Structural Analysis of Der p 1–Antibody Complexes and Comparison with Complexes of Proteins or Peptides with Monoclonal Antibodies

House dust mites are a common source of indoor allergens and a major cause of perennial asthma worldwide (1, 2). House dust mites are common in households around the world and can be found in beds, carpets, and soft furniture. Members of the Dermatophagoides genus feed on dander and small particles of shed skin, which is commonly present in households. Some of their digestive enzymes are potent proteases that are abundant in the feces of dust mites and are highly allergenic. Der p 1 is a cysteine protease and a major allergen (3). Chronic exposure to Der p 1 occurs by inhalation and may lead to the production of IgE Abs in susceptible atopic individuals. Der p 1 catalyzes the cleavage of the amide linkages in substrates such as α1-antitrypsin, the CD23 receptor on human B cells, the IL-2 receptor (CD25) on human T cells, and the Der p 1 propolypeptide sequence (4). Strong evidence suggests that Der p 1–related cleavage of these receptors contributes to its allergenicity (5, 6).

Structures of recombinant Der p 1 in both the proenzyme and mature forms were previously determined (7–9). The structure of natural Der f 1, which shares 81% sequence identity to Der p 1, was also determined (9). Additionally, structures of natural Der f 1 and natural Der p 1 in complex with the Fab fragment of the cross-reactive mAb 4C1 were also elucidated (10). In this study, we present the crystal structures of Der p 1, isolated from its natural source, complexed with the Fab fragment of 5H8 (Der p 1–5H8), Der p 1 complexed with the Fab fragment of 10B9 (Der p 1–10B9), and the Fab fragment of mAb 10B9 alone. Both 10B9 and 5H8 are species specific, whereas the 4C1 Ab is cross-reactive between Der p 1 from Dermatophagoides pteronyssinus and Der f 1 from Dermatophagoides farinae. This enabled the Der p 1 epitopes for mAbs 10B9, 5H8, and 4C1 to be compared with the corresponding surface on Der f 1 (9, 10). It was discovered that the Der p 1 epitopes, which bind 4C1 and 10B9 Abs, overlap and these two Abs compete for the same binding site (11). The 5H8 Ab, however, binds to the epitope located on a different side of Der p 1 and does not compete with 4C1 or 10B9 for binding (11). The binding interfaces of Der p 1 with mAbs 4C1, 5H8, and 10B9 with the binding interfaces of all currently known structures of complexes of proteins or peptides with mAbs were also compared.

Der p 1 was purified from D. pteronyssinus mite culture as described previously for Der f 1 (9, 10). Briefly, Der p 1 was purified from spent D. pteronyssinus mite culture extract (∼ 100 g/1 PBS) using affinity chromatography through a 4C1 mAb column. The mAbs 5H8 and 10B9 were fragmented commercially by GenicBio (Shanghai, China) and Strategic BioSolutions (Newark, DE), respectively. The fragmentation was performed using papain, and the resulting Fabs were purified by protein A affinity chromatography. The Fab from mAb 5H8 was further purified by gel filtration (Sephadex G-75). Both Der p 1–5H8 and Der p 1–10B9 complexes were prepared using the same protocol. In each case, the allergen was mixed with the Fab fragment of Ab in a 1:1 molar ratio and incubated at 4°C for 16 h for Der p 1–10B9 and 30 min for Der p 1–5H8. After incubation, the solution was concentrated using an Amicon Ultra concentrator (Millipore) with a 10,000-Da molecular mass cutoff and purified on a Superdex 200 column attached to an ÄKTA FPLC system (GE Healthcare). A solution composed of 10 mM Tris-HCl and 150 mM NaCl at pH 7.5 was used for gel filtration of both complexes. After gel filtration, fractions containing Der p 1–5H8 and Der p 1–10B9 were concentrated to ∼5 mg/ml. The 10B9 Fab fragment, used for crystallization of the Ab fragment alone, was also purified on a Superdex 200 using 10 mM Tris-HCl, 50 mM NaCl (pH 7.5). Prior to crystallization, the 10B9 Fab fragment was concentrated to 8 mg/ml.

Crystallization of Der p 1–10B9, Der p 1–5H8, and 10B9 was performed at 293 K. Crystals were grown using the hanging drop vapor diffusion method. The crystallization drops were a 1:1 mixture of the protein solution and the precipitant solution from the wells (100 mM MES, 10% [w/v] polyethylene glycol [PEG] 6000, 5% MPD at pH 7.0 for the Der p 1–5H8 complex, 100 mM sodium acetate, 8% [w/v] PEG 4000, 15% MPD at pH 4.5 for the Der p 1–10B9 complex, and in the case of 10B9, 100 mM sodium citrate, 15% [w/v] PEG 6000 at pH 5.5 was used). Prior to data collection, Der p 1–5H8, Der p 1–10B9, and 10B9 crystals were cryoprotected in LV CryoOil solution (MiTeGen, Ithaca, NY), a 1:1 mixture of Paratone-N and mineral oil, or well solution, respectively, then immediately cooled in liquid nitrogen.

