Genomic Analyses Reveal Genetic Adaptations to Tropical Climates in Chickens

Summary

The genetic footprints of adaptations to naturally occurring tropical stress along with domestication are poorly reported in chickens. Here, by conducting population genomic analyses of 67 chickens inhabiting distinct climates, we found signals of gene flow from Tibetan chickens to Sri Lankan and Saudi Arabian breeds and identified 12 positively selected genes that are likely involved in genetic adaptations to both tropical desert and tropical monsoon island climates. Notably, in tropical desert climate, advantageous alleles of TLR7 and ZC3HAV1, which could inhibit replication of viruses in cells, suggest immune adaptation to the defense against zoonotic diseases in chickens. Furthermore, comparative genomic analysis showed that four genes (OC90, PLA2G12B, GPR17 and TNFRSF11A) involved in arachidonic acid metabolism have undergone convergent adaptation to tropical desert climate between birds and mammals. Our study offers insights into the genetic mechanisms of adaptations to tropical climates in birds and other animals and provides practical value for breeding design and medical research on avian viruses.

Graphical Abstract

Highlights

• •Gene flow was detected from Tibetan chickens to Sri Lankan and Saudi Arabian chickens
• •Twelve genes showed chicken adaptations to both tropical desert and monsoon climates
• •Advantageous alleles of TLR7 and ZC3HAV1 showed immune adaptation in chickens
• •Convergent adaptation to tropical desert climate was found between birds and mammals

genomics; genetics; biological sciences; zoology.

Introduction

Chickens were domesticated from red jungle fowl (RJF) at least 4,000–4,500 years ago and have multiple origins within the geographic range of their wild ancestors in South and Southeast Asia (Fumihito et al., 1996; Larson and Fuller, 2014; Liu et al., 2006). Domestication of chickens is closely related to human activities. For instance, the complex and ancient interactions resulting from the Silk Route have facilitated maritime and terrestrial intercontinental translocation of several domestic chicken breeds among the Middle East and South, Southeast, and East Asia (Boivin and Fuller, 2009; Boulnois, 2004; Fuller and Boivin, 2009). After a long period of artificial and natural selection by humans, chicken breeds have evolved genetic adaptations to their local environmental conditions, such as hot and arid climates and high-altitude environments (Lawal et al., 2018; Wang et al., 2015). It is well known that the detection of genomic differences can shed light on the genetic basis of adaptation to diverse environments and provide insights into functionally important genetic variants (Andersson and Georges, 2004). The completion of the chicken genome sequence and genetic variation maps has promoted the exploration and research of chicken genetic mechanisms using whole-genome-based strategies (Hillier et al., 2014; Johnsson et al., 2012; Li et al., 2017a; Qanbari et al., 2019; Rubin et al., 2010; Wong et al., 2004). Conversely, considering their short reproductive and growth periods and wide distribution, chickens can be used as ideal models to study genetic adaptations to environments.

Environmental pressure is an important driver shaping the animal genome, which is also affected by pathogens (Fumagalli et al., 2011), disease (Sabeti et al., 2007), diet (Jiao et al., 2019), altitude (Huerta-Sanchez et al., 2013), and climate (Ai et al., 2015; Frichot et al., 2013; Lv et al., 2014). High altitudes are among the most challenging extreme environments, and high-altitude adaption has been widely studied in humans (Simonson et al., 2010), Tibetan chickens (TCs) (Wang et al., 2015; Zhang et al., 2016), yaks (Qiu et al., 2012), pigs (Li et al., 2013), ground tit (Qu et al., 2013), and sheep (Yang et al., 2016). Recently, genomic studies have also reported the genomic responses to acclimation to hot and arid environments in camels (Jirimutu et al., 2012; Wu et al., 2014), sheep (Yang et al., 2016), and goats (Kim et al., 2016). Although the acute heat response has been studied in commercial chickens (Lara and Rostagno, 2013), to the best of our knowledge, genetic adaptations to naturally occurring hot and arid environments in domestic chickens remain poorly known. In view of the current and future global effects of climate change, the genetic basis of adaptation to hot or arid climates in chickens is of practical value for breeding design and medical research (Diamond, 2002).

In this study, we performed a large-scale genomic analysis of chicken populations, including 36 high-altitude TCs from our work (Li et al., 2017a), 5 RJFs (Wang et al., 2015), 10 TCs (Wang et al., 2015), 11 Sri Lankan indigenous chickens (SL) (Lawal et al., 2018), and 5 Saudi Arabian indigenous chickens (SA) (Lawal et al., 2018), from habitats in harsh environments (high-altitude, tropical desert, and tropical monsoon island climates) (Figure 1A and Table S1). Genome-wide analysis for selective sweeps was performed to elucidate the genetic mechanisms of cutaneous melanin formation (thermotolerance), hormonal regulation, angiogenesis, vasodilation, and mitochondrial respiration, which may work together to enable chickens to adapt to tropical desert and tropic monsoon island climates. We also revealed the immune adaptation to the defense against zoonotic diseases in chickens. Additionally, we revealed convergent adaptation to tropical desert climate between birds and mammals. Our study adds to knowledge on genetic mechanisms of adaptations to tropical climates in vertebrates and is of fundamental importance for the potential application in functional genomics, breeding design, and the development of medical models for avian viruses.

Figure 1.

Phylogeny, Genetic Structure, and Inferred Migration Events of Chicken Breeds/Clusters

(A) Geographic variation of annual average temperature for three indigenous chickens.

