Structural, Functional, and Immunological Characterization of Profilin Panallergens Amb a 8, Art v 4, and Bet v 2*

Ragweed allergens affect several million people in the United States and Canada. To date, only two ragweed allergens, Amb t 5 and Amb a 11, have their structures determined and deposited to the Protein Data Bank. Here, we present structures of methylated ragweed allergen Amb a 8, Amb a 8 in the presence of poly(l-proline), and Art v 4 (mugwort allergen). Amb a 8 and Art v 4 are panallergens belonging to the profilin family of proteins. They share significant sequence and structural similarities, which results in cross-recognition by IgE antibodies. Molecular and immunological properties of Amb a 8 and Art v 4 are compared with those of Bet v 2 (birch pollen allergen) as well as with other allergenic profilins. We purified recombinant allergens that are recognized by patient IgE and are highly cross-reactive. It was determined that the analyzed allergens are relatively unstable. Structures of Amb a 8 in complex with poly(l-proline)10 or poly(l-proline)14 are the first structures of the plant profilin in complex with proline-rich peptides. Amb a 8 binds the poly(l-proline) in a mode similar to that observed in human, mouse, and P. falciparum profilin·peptide complexes. However, only some of the residues that form the peptide binding site are conserved.

The designation of "panallergen" usually refers to minor allergens, present in most eukaryotic cells, that are responsible for IgE cross-reactivity to a variety of allergenic sources (1). In addition to being ubiquitous, they share highly conserved sequences and three-dimensional structures, which subsequently satisfies the requirements for cross-recognition by IgE (1). Due to this cross-recognition, panallergens are thought to be indicators of sensitization to several pollens (2,3). Therefore, it is possible that panallergens could be used to predict cross/ poly-sensitization to a variety of pollen allergens (4).
The profilins found in weed pollens are of particular interest due to their increasing presence as sources of seasonal allergies (14,15). Weeds encompass a heterogeneous group of plants that are generally unwanted and invasive; thus, the term "weed" does not represent any particular botanical group or family (16). However, the clinically relevant allergens found in weed pollen, thus far, have been identified in plants belonging to only five different families, one of which is the Asteraceae family. The Asteraceae family includes thousands of plant species, but the important allergenic members are ragweed (Ambrosia), mugwort (Artemisia), feverfew (Parthenium), and sunflower (Helianthus) (15).
In this study, we structurally characterized Amb a 8, a ragweed profilin found in short or common ragweed (Ambrosia artemisiifolia), as well as the mugwort (Artemisia vulgaris) profilin Art v 4. Short ragweed, as well as giant ragweed, acts as a major elicitor of type 1 allergenic reactions in late summer and early fall, affecting over 15 million people in the United States and Canada (15). Amb a 8 has been found to have an 89% amino acid sequence identity with Art v 4 (4). In addition to their sequence identity, both Art v 4 and Amb a 8 have comparable IgE cross-reactivity, suggesting that either allergen could be used as a marker for profilin sensitization in weeds (4). Moreover, we compared molecular and immunological properties of Amb a 8 and Art v 4 with those of Bet v 2, a birch pollen profilin.

Results
Clustering Analysis of the Profilin Family-Sequences of proteins from the Pfam profilin family (Pfam00235), including sequences of the allergens classified in the AllFam database (17) of allergen families (AF051), were retrieved from the Pfam database (18). Clustering of the sequences was carried out to identify groups of similar sequences and to determine the distribution of known allergens and proteins with known structures within the data set. Clustering was performed based on their pairwise BLAST similarity scores (19) using a p value threshold of 1e Ϫ6 . Clustering at more stringent p values gave similar results that are in agreement with the high similarity of profilins from various sources with only a few subgroups clearly separated from the obtained supercluster ( Fig. 1).
Plant profilin sequences, including all of the allergen sequences classified in Allfam (with the exception of human profilin that is classified as an autoallergen), clustered together in a dense group in the central part of the diagram (represented by structures 1G5U (Hav b 8), 1A0K (Arabidopsis thaliana), and 1CQA (Bet v 2)). The remaining part of the supercluster (represented by structures 1ACF (Acanthamoeba castellanii), 1F2K (A. castellanii), 1K0K (Saccharomyces cerevisiae), 3D9Y (Schizosaccharomyces pombe), and 1PRQ (A. castellanii)) is composed of sequences from invertebrates and vertebral profilin 4. Furthermore, vertebral profilin 4 is more similar to invertebrate profilins than it is to vertebral profilins 1-3 (13) because it does not contain the poly(L-proline) binding site and does not interact with actin (20). Less similar sequences from different taxonomic groups, mostly Protists (e.g. Amoebozoa), are located peripherally but with clear connections to the central supercluster.
There are only two groups that are clearly separated. First, sequences of profilins from Alveolata, represented by structures of profilins from Apicomplexan parasites Toxoplasma gondii (PDB code 3NEC) 2 and Plasmodium falciparum (PDB code 2JKF), formed a subgroup clearly separated but still closely connected to the central supercluster (Alveolata). A second separate cluster is formed by human profilins 1, 2, and 3 and the corresponding profilins from other Vertebrata (Vertebrata 1-3) (represented by structures 2V8C (Mus musculus), 2VK3 (Rattus norvegicus), 1AWI (Homo sapiens), 1PNE (Bos taurus), and 1D1J (H. sapiens)) with a very closely connected but clearly distinguishable subgroup of sequences from viruses.
