Proteomic and in silico identification of potential allergenic proteins in cowpea (Vigna unguiculata L. Walp) seeds

Abstract

Cowpea (Vigna unguiculata L. Walp) seeds are rich in proteins (∼24%), thus they have been considered as a viable dietary protein substitute. Despite its advantages, allergenicity risks of cowpea seeds need to be taken into account. Herein, the protein expression and potential allergens in cowpea seeds were analyzed using LC/timsTOF Pro 2, PEAK studio and multiple in silico analysis. Based on functional classification using STRING analysis, the result revealed response to stimuli, e.g., oxidative stress and temperature, as the major cluster (187 proteins), followed by biosynthesis of secondary metabolite (117 proteins) and immune system (108 proteins). These suggest involvement of protein functions in maintaining homeostasis during seed development under stress conditions. By using webtool Allermatch™ and the Pfam database, 131 potential allergenic proteins were found from cowpea seeds. The findings revealed that cowpea seeds contain a number of recognized allergens, including endochitinase, beta-conglycinin, and vicilin, as well as numerous allergenic proteins not previously described, such as endochitinase 1B and 5-methyltetrahydropteroyltriglutamate–homocysteine methyltransferase (MetE). Additionally, use of the Kolaskar & Tongaonkar method predicted B-cell epitopes such as ³⁰VSGFGVI³⁶, ¹⁵²VPVLVGP¹⁵⁸ and ¹⁵³PVLVGPV¹⁵⁹, increasing the possibility of cowpea allergenicity. In conclusion, this study provides useful information on the potential allergens in cowpea seeds, providing a foundation for future cowpea allergenicity assessment including experimental IgE-binding or clinical validation.

Highlights

• •Proteomics and in silico analyses identified 131 possible allergens in cowpea.
• •The Immune Epitope Database showed 696 epitopes that linked to allergenic responses.
• •This study revealed novel cowpea allergens such as endochitinase 1B.
• •³⁰VSGFGVI³⁶ was the most potential B-cell epitope regions of endochitinase 1B.

1. Introduction

Plant proteins have recently become popular because of their high value for nutrition, low cost and their more environmentally friendly role than animal-based protein sources. Legumes including peanut, soybean, bean, pea as well as cowpea, have shown numerous benefits for planetary health. Moreover, higher legume consumption has been recommended by world dietary guidelines as Sustainable Healthy Diets (models of healthy eating that have low effect on the environment). However, the consumption rate is often low in various countries. One of the key elements is that the allergens in legumes may cause an allergic reaction in those who consume them.The food manufacturing and production sectors, and consumers all need to give careful thought to food allergy profile. For food authentication and allergen identification, sensitive, accurate, and effective detection techniques are essential. Therefore, research on the allergens found in specific legumes may be able to address this issue.

In numerous regions of worldwide, cowpea (Vigna unguiculata L. Walp) is regarded as an excellent source of vegetative protein. Cowpea is considered to originate in Africa (Gonçalves et al., 2016; Herniter et al., 2020; Karapanos et al., 2017; Van Damme, 2007), but it is also cultivated across Africa, Southeast Asia, Latin America and Southern United States, due to it can adapt to hotter regions with sufficient rainfall. According to U.S. Department of Agriculture (USDA), cowpea is a nutritious, rich in protein (∼24%), dietary fiber (∼11%), while low in lipids (<2%) (U.S. Department of Agriculture, 2024). In addition, compared to cereals and root crops, cowpea has two to four times the protein content and a greater lysine content (Gonçalves et al., 2016; Trehan et al., 2015). Cowpea is rich in potassium and contains several minerals such as zinc, magnesium, iron, calcium, phosphorus, selenium and sodium, while vitamins include vitamin C, A, K, and B (B3, B5, B9) (U.S. Department of Agriculture, 2024).

Due to its nutritional value and low-cost protein source, not only in South Africa, there is increasing consumer demand for cowpea worldwide. As a result, cowpea has been added to the list of legumes that can trigger allergic responses and classified as a “non-priority” legume, which is not yet required to have allergen labels, along with lentil, chickpea, and pea. However, there is the evidence of allergies to non-priority legumes (Hildebrand et al., 2021; Soller et al., 2021) thus bearing potential to become severe allergens if their consumption becomes increasingly frequent. In addition, there are concerns that certain individuals who are allergic to legumes may be at risk if cowpea is used in food (Abu Risha et al., 2024; Chentouh et al., 2022). To document the usage of cowpea seeds as food, more research into possible allergies in these seeds is necessary.

Among the four novel allergens discovered in cowpea (Chentouh et al., 2022), storage proteins, vicilins (also called 7S globulins) and albumins, were revealed to be crucial. IgE-cross-reactivity between cowpea, pea, and peanut was also assessed in children (n = 27, average age 6). To assess the subjects' basophil sensitivity and reactivity to legumes, basophil activation tests were conducted. The results demonstrated that most patients with legume-allergy were sensitized to cowpea. By using multiple in silico tools, endochitinase was revealed as potential allergen in cowpea (Cavas & Yilmaz-Abeska, 2023).Additionally, the allergenic proteins of cowpea reported from World Health Organization/International Union of Immunological Societies (WHO/IUIS) Allergen Nomenclature Sub-Committee and the Allergome database was Vig u 6. Taken together, the evidence has strengthened allergists to be aware of and assess the risk presented by cowpea allergy.

The composition and function of proteins are usually ascertained through proteomics, a comprehensive study of all proteins (the proteome). In addition to in silico, ELISA and PCR, proteomics has been emphasized as crucial instruments for determining food allergies and confirming the authenticity of food (Carrera et al., 2024). Proteomics is also crucial for allergy prediction since it can detect specific allergenic proteins in complex meals and legumes. For instance, proteomic methods have been used to characterize the decreased IgE binding capability in soybeans (Pi et al., 2022). The findings masked many soybean epitopes such as ALVTDADNVIPK of Gly m 4. Proteomics and in silico allergenicity prediction revealed 7S-vicilins and 11S legumins as potential allergens in lentil. As one of the most significant staple crops is cowpea, proteome research on cowpea seeds had been explored for agronomic performance (Ribeiro et al., 2023). Unfortunately, proteomics-based research on cowpea allergens is still lacking.

