Genomic Regions Associated with IgE Levels against Culicoides spp. Antigens in Three Horse Breeds

Insect bite hypersensitivity (IBH), which is a cutaneous allergic reaction to antigens from Culicoides spp., is the most prevalent skin disorder in horses. Misdiagnosis is possible, as IBH is usually diagnosed based on clinical signs. Our study is the first to employ IgE levels against several recombinant Culicoides spp. allergens as an objective, independent, and quantitative phenotype to improve the power to detect genetic variants that underlie IBH. Genotypes of 200 Shetland ponies, 127 Icelandic horses, and 223 Belgian Warmblood horses were analyzed while using a mixed model approach. No single-nucleotide polymorphism (SNP) passed the Bonferroni corrected significance threshold, but several regions were identified within and across breeds, which confirmed previously identified regions of interest and, in addition, identifying new regions of interest. Allergen-specific IgE levels are a continuous and objective phenotype that allow for more powerful analyses when compared to a case-control set-up, as more significant associations were obtained. However, the use of a higher density array seems necessary to fully employ the use of IgE levels as a phenotype. While these results still require validation in a large independent dataset, the use of allergen-specific IgE levels showed value as an objective and continuous phenotype that can deepen our understanding of the biology underlying IBH.


Introduction
The most common skin disease in horses, insect bite hypersensitivity (IBH), is the result of a cutaneous allergic reaction to salivary antigens from Culicoides spp. [1][2][3]. The affected horses show

Sampling
One research group studied Icelandic horses (n = 127) and Shetland ponies (n = 200). Individuals from both breeds were sampled in areas with different risk status for IBH in the Netherlands. Individuals were sampled while using a matched case-control design to avoid population stratification, as described by Schurink and colleagues [14,15]. All the individuals were inspected by one of the researchers and a veterinarian, while using a standardized scoring system and questionnaire to obtain detailed information regarding the history and management of IBH (for more details, see Schurink and colleagues [15]). Cases were defined as ponies showing mild to severe clinical signs, and controls were free of signs despite exposure to Culicoides spp. Therefore, controls were required to be at least four years of age (related to the average age of onset of IBH being between two and four years of age). Additionally, at the same premises, a case had to be present, to ensure the exposure to Culicoides spp. Paternal half-sibs were sought, to match cases and controls on genealogy as well. Most cases and controls were present for two or more years on their current premises to ensure constant management practices (>90%). The Shetland pony population consisted of 200 ponies (103 cases and 97 controls), while the Icelandic horse populations contained 127 horses (56 cases and 71 controls).
Belgian Warmblood horses (n = 223) were studied by the other research group and sampled during stable visits and competitions in the summer, as described by Peeters and colleagues [29]. Horse owners were asked about various environmental traits (e.g., soil and stable type) and horse traits (e.g., IBH status and body condition) by a trained investigator while using a questionnaire [33]. The cases were defined as horses showing or that had shown clinical signs. If the owner had never observed sings since the ownership of the horse began, the horse was classified as control. Additionally, controls were required to be at least 3 years of age and kept on premises with a case. Horses were selected for genotyping while using the exclusion protocol that is presented in Appendix A- Figure A1, leading to the selection of 107 controls and 116 cases. In short, a horse was excluded when IBH status was unknown or unclear and when it was not a purebred Belgian Warmblood horse. In selecting cases and controls, pedigree was considered (no full-sibs allowed), and an equal distribution between cases and controls across several traits was aimed for.
Information regarding the date of sampling, age, and sex was available in all breeds, information on coat color and size category was available for Shetland ponies, information on coat color and import status was available for the Icelandic horses, and information on vegetation, humidity, body condition, fitness, stable management, training frequency, and deworming frequency was available for Belgian Warmblood horses [14,15,29].
Concerning the Shetland ponies and Icelandic horses, the Board on Animal Ethics and Experiments from Wageningen University approved blood sample collection (experiments 2009055 and 2010109). For the Belgian Warmblood horses, the Ethical Committee for Animal Experiments of the Katholieke Universiteit Leuven approved blood sample collection (approval number P061-2012, date of approval 18 April 2012).

