Introduction

Increasing anthropogenic movement has led to increased introductions of non-native organisms (Pyšek et al. 2020). This can lead to ‘extinction by hybridisation’, where introgression leads to the replacement of a native form with its relative (Rhymer and Simberloff 1996; Randi 2008; Ottenburghs 2021). On the other hand, natural hybridisation can play a diversifying evolutionary role (Quilodrán et al. 2020), for example, via hybrid speciation (e.g. Lamichhaney et al. 2018). ‘Genetic rescue’ can increase population survival probability (Frankham 2015). Such cases suggest that anthropogenically induced hybridisation might have some positive impacts (Smith et al. 2022a).

After interbreeding with feral domestic pigeons Columba livia, the undomesticated Rock Dove (also Columba livia) was replaced by hybrids in many regions (Smith 2023a). For example, within the British Isles, whilst Outer Hebridean Rock Doves experience little introgression, around 10–20% of ancestry in Orkney’s doves now comes from feral pigeons, and this introgressed genetic material is distributed across the entire genome (Smith et al. 2022b). In Yorkshire and the Isle of Man, feral pigeons eventually replaced Rock Doves, and none of the latter remain in these regions today except as a small ancestry contribution in contemporary feral pigeon populations (Brown and Grice 2005; Morris and Sharpe 2021). Rock Doves with negligible evidence of introgression, such as those in the Outer Hebrides, persist as small populations and may have low genetic diversity. Whilst wild-feral hybridisation poses a risk to Rock Dove conservation, low levels of gene flow might bolster genetic diversity.

We use the Rock Dove populations of the British Isles to highlight the relationship between genomic admixture and diversity. We assess whether gene flow barriers prevent genomic homogenisation, whilst allowing sufficient gene flow to augment genetic diversity.

Methods

We used an existing genomic dataset consisting of whole-genome sequencing data from 232 pigeons (depth 4–10 ×) (Smith et al. 2022b). Some of these were wild Rock Doves (‘Outer Hebrides’, ‘Inner Hebrides & Arran’, ‘Highlands & Orkney’, ‘Shetland’, ‘Cape Clear Island’). Others were ‘domestic-type’ birds, including feral pigeons (‘England & Isle of Man’ and ‘Shetland Feral’) (Fig. 1a), as well as ‘captive Rock Doves’ (genetically similar to domestic breeds although probably admixed with other pigeon species) (ENA PRJEB52260), and various different domestic breeds common in captivity (GenBank PRJNA167554) (Shapiro et al. 2013). We used the ‘colLiv2’ genome (NCBI Assembly GCA_001887795.1) as the reference (Damas et al. 2017) and carried out mapping and data preparation as described in Smith et al. (2022a, b).

Fig. 1
figure 1

a Free-flying pigeon populations included in this study. ‘England & Isle of Man’ birds, and a subset of those from Shetland, are feral pigeons. All others are Rock Doves. b Q proportions for Rock Doves in this dataset (photograph: WJS), showing individuals ranked from lowest to highest ‘wild’ ancestry proportion. c Effective migration surfaces inferred by EEMS. Orange regions have reduced gene flow, and blue regions, higher gene flow. d Heatmap of pairwise FST values, with brighter reds indicating higher divergence

Full size image

We ranked wild-feral ancestry proportions (‘Q’) for pigeons (Smith et al. 2022b) calculated using NGSadmix v32 (Skotte et al. 2013). Then, to examine how population structure and introgression interact, we calculated FST for each pigeon population pair. This involved using ANGSD v0.933 to infer major and minor alleles from genotype likelihood data (-doMajorMinor 1), calculate allele frequency assuming major and minor are known (-doMaf 1) and use sites with minor allele frequency above 0.01 (-minmaf 0.01). We also set base and mapping quality thresholds of 20 and 30 (-minQ 20 -minMapQ 30), discarded ‘bad’ reads (-remove_bads 1), estimated posterior genotype probability (-doPost 2), estimated genotype likelihoods using the GATK model (-GL 2), used qscore recalibration (-baq 1) and kept sites with data from ≥ 10 individuals (-minInd 10). We extracted columns 1 and 2 from the resulting ‘mafs.gz’ file, before indexing the output with ‘angsd index’. For each population, we calculated allele frequency likelihoods with folded 2D site frequency spectra: setting base and mapping quality thresholds of 20 and 30 and discarding ‘bad’ reads. We estimated genotype likelihoods using the GATK model, using qscore recalibration, and calculated site allele frequency likelihood based on individual genotype likelihoods, assuming Hardy–Weinberg equilibrium (-dosaf 1). We calculated genome-wide FST using realSFS fst index, and estimated pairwise-weighted FST using ‘realSFS fst stats’.