Structural Analysis of Methylated Amb a 8 -Full-length, native Amb a 8.0101 (UNP Q2KN24) is composed of 133 residues with a molecular mass of 14,245 Da. The construct created for these studies contained a start methionine followed by a His 6 tag, tobacco etch virus cleavage site, and spacer region of residues GSG. The start methionine of Amb a 8 was removed in the construct, and the N-terminal region that is visible in the electron density is GSG followed by a serine, which corresponds to residue 2 of Amb a 8. Residue Ϫ2, the first glycine of the tag linker, to residue 133, a methionine, are visible in the electron density with average B factors of 11.9 Å 2 . The overall fold of Amb a 8 (Fig. 2) is similar to that of other profilins in that it is composed of two terminal ␣-helices, one short ␣-helix, and a ␤ hairpin that sandwich a central five-stranded antiparallel ␤ sheet. Amb a 8, which includes eight lysine residues, was subjected to reductive methylation, which was confirmed via MALDI TOF/TOF (Fig. 3). Electron density reveals that seven (residues 37, 45, 73, 88, 89, 97, and 98) of eight lysine residues were dimethylated. The eighth lysine (residue 54) has a disordered side chain that cannot be modeled. Amb a 8 also contains three cysteine residues, two of which participate in a disulfide bond (residues 95 and 117), and the third cysteine has clear electron density, indicating that it has been modified by ␤-mercaptoethanol (residue 13). Additionally, it appears that benzoic acid is bound to Amb a 8 at the surface of the molecule with an average B factor of 14.9 Å 2 . The source of the benzoic acid is unknown; however, its placement is based entirely on the electron density, which unquestionably fits that molecule.
Structural Analysis of Amb a 8 Bound with Poly(L-Proline) 14 -Unlike methylated Amb a 8, native Amb a 8 in the presence of poly(L-proline) 14 crystallized in the P2 1 space group and contains two subunits in the asymmetric unit cell. Clear electron density is visible for residues Ϫ1 to 133 in chain A and residues 0 -133 in chain B. The poly(L-proline) 14 chains that are bound to chains A and B contain 9 ordered proline residues. Both chain A and chain B have a cysteine residue that has been modified by ␤-mercaptoethanol at position 13, like the one that is seen for methylated Amb a 8; however, chain A has electron density for two positions of the modified cysteine residue. Chains A and B differ from one another in that chain A has a disulfide bridge between residues 95 and 117, like that of the methylated Amb a 8, whereas no disulfide bridge exists between 95 and 117 of chain B. Both chains A and B help stabilize the poly(L-proline) chains through interactions with the side chain residues from Trp-3, Tyr-6, His-10, Trp-35, and Tyr-127.
Structural Analysis of Amb a 8 Bound with Poly(L-Proline) 10 -Native Amb a 8 in the presence of poly(L-proline) 10 also crystallized in the P2 1 space group; however, it contains one subunit per asymmetric unit cell, unlike that of Amb a 8⅐poly(L-proline) 14 . Clear electron density is visible for residues 2-133 in chain A as well as all 10 proline residues of poly(L-proline) 10 . Similar to that of the Amb a 8⅐poly(L-proline) 14 structure, the Amb a 8⅐poly(L-proline) 10 structure also contains a disulfide link between Cys-95 and Cys-117. A superposition of Amb a 8⅐poly(L-proline) 10 on Amb a 8⅐poly(L-proline) 14 revealed an RMSD value of 0.46 Å. There are no significant differences between the Amb a 8⅐poly(L-proline) 14 and the Amb a 8⅐poly(Lproline) 10 structures. One minor difference between the two structures is the position of Gly-17. In the Amb a 8⅐poly(L-proline) 14 structure, Gly-17 is positioned toward the surface of the subunit, whereas in the Amb a 8⅐poly(L-proline) 10 structure, Gly-17 is 3.7 Å closer to the core of the protein. Due to the overall similarity and an RMSD value of 0.5 Å, for the remainder of the study, we will refer to the Amb a 8⅐poly(L-proline) 14 structure for analysis.
Structural Analysis of Art v 4 -Full-length, native Art v 4.0101 (UNP Q8H2C9) is also composed of 133 residues but has a molecular mass of 14,207 Da. Modifications made to the N  terminus of Amb a 8, described above, were also applied to the construct of Art v 4. The first visible residue of Art v 4 is serine of the spacer region, followed by a glycine of the spacer region and then serine, which corresponds to residue 2 of native Art v 4. Art v 4 has the same overall fold as other profilins, with a total of three ␣-helices and a seven-stranded antiparallel ␤ sheet. Residues Ϫ1, serine of the tag linker, to 133, a methionine, are visible in the electron density with average B factors of 17.9 Å 2 . Art v 4 also contains three cysteine residues, two of which form a disulfide bond (Cys-95 and -117), and one cysteine, like that of Amb a 8, that is modified by ␤-mercaptoethanol. Cys-95 is modeled at two positions with 117; however, it is apparent that Cys-95 may be modeled at a third position that does not allow for the formation of a disulfide bond. Due to software restrictions, this third conformation of Cys-95 has been omitted. Furthermore, the structure of Art v 4 contains one molecule of HEPES from the crystallization condition that is located at the surface of the protein.