Data were collected at the Structural Biology Center Collaborative Access Team 19-BM and 19-ID beamlines (12), as well as at the 21-ID-D beamline of the Life Sciences Collaborative Access Team at the Advanced Photon Source, Argonne National Laboratory. All structures were determined by molecular replacement using HKL-3000 (13), which incorporates MOLREP (14) and some of the programs included in the CCP4 package (15). The structure of the Fab fragment of mAb 10B9 was determined using the Fab fragment of mAb 4C1 (Protein Data Bank [PDB] code 3RVT) as a starting model. Structures of Der p 1 in complex with the Fab fragments of mAb 5H8 and 10B9 were determined utilizing the Der p 1 allergen complexed with the Fab fragment of mAb 4C1 (PDB code 3RVW) (10) as the start model. The sequences of the mAbs were obtained by sequencing reverse-transcribed mRNA isolated from hybridomas producing the mAb (9). The models were refined with COOT (16) and REFMAC5 (17). TLS groups used in the refinement of Der p 1 with the Fab fragment of mAb 10B9 were generated using the TLSMD Web server (18). All three structures were validated with MOLPROBITY (19) and ADIT (20).

Previous results showed that natural Der p 1 binds a calcium ion (10), and the metal bound by Der p 1–10B9 and Der p 1–5H8 complexes was also determined as calcium. Data collection and refinement statistics can be found in Table I. The atomic coordinates and structure factors for all the structures reported in the present study were deposited in the Protein Data Bank (www.rcsb.org) with the following accession codes: 4PP1 (Der p 1–5H8), 4PP2 (Der p 1–10B9), and 4POZ (10B9).

Sequences of the protein structures that are similar to the H and L chains of the Fab fragments of mAb 10B9 and 5H8 were obtained by a Basic Local Alignment Search Tool (BLAST) (21–23) search with an expectation value of 1e⁻¹⁰ against the pdbaa BLAST database (24). Structures of the most similar proteins were downloaded, and STRIDE (25) was used to determine their secondary structures. The Abs containing an α helix in the CDR H3 were superposed in PYMOL (26) and their CDRs were analyzed. The amino acids involved in the formation of hydrogen bonds between proteins were identified with Proteins, Interfaces, Structures and Assemblies (PISA) (27), and a cutoff distance of 3.3 Å for hydrogen bonds was chosen. The salt bridges between Der p 1 and the Fabs of 10B9 and 5H8 were identified by PISA (Gibbs free energy based) and VMD (distance based) (28). A salt bridge was considered to be formed when the distance between any of the oxygen atoms of acidic residues (carboxylate ions; acceptors) and the nitrogen atoms of basic residues (ammonium or guanidinium ions; donors) were within the cut-off distance of 4.0 Å (29–31).

Structures of Fab fragments of mAbs complexed with various macromolecules were obtained by running a PSI–BLAST search against the pdbaa database using the sequences of 5H8 and 10B9. Redundancy was removed and two datasets were created: the first, which contains structures with resolution of 3.0 Å or better, and the second, which contains structures with resolution of 3.5 Å or better. Preliminary analysis of the epitope and paratope content and hydrogen bonds between them were virtually identical for both datasets. All of the analyses were performed simultaneously on both datasets; however, the results from the first dataset, which contains 314 structures of Fab fragments in complex with protein or peptides solved at 3.0 Å resolution or better, were chosen for this study and referred to hereafter as the dataset. PISA was used to estimate the interfaces and hydrogen bonds formed by structures in the dataset. Amino acids were considered to be part of the interface when the calculated buried surface area for a particular amino acid was >10 Ų or was involved in forming hydrogen bonds.

The VMD program was used for structural conservation analysis between Der p 1 and Der f 1. Der p 1 (3F5V), Der f 1 (3D6S), Der f 1–4C1 (3RVV), Der p 1–4C1 (3RVW), Der p 1–10B9 (4PP2), and Der p 1–5H8 (4PP1) were the structures used for comparison.

RING (32) was used as a tool for analyzing residue interaction networks to describe the protein’s three-dimensional structure and the nature of interactions (e.g., hydrogen bonds, van der Waals contacts, π–cation interactions and π–π stacking interactions).

Residues on the protein’s surface have been identified with PYMOL. The pairwise protein sequence identity and similarity were calculated using the EMBOSS (33) package.

The frequencies of amino acids are shown per 100 surface-exposed amino acids found in all protein chains of Fab–protein or Fab–peptide complexes reported in the PDB database (as of 2013). Amino acid frequencies calculated this way were used to estimate the expected amino acid content of the epitopes and paratopes to compare it with the observed amino acid content of the interfaces.

A program was written in python (34) with the use of numpy (35), matplotlib (36), and mmlib (37) libraries to generate three types of plots: 1) bar plots for comparison between observed and expected distribution of the amino acids among epitopes and paratopes, 2) rainbow-colored bar plots showing the relative amount of surface area contributed to the Ag–Ab interface by each type of amino acid, and 3) array plots for displaying the number of hydrogen bonds between epitopes and paratopes as a grayscale. The values of the expected number of amino acids in the interfaces for the bar plots were derived from the frequencies of amino acid occurrences in all protein sequences found in BLAST nr database. The dataset obtained in the first step of the structural analysis was used for plots 2 and 3. The area contributed by a particular amino acid to the interface in the rainbow-colored plot was calculated with PISA. The number of hydrogen bonds in Ag–Ab complexes shown in plot 3 was also calculated with PISA. Some of the code was developed and used in a previous work (10).

Molecular Operating Environment (38) was used to verify the potential hydrogen bonds involved in creating the interfaces between Der p 1 and both Fab fragments of mAb 5H8 and 10B9.

