This was a prospective, double-blinded, placebo-controlled study. The protocol was approved by the Bavarian government under the number ROB-55-2-2532.Vet_02_20_20.

In total, 17 privately owned Icelandic horses, all living in the same region near Munich in Bavaria, Germany, with ages ranging from 4 to 28 years were included in the study. All of them were diagnosed with IBH based on clinical presentation. All showed at least 2 years of typical skin lesions, strong pruritus, hair loss, and, in some cases, secondary bacterial infections, especially at the mane, tail, and dorsal and ventral midline during the insect season. All were protected from insect bites by blankets and/or stabling and/or the use of a topical lotion containing 25% benzyl benzoate. Thirteen horses were treated orally with an herbal solution with a repellent effect according to the manufacturer’s instructions (Ökozon, Einhorn, Osterholz-Scharmbeck, Germany). No changes in the treatment of the individual horses were made during the whole duration of the study.

All horses were scored monthly by the same clinician and the owner from April until October during two subsequent years, consisting of the pre-treatment and the first treatment years (Supplementary Figure 1).

A standardized scoring system adapted from Olsen et al. (19) was used as an IBH lesion score. Briefly, each half of the horse's body was divided into 13 different zones, such as the head, hind legs, and neck. (19). In each zone, the clinical lesions such as crusts, lichenification, or excoriations were scored on a scale from 0 = healthy skin with no lesions to 3 = severe clinical lesions. The total score was calculated by adding the individual scores.

To determine a pruritus score, each horse owner had to score her/his horse twice weekly every fourth week during the insect season at the same time of the day using a modified scoring system by Friberg and Logas (20). Briefly, the owner had to observe the horse for 15 seconds (s) and record every pruritic act such as rolling, biting, stamping, scratching, or head shaking. Every act over a period of 15 s was awarded one point. The same occurred during the second subsequent period of 15 s in the same week. The sum of the points from both observations formed the monthly pruritus score.

In addition, the owner assessed the overall hair coat and skin quality monthly. The score was either 0 (healthy and normal), 1 (mild changes), 2 (mild to moderate changes), 3 (moderate changes), or 4 (severe changes).

In the first treatment year, blood samples were taken by the clinician during each visit. Blood was taken from the left jugular vein into serum tubes (Vacuette, Greiner Bio-One GmbH, Frickenhausen, Germany), centrifuged (5000 rpm) for 10 min at room temperature, and the serum was kept frozen until further evaluation.

The horses were assigned to a placebo (n = 8) and a verum group (n = 9) to have balanced groups with regards to stables, the origin of the horse [born in Iceland and exported to Germany (IS) or born and living in Germany (DE)], age, and genders (Table 1). The horse owners and the clinician who immunized the horses and evaluated the clinical signs were all blinded.

Horses were vaccinated subcutaneously into the left neck at the start of spring (March) and 4 and 16 weeks later. Nine horses were vaccinated with a pool of nine recombinant Culicoides allergens. The vaccine consisted of 20 μg each of eight Culicoides obsoletus (Cul o) recombinant (r-) allergens (Cul o 1P, Cul o 2P, Cul o 3, 5, 7, 8, 9, 11) and one Culicoides nubeculosus (Cul n) r-allergen (Cul n 4). All were produced in Escherichia coli, purified as described (6), and sterile filtrated before use. A mixture of 50 μg MPLA (MPLA-SM VacciGrade, InvivoGen, Toulouse, France, cat#vac-mpla2) and 500 μg aluminum-hydroxide-gel (Alhyhydrogel® 2%, InvivoGen, cat#vac-alu-50) was used as adjuvants as described previously (18). Allergens were chosen based on their relevance for IBH in Icelandic horses (6). The eight horses of the placebo group were vaccinated with 0.4 ml of sterile saline solution. All horses were examined prior to vaccination to ensure that the horses were healthy otherwise.

Based on the results from the first treatment year, the study was adapted and extended until July of the following year (second treatment year) on a slightly reduced number of horses with limited monitoring, only consisting of the assessment of the IBH lesion score in May and July by the clinician who was still blinded. In this second treatment year, the placebo (n = 7) and treatment horses (n = 9) were vaccinated in May and July as described above (Supplementary Figure 1).

Allergen-specific IgE was determined in the sera taken during the summer of the pre-treatment year using the same microarray as described in Novotny et al. (6). The results are displayed as a percentage of IgE-positive horses for each single allergen.

