Genetic Markers Associated with Field PRRSV-Induced Abortion Rates

In gilts and sows, the more severe clinical manifestation of porcine reproductive and respiratory syndrome virus (PRRSV) occurs in late gestation and can result in up to a 40% abortion incidence. Despite the known genetic component in resilience to PRRSV, there is scarce information regarding the abortive outcome of this disease. We tested the relationship between eight molecular markers (six from published studies and two identified in the present study in the HDAC6 gene) and the probability of abortion during a PRRSV outbreak, using data from two commercial Landrace x Large White sow farms with an incidence of abortion of 35% and 17%. From the markers tested, USP18_-1533G>A did not segregate in these populations, and CD163_c.3534C>T and HDAC6_g.2360C>T did not affect the abortion rate. In contrast, the minor allele of two markers in SSC4 (WUR1000125 in GBP1 and rs340943904 in GBP5), which lower viremia in growing pigs, and the major alleles of CD163_rs1107556229 and HDAC6_rs325981825 were associated with a lower probability of abortion during PRRSV outbreaks. The more striking result was for the MX1 gene, where the odds ratio of aborting versus not aborting was nine times lower in the sows homozygous for a 275-bp insertion than in the other genotypes. Interactions between markers were not relevant. All together, we bring here the first evidence that mutations in the host genome can predispose or protect from complete reproductive failure in sows infected with PRRSV.


Introduction
Pigs of all ages are susceptible to porcine reproductive and respiratory syndrome virus (PRRSV). The clinical presentation of PRRS varies greatly between herds and can range from asymptomatic to devastating disease [1]. In gilts and sows, PRRSV infection can cause reproductive failure particularly in early and late gestation, when the virus has the ability to cross the placental wall and infect the embryos. Pregnancy in pigs lasts for 114-116 days. Embryos dead prior to implantation are generally resorbed, and sows return to estrus. Embryos are also probably resorbed when death occurs between 14 and 35 days of gestation, causing irregular return to estrus if all embryos die or small litters if some of the embryos survive [2]. The more severe clinical manifestation of PRRSV occurs in late gestation and is characterized by abortions (up to a 40% abortion rate in late-pregnancy sows); early farrowings; fetal death; and the birth of weak, congenitally infected piglets, resulting in elevated preweaning mortality [3]. The formation of mummies, which is characteristic of PRRSV late-term infection, occurs by exsiccation of a dead fetus. Besides reproductive failure, clinical signs in pregnant sows and gilts are often mild or absent.
The mechanisms of transplacental infection and why virus transmission is blocked midpregnancy are still unclear. Pigs develop an incomplete diffuse epitheliochorial placenta without invasion. Neither

Pig Populations
Two populations of Landrace x Large White sows in commercial nursery farms in northeastern Spain were used for this analysis. Population 1 belonged to a 1500-PRRSV-naïve sow farm naturally infected with a field strain of the virus. The infection lasted for six weeks and resulted in an incidence of abortion close to 35% in two consecutive batches in the last third of the gestation period. Samples of jugular blood were collected from a subset of 180 sows (60 aborted and 120 not-aborted). No more data were available from this population. Population 2 was from a 1000-PRRSV-positive stable sow farm (following the classification proposed by Holtkamp in 2011 [17]) that underwent a sudden epidemic outbreak with an incidence of abortion close to 17% in the last third of the gestation period. Twelve sows aborted in the last third of the gestation period in this population, which were sampled along with 59 nonaborted females of the same batch of production to get a total of 71 samples. PRRSV genotype 1 (PRRSV-1), subtype 1, was involved in both clinical cases. The diagnosis was carried out by a laboratory specializing in pig diseases (http://www.gsplleida.net/ca) using standard operational procedures [18,19]. The homology between both viruses was probed to be quite low (89%), confirming that a different PRRSV strain was involved in each population [20].

DNA Isolation
Genomic DNA was isolated from blood by proteinase K/SDS lysis buffer, followed by phenol/ chloroform extraction and isopropanol precipitation using standard techniques [21]. The concentration and purity of DNA was assessed by spectrometry using a Nanodrop-100 (Fisher Scientific, Madrid, Spain) and was standardized at 10 ng/µL per sample.

