OAS1 Polymorphisms Are Associated with Susceptibility to West Nile Encephalitis in Horses

West Nile virus, first identified within the United States in 1999, has since spread across the continental states and infected birds, humans and domestic animals, resulting in numerous deaths. Previous studies in mice identified the Oas1b gene, a member of the OAS/RNASEL innate immune system, as a determining factor for resistance to West Nile virus (WNV) infection. A recent case-control association study described mutations of human OAS1 associated with clinical susceptibility to WNV infection. Similar studies in horses, a particularly susceptible species, have been lacking, in part, because of the difficulty in collecting populations sufficiently homogenous in their infection and disease states. The equine OAS gene cluster most closely resembles the human cluster, with single copies of OAS1, OAS3 and OAS2 in the same orientation. With naturally occurring susceptible and resistant sub-populations to lethal West Nile encephalitis, we undertook a case-control association study to investigate whether, similar to humans (OAS1) and mice (Oas1b), equine OAS1 plays a role in resistance to severe WNV infection. We identified naturally occurring single nucleotide mutations in equine (Equus caballus) OAS1 and RNASEL genes and, using Fisher's Exact test, we provide evidence that mutations in equine OAS1 contribute to host susceptibility. Virtually all of the associated OAS1 polymorphisms were located within the interferon-inducible promoter, suggesting that differences in OAS1 gene expression may determine the host's ability to resist clinical manifestations associated with WNV infection.


Introduction
The innate immune response confers host resistance by the recognition and limitation of viral infection and replication. Previous investigations of the innate immune response to West Nile virus (WNV) infection were conducted using inbred strains of naturally susceptible and resistant mice [1,2]. Using a positional cloning strategy, the Flavivirus resistance (Flv) gene was identified as 29,59-oligoadenylate synthetase 1b (Oas1b). The interferon (IFN)-induced OAS genes encode dsRNA-activated proteins which catalyze the synthesis of 29-59-linked oligoadenylate molecules (2-5A) from ATP [3][4][5]. The only known function of 2-5A molecules is to activate Ribonuclease L (RNASEL) for the degradation of cellular and viral RNA [3,[6][7][8].
OAS proteins are encoded by multiple genes, collectively referred to as the OAS gene cluster. Gene clusters vary between species in both number of genes and splice variants. The rodent Oas1 locus expanded to a family of 12 genes, while both the canine and bovine clusters contain duplications of OASL and OAS1 genes, respectively [9,10]. However, both equine and human OAS clusters contain single copies of each gene in the same orientation, OAS1-OAS3-OAS2 and a single OASL gene [9,11].
Previously, Yakub et al. found a SNP in human OASL associated with WNV susceptibility; however, this association was not replicated in a larger case-control study [12,13]. Case-control association studies are best implemented using homogenous populations, which reduces systematic bias from sample selection and minimizes the potential for false positive associations inherent within the population structures [14]. Lim et al. recently identified human OAS1 SNP rs10774671 in an association study of symptomatic and asymptomatic seroconverters [12]. In this study, case samples were compared to a control population of WNV false-positives. This SNP is located in an intron 5 splice site resulting in differential splicing and a protein product with diminished enzymatic activity. Taken together, data from both human and mouse studies support our investigation of equine candidate genes OAS1 and RNASEL.
Susceptibility to severe West Nile encephalitis among mammals is naturally variable [15]. Experimental infections in sheep [16], calves [17], pigs [18], and dogs [19] have shown these domestic species to be poor hosts for, or develop only mild clinical symptoms from WNV infection, thereby limiting their usefulness for genetic susceptibility/resistance studies. Horses however, are particularly susceptible to severe WNV infection, suffering clinical symptoms including fever, ataxia, paralysis and death [20]. Because many horses infected with WNV remain asymptomatic or present only mild symptoms, horses are an excellent model organism to test for genetic susceptibility using strictly phenotyped case and control populations. This advantage is complicated, however, by the inaccessibility of well-characterized case and control samples for retrospective study.
