28
Genomic prediction of host response to co-infection with PRRSV and PCV2b using a PRRSV-only infected training population
The objective of this study was to evaluate the accuracy of genomic prediction of host response to co-infection with porcine reproductive and respiratory syndrome virus (PRRSV) and porcine circovirus type 2b (PCV2b) using a PRRSV-only infected training population. The training dataset included 9 PRRS Host Genetics Consortium (PHGC) trials of PRRSV-infected commercial crossbred nursery pigs: 6 trials infected with the NVSL isolate (n=1055) and 3 trials with the KS06 isolate (n=528). PRRS viral load (VL; area under the curve [AUC] of log-PCR viremia from 0-21 days post-infection [dpi]), and weight gain (WG) from 0-42 dpi were evaluated. Effects of 40,813 SNPs were estimated using Bayes-B for the following training scenarios: NVSL trials, KS06 trials, and all trials, with or without genotype for the WUR SNP (previously associated with response to PRRSV infection) as a fixed effect, to evaluate predictive ability of the remainder of the genome. Data from an independent, unrelated PHGC trial in which commercial nursery pigs (n=203) were co-infected with PRRSV and PCV2b were used for validation. For the validation dataset, PRRS and PCV2b VL were defined as AUC from 0-21 and 0-42 dpi, respectively, and were pre-adjusted for fixed effects, including or excluding WUR. Accuracy of genomic prediction was assessed as the correlation between estimated and pre-adjusted phenotypes. Since the training population was infected with PRRSV only, PCV2b VL of the validation dataset was predicted using PRRS VL. Pedigree-based heritability estimates in the validation data were 0.17±0.15, 0.28±0.13, and 0 for WG, PRRS, and PCV2b VL, respectively. When WUR was not fitted as a fixed effect for the training data, prediction correlations were 0.14, 0.07, and 0.22 for PRRS VL, 0.11, 0.12, and 0.10 for WG, and ~zero for PCV2b VL when training on both isolates, NVSL, and KS06, respectively. When WUR was excluded from predictions, corresponding correlations were 0.12, 0.14, and 0.11 for WG, -0.03, -0.11, and 0.03 for PRRS VL, and -0.13, -0.14, and -0.03 for PCV2b VL. The latter negative correlations with PCV2b VL may suggest a negative genetic correlation between PRRS VL of a PRRSV-only infected population and PCV2b VL in a population co-infected with PRRSV and PCV2b. In conclusion, genomic predictions from a PRRSV-infected population appear to have some predictive ability for host response in pigs co-infected with PRRSV and PCV2b. This work was supported by the USDA ARS NIFA award 2012-38420-19286 and by PIC/Genus and Choice Genetics.
Keywords: Pigs, Porcine Reproductive and Respiratory Syndrome, Porcine Circovirus