Comparison of Amb a 8 and Art v 4-A superposition of Art v 4 on methylated Amb a 8 results in an RMSD of 0.6 Å. Overall, the two profilin structures superpose with very few structural differences. A slight difference between the two structures is located at the N terminus. Modified lysine residues 45, 88, and 89 of methylated Amb a 8 (PDB code 5EM1) have positions different from the lysine residues in the Art v 4 structure. Additionally, residue 89 has visible electron density for modeling at two positions. Another difference between Amb a 8 and Art v 4 is ligand location. Whereas both benzoic acid of methylated Amb a 8 and HEPES of Art v 4 bind at the surface of their respective structures, benzoic acid is located near residues 45 and 46 of methylated Amb a 8, whereas HEPES is located closest to residues 3 and 6 of Art v 4.
Comparison of Methylated Amb a 8 and Amb a 8 Bound with Poly(L-Proline) 14 -A superposition of chain A from methylated Amb a 8 on chain A from Amb a 8 with bound poly(L-proline) 14 resulted in an RMSD value of 0.7 Å. The largest difference is seen at the N terminus; this difference is due to the binding of the poly(L-proline). The N terminus of methylated Amb a 8 is located toward the surface of the molecule, whereas the N terminus of the Amb a 8 with bound poly(L-proline) is swung out to allow the proline chain to bind. The same changes are seen when comparing chain A from methylated Amb a 8 with chain B from Amb a 8 with bound poly(L-proline) 14 .
Structurally Similar Profilins-A Dali search was performed using the structures of either methylated Amb a 8 or Art v 4 as search models to determine what proteins are structurally similar to both profilins. Both searches, using methylated Amb a 8 or Art v 4 as the query, resulted in practically identical lists of similar structures. Also, since Amb a 8 and Art v 4 are so similar, all superpositions of related structures were performed using Amb a 8 as the reference structure. Profilin I from A. thaliana (PDB code 1A0K) was the most structurally related profilin that superposed on Amb a 8 with an RMSD value of 0.9 Å, over 130 aligned C␣ atoms, and a sequence identity of 63%. Aside from A. thaliana profilin I, profilins that are registered allergens from latex (Hev b 8, PDB code 1G5U), peanut (Ara h 5, PDB code 4ESP), and birch (Bet v 2, PDB code 1CQA) were the next most structurally related proteins to both Amb a 8 and Art v 4; Heb v 8 superposed on Amb a 8 with an RMSD value of 1.0 Å, over 129 aligned C␣ atoms, and a sequence identity of 69%; Ara h 5 superposed on Amb a 8 with an RMSD value of 1.0 Å, over 124 aligned C␣ atoms, and a sequence identity of 63%; Bet v 2 superposed on Amb a 8 with an RMSD value of 1.1 Å, over 117 aligned C␣ atoms, and a sequence identity of 74%.

Comparison of Mugwort Profilin Art v 4, Ragweed Profilin Amb a 8, and Birch Pollen Profilin
Bet v 2-Birch pollen profilin (Bet v 2) was compared with the plant profilins described in this study. A superposition of 1CQA on Art v 4 revealed an RMSD value of 1.1 Å, over 118 aligned C␣ atoms and a sequence identity of 77%, whereas a superposition of 1CQA on Amb a 8 revealed an RMSD value of 1.08 Å, over 117 aligned C␣ atoms, and a sequence identity of 74%. The region with the largest structural differences is located at the N-terminal ␣-helix (residues 1-12); both Art v 4 and Amb a 8 have the ␣-helix located toward the core of the protein, whereas the ␣-helix of Bet v 2 is swung out toward the periphery of the protein (Fig. 2). Another major difference between the weed pollen profilins and the tree pollen profilin described here is that Bet v 2 is missing residues 13-21, whereas Art v 4 and Amb a 8 have clear electron density for these residues. The remainder of the Bet v 2 protein is similar to Art v 4 and Amb a 8 with minor shifts located from residue 41 to 62, which includes a loop-␣-helix-loop motif, a loop region consisting of residues 68 -73, and the C-terminal ␣-helix consisting of residues 125-133.
Comparison of Plant Profilins and Profilins from H. sapiens-Profilin II from humans (PDB code 1D1J) was compared with the plant profilins described here; a superposition of 1D1J on Amb a 8 (PDB ID 5EM1) revealed an RMSD value of 1.80 Å, over 122 aligned C␣ atoms, and a sequence identity of 23%. The areas with the largest differences are located at the loop regions between ␣1 and ␤1, ␤4, and ␤5, and ␤5 and ␤6. Aside from profilin II from humans, we also compared plant profilins with human platelet profilin (PDB codes 1AWI and 1CF0) and human profilin-1 (PDB codes 3CHW and 2PAV) (Fig. 4). All four of the aforementioned structures contain polyproline peptides that are bound to the profilin molecule. A superposition of each individual human profilin on Amb a 8 with bound poly(Lproline) 14 (PDB code 5EV0) reveals RMSD values in the range of 1.7-1.8 Å over 122 aligned C␣ atoms and sequence identities all around 24%. Superposition revealed that all proline-rich peptides are located at the surface of the profilin in the same location as the poly(L-prolines) that bound to Amb a 8.
