Characteristic motifs for families of allergenic proteins

Abstract

The identification of potential allergenic proteins is usually done by scanning a database of allergenic proteins and locating known allergens with a high sequence similarity. However, there is no universally accepted cut-off value for sequence similarity to indicate potential IgE cross-reactivity. Further, overall sequence similarity may be less important than discrete areas of similarity in proteins with homologous structure. To identify such areas, we first classified all allergens and their subdomains in the Structural Database of Allergenic Proteins (SDAP, http://fermi.utmb.edu/SDAP/) to their closest protein families as defined in Pfam, and identified conserved physicochemical property motifs characteristic of each group of sequences. Allergens populate only a small subset of all known Pfam families, as all allergenic proteins in SDAP could be grouped to only 130 (of 9318 total) Pfams, and 31 families contain more than four allergens. Conserved physicochemical property motifs for the aligned sequences of the most populated Pfam families were identified with the PCPMer program suite and catalogued in the webserver Motif-Mate (http://born.utmb.edu/motifmate/summary.php). We also determined specific motifs for allergenic members of a family that could distinguish them from non-allergenic ones. These allergen specific motifs should be most useful in database searches for potential allergens. We found that sequence motifs unique to the allergens in three families (seed storage proteins, Bet v 1, and tropomyosin) overlap with known IgE epitopes, thus providing evidence that our motif based approach can be used to assess the potential allergenicity of novel proteins.

1. Introduction

The possibility that proteins from novel foods, drugs, or genetically modified organism may exhibit cross-reactivity with known allergens is of utmost concern to regulatory agencies, food scientists and physicians (WHO, 2003). Due to these considerations, it is important to be able to distinguish allergenic from nonallergenic proteins, and to predict potential IgE cross-reactivities (Aalberse, 2007; Breiteneder and Mills, 2006; Schein et al., 2007). Potential cross-reactive allergens often have very similar sequences (Aalberse and Stadler, 2006; Bonds et al., 2008). Thus, one of the first questions in determining potential cross-reactive foods is the degree of similarity between allergens. Allergens are referred to by names assigned by the Allergen Nomenclature Sub-Committee of the International Union of Immunological Societies (IUIS, www.allergen.org), based on the species/genus name of the source and the order they were identified (Chapman et al., 2007). This nomenclature system is independent of the biochemical and structural nature of the protein, and the names do not readily identify structural and sequence-based relationships among allergens. This means that, based on these names, one cannot easily identify the individual allergenic proteins in different organisms that could account for IgE cross-reactivity (Aalberse et al., 2001; Jenkins et al., 2005; Mirza et al., 2000; Schwietz et al., 2000).

Bioinformatics approaches and allergenic databases are now well established to identify molecular similarities of proteins as an explanation for clinically observed cross-reactivity from very different sources (Breiteneder and Mills, 2006; Brusic and Petrovsky, 2003; Furmonaviciene et al., 2005; Hileman et al., 2002; Schein et al., 2007; Thomas et al., 2005; Zorzet et al., 2004). The Structural Database of Allergenic Proteins (SDAP) (Ivanciuc et al., 2002, 2003) contains many sequence search tools that are seamlessly integrated in the design of the database. SDAP is user friendly and freely available on the Web to allergy researchers, food scientists and industrial engineers (http://fermi.utmb.edu/SDAP/). Allergy researchers can use SDAP primarily to determine food sources that might contain cross-reacting antigens. Regulators and industrial researchers can use the site tools to perform FASTA searches (Pearson, 1994) of allergenic proteins or sequence searches according to the WHO guidelines (Schein et al., 2006). FASTA searches are also helpful in clustering related allergens or suggesting the appropriate nomenclature for novel allergenic proteins. For example, cross-reactions in individuals allergic to the birch pollen allergen Bet v 1 with several fruits are a well-documented example of the pollen-food syndrome (Egger et al., 2006; Mittag et al., 2005), with symptoms ranging from local oral allergy syndrome to severe anaphylaxis. A FASTA search in SDAP quickly reveals that Bet v 1 has significant homology to the food allergens Pru av 1 from cherry (Bit score 160/Evalue 5.9e-35), Gly m 4 from soybean (Bit score 158/Evalue 3.1e-25) and Ara h 8 from peanut (Bit score 102/Evalue 4.3e-24) (Mittag et al., 2006), which could account for the cross-reactions. Pollen cross-reactivity may extend across a large number of species, and even to species from different continents (Midoro-Horiuti et al., 1999, 2003). Similar cross-reactivities among allergens with a high degree of identity have been observed for profilins, lipid transfer proteins, calcium-binding proteins, and pathogenesis-related proteins (Breiteneder and Mills, 2005b; Egger et al., 2006; Midoro-Horiuti et al., 2001; Weber, 2005). Other examples include the ficus-fruit syndrome related to the similarity of cysteine proteases in tropical fruits (Hemmer et al., 2004) or the IgE-based cross-reactivity of shrimp with other crustaceans and even non-edible arthropods such as cockroaches or dust mites due to the similarity of the muscle protein tropomyosin in these organisms (Ayuso et al., 2002a; Reese et al., 2006).

However, simple sequence similarity is not sufficient to conclusively predict IgE cross-reactivity. While short sequence elements can define an IgE epitope, short stretches of identical sequences are not long enough to predict with statistical significance cross-reactive IgE epitopes (Goodman, 2006; Ladics et al., 2006). The statistical significance can be substantially increased if the sequence is a motif that is common to many related known allergens, and is not found in related proteins that are non-allergens. Here, we define specific sequence regions with common physicochemical properties, PCP-motifs (Ivanciuc et al., 2004; Mathura et al., 2003) that may distinguish allergenic proteins.

