Deepest genomic division is between wild red junglefowl and chickens

Various analytical approaches indicated that the deepest population-genomic divergence in G. gallus is between chickens and their wild counterparts (Figs 1C and S1) [19,28]. Within our historic wild G. gallus samples, the subspecies bankiva forms a distinct cluster from the other subspecies (S1 and S2 Figs). For the remaining taxa, there seems to be an overall clinal geographic structure from Indian murghi to Southeast Asian spadiceus, with the latter being closer to domestic chickens (Figs 1C and S2). This result emerged irrespective of the inclusion of chickens from extremely different geographic backgrounds and pedigrees (S1 Fig and S2 Table). The phylogenomic tree constructed from >1 million genome-wide markers found all domestic individuals forming one well-supported clade within the G. gallus complex (S1 Fig). Using various approaches, we did not find any historic artifact biasing our data (S2 and S3 Figs).

Historic DNA reveals higher domestic introgression in modern wild junglefowl versus pre-Anthropocene samples

Principal component analysis (PCA) of almost 3 million genome-wide markers suggests that present-day wild populations possess more domestic alleles compared to populations from a century ago (Fig 1D). Here, we limited our analysis to only G. g. spadiceus, the putative ancestral subspecies domestic chickens originated from [14], to reduce artifacts of population structure from other subspecies (S2 Fig). Observed heterozygosity (H) has decreased drastically and Tajima’s D values have become positive from historic (H = 0.0470, Tajima’s D = -1.15996) to modern G. g. spadiceus (H = 0.00464, Tajima’s D = 0.470075), reflecting a loss of allelic diversity in modern G. g. spadiceus. The corresponding statistics for our panel of domestic chickens were mostly similar to modern G. g. spadiceus (H = 0.00503, Tajima’s D = -9.32).

Within G. g. spadiceus, modern samples seem to exhibit a clinal population structure from the south (Singapore) to the north of the subspecies range (northeastern India), with seemingly increasing domestic contributions (Figs 1D and S2). The purest examples of modern free-roaming junglefowl exhibit a genomic profile that is almost fully embedded with the historic wild population, while other modern free-roaming samples from throughout the range of spadiceus have profiles that cluster much more closely with domestics (Fig 1D), indicating substantial heterogeneity in the level of introgression, sometimes even within modern wild populations.

We quantified levels of excess domestic introgression into modern spadiceus populations by computing D-statistics using the topology (((spadiceus-historic, spadiceus-modern), chicken), Chinese bamboo partridge Bambusicola thoracicus) [29]. We carried out multiple D-statistics calculations using different subsets of populations across the range of spadiceus to ensure that population structure did not affect our tests. While these analyses are unable to evaluate the level of domestic introgression that was already present in pre-Anthropocene wild red junglefowl from one century ago, we showed that modern populations of the subspecies spadiceus exhibit a significant excess of such introgression as compared to populations from just before the Anthropocene (Fig 1B and S3 Table).

When inferring domestic versus wild ancestry across around 2.7 million SNPs using the program Struct-f4 [30], a pattern emerged in which a substantial proportion of modern individuals from throughout the range of spadiceus were characterised by a large amount of domestic introgression, reflecting our PCA results (Figs 1D and 2A). Global ancestry inference using the same program found most modern red junglefowl to be distinct from the other populations, with contributions from domestic chickens (20–50%) and historic red junglefowl (45–80% in the four genomically most “wild-type” junglefowl individuals from Singapore) (Fig 2B). Local ancestry inference with almost 150,000 single nucleotide polymorphisms (SNPs) using the program EILA (Efficient inference of local ancestry) [31] found the Singaporean population broadly heterogeneous, with percentages of domestic-introgressed alleles ranging from 5–97% depending on individual (Fig 2C). All individuals screened from northern Thailand (Chiang Mai), northeast India (Manipur) and southwest China (Yunnan) displayed percentages of domestic-introgressed alleles between 83%–96% (Fig 2C). When restricting EILA analysis to using only commercial farm breeds as the domestic reference, a similar trend emerged, but the percentages of domestic contribution fell to 30–70% (S4 Table and S4 Fig). While the accuracy of estimating local ancestry sources in the genome remain contentious, the trend obtained from our EILA analysis is reflective of our global ancestry estimation.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Ancestry inference of 15 representative modern red junglefowl from four localities across the range of Gallus gallus spadiceus.