We are interested in searching for bioactive peptides and proteins from edible plants, with legumes being of interest for this purpose. We found bioactive peptides from winged bean (unpublished data) and reported allergenic proteins from mature seeds and sprouts of purple winged bean (Subhasitanont et al., 2024). Our interest remains focused on the nutritional aspects including functional, bioactive value and allergenic proteins in legumes for usefulness and safety of consumers. The allergen profiles of the legume products are important for legume intake and finding methods to remove these proteins for safe consumption is still needed. As previously stated, there is a tendency for cowpea intake to rise along with concerns about cowpea allergy, and there is currently an absence of proteomic-based research on cowpea allergens. Herein, we were interested in employing proteomic techniques, followed by in silico methodology to search for the whole proteins and potential allergens of cowpea seeds. Our findings were published in an attempt of providing valuable information for planning and assessing future cowpea allergen validations.

2. Materials and methods

2.1. Legume materials

Cowpea seeds were purchased from Chua Yong Seng Co., Ltd. (Bangkok, Thailand).Until examination, the seeds were kept at 4 °C.

2.2. Protein isolation from cowpea seeds

Seeds were frozen using liquid nitrogen before grinding into powder with a ceramic pestle and ceramic mortar. Total proteins from powdered cowpea seeds (100 mg) were extracted with lysis buffer comprising urea (7 M), thiourea (2 M), Triton X-100 (2%), dithiotreitol (1%), phenylmethylsulfonyl fluoride (1 mM), and inhibitor mix of enzyme protease (Sigma-Aldrich, Saint Louis, MO, USA), and incubated at 25 °C for 1 h. After incubation, samples were centrifuged at 12,000 ×g for 20 min at 4 °C. Protein concentration was quantified using the Bradford protein assay kit (Bio-Rad Laboratories, Hercules, CA, USA). The bovine serum albumin was used as the standard.

2.3. 1D SDS-PAGE

A Hoefer Mini Vertical system (Hoefer Inc., Bridgewater, MA, USA) was used to separate ten micrograms of protein on a 12% polyacrylamide gel electrophoresis. Protein separation was performed using a constant current of 12 mA per gel until the migration distance reached 7 mm. The gels were then stained in a staining solution containing 0.1% Coomassie blue R-250 (Serva Electrophoresis GmbH, Heidelberg, Germany) for 16 h. All protein bands were removed from the gel after destaining, and they were then set up for in-gel tryptic digestion.

2.4. Proteomic analysis using mass spectrometry

The gel bands derived from SDS-PAGE of cowpea seed protein samples were digested using trypsin, following the procedures described in previous studies (Subhasitanont et al., 2024). Briefly, the bands were cut, minced and added 0.1 M NH₄HCO₃ in 50% acetonitrile at 30 °C for 20 min for destaining. After added 10 mM DTT, protein samples were incubated at 60 °C for reduction. After 45 min incubation, samples were incubated in the dark for 30 min at 25 °C using iodoacetamide at 100 mM for alkylation. After reagent removal, the gel samples were dried and the tryptic digestion was done by incubating 1:50 enzyme/protein at 37 °C overnight. Trifluoroacetic acid (TFA) was used to quench the digestion, and SpeedVac (Labconco Corporation, Kansas City, MO, USA) was used to dry peptide samples.

The dried peptide samples were desalted using ZipTip® (Merck KGaA, Darmstadt, Germany). SpeedVac was used to dry the samples after desalting. The acetonitrile solution (98%) with 0.1% formic acid was used to resuspend the samples. The samples with the amount 500 ng were injected onto nanoLC system in triplicate. An Aurora 25 cm column (IonOpticks, Collingwood, VIC, Australia) on a Dionex Ultimate 3000 RSLC Nano System was used to set up the total gradient for a 120-min runtime. The column was connected to a CaptiveSpray ionization source (Bruker Daltonics, Billerica, MA, USA) and a timsTOF Pro 2 mass spectrometer (Bruker Daltonics, Bremen, Germany).

The column oven (Sonation GmbH, Biberach, Germany) was maintained at 50 °C. The entire gradient was run using 200 nL/min flow rate. For data-dependent acquisition (DDA), full scans were set from 100 to 1700 m/z, with an ion mobility (1/K₀) dimension of 1.35 to 0.85 Vs/cm². The ion-mobility was recorded to separate precursors up to 10 parallel accumulation serial fragmentation (PASEF) MS/MS frames, excluding singly charged ions which are fully segregated in the mobility dimension. The threshold and target intensity were set at 1750 and 14,500 counts, respectively.

2.5. Protein identification and characterization

PEAKS® Studio Xpro 10.6 (Bioinformatics Solutions, Waterloo, ON, Canada) and the SwissProt database of Viridiplantae were used to identify the proteins. The following parameters were used when performing the database search: two missed cleavages during trypsin digestion; fragment mass error tolerance: 0.05 Da; parent mass error tolerance: 10.0 ppm. Minimum charge and maximum charge were set at 2 and 3, respectively. For variable modifications, oxidation (M), carbamidomethylation, and acetylation (N-terminal) were identified.

The characterization and protein–protein association networks of identified proteins were further investigated. The protein–protein association networks were analyzed from STRING (Search Tool for Retrieval of Interacting Genes/Proteins) version 12.0 (https://string-db.org, accessed on 11 Mar 2025) (Szklarczyk et al., 2023). Using Basic Local Alignment Search Tool (Blast®), the protein sequences from cowpea seeds were searched for homologous proteins with Arabidopsis thaliana. The clusters of protein-protein association networks obtained from STRING were examined to elucidate functional relationships and interaction networks.

