Genetic relationships of antibody response, viremia level, and weight gain in pigs experimentally infected with porcine reproductive and respiratory syndrome virus

Abstract

Genetic and antigenic variability between Porcine Reproductive and Respiratory Syndrome Virus (PRRSV) isolates has encumbered vaccine development. Here, the genetic basis of PRRSV antibody response was assessed using data from experimental infection trials of commercial crossbred weaner pigs across with one of two distinct PRRSV isolates, NVSL-97–7895 (~750 pigs) and KS-2006–72109 (~450 pigs). Objectives were to estimate the genetic parameters of antibody response, measured as the sample to positive ratio (S:P) of PRRSV N-protein specific IgG in serum at 42 d post infection (dpi); assess the relationship of S:P at 42 dpi with serum viremia and growth under infection; and identify genomic regions associated with S:P at 42 dpi. Estimates of heritability of S:P at 42 dpi for NVSL and KS06 were 0.31 ± 0.09 and 0.40 ± 0.10 and appeared to be under similar genetic control (genetic correlation 0.73 ± 0.39). Estimates of genetic correlations of S:P were generally weak with viral load (NVSL: −0.20 ± 0.18; KS06: −0.69 ± 0.20), measured as area under the curve of log₁₀ serum viremia from 0 to 21 dpi, and with weight gain (WG) from 0 to 42 dpi (NVSL: −0.38 ± 0.19; KS06: −0.08 ± 0.25). However, genetic correlations of S:P at 42 dpi with daily serum viremia and with 3-d WG revealed dynamic relationships, with S:P at 42 dpi having the strongest negative genetic correlations with daily viremia when IgG production starts (10–20 dpi), and negative genetic correlations with WG early after infection but positive later on. This suggests that animals that placed more emphasis on immune response early in infection reaped benefits of that later in infection by more effectively clearing the virus. The WUR10000125 SNP on SSC4, previously associated with response to PRRSV, did not have a significant effect on S:P at 42 dpi (P > 0.05) but genotype-specific genetic correlations of S:P with daily viremia and 3-d WG suggested that the lower WG of pigs with the unfavorable AA WUR10000125 genotype may be due to their utilization of a more energetically costly host response compared to pigs with the favorable genotype. Genome-wide association studies identified three SNPs in the Major Histocompatibility Complex associated with S:P that explained ~10 (NVSL) and 45% (KS06) of the genetic variance but were not associated with viremia or WG. In conclusion, antibody response to PRRSV infection is a possible biomarker for improved host response to PRRSV infection.

INTRODUCTION

Porcine reproductive and respiratory syndrome (PRRS) is a disease of global concern from an economic and animal welfare standpoint (Holtkamp et al., 2013). Development of an effective PRRS vaccine has been largely unsuccessful due to poor cross-protection to heterologous PRRS virus (PRRSV) strains (Darwich et al., 2010). Pigs genetically selected for improved response to PRRS may aid in its containment (Lewis et al., 2007; Hess et al., 2016; Dekkers et al., 2017).

A robust PRRSV-specific nonneutralizing antibody response, e.g., nucleocapsid (N)-specific IgG, is observed early in infection and sustained long after serum clearance of PRRSV (Chand et al., 2012). The role this response plays in providing protection to PRRSV infection is unclear (Lopez and Osorio, 2004); however, PRRSV N-protein specific antibody response was found to be favorably genetically correlated with sow reproductive performance during a PRRSV outbreak (Serao et al., 2014, 2016). N-specific antibody response may also be genetically correlated with performance in weaner pigs infected with PRRSV, although this relationship may change over time due to differential expression of host genes during the course of infection (Schroyen et al., 2015).

The PRRS Host Genomics Consortium (PHGC; Lunney et al., 2011; Dekkers et al., 2017) collected data on PRRSV N-protein specific IgG response at 42 d post infection (dpi), along with periodic measurements of body weight and serum viremia, on weaner age pigs that were experimentally infected with either NVSL-97–7895 (NVSL) or KS-2006–72109 (KS06) PRRSV isolates. The objectives of this study were to use these data to: 1) estimate genetic parameters of antibody response to PRRS infection; 2) assess the relationship of antibody response with the dynamics of viremia or growth following PRRSV infection; and 3) identify genomic regions associated with antibody response to PRRSV infection. Estimates were obtained separately for each isolate and were compared to assess whether the genetic basis of PRRSV N-protein specific IgG response was consistent across these two isolates.

MATERIALS AND METHODS

Study Design

A detailed description of the design, sample and data collection, and assays used in the PHGC trials can be found in Lunney et al. (2011). The Kansas State University Institutional Animal Care and Use Committee approved all experimental protocols for these trials. Each of the eight PHGC trials used in this study (PHGC 1–5 and 10–12) consisted of ~200 crossbred pigs from commercial genetic sources used in North America, as previously described in Hess et al. (2016). Pigs within a trial were from the same high health farm, except for two trials (5 and 12), which each included pigs from two farms. All farms were free of PRRSV, Mycoplasma hyopneumoniae, and swine influenza virus. After weaning, pigs were transported to Kansas State University at an average age of 21 d and randomly placed into pens of 10 to 15 pigs. Following a 7-d acclimation period, all pigs were experimentally infected, both intramuscularly and intranasally, with 10⁵ (TCID50) of either PRRSV isolate NVSL-97–7985 (PHGC trials 1 to 5) or KS-2006–72109 (PHGC trials 10 to 12). NVSL and KS06 are Type II (North American) PRRSV isolates 89% similar at the GP5 nucleotide sequence level (Ladinig et al., 2015). One breeding company supplied pigs of the same breed cross for four trials; PHGC 1, 2, and 3 were infected with NVSL and PHGC 11 with KS06. Each of the other trials used in this study consisted of pigs from a different breed cross.

Body weights were measured weekly from 0 to 42 d post infection (dpi). Serum viremia was measured in blood samples collected at 0, 4, 7, 11, 14, 21, 28, 35, and 42 dpi using a semi-quantitative TaqMan PCR assay for PRRSV RNA, as described in Boddicker et al. (2012). PRRSV N-protein-specific serum Immunoglobulin G (IgG) level was measured in blood samples collected at 42 dpi using a fluorescent microsphere immunoassay (FMIA; Christopher-Hennings et al., 2013; Stephenson et al., 2015). Pigs were euthanized at 42 dpi and ear notches were collected for DNA genotyping on Illumina’s Porcine SNP60 Beadchip (Ramos et al., 2009) v1 (San Diego, CA) at GeneSeek Inc. (Lincoln, NE) for all trials, except 11 and 12, which were genotyped on Porcine SNP60 Beadchip v2 (San Diego, CA) at Delta Genomics (Edmonton, Alberta).