(B) Phylogenetic tree constructed by the maximum likelihood (ML) algorithm. Different colors represent various breeds/clusters of chickens. Abbreviations of chicken populations: SL, Sri Lanka; SA, Saudi Arabia; RJF, red jungle fowl; TC1_Sub1, subpopulation 1 of Tibetan chicken population 1; TC1_Sub2, subpopulation 2 of Tibetan chicken population 1; TC2_Sub1, subpopulation 1 of Tibetan chicken population 2; TC2_Sub2, subpopulation 2 of Tibetan chicken population 2. See also Table S5 for more detail of each population.

(C) Principal-component analysis (PCA) plot for chickens. The fraction of the variance explained is 3.50% for eigenvector 1, 3.15% for eigenvector 2, and 2.19% for eigenvector 3.

(D) Phylogenetic network of the inferred migration events among the eight chicken breeds/clusters with six migration events. Migration arrows are colored according to their weight. The scale bar shows 10 times the average standard error of the entries in the sample covariance matrix.

(E) Genetic structure of chickens with K = 4 ancestral populations.

Results and Discussion

Characterization of Genomic Variants

After performing strict data quality filtering, a total of 1.56 Tb (terabases) of high-quality genome sequences from 67 chickens were mapped to the chicken reference genome (GRCg6a, GCF_000002315.6), resulting in an average of 94.74% non-gap genome coverage and ∼20.25-fold average depth (Table S2). We identified 12.78 million biallelic high-quality population genomic variants (PGVs) that contained 10.69 million SNPs and 2.09 million InDels with lengths ranging from 1 to 306 bp (Tables S3 and S4). Overall, more than half of the PGVs (52.63% and 52.65% for total SNPs and InDels, respectively) were located in intronic regions, followed by intergenic regions, with 39.28% and 40.00% for SNPs and InDels, respectively. The densities of the PGVs also showed a higher ratio in the promoter region than in the coding region (Figures S2 and S3). We annotated 180,729 SNPs (1.67% of the total) and 10,901 InDels (0.52% of the total) located within 15,370 protein-coding genes (89.31% of the total 17,209 genes). We identified a total of 55,480 missense, 6,798 frameshift, 629 stop-gain, and 92 stop-loss PGVs that caused amino acid changes, transcript elongation, or premature stopping, leading to structural changes in 11,407 genes (66.29% of total genes). We also observed an enrichment of ≤30 bp in coding sequences (95.13% of the total) that are multiples of 3 bp, which is expected to preserve the open reading frame (Figure S3).

The ratio of nonsynonymous to synonymous changes was estimated to be lower in chickens (0.45) than in other agricultural economic animals (Table 1), such as pigs (0.68) (Li et al., 2013), sheep (0.66) (Yang et al., 2016), and cattle (1.21) (Daetwyler et al., 2014), possibly indicating a weaker artificial selection accumulation effect in chickens than in other animals due to the short domestication time (Larson and Fuller, 2014). We also found a greater proportion of unique PGVs (13.73% and 1.84% for SNPs and InDels, respectively) in RJF than in other chicken groups (Table 1), which is consistent with the loss of genetic diversity during domestication (Larson and Fuller, 2014).

Table 1.

Summary of Genomic Variations in 67 Chickens

Category	Core Set	RJF	TC1_Sub1	TC1_Sub2	TC2_Sub1	TC2_Sub2	SL	SA
Number of individuals	67	5	6	19	8	13	11	5
Number of SNPs	10,691,650	8,409,391	8,254,044	10,313,588	8,456,758	9,727,520	9,341,260	7,631,063
Unique SNPs	Not applicable	1,154,609 (13.73%	253,399 (3.07%)	854,996 (8.29%)	405,079 (4.79%)	516,531 (5.31%)	560,476 (6.00%)	312,110 (4.09%)
Nonsynonymousa	56,026	39,294	41,938	53,591	43,016	50,360	47,525	37,518
Nonsyn/Syn ratiob	0.453	0.416	0.438	0.450	0.439	0.447	0.444	0.433
Number of InDels	2,085,509	1,179,662	952,921	1,389,263	1,018,893	1,224,798	1,165,978	877,956
Unique InDels	Not applicable	21,706 (1.84%)	4,002 (0.42%)	15,282 (1.10%)	7,947 (0.78%)	9,798 (0.8%)	10,494 (0.90%)	4,390 (0.50%)
Frameshift InDelsc	6,973	2,784	2,531	3,424	2,621	3,162	2,557	1,844

^aNonsynonymous variants include missense, stop-gain, and stop-loss variants.

^bNonsyn/Syn ratio indicates nonsynonymous to synonymous ratio.

^cFrameshift InDels include frameshift InDels and stop-gain and stop-loss variants.

Phylogeny and Introgression in Chickens

On the basis of the maximum likelihood tree and principal-components analysis (PCA), there were five distinct major clusters (i.e., TC1, TC2, SL, SA, and RJF) (Figures 1B and 1C). We further used an admixture model (FRAPPE) to explore the level of shared ancestry among individuals within each breed (Tang et al., 2005). As the ancestry component (K value) increased, we observed a simultaneous increase in the composition of ancestors in the RJF (Figure S4). This result is expected because the genetic materials of other chicken breeds were from components of RJF ancestry (Figure 1C). The subcluster ancestry composition showed considerable genetic difference between the TC1 and TC2 clusters, which is consistent with a previous report showing that multiple origins of TCs can be traced to two independent sources (Wang et al., 2015). We also observed that some SL individuals always maintained a single ancestry component and others showed a simultaneous increase in the composition of ancestors with increasing K value. We speculated that this single ancestor might be the result of a continent-island genetic model in Sri Lanka. We further identified four admixture category clusters (TC1_Sub1, TC1_Sub2, TC2_Sub1 and TC2_Sub2) on the basis of the two distinct TC clusters that were previously observed (Wang et al., 2015), according to phylogenetic analysis (Figure 1B). We found that the TC2_Sub2 cluster had more shared ancestry components with the SA and SL breeds than the other clusters (Figure 1E). Based on TreeMix (Pickrell and Pritchard, 2012), we observed signals of genetic flow from TCs to SA and SL breeds (Figures 1D and S5). Subsequently, the D-statistic (Durand et al., 2011; Green et al., 2010; Patterson et al., 2012) also indicated that the SA and SL breeds have a significant level of introgression with TC2_Sub2 chickens (Figure S6, Tables S6 and S7). Previous studies showed that Southwest China was one of the origin regions of chickens (Liu et al., 2006; Wang et al., 2020) and that trade activities have facilitated maritime and terrestrial intercontinental translocation of several domestic chicken breeds among the Middle East and South, Southeast, and East Asia (Figure S7) (Boivin and Fuller, 2009; Boulnois, 2004; Fuller and Boivin, 2009). Our admixture analysis supported the hypothesis that Chinese indigenous chicken breeds may have been transported to Sri Lanka and Saudi Arabia via the land and maritime Silk Route voyages (Fuller and Boivin, 2009).