2.6. In silico analysis for allergenicity and immune epitope prediction

The Allermatch™ (https://allermatch.org, accessed on March 14, 2025) (Fiers et al., 2004) was used to predict the allergenic potential of identified proteins in cowpea seeds. An 80-amino-acid sliding window alignment was used to compare sequences with known allergens, with a 35% identity cut-off. Allergens were categorized into protein families identified through the InterPro website and Pfam database version 37 (https://www.ebi.ac.uk/interpro, accessed on Apr 14, 2025) (Blum et al., 2025). Immune epitope mapping was performed using homologous peptides at a 90% identity level as determined by BLAST on the Immune Epitope Database (IEDB) website (https://www.iedb.org, accessed on June 28, 2025) (Vita et al., 2025). Host was set as human. Assay was set at B-cell, T-cell, and MHC. MHC restriction was set as any. Disease was set as allergic to cover all possible allergenic response. Kolaskar and Tongaonkar (KT) Antigenicity prediction algorithm within the IEDB was used for B-cell epitope prediction. All epitope sequences were aligned with peptide sequences of proteins identified from mass spectrometry using the Clustal Omega sequence analysis tool version 1.2.4 (https://www.ebi.ac.uk/jdispatcher/msa, accessed on June 28, 2025) (Madeira et al., 2024). The input sequence format was sequence fragments (FASTM/F). Protein structures of allergens were predicted using AlphaFold Protein Structure Database (https://alphafold.ebi.ac.uk/entry/%u, accessed on August 5, 2025).

3. Results and discussion

3.1. Proteins from cowpea seeds were identified using LC/MS/MS

To the best of our knowledge, the investigation of cowpea allergenicity is scarce yet described. There is no conclusive evidence of the severity of the issue, but parents in West Africa avoided giving their children cowpea as they became concerned about the legume's negative consequences (Akinyele & Akinlosotu, 1987). Six Indian individuals experienced cowpea allergy reactions when evaluated using the skin-prick test and IgE antibodies, and even at high protein concentrations, the seedless pods showed no allergenic activity (Rao et al., 2000). Allergic reaction was shown to be primarily caused by the albumin fraction of cowpea seeds, which has two major immunoreactive protein bands at 41 and 55 kDa (Rao et al., 2000). These proteins, however, were not identified. Four allergens have been identified in cowpeas thus far (vicilin, albumin-2 similar protein, and beta-conglycinin alpha and beta subunits) based on quantification of specific IgE by ELISA and ELISA IgE-inhibition assay and Basophil activation test (Chentouh et al., 2022). In addition to allergies, cowpea proteins were investigated using proteomics to increase their vital and stress tolerance (drought, pests) (Ribeiro et al., 2023). Nevertheless, there is still an absence of proteomics-based studies on cowpea allergens. Additionally, given that cowpea yield, harvested area, and dry grain output are rising in West Africa, followed by Asia, the Americas, and Europe (https://www.fao.org/faostat/en/; accessed on December 1, 2025), people in this region, particularly children (Akinyele & Akinlosotu, 1987; Chentouh et al., 2022), may be increasingly at risk for cowpea allergy (Abu Risha et al., 2024; Chentouh et al., 2022). Taken together, we thus aimed to used proteomics and multiple in silico analyses to predict possible allergens that might be concealed to be more found in an effort to provide useful information for future experimental planning and validations.

Proteins were isolated using lysis buffer. In order to remove detergents and salts from the samples, the SDS-PAGE of 10 μg proteins was performed for 10 min (Supplementary Fig. S1). Bands were cut, trypsinized and subjected to nanoLC/timsTOF Pro 2. The proteins were identified using PEAKS® Studio program (FDR ≤ 1%) as shown in Supplementary Table S1. There were 816 identified proteins from cowpea seeds. Number of peptides per protein was ranging from 1 to 26 peptides per protein. Percent sequence coverage was ranging from 1 to 87%. Searching in STRING software with multiple proteins was performed for functional classification (Fig. 1). Arabidopsis thaliana was used as a reference organism since it serves as the primary model plant, providing a universal reference for understanding plant protein functions and interactions, which are then mapped to other species via gene ontology (Reiser et al., 2024). A functional classification (PPI enrichment p-value <1 × 10⁻¹⁶) showed the total identified proteins were predominantly involved in response to stimulus including temperature, osmotic stress and oxidative stress (187 proteins, Fig. 1A), followed by biosynthesis of secondary metabolites (117 proteins, Fig. 1B), immune system (108 proteins, Fig. 1C) and heterocycle metabolic compound (105 proteins, Fig. 1D).

Fig. 1.

Protein-protein interaction analysis and functional classification of the proteomic-identified proteins of cowpea using STRING database network. Response to stimulus is the major group of functions (A), followed by biosynthesis of secondary metabolites (B), immune system (C) and heterocycle metabolic compound (D).