Across all five NVSL trials, 6% of pigs died before 42 dpi. Mortality rate was similar in the KS06 trials, with 7% of pigs dying before 42 dpi across the three trials. Dead pigs were necropsied and subsequent gross and microscopic pathology performed by a board-certified pathologist, who identified PRRS-associated disease as the major cause of mortality (Boddicker, 2013).

PRRSV N-Protein–Specific Antibody Response

To detect and semi-quantify antibodies or antigens in biological samples, the FMIA assay uses multiple fluorescent microspheres, with each bead set conjugated to different antigens or antibodies as the solid phase (Christopher-Hennings et al., 2013). Specifically, cloned PRRSV N protein was coupled to microspheres and IgG and IgM antibody response measured (Stephenson et al. 2015). Assays were conducted on a total of 21 96-well plates and results were reported as mean fluorescence intensity (MFI). A sample that fell within the linear phase of the standard curve (i.e., with an MFI value between 15,000 and 30,000 intensity) and that had ample serum was included as a positive control for every plate. A blank well (negative) was used as a negative control on each plate to correct for background noise. Resulting MFI values were then standardized to a sample to positive ratio (S:P), which was calculated for each animal as:

$$\text{S}:\text{P}=\frac{\text{Sample}-\text{Negative}}{\text{Positive Control}-\text{Negative}}$$ (1)

where sample is the MFI of an individual. Due to the high level of repeatability between technical replicates (Wang, 2013; Stephenson et al., 2015), serum from each animal was only measured once.

Models for Viremia and Weight Over Time

To capture the dynamics of changes in viremia over time, a single Wood’s curve (Eq. 2) or an Extended Wood’s Curve (Eq. 3) was fitted to the viremia data of each animal, as described by Islam et al. (2013) and Hess et al. (2016):

$$V\left(x\right)= a_1{x}^{b_1}{e}^{-c_{1}x}$$ (2)

$$V\left(x\right)= a_1{x}^{b_1}{e}^{-c_{1}x}+\max \left(0,a_2{\left(x-x_0\right)}^{b_2}{e}^{-c_2\left(x-x_0\right)}\right) + m$$ (3)

where V(x) is the level of viremia at x dpi (reported as log₁₀ of PRRSV RNA copies per ml of serum), parameter a₁ impacts the general level of viremia across time, b₁ is an indicator of the initial rate of increase to peak viremia, and c₁ is an indicator of the rate of decline after the peak and dominates the function as x → ∞. In the Extended Wood’s curve of Eq. 3, x₀ denotes the onset of the second phase of the profile, which was assumed to follow the same Wood’s shape as the primary phase and was thus defined by a second set of Wood’s curve parameters (Islam et al., 2013).

Wood’s curve parameter estimates $\widehat{a_1}, \widehat{b_1}, \widehat{c_1})$ were then used to obtain fitted viremia values for each pig for each day during infection. If the Extended Wood’s curve fitted a pig’s viremia profile better than the single Wood’s curve, as determined by AIC (Islam et al., 2013), that pig was classified as a rebound pig and estimates from the Extended model $\widehat{a_1}, \widehat{b_1}, \widehat{c_1},\widehat{t_0},\widehat{a_2}, \widehat{b_2}, \widehat{c_2})$ were used to obtain fitted values. Nonrebound pigs were further classified as cleared or persistent if their fitted log₁₀ viremia at 42 dpi was less than or greater than 1, respectively. For the trials used in this study, 22% of pigs from the NVSL trials and 6% of pigs from the KS06 trials were classified as rebound, and 33% and 53% of pigs infected with NVSL and KS06, respectively, were classified as persistent, with the remaining 45% and 41% classified as cleared. These classifications (cleared, rebound, persistent) will herein be referred to collectively as viremia status. Similar to previous analyses, we also evaluated viral load (VL), defined as the area under the Wood’s curve of log₁₀ viremia from 0 to 21 dpi (Hess et al., 2016), which is a measure of both the level of viremia and the rate and extent of clearance of the virus in the blood following infection, as:

$$V{L}_i=\underset{0}{\overset{21}{{\displaystyle \int }}}\widehat{a_{1i}}{x}^{\widehat{b_{1i}}}{e}^{-\widehat{c_{1i}}x} dx$$ (4)

Body weights were collected weekly and used to interpolate daily weights. A random regression model was fitted to the weight data of all pigs simultaneously, but separately for animals infected with NVSL and KS06, using second-order Legendre polynomials in the following model that was implemented in ASReml (Gilmour et al., 2009):

$$W_{ijklmop}\left(x\right)={\displaystyle \sum }_{n=0}^2{L}_{ni}(x)+P_j+A_k+S_l+{\displaystyle \sum }_{n=0}^2{L}_{ni}(x)\times R_m+{\displaystyle \sum }_{n=0}^2{L}_{ni}(t)\times A{n}_i+T{r}_o+ Pen{\left(Tr\right)}_{po}+{\epsilon }_{ijklmop}$$ (5)

where L_ni(x) denotes the n^th order Legendre polynomial at x dpi for individual i, with x ranging from 0 to 42 dpi, P is the parity of the dam (first, second, or later parities), A is the age of the individual at inoculation, S is the sex of the individual (barrow or gilt), and L_ni(t) × R is the interaction between the n^th order Legendre polynomial at x dpi and rebound status (0 or 1), to account for potential model differences in curve fittings between rebound and nonrebound pigs. Factors L_n(x), P, A, S and L_ni(t) × R were fitted as fixed effects, while L_ni(t) × An, Tr, and Pen(Tr) were included as random effects and denote the interaction of the n^th order Legendre polynomial at x dpi with animal, trial, and the interaction of trial and pen, respectively. Trial and Trial × Pen were included to capture systematic environmental effects. The interaction L_n(x) × An models an individual’s weight at each time point and was fitted using an unstructured variance–covariance structure for the polynomial parameters of an animal and independence of parameters between animals in order to capture both the genetic and permanent environmental effects for that animal. Residual variances were modeled separately for each measured dpi, to allow for changes in residual variance over time. Estimates from this model were used to obtain fitted values of each pig’s weight for each dpi (0–42) $\widehat{W(x)})$ using all coefficients estimated from model (5).