Primer Design and Genotyping Protocol
Six markers from published studies and two additional makers identified in the present study in the HDAC6 gene were selected to be genotyped in the resource populations. Seven of these markers were single-nucleotide polymorphisms (SNPs). Six of them were genotyped using qPCR-HRM (high-resolution melting) technology (Supplementary Table S1) as follows. Sequences of the corresponding genes were exported from the latest version of the pig genome in Ensembl (Sscrofa 11.1) and were used to design PCR primers with the Primer3plus tool (https://primer3plus.com/cgi-bin/dev/primer3plus.cgi; [22]) using standard qPCR parameters. PCR reactions were set up in a volume of 7 µL containing 1× Luminaris HRM Master Mix (Fisher Scientific), 0.2 µM of each primer, and 10-15 ng/DNA. Reactions were run in a QuantStudio 3 thermocycler (Fisher Scientific) using the following program: initial denaturation at 95 • C for 10 min, 40 cycles of 95 • C for 10 s and 60 • C for 1 min, followed by a slow ramp from 60 • C to 95 • C at a rate of 0.015 • C/s. Genotypes were assigned by comparing PCR melting curves using the HRM software (Fisher Scientific). The GBP1 SNP marker (rs80800372 or WUR1000125) was genotyped by allelic discrimination using allele-specific TaqMan probes labeled with FAM and VIC (Supplementary Table S1). For this, PCR reactions (7 µL) were set up with 1× TaqMan Universal PCR Master Mix (Fisher Scientific), a 1× Primer and Probe set, and 10-15 ng/DNA. Reactions were run in a QuantStudio 3 thermocycler using the following program: initial fluorescence detection at 60 • C for 1 min, denaturation at 95 • C for 10 min, 40 cycles of 95 • C for 10 s and 60 • C for 1 min, and final fluorescence detection at 60 • C for 1 min. Genotypes were assigned by comparing the accumulation of FAM and VIC fluorescence in each channel.
The eighth marker was a 275-bp insertion in the promoter of the MX1 gene, which was genotyped with an end-point PCR and electrophoresis in an agarose gel. A PCR reaction was set up in a volume of 15 µL containing 1× buffer, 0.2 mM dNTPs, 2 mM MgCl 2 , 0.4 µM of each primer, and 0.4 U of Taq polymerase (Bioline, London, UK). The primers are summarized in Supplementary Table S1.

HDAC6 Characterization and Genotyping
The HDAC6 gene sequence was exported from Ensembl as described above and was used for primer design with Primer3Plus using standard PCR parameters and with the desired PCR product size set in the 600-1200-bp range. Primer sets were designed to amplify five fragments of the gene that included most of the exons (Supplementary Table S2). This gene (~25 Kb) has 31 exons, most of them of a small size, that are spliced into five transcript variants leading to the production of two proteins differing in the last 50 amino acids. The PCR fragments were selected to include as much of the coding region as possible.
To describe new mutations in this gene, the five fragments were amplified and sequenced in 16 sows, 8

Statistical Analysis
The "abortion during PRRSV outbreak" outcome (aborted vs not-aborted) was analyzed using a binary logistic regression model on each genotype and population. The probability of aborting during a PRRSV outbreak was calculated as follows: where P(y = j) is the probability of occurrence of the abortion condition j in a specified situation i; α is a constant; and b' and x i are, respectively, the vector of coefficients in the logistic equation and the vector of independent variables in situation i. The genotypes for each individual investigated gene and the farm were included as independent variables. To test for potential associations and interactions between genotypes, the same model considering two genotypes at a time (all pairs of genotypes were tested) and their interactions was also used. The effects of each genotype and the corresponding odds ratios were tested using a χ 2 approximation to the asymptotic distribution of the likelihood ratio test. Analyses were performed with JMP Pro14 (SAS Institute Inc., Cary, NC, USA) software.

