An allergen peptide microarray had previously been designed, produced, and utilized to determine the peptide binding pattern of IgE, IgG, and IgG4 in serum or plasma from 15 allergic donors (20). Briefly, the microarray contained 16-mer peptides, with generally 15 amino acid overlap, of all allergens that at the time were present in the allergen database generated by the World Health Organization/IUIS Allergen Nomenclature Sub-Committee (http://www.allergen.org, accessed 2014-10-30) and of five additional proteins.

As previously described (26), samples were collected from 15 allergic subjects, of whom subjects 1–8 were subjected to AIT and subjects 9–15 were untreated controls. Donor 2, 3, 4, 7, and 8 were treated with subcutaneous injections with a grass pollen mix (Alutard®, ALK-Abelló, Hørsholm, Denmark), while the AIT used for donor 1, 5, and 6 contained no grass pollen allergens. Samples were obtained at treatment initiation, during vaccine up-dosing (at the stage when a dose of 10,000 SQ-U had been reached), and 1 and 3 years later (timepoint 1, 2, 3, and 4, respectively). The reactivity of these samples had previously been analyzed in-depth with respect to antibody responses against linear epitopes on pollen allergens of the beta-expansin and the PR-10 families (20).

Antibody binding patterns were evaluated using allergen peptide microarray slides. Quantification was, as described previously (20), done using Alexa Fluor 532 (for IgE) and Alexa Fluor 647 (for IgG4 and IgG) conjugated secondary antibodies and a NimbleGen MS200 scanner (Roche NimbleGen Inc.). Scanned images were converted into signal intensity values using GenePix Pro 5.1 image analysis program.

Candidate epitopes were determined as earlier described (20). In short, signal intensity values were log-transformed and candidate epitopes were defined by being represented by at least three adjacent peptides with a signal at least two standard deviations above the background signal for that sample.

To allow for comparison of homologous grass pollen allergens from different species and of isoallergens and variants, the amino acid sequences of all allergens belonging to the same group were aligned using MacVector Inc. (version 18.0.0). ClustalW was applied, using the Gonnet substitution matrix, open and extended gap penalty set to 10.00 and 0.20, respectively, and the slow pairwise alignment mode (Supplementary Table 1). Each peptide was assigned a number representing the position of its first amino acid, according to the alignment, and visualizations were made based on these numbers. Allergens belonging to grass pollen groups 2 and 3 as well as to groups 5 and 6 were aligned and visualized together.

Acidic ribosomal protein 1 allergens and acidic ribosomal protein 2 allergens, i.e., control allergens that were not included in the AIT, were also aligned and renumbered accordingly. Sequences of the human 60S acidic ribosomal protein 1 (P05386) and the human 60S acidic ribosomal protein 2 (P05387) were obtained from UniProt (27) and included in the alignments.

Differences in linear grass pollen-epitope recognition before and after AIT were evaluated for patients subjected to grass pollen AIT and compared with differences in epitope recognition during the same period for untreated patients and patients receiving AIT containing other allergens. First, non-reactive peptides, i.e., peptides that had not passed the candidate epitope filter, were set to zero. Background was calculated as the mean signal of a sample and subtracted from the signal intensity of all reactive peptides. Finally, the difference in signal between each peptide 3 years after treatment initiation (time 4) and the same peptide at the start of treatment (time 1) were calculated and visualized. For donors where the 3-year sample was missing, the 1-year sample was used instead. Additionally, the mean number of peptides with signal difference above 0 was compared for individuals who were or were not subjected to grass pollen vaccination. Mean values before and after AIT were calculated using peptide signal intensities of all grass pollen allergens of an allergen group and the number of peptides with higher mean value at time 4 (or in some cases time 3) compared to time 1 were counted, for each individual separately. Subsequently, the calculated values and Mann–Whitney U-test were used to compare the number of peptides with increased antigen binding over time in donors who had, or who had not, been subjected to grass pollen AIT. For some selected regions of group 5 and 6 grass pollen allergens, the mean signal intensities were also compared between the treatment initiation sample and the last collected sample. Average peptide signals for group 5 and 6 grass pollen allergens were calculated for each peptide in that region, each sample and donor separately. Differences between the time 1 and the time 3 or time 4 sample were evaluated using one-sided Wilcoxon signed-rank test, in order to identify individuals in whom the antibody response against the examined region increased during the investigated time period.

