Current Research in Food Science 7 (2023) 100578 Available online 25 August 2023 2665-9271/© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by- nc-nd/4.0/). Development of a new recombinant antibody, selected by phage-display technology from a celiac patient library, for detection of gluten in foods Eduardo Garcia-Calvo a, Aina García-García a,*, Santiago Rodríguez-Gómez a, Sergio Farrais b, Rosario Martín a,1, Teresa García a,1 a Departamento de Nutrición y Ciencia de los Alimentos, Facultad de Veterinaria, Universidad Complutense de Madrid, 28040, Madrid, Spain b Servicio de Medicina Digestiva, Hospital Universitario Fundación Jiménez Díaz, 28040, Madrid, Spain A R T I C L E I N F O Handling Editor: Dr. Yeonhwa Park Keywords: Gluten Celiac disease Phage display ELISA Immune library Recombinant Fab A B S T R A C T Gluten, a group of ethanol-soluble proteins present in the endosperm of cereals, is extensively used in the food industry due to its ability to improve dough properties. However, gluten is also associated with a range of gluten- related diseases (GRDs), such as wheat allergies, celiac disease, and gluten intolerance. The recommended treatment for GRDs patients is a gluten-free diet. To monitor adherence to this diet, it is necessary to develop gluten-detection systems in food products. Among the available methods, immunodetection systems are the most popular due to their simplicity, reproducibility, and accuracy. The aim of this study was to generate novel high- affinity antibodies against gluten to be used as the primary reactant in an enzyme-linked immunosorbent assay (ELISA) test. These antibodies were developed by constructing an immune library from mRNA obtained from two celiac patients with a high humoral response to gluten-related proteins. The resulting library (composed by 1.1x107) was subjected to selection against gliadin using phage display technology. Following several rounds of selection, the Fab-C was selected, and demonstrated good functionality in ELISA tests, presenting a limit of detection of 15 mg/kg for detection of gluten in spiked mixtures and food products. The methodology can discriminate gluten-free products according to the current legislation. 1. Introduction Gluten-related disorders (GRDs) have an estimated global prevalence of 5%. Nowadays, GRDs are considered a significant global health issue, mainly in Western countries (Taraghikhah et al., 2020). GRDs are clas sified by their etiopathology into three main groups: Allergy to gluten or its components, autoimmune diseases, and non-celiac gluten sensitivity. Celiac disease is the most prevalent within the second group, pre senting a 1- 2% prevalence in the Caucasian population (Peña and Rodrigo, 2015). The molecular mechanisms of celiac pathology are well described (Caio et al., 2019). The diagnosis of celiac disease is closely related to the autoimmune response described, and usually implies several tests like: excluding selective IgA deficiency, high serum levels of IgA and IgG anti-tTG, serum IgA and IgG anti-deaminated gliadin pep tide (DGP), serum IgA endomysial antibodies, and HLA-DQ typing (ge netic screening). In addition to the serological and genetic tests, the diagnosis is usually confirmed by a gut biopsy (Raiteri et al., 2022). Despite several therapies have been proposed to mitigate celiac disease (anti-inflammatory drugs, inhibitory monoclonal antibodies, gluten chelators …), even nowadays, a gluten-free diet is the best therapy that not only inhibits the autoimmune gut tissue destruction, but also con ducts to a correct restoration of the intestinal crypts and the digestive function (Itzlinger et al., 2018). Due to the high prevalence of GRDs and the efficacy of a gluten-free diet to treat them, it was necessary to develop systems for gluten detection in foods. Several methodologies have been developed (García-García et al., 2018), but immunoassays are the most used, due to their accuracy and simplicity. The main reactants of these methods are high-affinity antibodies to gluten. Polyclonal and monoclonal approaches have been used, outstanding the monoclonal antibody R5, obtained using a classical hybridoma strategy after mice immunization with gluten-containing flour extracts. A sandwich ELISA using the R5 antibody is the golden standard method for gluten detec tion in foodstuffs (Valdés et al., 2003). Despite monoclonal antibodies are widely used nowadays, new developments are needed, based on * Corresponding author. E-mail address: ainagarcia@ucm.es (A. García-García). 1 These authors share senior authorship. Contents lists available at ScienceDirect Current Research in Food Science journal homepage: www.sciencedirect.com/journal/current-research-in-food-science https://doi.org/10.1016/j.crfs.2023.100578 Received 18 June 2023; Received in revised form 28 July 2023; Accepted 24 August 2023 mailto:ainagarcia@ucm.es www.sciencedirect.com/science/journal/26659271 https://www.sciencedirect.com/journal/current-research-in-food-science https://doi.org/10.1016/j.crfs.2023.100578 https://doi.org/10.1016/j.crfs.2023.100578 https://doi.org/10.1016/j.crfs.2023.100578 http://crossmark.crossref.org/dialog/?doi=10.1016/j.crfs.2023.100578&domain=pdf http://creativecommons.org/licenses/by-nc-nd/4.0/ http://creativecommons.org/licenses/by-nc-nd/4.0/ Current Research in Food Science 7 (2023) 100578 2 novel technologies like recombinant antibodies, that present several advantages like scalability, customizability, fast and consistent pro duction, and avoidance of the expression of unwanted chains that occurs with hybridomas (Bradbury et al., 2018). The Fab (fragment antigen binding region) is a recombinant anti body fragment composed of two chains: the complete light chain (var iable and constant regions) and a heavy chain, that consists in the variable domain and only one constant region (CH1) of the mammal immunoglobulins, but Fabs lack the Fc (fragment crystallizable) region. This antibody format was chosen because it replicates the region that binds the antigens in natural antibodies, providing the complete para tope pocket. Also, Fabs are more stable than single chain formats due to the disulfide bond that joins the heavy and the light chains, and their smaller size (compared to complete immunoglobulins) enhances their diffusive properties and facilitates the recombinant expression in pro karyotic cells (Kang and Seong, 2020). The main aim of this work was to obtain new alternatives to gluten immunodetection in foods, based on the development of recombinant Fabs presenting high-affinity against gluten. 2. Materials and methods 2.1. Bacterial strains, plasmids, and growth media Escherichia coli XL1-Blue strain (recA1, endA1, gyrA96, thi-1, hsdR17, supE44 relA1, lac [F proAB, lacIqZΔM15, Tn10 (Tetr)]) (Agilent©, Santa Clara, CA, USA, ref #200150) was used for cloning, building the li braries, and for production of phage-displayed Fabs. Electrocompetent cells were in-house produced by growing E. coli XL1-Blue in Super-Broth medium (SB: 30 g/L tryptone, 20 g/L yeast extract, 10 g/L MOPS, pH 7) and concentrating by sequential cycles of centrifugation (3200 g, 4 ◦C) and resuspension in 10% glycerol (v/v) (PanReac AppliChem© Monza, Lombardy, Italy, CAS: 56-81-5) in water. After electroporation, cells were grown in SOC medium (Invitrogen™-Thermo Fisher©, Waltham, MA, USA, ref #15544-034). For DNA extraction before sequencing, bacteria were grown in Luria Broth (LB: 10 g/L tryptone, 5 g/L yeast extract, 10 g/L NaCl, pH 7). Agar plates were prepared with a 15 g/L agar concentration. The antibody chain coding fragments were cloned into the plasmid pComb3X (Andris-Widhopf et al., 2000). 2.2. Antigen preparations Gliadin-PWG (Arbeitsge-meinschaft Getreideforschung (AGF) e.V. DIGeFa GmbH, Detmold, Germany) was used as reference material for gluten detection in foods. Following the supplier’s recommendation, gliadin-PWG was dissolved in a 60% ethanol/water solution at 1 mg/mL and stored at room temperature protected from light. Gluten-like pro teins were also extracted from different flour cereal matrices such as wheat, rye, barley, corn, and rice, as well from a binary mixture of wheat, rye, and barley kernels prepared in a rice-based gluten-free matrix as previously described (García-García et al., 2020). Gluten-like proteins were extracted by dissolving 0.25 g of the finely ground sam ple in 10 mL of 60% ethanol, followed by shaking for 20 min at room temperature and centrifugation at 2000g for 10 min at room temperature. 