We used EEMS v0.0.0.9 to characterise gene flow (Petkova et al. 2016). We generated a genetic distance matrix in ANGSD for 132 free-flying individuals. Shetland’s feral pigeons were excluded as they fell within the same mapping unit as Shetland Rock Doves, which would inflate wild-feral migration estimates (Petkova et al. 2016). ANGSD was run: estimating genotype likelihoods with the SAMtools model, inferring major and minor from genotype likelihood data, calculating allele frequency assuming known major and minor alleles, and recording base frequency (-doCounts 1). This also involved printing the consensus base (-doIBS 2) and outputting a pairwise IBS matrix (-makeMatrix 1). We used qscore recalibration (-baq 2), set base and mapping quality thresholds of 30, and discarded bad reads and those not mapping uniquely (-uniqueonly 1). We kept sites with data from ≥ 71 individuals and skipped triallelic sites (-skipTriallelic 1). We used five EEMS runs, each with an MCMC chain of 10,000,000 iterations with a 20,000-iteration burn-in and 1800 thinning iterations. The number of geographic demes was set to 100, 150, 200, 250 and 300 (Petkova et al. 2016). Runs were combined using rEEMSplots in R v4.0.2 (R Development Core Team 2010).

To quantify heterozygosity, we used ANGSD: estimating genotype likelihoods using the SAMtools model, setting base and mapping quality thresholds of 20 and 30 and calculating site allele frequency likelihood (-dosaf 1), before estimating site frequency spectrum with realSFS (Nielsen et al. 2012). Heterozygosity was outputted as outlined at http://www.popgen.dk/angsd/index.php/Heterozygosity. We calculated inbreeding with default ngsF settings (Vieira et al. 2013), generating the input via ANGSD: estimating genotype likelihood with the SAMtools model, inferring major and minor and calculating allele frequency assuming known major and minor alleles, including only proper pairs (only_proper_pairs 1) and sites with minor allele frequency > 0.05. We excluded bad reads and those not mapping uniquely. We used qscore recalibration, set minimum base and mapping quality of 30 and 25, and kept sites with data from ≥ 115 birds and with a p value of 1e − 6 (-SNP_pval 1e − 6), outputting a genotype likelihood file (-doGlf 3). We tested for differences in heterozygosity and inbreeding between populations and pigeon types (wild Rock Dove, feral pigeon, domestic pigeon, captive ‘Rock Dove’) with Kruskal–Wallis tests, and followed up with pairwise Wilcoxon tests (Wilcoxon 1945) using a Benjamini–Hochberg correction (Benjamini and Hochberg 1995) to examine pairwise differences between populations and types. We calculated effect sizes using wilcox_effsize in R. To examine the impact of introgression score and pigeon type on both inbreeding and heterozygosity, we have used standard linear regression as implemented in the base stats package of R. To identify the best model, we used the MuMIn package (Barton 2022. MuMIn: multi-model inference—R package ver. 1.47.1. CRAN: The Comprehensive R Archive Network). This involved using its dredge function to test all possible model combinations originating from a model of inbreeding OR heterozygosity ~ introgression score × pigeon type. We used ΔAICc > 2 to indicate significant differences between models (Hurvich and Tsai 1989). For both heterozygosity and inbreeding, the best model was the full interaction model (see Supplementary Tables 5 and 6).

Results

Rock Dove Q scores do not drop below 0.75 (Fig. 1b). EEMS shows there is reduced gene flow across seas, moorland and mountains (Fig. 1c). Pairwise FST shows that where there is high introgression (e.g. Highlands & Orkney), individuals become genetically similar to feral pigeons (England & Isle of Man). However, genetic structure is also influenced by geographic distance, as seen for Cape Clear Island birds that are divergent from feral pigeons despite introgression (Fig. 1d).