Genomic Sweep Footprints in Chickens

The SA indigenous breed is well adapted to hot and arid environments because those breeding centers are located in the Eastern Province of Saudi Arabia and belong to a tropical desert climate zone with high solar radiation, very large diurnal ambient temperature differences, high daytime temperatures (ranging from 21.2°C to 50.8°C), frequent floating dust weather, and extremely low rainfall (74 mm per year) (Figures 1A and Table S1). The SL indigenous breed is from the Puttalam district of Sri Lanka, which is a tropical monsoon island country with an average annual rainfall of ∼1,000 mm and a temperature of 27°C, whereas TCs live in a temperate humid plateau climate in China. To identify genomic sweep footprints in chickens, we applied a modified population-branch statistic (PBS) method, which was very powerful in detecting incomplete selective sweeps over short divergence times (Simonson et al., 2010; Zhan et al., 2014). This approach was designed to take advantage of three outgroups (TC2_Sub2, TC2_Sub1 and RJFs), aiming to identify genomic regions under selection in SA/SL breeds. We identified 24.18 Mb (2.30% of the genome) and 15.58 Mb (1.48% of the genome), encompassing 723 and 464 positively selected genes (PSGs), in SA and SL breeds, respectively. Next, we employed three methods: log2 (θπ ratio) (Nei and Li, 1979), cross-population extended haplotype homozygosity (XP-EHH) (Sabeti et al., 2007), and cross-population composite likelihood ratio (XP-CLR) (Chen et al., 2010), although these methods could not remove noise signals from domestication if compared with RJFs (the ancestral population of domestic chickens). Ideally, for our purpose of detecting tropical adaptation, the best control group would be those breeds with a similar level of domestication and a distinct level of environmental temperature compared with study groups. We selected TC2_Sub2 as the control group, because our analyses of phylogeny and introgression showed that there is significant genetic mixing between SA/SL breeds and TC2_Sub2 cluster (Figures 1B and 1D). Thus, using the three methods, we performed a genome-wide selection sweep scanning with two comparisons, i.e., SA and SL breeds were compared with TC2_Sub2 cluster, respectively. Specifically, 376 and 396 PSGs were identified by XP-CLR, 233 and 305 from log₂ (θπ ratio), and 383 and 267 from XP-EHH in SA and SL breeds, respectively (Figure 2). Combined with the PBS method, only 87 and 45 PSGs were detected by all four methods in SA and SL breeds, respectively (Figures S8 and S9). The low number of overlapped PSGs could have resulted from different signatures of population variations in the four methods (Akey, 2009; Sabeti et al., 2006; Wang et al., 2016). Finally, we obtained 942 PSGs in SA breed and 923 PSGs in SL breed after integrating all candidate selection regions from the four methods, and 150 of them were found in both SA and SL breeds (Figure S10).

Figure 2.

Selective Sweep Signals in Two Indigenous Chickens

(A) Circular Manhattan plot of genome-wide selection sweep signals for SA breed compared with the TC2_Sub2 cluster.

(B) Circular Manhattan plot of genome-wide selection sweep signals for SL breed compared with the TC2_Sub2 cluster. From inside to outside, the circular Manhattan plots show the PBS value, log₂θπ ratio, XP-CLR score, and XP-EHH value. The thresholds for identifying candidate selection sweep regions were p < 0.01 (Z-test) for the PBS values and 1% outliers for log₂θπ ratio values, XP-CLR score, and XP-EHH values Twelve PSGs simultaneously present in the SA and SL breeds are shown in red.