2.3. Selection of celiac patients as suitable lymphocytes donors Prospective donors were patients attending the digestive medicine service of the Fundación Jiménez Díaz Hospital (Madrid, Spain), with symptoms of a prototypical celiac disease humoral response. The following diagnostic parameters were used to assess the patient eligi bility: IgG and IgA tissular anti-transglutaminase (anti-tTG), IgG and IgA anti-deaminated gliadin peptide (DGP). In addition, patients with selective IgA deficiency were not considered potential donors. It was essential that the patients would not have started a free-gluten diet to assure the maximum of gluten-reactive B-cells in their peripheral blood (Agardh et al., 2006). Once a preliminary search was done based on the diagnosis of celiac disease of the prospective donors, the humoral response profile to complete gliadin was characterized. The ELISA tests α-Gliatest S Chromo IgA (Eurospital, Trieste, Italy, ref #910796) and α-Gliatest S Chromo IgG (Eurospital, ref #910896) were used for that purpose, according to the manufacturers’ protocols. An in-house ELISA test was implemented to quantify the amount of each immunoglobulin sub-class expressed in the patients. Immunosorb plates (96-well Maxisorp™ plates, Thermo Fisher©, Waltham, MA, USA, ref #44-2404-21) were coated for 1 h at 37 ◦C with 100 μL of 3.3 μg/mL of gliadin PWG in 0.1 M carbonate/bicarbonate buffer pH 9.6. The coating solution was removed and, after three washing steps with PBS, 200 μL of blocking solution (3% BSA in PBS) was added to each well. The plate was incubated at room temperature for 1 h and washed three times with PBS-T (PBS containing 0.05% of Tween (Sigma©, Burlington, MA, USA, ref #P1379)). Then, 100 μL of patients’ serum dilution (1:200 for IgG1 and IgG2 tests and 1:50 for IgG3 and IgG4) in sterile PBS was added per well. After incubating the plate at room temperature for 2 h, three washing steps were performed with PBS-T. One hundred microliters of anti-human sub-isotypes (Cygnus Technologies LLC, Southport, USA, ref #IM151 to #IM154) diluted 1:1000 in blocking solution were added. The plate was incubated for 1 h at room temperature and washed three times with PBS-T. One hundred microliters of TMB (Sigma©, ref #T0440) was then added, and the reaction was stopped after 15–20 min with 50 μL of a diluted sulphuric acid solution. Absorbance readings were performed at 450 nm (FluoStar Optima™ from BMG labtech©, Ortenberg, Germany). The serum of a non-celiac person was used as negative control. The results obtained were normalized to consider the different dilution of the serum used for each isotype. The normalization index (i) was calculated as follows: i= ( AU 1 AU 2 ) x serum dilution Where: AU1 are the absorbance units of the donors’ serum samples and AU2 are the absorbance units of the non-reactive serum sample used as negative control. 2.4. Lymphocyte RNA preparation and retro-transcription Blood samples were collected from selected donors following the protocols approved by the human research ethics committee of Fundación Jiménez-Díaz (Madrid). After signing the informed consent, 350 mL of peripheral blood were extracted from each donor, and cells were counted in a Neubauer chamber after Trypan Blue staining. Total RNA was extracted by a standard method with TriZol LS reagent (Invi trogen™-Thermo Fisher©, ref #10296010). Once the RNA quality and quantity were checked by Agilent© 2100 Bioanalyzer (ref #G2939BA), it was used as a template in reverse transcription reactions (SuperScript IV®, Invitrogen™- Thermo Fisher©, ref #18091050), following the manufacturer protocol, but the reaction was performed at 52.5 ◦C for 1 h instead of 10 min. 2.5. Amplification and isolation of antibody chain coding fragments The antibody chain coding fragments were amplified in two rounds of PCR. The first reaction was performed for specific amplification of the genes, and the second just to add sequences flanking the restriction sites for cloning improvement. Light chains and Fab heavy chains (VH and CH1 domains) were amplified separately using the primer set developed by Barbas et al. (Barbas, 2001) and Table 1S. For the first round, the reactions were set up with the following proportions: 1.5 μL of cDNA; 60 E. Garcia-Calvo et al. Current Research in Food Science 7 (2023) 100578 3 pmol of each primer (synthesized by Eurofins genomics©, Luxemburg, Luxemburg); 8 μL of 2.5 mM dNTPs set (Thermo Fisher©, ref #AB0196); 10 μL of 10X AmpliTaqGold® buffer; 0.5 μL AmpliTaqGold® (Thermo Fisher©, ref #4311806) and molecular biology grade water (Sigma©, CAS: 7732-18-5) to a final volume of 100 μL. Amplification was per formed under the following cycling conditions: Enzyme activation at 94 ◦C for 5 min; 35 cycles with denaturation at 94 ◦C for 15 s, primer annealing at 56 ◦C for 15 s, primer extension at 72 ◦C for 90 s; and a final extension step at 72 ◦C for 10 min. The amplified DNA fragments were isolated by gel electrophoresis using UltraPure™ Low Melting Point agarose (Thermo-fisher©, ref #16520050), and purified with Nucleo Spin® gel DNA clean-up columns (Machery-Nagel©, Allentown, PA, USA, ref #740609). DNA was quantified in a Qubit® Fluorometer (Invitrogen Thermo-fisher©), and its quality was measured with a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies Inc., Montchanin, DE, USA). For the second round of PCRs (Table 1S), the reactions were set up with the same proportions of reactants as in the first round, but using 100 ng of purified first-round PCR product as a template, and the following cycling conditions: 5 min at 94 ◦C for enzyme activation; 15 cycles with denaturation at 94 ◦C for 15 s, primer annealing at 54 ◦C for 15 s, primer extension at 72 ◦C for 2 min and, a final extension step at 72 ◦C for 10 min. Then, the product purification was performed as in the previous round. 2.6. Library construction (clonation and transformation) The purified PCR products and the plasmid pComb3X were digested with the corresponding enzymes and then, ligated in a two-step process. In the first step, the light chain coding fragments and the plasmid were digested for 1 h at 37 ◦C with SacI and XbaI (New England Biolabs© Ipswich, MA, USA ref #R0156 and ref #R0145), following manufac turers’ set-up reactions, and then purified by agarose electrophoresis as previously described. Digested inserts and linearized vector were ligated (3:1 M ratio) overnight at 16 ◦C with T4 DNA ligase. The resulting ligation was ethanol precipitated, re-suspended in 20 μL of water, and transformed into high-efficiency electrocompetent Escherichia coli XL1- Blue. SOC recovery medium (3 mL) was then added to the trans formed cells before incubation at 250 rpm for 1 h at 37 ◦C. At this point, 10 μL of the culture was removed, diluted, and titrated in LB agar plates containing 100 μg/mL of carbenicillin. The rest of the culture was transferred to 12 mL of LB broth with 20 μg/mL of carbenicillin and 10 μg/mL of tetracycline and incubated for an additional hour. The culture was then expanded to 100 mL of LB (carbenicillin at 50 μg/mL and tetracycline at 10 μg/mL) and grown overnight at 37 ◦C. Phagemid DNA containing the immune light chain repertoire was purified by midi-prep (PureLink™ HiPure plasmid Miniprep Kit Invitrogen™-Thermo Fisher© ref #K210004) and digested alongside the heavy chain coding fragments with XhoI and SpeI (New England Biolabs© ref #R0146 and ref #R0133). After gel purification, the resulting fragments were cloned following the steps described above for the light chain. The ligation reaction (containing the Fab library) was ethanol precipitated and transformed into E. coli XL1-Blue. Following electroporation, 5 mL of recovering SOC medium was added, and the culture was shaken at 250 rpm for 1 h at 37 ◦C. Then, 10 μL of the culture was removed, and ten- fold dilutions were plated into SB plates (containing 100 μg/mL of carbenicillin) to calculate the repertoire size of the library. The following steps of culture amplification, phage induction and isolation were done using the protocol depicted (Barbas, 2001). 2.7. Selection of gliadin-binding phage-Fabs The panning procedure was performed as previously described (Garcia-Calvo et al., 2022). Briefly, 1 μg of antigen (gliadin PWG) diluted in coating solution (0.1 M carbonate/bicarbonate buffer, pH 9.6) was added per well in a microtiter plate. The plates were washed and blocked with PBS-BSA 3%, incubated for 1 h at 37 ◦C and washed three times with PBS. Then, 100 μL of isolated phages were added per well, and incubated at 37 ◦C for 2 h. Non-binding phages were washed out with PBS-T and binding ones were eluted with acidic conditions. The eluted phage-Fabs were amplified by infecting a culture of E. coli XL1-Blue, the helper phage VSCM13 was added and grown overnight. Phage-Fabs were precipitated with PEG8000 (polyethylene-glycol) and sodium chloride. The cycle re-started by adding the resulting precipi tated phages to another round of selection against the antigen. Three rounds of selection and amplification were performed for the present experiment, and an extra amplification of the third round was consid ered the fourth. 2.8. Characterization individual gliadin-binding phage-Fabs Individual colonies isolated after each round of panning from the tittering output plates were grown at a small scale of 4 mL of initial culture, infected with helper phage, and induced for a quick screening of clones by analyzing the supernatants containing the phage-Fabs. The clones with better features were produced at a bigger scale of 100 mL and PEG-precipitated. 