,761 SNPs) to preserve haplotype information and maximize the resolution of local ancestry 214 tracts. LAMP-LD was selected for its ability to infer local ancestry without the prior designation 215 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint of "pure" reference populations, making it particularly useful for providing initial estimates of 216 local ancestry in datasets with complex population structures. We divided the dataset into 217 chromosome-specific files using PLINK v1.09 (Purcell et al., 2007). Custom bash scripts were 218 used to configure LAMP -LD parameters for each of the 38 autosomal chromosomes (see Data 219 Accessibility section). 220 ELAI was then applied to refine local ancestry estimates, leveraging its flexibility in 221 handling dense SNP data and its ability to model complex population histories. The dataset was 222 divided into reference populations (individuals with high dingo and domestic dog ancestry, 223 respectively, identified based on the LAMP -LD results) and admixed populations, which 224 included admixed dingoes, known hybrids and Australian mixed -breed dogs. Genotype data 225 were then converted from PLINK to the 'bimbam' format required for ELAI input files. ELAI 226 was executed with parameters optimized for local ancestry inference: -mg was set to 10 to 227 specify the maximum generations since admixture, and -C was set to 2 to assume two ancestral 228 populations (dingoes and domestic dogs). The parameter -c was set to 10 to increase the 229 flexibility of the hidden Markov model in capturing complex local patterns of ancestry across 230 haplotypes, which enhances the model's ability to detect fine -scale introgression signals. 231 Additionally, -R was set to 45 to optimize the number of EM iterations and ensure convergence. 232 The results were summarized to visualize the mean proportions of dingo and dog ancestry across 233 the genome using ggplot2 in R (Wickham and Wickham, 2016). 234 Finally, we used the GHap package (Utsunomiya et al., 2016) to investigate the 235 distribution of extended haplotype blocks across the genome, providing complementary 236 information to the inferences obtained with LAMP -LD and ELAI. GHap focuses on identifying 237 and analyzing extended haplotype structures, enabling the detection of broader patterns of 238 recombination, shared ancestry, and structural signals of introgression that SNP -based local 239 ancestry methods may miss. Prior to the analysis, all genotypes were phased using Beagle v5.4 240 (Browning et al., 2021), with the burn -in parameter set to 10 and iterations to 1000, to ensure 241 accurate haplotype reconstruction. We applied GHap to the LD‐unpruned dataset, retaining all 242 loci to fully capture LD -based haplotype structure. An unsupervised GHap analysis was 243 performed using the elbow method to determine the optimal number of clusters (K), which was 244 found to be K = 2, corresponding to dingo and dog ancestral populations. Individuals with 245 ancestral purity exceeding 90% were then classified as non -admixed dingoes or dogs. This 246 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint information was used in a supervised GHap analysis to examine introgression events and further 247 refine estimates of genetic structure. The results were visualized with karyoplots and Manhattan 248 plots in R, highlighting the genomic distribution of ancestral contributions and identifying 249 potential introgression events. 250 251 2.3 Detection and Characterization of Introgressed Blocks 252 Potential adaptive introgression in the dingo population was investigated by analyzing regions of 253 elevated dog ancestry across the genome to identify candidate genes under adaptive 254 introgression. Ancestral allele dosage data obtained from ELAI were analyzed for all 255 chromosomes (1-38). SNP information files (“.snpinfo.txt”) and dosage files (“.ps21.txt”) were 256 processed to ensure consistent data dimensions, and dog introgression rates were calculated by 257 averaging dog dosage values across all dingoes and rescaling the resulting proportions to the [0, 258 1] range. Regions with elevated dog ancestry were identified using a conservative, chromosome -259 specific threshold, defined as SNPs with ancestry values exceeding three standard deviations 260 above the mean for each chromosome. This threshold was used as a descriptive criterion to 261 highlight genomic regions showing unusually high ancestry relative to the chromosomal 262 background, rather than as a formal SNP -wise hypothesis test, as local ancestry estimates are 263 highly autocorrelated along chromosomes. We chose this chromosome -specific approach, 264 following Pilot et al. (2021), rather than a single genome -wide cutoff, because each chromosome 265 acts as an independent recombination unit, and ancestry blocks are expected to vary in size and 266 distribution across chromosomes. Automated R scripts facilitated the visualization and 267 comparison of the elevated -ancestry segments across all chromosomes. Conversely, “ancestry 268 deserts” (regions of exceptionally low dog ancestry; Kim et al., 2018) were identified following 269 Sankararaman et al. (2014) as runs of ≥10 consecutive SNPs with <0.1% dog ancestry. 270 To support the interpretation of the chromosome 9 candidate region (see Results), we 271 additionally estimated Weir and Cockerham’s F ST in sliding windows along chromosome 9. FST 272 was calculated in 50 kb non -overlapping windows between dingoes and domestic dogs using 273 VCF-based genotype data. 274 To evaluate whether ancestry deserts harbor an excess of potentially deleterious dog -275 associated variants, we annotated variants within these regions using Ensembl Variant Effect 276 Predictor (VEP; McLaren et al., 2016). We focused on variants showing strong allele -frequency 277 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint differences between dogs and dingoes, retaining sites segregating in dogs but rare or absent in 278 dingoes (AF_dogs ≥ 0.05 and AF_dingoes ≤ 0.01). As a sensitivity analysis, we repeated this 279 using a more permissive threshold (AF_dogs ≥ 0.02; AF_dingoes ≤ 0.01). We summarized 280 predicted functional consequences and screened for missense and loss -of-function annotations 281 (including SIFT -deleterious missense and stop -gained, frameshift, and essential splice -site 282 variants). 283 Candidate introgression blocks were identified across different chromosomes using SNP 284 array data, and their genomic coordinates were mapped to the Canis familiaris reference genome 285 (canFam6; assembly ID: GCF_000002285.5) using the UCSC Genome Browser, which served 286 as a common coordinate and annotation framework. SNP positions from the Axiom Canine HD 287 array, originally defined on earlier canine genome assemblies, were converted to canFam6 288 coordinates using established UCSC liftover mappings prior to downstream analyses. Genes 289 located within these regions were identified based on canFam6 annotations, and their predicted 290 functions were manually verified using GeneCards (Safran et al., 2010). Chromosomes 291 displaying prominent or consistent introgression patterns, along with those containing isolated 292 peaks of interest, were selected for further analyses. To assess patterns of molecular evolution 293 within introgressed regions, BED files defining introgressed blocks were generated. Coding 294 sequences corresponding to genes located within these regions were extracted from the dingo 295

genome (Canis lupus dingo; GCA_003254725.2) using BEDTools v2.29.2 (Quinlan, 296 2014). Homologous coding sequences from the grey wolf reference genome (Canis lupus; 297 GCA_905319855.2) were included as an outgroup for comparative analyses. Introgressed 298 regions themselves were defined exclusively based on SNP array data. 299 Extracted sequences were converted to FASTA format, and custom Python scripts were 300 employed to correct frameshifts (see Data Accessibility section). Multiple sequences were 301 aligned using MAFFT v7.450 (Katoh and Standley, 2014) to ensure consistency across analyses. 302 To evaluate potential selection pressures acting on introgressed regions, dN/dS ratios were 303 calculated by comparing aligned coding sequences from the canFam6 reference genome and 304 domestic dogs, with the grey wolf (Canis lupus) included as an outgroup. The Ka/Ks ratio, 305 implemented in the R packages ape v5.8 (Paradis et al., 2018) and seqinr v4.2 -36 (Charif et al., 306 2023), was used to further evaluate selection pressures. Statistical significance was assessed 307 using z -tests, focusing on genes with dN/dS ratios that significantly deviated from neutrality 308 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint (dN/dS ≠ 1; Kryazhimskiy and Plotkin, 2008), thereby identifying candidate genes potentially 309 under positive selection. 310 To assess the extent of linkage disequilibrium (LD) within introgressed blocks, we 311 computed pairwise LD values (R²) using PLINK v1.9. LD calculations were performed 312 separately for each chromosome. R² values were extracted from the resulting LD matrices, and 313 their distribution was summarized for each chromosome to evaluate patterns of haplotype 314 structure and recombination. Histograms and summary statistics were generated in R to visualize 315 the spread of LD values, allowing us to determine whether introgressed regions exhibit extended 316 haplotype blocks, indicative of reduced recombination. 317 318 2.4 Introgression analysis based on D-statistics 319 Variant calling was performed jointly across all canid genomes analysed in the study, including 320 dingoes, domestic dogs, New Guinea Singing Dogs (NGSD), grey wolves, and the golden jackal, 321 generating a single multi -sample variant call set. Raw variants were called using bcftools 322 mpileup and bcftools call (Narasimhan et al., 2016) and filtered to retain high -confidence 323 biallelic SNPs (QUAL ≥ 30, DP ≥ 10). This joint VCF constituted the basis for all subsequent 324 phylogenetic and introgression analyses. 325 To reconstruct maximum likelihood phylogenetic trees for dingoes and domestic dogs, 326 we used IQ-TREE v3 (Nguyen et al., 2015; Minh et al., 2020). Genome -wide SNP data included 327 in the joint variant call set were available for 390 dingoes distributed across regional populations 328 and 400 domestic dogs (200 European purebred dogs, 42 Australian purebred dogs, 19 329 Australian mixed-breed dogs, and 139 European free -ranging dogs), as well as 44 grey wolves 330 (Canis lupus ), one golden jackal ( Canis aureus ) from Stefanovic et al. (2024), and two New 331 Guinea Singing Dogs (NGSD; NCBI accessions SRR7107989 and SRR7107990). The golden 332 jackal was used as the outgroup. SNPs for the two NGSD genomes were extracted from the same 333 joint variant call set using identical filtering criteria, ensuring full comparability across taxa. 334 Individuals identified as recent backcrosses (N = 3) based on local ancestry analyses (ELAI) 335 were excluded prior to phylogenetic reconstruction. Model selection was performed using the 336 ModelFinder Plus option based on the Bayesian Information Criterion (BIC), and branch support 337 was assessed using ultrafast bootstrapping with 1,000 replicates. Phylogenetic trees were 338 visualized using FigTree v1.4.4 (Rambaut, 2009). 339 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint To investigate signatures of introgression between dingoes and dogs, we estimated D 340 statistics (ABBA –BABA test) and the f 4-ratio using Dtrios from Dsuite v0.1 (Malinsky et al., 341 2021) across all autosomal chromosomes. Analyses were performed on the global dataset as well 342 as separately for each regional dingo population. We used chromosome -specific VCF files and 343 the previously inferred maximum likelihood phylogeny to define population relationships. 344 Individuals identified as recent backcrosses based on local ancestry (ELAI) were excluded from 345 the analysis. Briefly, the ABBA–BABA test examines introgression within a four-taxon topology 346 of (((P1, P2), P3), O), where O is the outgroup. The D statistic compares the frequency of ABBA 347 (derived allele shared between P2 and P3) and BABA (derived allele shared between P1 and P3) 348 site patterns. Under the null hypothesis of no gene flow between P3 and either P1 or P2, ABBA 349 and BABA patterns are expected at equal frequencies; significant deviations from this 350 expectation indicate excess allele sharing. 351 For our analyses, we focused on trios involving dingoes and domestic dogs, structured as 352 ((NGSD, Dingo), Dog), using the golden jackal as the outgroup. This configuration allows 353 detection of excess shared derived alleles between dingoes and domestic dogs beyond what is 354 expected from shared ancestry with NGSD, which represents a closely related lineage (Surbakti 355 et al., 2020; Souilmi et al., 2024). We acknowledge that the captive NGSD population derives 356 from a limited number of founders and has experienced substantial inbreeding and genetic drift. 357 However, despite this demographic history, NGSD remains the closest available non -dingo 358 lineage and provides a relevant phylogenetic reference for testing excess allele sharing. Although 359 NGSD may also experience introgression from domestic dogs of European origin, such gene 360 flow would be independent from the introgression in dingoes and its extent in captive NGSD 361 individuals analysed here is expected to be limited (Scarsbrook et al., 2025). Therefore, excess 362 allele sharing detected between dingoes and domestic dogs relative to NGSD is unlikely to be 363 explained by shared ancestral polymorphism alone. 364 We also applied this approach to the introgressed block identified on chromosome 27. 365 We extracted SNPs corresponding to this region using bcftools (Narasimhan et al., 2016) and re -366 ran Dtrios to test for excess allele sharing at these loci. This refined analysis assessed whether 367 excess allele sharing was elevated locally relative to genome -wide expectations, providing 368 support for localized dog-derived ancestry in the dingo genome. 369 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint To explicitly address the limitations of D -statistics in distinguishing introgression from 370 incomplete lineage sorting at local genomic scales (Martin et al., 2015), we additionally 371 quantified localized introgression across the chromosome 27 candidate block using the f_d 372 statistic implemented in Dsuite (Dinvestigate). f_d was estimated in sliding windows across the 373 block to assess whether excess allele sharing with domestic dogs was concentrated within this 374 region, providing a more robust measure of localized introgression independent of genome -wide 375

