Results
Genome sequencing and identification of variants. To detect genome-wide variation in three Chinese chicken breeds that have been conserved in situ , we genotyped 91 individuals using high-throughput sequencing (Figure 1). Alignment of 79.95 Gb of sequence data against the Gallus gallus 5.0 reference genome yielded > 8× average read depth (Table S1). In combination with genomic data obtained from the same three breeds conserved in ex situ programs (Zhang et al., 2018), a total of 5,070,414 variants were identified, including 4,709,112 SNPs and 361,302 short insertions and deletions (indels). Variants were evenly distributed across the genome (Figure S1a and Figure S2). 31.58% of the total SNPs were novel and had not been included in the dbSNP database at NCBI (Table S2 and Figure S1b). After removal of variations that did not meet quality criteria (MAF 0.01, HWE 10e-6), 1,518,758 SNPs remained for further analysis.
Population structure analysis. To investigate phylogenetic relationships and population structure amongst the 361 chickens, we constructed a neighbor-joining tree using a pairwise genetic distance matrix (Figure 2a) and performed principal component analysis (PCA) based on the variance-standardized genotype relationship matrix (Figure 2b). The neighbor-joining tree suggests that the samples form six major clusters that correspond to the three Chinese domestic chicken breeds, with further subdivision of each breed intoin-situ and ex-situ populations. This pattern was further confirmed by PCA. The first principal component (PC1, variance explained = 11.65%) successfully separated the Langshan chicken breed from the other groups. The second principal component (variance explained = 10.7%) separated all populations in the three chicken breeds (Figure S3). Notably, the PCA separated the in situ and ex situpopulations, especially for Beijing You chicken (Figure S3). To better understand p population ancestry, we used ADMIXTURE to estimate the number of ancestral populations (Alexander et al., 2009) and allowed population number (K) to vary from 2 to 9. The minimum estimated cross-validation error occurred at K=6 (Figure S4). These results suggest that the three Chinese domestic chicken breeds have distinct backgrounds and also differ between in-situ and ex-situpopulations, consistent with the results from the NJ tree and principal components analyses. The likelihood model based on K=6 resolves the three Chinese domestic chicken populations into six genetic clusters (Figure 2c). One individual from the in-situ conserved population of Beijing You chickens (YBYC) had a genetic background that was distinct from other individuals in the YBYC population, based on the NJ tree, PCA, and ADMIXTURE results. We therefore removed this individual from subsequent analyses.
Genomic diversity assessment. Analyses of genomic genetic variability parameters for the six sub-populations are presented in Figure 3 and Table 2. The parameters include observed (Ho) and expected (He) heterozygosity, allelic richness (AR), proportion of polymorphic SNPs (PN ), and inbreeding coefficient (F ). The genomic diversity in populations conserved in situ was higher than in those conserved ex situ . Observed heterozygosity (Ho) and expected heterozygosity (He) were similar for all three breeds in both in situ and ex situ conserved populations. For example, changes in genetic diversity between in situ conserved populations of the Beijing You chicken (YBYC, Ho = 0.2646, He = 0.2714) vs. ex situ (BYC15, Ho = 0.2729, He = 0.2658) were less than 5%. In contrast, allelic richness (AR) and proportion of polymorphic markers (PN ) for in situ conserved populations (AR = 1.209, PN = 0.7891) were higher than for ex situ (AR = 1.198, PN = 0.7258).
Estimation of inbreeding coefficients. To estimate the degree of inbreeding in in situ and ex situ conserved populations, we calculated FES andFROH across subpopulations. As expected,FES values increased while conservation procedures were maintained. This trend is also evident in the comparison of FES in in situ vs. ex situconserved chicken populations. Conservation practices have been applied for a longer period (conservation time; CT) for the in situpopulation than the ex situ population, and theFES values for the in situ population are correspondingly higher.
Since FROH is better at detecting both rare and common variants, we focused on this measurement in subsequent analyses. The inbreeding coefficient based on runs of homozygosity (FROH ) was relatively low, ranging from 0.0463 to 0.0958. Except for Langshan chickens, FROH inin situ conserved populations was lower thanFROH inex situ populations. The difference may be caused by the small size of the Langshan chicken in situ conserved population and the long conservation time (CT = 60 years). Inbreeding coefficients are compared for the current generation of all three chicken breeds in Figure 3 and Table 2.