Our work was predicated on previous studies which indicated that pollen and plant food allergens grouped to only a small number of all protein families (Breiteneder and Radauer, 2004; Jenkins et al., 2005; Radauer and Breiteneder, 2006); most of these families also contain non-allergenic proteins as well. The first step was to obtain a comprehensive assignment of all known allergens according to an existing classification scheme for protein families, Pfam (Version 22.0, http://pfam.sanger.ac.uk/)(Finn et al., 2006). These assignments have been made available on our SDAP web site. The major allergens belong to about 30 structural families, consistent with the results of others (Radauer et al., 2008). In order to discriminate the allergenic from the non-allergenic family members (Björklund et al., 2005; Brusic and Petrovsky, 2003; Furmonaviciene et al., 2005; Riaz et al., 2005; Schein et al., 2005a, 2007; Stadler and Stadler, 2003), we also determined common sequence motifs using our PCPMer program (Schein et al., 2005b,c). We show in three examples that motifs we defined as characteristic of allergens in a given Pfam coincided with previously determined IgE epitopes. The motifs thus represent a promising way to identify linear IgE epitopes that are likely to be responsible for IgE cross-reactivities. All sequence motifs for the major Pfam families with allergens can be obtained from our web server MotifMate (http://born.utmb.edu/motifmate/summary.php). The motifs can now be analyzed in screening sequence databases for potential IgE cross-reactivities (Aalberse, 2007; Hileman et al., 2002; Marti et al., 2007; Riaz et al., 2005; Saha and Raghava, 2006; Schein et al., 2007), or used in conjunction with 3D structural information on allergens to shed new light on the molecular determinants of allergenicity(Aalberse and Stadler, 2006; Breiteneder and Mills, 2006; Chapman et al., 2007; Jenkins et al., 2005; Oezguen et al., 2008).

2. Methods

2.1. Assignment of Pfam domains to all allergens

All allergen sequences from SDAP were searched in the Pfam A (Version 22.0, http://pfam.sanger.ac.uk/) (Finn et al., 2006) database for the matching family. Whenever the TrEMBL or SwissProt accession number of the allergen sequence was known, the Pfam assignment was made based on the corresponding accession number. Otherwise we performed BLAST searches to find related proteins to the SDAP allergen entry. The Pfam database has a collection of sequence alignments of related protein domains that were used to find Pfam domains for each allergen. Fragments of sequences without a significant match in Pfam where left unassigned. As a result of a direct match or individual BLAST searches, 594 out of 829 allergen protein sequences were grouped to their respective protein families and domains from Pfam A.

2.2. Generation of sequence motifs of allergens by MotifMate

MotifMate-PCP is a novel database and data mining tool developed by us to generate physicochemical property (PCP) motifs of allergens. PCP motifs were generated by our PCPMer web server (http://landau.utmb.edu:8080/WebPCPMer/). The motifs are based on the conservation of five physicochemical descriptors E₁-E₅ (Venkatarajan and Braun, 2001) in a multiple sequence alignment. The E₁-E₅ scale allows us to characterize motifs as protein regions where the side chains show conserved physicochemical properties, such as hydrophobicity, size or alpha-helical propensity, rather than strict sequence identity. We have tested the PCPMer motifs in other protein families to locate functional important regions and as meaningful fingerprints to find distantly related proteins (Mathura et al., 2003; Schein et al., 2002, 2005b,c).

We generated two types of motifs: one set of motifs that represent a complete Pfam family containing allergenic proteins, i.e. these are motifs generated from the multiple sequence alignment as archived in the Pfam database, and a second set of motifs using only the allergenic proteins in a family (prepared using ClustalW). Using Perl scripts, multiple sequence alignments of all Pfam families containing allergens were downloaded to a MySQL database, PCP motifs were generated and stored in the MySQL database. Sequence alignments of only allergenic proteins in a Pfam family were manually generated with ClustalW (Chenna et al., 2003). In that phase the protein sequences were cut to the region of the known Pfam domains. In addition, the allergen proteins for each family group were submitted to a pair-wise sequence search in SDAP to eliminate almost identical proteins or protein sequences from the same allergen source. Also, protein sequences with a sequence identity of only 20% or below to the other allergens from that group were eliminated.

3. Results

3.1. Main Pfam classes for allergens

The allergens in SDAP group to only 130 of the 9318 protein families from Pfam A, and of these 31 contain multiple allergenic proteins (Table 1). A list of the allergens in the 12 Pfam families most populated with allergens is given in Table 2. The complete classification of allergens is available on our SDAP web server (http://fermi.utmb.edu/SDAP/). For each family, we determined motifs that were common to all members, and, using separate alignments of the known allergens, those motifs that were unique to allergenic proteins.

Table 1.