processes. 376 377 2.5 Analysis of population-level variation in introgression using Bayesian genomic clines 378 Because population-level differences in introgression may not be fully captured by genome -wide 379 or predictive approaches, we additionally applied Bayesian genomic cline analyses to 380 characterize heterogeneity in introgression across loci within each population implemented in the 381 R package bgc -hm (Gompert et al., 2024). This approach models locus -specific deviations in 382 ancestry transitions along a hybrid index while explicitly accounting for genome -wide and 383 population-level variance. Analyses were conducted separately for each dingo population (Big 384 Desert, Central, East, South, West, and captive), using a shared set of ancestry -informative SNPs 385 derived from allele frequency differences between dingoes and domestic dogs. For each 386 population, individual hybrid indices were first estimated under a diploid genotype model using 387 the function est_hi. These estimates were then used as input for hierarchical genomic cline 388 models (est_genocl), which jointly estimate cline centers (α) and cline slopes (β) across loci 389 while allowing for population -specific dispersion parameters. To summarize introgression 390 heterogeneity within populations, we extracted posterior distributions of the dispersion of cline 391 centers (SDc) and cline slopes (SDv). Higher SDc or SDv values indicate greater locus -specific 392 heterogeneity in introgression, whereas lower values reflect more homogeneous ancestry 393 transitions across the genome. All models were run for 3,000 MCMC iterations using default 394 priors, and convergence was assessed visually. 395 396 2.6 Genetic and Environmental Drivers of Dingo-Dog Introgression Patterns 397 To analyze the potential role of environmental variables in shaping dingo -dog introgression 398 patterns, we first compiled topographic, environmental and landscape data potentially driving 399 these patterns (Table S2). Topographic (elevation) and environmental (19 bioclimatic variables) 400 data were obtained from the WorldClim database (Hijmans et al., 2005). These variables 401 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint represent annual trends (e.g., mean annual temperature, annual precipitation), seasonality (e.g., 402 temperature seasonality, precipitation seasonality), and extreme or limiting environmental factors 403 (e.g., temperature of the coldest and warmest months, precipitation of the driest and wettest 404 quarters). The WorldClim data were based on averages for the years 1970 –2000. Landscape data 405 were obtained from the Australian Bureau of Agricultural and Resource Economics and Sciences 406 data package (ABARES, 2022) and comprised land use and vegetation cover for Australia from 407 2010-2011 and 2015 -2016. This included land use change, trees, shrub and croplands cover. 408 Additionally, we incorporated regional human population data and dingo barrier fence data 409 obtained from the Australian Bureau of Statistics (ABS, 2023). Human footprint index data from 410 the Global Human Settlement Layer (GHSL; Schiavina et al., 2022) was also included and used 411 for calculating the distance to the nearest human settlement. We set all the variables at 30 412 seconds of the geographic coordinate system (~1 km²) of spatial resolution and obtained the 413 environmental data for each sampling location. Multicollinearity among predictors was 414 addressed by calculating pairwise correlations using the “pairs.panels” function from the psych 415 v2.1.9 R package (Revelle and Revelle, 2015), excluding variables with Pearson’s correlation 416 values of |r| > 0.70. 417 We then evaluated associations between genetic variation, local adaptation, and 418 environmental factors by testing for isolation by distance (IBD; Wright, 1943) and isolation by 419 environment (IBE). IBE is observed when positive correlation occurs between genetic and 420 environmental distances, independent of the effect of geographic distance (Wang and Bradburd, 421 2014). Pairwise genetic distances (Nei’s D) were estimated in R using the package hierfstat v0.5-422 7 (Goudet, 2005) with the "genet.dist" function and compared to geographic and environmental 423 distances using Mantel tests. Geographic distances were calculated as Haversine distances with 424 the “distm” function from the geosphere v1.5 –14 R package (Hijmans et al., 2017), while 425 environmental distances were computed as Euclidean distances using the “dist” function from 426 the stats v3.3.1 R package (R Core Team, 2013). Mantel tests were performed using Spearman 427 correlation with 9999 permutations in the vegan v2.5–7 R package (Oksanen et al., 2013). 428 We conducted partial Redundancy Analysis (pRDA ; Capblancq and Forester, 2021) to 429 partition the variance in genetic variation explained by climate, geography, and population 430 structure. The response variable consisted of individual genotypes (coded as allele counts: 0/1/2), 431 while explanatory variables were grouped into three sets: (1) climate -related environmental 432 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint variables (Table S2a), (2) proxies of genetic structure (population scores along PC1 and PC2), 433 and (3) geographic coordinates (latitude and longitude). The full pRDA model was fitted using 434 all explanatory variables simultaneously and tested for overall significance. A stepwise 435 procedure (ordiR2step) was applied only as an exploratory analysis to assess the relative 436 contribution of individual predictors, with significance determined by p < 0.01 and adjusted R² 437 values. Three separate pRDA models were performed to evaluate the independent contributions 438 of each set of variables: (1) an environment -only model conditioned on geography and 439 population structure, (2) a population structure -only model conditioned on geography and 440 environmental variables, and (3) a geography -only model conditioned on population structure 441 and environmental variables. This approach disentangled confounding effects and allowed us to 442 assess the relative contributions of each factor. This framework allows environmental 443 associations to be interpreted conservatively by explicitly accounting for spatial structure and 444 shared ancestry, which are major sources of confounding in genotype –environment association 445 analyses. 446 To investigate potential signals of local adaptation, we performed a second pRDA 447 specifically designed to detect genetic associations with environmental predictors. This analysis 448 used a response matrix of genotypes, with environmental predictors as explanatory variables and 449 population structure (PC1 and PC2) and geography (latitude and longitude) as conditioning 450 variables. The significance of RDA axes and individual environmental predictors was tested 451 using permutation tests (n = 999) with the anova.cca function in vegan. Outlier loci potentially 452 under selection were identified based on SNP loadings along the first three constrained axes, 453 applying a stringent filtering threshold of 3.5 standard deviations (p < 0.0005) (Forester et al., 454 2018). We manually checked for duplicate candidate loci associated with multiple RDA axes and 455 used Pearson’s correlation (r) to identify the strongest environmental predictor for each outlier 456 locus. This step aimed to detect loci potentially involved in local adaptation, linking genetic 457 variation to environmental drivers. As with other multivariate GEA approaches, results are 458 expected to depend on the set of environmental predictors considered; however, the pRDA 459 framework is well suited to identify robust associations with major climatic gradients while 460 controlling for non-environmental structure. 461 Random Forest (RF) analyses were used to investigate environmental drivers of variation 462 in introgression across dingo populations. Dog ancestry proportions were estimated at the 463 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint individual level using ELAI (see above) and used as a continuous response variable in all RF 464 models. We initially considered the full set of environmental and anthropogenic predictors 465 described above, including 19 bioclimatic variables, elevation, land -cover variables, and human 466 footprint metrics (26 predictors in total). To reduce redundancy among predictors, we filtered 467 variables based on pairwise correlations, iteratively removing highly correlated predictors (|r| > 468 0.8). This procedure resulted in a reduced but still comprehensive set of 16 predictors. No 469 variable selection based on the RDA results was applied, and RF analyses were therefore 470 conducted independently of the pRDA. RF regression models were fitted using the randomForest 471 R package (Liaw and Wiener, 2002), with 1,000 trees and default settings for the number of 472 variables considered at each split. Variable importance was quantified using permutation -based 473 importance measures (%IncMSE), which reflect the decrease in predictive accuracy when each 474 predictor is randomly permuted. Model performance was first evaluated using out -of-bag (OOB) 475 predictions. To further assess model generalizability and reduce the risk of overfitting, we 476 implemented a leave -one-population-out cross-validation (LOPO) approach. In this framework, 477 each population was sequentially excluded from model training and used exclusively for testing, 478 providing an explicit out -of-sample evaluation of predictive performance. Model accuracy was 479 quantified using mean squared error (MSE) and the coefficient of determination (R²). 480 481 3. Results 482 3.1 Population Structure and Admixture Patterns 483 The PCA of both dingo and dog populations of European origin revealed clear genetic 484 structuring, with dingoes forming a distinct cluster separate from domestic dogs. Free -ranging 485 and purebred dogs exhibited genetic differentiation but were more closely related to each other 486 than to dingoes (Figure 1 a). ADMIXTURE analyses (K = 12, Figure S 1) supported the PCA 487 results, showing that dingoes and domestic dogs formed separate clusters (Table S3a). 488 ADMIXTURE estimated the average dog admixture in dingoes at 0.09 and dingo admixture in 489 Australian mixed-breed and pure -breed dogs at 0.04 and 0.02, respectively. Dingo admixture in 490 European pure -breed dogs and European free -ranging dogs was estimated at 0.01 and 0.03, 491 respectively, which provides an estimate of a false positive rate. 492 When focusing exclusively on the dingo population, the PCA revealed significant genetic 493 differentiation among subpopulations, primarily aligned with geographic regions. Dingoes from 494 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint the Big Desert and Western Desert formed distinct genetic clusters, while subpopulations from 495 Northern, Southern, and Eastern Australia showed varying degrees of overlap, reflecting regional 496 differentiation and possible gene flow (Figure 1 b). ADMIXTURE analyses (Figure 1 c, Figure 497 S2) corroborated these patterns, highlighting the genetic isolation of desert populations and 498 shared ancestry among Northern and Eastern subpopulations. 499 3.2 Local Ancestry and Introgression Analysis 500 3.2.1. Local Ancestry Patterns 501 The combined results from LAMP -LD, ELAI, and GHap analyses revealed clear patterns of dog 502 introgression in dingoes across Australia (Supplementary Table S3). Both LAMP and ELAI 503 estimated the average dog admixture proportion at 0.15 (for ELAI, this estimate was achieved by 504 assuming that all individuals included in the reference population as pure dingoes had dog 505 admixture equal 0), while GHap estimated the admixture proportion at 0.12. All methods 506 identified significant population -level differences in introgression levels, showing that dingoes 507 from the Central and Western region exhibit markedly lower proportions of dog ancestry 508 compared to those from other regions, as well as the captive population, while the highest 509 admixture occurs in the Eastern Region (Figure 1d, Figure S 3; Table 1; Table S3b). All three 510