Genetic Adaptations to Tropical Climates in Chickens

Among the PSGs in the SA and SL breeds, we found significant enrichment of functional pathways (p < 0.05) that directly or indirectly influence several traits that are critical for survival under heat stress in tropical climates, e.g., 18 PSGs involved in the vascular smooth muscle contraction pathway (VSMC, 12 PSGs in SA breed and 9 PSGs in SL breed), 19 PSGs located in the melanogenesis pathway (13 in SA and 9 in SL); 27 PSGs located in the adrenergic signaling in cardiomyocytes pathway (12 in SA and 18 in SL), and 26 PSGs involved in the calcium signaling pathway (13 in SA and 17 in SL) (Tables S8 and S10). We found five PSGs (ADCY1, ADCY2, ADCY7, ADCY8, and ADCY9) associated with adenylyl cyclase (AC), affecting the formation of cAMP; these genes also play roles in the above four pathways (Figure 3). Adrenergic signaling in cardiomyocytes plays a pivotal role in regulating cardiac function in response to ever-changing environments. Systolic and diastolic function and heart rate are primarily controlled by the adrenergic and muscarinic systems in response to varying physiological demands (Schaub et al., 2006). The calcium signaling pathway acts as an intracellular messenger for signal transduction that is primarily modulated by adrenergic control of phosphorylation involved in intracellular Ca²⁺ handling (Schaub et al., 2006). We identified five PSGs (CACNA1B, CACNA1C, CACNA1G, CACNA1I, and CACNA1S) that are critical for calcium channel function, especially CACNA1C, which encodes the L-type voltage-gated calcium channel Cav1.2 and is under positive selection in both SA and SL breeds. This protein activates the VSMC pathway (Figure 3). The CAMK2D gene is under selection in chickens living in different tropical climates. It encodes calmodulin-dependent protein kinase-II (CaMKII), which is activated in response to adrenergic stimulation (Grimm and Brown, 2010), affects a variety of Ca²⁺-handling proteins (Schaub et al., 2006) and regulates melanin synthesis (Figure 3). In addition, we found that 68 PSGs are involved in central nervous system developmental processes, such as regulation of nervous system development, neurogenesis, neuron remodeling, cholinergic synapse, synaptic transmission regulation, cerebellar development, and axonal transport (Tables S9 and S11). The central nervous system of avian organisms exhibits high thermosensitivity in the control of thermoregulatory responses (Boulant and Dean, 1986). A deficit of thermal stimulation causes alterations, retardation, and deleterious effects in the central nervous system (Ahmed, 2005). Therefore, PSGs involved in these relevant functional ontologies suggested that nerve development, cutaneous melanin formation, hormonal regulation, angiogenesis, and vasodilation may have worked together to enable chickens to adapt to heat stress in tropical climates, reflecting common and local adaptations.

Figure 3.

Genetic Adaptations of Chickens to Tropical Climates

Dotted arrows indicate an indirect effect. Dotted lines indicate indirect inhibition. Solid arrows indicate activation. Solid lines indicate association. The PSGs simultaneously present in the SA and SL breeds are shown in red. The PSGs in SA breed are shown in brown. The PSGs in SL breed are shown in green. The names of the KEGG pathways are shown in purple.

PSGs Associated with Thermotolerance

The cutaneous melanin pigment plays a critical role in protecting animals against the harmful effects of solar radiation, and Wnt signaling plays a crucial role in orchestrating epidermal stratification. Of the 19 PSGs located in the melanogenesis pathway, three PSGs (WNT7B, WNT8B, and GNAO1) in SA breed and two (WNT7B and GSK3B) in SL breed are involved in Wnt signaling, which induces epidermal stratification and regeneration (Zhu et al., 2014); three PSGs (KIT, RAF1, and MAP2K1) located only in SA breed are involved in the MAPK signaling pathway, which is an upstream regulator of melanogenesis and melanoma angiogenesis (Figure 3). In particular, the interaction of KIT with MITF has been reported to confer resistance to melanosis and the ability to cope with solar radiation (Satzger et al., 2008). One PSG (LIG3) in SA breed was involved in “negative regulation of mitochondrial DNA replication” and “base-excision repair, DNA ligation” (Table S9); two PSGs (SIRT1 and MC1R) of SL were significantly enriched in “UV-damage excision repair,” of which SIRT1 was selected in SA breed simultaneously based on the XP-EHH value (Table S11). We found significantly higher (p < 10⁻¹⁶, Mann–Whitney U test) selective pressure in the SA breed than in the SL breed for SIRT1, indicating that the genome characteristics were sensitive to solar radiation (Figure S11).

PSGs Associated with the Circulatory and Respiratory Systems

The living climates of the SA and SL breeds are characterized by very large diurnal ambient temperature differences and high daytime temperatures caused by intense solar radiation (Table S1). Indigenous chickens use a variety of solutions to maintain core body temperature and internal homeostasis, such as convective and evaporative heat loss through diastolic blood vessels (Lara and Rostagno, 2013; Mutaf et al., 2009). Therefore, the balance between vascular contraction and relaxation plays a pivotal role in dissipating excessive heat through regulation and maintenance of blood pressure to maintain core body temperature and internal homeostasis in chickens. Eighteen PSGs involved in the VSMC pathway encode seven vasoconstrictors, eight vasodilators, two voltage-dependent calcium channel L-type alpha proteins, and one calcium-activated potassium channel subunit alpha protein. Notably, three genes (PRKCH, ADCY1, and CACNA1C) were simultaneously selected by SA and SL breeds and separately activated vasoconstrictors, vasodilators, and voltage-dependent calcium channels (Figure 4A). Furthermore, we performed quantitative real-time PCR (qRT-PCR) to investigate changes in gene expression level between heat stress and normal temperature using the C2C12 cell line. We found significantly higher expression levels of CACNA1C under heat treatments (Figure 4B). Subsequently, we analyzed the population genotype frequency in a large dataset of 845 Gallus gallus accessions (Wang et al., 2020). We first divided 845 accessions into four groups, of which 109 domesticated chickens live in tropical climates (TRCs) (Table S12), 112 are TCs, 475 are other domesticated chickens (non-tropical chickens, or NTRCs), and 149 are RJFs. Based on the recently released population SNP data of chickens (http://bigd.big.ac.cn/chickensd/), we screened out 50,582 SNPs within CACNA1C gene and up-/downstream 500-kb regions based on 845 accessions. We discovered directional increased or reduced genotype frequency through a priority order with TRCs → NTRCs → TCs, as demonstrated by a linear regression model (Transparent Methods); especially 147 mutations (29.05%) within top 1% outlier of absolute K value belonged to the region of CACNA1C gene (Figure 4C). Indeed, 11 PSGs in SA breed and 14 PSGs in SL breed were significantly enriched (p < 0.05) in GO terms “regulation of blood circulation” (Table S9). More specifically, the significantly enriched GO terms were “regulation of systemic arterial blood pressure by hormone” (RASL10B, DRD3 and RPS6KA2) and “blood vessel maturation” (MMP2 and ANKRD17) in SA breed, whereas GO terms included “regulation of blood vessel diameter” (ADRA2C, FGG, FGA, SMTNL2, DOCK5, GRIP2, UTS2R, and SCPEP1) in SL breed.