Phenotype-IgE Levels Determined with ELISA Test
Diagnostic ELISA tests were performed to determine the IgE levels (OD 450nm values) against Culicoides obsoletus (all breeds) and Culicoides nubeculosus (only Belgian Warmblood horses), as described by Peeters and colleagues [29], Schaffartzik and colleagues [25], and van der Meide and colleagues [26]. Belgian Warmblood horses were sampled from June through September to ensure exposure to Culicoides spp. Shetland ponies and Icelandic horses were sampled in September and October. More information regarding the recombinant allergens that were used in the different populations can be found in Table 1. The highest sensitivity (93.2%) in Shetland ponies and Icelandic horses was obtained while using a C. obsoletus whole body extract (WBE) [26]. In Belgian Warmblood horses, the highest sensitivity (70%) was obtained while combining IgE levels against Culn4, Culo1 (Genbank id JX512273), and Culo2 (Genbank id JX512274) with C. obsoletus WBE [29]. The obtained specificity of the ELISA test in the Belgian Warmblood horses using these allergens was 97% [29].

DNA Extractions, Genotyping, and Quality Control
For the Icelandic and Shetland samples, DNA was extracted, as described by Schurink and colleagues [14], and the samples were genotyped while using the Illumina equine HD chip containing 65,157 single-nucleotide polymorphisms (SNPs). For the Belgian Warmblood samples, DNA was extracted from blood samples (200 µL) while using the Qiagen kit. Belgian Warmblood samples were genotyped using the Affymetrix Axiom Equine Genotyping Array containing 670,796 SNPs. Quality control was separately assessed in each population using GenABEL in R [35]. SNPs with a minor allele frequency (MAF) <1% and call-rate <95% were discarded and a threshold for minimum call rate per horse was set to 90%. In the Icelandic horse population, 127 horses and 52,655 SNPs (80.8% of the SNPs) remained after quality control, while 199 ponies and 49,567 SNPs (76.1% of the SNPs) remained in the Shetland pony population. In the Belgian Warmblood horse population, 214 horses and 474,128 SNPs (70.7% of the SNPs) retained after quality control.  [29] and Culo1b (Genbank id KC339671) and Culo2b (Genbank ID KC339672) by van der Meide and colleagues [31]. 3 Combi1 contained allergens Culo3, 5, and 7 and Combi2 contained allergens Culo1b, 2b, 5, and 7 [32]. 4 Seven C. nubeculosus salivary gland recombinant allergens (Culn1, 3-5, 7, 8, 10) [25,29,36,37]. 5 WBE = whole body extracts from C. obsoletus biting midges, as constructed and tested by Peeters and colleagues [29]. 6 WBE = whole body extracts from C. obsoletus biting midges, as constructed and tested by van der Meide and colleagues [26]. 7 Thorax and head (TH) extracts of C. nubeculosus (body removed) [29].