3.2. Potential allergenic proteins in cowpea seeds were analyzed by in silico

Allermatch™ was used to examine cowpea seeds' potential allergenic proteins from total identified proteins, and the InterPro website and Pfam database were used to classify matching allergens into protein families. Under the criteria for percentage of best hit identity higher than 35, a total of 131 potential allergens were listed in Supplementary Table S2. We have successfully identified several potential allergens including endochitinase, endochitinase A2, beta-conglycinin alpha subunit, beta-conglycinin beta subunit and vicilins which were previously reported in cowpea (Bueno-Gavilá et al., 2019; Cavas & Yilmaz-Abeska, 2023; Chentouh et al., 2022). Remarkably, our proteomics and in silico analysis revealed 123 potential allergens that have never been reported in cowpea before, such as fructose-bisphosphate aldolase, cytoplasmic isozyme 2, UTP-glucose-1-phosphate uridylyltransferase 1, endochitinase 1B, and 5-methyltetrahydropteroyltriglutamate–homocysteine methyltransferase (MetE). According to a Luxembourg study, most patients with legume allergies were sensitive to cowpea proteins (Chentouh et al., 2022). Moreover, their results showed four allergens including beta-conglycinin alpha and beta subunits, vicilin, and albumin-2 like protein with potential cross-reactivity to other legumes such as pea and peanut. In this study, the results showed that proteins in cowpea were matched with numerous known allergens in legumes as well as other plants (Fig. 2). Corylus avellana, Triticum aestivum, Zea mays were accounted for the majority (14 allergens each), followed by Glycine max (10 allergens), Artemisia annua (8 allergens) and Vigna radiata (6 allergens) (Fig. 2A).These findings suggested that consumers with particular legume and/or plant allergies may be at higher risk while consuming nonpriority legumes like cowpea. For allergen family category (Fig. 2B), the major group was Hsp70 protein (27 allergens) followed by cupin (19 allergens), profilin (15 allergens) and minor groups such as chitinase class I (6 allergens), enolase (3 allergens) and catalase (1 allergen). Additionally, as classified by allergen name (Fig. 2C), the major group was Cor a 10.0101 (12 allergens), followed by Zea m 12 (10 allergens), Pers a 1.0101 (6 allergens), Gly m 5 (5 allergens), and Vig r 2 (4 allergens), in addition to other, minor groups such as Bet v 8.0101 (2 allergens), Tri a 31.0101 (2 allergens), and Ara h 12 (1 allergen). In an effort to decrease animal testing and speed up food safety and drug development, in silico allergen prediction tools such as Allermatch™ have been used to evaluate the possible allergenic risk of expressed proteins in crops (Zhou et al., 2025). In silico allergen prediction frequently overestimate allergenicity due to false positives, particularly when using basic sequence similarity criterion like 6 amino acid matches or > 35% identity over 80 amino acids. To increase specificity, lower false alarms, and more accurately predict potential allergens, further epitope mapping may be utilized (Jamakhani et al., 2018; Zhou et al., 2025). One useful source for epitope mapping is the Immune Epitope Database (IEDB) (www.iedb.org) (Liu et al., 2018; Vita et al., 2025). It lists T-cell and B-cell epitopes that have been mapped experimentally. In most of the legumes, allergenicity is triggered by IgE-mediated mechanism. Nevertheless, non-IgE mediated reactions, mechanisms that are mainly driven by eosinophil or T-cell, can emerge following consumption of the legumes (Bellanti, 2024; Verma et al., 2013). Additionally, MHC (major histocompatibility complex) ligands, short peptides produced from proteins that bind to MHC molecules, is also essential to T-cell-mediated immune responses, which includes those resulting from allergens. Therefore, not only IgE allergic reaction, cowpea also possibly involves the non-IgE allergic interaction between its allergens and T cells as well as MHC ligands eventually provoke allergenic symptoms.

Fig. 2.

Classification of allergens in cowpea using Allermatch™, InterPro website and Pfam database according to source (scientific name) (A), allergenic family (B), and allergen name (C).

To predict the allergenic epitopes in cowpea, following Allermatch™, InterPro, and Pfam database, all 131 potential allergens were further analyzed using IEDB based on plants and fruits. As a result, 695 potential allergenic epitopes were successfully predicted in the amino acid sequences of 52 out of 131 allergens (Table 1). The total number of potential allergenic epitopes in each protein according to IEDB was summarized in Supplementary Table S3.For example, there were 6, 1, 21, 1 and 4 of potential allergenic epitopes for 1-Cys peroxiredoxin PER1, Fructose-bisphosphate aldolase 4, Beta-conglycinin beta subunit 2, Heat shock 70 kDa protein 3 and Fructose-bisphosphate aldolase, respectively. All 695 potential epitopes were categorized based on source organism and molecule parent (Supplementary Table S3). Regarding to source organism, 352 epitopes matched to Arachis hypogaea and 201 epitopes to Glycine max, followed by minor groups such as Phleum pratense (41 epitopes), Betula pendula (24 epitopes) and Malus domestica (8 epitopes). Regarding to molecule parent, 341 epitopes (49.78% of total) matched to Ara h 1 followed by Gly m 6 (60 epitopes), P34 probable thiol protease (35 epitopes), beta-conglycinin alpha’ subunit (34 epitopes) and others as minor groups. According to these results, the majority of the predicted allergenic epitopes matched Ara h 1, a vicilin that is abundant in a variety of legumes such as peanut. Ara h 1 was the molecular basis for the clinically significant cross-reactivity between peanut and pea, according to a previous study (Wensing et al., 2003). In addition, both the Len c 1, a major allergen found in lentil, and Pis s 1, a major allergen found in pea, exhibit good conservation of the majority of the sequential B-cell epitopes identified on the C-terminus of the Ara h 1 allergen (Barre et al., 2005). Cowpea allergenicity may therefore be a concern for customers who are sensitive to Ara h 1 in other legumes.

Table 1.

List of potential allergenic cowpea proteins and peptides containing allergenic epitopes.