Parameter Estimation for Antibody Response

The model used for analysis of S:P was:

$$Y_{ijklmnopqr}=\mu +P_i+A_j+W_k+S_l+R_m+A{n}_n+L{i}_o+T{r}_p+ Pen{\left(Tr\right)}_{qp}+P{l}_r+{\epsilon }_{ijklmnopqr}$$ (6)

where Y is the dependent variable of S:P. Parity (P) sex (S), and rebound (R) were fitted as fixed class effects, as in model (5). Age (A) and weight (W) of the piglet at 0 dpi were included as linear covariates. Random effects included animal genetic effects (An), litter (Li), trial (Tr), Pen nested within trial (Pen(Tr)), the plate the assay was run on (Pl), and ε as the residual. Animal genetic effects were estimated using the genomic relationship matrix (G-matrix), including relationships between all animals, which was constructed using the VanRaden method (VanRaden, 2008), based on 48,164 autosomal SNPs (Build 10.2 http://www.ncbi.nlm.nih.gov/genome/guide/pig/, accessed 13 August 2015) that were segregating on both versions 1 and 2 of the Illumina Porcine SNP60 Beadchip. The heritability of S:P was estimated using model (6), separately for NVSL and KS06, using ASReml v3.0 (Gilmour et al., 2009). The phenotypic variance was obtained by summing the animal, litter, and pen within trial variance components. The genetic correlation between S:P for the two isolates was estimated using a bivariate model using Eq. 6, with S:P as the dependent variable separated by isolate.

A region on chromosome 4 of the swine genome (SSC4) was previously identified to be associated with both VL and WG in the PHGC trials (Boddicker et al., 2012) and a tag SNP, WUR10000125 (WUR), was identified that explained all of the genetic variance in this region for both traits (Boddicker et al., 2012, 2014a, 2014b). The numbers of animals with each genotype for WUR in the current study are given in Table 1. Model (6) was used to estimate LS means for S:P at 42 dpi for the two isolates, WUR genotype (AA or AB; BB animals were excluded for this analysis due to low frequency), and viremia status at 42 dpi. To estimate LS means for the two isolates, all data were analyzed jointly using model (6) with the effect of isolate added to the model. To estimate the association of WUR genotype with S:P, the interaction of isolate × WUR was also added to model (6). To estimate the association between viremia status and S:P, the interaction of isolate × viremia status was added to model (6) and rebound (R) was removed.

Table 1.

Least square means (SE) for antibody response at 42 d post infection by virus isolate, WUR genotype, and viremia status

Factor	Factor level	Isolate1
		NVSL (n = 746)	KS06 (n = 443)
Overall		0.87 (0.02)a	0.90 (0.04)a
WUR genotype2,3	AA	0.88 (0.02)a	0.90 (0.03)a
	AB	0.86 (0.03)a	0.90 (0.05)a
Viremia status	Cleared	0.84 (0.03)a	0.85 (0.04)a
	Rebound	0.90 (0.03)b	0.97 (0.05)b
	Persistent	0.85 (0.03)a	0.88 (0.04)a

¹Estimates within the same factor and isolate that share a letter are not significantly different from each other at P <0.05.

²NVSL: AA (n = 552), AB (n = 177), and BB (n = 17); KS06: AA (n = 382), AB (n = 60), and BB (n = 1).

³Animals with the BB genotype were not used for comparison due to low numbers.

Genetic Correlations of Antibody Response With Weight Gain and Viremia Over Time

Genetic correlations of S:P at 42 dpi with 3-d weight gain (WG; WG0-1, WG0-3, WG2-5, …, WG38-41) and with fitted viremia at every other dpi (i.e. 1, 3, 5, …, 41) were estimated in order to assess how the relationships of antibody response with WG and viremia changed over the course of infection. Genetic correlations were estimated using a bivariate model in ASReml that was the same as model (6) but without fitting plate for WG and viremia. The dependent variables were S:P at 42 dpi and either a fitted viremia at a given dpi or 3-d WG around a given dpi. For S:P and viremia, W in model (6) denotes weight at infection. In bivariate analyses of S:P with 3-d WG, estimated weights obtained from model (5) were fitted as the dependent variable for model (6), with W representing the fitted weight of the animal 3 d prior to the dpi fitted as the dependent variable. The exception was when WG from 0 to 1 dpi was assessed, for which weight at infection was fitted as a covariate. Estimated 42-d WG was also evaluated using model (6) by fitting the estimated weight at 42 dpi from model (5) as the dependent variable and weight at infection as the covariate W.

For NVSL-infected pigs, genetic correlations of S:P with viremia and WG were also estimated separately for animals with the AA vs. the AB genotype for the WUR SNP to assess the impact genotype at this SNP has on the relationships between these traits. The same analyses as described above in this section. The KS06 trials were not used for this analysis due to the limited number of animals. Animals with the BB genotyped were grouped with AB animals because of low numbers and because they were expected to perform similarly to AB animals, due to the dominant nature of this quantitative trait locus (Boddicker et al., 2014b). However, because models for the AB/BB animals did not converge, estimates of the genetic covariance between traits for these genotypes were obtained by assuming that the covariance between traits for the full NVSL data set (including AA, AB, and BB animals) was equal to a weighted sum of the covariances for the two genotype groups:

$$\text{Cov}{\left(S:P,T\right)}_{\text{Full}}=p_{AA}\text{Cov}{\left(S:P,T\right)}_{AA}+p_{AB/BB}\text{Cov}{\left(S:P,T\right)}_{AB/BB}$$ (7)

where T denotes either viremia or 3-d WG and p_AA and p_AB/BB refer to the proportion of animals with the AA and AB or BB WUR genotypes, respectively. The genetic correlation for AB/BB animals was then estimated by assuming equal genetic variances of S:P, viremia, and WG for AA and AB/BB animals.

Genome-Wide Association Study

Bayes B, as implemented in GenSel 4.73 (Fernando and Garrick, 2012), was used to test associations of the 60k SNPs with S:P based on the following model:

$$Y_{ijklmnopqr}=\mu +P_i+A_j+W_k+S_l+R_m+A{n}_n+L{i}_o+T{r}_p+ Pen{\left(Tr\right)}_{qp}+P{l}_r+{\displaystyle \sum }_{s=1}^t{z}_{ns}{\alpha }_s{\delta }_s+{\epsilon }_{ijklmnopqr}$$ (8)