As mentioned above, one advantage of a horse model is the ability to monitor both infection in control samples and WNVinduced encephalitis in case samples. In this report, we describe a two-stage association study of naturally occurring equine OAS1 and RNASEL mutations to investigate a potential role of these genes in the equine innate immune response to WNV infection. Because of the limited retrospective accessibility to adequately phenotyped samples with known pre-and post-infection status, our population sizes did not allow matching of case and control samples by breed. Although no breed-specific difference in susceptibility has been reported, our analysis showed the associations identified from these populations were not artifacts of the two most frequent breeds in our case population. Because most significantly associated SNPs were present in the OAS1 promoter region, we conducted reporter assays to measure the response of equine OAS1 promoter constructs to interferon stimulation by transient transfection using human transformed cell lines.

Defining case and control population samples
The control population consisted of 16 healthy, previously uninfected (naïve) horses of multiple breeds, including Thoroughbred (13), Quarterhorse (1), Paso Fino (1) and a single mixed breed horse. These unvaccinated horses were naturally infected with WNV by mosquito transmission during the height of the initial Florida epidemic of 2001, when only the NY99 strain was present. Control horses were monitored daily, yet failed to exhibit clinical symptoms. These healthy individuals tested positive for WNV infection and were therefore classified as subclinical seroconverters. An important characteristic of our case-control study, all horses included in the control population had an equal opportunity of being classified in the case population, had they displayed clinical symptoms post-infection.
Horses included in the case population were previously unvaccinated and naturally infected through mosquito transmission. Multiple breeds were present among the 44 case horses, the most common of which included Thoroughbred (12) and Quarterhorse (10), with 2 horses of unknown breed. All case horses developed clinical encephalitic symptoms diagnosed with veterinary treatment, ultimately requiring humane euthanasia. Veterinary examination noted multiple symptoms, the most common including forelimb and/or hindlimb ataxia. Diagnostic tests confirmed WNV infection by enzyme-linked immunosorbent assay (ELISA) or polymerase chain reaction (PCR).
An important characteristic shared by both the case and control populations, horses were unvaccinated prior to infection and all horses were infected by natural mosquito transmission during the early stages of the initial U.S. epidemic, between 2001 and 2002.

SNP genotyping and association analysis
Previously, we sequenced the coding regions of equine OAS1 and RNASEL as well as the OAS1 promoter from a random population and identified SNPs for case-control association [21]. In this study, we genotyped 49 equine OAS1 and RNASEL mutations in 16 control and 44 case samples. Genotype data were analyzed to identify statistically significant allelic (262) and genotypic (263) associations to WNV phenotype using the conservative Fisher's Exact test. Fifteen SNPs in OAS1 and three SNPs in exon 2 of RNASEL were significantly associated (262, p,0.05) with WNV susceptibility ( Table 1). The statistical associations of OAS1 mutations were not uniformly scattered throughout the entire gene but were concentrated in the upstream regulatory region, with twelve of the significantly associated mutations located in the OAS1 promoter and 59 untranslated region (UTR). Using the highly conservative Bonferonni threshold for statistical significance, only mutations of the OAS1 promoter (6 of 49) were associated (262, p,0.001) with WNV susceptibility. Next, we used odds ratios (OR) to measure the strength of the SNP associations ( Table 1). All significantly associated mutations had ORs greater than 1.0. Among significantly associated mutations of the OAS1 regulatory region, all but one had 95% OR confidence intervals greater than 1.0.
Deviation from Hardy-Weinberg Equilibrium (HWE) can be a useful tool to indicate errors in genotyping or population stratification [22]. To investigate this possibility, exact tests were used to determine if genotype frequencies in the control population deviated from HWE. Two SNPs (ss104806941 and ss104806963) failed the HWE test (Table 2), but neither SNP was associated with WNV susceptibility. Tests of HWE may also be informative for association studies when measured in the case population. Within our case population, many SNPs deviated significantly (p,0.05) from HWE (Table 2), and SNPs with the greatest deviation (lowest p-value) were in the OAS1 promoter. Exact tests for HWE in the control and case populations supported the significant associations of 6 of the 10 OAS1 promoter mutations.