2.9. Indirect phage-ELISA The specificity and affinity of the clones obtained were analyzed by indirect-phage ELISA. One hundred microliters of the suitable antigenic extracts were added per well, diluted in coating buffer (0.1 M carbon ate/bicarbonate buffer, pH 9.6), and the plate was sealed and incubated for 1 h at 37 ◦C. The coating solution was shaken out and 200 μL blocking solution (3% BSA in PBS) was added per well, and the plate was incubated for 1 h at 37 ◦C. Following 10 washing steps, 100 μL of diluted phage was added and plates were incubated for 2 h at 37 ◦C. The non- binding phages were washed away 10 times with PBS. Then, 100 μL of the secondary antibody (anti-phage protein VIII HRP conjugated from Sinobiological©, Beijing, China, ref #11973) diluted 1:5000 in blocking buffer was added. The plate was incubated for 1 h at 37 ◦C, and PBS washed 10 times. One hundred microliters of TMB (Sigma©, ref #T0440) was added, and the reaction was stopped after 15–20 min with 50 μL of a diluted sulphuric acid solution. Absorbance readings were performed at 450 nm (FluoStar Optima™ from BMG labtech©). This methodology was applied for process evaluation, using the polyclonal mixture of phage-antibodies obtained from each panning round, and for characterization of monoclonal phage-Fabs. 2.10. DNA isolation and sequencing Individual colonies from the titration plates of panning rounds three and four were picked up and grown overnight in 5 mL of LB medium supplemented with 100 μg/mL of carbenicillin. The plasmid DNA was isolated from the pellet using a mini-prep kit (GenElute™ plasmid miniprep kit from Sigma© ref #PLN70). The phagemids were sequenced by the Sanger method (Eurofins genomics©, Luxemburg, Luxemburg) using the primers ompAseq and g-back (Barbas, 2001) for light and heavy chain reading, respectively. The sequences were visualized, and preliminary analysis was done using Snapgene software (Dotmatics®, Boston, MA, USA). 2.11. Searching and characterization of the gluten-binding Fabs The following discovery pipeline was proposed, using indirect phage-ELISA for searching of strong gliadin binding candidates: 1) for the initial screening, 48 individual clones picked from panning outputs plates were grown, induced for production of phage-Fabs, and the su pernatants were tested in a phage-ELISA against 2 μg of gliadin PWG per well; 2) the clones showing the highest signals against gliadin were sequenced to discard repeated events; 3) specificity of the clones was characterized (by their response to gluten-containing flours like wheat, E. Garcia-Calvo et al. Current Research in Food Science 7 (2023) 100578 4 rye, and barley, and gluten-free flours like oats, corn, and rice), and their affinity against the objective antigen (gliadin-PWG) was analyzed in the range of 0-5 μg/mL; and 4) the top performance clones were also characterized for their response to the experimental binary mixture of wheat, rye, and barley kernels in a rice-based gluten-free matrix. Origin 8.0 software was employed to plot and analyze all the experimental data. The mean values of three independent determinations and stan dard derivation of each data set are shown in the figures. 2.12. Analysis of food products The applicability of the leading phage-Fab candidate to detect gluten in complex matrixes, such as food samples, was analyzed by indirect- phage ELISA. Gluten-like proteins from ten commercial products were extracted in a two-steps procedure: 0.25 g of the sample was dissolved in 2.5 mL of extraction buffer SENSISpec INgezim (Ingenasa®, Madrid, Spain ref #30.GLU.K.2) and shaken for 20 min in a rotating shaker. Then, 7.5 mL of 80% ethanol/water solution was added, and the mixture was shaken for 20 additional minutes. The mixture was centrifuged at 2000 g, 10 min at room temperature. The analysis was performed by indirect-phage ELISA using the same protocol previously explained. 2.13. In silico elucidation of Fab-C structure and its interaction with the target antigen In silico elucidation of the structure of the leading candidate (Fab-C) and the antibody-antigen interaction was done to have a deeper char acterization of the antibody. The primary data input was the DNA sequence from Fab-C. From this sequence, the following pipeline was implemented: 1) the sequence was launched to IMGT/V-QUEST, an alignment-based tool for antibody domain definition and classification (Brochet et al., 2008); 2) the structure of the Fab was predicted using the bioinformatics tools Abodybuilder2 (Dunbar et al., 2016) and IgFold (Ruffolo et al., 2022); 3) concordances and discrepancies between both models were discerned using the Matchmaker tool from ChimeraX (Pettersen et al., 2021); 4) once the Fab models were obtained, the in teractions with the target antigen (gliadin) were elucidated by computing epitope and paratope predictions with the tools Epipred (Krawczyk et al., 2014) and Antibody i-Patch (Krawczyk et al., 2013); and 5) based on the information about the antigen-antibody obtained from the previous analysis and, using both models of the Fab, a global simulation of the interaction with the antigen was calculated by dock ing, using HADDOCK 2.4 server (van Zundert et al., 2016). 3. Results and discussion 3.1. Antibody library design strategy Several approaches have been proposed in the last decades to generate high-affinity antibodies against gluten. Most of them used classical strategies, like immunizing an animal with cereal or recombi nant proteins and obtaining monoclonal antibodies secreted by hy bridoma cells. Some recent approaches were based in developing recombinant antibodies with a variety of objectives. On one hand, im mune libraries derived from patients aimed to profile their humoral response in celiac disease. On the other hand, camelid (Doña et al., 2010) or semisynthetic antibody libraries have been used as source of recombinant probes for detection of gluten (Garcia-Calvo et al., 2020). A novel approach is proposed in this work, the construction of a new, fully human immune library from peripheral blood of celiac patients, with the objective of translating the humoral features found in the selected donors to build a platform for antibody discovery, with the final goal of developing accurate systems for gluten detection in food. 3.2. Clinical features and evaluation of the humoral response of selected patients Building an immune library with enough diversity to allow isolation from of high-affinity Fabs against gluten proteins that are toxic to celiac patients required the capture of coding sequences for antibodies that are involved in the humoral response against the target antigen. To fulfill this requirement, potential suitable donors with high humoral response to several antigens related to celiac pathology that usually correlates with a high response to gliadin were searched at the digestive medicine service of the Fundación Jiménez-Díaz Hospital (Madrid). Initial pa rameters for selection of donors included anti-tTG and anti-DGP IgG and IgA titers. These serological markers are widely used because their presence is correlated with the intestinal damage, and their sensitivity and specificity in detecting untreated celiac disease are close to or above 95% (Leffler and Schuppan, 2010). The utilization of these markers, particularly circulating anti-tTG, for predictive purposes, encounters challenges in establishing a standardized and universally applicable cutoff point. The intricacy arises from the lack of harmonization among the diverse assay methods available, compounded by the preponderance of research focused on the pediatric population, as evidenced in many published works (Beltran et al., 2014). Consequently, in the selection process for this work, our clinical collaborators adhered to their hospi tal’s clinical guidelines, which dictated the cut-offs adapted to their population’s context. In contemporary clinical practice, the detection of specific antibodies against complete gliadin in peripheral blood, has waned as a biomarker due to its diminished sensitivity in comparison to the currently employed biomarkers (Singh et al., 2022). Nonetheless, for the present work, this detection method assumed critical significance, given the role of gliadin as the primary antigen for the selection process. Thoroughly characterizing the humoral response of potential donors against gliadin was paramount, as it enabled the translation of these immune charac teristics into the library. Also, it was very important to assure that the patients had not started a gluten-free diet, because with the removal of gluten from their diet the inhibition of the abnormal celiac humoral response is very strong and sharp (Sharma et al., 2020). Four patients were included in the study as potential donors (Table 1). Patient 2 was discarded because the levels of IgG Anti-tTG were low, and IgA anti-gliadin was bordering the levels for positive consideration. Patient 3 presented high humoral response levels and a clear celiac diagnosis but could not be included because started a gluten- free