where Y is the dependent variable of S:P and P, S, A, W, R, Tr, Pen(Tr), and Pl are as described for model (6) but with all effects fitted as fixed effects because GenSel does not accommodate random effects other than SNP effects; t is the number of SNPs analyzed; z is the genotype of SNP s for animal n (coded −1/0/1), α_s is the allele substitution effect for SNP s, and δ_s indicates whether SNP s was included (δ = 1) or excluded (δ = 0) in the model for a given iteration of the Monte Carlo Markov Chain (MCMC). The prior probability of δ = 0 was set equal to π = 0.99. A total of 50,000 iterations were run, including a burn-in of 5,000. A total of 0.19% of genotypes were missing and these were replaced with the mean genotype code for that SNP within a trial. To obtain MCMC samples of the breeding value for each animal for each nonoverlapping 1 Mb window of the genome, in each iteration of the MCMC, breeding values were calculated for each window and animal, using the sum of the animal’s genotypes multiplied by the sampled α_s and δ_s for SNPs in that window. These window-breeding values were then summed across the genome to get a sample of the overall breeding value of an individual. The additive genetic variance in each iteration was calculated as the variance of the sampled breeding values of all animals for that iteration. The percent genetic variance explained by a window was calculated by the variance of the sampled breeding values of all animals for that window divided by the variance of the sampled overall breeding values, multiplied by 100. The estimate of the percent genetic variance explained by a window was the posterior mean percent genetic variance explained by that window across all iterations (excluding the 5,000 burn-in; Garrick and Fernando, 2013). Windows that explained more than 5% of the genetic variance for S:P were deemed to have a strong association with antibody response. For the SNP or SNPs within each of these windows that had the highest posterior probability of inclusion (PPI), which was derived as the proportion of iterations of the MCMC chain that the given SNP was included in the model (Wolc et al., 2012), LS means were estimated by fitting SNP genotype as a fixed class effect in model (6).

RESULTS

Parameter Estimates for Antibody Response

Effects of isolate, WUR genotype, and viremia status.

The LS mean estimates for S:P for pigs infected with NVSL vs. KS06 were not significantly different from each other (P = 0.55; Table 1), nor was S:P significantly different between animals that had genotypes AA or AB for WUR, when infected with either NVSL (P = 0.28) or KS06 (P = 0.78). The effect of viremia status on S:P was significant, with rebound animals on average having a higher S:P ratio at 42 dpi than persistently infected or cleared animals, when infected with either NVSL or KS06 (P < 0.05).

Estimates of genetic parameters for S:P.

Moderate–high heritabilities were estimated for S:P at 42 dpi for both NVSL and KS06 infected animals (NVSL: 0.31 ± 0.09; KS06: 0.40 ± 0.10). The genetic correlation of S:P between isolates was estimated to be 0.73 ± 0.39. Although this is a moderately high correlation, the large SE prevented conclusions as to whether S:P can be considered the same trait during infection with these two PRRSV isolates. An attempt to improve the estimate by only analyzing trials with the same genetic background (i.e., same breeding company and same breed cross) was unsuccessful due to model convergence issues.

Correlations of Antibody Response With Viremia and Growth

Genetic and phenotypic correlations between S:P, VL, and WG were estimated separately for NVSL and KS06 (Table 2). All phenotypic correlations were negative and significantly different from zero (P < 0.05), with the exception of the phenotypic correlation between S:P and WG in KS06, which was estimated to be 0. Phenotypic correlations of VL and WG with S:P tended to be weak, while the phenotypic correlation between VL and WG was moderately negative for both NVSL and KS06. For NVSL, the genetic correlation between VL and S:P was not significantly different from zero (P = 0.27); however, a strong negative genetic correlation between VL and S:P was estimated for pigs infected with KS06 (−0.69 ± 0.20, P < 0.001). The genetic correlation between WG and S:P was moderately negative for pigs infected with NVSL (−0.38 ± 0.19, P < 0.05) but not significantly different from 0 for pigs infected with KS06 (P = 0.75). Genetic correlations between VL and WG were strongly negative (P < 0.01) for both virus isolates (−0.74 ± 0.10 for NVSL and −0.52 ± 0.17 for KS06).

Table 2.

Estimates of genetic and phenotypic correlations (SE) among viral load (VL), weight gain (WG), and antibody response (S:P) for pigs infected with NVSL or KS06

Traits	Genetic correlation		Phenotypic correlation
	NVSL	KS06	NVSL	KS06
VL, S:P	−0.20 (0.18)	−0.69 (0.20)	−0.09 (0.04)	−0.12 (0.05)
WG, S:P	−0.38 (0.19)	−0.08 (0.25)	−0.11 (0.04)	0.00 (0.06)
VL, WG	−0.74 (0.10)	−0.52 (0.17)	−0.33 (0.03)	−0.23 (0.05)

Relationship of Antibody Response With Viremia and WG Over the Course of Infection

Because it was suspected that the relationship of S:P with viremia or WG was more dynamic than was captured when using the overall measures VL and WG, these traits were broken down into viremia and WG at different time points during infection based on the Wood’s curve for daily viremia and 3-d WG obtained from the random regression model for weight. Phenotypic SD for daily viremia and 3-d WG were similar between pigs infected with NVSL and KS06 (Fig. 1A–D). Heritability estimates for daily viremia ranged from 0.12 (at 7 dpi) to 0.54 (at 13 dpi) for pigs infected with NVSL and had a similar range for KS06: 0.19 (at 27 dpi) to 0.47 (at 9 dpi). Heritability estimates for 3-d WG ranged from 0.11 (at 33 dpi) to 0.42 (at 1 dpi) for pigs infected with NVSL and from 0.17 (at 41 dpi) to 0.40 (at 19 dpi) for KS06.

Figure 1.

Estimates of heritabilities, phenotypic SD, and genetic and phenotypic correlations of antibody response at 42 dpi with daily viremia (A, B) and 3-d weight gain (C, D) at different dpi following infection with NVSL (A, C) or KS06 (B, D).

Genetic correlations of S:P at 42 dpi with daily viremia were estimated for all odd-numbered dpi. For NVSL, pre-peak viremia (before 7 dpi) was strongly positively correlated with S:P, with estimates of genetic correlations that were significantly (P < 0.05) different from 0 on days 1, 3, and 5, and that peaked at 5 dpi (0.58 ± 0.26) (Fig. 1A). After 5 dpi, the genetic correlation became negative, with estimates significantly (P < 0.05) less than 0 at 13 to 21 dpi and the lowest genetic correlation occurring at 17 dpi (−0.46 ± 0.19). However, for pigs infected with KS06, the genetic correlation between viremia and S:P was negative at all time points and significantly (P < 0.05) less than 0 from 5 to 33 dpi, with the lowest genetic correlation at 21 dpi (−0.78 ± 0.22) (Fig. 1B).

The genetic correlation of S:P at 42 dpi with 3-d WG changed from negative early in infection to positive late in infection for pigs infected with either NVSL or KS06 (Fig. 1C and D). In general, genetic correlations with 3-d WG were higher for KS06 than for NVSL (ranging from −0.20 ± 0.27 at 1 dpi to 0.78 ± 0.29 at 41 dpi). The only genetic correlations that were significantly (P < 0.05) different from 0 for KS06 were positive and occurred at 21 to 41 dpi (Fig. 1D). For NVSL, however, with estimates ranging from −0.48 ± 0.18 at 1 dpi to 0.39 ± 0.29 at 41 dpi, genetic correlations significantly (P < 0.05) lower than 0 were observed at 1 to 13 dpi, while none of positive correlations were significantly different from zero (Fig. 1C). For both daily viremia and 3-d WG, phenotypic correlations with S:P at 42 dpi tended to be weak, but showed a similar trend as the genetic correlations (Fig. 1A–D).