estimated dingo admixture in mixed-breed and pure -breed Australian dogs at less than 511 0.02 and less than 0.01, respectively, implying very limited introgression from dingoes to dogs 512 (Table S3). In European dogs, the estimate of dingo admixture was negligible, showing that the 513 false positive rate (i.e. detection of false admixture) is low for all three methods. 514 All three methods of local ancestry analyses highlighted chromosomes 9 and 27 as 515 genomic regions with particularly elevated dog ancestry (Figure 2 ), especially in Eastern and 516 Southern populations (Table S3b -d). A comparison of global versus local ancestry estimates 517 revealed that the three local methods (LAMP -LD, ELAI and GHap) produced highly consistent 518

(mean difference ≈ 1% per individual), whereas global ancestry proportions from 519 ADMIXTURE differed by 6 -7% on average from the local ancestry estimates (Table 1; Figure 520 S4). Comparison of ancestry estimates across genomic scales shows that both the mean and 521 variance of dog ancestry increase markedly on chromosomes 9 and 27 relative to genome -wide 522 estimates, with highly concordant patterns between LAMP and ELAI (Figure 2b). 523 3.2.2. Identification of Introgressed Regions 524 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint GHap analyses confirmed these chromosomes as primary regions of dog ancestry, where distinct 525 introgressed haplotype blocks were identified (Figure 2c-d), spanning 19.5 Mb on chromosome 9 526 and 8.9 Mb on chromosome 27. Notably, both chromosomes 9 and 27 exhibited elevated dog 527 ancestry in Eastern and Southern dingoes, with chromosome 27 also showing some introgression 528 in Central populations. These patterns suggest partial geographic structure in introgression. In 529 addition, isolated peaks of introgression were observed on chromosomes 13, 14, and 24. These 530 isolated peaks, though smaller in size, may represent additional regions of interest for adaptive 531 introgression (Tables S4, S5, Figure S5). 532 We explored linkage disequilibrium (LD) patterns in regions of introgression on 533 chromosomes 9 and 27, which harbored the most prominent dog -derived haplotype blocks. 534 Although LD (R 2) was calculated genome -wide for all chromosomes (mean = 0.103; 535 median = 0.049; max = 1.000), we focus here on chromosomes 9 and 27 due to their relevance. 536 On chromosome 9, the mean R 2 was 0.069 (median = 0.020; max = 1.000), and on chromosome 537 27 it was 0.055 (median = 0.018; max = 0.984). Crucially, over 95% of all SNP pairs with R 2 > 538 0.8 on chromosome 9 fall within our previously defined introgressed interval on that 539 chromosome, and over 90% of high -LD SNP pairs on chromosome 27 co -localize with its 540 candidate block. This strict spatial confinement of elevated LD to the introgressed regions 541 indicates that recombination has been less effective at breaking down these haplotypes locally, 542 potentially in combination with selection or recent demographic history. 543 To place these values in a genome -wide context, we summarized median LD (R² ≤ 500 544 kb) for all autosomes. Median LD values showed moderate variation across chromosomes but 545 largely overlapped (Figure S6). Chromosome 27 fell toward the lower end of the genome -wide 546 distribution, but did not differ significantly from the remaining autosomes when compared 547 against chromosomal medians (Wilcoxon rank-sum test, p = 0.053). 548 Using a fine -scale dog recombination map (Kidd, 2026), the chromosome 27 block 549 showed a length-weighted mean recombination rate of 1.39 cM/Mb. In a permutation test based 550 on 10,000 length -matched windows sampled across chromosome 27, the block rate fell well 551 within the chromosome -wide distribution (empirical p for unusually low recombination = 0.74; 552 Figure S7), indicating that the candidate region does not coincide with a recombination coldspot. 553 This provides an additional support for adaptive introgression as the potential cause of the 554 elevated introgression in this chromosomal block. 555 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint The fragment on chromosome 9 initially identified as an introgressed block was later 556 found to encompass a large chromosomal inversion previously described in domestic dogs but 557 absent in dingoes (Field et al., 2022). Using inversion coordinates provided by the authors and 558 converting them to the canFam6 reference genome with the UCSC liftOver tool (Raney et al., 559 2024), we confirmed that the inversion spans an interval fully contained within the candidate 560 region. This region does not show decreased genetic differentiation be tween dingoes and 561 domestic dogs (Figure S8), which would be expected in the case of intense introgression inferred 562 from the local ancestry patterns (Figure 2, Figure S3a) . The chromosome 9 inversion is present 563 in most but not all dogs breeds, so for this chromosomal region dogs that do not have th e 564 inversion are more similar to dingoes than to other dogs. This shared similarity between dingoes 565 and some dogs could have resulted in an incorrect signal of introgression in the local ancestry 566 analysis, with dogs that do not have the inversion being interpreted as donors of this variant to 567 dingoes (Figure S3a) . A lack of the inversion in dogs is rarer than presence , which is more 568 consistent with a scenario where dingoes are the donors of this variant to dogs. However, this in 569 turn is inconsistent with the pattern observed in other parts of chr omosome 9 (no introgression 570 from dingoes to dogs), which could be the reason why introgression from dogs to dingoes was 571 inferred. Field et al. (2022) showed that a lack of the inversion is an ancestral state (occurring in 572 grey wolves and dingoes ) and the inversion occurred recently during the dog breed 573 diversification process. Therefore, it is unlikely that the pattern observed in chromosome 9 574 reflects actual gene flow from dogs to dingoes, and we excluded this chromosomal region from 575 subsequent introgression analyses. 576 Besides the regions with elevated dog ancestry, we also detected dog ancestry deserts on 577 eight chromosomes (Table S 4, Figure S 9). Most of these ancestry deserts included protein -578 coding genes, indicating possible targets of negative selection. Notably, one of these regions 579 included Bone Morphogenetic Protein 4 (BMP4), a gene with well -established roles in 580 developmental processes (Ye et al., 2022). These ancestry deserts were generally short, 581 suggesting that strong, genome -wide barriers to introgression are limited. Dog -enriched variants 582 within ancestry deserts were predominantly non -coding (upstream, downstream, or intronic) and 583 showed no evidence of deleterious protein-coding changes. Across 21 variants under our primary 584 filter and 25 variants under a more permissive filter, we detected no missense or predicted loss -585 of-function variants. 586 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint 3.2.3. Functional Analysis of Genes Within Introgressed Regions 587 Within the introgressed block on chromosome 27, most genes showed signatures of purifying 588 selection based on dN/dS ratios significantly lower than 1 (Figure 4; Table S5). One gene, 589 OR8S3 (Olfactory receptor family 8 subfamily S member 8), exhibited a dN/dS ratio 590 significantly above 1, consistent with positive selection (p = 0.0012). Additional introgressed 591 regions on chromosomes 13, 14, and 24 also contained genes under purifying selection. 592 Specifically, GNPDA2 and KCTD8 (chr13), and SDHAF3 (chr14) showed significant results, 593 while ASNS, EMILIN3, and CHD6 were non -significant or marginally so. These analyses were 594 based exclusively on coding sequences; therefore, potential selective pressures on non -coding or 595 regulatory variants within the introgressed regions cannot be assessed with this approach. 596 Although the chromosome 9 region was later shown to represent an inversion rather than an 597 introgressed block, we nonetheless examined its protein -coding genes and found that two of 598 them -GJC1 (Gap Junction Protein Gamma 1) and TCAP (Titin -Cap)- exhibit signals of positive 599 selection (Table S6). 600 3.2.4. Phylogenetic and Genome-Wide Evidence of Introgression 601 For clarity of visualization and interpretation, the main phylogenies shown in Figure 3 were 602 reconstructed using a representative subset of five randomly selected individuals per population 603 or species, except for NGSD (n = 2) and the golden jackal (n = 1). The phylogenies reconstructed 604 using all available individuals are provided in the Supplementary Material (Figure S1 0). The 605 Maximum Likelihood tree reconstructed with IQ -TREE based on genome -wide SNP data 606 revealed reciprocally monophyletic clades for dingoes and domestic dogs (with domestic dog 607 cluster grouping purebred and mixed -breed dogs independent of geographic origin as well as 608 European free -ranging dogs). In contrast, New Guinea Singing Dogs (NGSD) did not form a 609 reciprocally monophyletic clade distinct from dingoes, instead clustering closely with them -610 consistent with their shared evolutionary history (Figure 3a, Figure S10a). The phylogenetic tree 611 for the introgressed region on chromosome 27 revealed clustering between domestic dogs and 612 dingoes, consistent with introgression (Figure 3b, Figure S10b). All populations except Big 613 Desert and captive dingoes had introgressed haplotypes in their gene pool, and these haplotypes 614 clustered with different European dog clades, implying that the introgression is a result of 615 multiple dingo-dog cross-breeding events. Domestic dogs did not have any haplotypes clustering 616 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint with dingo -specific haplotypes, implying no introgression from dingoes to dogs in this 617 chromosomal region. 618 In the phylogeny for the chromosome 9 region , some European domestic dogs clustered 619 closely with dingoes (Figure 3c, Figure S10c), while other dogs formed distinct clusters . This 620 pattern is consistent with the presence of a chromosomal inversion previously described in some , 621 but not all, dog breeds and absent in dingoes (Field et al. 2022). As such, the phylogenetic signal 622 in this region likely reflects haplotype similarity maintained by suppressed recombination within 623 the inversion, rather than introgression. 624 Additionally, ABBA -BABA analyses (D -statistics) provided strong genome -wide 625 evidence of introgression from domestic dogs into dingoes. Across all dingo populations, the D -626 statistic was 0.1547 (Z = 7.20, p < 1×10⁻¹²), with an f₄ -ratio of 0.192, indicating a substantial 627 excess of shared derived alleles between dingoes and domestic dogs (Table 2). When examined 628 individually, all dingo populations exhibited positive and highly significant D -statistics, with the 629 strongest signal detected in Eastern populations (D = 0.193, Z = 8.53, p = 2.3×10⁻¹⁶), followed 630 by those from the South (D = 0.179, Z = 7.08, p = 1.5×10⁻¹²), con gruent with our local ancestry 631