The most abundant Pfam A allergen families from SDAP

No	Pfam code	Pfam domain	No allergens
1	PF00234	Protease inhibitor/seed storage/LTP family	34
2	PF00235	Profilin	27
3	PF00036	EF hand	23
4	PF01357	Pollen allergen	19
5	PF00188	SCP-like extracellular protein	19
6	PF00407	Pathogenesis-related protein Bet v 1 family	16
7	PF00261	Tropomyosin	16
8	PF00190	Cupin	15
9	PF00061	Lipocalin/cytosolic fatty-acid binding protein family	12
10	PF03330	Rare lipoprotein A (RlpA)-like double-psi beta-barrel	12
11	PF00042	Globin	9
12	PF00544	Pectate lyase	9
13	PF00112	Papain family cysteine protease	8
14	PF00428	60s Acidic ribosomal protein	8
15	PF00082	Subtilase family	7
16	PF00314	Thaumatin family	7
17	PF01190	Pollen proteins Ole e 1 family	7
18	PF01620	Ribonuclease (pollen allergen)	7
19	PF00012	Hsp70 protein	6
20	PF00578	AhpC/TSA family	6
21	PF02221	ML domain	6
22	PF05922	Subtilisin N-terminal region	6
23	PF00089	Trypsin	5
24	PF00113	Enolase, C-terminal TIM barrel domain	5
25	PF00187	Chitin recognition protein	5
26	PF00273	Serum albumin family	5
27	PF03952	Enolase, N-terminal domain	5
28	PF00151	Lipase	4
29	PF00197	Trypsin and protease inhibitor	4
30	PF00295	Glycosyl hydrolases family 28	4
31	PF01630	Hyaluronidase	4

Table 2.

Classification of allergens in the 12 Pfam families most populated with allergens

Allergen	Source	Allergen	Source	Allergen	Source
PF00234: protease inhibitor/seed storage/LTP family
Amb a 6	Short ragweed	Ana o 3	Cashew nut	Ara h 2	Peanut
Ara h 6	Peanut	Ber e 1	Brazil nut	Bra j 1	Oriental mustard
Bra n 1	Rapeseed	Cor a 8	Hazelnut	Fag e 8kD	Common buckwheat
Gly m 1	soybean	Hev b 12	Rubber (latex)	Hor v 1	Barley
Hor v 21	Barley	Jug n 1	Black walnut	Jug r 1	English walnut
Lyc e 3	Tomato	Mal d 3	Apple	Ory s TAI	Rice
Par j 1	Parietaria judaica	Par j 2	Parietaria judaica	Pru ar 3	Apricot
Pru av 3	Sweet cherry	Pru d 3	European plum	Pru p 3	Peach
Pyr c 3	Pear	Ric c 1	Castor bean	Ses i 1	Sesame
Ses i 2	Sesame	Sin a 1	Yellow mustard	Tri a gliadin	Wheat
Tri a glutenin	Wheat	Tri a TAI	Wheat	Vit v 1	Grape
Zea m 14	Corn
PF00235: profilin
Ana c 1	Pineapple	Api g 4	Celery	Ara h 5	Peanut
Ara t 8	Mouse-ear cress	Bet v 2	Birch	Cap a 2	Bell pepper
Che a 2	Lamb's-quarters	Cor a 2	Hazelnut	Cuc m 2	Muskmelon
Cyn d 12	Bermuda grass	Dau c 4	Carrot	Gly m 3	Soybean
Hel a 2	Sunflower	Hev b 8	Rubber (latex)	Lit c 1	Litchi
Lyc e 1	Tomato	Mal d 4	Apple	Mer a 1	Mercurialis annua
Mus xp 1	Banana	Ole e 2	Olive	Par j 3	Parietaria judaica
Phi p 11	Timothy	Phl p 12	Timothy	Pru av 4	Sweet cherry
Pru p 4	Peach	Pyr c 4	Pear	Tri a profilin	Wheat
PF00036: EF hand
Aln g 4	Alder	Bet v 3	Birch	Bet v 4	Birch
Bos d 3	Domestic cattle	Bra n 1	Rapeseed	Bra n 2	Rapeseed
Bra r 1	Turnip	Che a 3	Lamb's-quarters	Cyn d 7	Bermuda grass
Cyp c 1	Common carp	Gad c 1	Cod	Gad m 1	Atlantic cod
Hom s 4	Human autoallergen	Jun o 4	Prickly juniper	Ole e 3	Olive
Ole e 8	Olive	Phl p 7	Timothy	Ran e 1	Edible frog
Ran e 2	Edible frog	Sal s 1	Atlantic salmon	Sco j 1	Chub mackerel
Syr v 3	Lilac	The c 1	Alaska pollock
PF01357: pollen allergen
Ara t expansin	Mouse-ear cress	Cyn d 1	Bermuda grass	Cyn d 15	Bermuda grass
Cyn d 2	Bermuda grass	Dac g 2	Orchard grass	Dac g 3	Orchard grass
Gly m 2	Soybean	Hol l 1	Velvet grass	Lol p 1	Rye grass
Lol p 2	Rye grass	Lol p 3	Rye grass	Ory s 1	Rice
Pha a 1	Canary grass	Phl p 1	Timothy	Phl p 2	Timothy
Poa p a	Kentucky blue grass	Tri a 3	Wheat	Tri a ps93	Wheat
Zea m 1	corn
PF00188: SCP-like extracellular protein
Cte f 2	Cat flea	Dol a 5	Yellow hornet	Dol m 5	White face hornet
Pol a 5	Wasp	Pol d 5	Mediterranean paper wasp	Pol e 5	Paper wasp
Pol f 5	Golden paper wasp	Pol g 5	Wasp	Sol i 3	Fire ant
Sol r 3	Black fire ant	Ves f 5	Hybrid yellowjacket	Ves g 5	German wasp
Ves m 5	Eastern yellowjacket	Ves p 5	Western yellowjacket	Ves s 5	Southern yellowjacket
Ves v 5	Yellowjacket	Ves vi 5	Yellowjacket	Vesp c 5	European hornet
Vesp m 5	Giant asian hornet
PF00407: pathogenesis-related protein Bet v 1 family
Aln g 1	Alder	Api g 1	Celery	Ara h 8	Peanut
Bet v 1	Birch	Car b 1	Hornbeam	Cas s 1	Chestnut
Cor a 1	Hazelnut	Dau c 1	Carrot	Gly m 4	Soybean
Mal d 1	Apple	Pet c PR10	Parsley	Pha v 1	Kidney bean
Pru ar 1	Apricot	Pru av 1	Sweet cherry	Pyr c 1	Pear
Tar o RAP	Common dandelion
PF00261: tropomyosin
Ani s 3	Herring worm	Cha f 1	crab	Chi k 10	Midge
Cra g 1	Pacific oyster	Der p 10	European house dust mite	Hal d 1	Abalone
Hel as 1	Brown garden snail	Hom a 1	American lobster	Lep d 10	Storage mite
Lep s 1	Silverfish	Met e 1	Greasyback shrimp	Mim n 1	Scallop
Pan s 1	Spiny lobster	Pen a 1	Brown shrimp	Per a 7	American cockroach
Per v 1	Tropical green mussel
PF00190: cupin
Ana o 1	Cashew	Ara h 1	Peanut	Ara h 3	Peanut
Ara h 4	Peanut	Ber e 2	Brazil nut	Cor a 11	Hazelnut
Cor a 9	Hazelnut	Fag e 1	Common buckwheat	Gly m Bd28K	Soybean
Gly m conglycinin	Soybean	Gly m glycinin G1	Soybean	Gly m glycinin G2	Soybean
Jug n 2	Black walnut	Jug r 2	English walnut	Ses i 3	Sesame
PF00061: lipocalin/cytosolic fatty-acid binding protein family
Aca s 13	Flour mite	Blo t 13	Dust mite	Bos d 2	Domestic cattle
Bos d 5	Domestic cattle	Can f 1	Dog	Can f 2	Dog
Equ c 1	Domestic horse	Fel d 4	Cat	Lep d 13	Storage mite
Mus m 1	Mouse	Rat n 1	Rat	Tyr p 13	Mould mite
PF03330: rare lipoprotein A (RlpA)-like double-psi beta-barrel
Ara t expansin	Mouse-ear cress	Cyn d 1	Bermuda grass	Gly m 2	Soybean
Hol l 1	Velvet grass	Lol p 1	Rye grass	Ory s 1	Rice
Pha a 1	Canary grass	Phl p 1	Timothy	Poa p a	Kentucky blue grass
Tri a ps93	Wheat	Zea m 1	Corn
PF00042: globin
Chi t 1	Midge	Chi t 2	Midge	Chi t 3	Midge
Chi t 4	Midge	Chi t 5	Midge	Chi t 6	Midge
Chi t 7	Midge	Chi t 8	Midge	Chi t 9	Midge
PF00544: pectate lyase
Amb a 1	Short ragweed	Amb a 2	Short ragweed	Cha o 1	Japanese cypress
Cry j 1	Japanese cedar	Cup a 1	Arizona cypress	Cup s 1	Common cypress
Jun a 1	Texas mountain cedar	Jun o 1	Prickly juniper	Jun v 1	Eastern red cedar