diet before the blood donation. Finally, patients 1 and 4 fulfilled all the requirements and donated blood samples. In addition to these clinical features, a serum sample from the donors (patients 1 and 4) was tested to quantify their response against native gliadin by IgG isotype, with the aim to translate the heterogeneous antibody expression of the patients to the library, instead of inserting equimolar constructs of the possible isotypes and sub-isotypes (Fig. 1S). Patients 1 and 4 presented a different IgG isotype response against gliadin, with the usual dominant expression of IgG1 in patient 1 (Nahm et al., 1987; Napodano et al., 2021). Nevertheless, patient 4 presented also a very high expression of IgG2 isotype, an unusual feature in celiac disease (Husby et al., 1986) but was considered appropriate for the li brary construction. To transfer the affinity of the humoral response of celiac patients to gluten proteins, the Fab sequences included in the library were ampli fied from the mRNA extracted from lymphocytes, which encode the rearranged sequences from the recombination of the V, D, and J genes. In addition, obtaining the lymphocytes from peripheral blood assures the amplification of matured antibodies against the objective antigen (Victora and Nussenzweig, 2022). E. Garcia-Calvo et al. Current Research in Food Science 7 (2023) 100578 5 3.3. Lymphocyte RNA preparation and retro-transcription Mononuclear cells isolated from patients’ blood were 1.2 x 108 cells/ mL from patient 1 and 1.8 x 108 cells/mL from patient 4. Total RNA was isolated with TriZol LS reagent, and tested for integrity, showing high RIN (RNA integrity number) values of 9.5/10 for patient 1 and 8.75/10 for patient 4. The total RNA from each patient was retro-transcribed using general oligodT-20 primers, obtaining the cDNA scaffold for PCR amplification. 3.4. Antibody chain coding fragments amplification and library construction Amplification of the fragments coding for the antibody chains was performed in two sequential rounds. Due to the amplification bias dur ing the PCR, some chains can be overrepresented in comparison with those expressed in the donors. To avoid this issue, the same PCR con ditions were used for all chains, so different quantities of each chain (gene family and isotype) were obtained for cloning instead of an equimolar pool of chains. Separate amplifications were done for each combination of family genes and isotypes (e.g., VH135-IgG1; VH4-IgA …). This approach is novel and unique, first because of the use of the Fab format while most of the previous works were ScFvs (Marzari et al., 2001; Rhyner et al., 2003), and also because of the amplification of as many VH genes as possible to generate a wide diversity, in contrast to previous studies that focused on certain VH families like VH4 (Sblattero et al., 2000) and VH5 (Not et al., 2011). Fab Heavy chain amplicons (Fig. 2S A-B) were amplified with different terminal restriction sites than the light chains (Fig. 2S C-D) for sequential cloning. The second round of PCRs was only used to expand the neighboring sequences to the restriction point (addition of 39 bp) to improve the cloning steps. The amplification was performed with fewer cycles than the first step in order to avoid alteration of the first-round chain composition (Fig. 2S E). This cloning strategy was used to avoid the more classical protocol of building the Fabs by OE-PCR (overlap-PCR) that implied three rounds of amplification (first of the V and C genes from cDNA, second an overlap of these genes, and third the fusion of the light and heavy chains by a second overlap) (Barbas, 2001). The methodology proposed for this work avoids one amplification step and overlaps by directly amplifying the light and heavy chains of the Fabs and cloning them into the plasmid. This strategy avoids PCR errors, misrepresentation of minority chains, and truncated overlaps (Omar and Lim, 2018). The Fab library was built through a two-step cloning procedure; firstly, the light chains were introduced into the pComb3X vector and transformed into E. coli XL1-Blue cells. Plasmid DNA containing the light-chain sub-library was isolated from the culture for the subsequent cloning of the Fab-heavy chains, resulting in the Fab library that was transformed into E. coli. After the electroporation and recuperation in SOC media, repertoire size (RZ) of the library was calculated, being 1.1 x 107 total clones. Some random clones were sequenced to assure that the Fab were correctly cloned, showing a self-ligation of 4%. The RZ > 107 and autoligation level (<5%) were good enough to continue with the panning protocol (Barbas, 2001). 3.5. Panning (selection and amplification) of gliadin binding phage-Fab The Fab-library was screened against gliadin by phage-display. Three rounds of panning were conducted with increasing stringency by incrementing the number of washing steps. The output titers were measured by plating dilutions of the E. coli XL1-Blue culture infected with the eluted phages in SB-agar supplemented with 100 μg/mL car benicillin (Fig. 1A). A 19-fold increase was observed in round two and then, it doubled in round three. Although measuring the output titers was an adequate way to follow panning performance, partial characterization of the overall affinity of the selected, amplified, and isolated phage-Fabs from the library was measured by indirect polyclonal phage-ELISA (Fig. 1B). A three-fold signal increase was observed after round one, and the maximum signal of the test was obtained after round three. Despite the insignificant increase in the transformants recovered in the first round, the polyclonal ELISA showed a clear increase in the overall affinity. On the contrary, the higher number of transformants recovered in the second round was not impaired with a noticeable increment of the affinity, which could be explained because a wide population of antibodies has been amplified with a mixture of affinities. Focusing on the third round, a strong increase of the colonies recovered, and also in the signal of the ELISA, means that the high- affinity phage-Fabs against gliadin dominate in the population, and a monoclonal binder could be easily isolated from the pool. 3.6. Selection of gliadin binders by monoclonal phage ELISA Forty-eight individual clones were picked from the tittering plates of the output phage population from last panning rounds (36 from R3 and 12 from R4). Induced culture supernatants (non-precipitated but con taining phage-Fabs) were 1:2 diluted and subjected to monoclonal phage-ELISA in immunoplates coated with gliadin-PWG at a concen tration of 2 μg/well. Twenty-nine out of the forty-eight clones produced absorbance signals higher than 1.5 and did not bind the negative control (BSA). Those clones were selected as candidates for further analysis (Fig. 1C). 3.7. Sequencing and classification of the selected positive clones The 29 clones identified in the monoclonal phage-ELISA as gliadin PWG binders, were sequenced to detect repetitive clones, to identify clonotypes and for their classification by gene family, class, and sub class. From the 29 sequenced clones, there were only 14 unique se quences, hereinafter named A-N (Fig. 2A). One of the unique sequences (Fab-C) was heavily repeated in round four (with 9 clones (31%) presenting that sequence; Fab-H was found in 4 (14%) of the sequenced clones; Fab-N in 3 (10%); Fab-G and Fab-K two times each (6.89% each) and the remaining ones were found once (3.34% each). Two clones belonged to the IgA2 subclass (C and H), five to IgG1 (A, D, G, L and N), four to IgG2 (B, E, J and I), and three to IgA1 (F, K and M) for the heavy chain. Regarding the light chains, there was a total dominance of kappa chains. These unique sequences were orga nized in clonotypes (the clones belonging to the same clonotype pre sented a rearrangement with the same germlines V(D)J gene segments) Table 1 Analysis of the humoral response of potential celiac patients to be included in this study. Patient IgA Anti-tTG (U/mL) IgG Anti-tTG (U/mL) Selective IgA deficiency (negative/positive) IgA Anti-DGP (U/mL) IgG Anti-DGP (U/mL) IgA Anti-native Gliadin (U/mL) IgG Anti-native Gliadin (U/mL) 1 5654 <0.8 negative 1934 22.6 168 550 2 97.6 <0.8 negative 44.8 37.2 19 96 3 6994 49.6 negative 2165.1 537.6 35.37 200.70 4 3238 2 negative 1971 228.50 96.56 169.27 Cut off >15 >15 >15 >15 >15 >50 E. Garcia-Calvo et al. Current Research in Food Science 7 (2023) 100578 6 using IgBLAST. The 14 different clones were classified into 8 clonotypes for the heavy chains: clonotype 1H (B, E, I and J), clonotype 2H (F, K and M), clonotype 3H (A and G), clonotype 4H (C), clonotype 5H (H), clonotype 6H (D), clonotype 7H (L) and clonotype 8H (N). Three clonotypes (3H, 6H and 7H) presented VH1 genes, four VH3 (clonotypes 1H, 2H, 4H and 5H), and the remaining (clonotype 8H) VH4 genes for the heavy chains (Fig. 2B). The variability of the V, D, J genes whose rearrangement produced the Fabs is shown in Fig. 2B, 2C and 2D. On one hand, the vast majority of the