(Table 2). These findings underscore geographic variation in the intensity of dog ancestry 632 across the dingo range. Focusing on the introgressed region on chromosome 27, the signal of dog 633 ancestry was even stronger (D = 0.263, Z = 3.35, p = 0.0008), with an f₄ -ratio of 0.556, 634 suggesting substantial localized introgression. Among populations, only trios with informative 635 ABBA-BABA sites and p < 0.05 were reported. Central dingoes exhibited the strongest signal 636 (D = 0.282, Z = 3.58, p = 0.0003), followed by those from the South and West (D = 0.235 and 637 0.210; Z = 2.83 and 2.69; p = 0.0046 and 0.0072, respectively). Across the chromosome 27 638 candidate block, f_d values were consistently elevated across most genomic windows (typically 639 ranging from ~0.4 to >0.6), indicating a strong excess of dog -derived ancestry concentrated 640 within this region. In contrast, f_d values decreased sharply outside the core of the block, 641 supporting the interpretation that introgression on chromosome 27 is spatially localized rather 642 than driven by genome-wide background processes. 643 3.2.5 Population-level variation in introgression revealed by Bayesian genomic clines 644 Bayesian genomic cline analyses revealed pronounced population -level differences in the 645 introgression patterns among dingo populations. Estimates of individual hybrid indices (HI) 646 confirmed substantial heterogeneity in overall levels of dog ancestry, with eastern and southern 647 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint populations showing higher mean HI values, whereas Western and Big Desert populations 648 exhibited lower average introgression (Table S7). 649 The populations also varied markedly in locus -specific introgression patterns. Eastern and 650 southern dingoes showed relatively low dispersion of genomic cline centers (SDc) and slopes 651 (SDv), indicating more homogeneous ancestry transitions across loci. In contrast, Western and 652 Central populations exhibited higher SDc values, consistent with greater locus -specific 653 heterogeneity in introgression across the genome. Captive dingoes showed intermediate levels of 654 dispersion, reflecting mixed ancestry patterns. Together, these results indicate that introgression 655 in dingoes is not only geographically structured in magnitude but also differs in its genomic 656 architecture among populations. Overall, populations with higher mean HI showed 657 comparatively lower SDc/SDv, whereas populations with lower mean HI showed higher 658 dispersion, although this pattern was not uniform across all populations. These patterns are 659 consistent with results from local ancestry inference and genome -wide introgression analyses, 660 reinforcing the conclusion that introgression dynamics vary spatially across Australia. 661 662 3.3 Influence of Environmental Variation on Spatial Introgression 663 Mantel tests revealed no significant relationship between genetic and geographic distances 664 (Figure S11a, r = 0.027, p = 0.1913). However, a positive correlation between genetic and 665 environmental distances (Figure S11b, r = 0.1515, p < 0.0001) supports genetic differentiation 666 driven by environmental gradients. 667 To identify the environmental variables most relevant to introgression patterns and 668 reduce redundancy among predictors, we conducted a partial Redundancy Analysis (pRDA) as 669 an intermediate step. This analysis revealed that environmental factors significantly contribute to 670 genetic differentiation between dingo populations, with the full model (climate, geography, and 671 population structure) explaining 9.7% of the genetic variation (p < 0.001; Table 3). Climate -672 specific variables were the strongest predictors, accounting for 4.9% of the variation, while 673 population structure and geography contributed 3.0% and 1.8%, respectively. Among the 674 environmental predictors, the maximum temperature of the warmest month (BIO5), mean 675 temperature of the driest quarter (BIO9), and the Human Footprint Index (HFPI) emerged as the 676 most influential variables (Figure 5). 677 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint In addition to identifying key environmental predictors, the pRDA highlighted loci 678 potentially involved in local adaptation. Across the second and third constrained axes, 524 679 outlier loci were associated with environmental variables, particularly temperature-related factors 680 (e.g., BIO5: 414 loci; BIO9: 353 loci) and vegetation -related features such as shrub and tree 681 cover (388 and 386 loci, respectively; Table S2e). These loci suggest possible genetic responses 682 to climatic and landscape features, supporting the hypothesis of environmental selection 683 pressures shaping genetic differentiation. 684 To directly evaluate how environmental variation relates to the proportion of dog 685 introgression, we used Random Forest models with individual -level dog ancestry as the response 686 variable. Across all individuals, introgression proportions ranged from 0.03 to 0.48, with a mean 687 of 0.16, and with substantial heterogeneity among populations. The global Random Forest model 688 explained a substantial proportion of variance in introgression (out -of-bag R² = 0.66; MSE = 689 0.0022). Variable importance rankings consistently identified climatic predictors as the strongest 690 contributors, particularly maximum temperature of the warmest month (BIO5), precipitation of 691 the coldest quarter (BIO19), and mean temperature of the wettest quarter (BIO8), followed by 692 additional temperature and precipitation -related variables (Figure S12). Anthropogenic 693 predictors, including HFPI and land -use metrics, contributed more modestly but consistently to 694 model performance. 695 Predictive performance varied markedly among populations when evaluated using a 696 leave-one-population-out cross-validation approach. Model performance was highest in southern 697 populations (R² = 0.69) and moderate in Western and Northern populations (R² = 0.34 and 0.32, 698 respectively), whereas predictive power was very low in Eastern, Central, and Big Desert 699 populations (R² ≤ 0.03). This spatial heterogeneity indicates that relationships between 700 environmental conditions and introgression are not uniformly transferable across regions. 701 To place these results in a biological context, we quantified effect sizes along observed 702 environmental gradients. Shifting BIO5 from the 10th to the 90th percentile of its empirical 703 range was associated with an average decrease of approximately 4.7% in predicted dog ancestry, 704 whereas an equivalent shift in BIO19 corresponded to an increase of approximately 3.9%. 705 Changes along BIO8 produced smaller effects (≈1%). These results indicate that individual 706 climatic gradients exert moderate effects on introgression levels, but their combined influence 707 can generate substantial spatial variation in hybridization across dingo populations. 708 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint 4. Discussion 709 4.1 Introgression Patterns in Australian Dingoes 710 Our study provides new insights into the genomic patterns of introgression between dingoes and 711 domestic dogs in Australia, emphasizing the complex interplay between gene flow and 712 environmental adaptation. By integrating high-resolution genomic data with ecological variables, 713 we highlight how environmental and anthropogenic factors are influencing the spatial and 714 genomic distribution of dog introgression in dingo populations. Population structure analyses, 715 including PCA and ADMIXTURE, confirmed regional differentiation among dingo populations 716 reported in prior studies (Cairns et al., 2023, Souilmi et al., 2024) as well as strong genetic 717 differentiation between dingoes and domestic dogs, with minimal overlap between clusters. 718 Nonetheless, admixture signals were detected, particularly in dingoes from Eastern and Southern 719 Australia. Local ancestry inference methods (LAMP -LD, ELAI, and GHap) further revealed 720 complex and heterogeneous introgression signatures across the genome. ABBA –BABA analyses 721 also supported the presence of genome -wide introgression, albeit with varying intensity across 722 geographic regions. Bayesian genomic cline analyses further indicate that populations differ not 723 only in the extent of introgression but also in its genomic architecture, with some populations 724 showing homogeneous ancestry transitions across loci and others exhibiting pronounced locus -725 specific heterogeneity. Genomic regions with overrepresented dog ancestry were identified on 726 four chromosomes, most prominently on chromosome 27. Importantly, this localized signal was 727 also supported by elevated f_d values across the chromosome 27 block, a statistic specifically 728 designed to detect introgression at fine genomic scales and to reduce the confounding effects of 729 incomplete lineage sorting (Martin et al., 2015). Taken together, our findings suggest historical 730 and ongoing dog introgression, with hybridization cases occurring infrequently but leading to 731 backcrossing, which has left variable genomic footprints across the dingo genome. 732 Differences in estimated ancestry proportions across studies likely reflect contrasts 733 between global clustering approaches and locus -specific ancestry inference methods. Our global 734 ancestry estimates are broadly consistent with those reported by Cairns et al. (2023), whereas 735 higher admixture proportions emerge when applying local ancestry inference. In contrast, global 736 clustering methods applied by Weeks et al. (2024) revealed minimal or no admixture. All the 737

we applied were consistent in showing the presence of dog admixture in dingoes, 738 although its proportions differed between the global ancestry estimate using ADMIXTURE (9% 739 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint at average) and local ancestry estimates (12 -15%). All the methods were also consistent in 740 identifying the lowest admixture proportions in the Central and Western populations and the 741 highest proportions in the Eastern, Southern and Captive populations. Our local ancestry 742 estimates are in line with estimates based on whole genome sequences using an ancient DNA 743 baseline (Scarsbrook et al., 2025). Very low (<0.0015) dingo admixture proportions in European 744 dogs inferred based on local ancestry methods show that the false positive rate is low. In 745 contrast, ADMIXTURE identified 3% of dingo ancestry in European free -ranging dogs, which 746 can be attributed to recent shared ancestry. These differences highlight how global clustering 747 approaches may dilute or underestimate localized introgression signals, whereas local ancestry 748 inference has greater power to detect both contemporary and historical gene flow. However, 749 when historical introgression has become pervasive across the genome, it may be interpreted as 750 ancestral variation rather than discrete admixture tracts. Local ancestry methods may be thus 751 considered as more precise, yet they may be prone to other sources of errors, such as those 752 resulting from the structural changes occurring in the process of dog evolution (see below). 753 Our integrative approach, combining local ancestry methods with phylogenomic 754 analyses, enabled us to detect introgression signals that were not apparent in previous SNP-based 755 studies. A key strength of our work lies in its broad geographic sampling, allowing for a 756 comprehensive assessment of hybridization dynamics across the continent. Below, we discuss 757 the environmental and anthropogenic factors shaping introgression patterns, highlight genomic 758 regions of interest that may reflect adaptive introgression, and consider the implications of these 759 findings for dingo conservation and genetic monitoring. 760 4.2 Environmental and Anthropogenic Drivers of Hybridization 761 Human-modified landscapes likely play a key role in facilitating hybridization between 762 Australian dingoes and domestic dogs . Indeed, our results detected significant interactions 763 between proximity to human settlements and introgression, which likely arise from higher 764 densities of free -roaming owned dogs ( Sparkes et al., 2022) , more intense lethal control 765 measures, or the effects of landscape alteration. Local ancestry analyses using LAMP -LD and 766 ELAI reveal clear geographic patterns of introgression, with dingoes in the Central, Western, and 767 Big Desert regions exhibiting considerably lower levels of dog ancestry compared to those in the 768 East and South. Notably, some of the differentiation between Western and Eastern dingo 769 populations predates European colonization, indicating long -standing historical isolation 770 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint alongside recent admixture (Souilmi et al., 2024 ). These findings are consistent with population 771 structure results and support the idea that introgression varies geographically across dingo 772 populations. While population structure likely reflects multiple historical and demographic 773 processes, differences in introgression patterns appear to contribute to the observed genetic 774 clustering. This interpretation aligns with previous genomic studies (Stephens et al., 2022; Cairns 775 et al., 2023; Weeks et al., 2024) and suggests that ecologically and geographically isolated 776 populations-such as those in the Central region-experience minimal admixture, likely due to both 777 reduced contact with domestic dogs and harsher environmental conditions that limit their 778 overlap. In contrast, dingoes from Eastern and Southern Australia – regions with higher human 779 footprint - show higher levels of introgression. 