Profilin Stability Determined Using Differential Scanning Fluorimetry-Thermal stability of Amb a 8, Art v 4, and Bet v 2 in the presence of L-proline or poly(L-proline) peptides was checked using differential scanning fluorimetry. For proline screening (Fig. 5), melting temperatures (T m ) were compared with protein stability in gel filtration buffer (10 mM Tris, pH 7.4, 150 mM NaCl) with no proline or proline oligomers. h-Amb a 8 was most stable in the presence of Pro 6 and Pro 10 (T m increase of 5°C) and did not show decreased stability in any of the conditions. Amb a 8 was most stable in the presence of Pro 10 (T m increase of 5°C) and least stable with Pro 14 (T m decrease of 1°C). h-Art v 4 was most stable in the presence of Pro 10 (T m increase of 2°C) and least stable in the presence of L-proline and Pro 6 (T m decrease of 8°C). Art v 4 was most stable in the pres-ence of Pro 10 and Pro 14 (T m increase of 4°C) and least stable in the presence of Pro 6 (T m decrease of 1°C). h-Bet v 2 was most stable in the presence of L-proline (T m increase of 8°C) and did not show decreased stability in any of the conditions. Bet v 2 did not show an increase in stability in any of the conditions and was most unstable in the presence of Pro 6 and Pro 10 (T m decrease of 3°C). In addition to L-proline or poly(L-proline) peptide screening, profilin stability was also measured in different concentrations of urea: 0, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0, 5.0, 6.0, or 8.0 M urea (Fig. 6). In the 0 M urea condition, protein was mixed 1:1 with gel filtration buffer (10 mM Tris, pH 7.4, 150 mM NaCl), which was used as the control. . Furthermore, profilin stability was also measured in different urea concentrations at pH 7.5; however, no significant differences were observed between urea concentrations that did not have their pH adjusted and concentrations that were adjusted to pH 7.5 (data not shown).
Immunologic Properties of Amb a 8, Art v 4, and Bet v 2-Sera from patients allergic to seasonal allergens, including trees, grasses, and weeds, were screened for the presence of anti-Art v 4-specific IgE antibody using ELISA. Sera from three patients with high IgE immunoreactivity to Art v 4 were selected for the comparative studies. All three sera contained IgE reacting with all three profilins: Art v 4, Amb a 8, and Bet v 2. The intensity of IgE reactivity to Art v 4 was more similar to that to Amb a 8 than Bet v 2. (Fig. 7 and Table 1). This was supported by an ELISA  14 . Amb a 8 residues are shown as yellow sticks, whereas poly(L-Pro) is shown in a ball-and-stick representation and labeled using one-letter codes. B, superposition of profilin complexes with proline-rich peptides. Only side chains of profilin residues involved in binding of proline-rich peptides are shown. The superposition involves Amb a 8 (yellow sticks), human profilin (cyan sticks; PDB entry 3CHW), mouse profilin (gray sticks; PDB entry 2V8F), and profilin from P. falciparum (purple sticks; PDB entry 2JKG). C, superposition of profilin complexes with proline-rich peptides. Only poly(L-Pro) and proline-rich peptide are shown. Poly(L-Pro) from the Amb a 8⅐poly(L-Pro) 14 complex is shown in ball-and-stick representation.  inhibition assay using soluble Art v 4 or Bet v 2 as an inhibitory allergen, and both were used at a concentration allowing for maximum inhibition (Fig. 8 and Table 2). Application of soluble Art v 4 completely abolished IgE binding to all three allergens tested. Although the application of soluble Bet v 2 completely abolished IgE binding to Bet v 2, it was able to inhibit IgE binding to Art v 4 or Amb a 8 in 19-49% only. No significant inhibition using an irrelevant allergen (Der p 1) could be demonstrated.

Discussion
Analysis of profilins' sequences revealed several distinct profilin groups (Fig. 1). The major cluster of sequences may be divided into two subclusters, and one of them contains all plant profilins, including all currently registered profilin allergens. Taking into account the high sequence similarity of plant profilins, it is not surprising that these allergens are highly crossreactive (14,(21)(22)(23). However, it is quite surprising that currently the only source of allergenic profilins is from plants. One could expect that at least some proteins that form the second part of the major cluster (i.e. invertebrates and vertebral profilin 4 due to their relatively high sequence similarity to allergenic plant profilins) should also be identified as allergens. Our clustering results are in agreement with previous studies of evolutionary origins of profilins showing that profilins 4 are most similar to invertebrate profilins and originated before vertebrate evolution (13). The proteins that were investigated here (Amb a 8.0101, Art v 4.0101, and Bet v 2.0101) have relatively low levels of sequence identity and similarity when compared with human profilins, and this fact correlates well with the presence of a separate cluster containing human proteins (Vertebrata 1-3) and other vertebrate profilins 1, 2, and 3. In terms of sequence, the plant profilins are also clearly distinct from profilins originating from Alveolata, represented, for example, by profilins from important human pathogens like T. gondii and P. falciparum. Although the profilins are often considered to be exclusively eukaryotic proteins, there is a group of profilins that originate from viruses. These proteins were shown to be similar to profilins 3 (13), and our results also reveal a distinct cluster of viral proteins next to the cluster Vertebrata 1-3.    Amb a 8.0101 and Art v 4.0101 have almost 90% sequence identity, and their structures are very similar as well (Fig. 2). Although crystallization of Art v 4.0101 did not require any protein modification, crystallization of Amb a 8 required methylation of lysine side chains or the addition of poly(L-proline). Comparison of crystal structures of Amb a 8, Art v 4, and Bet v 2 revealed a significant difference in conformation of the N-terminal fragment of Bet v 2 relative to the same fragments in the weed profilins. Such a difference may be caused by crystal packing; however, it is also possible that it illustrates a conformational change that is important for protein function. The second possibility is quite likely because the N-terminal part of the profilin participates in binding of proline-rich peptides. Besides Bet v 2 and Amb a 8 and Art v 4 (reported here), there are only two other allergenic profilins that have their structures determined: Ara h 5 and Hev b 8. All of these allergens have similar structures and sequences, which further explains high crossreactivity between proteins belonging to this group of allergens.