gliadin-reactive Fabs presented VH3 genes (64.3% of the clones) even surpassing the frequency found in a normal human antibody repertoire (approximately 50%). Surprisingly, the most expressed family in human antibodies, VH3- 23, (Sun et al., 2022), was only represented by one clonotype (2H), including by three Fabs. A special remark must be noticed on the selection of Fab-C, that presents the VH3-15 locus, a scarcely expressed gene (3% of the human repertory) but one of the most identified loci in gluten-reactive anti bodies in celiac disease (Steinsbø et al., 2014). On the other hand, although VH4 antibodies represent almost 30% of the human repertory, only one selected Fab presented this gene (Tiller et al., 2013). The light chains could be classified into eight clonotypes: clonotype 1L (A, C and E); clonotype 2L (J and K); clonotype 3L (F and G); clonotype 4L (B and M); clonotype 5L (H and I), clonotype 6L (D), clo notype 7L (L) and clonotype 8L (N) (Fig. 3G). Six of the clonotypes for the light chain presented Vk1 genes (clonotypes 1L, 3L, 4L, 5L, 6L and 7L), one Vk2 (clonotype 8L) and one Vk3 (clonotype 2L). The frequency of the V and J genes whose rearrangement produced the light chain of the selected Fab are depicted in Fig. 2F and Fig. 2G. Most of the clonotypes in the selected Fabs presented Vk1 (78.6%), that is also prevalent in the human repertory (approximately the 45% of kappa chains) (Tiller et al., 2013). The length of HCDR3 of the selected Fabs was remarkably high (16 ± 4.7 amino acids), especially in two clonotypes (1H and 4H) that presented more than 20 amino acids, an unusual feature (Fig. 2E), when the average in human antibodies is only 14.8 amino acids (Joyce et al., 2020). A high variability was found in the length and composition of HCDR3, in contrast to the shorter and less diverse LCDR3 (8.5 ± 0.76) (Fig. 2H). 3.8. Characterization of gliadin-binding clones Affinity and specificity of individual gliadin-binding clones were assessed by indirect-phage ELISA. The optimal phage-Fab concentration to be used as primary antibody in the phage-ELISA was 8 x 109 phage particles per well. A dose-response curve against increasing concentrations of gliadin- Fig. 1. Characterization of the panning progress for selection of gliadin-binding phage-Fabs. (A) Phage titration after each round of panning. (B) Polyclonal indirect phage-ELISA analysis of the phage-Fab li brary isolated after each round of panning. Notice that round 4 is referring to the amplification of output phages from round 3. (C) Monoclonal indirect phage- ELISA analysis of individual clones isolated from the third and fourth rounds of affinity enrichment against gliadin-PWG. Origin 8.0 software was used to plot and analyze the experimental data. Mean values of three independent determinations and standard derivation of each data set are shown. E. Garcia-Calvo et al. Current Research in Food Science 7 (2023) 100578 7 PWG (0–5 μg/mL) was obtained for each phage variant (Fig. 3A), Fab-C stands out from the other Fabs, with a linear range between 0 and 1.25 μg/mL, reaching a saturation plateau at 2.5 μg/mL. In addition, a specificity test was conducted with different cereal flours to evaluate the performance of the selected clones in an actual food matrix, and to discriminate the cereal species recognized by the phage-Fabs. Fab-J, Fab-F and Fab-K showed cross-reactivity for oats; Fab-J, Fab-F and Fab-L for corn; Fab-J and Fab-F for rice extracts, these clones were not considered for further characterization because of their cross-reactivity with cereals tolerated by people suffering from GRDs. The four strongest candidates (Fab-C, Fab-E, Fab-H and Fab-G), were tested against the gluten-like proteins extracted from a binary mixture containing wheat, rye, and barley kernels (0; 0.1; 0.2; 0.5; 1; 2.5; 5; 10 and 100 mg/g) in a rice-based gluten-free matrix, equivalent to 0; 10; 20; 50; 100; 250, 500, 1000 and 10000 mg/kg of gluten (Fig. 3B) (García-García et al., 2020). From the results obtained, the clone Fab-C clearly showed the best features: high response to gliadin-PWG in solution (purified antigen), and selectivity against gluten-containing flours (wheat, rye, and barley). Fab-C also showed a high performance for detection of gluten in the experimental binary mixture, with a LOD of 15 mg/kg of gluten (Fig. 3C), that fulfills the sorting capacity between gluten-free and gluten-containing foods set by current legislation (where gluten-free are considered when the product is containing less than 20 mg/kg of gluten) Fig. 2. Main features of the sequencing and classifi cation of the selected Fabs. (A) Phylogenetic tree of isolated Fabs. An alignment of sequencing results was conducted with SnapGene® software, computing MUSCLE algorithm and then the phylogenetic tree was represented using the Simple Phylogeny tool (ClustalW2 package, EMBL-EBI). (B-H) Antibody se quences were analyzed, and clustering was calcu lated, using IgBLAST tool (NCBI) and Origin 8.0 software for data representation. The pie charts represent the frequency of the genes, and the depicted number refers for the clonotypes that presented each gene: (B) VH genes identification and frequency. (C) JH genes identification and frequency. (D) D genes identification and frequency. (E) Heavy chain CDR3 length distribution per clonotype. (F) VK genes identification and frequency. (G) JK genes identifi cation and frequency. (H) Light chain CDR3 length distribution per clonotype. E. Garcia-Calvo et al. Current Research in Food Science 7 (2023) 100578 8 (Codex-Alimentarius, 2018). Despite the existence of other antibody-based tests in the market with lower LOD for gluten detection (Scherf and Poms, 2016), Fab-C exhibits interesting attributes that warrant further exploration and advancement. Notably, Fab-C complies with the legal standards of detection, having a single paratopic interaction with the antigen, in contrast to monoclonal antibodies that typically require two interactions (which can be increased to four in a sandwich ELISA format). This distinctive feature of Fab-C results in highly efficient binding with gliadin, emphasizing its potential as a valuable candidate for gluten detection. 3.9. Analysis of commercial food products Once the Fab-C was selected as the best gluten-binding probe, the next step was to demonstrate that it could be applied in an indirect- phage ELISA to detect gluten-like proteins extracted from market products, using the experimental binary mixture as standard curve (Fig. 3C). Ten commercial products were analyzed, five of them were labeled as gluten-free (like nuts or plant based products) and other five declared to contain gluten (like bakery products) (Table 2). After calculating gluten concentration, the products were sorted in two categories ac cording to the current legislation: negative or gluten-free (less than 20 mg/kg of gluten) and positive or gluten-containing product (with more than 20 mg/kg). All products tested were labeled according to the cur rent legislation, and the results were confirmed by an R5 sandwich ELISA (the standard method for gluten detection in food) (Scherf and Poms, 2016). These results confirm that phage-Fab-C can be used as the main reactant of an indirect phage-ELISA to differentiate gluten-free products. 3.10. In silico elucidation of the structure of Fab-C and its interaction with gliadin An in-silico approach was considered the best way to study the mo lecular interaction between the Fab-C and gliadin, because there are no experimental data available of high-resolution structures (obtained by X-ray crystallography or cryo-microscopy) for either the antigen (gliadin) or the antibody (Fab-C). In addition, Alpha-Fold is a high-accuracy tool for predicting the 3D structure of single-chain proteins (David et al., 2022). The first step for modeling the interaction was to obtain their mo lecular structures. The structure of a wheat alpha/beta gliadin has been recently resolved with Alpha-Fold and uploaded to the UNIPROT data base (IDS: P18573 or GDA9_WHEAT, presenting 307 aminoacids and a molecular weight of 35 KDa). These proteins are considered intrinsically disordered (IDPs) (Markgren et al., 2020) and the in silico approach is, to the date, the best solution of their structure. Characterization of the Fab structure faced several challenges. A first Fig. 3. Characterization and selection of gliadin-binding Fabs by phage-ELISA. (A) Comparative dose-response curve of the fourteen leading candidates against gliadin-PWG dilutions (0–5 μg/mL). (B) Dose-response curve of the four strongest candidates (Fab-C, Fab-E, Fab-H and Fab-G) using an experimental rice–based binary mixture with increasing presence of a proportional mixture of wheat, barley and rye in the range of 0.1 to 100 mg/g. (C) Calibration curve obtained for gluten detection with the experimental mixture and the Fab-C in an indirect phage-ELISA. Origin 8.0 software was used to plot and analyze the experimental data. Mean values of three independent determinations and standard derivation of each