Der p 1 with the Fab fragment of mAb 5H8 and Der p 1 with the Fab fragment of mAb 10B9 were purified, crystallized, and their structures were determined (Table I). Space groups were identified as P2₁2₁2 for the Fab fragment of mAb 10B9 and P2₁ for the Der p 1–5H8 and Der p 1–10B9 complexes. Previously, we were able to determine the crystal structure of Der p 1 complexed with the Fab fragment of mAb 4C1 (10). Fig. 1A shows the relative position of the Fab fragments of 4C1, 10B9, and 5H8 Abs in complex with Der p 1. More than 70% of the 10B9 binding epitope overlaps with the epitope that binds the 4C1 Ab; the epitope binding 5H8 Ab is located on a different part of Der p 1.

The electron density found in close proximity to Asn⁵² in each of the Der p 1 chains was interpreted as N-acetylglucosamine in the structures of both complexes. Additionally, it was discovered that the same, or similar, residue binds N-acetylglucosamine in the structures of group 1 dust mite allergens previously solved, where in the case of Der f 1 N-acetylglucosamine was bound by Asn⁵³ (9, 10). An exception was found in a structure where Asn⁵² was replaced by Gln⁵² (8). The N-acetylglucosamine binding site is far from the 4C1, 10B9, or 5H8 epitopes and does not interfere with the binding of these Abs.

Three well-ordered CDRs per Ab chain are involved in creating the interface areas (Supplemental Fig. 2). According to a relatively recent (2011) classification (39), the conformations of the CDRs in the Fab fragments of mAb 10B9 and 5H8 can be categorized into the same categories as 4C1 (10): CDR L1, L1-11-2; CDR L2, L2-8-1; CDR L3, L3-9-cis7-1; CDR H1, H1-14-1; CDR H2, H2-9-1.

Crystals of Fab 5H8 alone were not obtained. In the structure of the Der p 1–5H8 complex, the epitope binding 5H8 to Der p 1 is located far from the Der p 1’s active site. The buried surface area at the interface of Der p 1 and 5H8 is ∼910 Ų. The interface areas contributed by the H and L chains are ∼75 and 25%, respectively (Table II). Der p 1 epitope binding 5H8 Ab is formed by 19 aa, where 16 residues from Der p 1 interact with 15 residues from the H chain and 6 residues from Der p 1 interact with 7 residues from the L chain of 5H8 Ab. Amino acid composition of the epitope on the surface of Der p 1 (Fig. 1B–G) where the Fab of mAb 5H8 is bound differs significantly from those epitopes where the Fab of mAb 4C1 and 10B9 are bound. Fifteen hydrogen bonds were identified between Der p 1 and 5H8 Ab. Residue Arg⁵¹ of Der p 1 creates hydrogen bonds with both the L chain of V region of Ig (V_L) and H chain of V region of Ig (V_H) of mAb 5H8 (Gly⁹¹ and Asp¹⁰³, respectively). Some of the residues, such as Ser¹¹⁴ with Tyr³² and Tyr⁵⁰ and C-terminal Leu²²² with Lys⁹², form hydrogen bonds only with the V_L, whereas residues Gln⁵³, Ser⁵⁴, Gln⁸⁴, Tyr⁹³, Ala¹⁰⁸, Gln¹⁰⁹, Arg¹¹⁰, Gly¹¹², and Ile¹¹³ form hydrogen bonds only with the V_H of 5H8 Ab (Table II). Supplemental Figs. 3 and 4 demonstrate that hydrogen bond formation is quite specific, where a tyrosine residue from the Ab tends to form hydrogen bonds with an arginine residue from the Ag. This occurrence can be observed at the interfaces between Der p 1–4C1 and Der p 1–5H8, but not in Der p 1–10B9 complex (see Fig. 3 and our previous work in Ref. 10). Arg⁵¹ forms cation–π interactions with Tyr³² of V_L and Tyr¹⁰² of V_H. The amino acids that make up mAb 5H8’s paratope of the 5H8–Der p 1 interface are tyrosine, arginine, aspartate, glutamine, lysine, and a relatively high content of glycine. Phe¹¹¹ of Der p 1 forms π–π stacking interactions with Trp⁵² (edge-to-face orientation) and Tyr¹⁰² (face-to-edge orientation) of V_H, as well as Tyr⁴⁷ of Der p 1. The Fab fragment of mAb 5H8, which is similar to the Fab fragment of mAb 10B9, is also a species-specific Ab. It binds to a different fragment of Der p 1, and the epitope for this Ab does not overlap with the epitopes for the 10B9 and 4C1 Abs.

The epitope for the Fab fragment of mAb 10B9 is also located far from the catalytic site of the enzyme (Cys³⁴). The buried surface area at the interface of Der p 1 and 10B9 is 815 Ų, and the interface areas contributed by the H chain and L chain are ∼75% for the former and 25% for the latter (Table III). The Der p 1 epitope that binds 10B9 is formed by 17 amino acids, where 15 residues from Der p 1 interact with 11 residues from the H chain and 5 residues from Der p 1 interact with 5 residues from the L chain of 10B9.