To investigate potential false-positive SNP associations from over-represented case breeds, Quarterhorse (n = 10) and Thoroughbred (n = 12) case samples, together representing 50% of the population, were independently compared to the control population. Seven of the 8 mutations statistically significant in both casebreed analyses were in the equine OAS1 promoter (Table 3). Using the two most represented breeds of the case population, these analyses allow us to conclude that the SNP associations to WNV susceptibility are not artifacts attributable to breed specific allele frequencies in the major breeds of our case study population.

Haplotype assembly and association analysis
Fifteen SNPs genotyped in the promoter, 59UTR and exon 1 of equine OAS1 were used to infer haplotypes among case and control samples. From the assembled best reconstruction, we identified six tagSNPs (ss104806918, ss104806922, ss104806924, ss104806926, ss104806927 and ss104806931) with calculated mean percentage diversity explained (PDE) of 99.23% [23]. These tagSNPs, all associated with WNV susceptibility, were used to reconstruct haplotypes and conduct case-control comparisons. Haplotype frequencies were significantly different (p,0.01) between case and control populations. A single common haplotype (GACCGT) was assembled in 65.6% and 23.3% of control and case sample chromosomes, respectively. Fisher's Exact test showed deviations from this haplotype were significantly associated (p = 4.953 e-6) with susceptibility to severe WNV disease, with an odds ratio of 7.58 (95% CI = 2.88: 21.18). Five of the six alleles in this haplotype were found to be protective in our study, consistent with the increased haplotype frequency in the control population. This haplotype data supports the OAS1 promoter SNP associations to WNV susceptibility.
Equine RNASEL haplotypes were inferred from 42 horses genotyped at $75% of all RNASEL SNPs in order to minimize the effect of unknown genotypes. Six tagSNPs (ss104806949, ss104806954, ss104806955, ss104806958, ss104806959 and ss104806965) were identified with total mean PDE of 99.33%.
Haplotypes were re-constructed using these tagSNPs from the same 42 samples and, in contrast to OAS1, haplotype frequencies were not found to differ significantly between case and control populations (p = 0.53).

Interferon stimulation of equine OAS1 promoter
Since many of the SNPs associated with WNV susceptibility were present in the OAS1 promoter and because human OAS1 is induced by IFN-stimulated regulatory factors acting through an IFN-stimulated response element (ISRE) proximal to the transcription start site (TSS, Figure S1) [24], we conducted preliminary transient transfection experiments to determine if these mutations could alter IFN induction of the equine OAS1 promoter.
Functional assays of the OAS1 promoter by transient transfection should be conducted in equine cell lines derived from tissues involved in the early development of post-infection WNV disease, but such cell lines are currently unavailable. We therefore substituted two cell lines, 2fTGH and HepG2, that have been extensively used in studies of IFN and/or OAS [25][26][27][28][29].
Haplotypes of the proximal promoter of equine OAS1 were cloned upstream of the luciferase reporter coding region (Figure 1). Proximal promoter constructs were generated as deletions of the full-length clones mentioned below. These deletion constructs (EcOAS1D59_A-Luc and EcOAS1D59_B-Luc) lack the polymorphic microsatellite and sequence further upstream. These proximal promoter constructs were used in transient transfection assays of 2fTGH cells (a derivative of HT1080 cells) treated with 10 4 antiviral units (AVU) of interferon (IFN). Luciferase reporter activity 24 h after stimulation was 7-to 8-fold higher than basal levels (data not shown). Therefore, the proximal region from the TSS to the microsatellite (,518 bp) was found to be necessary and sufficient for equine OAS1 promoter responsiveness to IFN. This is the first direct observation of equine OAS1 promoter IFN responsiveness.