Characterization of Sequence Variants in Pig HDAC6
The five fragments of the HDAC6 gene were successfully sequenced in two groups of sows, which either aborted or not during a PRRSV outbreak. A sequence comparison identified 10 SNPs, 6 of them in the coding sequence ( Table 2). Three of these polymorphisms changed the sequence of the encoded protein at the amino acid positions 12, 360, and 503. Two protein intolerance prediction tools catalogued these mutations as potentially damaging for the functionality of the protein (Table 3), particularly the last two, as they were in the catalytic domain of the protein and in very conserved residues when compared to other species. Genotyping protocols were set up for the three mutations using an HRM-qPCR approximation. However, the Pro503His could not be genotyped with enough quality and was discarded from future analysis.
The position of the polymorphisms was calculated over the genomic DNA sequence taking the position of the start codon as a reference.

Allelic and Genotypic Frequencies of the Markers
The eight polymorphisms tested (Table 1) segregated in the two populations used for this study except for the USP18 -1533G>A marker, which was fixed at the G allele. This mutation, located in the promoter of the USP18 gene (ubiquitin specific protease 18), enhances the expression of this gene, probably by altering a binding site of the FOX transcription factor family [16]. In vitro USP18 overexpression can stop PRRSV infection in MARC-145 cells [14]. The fixed G allele agreed with data from Li et al. [23], who observed segregation of the A allele in three Asian pig breeds but not in Landrace, Duroc, or Yorkshire breeds.
Genotypic information from the two farms was pooled to calculate allelic frequencies. Overall, the minor allele frequencies were low or moderate, ranging from 0.19 for the GBP1 mutation to 0.38 for the g.2360C>T SNP in HDAC6. The distribution of genotypes is indicated in Table 4. The linkage disequilibrium between markers in the same chromosome was high for GBP1-GBP5 (D' = 0.89) and moderate and low for the two polymorphisms at HDAC6 (D' = 0.62) and CD163 (D' = 0.32).

Association Study between Markers and Abortion Rate
During the PRRSV outbreak, the probability of abortion was significantly different in homozygous sows for the minor alleles of GBP1, GBP5, CD163 (rs1107556229), MX1, and HDAC6 (rs325981825) ( Figure 1 and Supplementary Table S3; p < 0.05). For the GBP1 marker, the odds ratio for abortion versus not was 2.69 times higher in AA sows than in AG sows (Table 5). In the GBP5 marker, this odds ratio was 2.76 to 4.49 times higher in the GG genotype than in the TT and TG genotypes (Table 5; p < 0.05). Similarly, for the HDAC6_rs325981825 genotype, the GG sows had an odds ratio 2-2.5 times lower than in AA and AG sows (p < 0.05). A more extreme situation happened in the MX1 markers, where the homozygous animals with the minor allele showed an odds ratio for abortion versus not of 9 times less than the other two genotypes (p < 0.05). For the CD163_rs1107556229 marker, the minor allele (GG genotype) was associated with an odds ratio 1.96 to 3.6 times higher than in the AG and AA genotypes. Abortion rates did not differ between the CD163_c.3534C>T and HDAC6_g.2360C>T genotypes. On the whole, minor alleles for GBP1, GBP5, and MX1 markers and major alleles for CD163_rs1107556229 and HDAC6_rs325981825 had a protective role against total reproductive failure in sows exposed to a sudden outbreak of this disease. Interestingly, the effect of each genotype did not substantially change when they were analyzed in groups of two, with the exception of GBP1, whose impact on the abortion rate vanished when GBP5 was also included in the model. This reinforces the role of GBP5 in resilience to PRRSV compared to GBP1, whose effect should be explained by the strength of its linkage disequilibrium with GBP5. No significant interactions were observed besides that between GBP1_WUR10000125 and HDAC6_rs325981825 (p < 0.03), which, however, was not detected when GBP1 was replaced with GBP5. Therefore, interactions between markers did not play a relevant role in predicting the likelihood of abortion during a PRRSV outbreak. Figure 1. Abortion probability during a PRRSV outbreak by marker genotype (data pooled from the two studied populations). Within each marker, the genotypes with different superscripts indicate differences in the abortion ratio (p < 0.05).