The structures reported here of Amb a 8 in complex with poly(L-Pro) 10 or poly(L-Pro) 14 are the first structures of a plant profilin in complex with a proline-rich peptide. Previously reported profilin⅐proline-rich peptide complexes were formed by human, mouse, or P. falciparum proteins (Fig. 4). Analysis of the available complexes shows that the peptide binding site may be divided into two regions. In Amb a 8, the first region is formed by N-terminal residues (Trp-3, Tyr-6, and Trp-35), and the second region is formed by the C-terminal part of the protein (Gln-101, Tyr-127, Gln-131, and Met-133). Although the N-terminal region is highly conserved (Fig. 4B), with Trp-3 and Tyr-6 being present in all currently available complexes, the C-terminal regions are more variable. Additional insight into the binding of proline-rich peptides may be obtained by a comparison of structures of profilins in free and complexed forms. For example, the Hev b 8 molecule has the same conformation of the N-terminal helix as in the case of Amb a 8 and Art v 4; however, the conformation of Ara h 5 closely resembles an "open" conformation, as is observed in the Bet v 2 structure. Closer inspection of the two possible conformations observed for allergenic profilins clearly indicated that although a "closed" conformation of Amb a 8, Art v 4, and Hev b 8 is compatible with the formation of interactions necessary for the binding of proline-rich peptides, the "open" conformation most likely cannot mediate formation of the complex with the peptide. This is related to the fact that the conserved residues, Trp-3 and Tyr-6, are displaced by 6 -8 Å (as measured by the change of C␣ positions) in the "open" conformation and therefore are not able to participate in the formation of the groove that binds the proline-rich peptide. Such a groove has to have a shape compatible with the characteristic shape of the type II helix formed by the polyproline fragments. In fact, despite significant differences in the amino acid composition of the peptide binding site, the central part of the peptide is bound in a very similar mode by the ragweed, human, mouse, and P. falciparum profilins.
Despite the fact that profilins are widely distributed in pollens and food, most often they are minor allergens that cause relatively mild reactions. These mild reactions are mainly correlated to the low stability of profilins that are susceptible to denaturation in conditions used during food processing/ preparation. Only profilins that do not undergo high temperature treatment, like those in raw fruits, are considered to be the major allergens. Because the profilins studied in this paper originate from, and are present in, plant pollens, we decided to test their stability. The stability of Amb a 8, Art v 4 a, and Bet v 2 was studied using differential scanning fluorimetry (DSF), and we observed the influence of various compounds on the thermal stability of the proteins. Because our proteins were produced in a recombinant form, we also decided to check the influence of the His tag on protein stability. In all cases, the presence or absence of the purification tag affected protein stability; however, the effects depended on the protein tested (Figs. 5 and 6). Generally, polyproline peptides stabilized or, at least, did not destabilize Amb a 8 and Art v 4. The effect of the peptides on stability of h-Bet v 2 were less pronounced; however, Bet v 2 was clearly destabilized by L-proline and the poly(L-peptides). It is worth noting that the presence of free proline was detected in pollen of Asteraceae species (24). All proteins denatured in 1.5 M urea, which indicates that all proteins are quite unstable in the presence of this compound. Interestingly, h-Amb a 8, Amb a 8, Art v 4, and h-Bet v 2 were mildly stabilized by 0.5 M urea, and Bet v 2 was marginally stabilized by 1.0 M urea. Comparison of the recombinant profilins shows that h-Amb a 8 and h-Bet v 2 seem to be more stable than h-Art v 4 in the presence of L-proline or poly(L-proline), whereas Bet v 2 tends to be less stable than Amb a 8 or Art v 4. The latter observation may be explained by the fact that both Amb a 8 and Art v 4 may be additionally stabilized by a disulfide bridge formed by Cys-95 and Cys-117 (Amb a 8 numbering). There is no such bond in Bet v 2.0101 because the Cys-95 is replaced by Thr. Both Cys residues are conserved in Amb a 8.0101, Amb a 8.0102, Art v 4.0101, and Art v 4.0201. However, during the analysis of the crystal structures reported here, it was noticed that the formation of the disulfide linkage is not observed in all molecules. It is possible that it is caused by the E. coli strain used for protein production and that the formation of the disulfide bridges is not complete or that the disulfide bridges are easily damaged during the diffraction experiment (25).
The relatively low stability, as well as insufficient information on folding of profilins, may pose another problem. Namely, many researchers during purification of natural or recombinant profilins use a method that combines an affinity chromatography (binding to polyproline) that is later followed by elution with a solution of concentrated urea and refolding of the denatured profilin (7,26). However, only rarely is the refolding followed by use of CD spectroscopy or other experimental techniques that may confirm that the protein is properly folded (27). In most cases, the researchers simply assume that the refolding was successful and use the protein for immunological studies (14, 22, 28 -34). The use of unfolded protein with destroyed conformational epitopes may lead to errors, for example, in the antigenic characterization of proteins.