Additional promoter sequence containing the polymorphic microsatellite and upstream sequence was cloned upstream of the luciferase reporter coding region (Figure 1). Full-length promoter clone EcOAS1_A-Luc contains the alleles of the common haplotype previously mentioned (GACCG) while EcOAS1_B-Luc and EcOAS1_C-Luc contain the TATTG and TTTTG haplotypes, respectively. Additionally, each full-length clone contains a previously identified polymorphic microsatellite with repeat lengths of 9 (A), 16 (B) and 19 (C) [21].
Full-length constructs were transfected into 2fTGH cells and treated with different doses (10 2 to 10 4 AVU/mL) of IFN for 24 h. IFN stimulated activity of each OAS construct in a dosedependent manner (Figure 2). The variation between experiments was greater for the EcOAS1_B-Luc construct; however, the average fold induction across three replicates for the EcOAS1_A-Luc construct was ,2-and ,4-fold greater than   the EcOAS1_C-Luc construct when cells were treated with 10 3 AVU and 10 4 AVU, respectively. The greatest differences in fold induction between clones occurred when cells were treated with 10,000 AVU IFN (ANOVA, p = 0.026); however, little difference was seen between constructs when cells were treated with only 100 AVU for 24 h. Because the difference in fold induction between constructs may be an artifact of the high IFN concentration used and the long exposure time, full-length constructs were transfected into HepG2 liver cells, which have been used in previous studies of OAS expression, and treated for 6 h or 24 h [29]. Similar to the 2fTGH experiment, equine OAS1 promoter induction in HepG2 cells showed a dose-response to IFN at both 6 h and 24 h (Figure 3). Averaged across 4 replicates, the fold induction was similar between clones both at 6 h and 24 h when cells were treated with 10 3 AVU IFN (ANOVA, p.0.05). However, when HepG2 cells were treated with 100 AVU IFN for 6 h, construct EcOAS1_ A-Luc was induced at higher fold levels than the other clones (ANOVA, p,0.001). OAS is part of the immediate early response of cells to viral infection, and previous studies of OAS promoter function have included lower IFN doses and shorter exposure periods to mimic early times after infection [28]. These preliminary data suggest a similar model, and when considered with the promoter SNP and haplotype associations, suggest promoter mutations affecting equine OAS1 expression may contribute, in part, to susceptibility to severe WNV disease.

Discussion
The first evidence for an involvement of the OAS gene family in innate resistance to West Nile virus was provided using a mouse model [1,2]. However, the rodent OAS cluster contains numerous copies of the Oas1 gene and the comprehensive cluster structure is largely different from the human gene cluster. Another study recently provided evidence of a role for human OAS1 in innate resistance to WNV infection [12]. With our ability to closely monitor response to West Nile virus infection, we used the horse as a model to identify a potential association of the equine OAS1 gene to WNV resistance or susceptibility through genetic comparisons of case (susceptible) and control (resistant) animals subjected to similar pre-exposure conditions. These horses were phenotyped for their innate resistance and susceptibility to natural WNV infection. Furthermore, the equine OAS gene cluster is more similar to the human OAS cluster than any other known cluster in domesticated mammals.
We report associations of SNPs in the equine OAS1 gene with susceptibility to West Nile encephalitis. Because of the limited number of well-characterized horses, Fisher's Exact tests were used to identify significant differences in allelic and genotypic frequencies between case and control populations. Ten of the 18 susceptibility-associated SNPs (p,0.05) were identified within the regulatory region of equine OAS1. While these data suggest a potential role for equine OAS1 in innate resistance to WNV disease, identifying causal mutations is complicated by the highly variable nature of this region within our case population and the possibility of as yet unidentified mutations in strong linkage disequilibrium with those reported here.