Figure 4.

Genomic Regions with Strong Selective Sweep Signals in SA/SL Breeds

(A) Example of genes with strong selective signals in SA and SL breeds. Genomic regions located above the upper horizontal red, green, and yellow dashed lines and under the horizontal blue dashed line were termed regions with strong selective sweep signals (gray regions). Genome annotations are shown in the middle (gray bar, coding sequence; purple bar, gene body). The boundary of CACNA1C is marked in red.

(B) Quantitative real-time PCR (qRT-PCR) for the CACNA1C gene in C2C12 cells under 42°C for heat stress (HS) and 37°C for the normal control (NC) groups (n = 3 replicates per group; two-sided Student's t test (∗∗∗: p < 0.001, ∗∗: 0.001 < p < 0.01, ns: non-significant); data are represented as mean ± SEM).

(C) Genotype frequency analysis detects selection signals upstream and downstream 500 Kb around CACNA1C gene based on 845 chickens. Below left showed directional reduced frequency of 52 mutations and below right showed directional reduced frequency of 95 mutations across five chicken populations in CACNA1C gene.

(D) Selection signals in the SA breed for the VPS13C gene. The upper half is similar to (A).

(E) qRT-PCR for the VPS13C gene in C2C12 cells under HS and NC groups (n = 3 replicates per group; two-sided Student's t test (∗∗∗: p < 0.001, ∗∗: 0.001 < p < 0.01, ∗: 0.01 < p < 0.05, ns: non-significant); data are represented as mean ± SEM).

(F) P genotype frequency analysis detects selection signals upstream and downstream 500 Kb around VPS13C gene based on 845 chickens. Below left showed directional reduced frequency of seven mutations and below right showed directional reduced frequency of 52 mutations across five chicken populations in VPS13C gene.

Heat stress causes excessive oxygen demand in animals (Portner, 2001), leading to increased blood flow in circulation and optimized oxygen delivery. Consistent with this phenomenon in SA breed, we discovered that three PSGs (EPAS1, CREBBP, and HIF1AN) are significantly involved in the “cellular response to hypoxia,” whereas five PSGs (SIRT1, PPARD, HP1BP3, SDHD, and HIF1AN) were involved in the “cellular response to decreased oxygen levels” (Table S9). Among those genes, EPAS1 was detected by PBS (PBS value = 0.096; p = 0.007), and was known as hypoxia-inducible factor 2α (HIF-2α), and HIF1AN was involved in transcriptional repression through interaction with HIF1α, VHL, and histone deacetylases. Also, a unique set of regulatory targets was acted by each of HIF-1α and EPAS1 (Hu et al., 2003). It has been suggested that there is a close relationship between EPAS1 and the regulation of red blood cell production (Percy et al., 2008; Yi et al., 2010). Indeed, we found one PSG (KIT) that is involved in “cellular response to erythropoietin” (Table S9). One SNP of EPAS1 with the greatest genotype frequency difference between SA and TCs is located in the sixth intron, with a derived allele at 0% frequency in TCs and 70% in the SA breed, as well as ≤ 10% in SL breed and RJFs (Table S13). This is consistent with a previous finding that the intron 5 of EPAS1 harbored a highly differentiated SNP associated with erythrocyte abundance to positively influence adaptation to hypoxia (Yi et al., 2010). Similarly, a previous study also found that a non-synonymous SNP in the exon 7 of EPAS1 may contribute to the hypoxia adaptation of TCs (Li et al., 2017b). To test whether positive selection has acted on different regions of EPAS1 for coping with heat stress and high altitude, we analyzed genotype frequency of EPAS1 from 845 G. gallus accessions (Wang et al., 2020). In addition to one SNP in the sixth intron of EPAS1 identified in our dataset, two more SNPs located in the sixth intron with a significant frequency preference were found in tropical chickens using the large dataset of 845 accessions (Table S14). By contrast, the corresponding genotypes in TCs and RJFs were fixed, indicating that this trait was retained before the ancient divergence of chickens. However, after comparing these genotypes in 109 TRCs and 475 NTRCs with the background of RJFs and all 845 chickens, we observed significant deviations of p values using chi-square tests, indicating that the two SNPs in tropical chickens are not random and lineage-specific positive selection should have favored the tropical chickens (Table S15). Indeed, our PBS method was considered to be robust for identifying recent natural selection. Therefore, although the regulation mechanism of EPAS1 by the intronic mutations remains to be discovered, these findings supported that the naturally occurring tropical stress may have led to the alleles targeted by natural selection, which may confer a functionally relevant adaptation to excessive oxygen demand. This result also indicated that different mutations of same genes may have occurred to meet similar physiological needs to cope with different environment stresses.

Additionally, chickens can expel internal heat by evaporation of moisture by increasing the frequency of panting (Collier and Gebremedhin, 2015). Correspondingly, we found that one PSG (EDA) was involved in “trachea submucosa development” in SA breed (Table S9). These functions play an important role in the regeneration and repair of airway epithelial cells (Tata et al., 2013). Therefore, the synergistic action of these biological functions may have enabled chickens to cope with heat stress in the heart, a trachea burden, and vascular pressure.