Profilins were identified as novel allergens over 25 years ago, and the first profilin that was shown to be an allergen is Bet v 2 (birch pollen allergen) (35). Currently, there are almost 50 profilins recognized as allergens, and they represent one of the largest groups of panallergens (1). There is still an ongoing discussion of whether profilins are relevant allergens from the Structure of Amb a 8 and Art v 4 JULY (1, 22, 36 -38). Currently, it is clear that some profilins may be not only relevant pollen allergens but also important food allergens in some cases. For example, Cit s 2 (profilin from orange) and Cuc m 2 (profilin from melon) are relevant and major food allergens (31,39). In the case of Cit s 2 and Cuc m 2, the fact that oranges and melons most often are consumed in a raw form may be a very important factor because the allergens do not undergo denaturation before contact with a sensitized individual. The same is most likely also true in the case of profilins found in pollens; despite relatively poor stability in comparison with other food allergens, Amb a 8, Art v 4, and Bet v 2 may be intact upon entry into the human respiratory system.
The high level of sequence identity of allergenic profilins and common co-sensitization or co-recognition make the investigation of the role of individual profilins in induction of sensitization and triggering allergic symptoms difficult. In addition, studies of the molecular basis of profilin allergenicity are addi-tionally complicated by the fact that profilins form complexes not only with actin, but also with many other proteins by binding proline-rich peptides (6, 40 -42). The residues participating in the formation of protein⅐protein complexes are highly conserved, as is illustrated in Fig. 9A. Fig. 9B indicates residues that are identical between Amb a 8, Art v 4, and Bet v 2, which are also highly conserved among protein⅐protein complexes. For Cuc m 2 and Hel a 2 (profilin from sunflower pollen), it was experimentally demonstrated that the IgE biding epitopes overlap significantly with areas of the molecules participating in formation of the protein⅐protein complexes (Fig. 9, C and D) (43,44). The residues forming these epitopes in Cuc m 2 and Hel a 2 are also highly conserved in Amb a 8, Art v 4, and Bet v 2. The presence of IgE epitopes E3 (Cuc m 2) and 10A4 (Hel a 2) is somewhat surprising because the residues participating in the binding of proline-rich peptides are also conserved in human profilins (Figs. 8 and 9). However, it is possible that the conserved region only partially overlaps with IgE epitopes. FIGURE 9. A, sequence conservation of profilins calculated with ConSurf is mapped on the surface of Amb a 8. Actin (pink) was modeled using the structure of human profilin 1 in complex with actin (PDB entry 3CHW). Poly(L-Pro) is shown in a stick representation with carbon atoms marked in yellow. The figure shows three different orientations of the molecules corresponding to 0, 90, and 180°rotation along the y axis. B, residues that are identical in Amb a 8, Art v 4, and Bet v 2 are shown in blue. Residues that are not 100% conserved in these allergens are marked in gray. C, epitopes identified in Cuc m 2. The epitopes are labeled using a convention used in the original paper reporting these epitopes (44,74). D, epitopes identified in Hel a 2.0101. The epitopes are labeled using a convention used in the original paper reporting these epitopes (43).
The recombinant profilins were recognized by IgE from some allergic patients, which is consistent with previous reports (21,27,45,46). Several studies have demonstrated that IgE from patients sensitized to one profilin binds to profilins from all other sources (40,41). In many cases, detection of IgE directed to one plant profilin indicates sensitization to other plant profilins (47,48). However, other studies indicated only partial IgE cross-reactivity between profilins from different sources, and in some cases, sensitization to one profilin is not associated with sensitization to profilins from other sources (49,50). This may be related to structural differences between individual profilins and may also depend on the repertoire of IgE directed to species-specific epitopes of individual profilins (27).
Detected in the current study, differences in IgE binding between Art v 4, Amb a 8, and Bet v 2 indicate the existence of unique, species-specific epitopes in those profilins. Moreover, phylogenetically distant species, such as mugwort and birch, seem to contain more species-specific IgE-binding epitopes than closely related species, such as mugwort and ragweed. The inhibition of IgE binding to Art v 4 or Amb a 8 by Bet v 2 is less efficient than inhibition of IgE binding to nPhl p 12 (native Timothy pollen profilin) that was reported previously (27). Interestingly, the tested sera were collected from patients living in the area where mugwort and birch, but not ragweed, pollens are present. Therefore, it can be argued that similar intensity of IgE reactivity to Art v 4 and Amb a 8 does not reflect similar environmental exposure to those allergens but rather true cross-reactivity related to structural similarities between Art v 4 and Amb a 8 proteins. This is consistent with epidemiological studies that indicate that only a single allergenic source, probably that providing the highest profilin exposure, may be responsible for triggering IgE response to profilins (23). Moreover, IgE cross-reactivity to ragweed allergen Amb a 1 of sera derived from mugwort allergic patients living in the area with no ragweed exposure has been reported (46). In summary, our study provides evidence for the existence of a specific IgE response to well characterized profilins Art v 4 and Amb a 8 and demonstrates the differences in IgE binding between weed profilins and tree profilins, which are related to the structural differences of those proteins.