Subsequently, the total number of linear epitopes in all allergens belonging to either of three species of grass [Lolium perenne (English ryegrass), Dactylis glomerata (orchard grass), and Phleum pratense (timothy)] that are present in the mix used for AIT were evaluated. Alopecurus pratensis (meadow foxtail) and Festuca pratensis (meadow fescue) are also included in the grass pollen extract mix, but they had no available sequences in the allergen database used to design the peptide microarray and could therefore not be included in the analysis. The total number of continuous regions of peptides that were considered as reactive were calculated for each sample separately. To balance the impact of allergens with more than one known isoallergen or variant, mean values of number of epitopes were calculated for each allergen, based on all its isoallergens/variants, and used for calculation of the total number of epitopes of a sample. Finally, the number of linear epitopes recognized by antibodies in samples from donors who had, or who had not, been treated with grass pollen were compared and evaluated using Mann–Whitney U-test. The calculated values were based on six allergens (nine isoallergens/variants), four allergens, and 10 allergens (29 isoallergens/variants) originating in Lolium perenne, Dactylis glomerata, and Phleum pratense, respectively (Supplementary Table 1).

Selected epitopes were visualized on 3D structures of allergens [obtained from PDB (28)], using PyMol version 2.5.2 (29). In cases where the region of reactive peptides could be expected to contain multiple close-by epitopes, the entire region of amino acids covered by the reactive peptides was highlighted in the structure. In other cases, only amino acids that were present in all reactive peptides were expected to be part of the epitope and consequently marked in the structure. Structures with PDB id 2VXQ (30), 2M64 (31) (model 6), and 4BEH (model 11) (32) were used for Phl p 2, Phl p 5, and acidic ribosomal protein 1 and 2, respectively.

For analysis of epitope patterns in different variants of Phl p 5.01 and Phl p 5.02, the peptide signal intensity values were transformed into estimated values for each amino acid. Background values were calculated for each sample and allergen, as the mean signal intensity of all peptides that were considered as non-reactive. All non-reactive peptides were given the background signal level, to avoid single, high false positive signals to have an impact on estimated amino acid signal values in the next step. Finally, signal values were estimated for each amino acid by calculating a mean value of all peptides that were covering that amino acid and dividing by the background signal. Hence, most amino acid values were based on 16 peptide signal values, but the 15 amino acids closest to the N- and C-termini were based on fewer peptide signal values with the N- and C-terminal amino acids based solely on the first and last peptide signal, respectively.

We have recently developed an allergome-wide peptide microarray to study the IgE, IgG, and IgG4 responses to linear epitopes on some selected aeroallergens, for serum collected from allergic individuals during AIT (20). Here we exploited the collected data and the generated bioinformatic pipeline to evaluate the epitope recognition pattern against different grass pollen allergens in 15 allergic individuals, eight of which were subjected to AIT, five of whom with subcutaneous injections with a grass pollen mix (26). Linear epitopes recognized by IgG (Supplementary Figure 1) and IgG4 (Supplementary Figure 2) were identified in all examined groups of grass pollen allergens and generally seen already at time 1, i.e., before start of AIT. Antibodies of IgE isotype that bound allergen peptides were detected more rarely, and in most cases only after vaccination with grass pollen extract (Supplementary Figure 3). For grass pollen allergens belonging to group 7 and group 12, no linear epitopes were recognized by IgE at any timepoint.