Genome-Wide Association (GWA) Study
IgE levels (OD 450nm values) against Culicoides spp. were log transformed while using the natural logarithm, as preliminary analyses showed the presence of heterogeneous variance and deviations from a normal distribution (Appendix B- Table A1). These log transformed IgE levels were used as continuous traits in the GWA analyses. IgE levels against each allergen were separately tested, which resulted in 11 GWA studies in Belgian Warmblood horses and 10 GWA studies in both Icelandic horses as well as Shetland ponies.
In the Belgian Warmblood horse population, several covariates were obtained from the questionnaire that was performed at each sampling [33]: sex, age (in years), ageClass (categorical variable), period of sampling, date of sampling, vegetation at pasture, humidity around pasture, type of stabling, weight of the horse, fitness level of the horse, number of trainings per week, and number of parasite treatments per year. All of these variables were tested for having a significant effect on the IgE levels. While several variables were found to explain part of the variation observed in different IgE levels, only the variable "ageClass" had a significant effect on all IgE levels (except Culn1). Therefore, only this variable was included in the model in the downstream analyses. In the Shetland pony population, age, size category, coat color, month of sampling, and location of sampling were tested as potential factors or covariates having a significant effect on IgE levels. The age of the pony had a significant effect on several C. obsoletus specific IgE levels in the Shetland pony population, and therefore included in the model. In the Icelandic horse population, sex and age were tested as potential factors or covariates having a significant effect on IgE levels. However, none of these variables had a significant effect on IgE levels. Due to differences in the age of exposure to Culicoides spp., unequal sampling of cases and controls of imported Icelandic horses (in total 19 horses), and the subsequent confounding effect, only European born Icelandic horses were used to perform the GWA study (n = 127). The significance of factors and covariates was determined while using a linear model (command anova(lm,...) in the nnet package) in R software [38].
For each population, the genomic kinship matrix was computed and standard K-means clustering was performed to determine the number of potential subpopulations [39]. In the Belgian Warmblood horse population K = 3 was established, in the Shetland pony population K = 2, and K = 4 in the Icelandic horse population. Clustering within the populations did not coincide with the case-control status ( Figure 1). Additionally, no outliers were apparent in any of the studied populations while using the multidimensional scaling (MDS) plot ( Figure 1). GWA analyses were performed with the IgE levels against Culicoides antigens as a continuous trait using the GenABEL package in R [35,38]. To avoid spurious associations that may arise with unusual allele frequency differences between subpopulations [40], a mixed model-structured association approach ("mmscore" function) was employed with a correction for substructure through the use of the assigned cluster as a covariate [35]. The p-values were corrected for inflation factor lambda (Pc1df). The conservative Bonferroni corrected significance level was 1.01 × 10 −6 in Shetland ponies, 9.50 × 10 −7 in Icelandic horses, and 1.05 × 10 −7 in Belgian Warmblood horses (being α/n, where α is the desired significance level (0.05) and n is the number of SNPs that was tested). As none of the SNPs passed Bonferroni corrected significance threshold, the 50 SNPs that were most significantly associated with antigen-specific IgE levels per breed were presented. . Additionally, to investigate whether the higher density SNP-array in Belgian Warmblood horses rendered more significant results, a subset of SNPs was obtained representing the overlap between the 65 k Illumina equine HD chip and the 670 k Affymetrix Axiom Equine Genotyping Array. As a case study, a GWA analysis was performed in the Belgian Warmblood horse population using all SNPs (474,128 after quality control) and using the subset of SNPs for L_Culn7 (as most significant results were obtained using the IgE levels against this C. nubeculosus allergen).
Additionally, to investigate whether the GWA analysis using IgE levels against Culicoides spp. is indeed more powerful in detecting associated genetic variants as compared to a case-control GWA analysis, a case-control GWA analysis was performed in the Belgian Warmblood horse population using a structured association analysis ("qtscore") in the GenABEL package in R [35,38]. A case-control GWA study using the same Shetland ponies and Icelandic horses as in our study was already performed and presented in scientific literature [14,15]. The results from our GWA analyses using the IgE levels will be compared to the results that were obtained in these case-control studies [14,15].

Results
The locations of the 50 SNPs per breed most significantly associated with IgE levels against Culicoides spp. allergens in Belgian Warmblood horses, Icelandic horses, and Shetland ponies are presented in Figure 2. Genomic regions that are associated with IgE levels against Culicoides spp. allergens across two or all breeds were found on ECA2  Information on the p-value of these SNPs, allele frequency, and allele substitution effect of the minor allele, as well as which Culicoides spp. allergen it was associated with, are presented in Table A2 (Appendix C). Some Culicoides spp. allergens obtained more significant results (Appendix C- Table A2), where, in Shetland ponies, 11 out of 50 SNPs were associated with L_Culo4, 11 SNPs with L_Culo6, and 17 SNPs with L_Culo7. In Icelandic horses, 13 out of 50 SNPs were associated with L_Culo4. In Belgian Warmblood horses, 13 out of 50 SNPs were associated with L_WBE and nine SNPs with L_Culn7.
The GWA analysis in the Belgian Warmblood horse population using a subset of SNPs representing the density in the two other populations rendered less significant results as compared to the GWA analysis using all SNPs (474,128 after quality control; Appendix D- Figure A2a,b).
The results from the case-control GWA analysis in the Belgian Warmblood horse population are presented in Figure A3 (Appendix E). The case-control GWA analysis obtained less significant results when compared to the GWA analysis using IgE levels against Culicoides spp. allergens (Appendices C and D- Figure A2). P-value of the SNP most significantly associated with case-control status in the Belgian Warmblood horse population was 1.22 × 10 −5 (AX-103321942 on ECA13:728,323 bp).
A comprehensive list of genomic regions that were associated in any of the three investigated breeds can be found in Table A3 (Appendix F). Previously identified associations from case-control GWA analyses (in any equine breed) in proximity to the genomic regions identified in our study are also presented.