Genetic correlations of S:P at 42 dpi with viremia and WG across the experimental period were also estimated by WUR genotype for NVSL-infected pigs. When considering genetic correlations for animals with the AA genotype at the WUR SNP, the relationship of S:P with viremia and WG was more extreme than for the whole NVSL dataset (Figs. 1A and 2A). For AA animals, stronger negative correlations were observed between S:P and viremia than for all animals (Figs. 1A and 2A), with genetic correlations significantly different from zero from 11 to 27 dpi. Similar to using all NVSL-infected animals, the genetic correlation between S:P and viremia was lowest at 17 dpi but the genetic correlation at this dpi was much more negative for AA animals (−0.86 ± 0.24; Fig. 2A) than for all animals (−0.46 ± 0.19; Fig. 1A). Estimates of genetic correlations for AB and BB animals showed a consistent but generally low positive genetic correlation of S:P with viremia (Fig. 2A).

Figure 2.

Estimates of genetic correlations of antibody response at 42 dpi with viremia over time (A) and weight gain over time (B) for animals with the AA (red) or AB and BB (yellow) WUR genotype.

Animals with the AA genotype also had stronger negative genetic correlations of S:P at 42 dpi with WG than all animals, with a persistent negative correlation for AA animals of ~−0.6 from 1 to 17 dpi (P < 0.05), which then went positive at 29 dpi (Fig. 2B). Conversely, AB and BB animals started with a negative correlation between S:P and WG, which switched to positive at 11 dpi, and remained positive thereafter, with a maximum at 23 dpi of 0.57.

Genome-Wide Association Study

Genome-wide association studies (GWAS) for S:P ratio were conducted separately for the NVSL and KS06 data, as well as combined. All analyses showed strong associations in the Major Histocompatibility Complex (MHC) region on chromosome 7 (SSC7; Fig. 3), specifically for one SNP at the distal end of the MHC class I region and for two SNPs in the MHC class II region (Fig. 4). The 29 Mb window on SSC7 (MHC class II region) was identified as an important region in all three GWAS and explained 10.3% of the genetic variance for NVSL, 43.1% for KS06, and 29.9% for the combined analyses. The SNP ALGA0039771 (ALGA) consistently had a high PPI in the model (PPI) compared to other SNPs in this window (NVSL: 0.30; KS06: 1.00; Joint: 1.00). Another SNP in this window, MARC0058875 (MARC), had a high PPI for NVSL but a low PPI for the KS06 and combined GWAS (NVSL: 0.40; KS06: 0.01; Joint: 0.06). The joint analysis also identified the 26 Mb window on SSC7 to be associated with S:P ratio, explaining 13.8% of genetic variance, but this window explained less than 1% of genetic variance for the separate NVSL and KS06 analyses. The SNP DIAS0000349 (DIAS) in this window had a PPI equal to 1.00 for the joint analysis but a PPI of only 0.05 and 0.01 for the NVSL and KS06 analyses, respectively.

Figure 3.

Manhattan plots of percent of genetic variance explained by each 1 Mb window of the genome for S:P ratio in pigs infected with NVSL (A), KS06 (B), and the combined analysis (C).

Figure 4.

Linkage disequilibrium plot of for SNPs in the 26–30 Mb region of chromosome 7.

The LS means for S:P at 42 dpi were estimated for each SNP identified by the GWAS (Table 3). The DIAS SNP in the 26 Mb window on SSC7 had a significant association with S:P at 42 dpi for both NVSL and KS06 infected animals (NVSL: P < 0.0001; KS06: P = 0.0002), with animals with the AB genotype having significantly lower S:P than AA animals. The MARC SNP in the 29 Mb window on SSC7, which was identified only in the GWAS of animals infected with NVSL, had a significant association with S:P at 42 dpi for NVSL-infected animals (P < 0.0001) but not for KS06 infected animals (P = 0.16). Animals with the AA genotype for this SNP had significantly higher S:P than AB or BB animals. The ALGA SNP in the 29 Mb window on SSC7, which was identified in all GWAS, was associated with S:P for both NVSL and KS06 infected animals (P < 0.0001). In NVSL-infected pigs, animals with the AA genotype at ALGA had the highest S:P followed by AB animals. In KS06-infected pigs, however, a significant difference in S:P was not observed between animals with the AA and AB genotypes for ALGA but both were significantly different from BB individuals.

Table 3.

Least square means of antibody response at 42 d post infection with one of two isolates of the PRRS virus for three significantly associated genetic markers on chromosome 7

Marker (Chr; Pos)	Isolate	Genotype
		AA	AB	BB	Minor-allele frequency (allele)
DIAS (SSC7; 26 Mb)	NVSL	0.88 (0.04)a	0.77 (0.04)b	0.85 (0.06)ab	0.20 (B)
	KS06	0.89 (0.04)a	0.77 (0.05)b	0.69 (0.14)ab	0.07 (B)
MARC (SSC7; 29 Mb)	NVSL	0.93 (0.04)a	0.82 (0.04)b	0.78 (0.05)b	0.46 (A)
	KS06	0.86 (0.06)a	0.87 (0.04)a	0.86 (0.06)a	0.26 (A)
ALGA (SSC7; 29 Mb)	NVSL	0.93 (0.03)a	0.86 (0.03)b	0.77 (0.03)c	0.49 (B)
	KS06	1.00 (0.04)a	0.95 (0.03)a	0.78 (0.03)b	0.40 (B)

Estimates with different letters within a row are significantly different from each other at P <0.05.

Daily viremia and 3-d WG were also evaluated for an association with the three SNPs identified in the GWAS for S:P ratio in order to assess whether the observed genetic correlations of S:P with viremia or WG were due to the differences at the MHC (Supplementary Appendices 1–3). However, estimates of the effects of these SNPs were small for all daily viremia and 3-d WG traits and only significantly different from zero (P < 0.05) at a few time points (Supplementary Appendices 1 and 3).