780 Notably, our results revealed a strong signal of isolation by environment (IBE) but no 781 significant evidence of isolation by distance (IBD) in the present dataset. Although previous 782 studies have reported IBD patterns in dingoes (Stephens et al., 2022; Weeks et al., 2024), our 783 findings suggest that environmental heterogeneity may play a particularly important role in 784 shaping both population divergence and introgression at the spatial and genomic scale examined 785 here. In other canids, such as wolves, long -distance dispersal is common, but environmental 786 gradients remain key determinants of gene flow (Geffen et al., 2004; Pilot et al., 2006; Leonard 787 et al., 201 4). In dingoes, this interplay between dispersal potential and environmental variation 788 may similarly create localized barriers to gene flow, which influence both adaptation and 789 introgression outcomes. 790 Understanding the environmental and anthropogenic drivers of introgression is essential 791 to clarifying how hybridization shapes dingo populations in increasingly altered landscapes. 792 Dingoes inhabit a wide range of environments, and while direct assessments remain limited, 793 interactions between human activity and climatic conditions likely create varying opportunities 794 for contact and genetic exchange with domestic dogs (Cairns et al., 2020). For example, 795 ecological overlap in peri -urban areas and pastoral lands may increase dingo -dog encounters, 796 particularly where human -associated resources concentrate domestic dogs in areas that also 797 attract dingoes. These landscapes not only promote higher contact rates but also impose selective 798 pressures that may favor the retention of dog -derived alleles adaptive in human -dominated 799 landscapes in dingo genomes. 800 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Our Random Forest (RF) analyses highlight the importance of environmental gradients, 801 together with anthropogenic factors, in driving these patterns. Climatic variables emerged as the 802 strongest predictors of introgression, particularly maximum temperature of the warmest month 803 (BIO5) and precipitation -related variables (e.g. BIO19), suggesting that climate plays a central 804 role in shaping introgression dynamics. The Human Footprint Index (HFPI) also contributed 805 consistently to model performance, supporting the idea that hybridization and subsequent 806 introgression may be facilitated in areas with greater human disturbance. Together, these results 807 indicate that introgression patterns are strongly structured by climate, with human influence 808 acting as a context -dependent modifier. Additionally, climatic variables such as maximum 809 temperature of the warmest month (BIO5) and other temperature - and precipitation -related 810 factors likely shape hybridization dynamics by influencing habitat suitability, movement 811 patterns, and resource use. In Australia, climatic conditions strongly shape patterns of human 812 settlement, with milder climates supporting higher human densities. Because domestic dog 813 density closely tracks human density (Gompper , 2014), climate may indirectly modulate dingo -814 dog encounter rates by structuring dog abundance and human land use. In arid or seasonally dry 815 regions, dingoes may concentrate around permanent water sources, pastoral infrastructure or 816 areas with anthropogenic food subsidies, potentially increasing contact with dogs during 817 climatically stressful periods. 818 Dependence of hybridization rates on environmental variables have been reported across 819 a wide range of taxa, indicating that environmentally mediated hybridization is a general 820 phenomenon rather than a system -specific exception. In birds, human habitat disturbance has 821 been shown to increase hybridization rates between closely related species, as observed in black -822 capped and mountain chickadees, where landscape modification alters contact zones and mating 823 opportunities (Grabenstein et al., 2023). In freshwater fishes, hybridization outcomes among 824 trout species vary predictably with environmental context and historical demographic processes 825 linked to habitat alteration and management history (Mandeville et al., 2019). Similarly, 826 asymmetrical hybridization and spatial genetic structure in killifish hybrid zones have been 827 shown to reflect the combined influence of environmental gradients and landscape features 828 (Hardy et al., 2021). In plants, environmental heterogeneity and landscape structure can shape 829 mosaic hybrid zones and influence the maintenance of reproductive barriers (Faske et al., 2024). 830 Together, these studies reinforce the view that hybridization rates and genomic outcomes are 831 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint strongly contingent on extrinsic environmental conditions, paralleling the patterns we observe in 832 dingoes. Within canids, similar environmentally mediated patterns of admixture have also been 833 reported. Studies in North America have demonstrated that urbanization and landscape changes 834 significantly influence admixture between coyotes and wolves (Stronen et al., 2012), 835 underscoring the importance of human -mediated pressures in shaping hybridization outcomes in 836 wild canids (Pilot et al., 2021). 837 838 4.3 Adaptive Significance of Introgressed Haplotype Blocks 839 Building on the geographic patterns of introgression revealed through our analyses, we identified 840 a major introgressed block on chromosome 27 and several smaller peaks on chromosomes 13, 841 14, and 24. These regions may represent portions of the dingo genome where dog -derived alleles 842 have been retained by selection acting on functionally relevant genes. In most genes within the 843 chromosome 27 block and in the other smaller peaks, dN/dS values are below 1, suggesting 844 purifying selection. We note , however, that historical bottlenecks and reduced effective 845 population size can relax purifying selection and alter expectations for dN/dS ratios 846 (Kryazhimskiy and Plotkin, 2008 ), which cautions against interpreting elevated dN/dS values as 847 definitive evidence of selection. Historical bottlenecks experienced by dingoes (Kumar et al., 848 2023) can generate heterogeneous genomic patterns and accelerate the fixation of haplotypes, 849 particularly in regions of low recombination. In addition, heterogeneity in recombination rates 850 across chromosomes is known to shape genomic landscapes independently of selection (Burri et 851 al., 2015), and neutral variation in introgression rates has been widely documented in genomic 852 cline analyses (Gompert and Buerkle, 2011). 853 Although the signatures of natural selection should be treated with caution before they 854 can be tested at the phenotypic level, we detected one potential case of adaptive introgression: an 855 olfactory receptor gene, OR8S3, within the chromosome 27 block, exhibits a clear signature of 856 positive selection, suggesting that the introgressed allele may confer a sensory advantage. 857 Olfactory receptors (ORs) are central to foraging, social communication, and environmental 858 sensing in mammals (Robin et al., 2009), and wild canids often show strong selection on OR 859 repertoires in response to ecological pressures. Comparative data indicate that dingoes maintain a 860 larger, wolf‐like OR repertoire than modern dog breeds -likely reflecting continued reliance on 861 prey hunting (Mouton et al., 2025). In this context, the retention and selection of a dog -derived 862 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint OR8S3 variant could represent a fine -tuned sensory adaptation, perhaps aiding dingoes in 863 detecting novel anthropogenic cues or exploiting human-associated resources. 864 A region on chromosome 9 was flagged by all local -ancestry methods, but exhibited an 865 unusual introgression pattern (Figure S3a). We found this signal arose from a large inversion 866 segregating in dogs (Field et al., 2022) that suppresses recombination and conserves extended 867 haplotypes. Because this inversion is neither fixed in dogs nor present in dingoes (Field et al., 868 2022), we interpret the shared haplotypes as retained ancestral variation rather than true gene 869 flow and therefore we have excluded this region from our introgression analyses. 870 Linkage disequilibrium (LD) analyses of chromosomes 9 and 27 further revealed 871 heterogeneous patterns. Haplotype blocks identified as introgressed from dogs through local 872 ancestry inference coincided with regions of strong LD. This elevated LD may be due to reduced 873 recombination rates, which can arise from structural features such as inversions (as seen on 874 chromosome 9) or from selective sweeps maintaining favorable haplotypes. In contrast, LD was 875 generally lower outside the introgressed blocks, consistent with background recombination 876 eroding ancestral haplotypes over time. 877 In addition to introgressed haplotype blocks, we also identified eight ancestry deserts 878 across the genome (Table S6), defined as regions consistently reduced of dog ancestry. The 879 limited number and generally small size of these deserts indicate that resistance to dog 880 introgression does occur in the dingo genome but is relatively uncommon. This pattern is 881 consistent with the recent divergence between dingoes and domestic dogs, which likely 882 constrains the accumulation of strongly deleterious introgressed variants. Supporting this 883 interpretation, variants located within ancestry deserts were predominantly non -coding, and we 884 detected no missense or predicted loss-of-function mutations within these regions. The functional 885 and evolutionary significance of these ancestry deserts, including the roles of individual genes 886 located within them, remains an important topic for future investigation. Together, these results 887 suggest that purifying selection against dog -derived alleles is present but localized, acting on 888 specific genomic regions rather than imposing broad barriers to introgression. 889 4.4 Comparative Insights from Other Canids 890 Comparative evidence from other canids highlights the potential for adaptive introgression to 891 enhance resilience in challenging environments. In North America, for example, introgression 892 has enabled local adaptation in coyotes, where wolf -derived alleles contribute to ecological 893 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint versatility (vonHoldt et al., 2016). In Tibetan dogs, introgression of EPAS1 alleles from high -894 altitude wolves facilitates hypoxia tolerance (Miao et al., 2017; Wang et al., 20 14), while 895 European wolves may exhibit increased tolerance to human -modified landscapes following 896 hybridization with domestic dogs (Pilot et al., 2021; Sarabia et al. 2025; Lobo et al. 2025). 897 Although the functional consequences remain to be fully quantified, the dog -derived alleles 898 identified in dingoes may similarly confer adaptive benefits that improve survival in fragmented 899 or anthropogenically modified habitats. At the same time, similar as in our study, studies in other 900 species have identified genomic regions resistant to introgression (ancestry deserts ), often 901 associated with purifying selection against foreign alleles (Kim et al., 2018). 902 In a broader evolutionary perspective, our results emphasize that introgression can 903 simultaneously represent a source of beneficial genetic diversity and a threat to genomic 904 integrity. We also prov ide strong evidence that anthropogenic habitat modifications (quantified 905 as Human Footprint Index) are one of key factors enhancing introgression from domestic into 906 wild animals, which may both facilitate their adaptation to human -modified landscapes and 907 compromise their function in natural ecosystems. By deepening our understanding of these 908 processes, we can better assess how introgression contributes to adaptation in the face of rapid 909 environmental change and growing anthropogenic pressures. 