data set are shown. Table 2 Results obtained for detection of gluten in ten commercial food products using the indirect phage-ELISA with Fab-C, and the sandwich ELISA with R5 mono clonal antibody as reference method. Sample Product type Labeled as “gluten free” R5 Phage-Fab-C 1 Vegan sausages no + + 2 Vegan Deli no + + 3 Surimi no + + 4 Nuts no + + 5 Gluten-free pasta no + + 6 Macarrons yes – – 7 Apple cake yes – – 8 Free-sugar biscuits yes – – 9 Caramel biscuits yes – – 10 Sponge cake yes – – E. Garcia-Calvo et al. Current Research in Food Science 7 (2023) 100578 9 attempt to solve the structure using Alpha-fold program showed many inconsistencies, which could be caused because the Fab is a multi-chain protein. Then, alternative machine-learning antibody training tools were used, like IgFold and AbodyBuilder2. To compare both results, the structures were over-imposed with the ChimeraX alignment-based tool Matchmaker, visualizing a moderate discrepancy between the models regarding the spatial distribution of the HCDR3 (Fig. 4A). In an attempt to check which model could be closer to an actual experimental struc ture, molecules containing HCDR3 with similar length (20 to 23-mer) and amino acid sequences were searched using the SabDab database (Schneider et al., 2021), but they were not found. Thus, both structures were considered for further work. To identify the epitopes recognized by the Fab-C, the zones of the gliadin molecule with higher probability to produce contacts between de Fab and the antigen were assessed to assure better docking. The EpiPred tool was used for the identification of contacts between each of the Fab- C models and the antigen, obtaining the same results regardless the Fab- C model used. The main contacts are more likely to occur on the C- terminus, with three “hot zones” of several amino acids that are more likely to be part of the epitope: Q153 to V163; I200 to Q207 and E253 to Q261 (Fig. 4B). Once reliable structural models have been obtained (from a database for the antigen and generated de novo for the Fab) and a predicted epitope identified, a model of interaction by docking was computed using HADDOCK 2.4 tool, adjusting the contact zones to the CDRs for both Fab models and the predicted epitopes in the antigen. Several docking models were computed, and the best are described by their quality scores: for the IgFold model (RMSD = 0.6 ± 0.3 Ȧ; Z- score = -1.2) and for the Abodybuilder2 model (RMSD = 5.8 ± 0.5 Ȧ; Z- score = -2) (Kufareva and Abagyan, 2011). Root Mean Square Deviation (RMSD) represents the average distance between atoms in superimposed protein structures, and is a crucial quality parameter for docking models. The model produced by IgFold demonstrates a high level of resolution, which proves valuable in both docking (where an RMSD <2 indicates successful docking), and even for crystallography modeling, despite being a computational model in this case (López-Camacho et al., 2016). Based on these scores, the antigen-Fab interaction is better modeled using the IgFold Fab model. Thanks to this high-quality model, several likely interaction points between the antigen and the antibody were identified. As expected, the main paratopes are found in the heavy chain. The HCDR1 (Fig. 4C) interacts with proline and glutamine residues (position 250–251), and this PQ pattern has been identified in several peptides related to celiac toxicity (Morón et al., 2008). The HCDR2 (Fig. 4D) mainly interacts with glutamines in several positions within helixes (positions 203, 207, 251 and 254). This could mean a strength of this Fab, as gluten-related proteins are very rich in this amino acid. The abnormal length of the HCDR3 (Fig. 4E) enables it to have double contact with the antigen. The proximal side of the HCDR3 loop (posi tions 109–114) could interact with 153-157 positions of the antigen (with a couple of glutamines) and the distal side (positions 105–109) with the residues asparagine 257, leucine 258 and glutamine 261 in a parallel helix regarding the previous one. Although the contribution of Fig. 4. Computational model of the structure and interactions of Fab-C. (A) Fv structure prediction using two modeling programs: IgFold (blue) and SAbPred (brown). Comparison by superimposing the models obtained using the Matchmaker tool from ChimeraX. HCDR3 is flagged with a black arrow. (B) Representation of the epitope prediction of α/β-gliadin. The residues colored in red are more likely to be part of the conformational epitope. Rep resentation of the main likely interaction points be tween Fab-C and depicted α/β-gliadin (C to F). The colored residues are part of the CDRs (yellow and flagged with a black arrow) or the antigen epitopes (red) in the docking model: (C) HCDR1; (D) HCDR2; (E) HDR3; (F) LCDR3. (For interpretation of the ref erences to color in this figure legend, the reader is referred to the Web version of this article.) E. Garcia-Calvo et al. Current Research in Food Science 7 (2023) 100578 10 the light chain to the interaction seems to be much weaker, there is a possible contact between the LCDR3 and the glycine 159 and arginine 160 (Fig. 4F), which probably help the overall stability of the union. Although this results are simulations, thanks to the quality of the docking procedure, it could be concluded that Fab-C is a good candidate as binder to the gliadin C-terminal domain, something unusual compared to other gluten detection antibodies, like R5 or G12 that bind in the N-terminal region. In that case, the antibody-antigen interactions were elucidated in vitro, using peptide arrays composed of overlapping 15-mer gluten peptide sequences (Röckendorf et al., 2017). The different binding profiles could be useful for development of immuno assays, as it is highly likely that Fab-C and other available antibodies could be combined in an immunoassay with enhanced features due to a multiple epitope detection. 4. Conclusions This work presents the first construction and application (by phage display) of a fully human immune library with the final goal of obtaining a novel tool for gluten detection in foods. The new Fab library has been built from peripheral blood lympho cytes from celiac patients and, using phage-display technology, the Fab- C was selected as suitable probe for detection of gluten. The Fab-C is an antibody fragment with some uncommon features: it belongs to IgA2 subclass, which is directly correlated to celiac pathology; it also belongs to VH3-15 subfamily (not highly expressed in the human repertoire, but highly represented after the selection process), and it has a lengthy 21- mer HCDR3. The Fab-C has demonstrated to detect gluten in solution, and also in an experimental matrix and in food samples. A phage-ELISA functional test has been developed, with a LOD of 15 mg/kg, which can discriminate gluten-free products according to the current legislation. A quick, simple, and universal pipeline, based in novel prediction tools, has been developed. Its application resulted in the generation of the structure of Fab-C that was suitable to be used to calculate the fea tures of a proposed antibody-antigen interaction. These results showed that Fab-C is able to interact with the C-terminus of gliadin, another unexpected feature compared to other characterized antibodies (like R5 or G12). In essence, this work introduces an innovative methodology for crafting antibody fragments tailored for application in food analysis systems. By harnessing the distinctive immune response features observed in patients with celiac disease, a comprehensive and original approach emerges, integrating principles from healthcare, molecular evolution, and food safety domains. Funding This work was funded by Ministerio de Ciencia e Innovación Grant: PID2021-122925OB-I00. PRE2018-08427 fellowship of the Ministerio de Ciencia e Innovación granted for Eduardo Garcia-Calvo. Institutional Review Board statement The study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board (or Ethics Committee) of Hospital Universitario Fundación Jimenez Díaz (protocol code October 2018 and date of approval, 24 September 2019, Madrid, Spain). Informed Consent Statement: Informed consent was obtained from all subjects involved in the study. Data Availability Statement: Data is contained within the article. CRediT authorship contribution statement Eduardo Garcia-Calvo: Methodology, Investigation, experimenta tion, Validation, Formal analysis, Writing – original draft, All authors have read and agreed to the published version of the manuscript. Aina García-García: Investigation, Formal analysis, and writing and revision, All authors have read and agreed to the published version of the manuscript. Santiago Rodríguez-Gómez: Investigation, and, Formal analysis, All authors have read and agreed to the published version of the manuscript. Sergio Farrais: celiac patient selection, All authors have read and agreed to the published version of the manuscript. Rosario Martín: Funding acquisition, and project management, Project administration, All authors have read and agreed to the published version of the manuscript. Teresa García: Supervision, writing and revision, Funding acquisition, and project management, Project administration, All authors have read and agreed to the published version of the manuscript. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Data availability No data was used for the research described in the article. Acknowledgments We thank the Immunology and Cytogenetic Services of Fundación Jiménez Díaz Hospital (Madrid) for the management of the clinical samples used in this work. Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.crfs.2023.100578. References Agardh, D., Lynch, K., Brundin, C., Ivarsson, S.A., Lernmark, Å., Cilio, C.M., 2006. Reduction of tissue transglutaminase autoantibody levels by gluten-free diet is associated with changes in subsets of peripheral blood lymphocytes in children with newly diagnosed coeliac disease. Clin. Exp. Immunol. 144 (1), 67–75. https://doi. org/10.1111/j.1365-2249.2006.03036.x. Andris-Widhopf, J., Rader, C., Steinberger, P., Fuller, R., Barbas Iii, C.F., 2000. Methods for the generation of chicken monoclonal antibody fragments by phage display. J. Immunol. Methods 242 (1–2), 159–181. https://doi.org/10.1016/s0022-1759(00) 00221-0. Barbas, C.F., 2001. Phage Display : a Laboratory Manual. Cold Spring Harbor Laboratory Press. Beltran, L., Koenig, M., Egner, W., Howard, M., Butt, A., Austin, M.R., Patel, D., Sanderson, R.R., Goubet, S., Saleh, F., Lavender, J., Stainer, E., Tarzi, M.D., 2014. High-titre circulating tissue transglutaminase-2 antibodies predict small bowel villous atrophy, but decision cut-off limits must be locally validated. Clin. Exp. Immunol. 176 (2), 190–198. https://doi.org/10.1111/cei.12249. Bradbury, A.R.M., Trinklein, N.D., Thie, H., Wilkinson, I.C., Tandon, A.K., Anderson, S., Bladen, C.L., Jones, B., Aldred, S.F., Bestagno, M., Burrone, O., Maynard, J., Ferrara, F., Trimmer, J.S., Görnemann, J., Glanville, J., Wolf, P., Frenzel, A., Wong, J., Koh, X.Y., Eng, H.Y., Lane, D., Lefranc, M.P., Clark, M., Dübel, S., 2018. When monoclonal antibodies are not monospecific: hybridomas frequently express additional functional variable regions. mAbs 10 (4), 539–546. https://doi.org/ 10.1080/19420862.2018.1445456. Brochet, X., Lefranc, M.P., Giudicelli, V., 2008. IMGT/V-QUEST: the highly customized and integrated system for IG and TR standardized V-J and V-D-J sequence analysis. Nucleic Acids Res. 36, W503–W508. https://doi.org/10.1093/nar/gkn316. Caio, G., Volta, U., Sapone, A., Leffler, D.A., De Giorgio, R., Catassi, C., Fasano, A., 2019. Celiac disease: a comprehensive current review. BMC Med. 17 (1) https://doi.org/ 10.1186/s12916-019-1380-z. Codex-Alimentarius, 2018. Standard for Foods for Special Dietary Use for Persons Intolerant to Gluten. CSX, pp. 118–1979. https://www.fao.org/fao-who-codexalim entarius/codex-texts/list-standards/es/. David, A., Islam, S., Tankhilevich, E., Sternberg, M.J.E., 2022. The AlphaFold database of protein structures: a biologist’s guide. J. Mol. Biol. 434 (2), 167336 https://doi.org/ 10.1016/j.jmb.2021.167336. Dunbar, J., Krawczyk, K., Leem, J., Marks, C., Nowak, J., Regep, C., Georges, G., Kelm, S., Popovic, B., Deane, C.M., 2016. SAbPred: a structure-based antibody prediction server. Nucleic Acids Res. 44 (W1), W474–W478. https://doi.org/10.1093/nar/ gkw361. E. Garcia-Calvo et al. https://doi.org/10.1016/j.crfs.2023.100578 https://doi.org/10.1016/j.crfs.2023.100578 https://doi.org/10.1111/j.1365-2249.2006.03036.x https://doi.org/10.1111/j.1365-2249.2006.03036.x https://doi.org/10.1016/s0022-1759(00)00221-0 https://doi.org/10.1016/s0022-1759(00)00221-0 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref3 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref3 https://doi.org/10.1111/cei.12249 https://doi.org/10.1080/19420862.2018.1445456 https://doi.org/10.1080/19420862.2018.1445456 https://doi.org/10.1093/nar/gkn316 https://doi.org/10.1186/s12916-019-1380-z https://doi.org/10.1186/s12916-019-1380-z https://www.fao.org/fao-who-codexalimentarius/codex-texts/list-standards/es/ https://www.fao.org/fao-who-codexalimentarius/codex-texts/list-standards/es/ https://doi.org/10.1016/j.jmb.2021.167336 https://doi.org/10.1016/j.jmb.2021.167336 https://doi.org/10.1093/nar/gkw361 https://doi.org/10.1093/nar/gkw361 Current Research in Food Science 7 (2023) 100578 11 Garcia-Calvo, E., García-García, A., Madrid, R., Martín, R., García, T., 2020. From polyclonal sera to recombinant antibodies: a review of immunological detection of gluten in foodstuff. Foods 10 (1), 66. https://doi.org/10.3390/foods10010066. Garcia-Calvo, E., García-García, A., Rodríguez, S., Farrais, S., Martín, R., García, T., 2022. Construction of a Fab library merging chains from semisynthetic and immune origin, suitable for developing new tools for gluten immunodetection in food. Foods 12 (1), 149. https://doi.org/10.3390/foods12010149. García-García, A., Madrid, R., García, T., Martín, R., González, I., 2018. Use of multiplex ligation-dependent probe amplification (MLPA) for screening of wheat, barley, rye and oats in foods. Food Control 84, 268–277. https://doi.org/10.1016/j. foodcont.2017.07.037. García-García, A., Madrid, R., González, I., García, T., Martín, R., 2020. A novel approach to produce phage single domain antibody fragments for the detection of gluten in foods. Food Chem. 321, 126685 https://doi.org/10.1016/j.foodchem.2020.126685. Husby, S., Foged, N., Oxelius, V.A., Svehag, S.E., 1986. Serum IgG subclass antibodies to gliadin and other dietary antigens in children with coeliac disease. Clin. Exp. Immunol. 64 (3), 526–535. Itzlinger, A., Branchi, F., Elli, L., Schumann, M., 2018. Gluten-free diet in celiac disease—forever and for all? Nutrients 10 (11), 1796. https://doi.org/10.3390/ nu10111796. Joyce, C., Burton, D.R., Briney, B., 2020. Comparisons of the antibody repertoires of a humanized rodent and humans by high throughput sequencing. Sci. Rep. 10 (1) https://doi.org/10.1038/s41598-020-57764-7. Kang, T.H., Seong, B.L., 2020. Solubility, stability, and avidity of recombinant antibody fragments expressed in microorganisms. Front. Microbiol. 11 https://doi.org/ 10.3389/fmicb.2020.01927. Krawczyk, K., Baker, T., Shi, J., Deane, C.M., 2013. Antibody i-Patch prediction of the antibody binding site improves rigid local antibody–antigen docking. Protein Eng. Des. Sel. 26 (10), 621–629. https://doi.org/10.1093/protein/gzt043. Krawczyk, K., Liu, X., Baker, T., Shi, J., Deane, C.M., 2014. Improving B-cell epitope prediction and its application to global antibody-antigen docking. Bioinformatics 30 (16), 2288–2294. https://doi.org/10.1093/bioinformatics/btu190. Kufareva, I., Abagyan, R., 2011. Methods of protein structure comparison. Methods Mol. Biol. 857, 231–257. https://doi.org/10.1007/978-1-61779-588-6_10. Leffler, D.A., Schuppan, D., 2010. Update on serologic testing in celiac disease. Am. J. Gastroenterol. 105 (12), 2520–2524. https://doi.org/10.1038/ajg.2010.276. López-Camacho, E., García-Godoy, M.J., García-Nieto, J., Nebro, A.J., Aldana-Montes, J. F., 2016. In: A New Multi-Objective Approach for Molecular Docking Based on RMSD and Binding Energy. Springer International Publishing, pp. 65–77. https:// doi.org/10.1007/978-3-319-38827-4_6. Markgren, J., Hedenqvist, M., Rasheed, F., Skepö, M., Johansson, E., 2020. Glutenin and gliadin, a piece in the puzzle of their structural properties in the cell described through monte carlo simulations. Biomolecules 10 (8), 1095. https://doi.org/ 10.3390/biom10081095. Marzari, R., Sblattero, D., Florian, F., Tongiorgi, E., Not, T., Tommasini, A., Ventura, A., Bradbury, A., 2001. Molecular dissection of the tissue transglutaminase autoantibody response in celiac disease. J. Immunol. 166 (6), 4170–4176. https:// doi.org/10.4049/jimmunol.166.6.4170. Morón, B., Bethune, M.T., Comino, I., Manyani, H., Ferragud, M., López, M.C., Cebolla, Á., Khosla, C., Sousa, C., 2008. Toward the assessment of food toxicity for celiac patients: characterization of monoclonal antibodies to a main immunogenic gluten peptide. PLoS One 3 (5), e2294. https://doi.org/10.1371/journal. pone.0002294. Nahm, M.H., Scott, M.G., Shackelford, P.G., 1987. Expression of human IgG subclasses. Ann. Clin. Lab. Sci. 17 (3), 183–196. Napodano, C., Marino, M., Stefanile, A., Pocino, K., Scatena, R., Gulli, F., Rapaccini, G.L., Delli Noci, S., Capozio, G., Rigante, D., Basile, U., 2021. Immunological role of IgG subclasses. Immunology Investigations 50 (4), 427–444. https://doi.org/10.1080/ 08820139.2020.1775643. Not, T., Ziberna, F., Vatta, S., Quaglia, S., Martelossi, S., Villanacci, V., Marzari, R., Florian, F., Vecchiet, M., Sulic, A.