PSGs Associated with Mitochondrial Respiration

We also found that 15 PSGs and 5 PSGs are associated with the cellular component term “mitochondrial membrane” in SA and SL breeds, respectively (Tables S9 and S11). Under heat stress, oxygen levels in chicken body fluids may decrease, reflecting excessive oxygen demand (Portner, 2001). The adjustment of mitochondrial densities in addition to molecular or membrane adjustments appears crucial for maintaining aerobic scope and for shifting thermal tolerance. Of the eight PSGs (NDUFS3, NDUFA4, NDUFB5, NDUFA9, SDHD, VPS13C, PARK2, and PACRG) involved in the response to the mitochondrial respirasome based on the annotations in the PANTHER and UniProt databases, four (SDHD, VPS13C, PARK2, and PACRG) were under selection in both SA and SL breeds. VPS13C and PARK2 play a role in mitochondrial maintenance, such as in the regulation of mitochondrial respiration rates, suggesting compensatory adaptation aimed at preserving mitochondrial transmembrane potential levels (Lesage et al., 2016; Mortiboys et al., 2008). In our study, SDHD was involved in complex II of the mitochondrial electron transport chain, and this gene is responsible for transferring electrons from succinate to ubiquinone (Tables S9 and S11). We hypothesize that VPS13C regulates parkin-mediated stimulation of mitophagy in response to mitochondrial depolarization, thereby inhibiting the generation of reactive oxygen species and reducing irreversible mitochondrial damage (Figure 4D) (Youle and van der Bliek, 2012). The higher expression level of VPS13C in the C2C12 cell line under heat treatment may also support this hypothesis (Figure 4E). In addition, genotype frequency analysis of 845 accessions showed that 59 mutations (16.95%) within top 1% outlier of absolute K value are located within the region of VPS13C gene (Figure 4F; Transparent Methods) (Wang et al., 2020).

PSGs Associated with Defense against Zoonotic Diseases

Avian influenza and salmonellosis are caused by pathogens capable of infecting humans and animals and cause significant morbidity and mortality worldwide. Chickens are an important medical model for studying low-pathogenicity avian influenza viruses due to partial immunity owing to previous exposure (Seo et al., 2002). The Saudi Arabian climate, as a high-temperature climate, is conducive to the propagation of pathogenic bacteria in chicken coops, resulting in poultry that are vulnerable to infection. Cytokines are key players in the regulation of the immune response, particularly during infection, joint inflammation, and endocrinological autoimmune diseases (Evans, 1993). Among 942 PSGs of SA breed, we found that 60 were significantly ( < 0.05) involved in several functional categories related to infection and defense against zoonotic diseases, including the influenza A pathway (14 PSGs), the Salmonella infection pathway (11 PSGs), cytokine-cytokine receptor interaction (13 PSGs), immune system development (26 PSGs), and immune response (25 PSGs) (Tables S8 and S9). These functional terms indicate that the thermal climate can indirectly affect autoimmune regulation in animals through its effects on the environments that chickens inhabit.

Of these PSGs, TLR7 is a key component of innate and adaptive immunity and is involved in RNA virus recognition, especially in the recognition of highly pathogenic avian influenza viruses. Small antiviral compounds activate immune cells via the TLR7-MyD88-dependent signaling pathway. TLR7 may activate IRF7 through activation of MyD88, BTK, and TRAF6, thus inducing antiviral responses via the production of IFN-α. IFNGR2 interacts with IFNGR1 to form a receptor for the cytokine IFN-γ (Soh et al., 1994). A previous study reported that loxoribine induces antiviral gene expression, such as the expression of type I IFNs (IFN-α and IFN-β) and IFN-γ in primary chicken splenocytes, and can inhibit influenza A replication in vitro and in vivo in a dose-dependent manner (Stewart et al., 2012). The PSG IL12B is a cytokine that can act as a growth factor for T and NK cell activation, which can enhance the lytic activity of natural killer (NK)/lymphokine-activated killer cells, and stimulate the production of IFN-γ by resting peripheral blood mononuclear cells (Oppmann et al., 2000). Another PSG IL18 is a proinflammatory cytokine primarily involved in polarized T-helper 1 cell and NK cell immune responses (Tominaga et al., 2000). Another study found that members of the Toll-like receptor (TLR) family are critical for the recognition and clearance of Salmonella (Talbot et al., 2009; Vazquez-Torres et al., 2004). One consequence of Salmonella-induced TLR activation is the production of inflammatory cytokines and antimicrobial compounds, including pro–IL-1β, pro–IL-18, IFN-γ, TNF-α, and reactive oxygen species, which are critical mediators for the control of bacterial growth in host tissues (Eckmann and Kagnoff, 2001).

Remarkably, within the TLR7 gene, we found that two nearly fixed missense mutations in SA breed, i.e., Val121-Ile (G361-A) and Thr135-Ser (A403-T), have significantly higher frequencies than those observed in the other breeds (Figure 5A), possibly due to strong selective pressure. In addition, a similar pattern was identified in the gene ZC3HAV1, which has high breed genotype spectrum frequencies and harbors three missense mutations (Figure 5B). ZC3HAV1 encodes an antiviral protein that inhibits the replication of viruses by recruiting cellular RNA degradation machinery to degrade viral mRNAs (Hayakawa et al., 2011). In addition, our genotype frequency analysis of 845 accessions showed that 12 mutations within top 1% outlier of absolute K value are located within the region of these two genes (TLR7 and ZC3HAV1) (Wang et al., 2020) (Figures 5C–5F; Transparent Methods).

Figure 5.

Selective Sweep Signals with two PSGs Associated with Defense Against Zoonotic Diseases

(A) Selective signals in SA breed for the TLR7 gene. The upper half is similar to Figure 4A. Allele frequencies of variants within the TLR7 gene across seven chicken subclusters are shown at the bottom.

(B) Allele frequencies of five SNPs within the ZC3HAV1 gene across seven chicken subclusters.