910 911 4.5 Conservation Implications and Management Strategies 912 The intricate relationship between introgression and environmental adaptation presents both 913 critical challenges and unique opportunities for conserving Australian dingo populations. While 914 introgressed genes under positive selection may confer adaptive benefits -such as enhanced 915 survival in human -dominated landscapes-these potential advantages must be carefully weighed 916 against their ecological consequences. Notably, the acquisition of traits that facilitate utilization 917 of human -dominated areas and consumption of anthropogenic food could alter the dingo’s 918 functional role as an apex predator, potentially disrupting ecosystem dynamics. In our dataset, 919 signals of adaptive introgression appear limited and unlikely to compromise the functional 920 distinctiveness of dingoes. Moreover, we detected eight ancestry deserts, suggesting that 921 deleterious dog -derived alleles are likely purged by purifying selection but occur infrequently 922 and are rapidly eliminated in early -generation hybrids. Large population sizes are required for 923 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint efficient natural selection, therefore attempts to limit dog introgression by lethal control may 924 instead intensify the problem. 925 From a conservation perspective, the Central population exhibited the lowest admixture 926 across all methods, underscoring its value for conservation and as a reference “pure” dingo in 927 comparative studies. Conservation strategies must therefore adopt spatially explicit frameworks 928 that prioritize maintaining genetic diversity and the genetic integrity of dingo lineages while 929 acknowledging that introgression may not pose a major threat. Importantly, decisions regarding 930 population management-such as lethal control -should be guided by a nuanced understanding of 931 genetic ancestry and the functional relevance of introgressed traits, rather than simplistic purity 932 thresholds. Ultimately, by striking this balance, we can support the long-term viability of dingoes 933 across Australia’s diverse ecological landscapes and inform evidence -based approaches to 934 mitigate the genetic and ecological impacts of hybridization in wild canids. 935 936 Supporting Information 937 Supporting information can be found online in the Supporting Information section. It includes 938 the following files: a PDF file containing Supplementary Figures S2-S8, S11, S12, and 939 Supplementary Table 7 as well as all Supplementary Figure and Table legends; Supplementary 940 Tables S1-S6 as separate Excel files, and Supplementary Figures S1, S3, S9 and S10 as separate 941 PDF files. 942 943 Author contributions 944 MP and TN designed research, K MC performed lab work, BJN, KC and TMN contributed the 945 samples, AFM collected environmental data, COM and KD analyzed data, COM wrote the 946 manuscript, all the authors contributed to the manuscript revision. 947 948 Conflict of Interest 949 KMC is co -coordinator of the IUCN Species Survival Commission (SSC) Canid Specialist 950 Group’s Dingo Working Group and a voluntary scientific advisor to the Australian Dingo 951 Foundation and New Guinea Highland Wild Dog Foundation. 952 953 954 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Data Accessibility 955 The dataset that support the findings of this study, including PLINK files, as well as an HTML 956 file containing reproducible scripts and code used for the analyses, are openly available at 957 FigShare: https://figshare.com/s/0ef4bf5568234b8ad40f. The repository contains a README file to 958 guide users through the dataset structure and content, ensuring the analyses can be reproduced 959 accurately. 960 961

962 ABARES (2022). Land use of Australia 2010 –11 to 2015 –16, 250 m, Australian Bureau of 963 Agricultural and Resource Economics and Sciences, Canberra, September, CC BY 4.0. 964 DOI: 10.25814/7ygw-4d64 965 Abbott, R., Albach, D., Ansell, S., Arntzen, J. W., Baird, S. J., Bierne, N., ... & Zinner, D. 966 (2013). Hybridization and speciation. Journal of Evolutionary Biology, 26 (2), 229-246. 967 ABS (2023). Australian Bureau of Statistics. Regional population. 968 https://www.abs.gov.au/statistics/people/population/regional-population/latest-release. 969 Adavoudi, R., & Pilot, M. (2021). Consequences of hybridization in mammals: A systematic 970 review. Genes, 13 (1), 50. 971 Aguillon, S. M., Dodge, T. O., Preising, G. A., & Schumer, M. (2022). Introgression. Current 972 Biology, 32 (16), R865-R868. 973 Alexander, D. H., & Lange, K. (2011). Enhancements to the ADMIXTURE algorithm for 974 individual ancestry estimation. BMC Bioinformatics, 12, 1-6. 975 Amiri Ghanatsaman, Z., Wang, G. D., Asadollahpour Nanaei, H., Asadi Fozi, M., Peng, M. S., 976 Esmailizadeh, A., & Zhang, Y. P. (2020). Whole genome resequencing of the Iranian 977 native dogs and wolves to unravel variome during dog domestication. BMC Genomics, 21, 978 1-11. 979 Anderson, T. M., vonHoldt, B. M., Candille, S. I., Musiani, M., Greco, C., Stahler, D. R., ... & 980 Barsh, G. S. (2009). Molecular and evolutionary history of melanism in North American 981 gray wolves. Science, 323 (5919), 1339-1343. 982 Battilani, D., Ramos Madrigal, J., Hennelly, L. M., Gopalakrishnan, S., Vernesi, C., Mattucci, F., 983 ... & Caniglia, R. (2025). Whole ‐Genome Data to Investigate Recent and Historical Dog 984 Introgression Patterns in Italian Wolves. Ecology and Evolution, 15 (12), e72508. 985 Blumer, L. M., Burskaia, V., Artiushin, I., Saha, J., Camacho Garcia, J., Elkin, J., ... & Svardal, 986 H. (2024). Introgression dynamics of sex-linked chromosomal inversions shape the Malawi 987 cichlid adaptive radiation. bioRxiv, 2024-07. 988 Boronyak, L., Jacobs, B., & Smith, B. (2023). Unlocking lethal dingo management in 989 Australia. Diversity, 15 (5), 642. 990 Browning, B. L., Tian, X., Zhou, Y., & Browning, S. R. (2021). Fast two -stage phasing of large-991 scale sequence data. The American Journal of Human Genetics, 108 (10), 1880-1890. 992 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Burri, R., Nater, A., Kawakami, T., Mugal, C. F., Olason, P. I., Smeds, L., ... & Ellegren, H. 993 (2015). Linked selection and recombination rate variation drive the evolution of the 994 genomic landscape of differentiation across the speciation continuum of Ficedula 995 flycatchers. Genome Research, 25 (11), 1656-1665. 996 Capblancq, T., & Forester, B. R. (2021). Redundancy analysis: A Swiss Army Knife for 997 landscape genomics. Methods in Ecology and Evolution, 12 (12), 2298-2309. 998 Cairns, K. M., Brown, S. K., Sacks, B. N., & Ballard, J. W. O. (2017). Conservation implications 999 for dingoes from the maternal and paternal genome: Multiple populations, dog 1000 introgression, and demography. Ecology and Evolution, 7 (22), 9787-9807. 1001 Cairns, K. M., Nesbitt, B. J., Laffan, S. W., Letnic, M., & Crowther, M. S. (2020). Geographic 1002 hot spots of dingo genetic ancestry in southeastern Australia despite hybridisation with 1003 domestic dogs. Conservation Genetics, 21 (1), 77-90. 1004 Cairns, K. M. (2021). What is a dingo –origins, hybridisation and identity. Australian 1005 Zoologist, 41 (3), 322-337. 1006 Cairns, K. M., Crowther, M. S., Parker, H. G., Ostrander, E. A., & Letnic, M. (2023). Genome‐1007 wide variant analyses reveal new patterns of admixture and population structure in 1008 Australian dingoes. Molecular Ecology, 32 (15), 4133-4150. 1009 Castle, G., Kennedy, M. S. Allen, B. L. (2023). Stuck in the mud: Persistent failure of “the 1010 science” to provide reliable information on the ecological roles of Australian dingoes. 1011 Biological Conservation, 285, 110234. 1012 Charif, D., Lobry, J. R., Necsulea, A., Palmeira, L., Penel, S., Perriere, G., & Penel , M. S. 1013 (2023). Package ‘seqinr’. 1014 Corbett, L. K. (1995). The dingo in Australia and Asia. 1015 Crowther, M. S., Cairns, K. M., Van Eeden, L. M., & Letnic, M. (2021). Introgression does not 1016 influence the positive ecological and functional role of dingo populations. Australian 1017 Zoologist, 41 (3), 338-346. 1018 Doherty, T. S., Davis, N. E., Dickman, C. R., Forsyth, D. M., Letnic, M., Nimmo, D. G., ... & 1019 Newsome, T. M. (2019). Continental patterns in the diet of a top predator: Australia's 1020 dingo. Mammal Review, 49 (1), 31-44. 1021 Dobzhansky, T. (1937). Genetic nature of species differences. The American Naturalist , 71 1022 (735), 404-420. 1023 Du Sert, N. P., Hurst, V., Ahluwalia, A., Alam, S., Avey, M. T., Baker, M., ... & Würbel, H. 1024 (2020). The ARRIVE guidelines 2.0: Updated guidelines for reporting animal research. 1025 Elledge, A. E., Leung, L. K. P., Allen, L. R., Firestone, K., & Wilton, A. N. (2006). Assessing 1026 the taxonomic status of dingoes Canis familiaris dingo for conservation. Mammal 1027 Review, 36 (2), 142-156. 1028 Ellstrand, N. C., & Elam, D. R. (1993). Population genetic consequences of small population 1029 size: implications for plant conservation. Annual Review of Ecology and Systematics , 24 1030 (1), 217-242. 1031 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Faske, T. M., Agneray, A. C., Jahner, J. P., Osuna -Mascaró, C., Sheta, L. M., Richardson, B. A., 1032 & Parchman, T. L. (2024). Environment predicts the maintenance of reproductive isolation 1033 in a mosaic hybrid zone of rubber rabbitbrush. Evolution, 78 (2), 300–314. 1034 Fillios, M. A., Taçon , P. S. C. (2016). Who let the dogs in? A review of the recent genetic 1035 evidence for the introduction of the dingo to Australia and implications for the movement 1036 of people. J. Archaeol. Sci. Rep. 7, 782–792. 1037 Field, M. A., Yadav, S., Dudchenko, O., Esvaran, M., Rosen, B. D., Skvortsova, K., ... & 1038 Ballard, J. W. O. (2022). The Australian dingo is an early offshoot of modern breed 1039 dogs. Science Advances, 8 (16), eabm5944. 1040 Forester, B. R., Lasky, J. R., Wagner, H. H., & Urban, D. L. (2018). Comparing methods for 1041 detecting multilocus adaptation with multivariate genotype –environment associations. 1042 Molecular Ecology, 27 (9), 2215-2233. 1043 Geffen, E. L. I., Anderson, M. J., & Wayne, R. K. (2004). Climate and habitat barriers to 1044 dispersal in the highly mobile grey wolf. Molecular Ecology, 13 (8), 2481-2490. 1045 Giraud, T., Refrégier, G., Le Gac, M., de Vienne, D. M., & Hood, M. E. (2008). Speciation in 1046 fungi. Fungal Genetics and Biology, 45 (6), 791-802. 1047 Glen, A. S., Dickman, C. R., Soule, M. E., & Mackey, B. G. (2007). Evaluating the role of the 1048 dingo as a trophic regulator in Australian ecosystems. Austral Ecology, 32 (5), 492-501. 1049 Gompert, Z., & Buerkle, C. A. (2011). Bayesian estimation of genomic clines. Molecular 1050 Ecology, 20 (10), 2111-2127. 1051 Gompert, Z., DeRaad, D. A., & Buerkle, C. A. (2024). A Next Generation of Hierarchical 1052 Bayesian Analyses of Hybrid Zones Enables Model‐Based Quantification of Variation in 1053 Introgression in R. Ecology and Evolution, 14 (11), e70548. 1054 Gompper, M. E. (2014). The dog -human-wildlife interface: assessing the scope of the problem. 1055 In: Free-ranging Dogs and Wildlife Conservation, 1, 9-54. 1056 Goudet, J. (2005). Hierfstat, a package for R to compute and test hierarchical F‐statistics. 1057 Molecular Ecology Notes, 5 (1), 184-186 1058 Grabenstein, K. C., Otter, K. A., Burg, T. M., & Taylor, S. A. (2023). Hybridization between 1059 closely related songbirds is related to human habitat disturbance. Global Change Biology , 1060 29 (4), 955–968. 1061 Grant, B. R., & Grant, P. R. (1979). Darwin's finches: population variation and sympatric 1062 speciation. Proceedings of the National Academy of Sciences, 76 (5), 2359-2363 1063 Grant, B. R., & Grant, P. R. (2017). Watching speciation in action. Science, 355 (6328), 910 -1064 911. 1065 Guan, Y. (2014). Detecting structure of haplotypes and local ancestry. Genetics, 196 (3), 625 -1066 642. 1067 Hardy, A. L., Gaither, M. R., Lotterhos, K. E., Greaves, S., Cipolla, K. J., Kerns, E. V., Prieto 1068 Trujillo, A., & Gilg, M. R. (2021). Asymmetrical hybridization and environmental factors 1069 influence the spatial genetic structure of a killifish hybrid zone. Molecular Ecology, 30 1070 (13), 3215–3231. 1071 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Hijmans, R. J., Cameron, S. E., Parra, J. L., Jones, P. G., & Jarvis, A. (2005). Very high 1072 resolution interpolated climate surfaces for global land areas. International Journal of 1073 Climatology: A Journal of the Royal Meteorological Society, 25 (15), 1965-1978. 1074 Hijmans, R. J., Williams, E., Vennes, C., & Hijmans, M. R. J. (2017). Package 1075 ‘geosphere’. Spherical trigonometry, 1(7), 1-45. 1076 Hohenlohe, P. A., Funk, W. C., & Rajora, O. P. (2021). Population genomics for wildlife 1077 conservation and management. Molecular Ecology, 30 (1), 62-82. 1078 Kato, S., Arakaki, S., Nagano, A. J., Kikuchi, K., & Hirase, S. (2024). Genomic landscape of 1079 introgression from the ghost lineage in a gobiid fish uncovers the generality of forces 1080 shaping hybrid genomes. Molecular Ecology, 33 (20), e17216. 1081 Katoh, K., & Standley, D. M. (2014). MAFFT: iterative refinement and additional methods. 1082 Multiple Sequence Alignment Methods, 131-146. 1083 Kidd, J. M. (2026). Fine -scale recombination rates inferred using the canFam4 assembly are 1084 strongly correlated with previous maps of dog recombination. Mammalian Genome, 37 (1), 1085 12. 1086 Kinneberg, V. B., Lü, D. S., Peris, D., Ravinet, M., & Skrede, I. (2023). Introgression between 1087 highly divergent fungal sister species. Journal of Evolutionary Biology, 36 (8), 1133-1149. 