Although no difference in susceptibility to WNV infection between major equine breeds in the United States has been reported, we investigated the potential for false-positive associations resulting from the two most frequent breeds of our case population, Quarterhorse and Thoroughbred. Both breeds, together representing 50% of the case population, were individually compared to the control population which is almost entirely     composed of Thoroughbreds. Promoter polymorphism associations remained significant for each breed, indicating that the associations reported here are not artifacts based on skewing from the major case population breeds. While the multiple SNP associations of the OAS1 promoter with WNV susceptibility are likely due to linkage disequilibrium, our association data also suggest a potential functional mechanism by which OAS1 expression in response to infection might, in part, confer resistance to WNV. Five tagSNPs were identified and genotyped in each promoter construct, with clone EcOAS1_A-Luc representing 65.6% and 23.3% of control and case population haplotypes, respectively. Previous studies have shown that expression of the interferon-inducible murine Oas1b during the early stages of WNV infection (6-9 h post-infection) greatly reduced virus production compared to later timepoints [30]. Using transient transfection reporter assays, we investigated potential effects of the promoter mutations on interferon responsiveness. A dose-response was observed for both 2fTGH and HepG2 cells, two cell lines previously utilized in studies of interferon responsiveness and/or OAS [25][26][27][28][29]. Furthermore, when HepG2 cells were treated with 100 AVU IFN for 6 h, IFN responsiveness was greater in the clone with the common control-population haplotype compared to the others, whose promoter haplotypes were seen more frequently in the case population. Although the differences in fold induction are not dramatic, they are statistically significant and warrant future studies that may help determine how physiological levels of OAS1 expression affect the host response to WNV infection.
WNV infects several tissues during the early viremic phase prior to infection of the central nervous system. As a result, the polymorphisms in the OAS1 upstream region that are associated with susceptibility to WNV may have a greater functional relevance on OAS1 expression in other cell lines than those used here. Cell type has been shown to have a pronounced effect on the induction of the p69 isoform of OAS2 by IFNB [31]. Specifically, expression levels of p69 OAS were substantially higher in lymphoid Daudi cells than in human fibrosarcoma HT1080 cells, the parental cell line of the 2fTGH cells used in these studies [31,32]. Alternatively, moderate differences in OAS1 expression may be more readily detectable by quantitative measures of OAS1 mRNA in WNV-infected cells from horses having either susceptible or resistant genotypes, as was shown in previous studies of WNV-infected mouse embryo fibroblasts [33]. Such infectivity based assays have the advantages of relying on endogenous IFN levels and on the endogenous OAS1 promoter. The increased sensitivity of such approaches make them more capable of measuring subtle differences in OAS1 promoter activity that may be important for inhibiting viral replication during the initial stage of infection. Unfortunately, these types of infectivity based assays are not easily implemented in a retrospective case-control study where cases are selected after they have been diagnosed with fatal WNV disease. Therefore, experiments to test the effects of individual OAS1 promoter variants will have to await the establishment of equine cell lines derived from a variety of tissues that are infected early in WNV disease progression.
While naturally occurring mutations have demonstrated a central role for Oas in murine resistance to WNV infection, the radically different composition of the mammalian OAS gene clusters make it difficult to extend this conclusion. Our results demonstrate that OAS1 contributes to naturally occurring WNV susceptibility in a mammal that a) has a very similar OAS gene cluster to humans and b) may be more amenable to in vivo investigations of the OAS1 response to WNV infection. complete medium during transfection and subsequent treatment periods.

Statistical Analysis
Statistical association analyses were conducted using STATA 9 [23] software and the R Statistical Environment [36]. Allelic associations were conducted using Fisher's Exact tests on 262 tables. Fisher's Exact tests were conducted on 263 tables to identify genotypic associations. Significance is reported with a = 0.05. Haplotype associations were computed using a 262 design by comparing single haplotypes to all others. Case-control haplotype frequency analysis was also conducted using Phase v2. Figure S1 Local alignment of human and horse OAS1 promoters. ClustalX alignment of human (1,036 bp) and equine (1,091 bp) OAS1 promoters and 59UTR. Equine OAS1 was sequenced from CHORI BAC 100:I10 as previously described [21]. Identical sequences are designated with a star (*). The previously identified human interferon-stimulated regulatory element (ISRE) is double-underlined [24]. Significantly associated SNPs are outlined in blue with tagSNPs outlined in red. Found at: doi:10.1371/journal.pone.0010537.s001 (0.02 MB DOC)