(A) Multi-Dimensional Scaling plot from Struct-f4 (K = 4) assuming no admixture based on 2,603,726 SNPs. Modern red junglefowl samples are highlighted with a black frame; historic red junglefowl samples are to the right of the modern ones (teal), and domestic chicken samples to the left (green). The color legend indicates which population each sample is assigned to regardless of actual identity (RJF = red junglefowl). (B) Individual ancestry profiles from Struct-f4 (K = 4) based on 2,603,726 SNPs. Colors refer to genomic contributions as inferred in 2A. ‘Out’ refers to the outgroups used in the analysis. (C) Local ancestry inference based on 143,526 SNPs using Efficient inference of local ancestry (EILA). Only the top 15 chromosomes (Chr 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 12, 13, 14, 16, 20) with the most SNPs called are illustrated. Blue coloration refers to domestic ancestry and red coloration refers to wild ancestry. The percentage of genome-wide SNPs with domestic ancestry is given on the bottom right.

https://doi.org/10.1371/journal.pgen.1010551.g002

Identification of possible domestication loci under selection in chickens

Following from our evidence that domestic introgression into wild red junglefowl has increased in modern times, we sought to find genomic areas of elevated divergence between chickens and historic junglefowl. We performed scans of genomic divergence across the chicken reference genome (RefSeq Assembly Accession: GCF_000002315.6) in 50kb sliding windows using D_XY as a measure of differentiation, comparing domestics to wild historic samples of each of the five red junglefowl subspecies (Fig 3). We also used F_ST as another measure of differentiation, but preliminary plot inspection indicated an excess of potential domestication loci which may be an artifact of the difference in sample size among populations (S5 Fig). We chose a wide range of domestics from farm breeds to village chickens to reduce signals associated with post-domestication breed formation and adaptive differentiation. Most genomic divergence between domestics and wild individuals was at a roughly constant background level, occasionally disrupted by so-called ‘islands of extreme genomic divergence’ [32,33]. We harvested regions with D_XY sliding window values that exceed the 99^th percentile for each pairwise comparison and in this way identified 883 potential domestication-related genic regions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Positive selection on genomic islands of differentiation between chickens and wild red junglefowl.

(A) D_XY (pairwise divergence) between historic wild individuals of each red junglefowl subspecies versus modern chickens in 50kb sliding windows with a step size of 10kb across the chicken reference genome. Alternating hues of blue denote different chromosomes and the horizontal black dotted lines denote the 99^th percentile of D_XY of autosomes (bottom) and the Z chromosome (top). Arrows indicate genic regions located within genomic islands of high differentiation between chickens and junglefowl which are under selection in domestic chickens; their colors refer to the main function of genes following (B). (B) Branches under positive selection that were inclusive of domestic chickens calculated using aBSREL (red: uncorrected p-value ≤ 0.05) for genes that are highly divergent between chickens and wild junglefowl. Seven outgroups across the avian phylogeny were used in this analysis (signatures of selection in outgroups not depicted, see S6 Fig).

https://doi.org/10.1371/journal.pgen.1010551.g003

These loci were tested for positive selection using aBSREL with seven outgroups across the avian phylogeny [34]. A total of 112 genes were found to be under selection in the Gallus gallus complex, with a subset of eight solely in domesticated chickens not including selection signals in the outgroups (Figs 3 and S6 and Tables 1 and S5). Genes identified as deeply divergent and under selection between domestics and historic wild samples of various subspecies play an important role in such processes as vision maintenance (CERKL [35]), reproduction (CFAP97 [36], MUS81 [37]), metabolism (C2CD5 [38]), nervous system signaling (CAPS2 [39]) and development (DYNC2H1 [40], SLTM [41], GPATCH8 [42]) (Fig 3 and Table 1). With the exception of CERKL, which emerged as under selection in both wild spadiceus and chickens, the other genes we identified are only under selection in domestic chickens (Table 1). These eight genes also showcase positive selection in the outgroups (S6 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Function of genes located on genomic islands of high differentiation between chickens and junglefowl which are under selection in chickens.