The mechanism through which AIT alters the progression of allergic disease is often suggested to include increased levels of protective allergen-specific IgG (in particular IgG4) (7, 8, 15). We examined if such increased levels of IgG/IgG4 could be seen against linear grass pollen epitopes by comparing the difference in estimated signal level for each peptide between the first and last sample collected for each individual, i.e., at treatment initiation and at 1 or 3 years later. Indeed, the levels of IgG and IgG4 against many linear epitopes increased for individuals vaccinated against grass, while increased levels of IgE were more rarely seen (Supplementary Figures 4–6). In individuals who had been subjected to AIT only containing pollen extract from other sources than grass, or who had not been subjected to AIT at all, the level of antibodies against linear grass pollen epitopes did generally not increase over time (Supplementary Figures 4–6). Indeed, statistical comparison of the number of peptides with increased signal intensity over time showed that these were often more common in donors who had been subjected to grass pollen AIT compared to donors who had not (Supplementary Figures 4–6). As an example, grass pollen allergens belonging to group 5 and 6 had multiple epitopes for which IgG4 appeared to increase after vaccination, for example in the peptides starting at amino acid 21–53 (epitope A) and at amino acid 182–196 (epitope B). We further examined the peptide signal intensities for these reactive regions identified in individual 2, 3, and 4. As expected, the IgG4 signals against the relevant peptides increased significantly during the treatment compared to baseline levels in all three individuals and against both peptides, except for donor 4 and epitope B (Figure 1A; Supplementary Figures 7, 8). Total epitope-specific IgG levels also often increased, but in many cases from higher starting levels and to a lower degree, indicating both an increased production of IgG against the examined epitopes and a shift in the relationship between IgG4 and other types of IgG.

As exemplified by these epitopes, AIT can introduce antibody responses against linear epitopes not recognized by that isotype before treatment. In order to examine if vaccination generally induced production of antibodies of IgG, IgG4, and IgE isotype against novel epitopes, we also studied the immune response against the total number of linear epitopes in allergens from grass species present in the used vaccine before, during, and after AIT (Figure 2). Before treatment (time 1), the number of epitopes were similar in individuals that were to and in individuals that were not to be subjected to grass AIT. During treatment, the number of epitopes recognized by IgG, IgG4, and IgE isotypes increased, and became significantly higher than for the groups of subjects not subjected to AIT after 1 year, i.e., at timepoint 3 (Figure 2). After 3 years (timepoint 4), significant difference between the groups were only seen for epitopes recognized by IgG, likely due to a lower number of samples collected in the unvaccinated group at this timepoint compared to at timepoint 3.

As the level of IgG against certain epitopes increased during grass pollen AIT, we wanted to examine if the overall profile of epitopes recognized by IgG remained stable for each subject or if AIT introduced some common profile patterns. Hierarchical clustering of epitope profiles against Phl p 5.0101 (Figure 3), Phl p 2.0101, and Phl p 6.0101 (Supplementary Figure 9), showed that IgG and IgG4 peptide recognition were more similar in samples of the same subject collected over time than in samples of different individuals, independently if the patient had been subjected to grass pollen AIT or not, implying that allergen-epitope profiles are highly individual.

Still, some shared patterns could be seen in many of the vaccinated individuals. One epitope (epitope B), found at Phl p 5.0101, was not recognized by IgG or IgG4 from any of the studied individuals at time 1 (Figure 1B). During grass pollen AIT, however, four out of five individuals developed IgG and IgG4 against this epitope. One of these, subject 2, expressed IgE against the same epitope during the entire time course, i.e., even before treatment initiation. For individuals not subjected to AIT containing grass pollen, no antibodies could be identified against this epitope (Figure 1B).

With peptides from 731 different allergens, the microarray platform allows for a detailed study of antibody cross-reactivity. Comparison of antibody binding patterns between allergens from different grass species, belonging to group 2 and 3 or to group 5 and 6, showed that while some linear epitopes were unique for one single grass pollen allergen, many were seen in two or more allergens of the same group (Supplementary Figures 1–3). Epitope B at Phl p 5.0101, for which IgG and IgG4 were commonly developed during AIT, could also be identified in other grass pollen allergens of group 5 and 6, including the isoallergen Phl p 5.0201 (Figure 1C), even though Phl p 5.0101 (and its related variants Phl p 5.0102-Phl p 5.0109) has a two amino acid deletion (Figure 1D) within this region. In comparison with Phl p 5.0101, multiple individuals expressed IgG against this epitope on Phl p 5.0201 already at time 1, suggesting that it is not the same IgG binding to epitope B on these two isoallergens. Indeed, a more detailed study of this region showed that the binding patterns were slightly different between different isoallergens, and that the region more likely carries two, slightly overlapping epitopes: epitope B1 which is found on group 5 and 6 grass pollen allergens with the amino acid deletion and epitope B2 which is found on allergens lacking this deletion (Figures 1A,D; Supplementary Figure 8). Notably, some individuals express antibodies against only one of these epitopes, such as subject 4 (only recognizes epitope B1) and subject 11 (only recognizes epitope B2), while others (e.g., subject 2 and 3) express antibodies against both epitopes. The antibodies recognizing epitope B1 and B2 often also bind homologous allergens from other grass species that contain (Hor v 5.0101 and Sec C 5.0101) or do not contain (Lol p 5.0101, Lol p 5.0102, Hol l 5.0101, Hol l 5.0102, Poa P 5.0101) the two amino acid deletion, respectively (Supplementary Figure 8). A major part of epitope B1 and the entire epitope B2 reside in the flexible linker that connects the allergens' N- and C-terminal domains (Figure 1E).