The uncomplexed Fab fragment of mAb 10B9 crystallized in the orthorhombic space group P2₁2₁2. The overall conformation is almost identical to the Fab fragment of mAb 10B9 that crystallized in complex with Der p 1, but there are conformational differences between the CDRs. A superposition of C_L, C_H, V_L, V_H from the 10B9 crystal and corresponding chains from 10B9 from the complex with Der p 1 have a root mean square deviation of 0.8 Å. CDRs L1 (residues 23–34), L2 (residues 49–56), L3 (residues 88–97), H1 (residues 23–36), and H2 (residues 50–67) have the same side chain conformations and almost identical placement of Cα atoms in both the nonbonded and bound Ab structures. However, CDR H3 (residues 98–110) has a very rare conformation in the binding area, an α helix. A comparison of uncomplexed 10B9 with Der p 1–10B9 complex revealed that the CDR H3 undergoes a few significant conformational changes upon binding to Der p 1. The change is especially well visible for residues Phe¹⁰⁰, Leu¹⁰¹, Thr¹⁰², Thr¹⁰³, and Asn¹⁰⁵, which adopt a well-defined α helical conformation (one and a half turn α helix as shown in Supplemental Fig. 1B). The loop placement during the binding of Der p 1 is shown in Supplemental Fig. 1A. The overall conformation of the Fab fragment of mAb 10B9 is not significantly changed, in comparison with the uncomplexed form, but rather all the CDR regions are shifted upon binding to Der p 1. There are two salt bridges formed by Der p 1 and 10B9. The first is formed by Asp¹⁹⁸ and Arg⁵³ of the V_L and the second is formed between Glu¹³ and Arg⁹⁸ of the V_H. The asymmetric unit contains two copies of Der p 1–10B9 complex; 14 hydrogen bonds have been identified in the first copy and 13 in the second copy of the complex. The hydrogen bonds between Der p 1 and V_L of 10B9 involve the following residues: Gln¹⁸ of Der p 1 and both Gly⁹¹ and Asn⁹² of 10B9 (asparagine forms a hydrogen bond only in one of the copies of Der p 1–10B9 complex), Asp¹⁹⁸ of Der p 1 and Arg⁵³ (as well as Tyr¹⁰⁶ of V_H) of 10B9,and Asn¹⁹⁹ of Der p 1 and Arg⁵³ of 10B9. Arg²⁰ was found to form a hydrogen bond with Ser³⁰ of the V_L, but only in one of the copies of the complex. The hydrogen bonds between Der p 1 and V_H of 10B9 involve, respectively, Ala¹² and Tyr⁵³, Glu¹³ and Arg⁹⁸, Asp¹⁵ and Tyr¹⁰⁷, Asn¹⁷⁹ and both Ser³¹ and Gly³², Gly¹⁸² and Ser³¹, and Asp¹⁹⁸ and Tyr¹⁰⁶ (Table III). Gln¹⁸¹ was found to form a hydrogen bond with Ser³¹ in only one of the copies of the Der p 1–10B9 complex. There is a possibility that two Der p 1 residues form cation–π interactions—Gln¹⁸ with Tyr¹⁰⁷ and Arg¹⁷ with Tyr¹⁰⁶ of the V_H of 10B9—but a higher resolution structure of Der p 1–10B9 complex would be required to verify these interactions.

Even though the epitopes for both 4C1 and 10B9 largely overlap, the relative position of the Abs is not shifted, but rather the mAb 4C1 binds to its epitope in a position that is “rotated counterclockwise” by a little more than 90° with respect to the position assumed by 10B9 upon binding to its epitope on Der p 1 (Fig. 1B, 1C). The residues of Der p 1 that participate in hydrogen bonding with both 10B9 and 4C1 are Glu¹³, Arg¹⁷, Gln¹⁸, and Asp¹⁹⁸. Residues specific to hydrogen bond formation with 4C1 are Arg²⁰ and Arg¹⁵⁶, whereas Ala¹², Asp¹⁵, Gln¹⁸¹, Gly¹⁸², Tyr¹⁸⁵, and Asn¹⁹⁹ participate in hydrogen bonding with 10B9. A surface-exposed tyrosine residue of the Fab fragment of mAb 4C1 forms hydrogen bonds with an arginine residue on the surface of the Ag, which is observed at the Der p 1 and 4C1 interface (Supplemental Figs. 3A, 3B, 4A, 4B as well as our previous work in Ref. 10). These bonds are not observed in the Der p 1–10B9 complex. However, hydrogen bonds are present between a surface tyrosine on the Ab and an aspartate residue on the surface of Der p 1. Furthermore, such hydrogen bonds are not present at the interfaces formed by the Fab fragments of mAb 4C1 or 5H8 and Der p 1. The interface between 10B9 and Der p 1 comprises several hydrogen bonds that are formed between serine and arginine as well as asparagine/glutamine and glycine. In the 4C1 complex, however, the interface is formed between arginine, aspartate, serine, or threonine from 4C1 and tyrosine, arginine, aspartate, glutamine, or glutamate on Der p 1, with the most abundant linkage being between tyrosine from 4C1 and arginine from Der p 1 (figure 3a of Ref. 10).

There is a conformational change of the Der p 1 loop from residue 179 to 185 (root mean square deviation of 0.7 Å, where overall root mean square deviation is 0.6 Å) calculated between Der p 1 complexed with 4C1 and Der p 1 complexed with 10B9. This conformational change, however, is not observed between uncomplexed Der p 1 and the Der p 1–10B9 complex. This loop provides strong interactions with the paratope on the H chain of the Fab fragment of mAb 10B9, but it is on the edge of the epitope and does not form any substantial interactions with the L chain of the Fab fragment of mAb 4C1 (Supplemental Fig. 1A). Therefore, the conformational change might be induced upon binding of water molecules between the 4C1 Ab and Der p 1, which would therefore create bridging interactions between Asn¹⁷⁹, Ser¹⁸⁰, and Tyr¹⁸⁶ with Glu¹⁰⁶ of CDR H3 (as described in Ref. 10). However, this rearrangement is located on a peripheral part of the Der p 1–4C1 interface, and other variables might contribute to the rearrangement.