No.	Accessiona	Descriptiona	Allergenic familyb	Allergen nameb	Scientific namec	Best hit identity (%)c	% of hitsc	Identical amino acids (%)c	Epitope exampled		Peptide sequencee
									Epitope ID	Epitope sequence
1	A2SZW8	1-Cys peroxiredoxin PER1	C-terminal domain of 1-Cys peroxiredoxin	Tri a 32.0101	Triticum aestivum	88.8	100	81.2/229	174,083	DVIPANWKPD	–
2	P93263	5-methyltetrahydropteroyltriglutamate–homocysteine methyltransferase	Cobalamin-independent synthase	Sal k 3.0101	Kali turgidum	100	100	91.4/756	228,233 228,232 228,834	GPVTILNWSFVRNDQ GPGVYDIHSPRIPSK TKLDSEIKSWLAFAA	KGMLTGPVTILNWSFVRN KYGAGIGPGVYDIHSPRI KSWLAFAAQKV
3	P43237	Allergen Ara h 1 clone P17	Cupin	Ara h 1	Arachis hypogaea	100	100	91.6/597	434,773	VAKISMPVNTPGQFEDFFPASSR	KISMPVNTPGQFEDFFPASSRD
4	P43238	Allergen Ara h 1 clone P41B	Cupin	Ara h 1	Arachis hypogaea	100	100	92.6/606	434,773	VAKISMPVNTPGQFEDFFPASSR	KISMPVNTPGQFEDFFPASSRD
5	P13917	Basic 7S globulin	Xylanase inhibitor	Pru du gamma-conglutin	Prunus dulcis	70	94.54	49.6/413	913,233 913,590	FSTSSLHSH LASHFGLQ	RVGFSTSSLHSHGVKC RNTQGVAGLGHAPISLPNQLASHFGLQRQ
6	P11827	Beta-conglycinin alpha’ subunit	Cupin	Gly m 5	Glycine max	100	100	99.2/597	181,524 181,455	YYVVNPDNNENLRLITLAIPVNKPGRFES SEDKPFNLRSRDPIYSNKLGKFFEITPEKN	RVPAGTTYYVVNPDNDENLRM KTISSEDKPFNLRS
7	P0DO16	Beta-conglycinin alpha subunit 1	Cupin	Gly m 5	Glycine max	100	98.1	100/543	914,189 913,596 913,402	TTYYVV LEASYDT IFLSIVD	RVPSGTTYYVVNPDNNENLRL RNILEASYDTKFEEINKV RDLDIFLSIVDMNEGALLLPHFNSKA
8	P0DO15	Beta-conglycinin alpha subunit 2	Cupin	Gly m 5	Glycine max	100	98.1	100/543	914,189 913,596 913,402	TTYYVV LEASYDT IFLSIVD	RVPSGTTYYVVNPDNNENLRL RNILEASYDTKFEEINKV RDLDIFLSIVDMNEGALLLPHFNSKA
9	P25974	Beta-conglycinin beta subunit 1	Cupin	Gly m 5	Glycine max	100	100	99.3/416	913,403 912,964 913,847	IFLSSVDIN AYPFVV NQRESYFVDAQPKK	RDLDIFLSSVDINEGALLLPHFNSKA RAELSEDDVFVIPAAYPFVVNATSNLNFLAFGINAENNQRN RESYFVDAQPQQKE
10	F7J077	Beta-conglycinin beta subunit 2	Cupin	Gly m 5	Glycine max	100	100	99/416	99,372 913,847	KEGALMLPHFNSKAM NQRESYFVDAQPKK	RDLDIFLSSVDINEGALLLPHFNSKA RESYFVDAQPQQKE
11	P29357	Chloroplast envelope membrane 70 kDa heat shock-related protein	Hsp70 protein	Cor a 10.0101	Corylus avellana	82.5	100	62.6/633	228,761	SIKNAVVTVPAYFND	KNAVVTVPAYFNDSQRQ
12	P06215	Endochitinase	Chitinase class I	Pers a 1.0101	Persea americana	81.27	100	75.8/293	6914	CRCEGLRMMMMRMQ	–
13	Q06013	Endochitinase 1B (Fragment)	Chitinase class I	Pers a 1.0101	Persea americana	43.76	100	76.1/46	432,876	VKVEIINGGLTL	–
14	P21226	Endochitinase A2	Chitinase class I	Pers a 1.0101	Persea americana	88.8	100	75.3/299	104,744	CSQWGWCGST	–
15	Q9FSY7	Endoplasmic reticulum chaperone BiP	Hsp70 protein	Cor a 10.0101	Corylus avellana	100	100	100/640	228,878	VEIIANDQGNRTTPS	KNGHVEIIANDQGNRI
16	P42896	Enolase	Enolase	Hev b 9	Hevea brasiliensis	100	100	95.1/445	561,050 228,534 168,618	TDFGVFRAAVPSGAS NALLLKVNQIGSVTE GAGWGVMVSHRSGET	RAAVPSGASTGIYEALELRD KVNQIGSVTESIEAVRM RAGWGVMASHRS
17	P26301	Enolase 1	Enolase	Hev b 9	Hevea brasiliensis	97.5	100	88.8/446	228,534 168,618	NALLLKVNQIGSVTE GAGWGVMVSHRSGET	KVNQIGSVTESIEAVRM RAGWGVMASHRS
18	Q9LEI9	Enolase 2	Enolase	Hev b 9	Hevea brasiliensi	100	100	100/445	168,618	GAGWGVMVSHRSGET	RAGWGVMASHRS
19	Q9ZU52	Fructose-bisphosphate aldolase 3, chloroplastic	Fructose-bisphosphate aldolase class-I	Art an 14.0101	Artemisia annua	66.2	95.51	55.2/353	228,492	LQHISGVILFEETLY	–
20	F4KGQ0	Fructose-bisphosphate aldolase 4, cytosolic	Fructose-bisphosphate aldolase class-I	Art an 14.0101	Artemisia annua	92.5	100	86.2/349	228,703	RCAYVTEVVLAACYK	RVLAACYKA
21	Q9SJQ9	Fructose-bisphosphate aldolase 6, cytosolic	Fructose-bisphosphate aldolase class-I	Art an 14.0101	Artemisia annua	91.2	100	84.4/358	228,314 228,957	IGKRFASINVENVED WFLSFSFGRALQQST	RLASINVENVESNRRA RALQQSTLKT
22	Q9LF98	Fructose-bisphosphate aldolase 8, cytosolic	Fructose-bisphosphate aldolase class-I	Art an 14.0101	Artemisia annua	88.8	100	81.6/358	228,314 228,957	IGKRFASINVENVED WFLSFSFGRALQQST	RLASINVENVESNRRA RALQQSTLKT
23	O65735	Fructose-bisphosphate aldolase, cytoplasmic isozyme	Fructose-bisphosphate aldolase class-I	Art an 14.0101	Artemisia annua	90	100	82.7/359	228,957	WFLSFSFGRALQQST	KKPWTLSFSFGRA
24	P46257	Fructose-bisphosphate aldolase, cytoplasmic isozyme 2	Fructose-bisphosphate aldolase class-I	Art an 14.0101	Artemisia annua	90	100	82.2/359	228,980	WFLSFSFGRALQQST	KKPWTLSFSFGRA
25	P34922	Glyceraldehyde-3-phosphate dehydrogenase, cytosolic	Glyceraldehyde 3-phosphate dehydrogenase	Tri a 34.0101	Triticum aestivum	96.2	100	84.2/336	228,432 559,972	KVINDRFGIVEGLMT KIGINGFGRIGRLVL	RFGIVEGLMTTVHSITATQKT KIGINGFGRI