1088 Kim, B. Y., Huber, C. D., & Lohmueller, K. E. (2018). Deleterious variation shapes the genomic 1089 landscape of introgression. PLoS Genetics, 14 (10), e1007741. 1090 Kryazhimskiy, S., & Plotkin , J. B. (2008). The population genetics of dN/dS. PLoS Genetics, 4 1091 (12), e1000304. 1092 Kuijper, D. P., Sahlén, E., Elmhagen, B., Chamaillé -Jammes, S., Sand, H., Lone, K., & 1093 Cromsigt, J. P. G. M. (2016). Paws without claws? Ecological effects of large carnivores in 1094 anthropogenic landscapes. Proceedings of the Royal Society B: Biological Sciences , 283 1095 (1841), 20161625. 1096 Kumar, M., Conroy, G., Ogbourne, S., Cairns, K., Borburgh, L., & Subramanian, S. (2023). 1097 Genomic signatures of bottleneck and founder effects in dingoes. Ecology and Evolution , 1098 13 (9), e10525. 1099 Leonard, J. A. (2015). Ecology drives evolution in grey wolves. Evolutionary Ecology Research, 1100 16 (6), 461-473. 1101 Letnic, M., & Koch, F. (2010). Are dingoes a trophic regulator in arid Australia? A comparison 1102 of mammal communities on either side of the dingo fence. Austral Ecology, 35 (2), 167 -1103 175. 1104 Liaw, A., & Wiener, M. (2002). Classification and regression by randomForest. R new s, 2(3), 1105 18-22. 1106 Lobo, D., Morales, H. E., Van Oosterhout, C., López -Bao, J. V., Silva, P., Llaneza, L., ... & 1107 Godinho, R. (2025). Ancient dog introgression into the Iberian wolf genome may have 1108 facilitated adaptation to human-dominated landscapes. Genome Research, 35(3), 432-445. 1109 McLaren, William, et al. "The ensembl variant effect predictor." Genome Biology 17.1 (2016): 1110 122. 1111 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Mallet, J. (2007). Hybrid speciation. Nature, 446 (7133), 279-283. 1112 Malinsky, M., Matschiner, M., & Svardal, H. (2021). Dsuite‐Fast D‐statistics and related 1113 admixture evidence from VCF files. Molecular Ecology Resources, 21 (2), 584-595. 1114 Mandeville, E. G., Walters, A. W., Nordberg, B. J., Higgins, K. H., Burckhardt, J. C., & Wagner, 1115 C. E. (2019). Variable hybridization outcomes in trout are predicted by environmental 1116 context and historical demographic processes. Molecular Ecology, 28 (16), 3738–3755. 1117 Martin, S. H., Davey, J. W., & Jiggins, C. D. (2015). Evaluating the use of ABBA –BABA 1118 statistics to locate introgressed loci. Molecular Biology and Evolution, 32 (1), 244-257. 1119 Mayr, E. (1942). Birds collected during the Whitney South Sea Expedition. 48, Notes on the 1120 Polynesian species of Aplonis. American Museum novitates; no. 1166. 1121 McFarlane, S. E., & Pemberton, J. M. (2019). Detecting the true extent of introgression during 1122 anthropogenic hybridization. Trends in Ecology & Evolution, 34 (4), 315-326. 1123 Miao, B., Wang, Z., & Li, Y. (2017). Genomic analysis reveals hypoxia adaptation in the Tibetan 1124 mastiff by introgression of the gray wolf from the Tibetan Plateau. Molecular Biology and 1125 Evolution, 34 (3), 734-743. 1126 Minh, B. Q., Schmidt, H. A., Chernomor, O., Schrempf, D., Woodhams, M. D., Von Haeseler, 1127 A., & Lanfear, R. (2020). IQ -TREE 2: new models and efficient methods for phylogenetic 1128 inference in the genomic era. Molecular Biology and Evolution, 37 (5), 1530-1534 1129 Morrill, K., Hekman, J., Li, X., McClure, J., Logan, B., Goodman, L., ... & Karlsson, E. K. 1130 (2022). Ancestry -inclusive dog genomics challenges popular breed 1131 stereotypes. Science, 376 (6592), eabk0639. 1132 Mouton, A., Bird, D. J., Li, G., Craven, B. A., Levine, J. M., Morselli, M., ... & Murphy, W. J. 1133 (2025). Genetic and anatomical determinants of olfaction in dogs and wild 1134 canids. Molecular Biology and Evolution, 42 (3), msaf035. 1135 Narasimhan, V., Danecek, P., Scally, A., Xue, Y., Tyler -Smith, C., & Durbin, R. (2016). 1136 BCFtools/RoH: a hidden Markov model approach for detecting autozygosity from next -1137 generation sequencing data. Bioinformatics, 32 (11), 1749-1751. 1138 Newsome, T. M., Dellinger, J. A., Pavey, C. R., Ripple, W. J., Shores, C. R., Wirsing, A. J., & 1139 Dickman, C. R. (2015a). The ecological effects of providing resource subsidies to 1140 predators. Global Ecology and Biogeography, 24 (1), 1-11. 1141 Newsome, T. M., Ballard, G. A., Crowther, M. S., Dellinger, J. A., Fleming, P. J., Glen, A. S., ... 1142 & Dickman, C. R. (2015b). Resolving the value of the dingo in ecological 1143 restoration. Restoration Ecology, 23 (3), 201-208. 1144 Newsome, T. M., Fleming, P. J., Dickman, C. R., Doherty, T. S., Ripple, W. J., Ritchie, E. G., & 1145 Wirsing, A. J. (2017). Making a new dog?. BioScience, 67 (4), 374-381. 1146 Nguyen, L. T., Schmidt, H. A., Von Haeseler, A., & Minh, B. Q. (2015). IQ -TREE: a fast and 1147 effective stochastic algorithm for estimating maximum -likelihood phylogenies. Molecular 1148 biology and evolution, 32 (1), 268-274. 1149 Olender, T., Fuchs, T., Linhart, C., Shamir, R., Adams, M., Kalush, F., ... & Lancet, D. (2004). 1150 The canine olfactory subgenome. Genomics, 83 (3), 361-372. 1151 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Oksanen, J., Blanchet, F. G., Kindt, R., Legendre, P., Minchin, P. R., O’hara, R. B., ... & 1152 Oksanen, M. J. (2013). Package ‘vegan.’ Community Ecology package , version, 2 (9), 1 -1153 295. 1154 Paradis, E., & Schliep, K. (2019). ape 5.0: an environment for modern phylogenetics and 1155 evolutionary analyses in R. Bioinformatics, 35(3), 526-528. 1156 Parr, William CH, et al. "Cranial shape and the modularity of hybridization in dingoes and dogs; 1157 hybridization does not spell the end for native morphology." Evolutionary Biolog y 43.2 1158 (2016): 171-187. 1159 Patterson, N., Price, A. L., & Reich, D. (2006). Population structure and eigenanalysis. PLoS 1160 Genetics, 2 (12), e190. 1161 Pilot, M., J ędrzejewski, W., Branicki, W., Sidorovich, V. E., J ędrzejewska, B., Stachura, K., & 1162 Funk, S. M. (2006). Ecological factors influence population genetic structure of European 1163 grey wolves. Molecular Ecology, 15 (14), 4533-4553. 1164 Pilot, M., Greco, C., vonHoldt, B. M., Randi, E., Jędrzejewski, W., Sidorovich, V. E., ... & 1165 Wayne, R. K. (2018). Widespread, long‐term admixture between grey wolves and 1166 domestic dogs across Eurasia and its implications for the conservation status of hybrids. 1167 Evolutionary Applications, 11 (5), 662-680. 1168 Pilot, M., Moura, A. E., Okhlopkov, I. M., Mamaev, N. V., Manaseryan, N. H., Hayrapetyan, V., 1169 ... & Bogdanowicz, W. (2021). Human‐modified canids in human‐modified landscapes: 1170 The evolutionary consequences of hybridization for grey wolves and free‐ranging domestic 1171 dogs. Evolutionary Applications, 14 (10), 2433-2456. 1172 Pollock, T. I., Hunter, D. O., Hocking, D. P., & Evans, A. R. (2022). Eye in the sky: observing 1173 wild dingo hunting behaviour using drones. Wildlife Research, 50 (3), 212-223. 1174 Purcell, S., Neale, B., Todd -Brown, K., Thomas, L., Ferreira, M. A., Bender, D., ... & Sham, P. 1175 C. (2007). PLINK: a tool set for whole -genome association and population -based linkage 1176 analyses. The American Journal of Human Genetics, 81(3), 559-575. 1177 Quinlan, A. R. (2014). BEDTools: the Swiss‐army tool for genome feature analysis. Current 1178 Protocols in Bioinformatics, 47 (1), 11-12. 1179 R Core Team (2013) R: A Language and Environment for Statistical Computing. R Foundation 1180 for Statistical Computing, Vienna. http://www.R-project.org/ 1181 Rambaut, A. (2009). FigTree. Tree figure drawing tool. http://tree. bio. ed. ac. 1182 uk/software/figtree/. 1183 Raney, B. J., Barber, G. P., Benet -Pagès, A., Casper, J., Clawson, H., Cline, M. S., ... & 1184 Haeussler, M. (2024). The UCSC genome browser database: 2024 update. Nucleic Acids 1185 Research, 52 (D1), D1082-D1088. 1186 Revelle, W., & Revelle, M. W. (2015). Package ‘psych’. The comprehensive R archive network , 1187 337 (338), 161-165. 1188 Rieseberg, L., & Wendel, J. (1993). lntrogression and Its Consequences in Plants. 1189 Robin, S., Tacher, S., Rimbault, M., Vaysse, A., Dréano, S., André, C., ... & Galibert, F. (2009). 1190 Genetic diversity of canine olfactory receptors. BMC Genomics, 10, 1-16. 1191 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Safran, M., Dalah, I., Alexander, J., Rosen, N., Iny Stein, T., Shmoish, M., ... & Lancet, D. 1192 (2010). GeneCards Version 3: the human gene integrator. Database, 2010, baq020. 1193 Sankararaman, S., Sridhar, S., Kimmel, G., & Halperin, E. (2008). Estimating local ancestry in 1194 admixed populations. The American Journal of Human Genetics, 82 (2), 290-303. 1195 Sankararaman, S., Mallick, S., Dannemann, M., Prüfer, K., Kelso, J., Pääbo , S., ... & Reich, D. 1196 (2014). The genomic landscape of Neanderthal ancestry in present -day 1197 humans. Nature, 507 (7492), 354-357. 1198 Sarabia, C., Salado, I., Fernández-Gil, A., vonHoldt, B. M., Hofreiter, M., Vilà, C., & Leonard, J. 1199 A. (2025). Potential Adaptive Introgression From Dogs in Iberian Grey Wolves ( Canis 1200 lupus). Molecular Ecology, e17639. 1201 Scarsbrook, L., Cairns, K. M., Mitchell, K. J., Bougiouri, K., Evin, A., Harris, A. C., ... & Fillios, 1202 M. A. (2025). The impacts of European arrival on Australian dingoes. Proceedings of the 1203 National Academy of Sciences, 122(48), e2421749122. 1204 Schiavina, M., Melchiorri, M., Pesaresi, M., Politis, P., Freire, S., Maffenini, L., ... & Kemper, T. 1205 (2022). GHSL data package 2022. Publications Office of the European Union: 1206 Luxembourg. 1207 Seehausen, O. (2004). Hybridization and adaptive radiation. Trends in Ecology & Evolution , 19 1208 (4), 198-207. 1209 Singhal, S., Derryberry, G. E., Bravo, G. A., Derryberry, E. P., Brumfield, R. T., & Harvey, M. 1210 G. (2021). The dynamics of introgression across an avian radiation. Evolution Letters , 5 1211 (6), 568-581. 1212 Souilmi, Y., Wasef, S., Williams, M. P., Conroy, G., Bar, I., Bover, P., ... & Mitchell, K. J. 1213 (2024). Ancient genomes reveal over two thousand years of dingo population 1214 structure. Proceedings of the National Academy of Sciences, 121 (30), e2407584121. 1215 Sparkes, J., Körtner, G., Ballard, G., & Fleming, P. J. (2022). Spatial and temporal activity 1216 patterns of owned, free -roaming dogs in coastal eastern Australia. Preventive Veterinary 1217 Medicine, 204, 105641. 1218 Spatola, G.J., Buckley, R.M., Dillon, M., Dutrow, E.V., Betz, J.A., Pilot, M., Parker, H.G., 1219 Bogdanowicz, W., Thomas, R., Chyzhevskyi, I. , Milinevsky, G., Kleiman, N., Breen, M., 1220 Ostrander, E.A., Mousseau, T.A. (2023). The dogs of Chernobyl: Demographic insights 1221 into populations inhabiting the nuclear exclusion zone. Science Advances , 9 (9), 1222 p.eade2537. 1223 Stebbins, G. L. (1959). The role of hybridization in evolution. Proceedings of the American 1224 Philosophical Society, 103 (2), 231-251. 1225 Stefanović, M., Bogdanowicz, W., Adavoudi, R., Martínez -Sosa, F., Doan, K., Flores -1226 Manzanero, A., ... & Pilot, M. (2024). Range -wide phylogeography of the golden jackals 1227 (Canis aureus ) reveals multiple sources of recent spatial expansion and admixture with 1228 dogs at the expansion front. Biological Conservation, 290, 110448. 1229 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Stephens, D., Wilton, A. N., Fleming, P. J., & Berry, O. (2015). Death by sex in an Australian 1230 icon: A continent‐wide survey reveals extensive hybridization between dingoes and 1231 domestic dogs. Molecular Ecology, 24 (22), 5643-5656. 1232 Stephens, D., Fleming, P. J., Sawyers, E., & Mayr, T. P. (2022). An isolated population reveals 1233 greater genetic structuring of the Australian dingo. Scientific Reports, 12 (1), 19105. 1234 Stronen, A. V., Tessier, N., Jolicoeur, H., Paquet, P. C., Henault, M., Villemure, M., ... & 1235 Lapointe, F. J. (2012). Canid hybridization: Contemporary evolution in human‐modified 1236 landscapes. Ecology and Evolution, 2 (9), 2128-2140. 1237 Stull, G. W., Pham, K. K., Soltis, P. S., & Soltis, D. E. (2023). Deep reticulation: the long legacy 1238 of hybridization in vascular plant evolution. The Plant Journal, 114 (4), 743-76 1239 Surbakti, S., Parker, H. G., McIntyre, J. K., Maury, H. K., Cairns, K. M., Selvig, M., ... & 1240 Ostrander, E. A. (2020). New Guinea highland wild dogs are the original New Guinea 1241 singing dogs. Proceedings of the National Academy of Sciences, 117 (39), 24369-24376. 1242 Tensen, L., & Fischer, K. (2024). Evaluating hybrid speciation and swamping in wild carnivores 1243 with a decision‐tree approach. Conservation Biology, 38 (1), e14197. 1244 Thomas, C. D. (2015). Rapid acceleration of plant speciation during the Anthropocene. Trends in 1245 Ecology & Evolution, 30 (8), 448-455. 1246 Utsunomiya, Y. T., Milanesi, M., Utsunomiya, A. T., Ajmone-Marsan, P., & Garcia, J. F. (2016). 1247 GHap: an R package for genome-wide haplotyping. Bioinformatics, 32 (18), 2861-2862. 1248 vonHoldt, B. M., Kays, R., Pollinger, J. P., & Wayne, R. K. (2016). Admixture mapping 1249 identifies introgressed genomic regions in North American canids. Molecular Ecology, 25 1250 (11), 2443-2453. 1251 Wang, I. J., & Bradburd, G. S. (2014). Isolation by environment. Molecular Ecology, 23 (23), 1252 5649-5662. 1253 Wang, G. D., Fan, R. X., Zhai, W., Liu, F., Wang, L., Zhong, L., ... & Zhang, Y. P. (2014). 1254 Genetic convergence in the adaptation of dogs and humans to the high -altitude 1255 environment of the Tibetan plateau. Genome Biology and Evolution, 6 (8), 2122-2128. 1256 Weeks, A. R., Kriesner, P., Bartonicek, N., van Rooyen, A., Cairns, K. M., & Ahrens, C. W. 1257 (2024). Genetic structure and common ancestry expose the dingo -dog hybrid myth. 1258 Evolution Letters, qrae057. 1259 Wheeldon, T. J., Rutledge, L. Y., Patterson, B. R., White, B. N., & Wilson, P. J. (2013). Y‐1260 chromosome evidence supports asymmetric dog introgression into eastern coyotes. 1261 Ecology and Evolution, 3 (9), 3005-3020. 1262 Wickham, H., Chang, W., & Wickham, M. H. (2016). Package ‘ggplot2’. Create elegant data 1263 visualisations using the grammar of graphics. Version, 2 (1), 1-189. 1264 Wilton, A. N. (2001). DNA methods of assessing dingo purity. In A Symposium on the 1265 Dingo (pp. 49-56). PO Box 20, Mosman NSW 2088, Australia: Royal Zoological Society 1266 of New South Wales. 1267 Wright, S. (1943). Isolation by distance. Genetics, 28(2), 114. 1268 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Ye, Y., Jiang, Z., Pan, Y., Yang, G., & Wang, Y. (2022). Role and mechanism of BMP4 in bone, 1269 craniofacial, and tooth development. Archives of Oral Biology, 140, 105465. 1270 Zhang, S. J., Wang, G. D., Ma, P., Zhang, L. L., Yin, T. T., Liu, Y. H., ... & Zhang, Y. P. (2020). 1271 Genomic regions under selection in the feralization of the dingoes. Nature 1272 Communications, 11 (1), 671. 1273 1274 1275 1276 1277 1278