Phl p 5 has two known isoallergens (Phl p 5.01 and Phl 5.02) that are further divided into nine and seven different known variants (Phl p 5.0101-Phl p 5.0109 and Phl p 5.0201-Phl p 5.0207). Peptides from each of these variants have been included in the peptide microarray, thereby enabling us to examine the antibody binding pattern deviations between the different isoforms. We transformed the peptide signal intensity data into estimated signal intensities for each amino acid, to better follow the effect of small sequence differences on antibody recognition. Differences between variants, even if being limited to only one amino acid residue, often had a clear effect on the binding of antibodies, resulting in either decreased or completely absent binding signals (Supplementary Figures 10, 11). Similar as for epitope B1 and B2, binding patterns varied to a great extent between different individuals, as highlighted by for example epitope C, which covers residue 79 (Supplementary Figure 12) in Phl p 5.0101-Phl p 5.0109 and is situated in one of the alpha helices of the allergens' N- terminal domain (Figure 1E). This position may carry either an alanine or an aspartate amino acid. Antibodies found in subjects 2, 12, and 15 recognize peptides with an aspartate in this position but bind peptides with alanine to a lower degree (with signals too low for passing the cut-off) or not at all. Antibodies found in other individuals (3 and 4) instead prefer peptides with an A79. In one individual (donor 7) IgG signals were only seen for D79 peptides, while IgG4 signals only could be measured for A79 peptides.

We and others (23–25) have previously identified antibodies that represent a so-called stereotyped/public IgE response against the timothy grass pollen allergen Phl p 2. Stereotyped antibodies develop in multiple individuals in response to a particular antigen, in this case Phl p 2, and feature a set of similar, common sequence characteristics. The binding of antibodies with these features has been associated with a discontinuous epitope (30), located mainly within a four-stranded β-sheet of the Phl p 2 allergen, and shown to largely inhibit the binding of IgE from grass pollen-allergic subjects to Phl p 2 (30). Now, we examined the generated microarray peptide data aiming to identify epitopes that are located elsewhere in Phl p 2, and that thus may be of relevance in the IgE crosslinking of FcεRI on effector cells in the presence of dominant stereotyped antibodies. The binding of IgE to linear peptides could only be identify in one individual, but the same epitope (epitope D) was also bound by IgG4 in two subjects (Figures 4A,B). Additionally, two major epitopes (epitope E and F) found in seven and two subjects, respectively, were seen in the IgG data. These two epitopes were found at a distance from both each other and the previously determined epitope of Phl p 2 stereotyped IgE, while epitope D partly overlapped with the conformational epitope (Figure 4C). Additional linear epitopes were seen in group 2 and 3 allergens of other species (Supplementary Figures 1, 2) including some that based on homology to the structure of Phl p 2 reside at a distance from the epitope recognized by stereotyped antibodies (data not shown).

Epitope B1 and B2 on grass pollen allergens belonging to group 5 were situated in the flexible linker of the allergen (Figure 1). We also examined another type of allergens with long, flexible arms; the acidic ribosomal protein P1 and P2. All examined individuals expressed IgG against linear epitopes on these two groups of allergens (Supplementary Figure 13), but no epitopes bound by IgE could be identified (data not shown), and only very limited IgG4 response was seen (Supplementary Figure 14). Much of the IgG reactivity was identified in one distinct region of allergens belonging to acidic ribosomal protein P1 (peptides starting at position 52–65, named epitope G) (Figures 5A,C) and in one distinct region of allergens belonging to acidic ribosomal protein P2 (peptides starting at position 56–74, named epitope H) (Figures 5B,D). Indeed, both these epitopes were found in the flexible arms of the allergen, close to their N-terminal domains (Figure 5E).