26	Q75KH3	Glucose and ribitol dehydrogenase homolog	Enoyl-(Acyl carrier protein) reductase	No Allergen Name	Sesamum indicum	81.2	100	70.7/304	148,867	QQQQFPPQQP	–
27	P04776	Glycinin G1	Cupin	Gly m 6	Glycine max	100	67.07	100/287	2,248,894	NRIESEGGYIETWNP	RIESEGGLIETWNPNNKPFQCAGVALSRC
28	P04405	Glycinin G4	Cupin	Gly m 6	Glycine max	100	66.26	100/278	2,265,315	RFYLAGNQEQEFLKY	SLENQLDQMPRRFYLAGNQEQEFLKYQQEQG
29	P26413	Heat shock 70 kDa protein	Hsp70 protein	Cor a 10.0101	Corylus avellana	82.5	100	63.7/626	560,688 228,878 228,761	QPGVLIQVYEGERAM VEIIANDQGNRTTPS SIKNAVVTVPAYFND	KEQIFSTYSDNQPGVLIQVFEGERA RVEIIPNDQGNRT KNAVVTVPAYFNDSQRQ
30	P22954	Heat shock 70 kDa protein 2	Hsp70 protein	Cor a 10.0101	Corylus avellana	83.8	100	62.7/632	228,761	SIKNAVVTVPAYFND	KNAVVTVPAYFNDSQRQ
31	O65719	Heat shock 70 kDa protein 3	Hsp70 protein	Cor a 10.0101	Corylus avellana	83.8	100	64.5/633	560,688	QPGVLIQVYEGERAM	KEQVFSTYSDNQPGVLIQVYEGERA
32	Q6Z7B0	Heat shock 70 kDa protein BIP1	Hsp70 protein	Cor a 10.0101	Corylus avellana	98.8	100	91.4/584	228,878	VEIIANDQGNRTTPS	KNGHVEIIANDQGNRI
33	P27322	Heat shock cognate 70 kDa protein 2	Hsp70 protein	Cor a 10.0101	Corylus avellana	81.2	100	63.2/638	228,310	IEIDSLFEGIDFYST	–
34	Q948T6	Lactoylglutathione lyase	Glyoxalase/Bleomycin resistance protein/Dioxygenase superfamily	Ory s Glyoxalase I	Oryza sativa subsp. japonica	100	100	100	228,389	KIASFLDPDGWKVVL	–
35	Q42434	Luminal-binding protein	Hsp70 protein	Cor a 10.0101	Corylus avellana	97.5	100	89.9/641	228,878	VEIIANDQGNRTTPS	KVEIIANDQGNRI
36	P83434	Non-specific lipid-transfer protein 1	Protease inhibitor/seed storage/LTP family	Pha v 3.0101	Phaseolus vulgaris	76.23	100	76.9/91	50,121	KISTSTNCNSIN	PYKISTSTNCATVK
37	P25986	Pathogenesis-related protein 2	Pathogenesis-related protein Bet v 1 family	Vig r 1.0101	Vigna radiata	88.8	100	85.2/155	44,215	NIEGNGGPGTIK	KSVEIVEGNGGPGTIKK
38	Q8W171	Peptidyl-prolyl cis-trans isomerase 1	Cyclophilin type peptidyl-prolyl cis-trans isomerase/CLD	Cat r 1	Catharanthus roseus	95	100	86.5/171	953,442 954,519	DFTAGNGTGGESIYGSKFADENFV IYGSKFADENFVKKHTGPGILSMA	KFADENFVKKH KFADENFVKK
39	Q38900	Peptidyl-prolyl cis-trans isomerase CYP19–1	Cyclophilin type peptidyl-prolyl cis-trans isomerase/CLD	Dau c CyP	Daucus carota	90	100	84.5/168	169,871	TNGSQFFITTVVTSW	–
40	Q9SP02	Peptidyl-prolyl cis-trans isomerase CYP20–1	Cyclophilin type peptidyl-prolyl cis-trans isomerase/CLD	Dau c CyP	Daucus carota	83.8	100	70.5/166	228,762	SIYGAKFADENFIKK	KFADENFKL
41	Q38867	Peptidyl-prolyl cis-trans isomerase CYP19–3	Cyclophilin type peptidyl-prolyl cis-trans isomerase/CLD	Dau c CyP	Daucus carota	87.5	100	76.5/170	169,871	TNGSQFFITTVVTSW	–
42	Q8GT39	Profilin	Profilin	Pru p 4.02	Prunus persica	100	100	100/130	126,519	KYMVIQGEPGAVIRG	KYMVIQGEPGAVIRG
43	Q9SH52	Serpin-Z1	Serpin	Tri a 33	Triticum aestivum	53.8	79.08	41.8/395	1,717,741	ADHPFLFCIKHIATN	KIDFVADHPFLFLIRE
44	Q9ZQR6	Serpin-Z2	Serpin	Tri a 33	Triticum aestivum	55	71.65	39.8/394	1,717,741	ADHPFLFCIKHIATN	KIDFVADHPFLFLIRE
45	P57751	UTP–glucose-1-phosphate uridylyltransferase 1	Pathogenesis-related protein Bet v 1 family	Ara h 8.0101	Arachis hypogaea	37.5	3.32	37.3/83	228,900 228,241 228,327	VKVLQLETAAGAAIR GRFKSIPSIVELDSL IKRLVEADALKMEII	KVLQLETAAGAAIRF KSIPSIVELDSLKV KLVEADALKM
46	Q9SP02	Peptidyl-prolyl cis-trans isomerase CYP20–1	Cyclophilin type peptidyl-prolyl cis-trans isomerase/CLD	Dau c CyP	Daucus carota	83.8	100	70.5/166	228,762 954,519	SIYGAKFADENFIKK IYGSKFADENFVKKHTGPGILSMA	KFADENFKL KFADENFKL
47	P80463	Phaseolin	Cupin	Vig r 2	Vigna radiata	72.5	99.14	52.8/430	913,196	FFLSSTEAQ	–
48	P07219	Phaseolin, alpha-type	Cupin	Vig r 2	Vigna radiata	75	100	57.4/430	913,196	FFLSSTEAQ	–
49	P02853	Phaseolin, beta-type	Cupin	Vig r 2	Vigna radiata	75	100	57.1/424	913,196	FFLSSTEAQ	–
50	Q02028	Stromal 70 kDa heat shock-related protein, chloroplastic	Hsp70 protein	Cor a 10.0101	Corylus avellana	76.2	80.54	49.1/635	228,761	SIKNAVVTVPAYFND	KAVVTVPAYFNDSQRT
51	P27084	Superoxide dismutase [Mn], mitochondrial	Iron/manganese superoxide dismutases	Hev b 10.0101	Hevea brasiliensis	83.8	100	80.2/202	228,282	HAYYLQYKNVRPDYL	KNVRPDYLKN
52	P12863	Triosephosphate isomerase, cytosolic	Triosephosphate isomerase	Tri a 31.0101	Triticum aestivum	93.8	100	88.1/253	560,310 228,184	MGRKFFVGGNWKMNG GESSEFVGDKVAYAL	KFFVGGNWKC KVAYALSQGLKV
53	P13918	Vicilin	Cupin	Len c 1	Lens culinaris	96.2	100	85.4	100,038	FGKLFEVKPDKKNPQLQDLD	KFFEITPEKN