Only CERKL was found to be under selection in both chickens and wild spadiceus; the remainder were found to be under selection uniquely in chickens. Values are rounded to 3 significant figures (s.f.).

https://doi.org/10.1371/journal.pgen.1010551.t001

By comparing the genome position of the eight genes under selection (Table 1) with our local ancestry results (Fig 2C), we deduced the most likely origin of these genes in the 15 modern red junglefowl individuals included. Out of the eight genes, we were unable to infer the local ancestry for MUS81 and GPATCH8 due to a lack of SNPs in the surroundings of their genomic location in EILA. For the other six genes, the majority of all 15 modern red junglefowl screened possessed a copy of domestic ancestry (SLTM– 73.3%, CFAP97–80%, CAPS2–66.7%, C2CD5–60%, DYNC2H1–66.7%) with only one exception (CERKL– 20%), underpinning the veracity of our approach of using historic red junglefowls as a reference. If these modern individuals had been chosen for domestic selection scans, these genes would not have been flagged as potential candidates as they have already been homogenized in modern junglefowl via domestic introgression.

Increasing levels of domestic introgression threaten the modern wild genotype of G. gallus

The deep genetic rift between domestics and all wild junglefowl may appear unusual because the different subspecies of red junglefowl have had hundreds of thousands of years to differentiate, whereas domestication is only thought to date back approximately 4,000 to 10,000 years [27,43,44] (Figs 1 and S1). This rift is indicative of pronounced long-term gene flow among all adjoining continental subspecies of the red junglefowl into historic times, preventing the formation of any deep population-genomic barriers between wild populations. Furthermore, this domestic-wild rift is likely an effect of a strong bottleneck during domestication similar to other domestic animals [45,46]. This clear distinction between our historic G. gallus individuals and modern chickens suggests that there may have been some form of spatial isolation between chickens and the historic wild junglefowl selected for this study, making the latter an appropriate representative of the wild genotype in further comparisons.

The substantial increase in domestic introgression over the span of only a few decades does not bode well for safeguarding the wild allelic diversity of G. gallus [20,47,48]. This intensification of domestic admixture is likely driven by anthropogenic change to the native environment of G. gallus, which has experienced a high level of habitat loss following the relentless human encroachment into tropical Asia’s last wilderness areas in recent decades [49–51]. While the increase in domestic introgression is heterogenous throughout the range, all modern populations will–by now–be in proximity to released domestic chickens that are free to interbreed with them. Our data suggest that intensifying introgression from highly homogenised domestic stock into the diverse wild stock has led to an overall loss of heterozygosity by an order of magnitude. This result may be counter-intuitive at first, given that outbreeding and introgression are theoretically expected to lead to increasing heterozygosity. However, chickens, despite their great phenotypic diversity and population size, are known to have undergone an extreme bottleneck and are genomically impoverished akin to other domestic animals, as reflected by their low heterozygosity and Tajima D values in our study. It is likely the introduction of highly homogenised domestic alleles into the wild red junglefowl gene pool that has caused wild-type genetic diversity to plummet so precipitously.

As a caveat, our conservative assignment of some samples of questionable provenance to ‘domestic village chickens’ (see methods) may have led to an overestimate of the absolute level of domestic introgression. Estimation of local ancestry using only commercial farm samples reduced the domestic contribution by about 30% in some samples (S4 Table). However, this potential bias is incompatible with the stark contrast between the clear genomic division between historic junglefowl and chickens as compared to the more blurred division when modern junglefowl are added (Fig 1), which is clearly due to excess introgression in modern samples. A previous study has found a correlation between plumage and genomic makeup in a chicken-junglefowl cline [19], and while we have photographic evidence only for some of the modern red junglefowl included in this study, this evidence suggests that the samples that cluster closer to domestics do exhibit some phenotypic differences from the ancestral wild phenotype (S7 Fig).

An increase in domestic introgression would allow the wild populations to continue to persist. Indeed, there are documented cases of adaptive introgression aiding wild ancestral populations of domesticated animals in their survival (reviewed in [52]). At the same time, it is unknown whether such a positive evolutionary development would apply in junglefowl. More localized studies on chicken-junglefowl interactions [19] and studies concentrating on single traits [21] have shown that at least some of the domestic admixture into junglefowl may fairly be characterised as adaptive introgression. However, an extensive phenotypic dataset combined with genomic data would be required to conclusively address this question. In the meantime, the steep reduction in genetic diversity following in the wake of domestic introgression (see above) raises concerns that the admixture process may lead to a dangerous reduction of allelic diversity in the wild type.