Populations at different locations differ in level of inbreeding (Kruskal–Wallis test, χ2 = 125.68, n = 9, df = 8, P < 0.001, Fig. 2a) and heterozygosity (Kruskal–Wallis test, χ2 = 111.16, n = 9, df = 8, P < 0.001; Fig. 2b). Pairwise comparisons using Wilcoxon tests show (Supplementary Tables 1 and 2) that this stems from significant differences between various different locations. Pigeon types also differ in inbreeding (Kruskal–Wallis test, χ2 = 111.62, n = 4, df = 3, P < 0.001) and heterozygosity (Kruskal–Wallis test, χ2 = 95.946, n = 4, df = 3, P < 0.001). Wilcoxon pairwise tests show that this is driven by differences among almost all pairs of pigeon types (Supplementary Tables 3 and 4). Domestic breeds and ‘captive Rock Doves’ have lower genetic diversity than free-flying populations. Within free-flying populations, those that experience hybridisation have higher genetic diversity than feral pigeons or Outer Hebridean (i.e. ‘pure’ or nearly pure) Rock Doves. The impact of increasing domestic ancestry on inbreeding (Fig. 2c) and heterozygosity (Fig. 2d) differed between Rock Doves and ‘domestic-type’ birds. In the latter, inbreeding increases with higher ‘introgression score’ (i.e. domestic type ancestry) (estimate = 0.372, P < 0.001), while heterozygosity decreases (estimate = − 0.002, P < 0.001). However, this relationship is significantly different in Rock Doves, for both inbreeding (estimate = − 0.791, P < 0.001) and heterozygosity (estimate = 0.004, P < 0.001), where inbreeding decreases and heterozygosity increases with the introgression score.

Fig. 2
figure 2

a Inbreeding varies with region and pigeon type. Red lines indicate the mean values. b Individual genome-wide heterozygosity varies with region and pigeon type. Red lines indicate the mean values. c As domestic-type ancestry levels increase, individual inbreeding decreases in Rock Doves but increases in domestic-type pigeons. d As domestic-type ancestry levels increase, individual genome-wide heterozygosity increases in Rock Doves but decreases in domestic-type pigeons

Full size image

Discussion

Gene flow and genetic differentiation patterns align with observations that feral pigeons are unlikely to cross seas, and are rare in mountains and moorlands (Lack 2010) (Fig. 1c). Such natural barriers to feral pigeon dispersal protect the least introgressed Rock Dove populations in the Outer Hebrides (Smith et al. 2022b). Escapees from captive domestic collections, or feral pigeons riding on ferries, may still threaten such populations (Smith 2023a). It is also possible that other factors may contribute to gene flow limitation. For example, inversions which may occur during the domestication process (Imsland et al. 2012) may cause further isolation through pre-mating isolation or hybrid breakdown (Fishman and Sweigart 2018). Nevertheless, such genomic limitations are clearly unable to prevent extinction by hybridisation in the long term in this system, given the global decline of Rock Dove populations (Johnston et al. 1988).

Introgression correlates with increased genomic diversity in Rock Doves (Fig. 2). Among free-flying populations, both feral pigeons (having undergone both domestication and feralisation) and Outer Hebridean Rock Doves (an unadmixed small population) have lower genetic diversity than hybrid populations. The low levels of genetic diversity in feral/domestic and Rock Dove populations fall within the range which is usual for endangered species (below 0.004 genome-wide heterozygosity). This value comes from a study of mammals comprising representatives of around 80% of mammalian families—although the relationship between conservation status and genetic diversity is weak (Genereux et al. 2020; Teixeira and Huber 2021). Among hybrid populations, regional variation in genomic diversity could be explained by variation in levels of wild-feral gene flow. Interbreeding between a rare outbred and an abundant inbred lineage could lead to heightened inbreeding within the outbred lineage. However, in this case, both the wild and feral/domestic populations are (for different reasons) inbred, and likely fixed for differing alleles. Rock Doves and feral pigeons are therefore likely to produce heterozygous offspring.

The potential positive impact of hybridisation may be a temporary effect. Feral pigeons replaced Rock Doves in Yorkshire and the Isle of Man in recent decades (Smith 2023a), and these birds have the low genetic diversity associated with more long-standing feral populations. The putatively temporary benefit of wild-feral introgression is most likely due to eventual genetic swamping (assessments of phenotype suggest that genomic replacement of Rock Doves with feral pigeons has happened at these sites (Smith 2023b)), but selection against introgressed material could be involved to some extent and has been identified as important in other species (Moran et al. 2021). The complete loss of unadmixed Rock Doves would be undesirable if we are to maintain genomic diversity within Columba livia (Smith et al. 2022a, b; Smith 2023a).

Conclusion

In the British Isles’ Rock Dove populations, birds with admixed wild-feral ancestry have higher genetic diversity than unadmixed individuals. This could facilitate persistence of relictual wild populations, but only if hybridisation is kept at low levels. Further research characterising bird longevity, ecology and reproductive success would be needed to reveal links between individual introgression status, genomic diversity, and fitness in populations of domestic, feral, wild and hybrid origin.