1279 We thank Prof. Josephine Pemberton and two anonymous reviewers for their constructive 1280 comments on the manuscript. 1281 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Table legends 1282 Table 1. Sample size (N) and mean proportion of dog ancestry in dingoes, hybrids and reference 1283 dog groups, as estimated by four analytical methods: ADMIXTURE (global ancestry), GHap 1284 (haplotype‐block sharing), LAMP (LD‐based local ancestry) and ELAI (two‐layer local 1285 ancestry). Values represent the average fraction of dog‐derived genome across individuals in 1286 each population. 1287 Table 2. Summary of ABBA –BABA (D -statistic) tests of introgression between dingoes and 1288 European dogs, conducted with Dsuite on (i) whole‐genome VCF data and (ii) a candidate 1289 introgressed block on chromosome 27. In each test, P1 was the New Guinea singing dog (NGSD; 1290 the sister lineage to dingoes), P2 was either all dingoes combined or one of six regional dingo 1291 populations (East, South, Captive, Big Desert, Central, West), and P3 was the pure breed 1292 European dog samples (Eu. Dog), with the jackal as an outgroup. For each trio we report 1293 Patterson’s D (excess allele sharing between P2 and P3 versus P1 and P3), its Z‐score (block‐1294 jackknife), two‐tailed p-value, the f₄-ratio (proportion of P3 ancestry in P2), and counts of BBAA 1295 (derived allele shared by P1+P2), ABBA (shared by P2+P3) and BABA (shared by P1+P3) site 1296 patterns. In the genome‐wide analyses, all D values are significantly positive (p < 0.001), 1297 indicating gene flow between dingoes and European dogs. For chromosome 27, only trios with 1298 informative ABBA/BABA sites and p < 0.05 are shown. Elevated D and f₄ -ratio values on 1299 chromosome 27 relative to genome‐wide averages highlight a localized introgression signal in 1300 this candidate region. 1301 Table 3. Results of partial Redundancy Analysis (partial RDA) examining the influence of 1302 climate, genetic structure, and geography on genetic variation. The analysis decomposes the 1303 variation into components attributable to climate, genetic structure, and geography. The 1304 proportion of explainable variance represents the total constrained variation explained by the full 1305 model. 1306 1307 1308 1309 1310 1311 1312 1313 1314 1315 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Tables 1316 1317 Table 1. 1318 1319 Population N ADMIXTURE GHap LAMP ELAI All dingoes 390 0.091 0.123 0.151 0.151 Central 36 0.006 0.030 0.041 0.031 West 109 0.037 0.061 0.079 0.075 Big Desert 15 0.016 0.110 0.127 0.130 North 49 0.116 0.149 0.181 0.186 South 66 0.132 0.162 0.203 0.197 East 79 0.166 0.183 0.224 0.228 Captive 36 0.102 0.169 0.190 0.198 Dingo-dog F1 hybrid 2 0.557 0.568 0.588 0.597 Australian mixed-breed dogs 21 0.956 0.985 0.987 0.989 Australian pure-bred dogs 33 0.980 0.993 0.993 0.998 European pure-bred dogs 226 0.989 0.999 0.998 1 European free-ranging dogs 116 0.968 0.999 0.998 1 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint 1332 Table 2. 1333 1334 1335 P1 P2 P3 D Z p-value f4 BBAA ABBA BABA Whole genome NGSD Dingo (all) Eu. Dog 0.154 7.199 6.03E-13 0.192 1703.34 725.536 531.154 NGSD Dingo East Eu. Dog 0.192 8.528 2.30E-16 0.258 1647.38 806.038 545.697 NGSD Dingo South Eu. Dog 0.179 7.077 1.47E-12 0.235 1672.9 778.901 541.886 NGSD Dingo Captive Eu. Dog 0.172 6.654 2.84E-11 0.220 1681.39 756.018 533.922 NGSD Dingo Big Desert Eu. Dog 0.160 4.464 8.02E-06 0.207 1674.29 756.603 547.059 NGSD Dingo Central Eu. Dog 0.139 6.325 2.53E-10 0.166 1736.97 685.937 518.038 NGSD Dingo West Eu. Dog 0.118 5.932 2.99E-09 0.140 1738.07 665.94 524.573 Chromosome 27 block NGSD Dingo (all) Eu. Dog 0.262 3.345 0.0008 0.556 34.826 28.190 16.461 NGSD Dingo Central Eu. Dog 0.282 3.577 0.0003 0.621 33.992 29.805 16.689 NGSD Dingo South Eu. Dog 0.234 2.834 0.0045 0.498 33.200 27.667 17.141 NGSD Dingo West Eu. Dog 0.209 2.688 0.0071 0.385 39.395 23.437 15.302 1336 1337 1338 1339 1340 1341 1342 1343 1344 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint 1345 Table 3. 1346 1347 Partial RDA models Variance R2 p value Proportion of explainable variance Proportion of total variance Full model: F ~ clim. + geog. + struct. 339.0 9.74E-02 0.001 1 0.097 Pure climate: F ~ clim. | (geog. + struct.) 160 4.86E-02 0.001 0.471 0.049 Pure structure: F ~ struct. | (clim. + geog.) 104 2.77E-02 0.001 0.307 0.028 Pure geography: F ~ geog. | (clim. + struct.) 67.4 1.79E-02 0.001 0.198 0.018 Confounded climate/structure/geography 408 Total unexplained 3210 Total variance 3720 1348 1349 1350 1351 1352 1353 1354 1355 1356 1357 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Figure legends 1358 1359 Figure 1. Population structure and genetic differentiation of dingoes and domestic dogs 1360 across Australia. (a) Principal Component Analysis (PCA) including dingoes, domestic dogs, 1361 and putative hybrids. Dingoes are shown in yellow, Australian domestic dogs (purebred and 1362 mixed-breed) in green, European free-ranging dogs in dark blue, European purebred dogs in light 1363 blue, and hybrids in red. The PCA highlights the strong genetic differentiation between dingoes 1364 and domestic dogs, as well as individuals with intermediate genotypes. The percentage of 1365 variance explained by each axis is indicated. (b) PCA of dingo populations only, colored by 1366 geographic group (Big Desert, Captive, Central, East, North, South, and West), showing fine -1367 scale population structure within Australia. (c) Geographic distribution of dingo populations 1368 across Australia. Pie charts at each sampling location represent the proportion of ancestry from 1369 the four genetic clusters (K = 4) inferred by ADMIXTURE (see Figure S2). (d) Geographic 1370 distribution of local ancestry inferred with ELAI. Pie charts represent the estimated proportion of 1371 dingo (green) and domestic dog (blue) ancestry for each individual, illustrating the spatial 1372 heterogeneity of introgression across Australia. 1373 Figure 2. Dog ancestry across chromosomes inferred using haplotype sharing and local 1374 ancestry analyses. (a) Genome-wide distribution of haplotype blocks shared between dingoes 1375 and domestic dogs inferred with GHap. The y-axis indicates the number of dog-shared haplotype 1376 blocks per genomic window. Chromosomes 9 and 27 show pronounced peaks, indicating regions 1377 with elevated apparent dog ancestry. (b) Local ancestry inference along chromosome 9 based on 1378 ELAI. The y -axis represents the proportion of dog ancestry along the chromosome. The black 1379 dot–dash line indicates the genome -wide average introgression level, and the blue dashed line 1380 indicates the chromosome-specific mean. The region initially identified as a putative introgressed 1381 block encompasses a large chromosomal inversion previously described in domestic dogs (Field 1382 et al., 2022), whose coordinates are indicated by the arrow and likely explain the observed 1383 ancestry pattern. (c) Local ancestry inference along chromosome 27 inferred with ELAI. Line 1384 types and thresholds are as in panel (b). (d) Comparison of local ancestry estimates across 1385 genomic scales. Distribution of individual -level proportions of dog ancestry in dingoes inferred 1386 using LAMP and ELAI at three genomic scales: genome -wide, chromosome 9, and chromosome 1387 27. Points represent individuals (jittered for visibility). Boxplots show medians and interquartile 1388 ranges; white circles indicate means and vertical bars represent ±1 standard deviation. Both mean 1389 ancestry and variance increase markedly on chromosomes 9 and 27 relative to genome -wide 1390 estimates. 1391 Figure 3. Phylogenetic analyses. (a) Maximum Likelihood phylogeny reconstructed with IQ -1392 TREE using genome -wide SNP data from a subset of five randomly selected individuals per 1393 population or species, except for New Guinea Singing Dogs (NGSD; n = 2) and the golden 1394 jackal (outgroup; n = 1). The tree shows species - and population -level clustering. Ultrafast 1395 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint bootstrap support values (UFBS) are shown at the nodes, and the scale bar represents the 1396 expected number of nucleotide substitutions per site. (b) Maximum Likelihood phylogeny for the 1397 introgressed block on chromosome 27, using the same subset of individuals. Clustering of 1398 domestic dogs and dingoes in this region is consistent with localized introgression. The full 1399 phylogenies including all individuals are provided in the Supplementary Material ( Figure S10 a-1400 c). (c) Chromosome 9 inversion region: ML phylogeny built on SNPs within the inversion 1401 detected on chromosome 9. Some European domestic dogs cluster closely with dingoes 1402 mirroring the pattern from LAMP‐LD local ancestry analyses. 1403 Figure 4. Genomic regions of interest in Chromosomes 9 and 27. (a) Detailed view of 1404 Chromosome 9, highlighting genes located within the inverted region that initially appeared as 1405 an introgressed block. Two genes -GJC1 and TCAP-showed signals of positive selection and are 1406 marked with an asterisk (*). (b) Detailed view of Chromosome 27, where a well -defined 1407 introgressed block was identified. One gene -OR8S3-was under positive selection, also indicated 1408 by an asterisk (*). Most genes in both regions were under purifying selection, with dN/dS ratios 1409 below 1. For a full list of genes, selection statistics, and associated p-values, see Table S5. 1410 Figure 5. Environmental drivers of genetic variation. Redundancy Analysis (RDA). Each 1411 point is an individual colored by population (Big Desert, Central, East, North, South, West, 1412 Captive). Environmental predictors are shown as arrows: popgrid: regional human population 1413 density, human_settlements: proximity to built‐up areas, trees: forest cover, fences: livestock 1414 fencing density, landuse: proportion of modified land types, crops: agricultural land cover, 1415 shrubs: shrubland cover, elev: elevation, BIO3, BIO5, BIO9: bioclimatic variables 1416 (isothermality, max temperature of warmest month, mean temperature of driest quarter, 1417 respectively). Arrow direction and length indicate the variables’ loadings on RDA1 and RDA2; 1418 longer arrows denote stronger associations with dingo genetic structure. 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Figure 1. 1429 1430 1431 1432 1433 1434 1435 1436 1437 1438 1439 1440 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint Figure 2. 1441 1442 1443 1444 1445 1446 1447 1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint 1465 Figure 3. 1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint 1477 Figure 4. 1478 1479 1480 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint 1481 Figure 5. 1482 1483 1484 1485 1486 1487 .CC-BY-NC-ND 4.0 International licenseavailable under a (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made The copyright holder for this preprintthis version posted March 24, 2026. ; https://doi.org/10.64898/2026.03.22.713106doi: bioRxiv preprint