Discussions
IBH is a serious welfare concern in several horse breeds, and several studies have tried to identify the genetic background of IBH in order to gain a better understanding of the disease and optimize selection strategies [14][15][16][17][18]. The misclassification of IBH can occur as clinical signs are not always visible, depending on exposure to Culicoides spp., and it may originate from some other cause, making a typical case-control GWA study less powerful [41][42][43] when compared to a continuous, independent, and objective trait that reflects the underlying sensitivity. Our study used a diagnostic ELISA test with high sensitivity and specificity to assess the IgE levels against Culicoides spp. antigens, resulting in an objective, independent, and quantitative phenotype as input for a GWA study [34,44]. The classical diagnosis of IBH combines the history of the horse and the observation of clinical signs that follow a seasonal pattern [2,3,45]. As this might lead to the failure to detect affected horses, a diagnostic test could give a more reliable observation of the horse's sensitivity to Culicoides spp. if they have been exposed to the midges. Several diagnostic tests are available, such as the measurement of inflammatory mediators [46,47], the measurement of wheal size after intradermal allergen challenge [24,48], and the use of an ELISA test to analyze allergen-specific IgE levels in the serum of horses [25,26,29]. These tests mostly use recombinant allergens from different species of Culicoides, and they were able to differentiate between cases and controls in several breeds. While using a C. obsoletus WBE, a sensitivity of 93.2% in Shetland ponies and Icelandic horses was reached with a specificity of 90% [26]. In Belgian Warmblood horses a sensitivity of 70% and a specificity of 97% was obtained combining Culn4, Culo1, and Culo2 with IgE levels against C. obsoletus WBE [29]. In contrast to the classical IBH case-control scoring system, the IgE levels from the allergen-specific ELISA test provided a continuous and objective phenotype that allowed for a more optimal and powerful GWA analysis [34,45]. When compared to the case-control GWA analysis that was performed in our study, more significant associations were identified when using IgE levels. Van der Meide and colleagues [26] showed that there is variation in IgE levels within cases and within controls, and IgE levels against Culicoides spp. antigens were correlated with the severity of IBH clinical signs [30]. Similarly, in humans, a clear correlation has been established between the total IgE levels or allergen-specific IgE levels and severity of eczema signs [49]. Provided that the variation in IgE levels against Culicoides spp. is related to IBH sensitivity, IgE levels will be more accurate in representing this sensitivity when compared to a case-control classification. Moreover, through using IgE levels from an ELISA test with high specificity and sensitivity misdiagnosis is limited and a more direct genotype to phenotype relationship is established.
Associations with IgE levels were found on most chromosomes, where across-breed associations were identified on ECA2, ECA3, ECA6, ECA7, ECA15, and ECA20. The complex nature of IBH supports the selection of any SNP surpassing a reasonable significance threshold, even though no SNP reached genome-wide significance based on the strict Bonferroni correction [50][51][52][53]. Although a common genetic background is to be expected across different breeds concerning a Th2 driven IgE production and induction of allergic diseases, selection criteria to sample cases and controls, allergens used to determine the IgE levels, and the SNP density differed between the three breeds under study, which might have contributed to a few regions being associated across breeds. Even with the use of the same density chip in the Shetland pony and Icelandic horse, certain regions in the genome might contain different SNP densities, as each breed underwent a separate quality control. In addition, linkage disequilibrium (LD) between SNPs and a causal mutation might differ between the breeds [54].
When comparing the different GWA analyses on IBH using a case-control approach that was based on clinical signs [14][15][16][17][18] and our quantitative trait being antigen-specific IgE levels against Culicoides spp. revealed several associated genomic regions across various chromosomes in common (Appendix F). For instance, the identified region on ECA6 in Belgian Warmblood horses (44)(45)(46) was in close proximity to regions that were significantly associated with IBH in Shetland ponies [15] and Icelandic horses [17]. Other regions that were associated with IgE levels and IBH status across two or more unrelated breeds were found on ECA1:  [17]), no obvious candidate genes that are involved in the Th2 driven IgE production and induction of allergic disease were detected in close proximity to the identified regions significantly associated with allergen-specific IgE levels or previous GWA analyses based on case-control status. Although several previously identified genomic regions that are associated with IBH were confirmed, new regions of interest were also found. These new regions might potentially represent additional or other pathways identified due to the use of the more objective and quantitative IgE levels, as compared to case-control status representing the presence or absence of clinical signs.
The GWA analyses in Belgian Warmblood horses resulted in more significant associations when IgE levels were used as compared to a case-control status. These analyses allowed for a direct comparison, as the same horses were used. The identified associations based on case-control status in the same Shetland ponies as investigated in our study were less significant [15]. However, similar significance levels as in our study were observed in Exmoor ponies and Friesians horses [17,18] while using the same SNP density as in Belgian Warmblood horses. The Exmoor pony and Friesian horse populations represent inbred horses breeds and larger samples were used [17,18] as compared to our breeds. Inbreeding likely resulted in higher LD [55] and, when combined with the larger sample size, more power to detect causal variants. A comparison with results from Shrestha and colleagues [16] and Schurink and colleagues [14] is quite challenging, as a Bayesian approach was employed, presenting the percentage of genetic variance explained by 1 Mb windows. The use of allergen-specific IgE levels showed value as a continuous and objective phenotype.
The ELISA test applied in Belgian Warmblood horses used a different set of Culicoides antigens (several recombinant allergens of C. nubeculosus and C. obsoletus) as compared to the test applied in Shetland ponies and Icelandic horses where only allergens of C. obsoletus were used [26,[29][30][31][32]. Anderson and colleagues [56] showed that cases reacted to all the extracts of Culicoides spp., even when cases had not been exposed to most of the species, which suggested that the allergen(s) were present in all the investigated extracts (native C. obsoletus, C. cockerellii, C. imicola, C. biguttatus, C. variipennis, and non-native C. obsoletus). However, using the diagnostic test in Belgian Warmblood horses, different reactions were observed to several recombinant C. nubeculosus and C. obsoletus antigens, where generally C. obsoletus allergens best differentiated between cases and controls [29]. Hardly any C. nubeculosus and C. sonorensis were found feeding of the horses in the Netherlands [57]. Correspondingly, the observed IgE levels against C. nubeculosus and C. sonorensis allergens were lower when compared to levels against C. obsoletus [26]. The use of different allergens might have contributed to the differences in results that were obtained in Belgian Warmblood horses as compared to Shetland ponies and Icelandic horses.
While using allergen-specific IgE levels against Culicoides spp. in the serum of horses as phenotype for a GWA analysis has allowed for the identification of several new genomic regions of interest. Previous studies used lower density SNP arrays and/or a case-control GWA study design, and they were only able to explain a small part of the IBH heritability [14][15][16][17][18]58]. LD in most breeds declines rapidly, reaching a background level in only 1 or 2 Mb, which indicates that the use of lower density chips will be unable to detect causal mutations at a distance of 2 Mb or more from the genetic marker [59]. In combination with a lower SNP density being employed in Shetland ponies and Icelandic horses, and the use of other Culicoides antigens, this might explain why fewer regions were significantly associated in these breeds as compared to the Belgian Warmblood horse breed [58][59][60]. Indeed, when a subset of SNPs was selected in the Belgian Warmblood horses, which represented the same density and SNPs being tested in Shetland ponies and Icelandic horses, the obtained associations were less significant (Appendix D). Therefore, our results seem to indicate that it might require a higher density array to take full benefit of using these IgE levels as a more objective, independent, and quantitative phenotype.