The interfaces created between Der p 1 and 10B9 or Der p 1 and 5H8 (both species-specific mAbs) have larger contact areas than do the interfaces between Der p 1 and 4C1 and Der f 1 and 4C1 (cross-reactive mAb) (∼910 and ∼845 Ų versus ∼745 and ∼760 Ų, respectively; calculated by PISA). As mentioned previously, both 10B9 and 5H8 are species specific, whereas the 4C1 Ab is cross-reactive between Der p 1 from D. pteronyssinus and Der f 1 from D. farinae. The deletion of serine in the eighth position of Der p 1 causes an offset by one residue in the amino acid numbering between Der p 1 and Der f 1, and thus Der f 1 numbering of amino acid residues is used in comparisons of Der f 1 with Der p 1 for both allergens in this section.

The Q_H algorithm was used as the scoring function on structures superimposed in VMD. The calculated Q_RES values were assigned for every residue, and the surface was marked in such a way that the amino acids with side chains that had the most similar orientation and shape were colored blue through white to red (which indicates that the orientation and shape of the side chains were different) (Fig. 1E–G).

The 4C1 Ab is cross-reactive between two dust mite allergens, Der p 1 and Der f 1, whereas 10B9 is specific for Der p 1. To explain this observation, we compared the residues of Der p 1 that form interactions with 10B9 to corresponding residues from Der f 1. The residues that form the 10B9 binding epitope on Der p 1 are partially conserved in Der f 1. In addition to the surrounding amino acids not directly involved in the interaction with 10B9, the major differences between Der p 1 and Der f 1 in the 10B9 binding epitope include a substitution of residues Ser¹³, Ser¹⁹, Ser¹⁸⁰, Thr¹⁸¹, and Ser²⁰⁰ in Der f 1 by Ala, Gln, Asn, Ala, and Asn, respectively, in Der p 1 (Figs. 1F, 2C), which allows for hydrogen bond formation between Der p 1 and 10B9 (Table III). However, a lack of these hydrogen bonds is not the major factor preventing the binding of 10B9 to Der f 1. A comparison of the available Der p 1 and Der f 1 structures reveals that there is a significant difference in conformation and composition of the 179–183 region (Asn-Ala-Gln-Gly-Val) of Der p 1 and corresponding 180–184 fragment (Ser-Thr-Gln-Gly-Asp) of Der f 1. The different conformations of these regions results in a significant shift of the central Gln position, with corresponding Cα atoms being shifted by ∼3 Å. This shift results in Gln¹⁸¹ in Der p 1 being “parallel” to the allergen’s surface, whereas in the case of Der f 1, this residue is “perpendicular” to the allergen’s surface. A superposition of the Der p 1–10B9 complex and Der f 1 shows that the perpendicular conformation of Gln¹⁸² in Der f 1 results in a steric clash with Ser³¹ from the H chain of the Ab. The density map, which is of good quality around Gln¹⁸¹, allows for the assumption that a small change in the amino acid composition results in different conformations of this region in Der p 1 and Der f 1.

The 5H8 binding epitope area on Der p 1 is mostly conserved in Der f 1 (15 of 24 aa residues), but there are several differences between them (Table II). Residues Thr⁵⁴ and Ser⁹⁴ of Der f 1 are substituted by Gln and Tyr, respectively, in Der p 1, thus making it impossible for Der f 1 to form hydrogen bonds with the Fab fragment of mAb 5H8 (Fig. 2D). Der f 1 amino acid residues Thr⁵⁴, Glu⁹¹, Ser⁹⁴, Ser¹⁰⁹, and Tyr¹¹² are substituted by Gln, Gln, Tyr, Ala, and Gly residues, respectively, in Der p 1, which creates a nonconserved spot in the central part of the 5H8 epitope surface area (white spot in the central part of Fig. 1G) and therefore changes the shape of the surface of the 5H8 binding epitope (Fig. 2D). Some of the amino acids in Der f 1 (Gln⁸⁶, His¹¹¹, and Met²²³) that correspond to the 5H8 binding epitope on Der p 1 are substituted (His, Arg, and Leu), but in the Der p 1–5H8 complex all the interactions are mediated by the main chain atoms for these residues and the side chains do not create sterical clashes. Therefore, these residues should not interfere with the possibility of 5H8 binding to Der f 1.

The methods used to investigate Der p 1 complexes with Abs made it possible to not only analyze these complexes, but all nonredundant complexes of Fab fragments of Abs with proteins or peptides available in the PDB of resolution ≥3.0 Å (n = 314 as of 2013). The particular amino acids found in the protein–Ab or peptide–Ab interfaces are shown in Supplemental Fig. 3G and 3I. However, the observed frequency of the amino acids on the surface of the epitopes, in general, is very similar to the expected frequency of these amino acids on the surface of proteins from the dataset, with charged or polar residues more frequent and hydrophobic residues less frequent. Serine and threonine are the exception, however, with their occurrence being essentially 2-fold less frequent in epitopes than in other parts of the surface of proteins from the dataset. Glutamine residues in the epitopes are approximately one third more frequent than expected. The observed frequency of the amino acids in the paratopes differs significantly from the expected frequency, with tyrosine being 6-fold more frequent and tryptophan being 3-fold more frequent in the paratope interfaces than on the surface of proteins from the dataset (Table IV). The observed value of serine frequency in the paratopes of Fab–protein complexes is almost equal to the expected serine frequency on the protein’s surface (Table IV). This frequency of serine is not observed in the paratopes of the Abs interacting with peptides and is half of the expected value (Supplemental Fig. 3J, Table IV). Among charged amino acids, only arginine and histidine are more frequent than expected in the paratopes of proteins than on the surface. The frequency of lysine is lower than expected for protein–Ab complexes or almost equal to the expected value for peptide–Ab complexes. We have also analyzed amino acids from epitopes and paratopes with respect to their contribution to the interface area; this analysis is detailed in the Supplemental Fig. 4.