3.1.1. PF00234: protease inhibitor/seed storage/LTP family

This domain (InterPro IPR003612) is found in plant lipid transfer proteins, seed storage proteins, and trypsin-alpha amylase inhibitors. The domain forms a four-helical bundle in a right-handed superhelix with a folded leaf topology, which is stabilized by disulfide bonds, and which has an internal cavity. Allergens from the lipid transfer protein (LTP) family are highly resistant to both heat treatment and proteolytic digestion, and are particularly important in the Mediterranean area (Breiteneder and Mills, 2005a; Salcedo et al., 2004). Three-dimensional structures are known for three allergens from this family, namely Pru p 3 (2ALG, Fig. 1A), Hor v 1 (1JTB), Zea m 14 (1MZM). The molecular determinants of allergenicity for this family may be extracted from the known IgE epitopes, for Ara h 2 (Stanley et al., 1997), Jug r 1 (Robotham et al., 2002), Par j 1 (Asturias et al., 2003), and Par j 2 (Asturias et al., 2003). The T-cell epitopes are known only for Ara h 2 (Glaspole et al., 2005).

Fig. 1.

PDB structures for allergens from the most abundant Pfam families: (A) Pru p 3 (PF00234, protease inhibitor/seed storage/LTP family; 2ALG); (B) Hev b 8, (PF00235, profilin; 1G5U); (C) Bet v 4, (PF00036, EF hand; 1H4B); (D) Phl p 2, (PF01357, pollen allergen; 1WHO); (E) Ves v 5, (PF00188, SCP-like extracellular protein; 1QNX); (F) Bet v 1, (PF00407, pathogenesis-related protein Bet v 1 family; 1BV1); (G) Bos d 5, (PF00061, lipocalin/cytosolic fatty-acid binding protein family; 1GXA); (H) Phl p 1, (PF03330, rare lipoprotein A (RlpA)-like double-psi beta-barrel; 1N10); (I) Chi t 1, (PF00042, globin; 1ECO); (J) Jun a 1, (PF00544, pectate lyase; 1PXZ).

