Newcastle disease (ND) is a global threat to domestic poultry, especially in rural areas of Africa and Asia, where the loss of entire backyard local chicken flocks often threatens household food security and income. To investigate the genetics of Ghanaian local chicken ecotypes to Newcastle disease virus (NDV), in this study, three popular Ghanaian chicken ecotypes (regional populations) were challenged with a lentogenic NDV strain at 28 days of age. This study was conducted in parallel with a similar study that used three popular Tanzanian local chicken ecotypes and after two companion studies in the United States, using Hy-line Brown commercial laying birds. In addition to growth rate, NDV response traits were measured following infection, including anti-NDV antibody levels [pre-infection and 10 days post-infection (dpi)], and viral load (2 and 6 dpi). Genetic parameters were estimated, and two genome-wide association study analysis methods were used on data from 1,440 Ghanaian chickens that were genotyped on a chicken 600K Single Nucleotide Polymorphism (SNP) chip. Both Ghana and Tanzania NDV challenge studies revealed moderate to high (0.18 – 0.55) estimates of heritability for all traits, except viral clearance where the heritability estimate was not different from zero for the Tanzanian ecotypes. For the Ghana study, 12 quantitative trait loci (QTL) for growth and/or response to NDV from single-SNP analyses and 20 genomic regions that explained more than 1% of genetic variance using the Bayes B method were identified. Seven of these windows were also identified as having at least one significant SNP in the single SNP analyses for growth rate, anti-NDV antibody levels, and viral load at 2 and 6 dpi. An important gene for growth during stress, CHORDC1 associated with post-infection growth rate was identified as a positional candidate gene, as well as other immune related genes, including VAV2, IL12B, DUSP1, and IL17B. The QTL identified in the Ghana study did not overlap with those identified in the Tanzania study. However, both studies revealed QTL with genes vital for growth and immune response during NDV challenge. The Tanzania parallel study revealed an overlapping QTL on chromosome 24 for viral load at 6 dpi with the US NDV study in which birds were challenged with NDV under heat stress. This QTL region includes genes related to immune response, including TIRAP, ETS1, and KIRREL3. The moderate to high estimates of heritability and the identified QTL suggest that host response to NDV of local African chicken ecotypes can be improved through selective breeding to enhance increased NDV resistance and vaccine efficacy.
All experimental protocols used in this study were approved by the University of California, Davis Institutional Animal Care and Use Committee (#17853). This study was designed similar to a parallel study conducted in Tanzania as described by Walugembe et al. (2019b). In brief, local breeder chickens were randomly sampled from three ecological zones of Ghana, namely Interior Savannah (IS), Coastal Savannah (CS), and Forest (FO) zones. Breeder birds were vaccinated for various poultry diseases, following recommendations by project veterinary personnel, with the exception of receiving no NDV vaccine. Local breeder chickens were grouped for natural mating to produce 25 sire half-sib families per ecotype, using a mating ratio of 1 sire to 8 dams to generate chicks for the NDV challenge study. Challenge experiments were conducted for a total of 1,440 chicks (411 IS, 511 CS, and 518 FO) from hatch to 38 days of age (doa) across four replicates (hatches). All chicks were raised under similar conditions with ad libitum access to feed and water. Chicks were challenged with a live attenuated type B1 LaSota lentogenic NDV strain at 28 days of age and evaluated for pre- and post-infection growth rate, antibody level in at 10 dpi, and viral load from tears at 2 and 6 dpi, as described by Walugembe et al. (2019b). Blood samples were collected using Whatman FTA cards (Sigma-Aldrich, St. Louis, MO, United States) from chicks before challenge. Genotyping and genotype quality control were as described by Walugembe et al. (2019b). A total of 403,165 SNPs with call rate > 99% and minor allele frequency > 0.05 remained after filtering (Table 1). Imputation of missing genotypes (<1% after quality control) for the 403,165 SNPs was performed using Fimpute (Sargolzaei et al., 2014). Because dams were housed in group pens, the 1,440 birds were assigned to half- and full-sib families based on their genomic relationships, with cutoffs of 0.18–0.37, 0.38–0.77, and <0.18 for half-sibs, full-sibs, and less related individuals, respectively, which were determined based on the distribution of the genomic relationships among the 1440 birds. Genotype quality metrics provided by Affymetrix and the requirements used in quality control filtering. The combined genotype data from the current study and that from the three Tanzanian local chicken ecotypes from the parallel study (Walugembe et al., 2019b) were used to examine the population structure within and across the two country populations using the PLINK v1.9 software (Chang et al., 2015). Shared ancestry of the Ghanaian local chicken ecotypes was explored using the Admixture software (Alexander et al., 2009), allowing the number of ancestral subpopulations to range from 1 to 6, with 3 ancestral subpopulations giving the lowest cross-validation error. The ancestral subpopulation proportions generated by admixture analyses for each individual bird were used as covariate effects in the downstream genetic analyses. Variance component estimates and heritabilities were performed using ASReml 4 (Gilmour et al., 2015), as described by Walugembe et al. (2019b), with some modifications to the model fitted to account for the presence of three rather than two ancestral subpopulations. The univariate mixed linear animal model was; where Y is the dependent phenotype variable: pre- and post-infection growth rate, antibody at 10 dpi, viral load at 2 dpi, and viral load at 6 dpi. Fixed effects included death prior to 10 dpi (D = 0/1), replicate (R = 1 to 4), and sex (S = male/female). Two covariates, ancestral subpopulation proportions (C and P) obtained from admixture analyses were fitted. Random effects included animal genetic effects (A) with a genomic relationship matrix computed based on method 1 of VanRaden (2008), dam (M) to account for maternal effects, and residuals (e). The dam effect (based on assigned full-sib families). The dam effect was removed for some traits for which it explained no variance. For viral load at 2 and 6 dpi, and antibody at 10 dpi, qPCR plate (55 and 57 plates, for 2 and 6 dpi, respectively) and replicate plate (40), respectively, were added as fixed effects. Phenotypic variance was obtained as the sum of variance due to animal, dam, and residuals. Heritability was computed as the ratio of the estimates of animal to phenotypic variance. Bivariate animal models were used to estimate pairwise phenotypic and genetic correlations between traits, with the same fixed and random effects as specified for the univariate analyses. Two whole genome association analysis methods were used for GWAS, as described by Walugembe et al. (2019b). These two method were utilized to identify any overlap in results to better understand the sensitivity of the underlying model assumptions. Briefly, method Bayes B (Habier et al., 2011), as implemented in the Gensel software (Fernando and Garrick, 2008), was used to estimate the genetic variance accounted for each one megabase (Mb) window of SNPs. Model (1) was used and effects were fitted as described in Walugembe et al. (2019b). One Mb regions that explained more than 1% of the genetic variance explained by all SNPs across the genome were considered significant. Gene annotation for the 1-Mb windows was completed using the Genome Data Viewer in NCBI on a Gallus gallus 5 genome version1. The R package GenABEL was used to identify single SNPs associated with the various NDV response traits (Aulchenko, 2015) using a hierarchical generalized linear model (Rönnegård et al., 2010). Model (1) was used for single-SNP analyses, with genotype at each SNP included as a fixed effect, one at a time. Genome-wide significance thresholds of 20 and 10% were derived based on a modified Bonferroni correction, as 0.2 or 0.1 divided by the number of independent tests (Rowland et al., 2018; Saelao et al., 2019; Walugembe et al., 2019b). A more relaxed significance threshold was used to reduce the number of false negatives and allow results to be compared to those from the Bayesian analysis and with results from previous and future studies.
N/A
DIMA AI Care