Supplemental Fig. 4 shows the area contributed to the interfaces by particular amino acids. The amino acids found in the epitopes of proteins and peptides generally contribute more area to the interfaces than amino acids found in the paratopes. Residues that contribute the most area to the protein–Ab interfaces found in the epitopes are arginine, histidine, lysine, methionine, phenylalanine, and tryptophan, whereas the residues that contribute the most area to the protein–Ab interfaces found in the paratopes are arginine, methionine, phenylalanine, and tryptophan. The amino acids found in the epitopes of the peptide–Ab interfaces contribute even more area to the interface as compared with the protein–Ab interface, but this observation can be explained by the small size of peptides. In the paratopes of the peptide–Ab interfaces, methionine contributes the most area. Curiously, and in contrast with the larger area contribution of methionine in all of these interfaces, the observed frequency of methionine, found in both the epitopes and paratopes, is half the expected frequency of methionine found on the surface of proteins in the dataset, with the exception of methionine found in the peptide epitopes (Table IV).

The interface area of Der p 1 complexed with the Fab fragments of mAb 4C1, 10B9, or 5H8 (Supplemental Figs. 3, 4) is composed of almost every charged or polar amino acid (tyrosine, threonine, arginine, serine, glutamine, aspartate, glutamate, and aspartate). However, hydrophobic amino acids such as leucine and isoleucine are also present. The amino acids that form the interface, which are found in any of the three Abs, include mostly tyrosine, arginine, threonine, and serine. Tyrosine is one of the main amino acids that contribute the most surface area to the interfaces in both the epitopes of Der p 1 and the paratopes of the Abs. The number of tyrosines is approximately 5- (5H8) to 10-fold (10B9 and 4C1) higher in the paratopes than the average tyrosine frequency on protein surfaces of Fab–protein and Fab–peptide complexes (Table IV). Arginine is significantly more abundant in the epitopes of Der p 1 and the paratopes of all three Abs. In the paratope of the Fab fragment of mAb 4C1, arginine is 8-fold more frequent than it is on protein surfaces of Fab–protein and Fab–peptide complexes (Table IV). The interface area on the Der p 1 epitopes contains a significantly higher frequency of arginine, aspartate, glutamate, and glutamine than the general frequency of these amino acids on the surface of the proteins from the dataset. Arginine is 4- to 6-fold more frequent, aspartate is 3- to 4-fold more frequent, glutamate is 2-fold more frequent, and glutamine is 2.5- to 4-fold more frequent than what is expected. Asparagine, despite the physicochemical similarity to glutamine, is only present in the paratope of the 4C1 Ab in twice the expected frequency, and is absent in the 10B9 and 5H8 paratopes. A rather unusual observation, in comparison with other Abs, is that the paratope of the 4C1 Ab contains 8-fold more arginine than the usual arginine frequency on protein surfaces, and it is much higher than the general arginine content in the paratopes, as well as in the paratopes of 10B9 and 5H8 (Table IV).

The hydrogen bonds formed between glycines of the Fab fragment of mAb 5H8 or the Fab fragment of mAb 10B9 and the Der p 1 allergen are quite common in comparison with the hydrogen bonds created between all Abs and all proteins or peptides (Fig. 3). The differences between amino acids involved in hydrogen bonds between the paratopes of the Fab fragments of mAbs 4C1, 10B9, and 5H8 and Der p 1 allergen are significant. Most of the hydrogen bonds in the paratope of 10B9 are formed by serine with arginine, asparagine, glutamine, or glycine and by tyrosine with alanine or aspartate on Der p 1 (Fig. 3A). Hydrogen bonds on the 5H8 paratope are formed by aspartate with arginine, glutamine, or glycine, by glutamine with arginine or asparagine, and by tyrosine with serine or arginine on the surface of Der p 1 (Fig. 3B). Quite rare hydrogen bonding interactions can be observed at the Der p 1 epitope and 5H8 paratope interface. Namely, the hydrogen bonds are created between the main-chain atoms of Ile⁸³ and Ile¹¹³ of Der p 1 and side chain of Arg¹⁰¹ from 5H8. Additionally, Ile¹¹³ from Der p 1 and Asp¹⁰³ from 5H8 also form hydrogen bonds. Surprisingly, there are also not common hydrogen bonds at the interface between Der p 1 and 5H8—formed between glutamines from Der p 1 and arginines and aspartates from 5H8 Ab.

The amino acid composition of the Der p 1 epitopes that bind to the Fab of mAb 10B9, 5H8, and 4C1 does not contain all of the standard amino acids. The amino acids that are missing or underrepresented are mostly nonpolar; however, some polar or charged amino acids may not be present in every epitope. Aspartate and glutamate are not present in the 5H8 binding epitope, but they are very common in the 4C1 and 10B9 binding epitopes. Serine is absent in the 4C1 binding epitope, but it is present in other epitopes. The 4C1 binding epitope contains threonine, which is absent in both 5H8 and 10B9 binding epitopes. Valine and lysine are absent in all of the three epitopes. Alanine is overrepresented in the 4C1 and 10B9 binding epitopes and is present in equal amounts to the expected value in the 5H8 binding epitope on Der p 1. The content of isoleucine, despite being a nonpolar amino acid, is significantly higher than expected in the 4C1 and 5H8 epitopes, but isoleucine is absent in the 10B9 epitope. Even though Der p 1 epitopes that bind 4C1 and 10B9 mostly overlap, their amino acid composition is different. The epitope that binds 10B9 contains glycine and serine in lower than expected quanitities and histidine in higher than the expected amount. However, it does not contain either threonine or isoleucine, both of which are present in the 4C1 binding epitope.