“-” indicates “not detected”. Black bold and red bold alphabet indicate match and miss-matched, respectively.

^aAccession and description were obtained from the LC-MS/MS analysis.

^bAllergenic families and allergen name were identified through the InterPro website and Pfam database version 37.

^cBest hit identity, % of hit and identical amino acids (%) were obtained from Allermatch™.

^dAllergenic epitope sequences were predicted by the database IEDB (https://www.iedb.org/).

^ePeptide sequences with at least six contiguous amino acids shared to with allergenic epitope sequence were obtained from the LC-MS/MS analysis.

Among the 131 potential allergens, endochitinase family was selected for further discussion, since previous investigations revealed that only endochitinase was found to be potential allergen in cowpea, but not endochitinase 1B. Moreover, MetE was also particularly worthy of discussion because there are currently few studies on this allergen and none on cowpea.

Chitinases are glycosidic hydrolases that catalyze the breakdown of chitin by hydrolyzing the β-1,4 glycosidic links in the chitin chain (Chen et al., 2020). Chitinase enzymes are classified into two categories according to their mode of action: endochitinase and exochitinase. Exochitinase cleaves chitin at the reducing or non-reducing terminal ends, while endochitinase randomly breaks the internal bonds. Plant chitinases are part of pathogenesis-related proteins as they can degrade fungal cell walls thus play role as seed defense and growth (Vaghela et al., 2022). Chitinases have been shown to play a role in a number of abiotic stress reactions in plants, including wounding, osmotic pressure, cold, heavy metal stress, and salt (Vaghela et al., 2022). Cowpea chitinase was purified in 1996 (Gomes et al., 1996). Endochitinase proteins including 1, 1B and A2 were classified as hevein-like and class I/II chitinase family, shown in cowpea seeds from our study. Of these, endochitinase was computational predicted in cowpea based on the similarity to protein endochitinase class I found in Persea americana (Cavas & Yilmaz-Abeska, 2023). However, endochitinase 1B has not been reported before. Therefore, endochitinase 1B is one of the potential novel cowpea allergenic proteins in endochitinase family found from our study. Search results from the IEDB database showed 3 potential allergenic epitopes of endochitinase 1B (Supplementary Table S3). The mapping of potential allergenic epitopes of endochinase and endochitinase 1B using IEDB as shown in Fig. 3A. The result showed that endochitinase 1B exhibited overlapping regions with multiple T-cell epitopes. For example, ³⁴VITNIINGGLDC⁴⁶ matched with epitope ID 432876 (¹⁴⁵VKVEIINGGLTL¹⁵⁶ of Amb a 1) which associated with proliferative T-cell response (Jahn-Schmid et al., 2010). The result also revealed IINGGL as possible conserve epitope region between endochitinase and endochitinase 1B. The B-cell epitope region for endochitinase 1B was further predicted using Kolaskar & Tongaonkar (KT) method within IEDB (Fig. 4A and C). The result highlighted ⁸PSSHDVI¹⁴ and ²⁹RVSGFGVI³⁶ as the most potential antibody recognized antigenicity regions. Three-dimensional model of endochitinase 1B was then predicted using AlphaFold. On a scale of 0 to 100, AlphaFold generates a per-residue estimate of its confidence, named as predicted Local Distance Difference Test (pLDDT). The AlphaFold-predicted three-dimensional model of endochitinase 1B showed model confidence with the average pLDDT score of 98.12. This level of structural confidence was considered as very high thus suitable for the prediction of epitope localization. After located the predicted epitopes in three-dimensional model, the result revealed that ²⁹RVSGFGVI³⁶ was found at random coiled and α-helix structure, while ⁸PSSHDVI¹⁴ was at an α-helix structure. The top three predicted residues were ³⁰VSGFGVI³⁶, ⁸PSSHDVI¹⁴, and ²⁹RVSGFGV³⁵, with scores of 1.110, 1.085, and 1.070, respectively.

Fig. 3.

Multiple regions of potential T-cell allergenic epitopes in endochitinase, endochitinase 1B and 5-methyltetrahydropteroyltriglutamate—homocysteine methyltransferase (MetE) predicted by Immune Epitope Database (IEDB). Sequence alignment was performed using Clustal Omega (EMBL-EBI). See full region alignment in Supplementary Figure S2, Supplementary Figure S3.

Multiple regions of potential T-cell allergenic epitopes in endochitinase, endochitinase 1B and 5-methyltetrahydropteroyltriglutamate—homocysteine methyltransferase (MetE) predicted by Immune Epitope Database (IEDB). Sequence alignment was performed using Clustal Omega (EMBL-EBI). See full region alignment in Supplementary Fig. S2 and S3.

Fig. 4.