-M., Ferrara, F., Bradbury, A., Sblattero, D., Ventura, A., 2011. Cryptic genetic gluten intolerance revealed by intestinal antitransglutaminase antibodies and response to gluten-free diet. Gut 60, 1487–1493. https://doi.org/10.1136/gut.2010.232900. Omar, N., Lim, T.S., 2018. Construction of naive and immune human Fab phage-display library. Methods Mol. Biol. 1701, 25–44. https://doi.org/10.1007/978-1-4939- 7447-4_2. Peña, A.S., Rodrigo, L., 2015. Epidemiology of celiac disease and non-celiac gluten- related disorders. In: Advances in the Understanding of Gluten Related Pathology and the Evolution of Gluten-free Foods. Omniascience. Pettersen, E.F., Goddard, T.D., Huang, C.C., Meng, E.C., Couch, G.S., Croll, T.I., Morris, J. H., Ferrin, T.E., 2021. UCSF ChimeraX: structure visualization for researchers, educators, and developers. Protein Sci. 30 (1), 70–82. https://doi.org/10.1002/ pro.3943. Raiteri, A., Granito, A., Giamperoli, A., Catenaro, T., Negrini, G., Tovoli, F., 2022. Current guidelines for the management of celiac disease: a systematic review with comparative analysis. World J. Gastroenterol. 28 (1), 154–176. https://doi.org/ 10.3748/wjg.v28.i1.154. Rhyner, C., Weichel, M., Hübner, P., Achatz, G., Blaser, K., Crameri, R., 2003. Phage display of human antibodies from a patient suffering from coeliac disease and selection of isotype-specific scFv against gliadin. Immunology 110 (2), 269–274. https://doi.org/10.1046/j.1365-2567.2003.01728.x. Röckendorf, N., Meckelein, B., Scherf, K.A., Schalk, K., Koehler, P., Frey, A., 2017. Identification of novel antibody-reactive detection sites for comprehensive gluten monitoring. PLoS One 12 (7), e0181566. https://doi.org/10.1371/journal. pone.0181566. Ruffolo, J.A., Chu, L.-S., Mahajan, S.P., Gray, J.J., 2022. Fast, Accurate Antibody Structure Prediction from Deep Learning on Massive Set of Natural Antibodies. Cold Spring Harbor Laboratory. https://doi.org/10.1101/2022.04.20.488972. Sblattero, D., Florian, F., Not, T., Ventura, A., Bradbury, A., Marzari, R., 2000. Analyzing the peripheral blood antibody repertoire of a celiac disease patient using phage antibody libraries. Hum. Antibodies 9, 199–205. https://doi.org/10.3233/HAB- 2000-9402. Scherf, K.A., Poms, R.E., 2016. Recent developments in analytical methods for tracing gluten. J. Cereal. Sci. 67, 112–122. https://doi.org/10.1016/j.jcs.2015.08.006. Schneider, C., Raybould, M.I.J., Deane, C.M., 2021. SAbDab in the age of biotherapeutics: updates including SAbDab-nano, the nanobody structure tracker. Nucleic Acids Res. 50 (D1), D1368–D1372. https://doi.org/10.1093/nar/gkab1050. Sharma, N., Bhatia, S., Chunduri, V., Kaur, S., Sharma, S., Kapoor, P., Kumari, A., Garg, M., 2020. Pathogenesis of celiac disease and other gluten related disorders in wheat and strategies for mitigating them. Front. Nutr. 7 https://doi.org/10.3389/ fnut.2020.00006. Singh, P., Singh, A.D., Ahuja, V., Makharia, G.K., 2022. Who to screen and how to screen for celiac disease. World J. Gastroenterol. 28 (32), 4493–4507. https://doi.org/ 10.3748/wjg.v28.i32.4493. Steinsbø, Ø., Dunand, C.J.H., Huang, M., Mesin, L., Salgado-Ferrer, M., Lundin, K.E.A., Jahnsen, J., Wilson, P.C., Sollid, L.M., 2014. Restricted VH/VL usage and limited mutations in gluten-specific IgA of coeliac disease lesion plasma cells. Nat. Commun. 5 (1) https://doi.org/10.1038/ncomms5041. Sun, Z., Li, W., Mellors, J.W., Orentas, R., Dimitrov, D.S., 2022. Construction of a large size human immunoglobulin heavy chain variable (VH) domain library, isolation and characterization of novel human antibody VH domains targeting PD-L1 and CD22. Front. Immunol. 13, 869825 https://doi.org/10.3389/fimmu.2022.869825. Taraghikhah, N., Ashtari, S., Asri, N., Shahbazkhani, B., Al-Dulaimi, D., Rostami- Nejad, M., Rezaei-Tavirani, M., Razzaghi, M.R., Zali, M.R., 2020. An updated overview of spectrum of gluten-related disorders: clinical and diagnostic aspects. BMC Gastroenterol. 20 (1) https://doi.org/10.1186/s12876-020-01390-0. Tiller, T., Schuster, I., Deppe, D., Siegers, K., Strohner, R., Herrmann, T., Berenguer, M., Poujol, D., Stehle, J., Stark, Y., Heßling, M., Daubert, D., Felderer, K., Kaden, S., Kölln, J., Enzelberger, M., Urlinger, S., 2013. A fully synthetic human Fab antibody library based on fixed VH/VL framework pairings with favorable biophysical properties. mAbs 5 (3), 445–470. https://doi.org/10.4161/mabs.24218. Valdés, I., García, E., Llorente, M., Méndez, E., 2003. Innovative approach to low-level gluten determination in foods using a novel sandwich enzyme-linked immunosorbent assay protocol. Eur. J. Gastroenterol. Hepatol. 15 (5), 465–747. https://doi.org/10.1097/01.meg.0000059119.41030.df. van Zundert, G.C.P., Rodrigues, J.P.G.L.M., Trellet, M., Schmitz, C., Kastritis, P.L., Karaca, E., Melquiond, A.S.J., van Dijk, M., de Vries, S.J., Bonvin, A.M.J.J., 2016. The HADDOCK2.2 web server: user-friendly integrative modeling of biomolecular complexes. J. Mol. Biol. 428 (4), 720–725. https://doi.org/10.1016/j. jmb.2015.09.014. Victora, G.D., Nussenzweig, M.C., 2022. Germinal centers. Annu. Rev. Immunol. 40 (1), 413–442. https://doi.org/10.1146/annurev-immunol-120419-022408. E. Garcia-Calvo et al. https://doi.org/10.3390/foods10010066 https://doi.org/10.3390/foods12010149 https://doi.org/10.1016/j.foodcont.2017.07.037 https://doi.org/10.1016/j.foodcont.2017.07.037 https://doi.org/10.1016/j.foodchem.2020.126685 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref15 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref15 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref15 https://doi.org/10.3390/nu10111796 https://doi.org/10.3390/nu10111796 https://doi.org/10.1038/s41598-020-57764-7 https://doi.org/10.3389/fmicb.2020.01927 https://doi.org/10.3389/fmicb.2020.01927 https://doi.org/10.1093/protein/gzt043 https://doi.org/10.1093/bioinformatics/btu190 https://doi.org/10.1007/978-1-61779-588-6_10 https://doi.org/10.1038/ajg.2010.276 https://doi.org/10.1007/978-3-319-38827-4_6 https://doi.org/10.1007/978-3-319-38827-4_6 https://doi.org/10.3390/biom10081095 https://doi.org/10.3390/biom10081095 https://doi.org/10.4049/jimmunol.166.6.4170 https://doi.org/10.4049/jimmunol.166.6.4170 https://doi.org/10.1371/journal.pone.0002294 https://doi.org/10.1371/journal.pone.0002294 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref27 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref27 https://doi.org/10.1080/08820139.2020.1775643 https://doi.org/10.1080/08820139.2020.1775643 https://doi.org/10.1136/gut.2010.232900 https://doi.org/10.1007/978-1-4939-7447-4_2 https://doi.org/10.1007/978-1-4939-7447-4_2 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref31 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref31 http://refhub.elsevier.com/S2665-9271(23)00146-6/sref31 https://doi.org/10.1002/pro.3943 https://doi.org/10.1002/pro.3943 https://doi.org/10.3748/wjg.v28.i1.154 https://doi.org/10.3748/wjg.v28.i1.154 https://doi.org/10.1046/j.1365-2567.2003.01728.x https://doi.org/10.1371/journal.pone.0181566 https://doi.org/10.1371/journal.pone.0181566 https://doi.org/10.1101/2022.04.20.488972 https://doi.org/10.3233/HAB-2000-9402 https://doi.org/10.3233/HAB-2000-9402 https://doi.org/10.1016/j.jcs.2015.08.006 https://doi.org/10.1093/nar/gkab1050 https://doi.org/10.3389/fnut.2020.00006 https://doi.org/10.3389/fnut.2020.00006 https://doi.org/10.3748/wjg.v28.i32.4493 https://doi.org/10.3748/wjg.v28.i32.4493 https://doi.org/10.1038/ncomms5041 https://doi.org/10.3389/fimmu.2022.869825 https://doi.org/10.1186/s12876-020-01390-0 https://doi.org/10.4161/mabs.24218 https://doi.org/10.1097/01.meg.0000059119.41030.df https://doi.org/10.1016/j.jmb.2015.09.014 https://doi.org/10.1016/j.jmb.2015.09.014 https://doi.org/10.1146/annurev-immunol-120419-022408 Development of a new recombinant antibody, selected by phage-display technology from a celiac patient library, for detectio ... 1 Introduction 2 Materials and methods 2.1 Bacterial strains, plasmids, and growth media 2.2 Antigen preparations 2.3 Selection of celiac patients as suitable lymphocytes donors 2.4 Lymphocyte RNA preparation and retro-transcription 2.5 Amplification and isolation of antibody chain coding fragments 2.6 Library construction (clonation and transformation) 2.7 Selection of gliadin-binding phage-Fabs 2.8 Characterization individual gliadin-binding phage-Fabs 2.9 Indirect phage-ELISA 2.10 DNA isolation and sequencing 2.11 Searching and characterization of the gluten-binding Fabs 2.12 Analysis of food products 2.13 In silico elucidation of Fab-C structure and its interaction with the target antigen 3 Results and discussion 3.1 Antibody library design strategy 3.2 Clinical features and evaluation of the humoral response of selected patients 3.3 Lymphocyte RNA preparation and retro-transcription 3.4 Antibody chain coding fragments amplification and library construction 3.5 Panning (selection and amplification) of gliadin binding phage-Fab 3.6 Selection of gliadin binders by monoclonal phage ELISA 3.7 Sequencing and classification of the selected positive clones 3.8 Characterization of gliadin-binding clones 3.9 Analysis of commercial food products 3.10 In silico elucidation of the structure of Fab-C and its interaction with gliadin 4 Conclusions Funding Institutional Review Board statement CRediT authorship contribution statement Declaration of competing interest Data availability Acknowledgments Appendix A Supplementary data References