3.1.2. PF00235: profilin

Profilin (InterPro IPR002097) binds to monomeric actin in a 1:1 ratio and prevents the polymerization of actin into filaments. Three-dimensional structures for allergens in this class are available for Ara t 8 (3NUL), Bet v 2 (1CQA), and Hev b 8 (1G5U, Fig. 1B).

3.1.3. PF00036: EF hand

This family collects calcium-binding proteins (InterPro IPR002048) that contain a common domain known as the EF-hand. The EF-hand motif has a 12 residue loop flanked on both side by a twelve residue alpha-helical domain. The proteins from this class may be signaling proteins (calmodulin, troponin C) or buffering/transport proteins (calbindin D9k). PDB structures are available for Bet v 4 (1H4B, Fig. 1C), Che a 3 (1PMZ), and Phl p 7 (1K9U).

3.1.4. PF01357: pollen allergen

This family (InterPro IPR007117, Pollen allergen/expansin, C-terminal) contains expansins, proteins that mediate cell wall extension in plants. Expansins allow wall polymers to slide by breaking hydrogen bonds that keel together the wall constituents. Grass pollen allergens are the main allergens from this family (Table 2). PDB structures are available for Phl p 1 (1N10) and Phl p 2 (1WHO, Fig. 1D).

3.1.5. PF00188: SCP-like extracellular protein

This family (InterPro IPR001283, Allergen V5/Tpx-1 related) includes venom antigen 5 from wasps (Dol a 5 from the yellow hornet Dolichovespula arenaria, Dol m 5 from the white face hornet Dolichovespula maculata, Pol a 5 from the paper wasp Polistes annularies, Pol d 5 from the Mediterranean paper wasp Polistes dominulus, Pol e 5 from the paper wasp Polistes exclamans, Pol f 5 from the paper wasp Polistes fuscatus, Pol g 5 from the paper wasp Polistes gallicus, Ves f 5 from the downy yellowjacket Vespula flavopilosa, Ves g 5 from the German yellowjacket Vespula germanica, Ves m 5 from the Eastern yellow jacket Vespula maculifrons, Ves p 5 from the Western yellowjacket Vespula pennsylvanica, Ves s 5 from the Southern yellowjacket Vespula squamosa, Ves v 5 from the common yellowjacket Vespula vulgaris, Ves vi 5 from the wasp Vespula vidua, Vesp c 5 from the European hornet Vespa crabo, Vesp m 5 from the giant asian hornet Vespa mandarina) and venom antigen 3 (Sol i 3 from the fire ant Solenopsis invicta and Sol r 3 from the black imported fire ant Solenopsis richteri), which both are potent allergens that mediate allergic reactions to insect stings of the Hymenoptera family. The structure (1QNX, Fig. 1E) and T-cell epitopes of Ves v 5(Bohle et al., 2005) are known.

3.1.6. PF00407: pathogenesis-related protein Bet v 1 family

The most important allergen from this class (InterPro IPR000916) is the major white birch (Betula verrucosa) pollen antigen. Bet v 1, which is the main cause of type I allergies observed in spring. The Bet v 1 allergens are formed by six anti-parallel betastrands and three alpha-helices. Four of the beta-strands dominate the global fold, and two of the helices form a C-terminal amphipathic helical motif. The family contains pathogenesis-related (PR-10) allergens (Midoro-Horiuti et al., 2001), such as Aln g 1, Api g 1, Ara h 8, Bet v 1, Cor a 1, Dau c 1, Gly m 4, Mal d 1, Pru ar 1, Pru av 1, and Pyr c 1. PDB structures are reported for Api g 1 (2BK0), Bet v 1 (1BV1, Fig. 1F), and Pru av 1 (1E09). The conformational IgE epitopes of Bet v 1 were identified (Mirza et al., 2000).

3.1.7. PF00261: tropomyosin

Tropomyosins (InterPro IPR000533) are alpha-helical proteins that form a coiled-coil structure of two parallel helices containing two sets of seven alternating actin-binding sites. In striated muscles, tropomyosin regulates the muscle contraction by mediating the interactions between the troponin complex and actin. Allergies to crustaceans, such as shrimp, crab, crawfish and lobster, are mainly induced by tropomyosin (Reese et al., 1999). IgE epitopes are known for the shrimp allergens Pen a 1 (Ayuso et al., 2002a) and Pen i 1 (Shanti et al., 1993). The high conservation of tropomyosin sequences among invertebrates explains why the cross-reactivity of allergens from shellfish and mollusks are often cross-reactive (Chu et al., 2000; Jeong et al., 2006). However, vertebrate tropomyosins are not known to be allergenic.

3.1.8. PF00190: cupin

The cupin family (InterPro IPR006045) contains the conserved barrel domain of the cupin superfamily (cupa is the Latin term for a small barrel), and is comprised of 11S and 7S plant seed storage proteins. The IgE epitopes for five members of this family are reported in the literature: Ara h 1 (Shin et al., 1998), Ara h 3 (Rabjohn et al., 1999), Fag e 1 (Yoshioka et al., 2004), Gly m glycinin G1 (Beardslee et al., 2000) and Gly m glycinin G2 (Helm et al., 2000).

3.1.9. PF00061: lipocalin/cytosolic fatty-acid binding protein family

Lipocalins (InterPro IPR000566) are proteins that transport small hydrophobic molecules, such as lipids, retinoids, and steroids. The fold is an eight-strand anti-parallel beta-barrel enclosing the binding site. The structures of several allergens from this family are known: Bos d 2 (1BJ7), Bos d 5 (1GXA, Fig. 1G), Equ c 1 (1EW3), Mus m 1 (1MUP) and Rat n 1 (2A2U).