The comparative analysis of interfaces formed by Der p 1 with the Fab fragments of mAb 4C1, 10B9, and 5H8 required an investigation of the content of the interfaces and hydrogen bonds formed by all Fab fragments of Abs complexed with proteins or peptides. Analysis of Ag–Ab interfaces showed (Supplemental Figs. 3, 4) that tyrosine is found >5-fold more often on the surface of the paratopes than on the surface of proteins in general (Table IV). It strongly suggests the important structural role of tyrosine in forming interfaces and its presence in interface areas found in all complexes of Fab fragments of Abs with proteins, peptides, or nucleic acids (unpublished data) and it is consistent with earlier studies (40). Tryptophan, which shares some of the properties of tyrosine, is found to be three times more frequent on the surface of paratopes than on the surface of proteins in general. A higher concentration of aromatic amino acids at the binding interfaces is consistent with a previous report (41). The quantitative analysis of the binding interfaces showed that aromatic amino acids found in paratopes tend to contribute significantly more surface area to the interfaces than do other amino acids found in paratopes, with the exception of methionine. Methionine, despite its very low frequency of occurrences in Ag–Ab interfaces, usually contributes a significant surface area to the interface when present. Serine and threonine, despite having polar side chains, are found to be 2-fold less frequent in the protein–Ab or peptide–Ab interfaces, as in other parts of proteins. The observed frequency of serine in the paratopes, however, is almost equal to the expected value.

Most human IgE Abs (>80%) are directed against cross-reacting epitopes on Der p 1 and Der f 1 (42, 43). The mAb analyzed, including the cross-reactive mAb 4C1 and species-specific mAb 10B9 and 5H8, partially inhibited IgE Ab binding to Der p 1 (11, 42, 44). The three epitopes identified here on Der p 1, by x-ray crystallography, provide the basis for an ongoing analysis of antigenic determinants involved in IgE Ab binding.

The unique α helical loop in CDR H3 of 10B9 Ab, which is able to fit into a cavity on Der p 1’s surface, is formed by a conformational change that occurs upon binding to Der p 1. The α helix in CDR H3 is present in <20 out of ∼1000 (∼2%) Ab structures reported to the PDB. This form of secondary structure is very rare, and its role has been described as inhibitory (it competes for binding site by mimicking the structure of the other protein) (45), as affecting Ag recognition (46, 47), or it can form a tentacle that provides a large interaction surface for Ag binding (47). In almost all other structures of Fab fragments with Abs, the CDR H3 forms a random coil loop. The function of the CDR H3 loop of 10B9 can be explained as affecting Ag recognition and providing a backbone for the formation of the Der p 1-10B9 interface, by increasing the size of the paratope area and participating in forming hydrogen bonds. The interaction between 10B9 and Der p 1 is an example of “induced fit” protein interactions rather than “lock-and-key.” Considering the lock-and-key interaction of Der p 1 with 4C1 Ab described previously (10), this study shows that both types of Ag–Ab interactions can occur even within a similar area on the Ag surface.

The Der p 1 epitope binding 5H8 Ab may be affected by sequence polymorphisms (48) in a few sites (Ala¹⁰⁸ and Ile¹¹³). However, structural analysis reveals that these amino acids are involved in the interface creation by their main-chain atoms, and their variance most likely does not interfere with 5H8 binding. Both 10B9 and 4C1 epitopes do not contain residues that are associated with sequence polymorphisms (48). A comparison of mature pro-rDer p 1–Ab complex structures with the structure of pro–Der p 1 clearly indicates that the presence of the propeptide does not interfere with the binding of 4C1, 10B9, and 5H8 Fab fragments to the allergen.

Structural studies of Der p 1 and Der f 1 allergens allowed us to directly compare the epitopes that bind the Fab fragments of mAb 4C1, 10B9, or 5H8. Such a comparison provides detailed insight, at the molecular level, about which of these allergen surfaces are responsible for the fact that 10B9 and 5H8 Abs are Der p 1 specific, whereas 4C1 is cross-reactive and binds both Der f 1 and Der p 1 (10). The results in the present study indicate which of the Der p 1 amino acids, responsible for binding 10B9 and 5H8 Abs, are different between Der p 1 and Der f 1, rendering the latter impossible to bind 10B9 and 5H8. Residues that were different between 10B9 and 5H8 binding epitopes on Der p 1, as well as the corresponding surface area of Der f 1, are mostly charged and polar. However, there are a few substitutions of nonpolar amino acids such as alanine to small polar amino acids such as threonine and serine that may affect their attractions with the amino acids found in the paratopes of 10B9 and 5H8 Abs and therefore contribute to species specificity. In Der f 1, the number of amino acids that differed from the corresponding residues that form hydrogen bonds between 10B9 or 5H8 and Der p 1 was significant. These substitutions apparently contributed to the lack of binding of these two Abs to the Der f 1 allergen, despite a high sequence identity (∼75% at the surface level) and similarity (∼83% at the surface level) between these two allergens. These results indicate that relatively small differences in the allergen surface may impair Ab binding. For example, such changes may involve the presence of a few nonconserved residues, which were observed in the case of the Der f 1 region that corresponds to the 5H8 epitope in Der p 1. However, these differences may be more subtle as it occurs on the surface of Der f 1, in the region corresponding to the 10B9 epitope in Der p 1. In this case, Der f 1 has smaller residues than the corresponding residues in Der p 1, and it therefore cannot form many binding interactions with 10B9. Additionally, to the “missing” interactions, the conformation of the 180–184 fragment (Ser-Thr-Gln-Gly-Asp) in Der f 1 is not compatible with 10B9 binding, as the main chain of Gln¹⁸² generates a steric clash between the allergen and the Ab.