The loss of domestication markers due to domestic introgression

The quest for domestication markers is important from two perspectives: (1) it highlights genes and other genomic motifs which render chickens what they are, and which help us define the most important traits underlying domestication; (2) it assists in future attempts to diagnose wild populations which have experienced little admixture with domestic alleles and retained some of the most important genomic architecture characterizing the wild genotype. The search for these loci has been a hot biological topic for decades [14,15,25] and has routinely involved comparisons between modern wild and domestic populations [14,15,25]. While some domestication markers selected for in domestics may be selected against in wild population, allelic homogenization following extensive domestic introgression into modern wild populations could mask many other loci that have played an important role in the original domestication process. The high proportion of modern red junglefowl possessing the domestic copy of genes under selection in domestic chickens is consistent with such a scenario (Table 1 and Fig 2). While we do not possess ancient genomes of red junglefowl, genes identified using historic G. gallus samples as opposed to modern ones would suffer this problem to a much lower extent due to their less pronounced levels of introgression. As such, the analysis of historic G. gallus samples is crucial to shedding light on the differences between the domestic chicken and its wild counterpart.

Genes involved in the nervous system are dissimilar between red junglefowl and chickens

Our study identifies four genes involved in the nervous system that are selected for in chickens but not their wild counterpart. Both DYNC2H1 and SLTM code for proteins involved in the Sonic hedgehog pathway that are critical in the formation of organs and the central nervous system during development by regulating neural crest cells [53–55]. We also identified two other genes with relevance to the nervous system–CAPS2 modulates synaptic activities [39] and C2CD5 regulates body weight and appetite in the hypothalamus [38]. These properties are not surprising given that tameness is one of the main characteristics being selected for during domestication [56,57].

Chickens are known to have poorer vision compared to their wild counterparts [58], with a handful of vision genes under positive selection that have been proposed to cause this phenotype [59]. Here we add another vision gene, CERKL, that plays a role in reducing oxidative damage on the retina, which is equally under positive selection [35]. Interestingly, wild spadiceus also seem to be experiencing positive selection in this gene, either due to (1) introgression of the domestic copy of CERKL into wild spadiceus, or (2) because spadiceus may naturally be endowed with a weaker vision.

A previous study suggested that TSHR, a gene encoding for the thyroid-stimulating hormone receptor protein, could be a domestication marker based on its near fixation in chickens compared to G. g. spadiceus [15], a result contradicted by later research suggesting that this pattern is merely a by-product of increasing allele frequency in wild populations [14]. This latter gene is also not flagged in our analyses.

Using genomes of historic specimens from throughout the range of red junglefowl, we were able to (1) demonstrate an intensification of introgression from chickens into modern red junglefowl across the Anthropocene and (2) identify eight candidate genes under selection in chickens. Given that our historic red junglefowl samples exhibited less domestic admixture compared to modern fowl, we were able to identify possible genes in our selection scan that would remain intractable in comparisons involving only modern junglefowl, as the latter’s genomes are frequently homogenized due to extensive domestic introgression. These candidate genes will be critical for future genomic enquiries to add to our understanding of the functional differences between domestic chickens and wild junglefowl.

Population sampling

We obtained 45 dried toepad samples of G. gallus collected between 1874 and 1939 across its native range preserved at the Natural History Museum at Tring (UK) and Lee Kong Chian Natural History Museum (Singapore) (Fig 1 and S1 Table). The level of domestic introgression in historic samples is likely lower compared to modern-day samples given that habitat encroachment in their native range was less severe in the past, allowing us to use them as a reference point to quantify excess modern introgression. We also incorporated an additional 69 samples of red junglefowl and various chicken breeds available from GenBank, including eight free-roaming G. g. spadiceus from Singapore, seven farm individuals and two historic samples from peninsular Malaysia from Wu et al. (2020) [19] along with 8 G. g. spadiceus from Wang et al. (2020) [14] (S2 Table), adding them to our dataset of 45 historic G. gallus for a total of 114 samples. We additionally included sequences of one Bambusicola thoracicus individual from GenBank as an outgroup [60] (S2 Table), leading to a total dataset of 115 samples. It is difficult to determine the origin of red junglefowl samples from online databases without phenotypic validation. Therefore, we have opted in favor of a conservative approach whereby samples labelled G. gallus of uncertain but likely domestic origin were labelled as village chickens (S2 Table) and only kept those samples confidently labelled as modern red junglefowl.

Data generation and basic processing