3.1.10. PF03330: rare lipoprotein A (RlpA)-like double-psi beta-barrel

The rare lipoprotein A (RlpA) fold (InterPro IPR005132) is found in bacterial and eukaryotic lipoproteins, and represents a double-psi beta-barrel fold. This domain may be found in the N-terminal part of several pollen allergens. The 3D structure of only one allergen, Phl p 1 (1N10, Fig. 1H), is known.

3.1.11. PF00042: globin

Globins (InterPro IPR000971) are heme-containing proteins involved in binding and/or transporting oxygen. Hemoglobin is a protein that in vertebrates transports oxygen from lungs to other tissues, containing two alpha and two beta chains with the characteristic three-dimensional globin fold. Monomeric and dimeric hemoglobins have been identified as major allergenic components in insects. The antigenic determinants of this family from Chironomus thummi thummi (midge) have been characterized as regions with dominant polar amino acids and high flexibility (Baur et al., 1986). The global fold of the monomeric allergen Chit 1 is shown in Fig. 1I (PDB code 1ECO).

3.1.12. PF00544: pectate lyase

Pectate lyase (InterPro IPR002022) is an enzyme involved in the cleavage of pectate, which occurs during the maceration and rotting of plant tissue. This family contains several major pollen allergens, such as those from short ragweed (Ambrosia artemisiifolia), Amb a 1 and Amb a 2. The most common pollen allergen in Japan is Cry j 1, a glycoprotein from the Japanese cedar (Cryptomeria japonica). Other cedar allergens are Jun a 1 (Juniperus ashei, mountain cedar), Jun o 1 (Juniperus oxycedrus, prickly juniper), Jun v 1 (Juniperus virginiana, eastern red cedar). Pollen from several cypress species contains allergens homologous with pectate lyase, namely Cup a 1 (Cupressus arizonica, cypress), Cup s 1 (Cupressus sempervirens, common cypress), Cha o 1 (Chamaecyparis obtuse, Japanese cypress). The IgE epitopes are known for Cry j 2 (Tamura et al., 2003) and Jun a 1 (Midoro-Horiuti et al., 2003, 2006), and the T-cell epitopes were identified for Cha o 1 (Sone et al., 2005), Cry j 1 (Sone et al., 1998), and Cry j 2 (Sone et al., 1998). The structure of one allergen for this family has been deposited in PDB: Jun a 1 (1PXZ, Fig. 1J) (Czerwinski et al., 2005). All allergens from this family have similar sequences and there are significant cross-reactivities to food allergens (Bonds et al., 2008). Schwietz et al. studied the in vivo and in vitro cross-reactivity between pollen extracts of mountain cedar and 11 other Cupressaceae species, one Taxodiaceae species (Japanese cedar), one Pinaceae species, and an angiosperm, and found that the 12 Cupressaceae and the Japanese cedar are cross-reactive (Schwietz et al., 2000).

3.2. Sequence motifs characteristic of allergens may correlate with cross-reactivity

Proteins in the same Pfam class are homologous, are expected to share a similar 3D-structure, and often have common biochemical functions (Finn et al., 2006). High overall sequence similarity is a good indicator of cross-reactivity (Aalberse, 2007). However, as antibodies bind to surface patches of folded proteins, cross-reactivity may be better indicated by matching specific areas of the protein structure rather than just the global fold. To differentiate local sequence areas of known allergens, we first generated sequence motifs that are characteristic for the complete family of all those Pfam classes that contain allergens. These “Full-Pfam motifs” can be used to classify novel proteins, and to determine whether it belongs to a Pfam with many allergenic members. In addition, “Allrg-Pfam” motifs were defined that were derived from alignments of only the allergens within each protein family. This procedure allows us to distinguish allergen specific motifs from those that are common to all proteins in the family. All Full-Pfam sequence motifs are publicly available on our MotifMate web server (http://born.utmb.edu/motifmate/summary.php).

We next compared the motifs that were specific for the allergens with known IgE epitopes, to see if there was a correlation that could account for clinically significant cross-reactivities between allergens. Three major Pfam families were chosen: the seed storage proteins (a subset of PF00234), the pathogenesis-related protein Bet v 1 family (PF00407) and tropomyosin (PF00261). The motifs common to the allergen members of each family were compared to known IgE epitopes (Table 3). Motifs of Full-Pfam and Allrg-Pfam in equivalent positions in the Pfam domains are listed on the same line and referred to with the number in column 1. For the seed storage proteins, there are five motifs in Allrg-Pfam (numbered 1, 3-6) and four Full-Pfam motifs (2, 3, 4, 7). Motifs 1, 5 and 6 are unique to the allergens (Allrg-Pfam). A novel protein that contained some or all of the Full-Pfam motifs would probably be a member of this Pfam. If there were a significant match to motifs 1, 5 or 6 that characterized the allergenic proteins, it would be also flagged as potentially allergenic. The only representative of this family for which IgE epitopes have been determined is the walnut allergen Jug r 1. The epitope QGLRGEEMEEMV (Robotham et al., 2002) partially overlaps with motif 6 that is characteristic of allergens (bold letters in Table 3). This suggests that this common sequence could play a role in observed clinical cross-reactivities among allergens of this protein family (Comstock et al., 2004; Goetz et al., 2005; Robotham et al., 2005).

Table 3.