DISCUSSION

Antibody Response to PRRSV Infection

A major challenge with containing PRRS is the level of genetic and antigenic variability observed between PRRSV isolates (Darwich et al., 2010). Antibodies produced by the host can be categorized into nonneutralizing antibodies, which bind to the virus in order to tag it for virolysis or viral clearance via phagocytosis (Burton, 2002), and neutralizing antibodies, which directly inhibits the function of the virus in some manner (Lopez and Osorio, 2004). Following PRRSV infection, the earliest and strongest antibody response is a nonneutralizing IgG response against the virus nucleocapsid (N) protein (Chand et al., 2012), which was the antibody response measured in this study. Virus neutralizing antibodies are typically weak and delayed during PRRSV infection. Once an antibody is produced that successfully neutralizes the virus, this response is very effective (Lopez and Osorio, 2004), but generally only to homologous and not heterologous isolates of PRRSV. Thus, a previous natural infection or vaccination with a modified live virus (MLV) generally results in only homologous protection against closely related PRRSV isolates (Huang and Meng, 2010). Understanding the host’s ability to mount a nonneutralizing PRRSV-specific antibody response, such as PRRSV N-protein specific IgG, may afford a better understanding of host response to PRRSV infection that is not specific to the PRRSV isolate, and may also be correlated with subsequent neutralizing antibody response.

In this study, antibody response was measured using FMIA, a protein-detection assay that utilizes fluorescent microspheres as the binding surface for antigen-antibody complexes, rather than the standard commercialized technique for detection of antibodies against PRRSV, which is an ELISA. Commercial ELISA assays typically are designed around a 0.4 S:P ratio as a positive–negative cutoff and are, therefore, not designed to be a quantitative measure of antibody activity. FMIA has been shown to have higher specificity and sensitivity than ELISA (Langenhorst et al., 2012). Balancing sensitivity and specificity, Wang (2013) set the threshold for identifying PRRSV infected animals for FMIA at 0.2. Wang (2013) also showed that the FMIA assay was able to detect antibodies sooner after infection than ELISA, with 4% of animals tested showing antibodies at 4 dpi based on FMIA, compared to 0% when using the ELISA kit. Another advantage of FMIA over ELISA is that it allows for uniform detection of multiple antigens or antibodies simultaneously within a small volume of a single sample.

Serum antibody levels were only assayed at 42 dpi, the end of the experimental trial. It has been reported that antibodies against PRRSV N-protein peak after 28 dpi and remain high for several months after infection, with low levels of antibody still detected in some animals for up to a year after infection (Yoon et al., 1995; Loemba et al., 1996; Johnson et al., 2011). Thus, antibody at 42 dpi is expected to measure the maximum antibody level achieved. Further research that includes assessing earlier samples at, e.g., 14 and 28 dpi, would be useful to gain a better understanding of the genetic and biological processes that are involved with the kinetics of antibody response, such as the time after infection when antibodies start to be produced, the rate at which antibodies are produced, and when the antibody level plateaus. At 14 dpi, most animals are expected to have seroconverted (Cano et al., 2007) but this is still an early time point for antibody response. An intermediate time point (e.g., 28 dpi) between seroconversion and the predicted plateau at 42 dpi would allow assessment of the rate of antibody accumulation (e.g., area under the curve).

The FMIA assay used in this study also measured IgM (Wang, 2013), but they were either undetectable or low because the animals had already undergone class switching by 42 dpi (Di Noia, 2015). This is consistent with previous reports on the kinetics of IgM antibody response to PRRSV infection (Loemba et al., 1996; Venteo et al., 2012). Measurement of antibody levels at 14 dpi would also allow IgM antibody levels to be evaluated to confirm their earlier production. IgM levels are expected to be low for most pigs at 28 dpi (Venteo et al., 2012).

Comparison of NVSL and KS06 Isolates

NVSL and KS06 are two genetically distinct PRRSV isolates (Ladinig et al., 2015; Trible et al., 2015). KS06 was isolated from a farm following an abortion storm. Among PRRSV isolates, NVSL is considered highly virulent (Truong et al., 2004). Comparison of reproductive performance of pregnant gilts following infection with these two isolates showed that they had similar virulence (Ladinig et al., 2015). However, in our studies in growing pigs, KS06 resulted in lower viral burden and less stunted growth than NSVL (Hess et al., 2016). Johnson et al. (2004) found a strong positive correlation between virulence of the PRRSV and antibody response. However, S:P levels at 42 dpi were not significantly different between pigs that were infected with NVSL vs. KS06 in our study, perhaps because antibody was only measured at 42 dpi or because these two isolates are not different enough in virulence to observe a difference in immunogenicity.

Estimates of heritability of S:P at 42 dpi were 0.31 ± 0.09 for NVSL and 0.40 ± 0.10 for KS06. These are similar to the estimates of heritability of S:P in blood samples collected from a commercial multiplier sow herd approximately 46 d after it experienced a PRRS outbreak (Serao et al., 2014; Serao et al., 2016), which used the commercial ELISA kit.

To assess whether the host genetics influencing S:P is similar when infected with different isolates of PRRSV, the genetic correlation between NVSL S:P and KS06 S:P was estimated. Unfortunately, due to large SE, this analysis was inconclusive. Supporting a high-genetic correlation of S:P between NVSL and KS06, a region of the genome was identified by GWAS that had an effect on S:P for both isolates, and explained 10% of the genetic variance for NVSL and 43% for KS06. It is conceivable that a substantial portion of the polygenic variance is also common to both isolates, as was identified for VL and WG during PRRSV infection (Hess et al., 2016; Waide et al., 2016).

Antibody Response in Cleared, Rebound, and Persistently Infected Pigs

Populations of PRRSV within an individual include subpopulations termed quasispecies, which are characterized by subtle variations from the consensus sequence, and these may be responsible for PRRSV escape from immune response (Rowland et al., 1999; Costers et al., 2010; Evans et al. 2017). Rebound in viremia is likely due to immune escape by the virus, which the pig recognizes as re-exposure to the virus. Rebound pigs had the highest S:P at 42 dpi. Therefore, pigs that have a rebound in viremia are expected to have greater stimulation of memory B cells, resulting in a higher level of antibody than pigs without a rebound (Kurosaki et al., 2015). It is also possible that pigs with higher antibody levels put more selective pressure on the virus and are, therefore, more likely to show rebound (Costers et al., 2010; Delisle et al., 2012; Trible et al., 2015). Whether rebound is the cause or effect of higher antibody levels at 42 dpi is unclear because antibody response was only measured at the end of the experimental period.

Animals that had persistent viremia had similar S:P at 42 dpi as pigs that had cleared the virus by 42 dpi, although the former were expected to have lower S:P due to a failure to mount a successful immune response. It may be that animals with persistent viremia had a delayed antibody response but a similar antibody plateau level at 42 dpi (Yoon et al., 1995; Loemba et al., 1996; Johnson et al., 2011). Measuring antibody response at multiple time points would help to elucidate these relationships between viremia status and antibody response.