Experimental Procedures
Cloning, Expression, and Purification-Recombinant Amb a 8.0101, Art v 4.0101, and Bet v 2.0101 were ordered from DNA 2.0 (Menlo Park, CA) in expression vector pJexpress411 with kanamycin resistance. The recombinant DNA was designed with a cleavable polyhistidine tag (MHHHHHHSSGVDL-GTENLYFQS2GSG, where the arrow marks a tobacco etch virus cleavage site) for ease of purification and cleavability to aid in crystallization. This plasmid was then transformed into E. coli strain BL21 (DE3). Cells were grown in LB broth at 37°C to an A 600 of 0.6 -0.8 and then induced with isopropyl ␤-D-1thiogalactopyranoside to a final concentration of 400 M. Following induction, the cells were incubated overnight with shaking at 16°C. Cells were harvested by centrifugation for 30 min at 10,500 ϫ g at 4°C and then resuspended in 50 mM Tris, pH 7.4, 500 mM NaCl, 10 mM imidazole, 2% glycerol, and 20 mM ␤-mer-captoethanol (lysis buffer) supplemented with protease inhibitor tablets (Pierce) used according to the manufacturer's directions and then disrupted by sonication using a Branson Sonifier 450 sonicator. Homogenate was centrifuged for 30 min at 10,500 ϫ g, and a clear, yellow supernatant was separated from the pelleted cellular debris. The crude extract was then loaded onto a 12 ϫ 1.5-cm, 5-ml bed volume column of nickel-nitrilotriacetic acid-agarose medium (Pierce) previously equilibrated in wash buffer (50 mM Tris, pH 7.4, 500 mM NaCl, 30 mM imidazole, 2% glycerol, and 20 mM ␤-mercaptoethanol). Following a washing step with wash buffer, protein was eluted from the medium using elution buffer (50 mM Tris, pH 7.4, 500 mM NaCl, 250 mM imidazole, 2% glycerol, and 20 mM ␤-mercaptoethanol). Fractions were analyzed by SDS-PAGE, and those fractions containing protein were pooled and dialyzed against 10 mM Tris, pH 7.4, 150 mM NaCl, and 5 mM ␤-mercaptoethanol overnight at 4°C. Protein was concentrated using an Amicon Ultra concentrator (Millipore) with a 10,000 Da molecular mass cut-off and then purified on a Superdex 200 column attached to an ÄKTA Pure FPLC system (GE Healthcare). A solution composed of 10 mM Tris-HCl and 150 mM NaCl at pH 7.4 was used for gel filtration of both proteins. Fractions containing protein were pooled, and the concentration was determined using the Bradford method (51).
His tag cleavage was achieved by supplementing pure protein (1-2 mg/ml) with tobacco etch virus protease in a 1:100 (w/w) protease/protein ratio and then dialyzed overnight at 4°C. In this work, recombinant proteins with a His tag attached are referred to as h-profilin (i.e. h-Amb a 8, h-Art v 4, h-Bet v 2), and recombinant proteins with the His tag cleaved off are referred to with the name of the profilin (i.e. Amb a 8, Art v 4, and Bet v 2, respectively). Following dialysis, the protein-tobacco etch virus solution was loaded onto nickel-nitrilotriacetic acidagarose medium equilibrated in wash buffer and allowed to incubate with the medium for 30 min at 4°C. Subsequent to incubation, cleaved protein was eluted with wash buffer, concentrated, and then loaded onto a Superdex 200 column attached to an ÄKTA Pure FPLC gel filtration system (GE Healthcare) in a buffer containing 10 mM Tris-HCl and 150 mM NaCl at pH 7.4. Fractions containing cleaved protein were pooled and concentrated to ϳ10 mg/ml. For studies of profilin bound to poly(L-proline), each profilin was titrated with L-Pro 14 or L-Pro 10 (NeoBiolabs, Cambridge, MA) in a 1:15 ratio at 30°C with stirring for 5 h.
Crystallization, Data Collection, and Processing-Crystallization experiments for Amb a 8, Amb a 8⅐poly(L-proline) 14 after titration, Amb a 8⅐poly(L-proline) 10 after titration, and Art v 4 were performed at 293 K using the sitting-drop vapor diffusion method and MRC 2-drop 96-well crystallization plates (Molecular Dimensions, Altamonte Springs, FL). A solution of recombinant, His tag-free protein was mixed with well solution (0.1 M HEPES, pH 6.5-8.5, and 0.5-1.44 M sodium citrate) in a 1:1 ratio. Crystals of native Art v 4 appeared within 48 h; however, no native Amb a 8 crystals appeared. Only after subjecting native Amb a 8 to reductive methylation, according to previous protocols (52,53), or titrating Amb a 8 with poly(L-proline) 14 or poly(L-proline) 10 did crystallization occur. The level of methylation was confirmed using a Bruker Ultraflex II MALDI TOF/ Structure of Amb a 8 and Art v 4 JULY 22, 2016 • VOLUME 291 • NUMBER 30

JOURNAL OF BIOLOGICAL CHEMISTRY 15455
TOF mass spectrometer (Billerica, MA) (Fig. 3). Crystals were cryo-protected with well solution followed by immediate cryocooling in liquid N 2 . Data for Amb a 8, Amb a 8⅐poly(L-proline) 14 , Amb a 8⅐poly(L-proline) 10 , and Art v 4 were collected from single crystals at 100 K at the Southeast Regional Collaborative Access Team (SER-CAT) 22BM, SER-CAT 22ID, Structural Biology Center 19BM (54) and the Life Sciences Collaborative Access Team (LS-CAT) 21-ID-G beamlines, respectively, at the Advanced Photon Source, Argonne National Laboratory (Argonne, IL). Data were processed with the HKL-2000 software package (55). Methylated Amb a 8 crystallized in the primitive orthorhombic space group P2 1 2 1 2 1 , Amb a 8 with bound poly(L-proline) 14 and poly(L-proline) 10 crystallized in the primitive monoclinic space group P2 1 , and Art v 4 crystallized in the primitive orthorhombic space group of P2 1 2 1 2. Data collection statistics are reported in Table 3.