(C) Genotype frequency analysis detects selection signals upstream and downstream 500 Kb around TLR7 gene based on 845 chickens.

(D) Genotype frequency analysis detects selection signals upstream and downstream 500 Kb around ZC3HAV1 gene based on 845 chickens.

(E) Regional plots for three directional mutations across five chicken populations in TLR7 gene.

(F) Regional plots for nine directional mutations across five chicken populations in ZC3HAV1 gene.

Thus, the immune adaptation of the chickens was shaped by the hot environment for resistance to invasion by pathogens. Our findings suggest the importance of specific advantageous alleles for the defensive response to pathogens in chickens.

Convergent Evolution to Tropical Desert Climate between Birds and Mammals

Intriguingly, we identified seven PSGs (CYP2J19, CYP2J21, CYP2J22, CYP2J23, PLA2G12B, TBXAS1, and HPGDS) that are significantly enriched in the pathway of “Arachidonic acid metabolism” and three PSGs (OC90, DRD3, and PLA2G12B) related to the GO term “arachidonic acid secretion (GO: 0050482, p < 0.05)” specific in SA breed (Tables S8 and S9). The arachidonic acid metabolism pathway plays a key role in converting arachidonic acid to hydroxyeicosatetraenoic acids (HETE) and exhibits strong expression in the kidney (Rouault et al., 2003). Members of the CYP2J family could help convert arachidonic acid into 19(S)-HETE, which has been demonstrated to be a potent vasodilator of renal preglomerular vessels for stimulating water reabsorption (Carroll et al., 1996; Croft et al., 2000; Escalante et al., 1991). In addition, CYP2J2 is regulated by high-salt diet and its suppression can lead to high blood pressure (Carroll et al., 1996). We also observed specific genotype blocks of the two genes (OC90 and PLA2G12B) in SA breed, which have significantly higher or lower frequencies than those observed in the other breeds (Figures 6A and 6B). Indeed, PLA2G12B catalyzes the conversion of arachidonic acid to 20-HETE eicosanoid, which is a potent vasoconstrictor produced in vascular smooth muscle cells that depolarizes and stimulates angiogenic responses in vivo (Figure 6C) (Miyata and Roman, 2005). Consistently, we also identified in the SA breed two PSGs (KIF26B and AR) involved in urogenital system development and one PSG (CRH) harboring positive regulation of defecation (Table S9), which may be involved in the response to renal stress. In conclusion, those genes are important for the adaptation of chickens to tropical desert climates. Additionally, the arachidonic acid metabolism pathway was reported to be important for desert adaptation in the Bactrian camel (Jirimutu et al., 2012) and sheep (Yang et al., 2016), indicating possible convergent evolution between chickens and mammals for inhabiting similar tropical desert climates.

Figure 6.

Adaptively Convergent Genes Linked to Arachidonic Acid Metabolism

(A) Selective signals around the PLA2G12B locus. Genomic regions located above the upper horizontal red, green, and yellow dashed lines and under the horizontal blue dashed line were termed regions with strong selective signals for SA breed (gray shading). The structure of this gene is shown in the middle of the figure panel (gray bar, coding sequence; purple bar, gene body), and the region of this gene showing selection signals is marked in red. Allele frequencies of variants within the PLA2G12B gene across seven chicken subclusters are shown at the bottom.

(B) Allele frequencies of five SNPs within the OC90 gene across seven chicken subclusters.

(C) Schematic diagram showing the metabolic pathways of arachidonic acid and its derivatives. The left panel represents a part of the arachidonic acid metabolic pathway. The upper right represents the Reactome R-HSA-391937 pathway for derivatives, and the bottom right indicates the GO Ancestor chart of derivatives.

(D) Phylogenetic tree for the 11 species of birds and mammals used for identification of molecular convergence.

(E) A tree model illustrating the counting of the numbers of observed and expected molecular convergences between two thick branches. For a given position, the amino acids at nodes 0–4 are indicated by S0–S4. The relevant branch lengths are indicated by the b values.

(F) Negative correlation between the observed number of molecular convergences relative to the expected number (R) and the genetic distance between the two branches concerned. Each dot represents one branch pair, and different colors show the results under different substitution models. The R values under the JTT-f_gene are based on 4,844 proteins. Genetic distance is the number of amino acid substitutions per site between the two younger ends of the two branches considered. Lines show linear regressions.

Molecular convergence is viewed as evidence for common adaptations in divergent lineages. We used 11 species, including six birds and five mammals (Figure 6D), to further identify convergent amino acid substitutions for tropical climate adaptation by applying an established method for detecting convergent and parallel evolution (Zhang and Kumar, 1997). Of these species, three also experience desert stress, such as the southern ostrich (Struthio camelus australis) (Zhang et al., 2015), which is abundant in dry areas with low rainfall in Africa; the yellow-throated sandgrouse (Pterocles gutturalis), which was collected from the Sharjah Breeding Center in the United Arab Emirates (Zhang et al., 2014); and Bactrian camels (Jirimutu et al., 2012), which live in desert regions. We first replaced the bases of gene sequences from the reference genome with the corresponding fixed genotypes in SA chickens, where the fixed genotype was defined if the allele frequency was more than 0.8. We identified 4,844 single-copy orthologs using OrthoMCL (Li et al., 2003) and then compared the observed number of convergent sites with the random expected number for each gene under the JTT-f_gene amino acid substitution models (Figures 6D and 6E) (Zou and Zhang, 2015). As expected, the ratio of molecular convergence between the numbers of observed and expected values declined with the increase in genetic distance between the two branches compared (Figure 6F). Only genes with a significantly higher number of observed convergent sites were considered to be truly under convergent adaptation (p < 0.05, Poisson test) (Thomas and Hahn, 2015). Subsequently, we screened convergent genes harboring nonsynonymous amino acids on the basis of PGVs in SA breed. We observed 11, 32, and 37 adaptively convergent genes (ACGs) between SA chickens and Bactrian camels, the southern ostrich, and the yellow-throated sandgrouse, respectively. Gene function enrichment analyses revealed several significant terms involved in the regulation of the response to stimulus and stress (Table S16), especially terms related to “immune system development,” “cardiac muscle adaptation,” “glomerular development,” and “blood vessel development.”