Discussion
In previous studies led by the PRRS Host Genomic Consortium (PHGC), large sample sizes of the pregnant gilt model enabled an exceptional opportunity to identify new genomic regions associated with reproductive PRRS outcomes [3,8,[24][25][26]. Twenty-one candidate genomic regions across 10 chromosomes were found to be significantly associated with fetal viability, fetal death, and viral load in the fetal thymus. Together, these regions accounted for a biologically relevant portion of the overall genetic variation. Seven of these overlapped with previously reported quantitative trait loci (QTLs) for pig health and reproduction [27]. However, it was not feasible to describe any marker associated with abortion because the abortion rate in this experimental model was unexpectedly low. Recently, Scanlan and coworkers [28] estimated the heritability of abortion rates, which was low (0.07) in healthy farms but rose to 0.17 during PRRSV outbreaks. In this context, our research work provides the first scientific evidence that the genetic background of the sow could be linked with the probability of abortion after a PRRSV infection.
The first two molecular markers tested were rs80800372 (i.e., WUR10000125) in GBP1 and rs340943904 in GBP5 (Table 1). WUR10000125 was the first molecular marker associated with PRRSV resistance and productivity. A genome-wide association study identified a genomic region in SSC4, represented by WUR10000125, which explained 15.7% of the genetic variance in viral load and 11.2% of weight gain in weaned piglets directly challenged with a very pathogenic American (PRRSV-2) strain [7]. The estimated effects for this region were favorably and nearly perfectly correlated: That is, pigs with a low virus load exhibited greater weight gain. The WUR10000125 mutation is a single-nucleotide polymorphism (SNP) located in the 3'UTR region of the GBP1 gene, a response gene to type II interferons. The G allele modifies a close polyadenylation signal in the GBP1 gene, reversing the proportion of two alternative transcripts [29]. The clinical and productive outcomes of this marker have been validated in experimental [7,30] or natural challenges [8,18] with PRRSV-1 [18] and -2 in growing animals [7,10,30,31]. However, the causality of this region of chromosome 4 is currently attributed to a nearby gene, GBP5, where an intronic SNP (rs340943904) (also included in our study) generates a new splicing acceptor site, changing the proportion of two alternative transcripts [9]. This gene encodes for a protein involved in the inflammasome assembly during innate immune responses [7,9].
The effect of the WUR10000125 genotype on reproductive traits is controversial. An initial study found that sows carrying the favorable G allele had more piglets born alive and weaned than homozygous AA sows in uninfected farms [12]. However, this result was not replicated in a subsequent experimental infection using a PRRSV-2 strain. The WUR10000125 genotype in both gilts and fetuses was associated neither with fetal death/viral load [32] nor with reproductive performance during a PRRSV outbreak in a commercial multiplier sow herd [11]. However, we describe here that, as for the respiratory form of the disease, the G allele was outstanding as a protective factor against abortion ( Figure 1). The difference in abortion rate probability was more evident in the closely linked GBP5 marker. In this case, the G allele showed an additive pattern, as described previously for WUR10000125 in the respiratory form of the disease [7,18]. Thus, the genotype GG was more susceptible to abortion (probability of abortion 30%) than GT and TT sows (the less represented genotype), where the probability of abortion was 13% and 3%, respectively. Our results complement previous studies on the effect of the SSC4 region on reproductive traits, where abortion rates were not included. It must be highlighted that, by studying the abortion rate and not the reproductive performance of nonaborted sows, we probably discriminated the sow population most susceptible to the virus.
Our study also included two markers in the CD163 gene (Table 1), which encodes the membrane receptor used by the virus to enter macrophages and initiate infection [1]. Several natural mutations have been described in this gene, some of which have been related to a better response capacity of pigs to the virus. For example, the CC genotype of the CD163_c.3534C>T polymorphism has been associated with lower levels of IgG [33] and viremia and enhanced weight gain [11] after a PRRSV challenge. The second polymorphism selected in this gene, c.2494G>A (rs1107556229), is located at the end of exon 10 and is a synonymous variant catalogued as a splice region variant by the variant effect predictor (VEP) tool. Our results clearly show that the different variants of CD163_rs1107556229 were linked with the probability of abortion in infected sows (Table 5). Scavenger receptor CD163 is a key entry mediator for PRRSV, and complete [34] or extracellular domain [35] deletion by genetic editing of this receptor makes pigs fully resistant to PRRSV infection. Thus, mutations in CD163 could imply a significant change in the susceptibility of the pigs to PRRSV infection. Our results totally agree with this information in the sense that sows with the rs1107556229 variant, naturally present in nature, had a different clinical outcome when abortion was the end-point measure (Table 5 and Figure 1).