Genetic Correlations of Antibody Response With WG and Viremia

Resource allocation theory relates traits competing for the limited energy resources that an animal has to the trade-offs that are often observed between traits (Van Noordwijk and Dejong, 1986; Roff, 2007) and can explain negative relationships between traits. Genetic correlations between traits are commonly used to assess trade-offs between traits at the genetic level (Roff, 2007). Studies investigating growth and immune function have observed trade-offs between these traits through negative genetic relationships (van der Most et al., 2011). Selection for production traits has resulted in reduced immunological fitness in multiple species of livestock (Rauw et al., 1998; van der Most et al., 2011). However, selection for improved immune response has been shown to have less of a negative impact on growth, perhaps because immune response is less costly than growth (Coop and Kyriazakis, 1999; Wilkie and Mallard, 1999; van der Most et al., 2011) or because healthy animals do not divert as many resources to immune response (Rauw, 2012).

In this study, weak genetic correlations of S:P at 42 dpi with VL and WG42 were observed. VL and WG42 are both summary measures of the cumulative effect of infection over time. The relationship between immune response and growth is, however, expected to change over time, as the animal places different emphasis on these processes throughout the course of infection (Coop and Kyriazakis, 1999; Doeschl-Wilson et al., 2009; Hess et al., 2016). Therefore, the low genetic correlations of S:P with VL and/or WG42 could be a consequence of the averaging of genetic correlations between traits at different time points.

Expression of an animal’s genetic merit for growth and disease resistance depends on the availability of nutritional resources (Doeschl-Wilson et al., 2009; Rauw, 2012). Infection often results in the pig eating less, which results in a competition for resources between growth and immune response within the animal. If resources are preferentially allocated toward growth, the animal may maintain growth but not recover from infection and consequently have lower growth later in infection. If resources are preferentially allocated toward immune response, the animal may sacrifice growth, but if it results in clearing the infection more quickly, then the animal will return to homeostasis sooner and growth will benefit in the long term, as demonstrated in a simulation study by Doeschl-Wilson et al. (2009).

Our results show that the genetic correlations of S:P at 42 dpi with 3-d WG and viremia levels indeed changed substantially over time. The strongest negative genetic correlation estimates between antibody response and daily viremia were observed when IgG antibody production is expected to start (10–20 dpi) (Chand et al., 2012; Venteo et al., 2012), which suggests that animals that placed more emphasis on immune response early in infection reaped the benefits of this later in infection by more effectively clearing the virus. This was supported by S:P at 42 dpi having negative genetic correlation estimates with WG early after infection but positive later on. These results are consistent with the simulation study by Doeschl-Wilson et al. (2009).

In NVSL-infected animals, the genetic correlation of S:P at 42 dpi with WG shifted from negative to positive shortly after 25 dpi, which corresponds to the time when pigs started to clear NVSL virus from blood. Thus, the transition from negative to positive genetic correlations of antibody response with WG in NVSL-infected pigs may be due to the pig recovering from infection. During infection, animals experience cachexia, presumably to promote an effective immune response (Exton, 1997; Kyriazakis et al., 1998; Langhans, 2000). Once the animal has produced an effective immune response, the animal will either return to a normal feed intake level, resulting in a return to homeostasis (Kyriazakis et al., 1998), or increase feed intake, resulting in compensatory growth (Diaz et al., 2005).

In KS06 infected pigs, the genetic correlation of S:P at 42 dpi with WG was weak and negative early in infection but rapidly changed to positive at 7 dpi. The genetic correlation of S:P at 42 dpi with daily viremia was negative at all time points for animals infected with KS06, whereas NVSL-infected pigs showed a strong positive genetic correlation between S:P and viremia early in infection, which switched to a negative correlation at ~7 dpi, and was most negative between 14 and 21 dpi, when production of IgG is expected to start. Previous analyses of these viremia and weight data showed that WG were greater for animals infected with KS06 than for pigs infected with NVSL (Hess et al., 2016). Thus, differences in genetic correlations of S:P at 42 dpi with either WG or viremia when comparing NVSL and KS06 infections may be due to KS06 being less virulent and, thus, requiring less energy to be diverted to combat the infection, which could be reflected in the weaker genetic correlations of S:P with early WG traits for KS06.

Impact of the SSC4 QTL on the Relationship of S:P With Viremia and WG

The WUR10000125 SNP did not have a direct effect on the level of antibody produced in NVSL- or KS06-infected animals. The genotype at this SNP did, however, appear to have an impact on the genetic correlation of S:P with both viremia and WG under infection, at least for pigs infected with NVSL; this was not tested for KS06 due to the limited number of animals. Compared to all NVSL-infected animals, genetic correlations of S:P at 42 dpi with viremia were more negative for pigs with the AA genotype for WUR compared to AB/BB pigs, and genetic correlations with WG were more persistently negative over time for AA pigs. While genetic correlations of S:P with viremia and WG could not be estimated directly for AB/BB animals, estimates of genetic correlations for these animals derived by comparing all animals to AA animals, revealed that S:P of AB/BB animals generally had high positive genetic correlations with WG and lower genetic correlations with viremia, compared to their AA counterparts.

The GBP5 gene, for which WUR is a tag SNP, is involved in formation of the inflammasome and pigs with the AA WUR genotype are predicted not to produce functional GBP5 protein (Koltes et al., 2015). The GBP5 protein binds to NLRP3 to promote ASC protein oligomerization and inflammasome assembly to activate caspase-1 (Shenoy et al., 2012). This pathway has been shown to induce an increase in IgG response (Kumar et al., 2009; Ellebedy et al., 2013). Gene expression profiling studies have shown that AA animals maintain higher expression of immune-related genes longer after infection than their AB counterparts (Schroyen et al., 2015). Furthermore, genes that were differentially expressed between 4 and 0 dpi showed greater enrichment of genes associated with humoral immune response for AA vs. AB pigs (Schroyen et al., 2015). This suggests that animals with the AA genotype require more energy to produce an effective antibody response than AB animals (either earlier, greater quantity, or both). An effective antibody response reduces viremia over time but the production of these antibodies may also come at a cost of lower WG. Lee (2006) argued that patterns in constitutive immunity among individuals depend on the cost of the defense and that individuals at the nonspecific inflammatory end of the innate immune response spectrum need to maintain higher levels of constitutive protection in order to minimize the use of costly inflammatory responses. Therefore, AA animals may also rely more heavily on alternate, likely nonconstitutive, inflammatory responses than AB or BB animals, resulting in a greater energetic cost for fighting PRRSV infection. Thus, the higher VL and lower WG observed in AA animals compared to AB animals (Boddicker et al., 2014b; Hess et al., 2016) may be due to the greater time or energetic resources AA animals require to mount an effective alternative inflammatory response. Gene expression studies have shown pigs with different genotypes at WUR differ in gene expression pathways (Schroyen et al., 2016).