Structure Determination and Refinement-Both Amb a 8 and Art v 4 were solved by molecular replacement using models of each respective protein generated from SWISS-MODEL (56) using birch pollen profilin (PDB code 1CQA) as a template. The structure of Amb a 8⅐polyproline was solved by molecular replacement using the native structure determined here as a start model. Molecular replacement was performed with HKL-3000 (57) integrated with MOLREP (58) and selected programs from the CCP4 package (59). Models were rebuilt using COOT (60) and refined with REFMAC (61). TLS was used in the final stages of refinement, and TLS groups were determined using the TLSMD server (62). MOLPROBITY (63) and ADIT (64) were used for structure validation.
DSF-Enzyme stability was observed following the Thermo-Fluor method (65), using a CFX96 Touch TM real-time PCR detection system (Bio-Rad) (66). Sypro Orange (excitation, 280 and 450 nm; emission, 610 nm) in DMSO (Sigma-Aldrich) was added to the profilins (in 10 mM Tris, 150 mM NaCl, pH 7.4) at 1.0 mg/ml in a 1:1000 ratio. Protein (with His tag and without His tag) was screened against 10.0 mM L-proline, poly(L-proline) 6 , poly(L-proline) 10 , or poly(L-proline) 14 and against a urea gradient from 0 to 8.0 M urea. The screening condition and protein were mixed in a 1:1 ratio in a 96-well hard shell PCR plate (Bio-Rad). The plates were sealed with Microseal "B" adhesive seals (Bio-Rad) and heated in a CFX96 real-time PCR instrument (Bio-Rad) from 30 to 90°C in 2.0°C increments for 1 min. Proline and polyproline screening was run in triplicate, and urea screening was run in duplicate.
Sequence Similarity-based Clustering-The sequences of Profilin family proteins representing Pfam00235 were downloaded from the Pfam database. To visualize pairwise similarities between profilins and identify groups of more closely related sequences, we used CLANS (cluster analysis of sequences), a Java utility that applies a version of the Fruchterman-Reingold graph layout algorithm (67). CLANS uses the p values of highly scoring segment pairs obtained from an N ϫ N BLAST search to compute attractive and repulsive forces between each sequence pair in a user-defined data set. A two-dimensional representation of sequence subgroups was achieved by randomly seeding the sequences in the arbitrary distance space and then moving them within this environment according to the force vectors resulting from all pairwise interactions; the process was repeated until convergence.
Other Computational Calculations-The SSM algorithm (68) in COOT was used to superpose protein models. All figures containing protein structures were prepared with PyMOL (69) Crystallographic data and refinement statistics for Amb a 8, Amb a 8⅐poly(L-Pro) 14 , Amb a 8⅐poly(L-Pro) 10

, and Art v 4
Values in parenthesis refer to the highest resolution shell. AU, asymmetric unit.
Amb a 8 Amb a 8 ϩ poly (L-Pro) 14 Amb a 8 ϩ poly (L-Pro) 10 Art v 4 or Chimera (70). PDBePISA was used to calculate probable quaternary structure (71). The Dali server (72) was used to identify structurally similar proteins. The ConSurf server was used to map sequence conservation on the structure of Amb a 8 (73). ELISA-Fifty mugwort, grass, and/or birch allergic patients participated in a study approved by the Local Bioethics Committee (R-I-002/36/2015), and all patients signed the informed consent. Their sera were screened for IgE reactivity to recombinant Art v 4. The patients selected for the experiment (n ϭ 3) were allergic to mugwort allergen Art v 4, and none of the studied patients were allergic to ragweed by skin prick testing.

Data collection
Ninety-six-well MaxiSorp microtiter plates (Nunc, Roskilde, Denmark) were coated with 100 l of Art v 4, Amb a 8, or Bet v 2, each at a final concentration of 10 g/ml diluted in coating buffer (25 mM carbonate/bicarbonate buffer, pH 9.4), and incubated overnight at 4°C. After washing with TBST buffer, the plates were blocked with TBST ϩ 1% BSA for 1 h at room temperature. Subsequently, the plates were washed in TBST, and serial dilutions of patients' sera in TBST ϩ 1% BSA (1:5, 1:25, 1:125, and 1:625) were loaded on the plates and incubated for 2 h at room temperature. The plates were thoroughly washed in TBST, and a monoclonal anti-human IgE antibody conjugated to alkaline phosphatase diluted 1:1000 in TBST ϩ 1% BSA (BD Pharmingen, Heidelberg, Germany) was applied and incubated for 2 h at room temperature. Following washing, bound antibodies were detected by incubation with p-nitrophenyl phosphate (Sigma Fast p-nitrophenyl phosphate tablet sets, Sigma-Aldrich). The reaction was quantified by measuring color intensity at 405 nm. To evaluate specificity of the test in all experiments, negative controls were used: no allergen immobilized on the plate, no serum applied, irrelevant allergen immobilized on the plate, or no anti-IgE added. All samples were run in triplicate, and the mean value was used for analysis. The cut-off for positive values was determined using OD values of 10 non-allergic patients' sera and calculated from their mean OD plus three S.D. values.
For the ELISA inhibition assay, in addition to regular ELISA performed as described above, serial dilutions of patients' sera were preincubated for 30 min at room temperature with an allergen dissolved in TBST ϩ 1% BSA at concentrations up to 30 g/ml. The concentration of soluble allergens was well above the minimal concentration necessary for maximum inhibition as determined by preliminary dose-response studies (not shown). The samples were then loaded on a plate, and the procedure was continued as described for the ELISA above.