The amino acid composition of the Der p 1 epitopes for 4C1, 10B9, and 5H8 Abs demonstrates that in general there is little preference for particular amino acids. However, all polar or charged amino acids are much more frequent than those that are nonpolar. The surface of the paratope is built from polar amino acids such as tyrosine or serine, possibly to increase sterical compatibility and to compensate for the epitope’s charge. In contrast to 10B9 and 5H8, the paratope of the 4C1 Ab has a very high arginine content, which, by adding a positive charge to the Der p 1 binding paratope, may partially contribute to its cross-reactivity.

Analysis of Der p 1 and Der f 1 allergen–Ab complexes prompted us to investigate all protein–Ab and peptide–Ab complexes that have their structures determined. We have developed a unique methodology that allows analysis of the properties of epitopes and paratopes as: relative frequency of amino acids, areas of interactions for particular amino acids, and hydrogen bonding interactions. This methodology allows for the generation of plots highlighting diverse aspects of epitopes and paratopes. The analysis of the amino acid composition of the interfaces in the structures deposited to PDB (Table IV, Supplemental Fig. 3) showed that none of the amino acids is significantly overrepresented in the epitopes. However, amino acid composition of the epitopes shows bias toward polar (with the exception of serine and threonine) and charged amino acids. Such bias is observed for both proteins and short peptides. The observed frequency of serine and threonine is half of the expected value. One could expect such results for proteins, taking into account that the residues forming epitopes are on their surface, but this trend is not so obvious for peptide–Ab complexes. The amino acid composition of the paratopes is also highly biased toward polar (except for serine and threonine) and charged amino acids with significant absence of the nonpolar amino acids. It is also striking that tryptophan and especially tyrosine residues are significantly overrepresented in CDRs of the Abs. A special role of the tyrosine was already noticed (40) and most likely could be explained by the chemical nature of this amino acid that allows it to participate in many types of bonding interactions.

The analysis of the Ag–Ab interfaces revealed a preference for certain amino acids that are involved in creating hydrogen bonds between Abs and Ags. The amino acids that are most often involved in hydrogen bond formation on the surface of the paratopes are tyrosine (22%), serine (14%), aspartate (13%), arginine (10%), and asparagine (9%). On the surface of the epitopes, most of the hydrogen bonds are created by arginine (15%), glutamate (11%), aspartate (10%), lysine (10%), serine (9%), and asparagine or glutamine (both are 8% each). A closer look at the surface area of the Ag–Ab interfaces indicates that the number of amino acids is lower for the Ags and higher for the Abs. The relative area contributed to the interfaces by particular amino acids, however, is quite the opposite, being higher for the Ags and lower for the Abs. This is consistent with the fact that, in general, Abs must have the possibility to evolve and quickly adjust to recognize a wide variety of Ags (clonal selection and affinity maturation). Having several residues in the binding site provides greater flexibility to precisely recognize different epitopes. Such variability is not needed at the level of the epitopes for the immune response to be flexible. It may not even be desired in the design of molecules that no longer bind the Ab, as the mutation of just a few residues, which contribute a large surface area, might be enough to prevent undesired recognition by Abs.

There are various methods that currently exist for the prediction of linear (49–51) and conformational (52–55) epitopes. The linear epitope prediction algorithms do not provide enough value, as most of the epitopes in globular proteins are discontinuous or nonlinear and brought together during protein folding (56, 57). The algorithms used in these methods take into account various properties of the epitopes such as the hydrophilicity scale, protein secondary structure, and protein flexibility, but their accuracy is usually ∼80%. There is no epitope prediction algorithm that takes into account the surface contributed to the interface area by particular amino acids. The information obtained from the contribution of surface area and preferences in creating hydrogen bonds between Ags and Abs may be used alongside sequences and structures as an element of creating new epitope prediction tools with very high prediction accuracy.

The determination of the three-dimensional structure of allergen–Ab complexes reveals conformational epitopes, which are the main epitopes in inhalant allergens, such as Der p 1 (58). This approach for mapping Ab epitopes represents a major advantage, compared with previously used strategies, which are based on the identification of Ab-binding peptides that only contain linear epitopes. Therefore, the results of our structural studies of group 1 mite allergens may be used not only to design experiments allowing for the identification of IgE binding epitopes, but also to design recombinant versions of allergens for application in immunotherapy (2).

A.P., J.G., L.D.V., and M.D.C. are employees of Indoor Biotechnologies, Inc., and M.D.C. is also founder and co-owner of the company. The work in the present study was sponsored by a multiple principal investigator R01 grant to Indoor Biotechnologies, Inc. (principal investigators A.P. and M.D.C.) in collaboration with the University of South Carolina (principal investigator M.C.) and the University of Virginia. The remaining authors have no financial conflicts of interest.