Further analyses are needed for understanding the differences in response between the animals with different genotypes at this QTL. Body temperature and feed intake could be included as additional traits to explore response to PRRSV infection, which may provide insight into the early innate immune response to PRRSV infection of AA vs. AB animals. Also, cytokines could be investigated as other blood parameters involved in response to infection, while cytokine levels at day 0 might serve as useful predictors of response to infection.

Genes Affecting Antibody Response

The MHC, known as the Swine Leukocyte Antigen Complex in pigs, plays a crucial role in the host’s ability to mount an antibody response to infection (Lunney et al., 2009). The MHC is a gene-rich region that contains a cluster of genes associated with the immune system. One of the better characterized functions of the gene products of the MHC is the processing and presentation of antigens to T cells but many other immune-related genes and genes with unknown functions are also located in this region (Janeway et al., 2001). Between-breed differences in antibody response to PRRSV infection have been identified (Halbur et al., 1996; Petry et al., 2007). Wimmers et al. (2009) found an association of variants at the DQB MHC class II gene with antibody response at 10 d after vaccination with a modified live PRRS vaccine in an F2-population of reciprocal Berlin Miniature and Duroc crossbred piglets. Two previous studies (Serao et al., 2014, 2016) identified two large QTL on SSC7 for S:P measured by ELISA in a PRRSV outbreak sow herd. One of these QTL was in the MHC class II region and explained 25% of the genetic variation in antibody response.

As expected, the GWAS in our study also identified the MHC as being associated with antibody response at 42 dpi, using either the NVSL, KS06, or combined datasets. The 1 Mb window at 29 Mb on SSC7 explained 10.3% of the genetic variance for NVLS, 43.1% for KS06, and 29.9% for the combined analyses. The SNP DIAS in the 26 Mb window was identified as being associated with S:P in the joint analysis (explaining <1% of genetic variance for NVSL and KS06 and 13.8% for the joint analysis). DIAS may not have been identified in the NVSL GWAS due to potential nonadditive effects, while it may not have been identified in the KS06 GWAS due to a low minor-allele frequency and a corresponding low number of BB animals, resulting in a greater level of shrinkage in the Bayes B analysis that was employed in the GWAS. Two of these SNPs, ALGA and MARC, are located in the MHC class II region of SSC7, while DIAS is located at the distal end of the MHC class I region of SSC7. The function of MHC I and II molecules is to bind peptide fragments from pathogens and present these on the surface of the cell for recognition by the appropriate T cells (Janeway et al., 2001). Once the antigen is presented on the surface with an MHC class II molecule, it is recognized by CD4+ T cells, which release cytokines that amplify the immune response, for example, by stimulating B-cell differentiation. This B-cell differentiation into plasma B cells is responsible for the production of antibodies (Kurosaki et al., 2015). Therefore, it is not surprising that the MHC class II region shows a strong association with S:P. One SNP in this region, MARC, was previously identified as being associated with S:P ratio in sows that experienced a PRRS outbreak (Serao et al., 2014; Serao et al., 2016). Interestingly, in our study, this SNP was only associated with S:P for pigs infected with NVSL. The location of DIAS at the MHC I/III juncture makes it difficult to provide a good candidate gene for this SNP. Overall, the linkage disequilibrium in this region was low and the linkage disequilibrium between ALGA, MARC, and DIAS was also low. This suggests that these SNPs likely mark independent QTL.

Despite the strong genetic correlations of S:P with both viremia and WG, the SNPs identified for S:P by the GWAS either did not have a significant association with viremia and WG or only a small effect, indicating that these QTL play a role in antibody response but not in viremia or WG. GWAS of VL and WG following infection with NVSL (Boddicker et al., 2012; Waide et al., 2016; Hess et al., 2016) and KS06 (Waide et al., 2016; Hess et al., 2016) have concluded that the majority of the genetic variance of these traits is polygenic, as was also found for antibody response in this study. Taken together, these results suggest that the majority of the genetic correlation of antibody response with VL and WG is explained by many QTL with small effects spread across the genome, rather than by the QTL located within the MHC. If this is the case, genomic selection for improved resistance to PRRSV may be more effective at improving multiple aspects of host response to PRRSV infection, than selection on a limited number of SNPs in the MHC region.

CONCLUSIONS

Antibody response, measured by PRRSV N-protein specific serum IgG levels at 42 dpi, was estimated to have moderate–high heritability for both the NVSL and KS06 PRRSV isolates. The genetic correlation of antibody response between isolates was also moderate to high, suggesting similar host genetic control, although this estimate had a large SE. The results of this study are consistent with resource allocation theory, in that pigs that put more energy into fighting infection early on have higher antibody levels and lower viremia during later stages of infection. This is at a cost of early growth but showed a benefit in later growth, presumably as animals recovered from infection. Animals with the unfavorable AA genotype at the WUR locus on SSC4 showed a stronger negative correlation of antibody response with viremia and a more persistent negative genetic correlation with growth than AB/BB animals. This suggests that AA animals require more energy to fight infection, resulting in lower growth. Genetic correlations of antibody response with either viremia or WG likely also depend on the virulence of the PRRSV isolate.

Viremia measurements early after PRRSV infection in an outbreak are typically not available. However, our results show that S:P measured at 42 dpi has a moderate–high genetic correlation with PRRSV viremia and therefore may be an appropriate indicator of host response in an outbreak herd after the virus may have already cleared from serum. Antibody levels can be obtained from commercially available kits designed for diagnostic purposes (e.g., the IDEXX PRRS X3 Ab Test). However, the genetic control of host antibody response to infection must be assessed at more time points in order to explore how antibody response changes over time and how this correlates with changes in other traits over time. GWAS identified three SNPs associated with antibody response in the MHC region, but these SNPs were not significantly associated with viremia or WG.

Response to vaccination may provide an attractive alternative to natural infection for identifying pigs with increased resistance to PRRS, because all animals can be vaccinated at the same dose/age/time and, thus, response phenotypes (e.g., viremia, WG, S:P, etc.) can be collected with higher consistency. Measuring PRRS resistance based on response to vaccination may have the added benefit of reducing the circulating level of PRRSV during natural infection, as the vaccine provides partial protection to subsequent PRRSV exposure (Loving et al., 2015). Further studies are, however, needed to estimate the genetic correlation between response following vaccination and response following natural PRRSV infection. It is also important to assess the desirable, or optimal, level of antibody response that would increase resistance or decrease susceptibility to PRRSV, while not negatively impacting the ability of the pig to combat other infections.

SUPPLEMENTARY DATA

Supplementary data are available at Journal of Animal Science online.