AutoMotifs for allergens and entire Pfam families for seed storage proteins, Bet v 1-related family, and tropomyosin

No	Allergens only		Entire Pfam family
	E	Motif	E	Motif
PF00234: protease inhibitor/seed storage/LTP family (subgroup B)-seed sequence: Jug r 1
1	1.7	1 CQYYLR 6
2			0.5	10 RSGGYDED 17
3	1.8	26 CCQQLS 31	0.9	26 CCQQLSQI 33
4	2.0	37 CQCEGLR 43	0.5	37 CQCEGL 42
5	1.7	49 QQQQ 52
6	1.8	59 EMEEMVQSA 67
7			1.2	67 ARDLPKEC 74
PF00407: pathogenesis-related protein Bet v 1 family-seed sequence: Bet v 1
1	2.0	6 ETETTSVIPA 15
2			1.3	15 AARLFKA 21
3	2.0	31 PKVAP 35	1.2	25 DGDNLFPKVAP 35
4	2.0	42 ENIEGNGGPGTIK 54	1.8	46 GNGGPG 51
5	1.8	69 DRVDEVD 75	1.5	68 KDRVDEVD 75
6	1.7	81 YNYSVIEGGPI 91
7	2.0	110 GGSILK 115	1.5	110 GGSILK 115
8	2.0	120 YHTKG 124	1.0	120 YHTKGD 125
9			0.7	129 KAEQVKASK 137
10	2.0	145 RAVESYLLAH 154	1.2	145 RAVESYLLAH 154
PF00261: tropomyosin-seed sequence: Pen a 1
1	2.0	7 ENDLD 11
2	1.8	14 QESL 17
3	2.0	20 ANIQ 23	0.8	20 ANIQLV 25
4	2.0	33 NAEGEVA 39
5			1.0	39 AALNRR 44
6	2.0	47 LLEEDLERSEER 58	1.0	54 RSEER 58
7	2.0	65 KLAEASQAADESERMRKVLE 84	1.4	62 ATTKLAEASQAADE 75
8			1.5	79 MRKVLENR 86
9	2.0	90 DEERMDALENQLKEAR 105	1.5	90 DEERM 94
10			1.6	98 ENQLKEA 104
11	2.0	108 AEEADRKYDEVARKLAMVEADLERAEERAE 137	1.5	108 AEEADRKYDEVA 119
12			1.2	130 ERAEERAETGE 140
13	2.0	145 ELEEELRVVGNNLKSLEVSEEKANQRE 171	1.1	147 EEELR 151
14			1.0	155 NNLKS 159
15			1.1	166 KANQREEAYK 175
16	2.0	174 YKEQIKTL 181
17	2.0	184 KLKAAEARA 192	1.4	186 KAAEARAEFAE 196
18	2.0	195 AERSVQKLQKEVDRLEDELVNEKEKYK 221	1.2	201 KLQKEVDRLE 210
19	2.0	225 DELD 228

Similarly, unique Allrg-Pfam motifs 1, 3-8 and 10 characterize allergens in the Bet v 1 family (Table 3). Here again, a conformational IgE epitope, 42ENIEGNGGPGT52 70R 72D 76H 86I 97K (Mirza et al., 2000) correlates with sequences within these Allrg-Pfam motifs. The entire linear part of the IgE epitope is found in the Allrg-Pfam motif 4, and the individual residues 70R, 72D and 86I are in motifs 4 and 5. The cross-reactivites observed between allergens from this family (Aalberse et al., 2001; Kazemi-Shirazi et al., 2000; Wensing et al., 2002) may be explained by the conservation of this physico-chemical profile for the Bet v1 IgE epitope across all these allergens. As in the first example, the experimentally documented IgE epitope sequence correlates better with motifs derived from the known allergens than for those that characterize the whole Pfam class.

Numerous studies have related the similar structures of members of the tropomyosin family to clinically significant cross-reactions (Ayuso et al., 2002b; Chu et al., 2000; Fernandes et al., 2003; Jeong et al., 2006; Wild and Lehrer, 2005; Zhang et al., 2006). We previously demonstrated that tropomyosin allergens are difficult to discriminate from non-allergenic tropomyosins with the current web servers for allergenicity prediction (Schein et al., 2007). In this work, we found that Allrg-Pfam and Full-Pfam motifs showed distinctions between the two groups. MotifMate identified 19 common motifs in the highly conserved sequences of tropomyosins. Five of these, 1, 2, 4, 16 and 19, are characteristic of the allergenic family members. We then mapped the sequences of nine linear IgE motifs that were identified for the shrimp tropomyosin allergen, Pen a 1 (Ayuso et al., 2002a). While areas of the epitopes are found in motifs common to all tropomyosins, the sequences for the most part correlate with the Allrg-Pfam motifs that are specific for the allergenic tropomyosins. In particular, the Allrg-Pfam motifs 1 and 19 match well to epitope sequences. These three examples all indicate that distinguishing local areas of conserved physicochemical properties that are common to allergenic members of the same Pfam can be useful in predicting determinants of IgE reactivity, and potential cross-reactivity.

4. Discussion

One major goal of our SDAP database is to provide a rapid way for researchers to identify common features of allergenic proteins, as a basis for identifying proteins that could be expected to cause cross-reactions in patients. The sequence comparison tools in SDAP can be used for that purpose. However, grouping all the allergens in SDAP according to protein families within Pfam (Tables 1 and 2) now makes this determination even faster, and more accurate as distinct domains of the allergens are assigned to different Pfam. Further, this grouping allowed us to define a series of known 3D protein structures (Fig. 1) to characterize the folds of the majority of allergens. We also derived new sequence specific motifs of proteins in those protein families with a large number of allergens and demonstrated that we also can generate specific motifs that can distinguish them from homologous but non-allergenic proteins in the same protein family (Table 3). Finally, we could show that specific motifs did indeed correlate with allergenicity, as they corresponded to experimentally determined linear IgE epitopes for three different examples.