B-cell epitope prediction of endochitinase 1B and 5-methyltetrahydropteroyltriglutamate–homocysteine methyltransferase (MetE). Panel A and B represent protein structure of endochitinase 1B and MetE, respectively, predicted by AlphaFold database. The B-cell epitope regions predicted by the Kolaskar & Tongaonkar (KT) algorithm within IEDB are indicated by the dashed red lines and yellow highlight in protein structures. The KT antigenicity maps of endochitinase 1B and MetE are displayed in panels C and D, respectively. The residues in yellow regions indicate potential antigenic areas. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

MetE, also known as methionine synthase (cobalamin-independent), plays a crucial role in one‑carbon metabolism by converting the amino acid homocysteine into methionine, which is essential for protein synthesis. MetE can be found in plant pollens such as in Catharanthus Roseus and Kali turgidum. Furthermore, MetE has been found in white mulberry pollen, which can cause airborne allergies (Çetereisi et al., 2019). However, few reports show existent of MetE in legumes. Consequently, since MetE was predicted as potential allergen in cowpeas in our study, more research on MetE in cowpeas may identify its new allergens that improve health risk management. IEDB predicted 9 possible allergenic epitopes of MetE (Fig. 3B, Supplementary Table S3). Each epitope ID was expected to be associated with T-cell allergies. For example, ³⁴⁰TKLDDEIKSWLAFAA³⁵³ of MetE matched with epitope ID 169849 (³⁴¹TKLDDEIKSWLAFAA³⁵⁵ of Sal k 3 pollen allergen) which associated with IL-5 release from T cell (Oseroff et al., 2012). The KT method showed a high score for the MetE putative B-cell epitope area (Fig. 4C and D). The AlphaFold-predicted three-dimensional model of MetE showed model confidence with the average pLDDT score of 96.19 (very high). With 13 amino acids, the predicted epitope area with the greatest score was ¹⁵²VPVLVGPVSYLLL¹⁶⁴ which was located on β-sheet and α-helix structure. The top three predicted residues among the ¹⁵²VPVLVGPVSYLLL¹⁶⁴ region were ¹⁵²VPVLVGP¹⁵⁸, ¹⁵³PVLVGPV¹⁵⁹, and ¹⁵⁸PVSYLLL¹⁶⁴, with scores of 1.200, 1.200, and 1.196, respectively. Notably, as MetE is known as a pollen allergen, inhaling cowpea seed dust or other airway exposure possibly trigger allergenic responses (Khan et al., 2023; Quirce & Diaz-Perales, 2013). Thus, MetE should not be ignored as a possible cowpea allergen, according to this study.

Pollen allergens can cross-react with seed proteins, such those found in Artemisia species, because of their similar three-dimensional structures (epitopes) namely panallergens. This is especially true within protein families including profilins, PR-10 (Bet v 1 homologs), and lipid transfer proteins (Haidar et al., 2025). Furthermore, profilins are known as significant panallergens in Pollen-Food Allergy Syndrome (PFAS) via common immune recognition (Haidar et al., 2025; Pablos et al., 2019). An allergic reaction develops when the immune system, which has previously become sensitized to a pollen protein, recognizes a similar protein in a seed as the initial threat. Therefore, we hypothesized that people who are allergic to pollens may cross-react with cowpeas because we also predicted profilin, a critical panallergen, as a potential allergen in cowpeas (Table 1).

4. Conclusions

We are interested in proteins from edible plants and selected legumes since they are good sources of plant proteins. However, consumers with allergies may be at risk from novel protein sources. Given the growing global demand for cowpea products, we used proteomics to identify the entire expressed proteins and a number of in silico approaches and predicted potential allergens in cowpea. The potential novel allergens were predicted through bioinformatics analyses in cowpea, including endochitinase 1B and 5-methyltetrahydropteroyltriglutamate–homocysteine methyltransferase. Additionally, we identified a substantial number of cowpea's potential allergens that shared a high degree of similarity with numerous other legumes and pollen. It is the evident that in silico used in this study are unable to predict potential allergen with 100% accuracy. For additional validations to verify the presence of cowpea allergens predicted by our study, allergists and researchers may find this information instructive.

CRediT authorship contribution statement

Daranee Chokchaichamnankit: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Conceptualization. Pantipa Subhasitanont: Writing – review & editing, Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Conceptualization. Toollayaporn Audsasan: Writing – review & editing, Methodology, Investigation. Jisnuson Svasti: Writing – review & editing, Resources, Project administration, Conceptualization. Theetat Ruangjaroon: Writing – review & editing, Writing – original draft, Validation, Resources, Methodology, Investigation, Formal analysis, Conceptualization. Chantragan Srisomsap: Writing – review & editing, Writing – original draft, Visualization, Validation, Supervision, Resources, Project administration, Methodology, Investigation, Funding acquisition, Formal analysis, Conceptualization.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could influence the work reported in this paper.

Appendix A. Supplementary data

Supplementary Figure S1.

The SDS-PAGE of 10 μg proteins from cowpea seeds was performed for 10 minutes. Coomassie blue R-250 was used for staining. Protein bands (red box region) were cut from the gel and prepared for in-gel tryptic digestion and proteomics analysis. M: Molecular weight standards.

Supplementary Figure S2.

Full region alignment depicts multiple regions of potential T-cell allergenic epitopes in endochitinase and endochitinase 1B predicted by Immune Epitope Database (IEDB). Sequence alignment was performed using Clustal Omega (EMBL-EBI).

Supplementary Figure S3.

Full region alignment depicts multiple regions of potential T-cell allergenic epitopes in 5-methyltetrahydropteroyltriglutamate—homocysteine methyltransferase (MetE) predicted by Immune Epitope Database (IEDB). Sequence alignment was performed using Clustal Omega (EMBL-EBI).

Supplementary material 4

The LC-MS/MS list of total proteins in cowpea seeds.

Supplementary material 5

The potential allergens identified by Allermatch^TM and classified into allergen family by Pfam database in cowpea seeds.

Supplementary material 6

Allergenic epitope prediction of cowpea seeds using the Immune Epitope Database (IEDB).

Data availability

No data was used for the research described in the article.