Serum levels of allergen-specific antibodies were measured in the sera taken at 10 different time points (March, April, May, June, July, August, September, and October in 2021, and May and July in 2022) in all horses using the most relevant Culicoides r-allergen included in the AIT (Cul o 8) and, for comparison, a Culicoides r-allergen not included in the AIT (Cul o 10). The enzyme linked immunosorbent assay (ELISA) was carried out as described previously (18, 21) but using 384-well plates (Immunolon 4HBX, Thermo Fisher Scientific, Rochester, NY, USA). Accordingly, volumes of 50 μl per well were used, except for blocking, where 80 μl were applied. For IgE detection, samples were run in duplicates, while triplicates were used for IgG determination. Briefly, plates were coated with the recombinant Culicoides r-allergens Cul o 8 and Cul o 10, expressed in insect cells and purified as described (18), diluted to 1 μg/ml in 0.2M carbonate–bicarbonate buffer, pH 9.4 (Thermo Fisher Scientific). Non-specific binding sites were blocked with 5% dried milk powder and 5% Tween 20 in phosphate buffered saline (PBS). Sera from the study horses as well as positive and negative control sera were diluted in blocking buffer, 1:10 for IgE and 1:800 for IgG determination, added to the plates, and incubated overnight at 4°C. After washing, anti-equine IgE [clone 3H10; (22)] or anti-equine IgG1, IgG4/7, or IgG5 were added (23). After incubation and washing, an alkaline phosphatase (AP) labeled goat anti-mouse IgG (Jackson ImmunoResearch, Ely, UK, Cat# 115-055-071, RRID: AB_2338535) was added. For pan IgG detection, an AP-labeled goat anti-horse IgG-Fc (Jackson ImmunoResearch Cat# 108-055-008, RRID: AB_2337493) was used. Plates were developed with p-nitrophenyl phosphate (PNPP) in 10% diethanolamine buffer, pH 9.8. Absorbance was measured at 405 nm. Blank values were subtracted and mean values were calculated.

The ability of the horse sera to block IgE binding of IBH-horses to Cul o 8 was tested in an inhibition ELISA performed as previously described (21). Briefly, the ELISA plates were coated with Bac Cul o 8 and blocked as described above. Three pools of sera from the AIT or placebo group containing the same amount of serum from each of the horses within each group, taken before vaccination (March), after the third vaccination (August of the first treatment year), and in the second treatment year (July) were serially twofold diluted (1:10–1:160) and added to the plate. After incubation for 1 h at 37°C, serum from an IBH-affected horse with high IgE to Cul o 8 was added. The ELISA was then carried out as described above. The percentage of inhibition for each dilution of the pre- and post-vaccination serum pools was calculated.

Graphs and statistical analyses were done in GraphPad Prism 10.2.2 (www.graphpad.com). As the data were not always distributed normally (Shapiro–Wilk Test), means and standard deviation or medians and interquartile range were plotted. Clinical lesion score, pruritus, and owner assessment scores were calculated for each month.

A comparison of the scores between the AIT and placebo groups was carried out for each month using a t-test (Figure 1). The response to treatment was assessed in analogy to Fettelschoss-Gabriel et al. (24). The average lesion score during the IBH season (May–October) was calculated for each horse. The percentage of horses achieving >50% improvement of the mean clinical lesion score was determined and compared between the AIT and placebo groups using a two-sided Fisher's exact test (Figure 2A). The same procedure was used to assess the treatment response in the second treatment year by using the average lesion scores of May and July of the pre- treatment and second treatment years (Figure 3B). Furthermore, for each month, the difference in scores between the treatment and pre-treatment years (Δ lesion score) was calculated and subsequently, the average of the Δ lesion scores (May–October) from the horses in the AIT group was compared to those treated with placebo using the Mann–Whitney test (Figure 2B). Average IBH lesion scores of May and July of the pre-, first, and second treatment years were calculated for each horse and compared using a paired ANOVA test with the Holm–Sidak correction for multiple comparisons (Figure 3A). Owner general assessments and pruritus were compared similarly. Furthermore, Spearman’s rank correlation coefficient between the IBH lesion score and the owners’ skin assessment and pruritus scores was calculated.

Statistical analysis of the serum antibody levels was performed using an RStudio software package (version 2023.06.1 + 524; www.r-studio.com). To test the global effect of the factors “group” (AIT/placebo) and “time point” on the antibody levels, a repeated measures analysis of variance was performed. Given the non-normally distributed data, a non-parametric test method according to Brunner et al. was used (25). The NparLD software package was used to analyze the individual antibody class levels to the individual Culicoides recombinant allergens Cul o 8 and Cul o 10. The NparLD software package employs robust rank-based methods for analyzing longitudinal data in factorial settings (26). ANOVA-type test statistics with Box approximation (ANOVA test) were calculated to assess group and time effects, as well as their interactions.

p-values ≤0.05 were considered significant.

In both the placebo and treatment group, there was a reduction in the IBH lesion score in the treatment year compared to the pre-treatment year. There were no significant differences between the groups for any of the single time points in the pre- and first treatment years, although, in September and October, the mean IBH score started to diverge between the groups (Figure 1). Assessment of the IBH lesion score showed that 67% of the AIT and 25% of the placebo horses reached >50% improvement in the first treatment year (Figure 2A, ns). The decrease (Δ) in the average IBH lesion score between the treatment and pre-treatment year was significantly larger in the AIT compared to the placebo group (median = −8.83 in the AIT vs. −2.75 in the placebo group, p < 0.05; Figure 2B).