The next gene marker tested was a 275-bp insertion in the proximal promoter of the MX1 gene (MX Dynamin Like GTPase 1), which enhances the transcriptional activity of this gene, mediating the humoral response to infection with PRRSV [13]. As with GBP1 and GBP5, MX1 encodes a guanosine triphosphate (GTP)-metabolizing protein, whose expression is induced by interferons. MX1 has strong antiviral activity against a wide range of RNA viruses and some DNA viruses through binding and inactivation of their ribonucleocapsid. The antiviral effect of MX1 against the influenza virus has been well documented. Mice carrying the defective MX1 gene show greater susceptibility to influenza A infection compared to mice carrying the functional protein [36]. MX1 also conferred protection against the influenza A virus, classical swine fever virus, and vesicular stomatitis virus in studies performed in vitro [37]. The mutation studied here was associated with higher MX1 expression in vitro [13]. In our data, this mutation showed the strongest association with a low abortion rate ( Table 5). The effect was completely recessive, where sows with two copies of the insertion had an abortion probability of 3% during the outbreak, while in the other sows the probability was~25% ( Figure 1). As with the favorable alleles of GBP1 and GBP5, the MAF of this allele was low (<0.20, Table 4), which raises the question of whether there was an indirect selection of these alleles representing three genes with very similar functionalities. Selection in nucleus farms is performed in clean environments with a lower pathogen load than in standard field conditions. This finding supports also studying the possible effect of these markers in healthy production farms and in other production traits.
The last gene studied was HDAC6. This histone deacetylase gene was selected for its role in macrophage differentiation. Initially, histone deacetylases were characterized as enzymes that removed acetyl groups from histones, establishing a silent chromatin structure. However, HDACs have recently been shown to act over a wider spectrum of substrate proteins, involved in a range of cellular processes that extend beyond epigenetic labels. Histone deacetylases are critical regulators in macrophage differentiation and in maintaining M1 or M2 macrophage functions in Th1 and Th2 T-cell responses, respectively. The balance between the contrasting cytokine profiles of M1 and M2 macrophages regulates many immune checkpoint modulators [38,39]. HDAC6 overexpression in cells and in transgenic mice enhances resistance to viral infection with the human acquired immunodeficiency virus (HIV-1), influenza A virus, and vesicular stomatitis virus [40]. Recently, HDAC6 overexpression in transgenic pigs reduced the viral load of PRRSV-challenged animals and had extended survival time and fewer clinical signs than wild-type controls [15]. We identified three missense mutations in the coding region of this gene (Table 2) with potential damaging effects on the function of the protein, two of them directly affecting the deacetylase domain of the protein ( Table 3). The two markers analyzed had intermediate allelic frequencies, with MAF close to 0.40 ( Table 4). The odds ratios of aborting versus not aborting during an outbreak were 2.3 to 4 times greater in sows of the rs325981825 GG genotype than in the other genotypes ( Table 5). The probability of abortion followed an additive pattern, similar to GBP5_rs340943904, where the presence of the A allele raised the probability of abortion from 16% in GG sows to 27% and 40% in AG and AA animals, respectively (Figure 1). Given the intermediate frequencies, this marker responded better to selection than the other three favorable alleles did, which were present at a much lower frequency.

Conclusions
In conclusion, we bring here the first evidence that mutations in the host genome can predispose or protect from complete reproductive failure in sows infected with a field strain of PRRSV. Although transcriptomic and genomic scans have studied the genetic component of spontaneous abortions in pigs [41] and cattle [42], abortion as an outcome of an infectious disease has not been explored from a genetic point of view in pigs. The low or intermediate allelic frequency of some protective alleles and potential interactions between some of them will need to be tackled in future studies.
Supplementary Materials: The following are available online at http://www.mdpi.com/1999-4915/11/8/706/s1, Table S1: Primers and probes used in the genotyping experiment; Table S2: Primers used to sequence the HDAC6 promoter and coding regions; Table S3: Comparison of the probability of abortion during a PRRSV outbreak by marker genotype in the sow population studied. Funding: This research and the APC were partially funded by FEDER projects COMRDI16-1-0035-03 and RTI2018-097700-B-I00 from the Spanish Ministry of Science, Innovation, and Universities.