Conclusions and Implications
Using a diagnostic ELISA test to assess the IgE levels against Culicoides spp. allergens provided a more objective, independent, and quantitative phenotype (as compared to IBH case or control) to be used as input for a GWA study. The GWA analyses employing IgE levels as a phenotype seem to have more power than the GWA analyses that used case-control status as the phenotype. Our GWA results confirmed the complex nature of IBH in horses, affirming several previously identified genomic regions that were associated to IBH and identifying new regions of interest, potentially representing additional or other pathways that are involved in IgE levels as compared to the case-control status. Our findings also seem to indicate that the use of a higher density array might be necessary to fully employ the use of allergen-specific IgE levels. Although our results need to be verified in a larger dataset, the use of allergen-specific IgE levels and the availability of high density SNP genotypes allowed for the identification of new candidate regions that are possibly important in the aetiology of IBH and contributing to our understanding of its biology.       ) against Culicoides spp. were log transformed using the natural logarithm in Icelandic horses and Shetland ponies and using the negative natural logarithm (−ln(OD 450nm values)) in Belgian Warmblood horses. 2 Similar nomenclature but different proteins were used: Culo1a (Genbank id JX512273) and Culo2a (Genbank ID JX512274) by Peeters and colleagues [29] and Culo1b (Genbank id KC339671) and Culo2b (Genbank ID KC339672) by van der Meide and colleagues [31]. 3 Combi1 contained allergens Culo3, 5, and 7 and Combi2 contained allergens Culo1b, 2b, 5, and 7 [32]. 4 Seven C. nubeculosus salivary gland recombinant allergens (Culn1, 3-5, 7, 8, 10) [25,29,36,37]. 5 WBE = whole body extracts from C. obsoletus biting midges, as constructed and tested by Peeters and colleagues [29]. 6 WBE = whole body extracts from C. obsoletus biting midges, as constructed and tested by van der Meide and colleagues [26]. 7 Thorax and head (TH) extracts of C. nubeculosus (body removed) [29].