The AIT effect was more pronounced in the second treatment year. Both in May and July, the IBH lesion score was significantly lower in the AIT compared to the placebo group (Figure 1). In the placebo group, the average IBH lesion score slightly increased from the first to the second treatment year (Figure 3A, ns). In contrast, the average IBH lesion score in the AIT group significantly decreased between the first and second treatment years (Figure 3A) and the variability of the scores was lower in the AIT group compared to the placebo group. Accordingly, 89% of the AIT horses but only 14% of the placebo horses achieved >50% improvement in the second compared to the pre-treatment year (Figure 3B, p < 0.01).

The pruritus score decreased significantly in the treatment compared to the pre-treatment year in both the placebo and AIT groups, but the difference was more pronounced in the AIT group (Figure 4A). Owner assessment of the hair coat and skin showed a significant reduction of the score in the AIT group in the treatment compared to the pre-treatment year (p < 0.01). A reduction in the score was observed in all AIT horses (Figure 4B). In the placebo group, there was much more variability in score changes between the pre-treatment and treatment years. The difference between the treatment and pre-treatment years was not significant in the placebo group. Direct comparison between the AIT and placebo groups for pruritus or haircoat and skin scores did not result in any significant differences between the groups.

There was a strong correlation between the owner skin assessment and the IBH lesion scores (Spearman’s rank correlation coefficient = 0.74, p < 0.0001), while the correlation between the pruritus and the IBH lesion scores was moderate (0.49, p < 0.0001).

In the summer of the pre-treatment year, between 71% and 93% of the horses had positive IgE values to the single Culicoides allergens included in the vaccine, except for Cul o 5, where only 43% of the study horses had a positive IgE result (Figure 5A). The horses also showed varying sensitization patterns to the other eight Culicoides obsoletus allergens of the microarray not included in the vaccine (Figure 5B).

A similar pattern of sensitization to the vaccine allergens was observed in the AIT and placebo groups, except for Cul o 5 to which only 20% of the AIT vs. 80% of the placebo group were IgE positive (Supplementary Figure 2). The AIT horses were sensitized to a mean number of 6.9 of the nine vaccine allergens (range 0–9). The placebo horses were sensitized to almost the same number of allergens (mean = 7.4 allergens, range 2–9 allergens).

As shown for the AIT allergen Cul o 8 (Figure 6A), IgE serum levels did not differ significantly between the placebo and AIT groups (Brunner–Langer model). However, there was a significant variation over time in Cul o 8-specific IgE levels (p ≤ 0.05) as well as a significant (p ≤ 0.05) interaction between group and time: changes over time in IgE levels showed a different pattern in the AIT group and in the placebo group. In the AIT group, IgE levels to Cul o 8 had already increased 4 weeks after the first immunization, i.e., in April, and decreased in May, after the second immunization. In the placebo group, IgE started to increase only in May, i.e., 1 month later than in the AIT group, probably because of exposure to Culicoides. Later, median IgE levels remained similar in both groups until May of the next year, but then they decreased in the AIT group and increased in the placebo group. IgE levels to Cul o 10, an allergen not included in the vaccine, were higher in the AIT group but the difference was not significant. Cul o 10-specific IgE also varied over time (p ≤ 0.01) but the changes were parallel in both groups (Figure 6A) and thus there was no significant interaction between group and time in the Brunner–Langer model.

IgG levels increased after immunization in the AIT group (Figure 6B), and, consequently, IgG levels to Cul o 8 were significantly higher in the AIT compared to the placebo group (Brunner–Langer model, p < 0.05). IgG to Cul o 8 varied significantly over time (p < 0.001) and the interaction between group and time was significant (p < 0.05). IgG levels to Cul o 10 did not differ significantly between the groups, as shown in Figure 6B. They only varied slightly over time (p < 0.01) and no interactions between group and time were observed.

The IgG subclass response was investigated for Cul o 8 (Figure 7). Significant effects of group (p < 0.05) and time (p < 0.00001) were observed for the three tested subclasses IgG1, IgG 4/7, and IgG5. After an increase following the immunizations, all subclasses decreased until the following May. The immunization in May 2022 resulted again in an increase in all IgG subclasses. Significant interactions between the treatment group and time were observed for IgG1 and IgG4/7 (p < 0.05). When comparing the antibody levels to Cul o 8 at the single time points between the AIT and placebo groups, IgG1 levels were only significantly higher in the AIT compared to the placebo group in April and May, while IgG4/7 levels were significantly higher in the AIT group at all times points following the first AIT. This was similar for IgG5, except for July in the 1st treatment year. Interestingly, IgG4/7 was the only antibody class where the difference between the groups reached significance after Bonferroni correction for multiple comparisons.

The blocking activity of the sera was investigated for the allergen Cul o 8. AIT-induced antibodies were able to block IgE binding to this allergen. The blocking activity of the serum pool of the AIT group increased from 1.5% in the pre-immune sera to 43% after the third immunization and was similar in July of the second treatment year. The blocking activity of the serum pools taken after the AITs was dilution-dependent, i.e., it decreased almost linearly between the 1:10 and 1:160 serum dilutions. Some blocking activity was seen with the serum pools of the placebo group, but it was random and not dilution-dependent (Figure 8).