Among the 11 ACGs detected between SA chickens and Bactrian camels, GPR17 has been identified as a dual uracil nucleotide and cysteinyl-leukotriene (CysLT) receptor and is involved in the Reactome pathway named “UDP/CysLT receptor can bind cysteinyl leukotrienes” (R-HSA-391937) (Ciana et al., 2006). CysLTs are derived from the ubiquitous membrane constituent arachidonic acid and are members of a large family of molecules known as eicosanoids. Another convergently evolved gene, TNFRSF11A, functions upstream of the biological process “positive regulation of eicosanoid secretion” (GO: 0032305) (Figure 6C). These findings confirmed that arachidonic acid metabolism was an important molecular adaptive convergence mechanism for the tropical desert climate (Table S16). The kidney is a key tissue for regulating water retention and reabsorption. Interestingly, we found several genes associated with kidney size and function, such as ASXL1, which are involved in “regulation of kidney size” (GO: 0035564) in the southern ostrich, and KLF15, which is involved in “glomerular epithelium development” (GO: 0072010) in the yellow-throated sandgrouse. Additionally, GPR17 and TNFRSF11A were also involved in the biological process “regulation of inflammatory response” (GO:0050727), and similar convergence mechanisms were also found in the southern ostrich and the yellow-throated sandgrouse, which indicated that heat stress could increase inflammatory signaling (Table S16) (Ganesan et al., 2016; Pearce et al., 2013).

Conclusions

In sum, selective sweep analyses of the chicken genomes revealed a variety of important genes, pathways, and GO categories associated with genetic adaptations of chicken to tropical desert/monsoon island climates. Specifically, 12 PSGs (ADCY1, CACNA1C, CAMK2D, PACRG, PARK2, PRKCH, SDHD, SIRT1, WNT7B, TBXAS1, IL18, and VPS13C) were detected to play roles in chicken adaptations to both tropical desert and tropical monsoon island climates. Our study revealed that nerve development, cutaneous melanin formation, hormonal regulation, angiogenesis, and vasodilation may have worked together to enable chickens to adapt to heat stress. We also discovered different mutations of the same EPAS1 gene to cope with different environment stresses. Furthermore, a number of PSGs are functionally related to the immune adaptation to the defense against zoonotic diseases and water reabsorption in the tropical desert climate; some of these findings were confirmed by analysis of population genotype frequency in a large dataset of 845 chicken accessions just released. In addition, we provided insights into the genetic mechanism of water reabsorption that have resulted in adaptive convergence between birds and mammals inhabiting tropical desert climate. In view of the current and future global effects of climate change, our study adds to knowledge on genetic mechanisms of adaptations to tropical climates in vertebrates and also has inestimable value for breeding heat-, drought-, and stress-tolerant chicken lines and/or breeds, as well as for medical research related to zoonotic diseases.

Limitations of the Study

To make up for the adverse effect caused by the small sample size, we combined four classic methods to independently identifying genome-wide selection signals for two chicken breeds inhabiting both types of tropical climates; this could reduce the false-positive rate caused by the deviation of a single method. We subsequently revealed genetic adaptations to tropical climates in chickens of two populations inhabiting two similar tropical climates and confirmed some of these results by analysis of the population genotype frequency in a large dataset of 845 chicken accessions just released. Furthermore, we employed comparative genomic analysis between six birds and five mammals, combining with population genomics analysis, to explain shared genetic mechanisms associated with adaptations to the tropical desert climate from divergent lineages. Due to the protection of local chicken breeds abroad, it is difficult to obtain samples with similar genetic backgrounds. In the future, more samples from diverse breeds are needed to further understand the genetic mechanisms of adaptations to tropical climates in chickens or other vertebrates.

Resource Availability

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Huabin Zhao (huabinzhao@whu.edu.cn).

Data Availability

The sequence data of 67 chicken genomes in this study are available under the NCBI accession numbers of SRP067615, PRJNA241474, and PRJNA453469.

Materials Availability

This study did not generate new reagents or other materials.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.

Published: November 20, 2020

Supplemental Information

Table S9. Gene Ontology Enrichment of PSGs in SA Breed from the Comparison between SA and TC2_Sub2. Related to Figures 2–6 (See the Excel File for This Table)

PANTHER is used for Gene Ontology enrichment analysis of GO biological process (GO-BP), molecular function (GO-MF), and cellular component (GO-CC) terms. Ontologies with a cutoff value of 0.05 for p values as being biologically significant.

Table S11. Gene Ontology Enrichment of PSGs in SL Breed from the Comparison between SL and TC2_Sub2. Related to Figures 2–4 (See the Excel File for This Table)

PANTHER is used for Gene Ontology enrichment analysis of GO biological process (GO-BP), molecular function (GO-MF), and cellular component (GO-CC) terms. Ontologies with a cutoff value of 0.05 for p values as being biologically significant.

Table S16. Gene Function Enrichment for Significantly Convergent Genes in Two Divergent Lineages. Related to Figure 6 (See the Excel File for This Table)

The three comparison groups are SA Chicken versus Bactrian camels, SA Chicken versus Southern Ostrich, and SA Chicken versus Yellow-throated Sandgrouse. Ontologies with a cutoff value of 0.05 for p values as being biologically significant.