However, the conserved motifs that are characteristic only for allergens from a Pfam family are not restricted to the set of IgE epitopes. These motifs may be buried, in which case they represent structurally important residues. Alternatively, the group of residues could give the necessary conformational flexibility to an antigenic site, thus distinguishing them from the rest of the family. The results reported in Table 3 demonstrate that allergens within a Pfam family have distinct conserved regions as compared to the entire Pfam family.

Our data correlate well with previous attempts to group allergenic proteins according to common sequences, structures (Aalberse, 2000), and functional classes (Breiteneder and Ebner, 2000; Ebner et al., 2001; Midoro-Horiuti et al., 2001). Our finding demonstrate novel applications of allergen classifications (Breiteneder and Mills, 2005b; Breiteneder and Mills, 2006; Jenkins et al., 2005; Radauer and Breiteneder, 2006; Radauer et al., 2008), and allowed us to also analyze finer details of the sequences that correlate with allergenicity. As most of the known allergens can be grouped to only 31 Pfam, this indicates that bioinformatics approaches should be useful for predicting allergenicity for novel proteins. The MotifMate approach outlined here indicates further that sequence fingerprints of allergens and non-allergens within each Pfam family could provide a useful tool to predict cross-reactivity of allergens with similar sequences.

These findings represent a considerable advance over the original decision tree for combining computational and experimental tests to determine whether a protein is a potential allergen (WHO, 2000, 2001, 2003). There, cross-reactivity was predicted based on overall sequence similarity of 35% of sequence identity in a window of 80 residues, from FASTA alignments, or on identical regions, as short as six to eight residues, in the protein sequences. While the FAO/WHO procedure is available in SDAP (Schein et al., 2006, 2007), our results and those of others (e.g. Hileman et al., 2002) indicated that too many non-allergenic proteins are detected by the suggested thresholds. We suggest that these guidelines be modified, to use more sophisticated comparisons to the known allergen sequences, and particularly allergen specific motifs such as those we define here.

Others have also suggested that motif-based methods could identify allergenic proteins more specifically. The MEME protein motifs (Mari, 2005; Saha and Raghava, 2006; Stadler and Stadler, 2003), for example, have average lengths of 50 residues (Marti et al., 2007), and are thus not as specific as the physicochemical properties motifs we were able to extract. Our motifs correspond better to the length expected for epitopes. The MotifMate motifs can be used to filter large genomic databases directly, either before or after a preliminary classification to eliminate all sequences that do not belong to Pfam families in SDAP. In this way, a large number of sequences will be eliminated from the first step, without time-consuming computations.

The advantages of a motif-based approach are clear from the examples presented above. Our MotifMate comparisons discriminated between allergenic and non-allergenic tropomyosins, a difficult task as allergenic tropomysosins, from mite (Der p 10) and shrimp (Pen a 1), are highly similar (80.28%, 228/284, E score 7.2e-73) to the mammalian homologs that are not allergenic (Schein et al., 2007). Of four programs tested for their ability to distinguish four allergenic tropomysosins from four non-allergens, only WebAllergen (Riaz et al., 2005) found that while all eight proteins have five wavelet allergenic motifs (Li et al., 2004) in common, the allergenic tropomyosins have several additional wavelet motifs that may distinguish them. Both Allermatch (Fiers et al., 2004), which applies the FAO/WHO allergenicity guidelines (WHO, 2000, 2001, 2003) and AlgPred (a support vector machines classifier) (Saha and Raghava, 2006) found all eight tropomyosins to be allergens, while MEME motifs (Stadler and Stadler, 2003) predicted all eight to be non-allergens.

We should at this point note that PCPMer motifs, Allrg-Pfam and Full-Pfam, are numerical vectors based on the E₁-E₅ physico-chemical properties. They have been translated, for convenience, into representative amino acid sequences in Table 3. They can be used, in combination with other SDAP tools, to compare the physicochemical properties of these motifs to those of novel proteins.

5. Conclusions

The identification of potential allergenic proteins is usually done by global sequence similarity searches. Tools to do overall similarity searching are now incorporated in SDAP. The classification of allergens into Pfam domains reveals the structural relationship between various allergens, thus providing a basis for identifying allergenic determinants. Our results show that allergens can be represented by a small fraction of possible protein families and folds. Out of the 9318 protein families from Pfam, only 130 families are currently listed for all allergens in SDAP, and 31 families contain more than four allergens. The most populated Pfam families are protease inhibitor/seed storage/lipid transfer protein, profilin, EF hand, group I pollen allergens, SCP-like extracellular protein, pathogenesis-related protein Bet v 1 family, tropomyosin, and cupins. Details for the Pfam classification of all allergens can be accessed from the SDAP web site (http://fermi.utmb.edu/SDAP/). The sequence motifs characteristic for Pfam classes are available via our web server MotifMate (http://born.utmb.edu/MotifMate/). Those motifs represent sequence-based fingerprints that characterize the major Pfam families with allergens. In addition we also showed, for three important Pfam classes that contain many allergens, how specific motifs correspond to known IgE epitopes. These allergen-specific motifs are the basis of an original method to predict the potential allergenicity of novel proteins and clinical cross-reactivity.