Calculation of nucleotide diversity The results of a nucleotide diversity (Pi) survey are shown for the three breeds in Figure 4a. The YLSC (Pi = 0.000112582) had the highest average nucleotide diversity amongst the 12 subpopulations, followed in descending order by YBEC, LSC15, LSC12, YBYC, BEC10, BEC07, LSC10, BYC07, BYC10, BEC15, and BYC15. Pi was markedly higher in in situconserved populations than in ex situ in all three chicken breeds, and highly significant differences (P<0.001) were observed between populations within breeds.
Linkage disequilibrium decay Differences in LD decay between in situ and ex situ conserved populations are shown in Figure 4b. The highest maximum average LD (r2 = 0.2235) was observed in Beijing You chickens (BYC15), and the lowest (r2 = 0.1806) occurred in Baier Yellow chickens (YBEC). Compared to the current generation of ex situ conserved populations (BYC15, BEC15, and LSC15), maximum average LD values were lower in the in situconserved Beijing You chicken and Baier Yellow chicken populations, while higher values were observed in Langshan chickens. This may indicate that YBYC and YLSC have greater genetic diversity than BYC15 and LSC15. As expected, LD declined as the physical distance increased between pairwise SNPs. As shown in Figure 4b, LD decay inin situ conserved populations declined markedly compared withex situ populations for Beijing You chicken and Baier Yellow chicken. In contrast, LD decay. is similar in in situ andex situ conserved populations for Langshan chickens. Using Beijing You chickens as an example, r2decreased by half (from 0.1982 to 0.0991) over a span of 11.84 kb in thein situ group, while LD decayed by half over a span of 14.68 kb in the ex situ conserved population (BYC15).
Estimation of population differentiation using Fst To estimate population differentiation, we calculated pairwiseFST values across the sub-populations (Table S3). Values ranged from 0.004826 to 0.1508. FST values for all pair-wise comparisons are shown in Figure 5. For all three breeds, FST values amongst three successive generations were lower than 0.05. This result indicates that no or little genetic differentiation has occurred in the conserved populations from one generation to the next. Significant or moderate genetic differentiation is observed between breeds, and the maximumFST value was calculated between LSC15 and BYC15 (FST = 0.1508). Notably,FST values between in situ and ex situ conserved populations for all three breeds were greater than 0.05. In the case of the Beijing You chicken, FSTvalues have increased with time of conservation, and the maximumFST value was 0.1379 between BYC15 and YBYC. Overall, moderate genetic differentiation has occurred in in situand ex situ conserved populations among the three chicken breeds.
Effective population size (Ne) Ne is an important measure in conservation genetics, and conservation efforts strive to increase Ne. In order to estimate current Ne for conserved Chinese domestic chicken breeds, we used NeEstimator v2 (Do et al., 2014b), which applies a method based on linkage disequilibrium (LD) to calculate Ne using whole-genome SNPs markers. Effective population size (Ne) was estimated for autosomal chromosomes gga1 through gga28 (Table S4). Ne ranged from 2.7 to 167.4, with a mean of 43.81. Amongst macro-chromosomes (gga1-gga5), BEC15 exhibited the smallest estimated Ne (50.96), suggesting that BEC15 is a limited pool, whereas YBEC had the largest value (130.28), suggesting much higher genetic diversity. Importantly, Ne in in situ conserved populations was higher than in current generations of ex situconserved populations (Figure 6).
Runs of homozygosity The abundance and genomic distribution of ROH provide information about the demographic history of a livestock species. ROH were identified in the genomes of all in situ and ex situ conserved populations (Table S5). A genome-wide survey for autozygosity was conducted to identify regions with signatures that reflect ancient or recent inbreeding effects. Using estimates of FROH , maximum values were found in Beijing You chickens subjected to in ex situ conservation. In contrast, the minimum values occurred in Baier Yellow chicken breeds enrolled in in situ conservation programs (Table 3). BYC15, the current generation in an ex situ conserved population, had the highest level of inbreeding (0.1018). As expected, YBYC in the in situ conservation population had a lower level of inbreeding (0.0777) than BYC15. YBEC had the lowest level of inbreeding (0.0463) amongst all populations. However, within the Langshan chicken breed, YLSC (FROH = 0.0745) had a higher level of inbreeding than LSC15 (FROH = 0.0604).
All ROH were then assessed to determine whether any populations exhibit evidence of recent inbreeding. For BYC and BEC, the ex situconserved populations had longer ROH and lower genomic diversity than the in situ conserved populations in these breeds (Figure 7a). In contrast, the in situ conserved LSC population had a higher level of inbreeding than the ex situ conserved population. We also mapped ROH to the genome, and found that the homozygosity segments in populations subjected to in situ vs. ex situ conservation were distributed differently (Figure 7b).