Mitigating the Affinity/Specificity Trade-Off: Production of Monoclonal Antibodies with High Affinity and Fine Specificity to Sulfonamides by Ligand- and Receptor-Based Rational Hapten Design

Yingjie Zhang; Chenglong Li; Xiya Zhang; Changfei Duan; Qing Shen; Weilin Wu; Xuezhi Yu; Kai Wen; Jianzhong Shen; Zhanhui Wang

doi:10.1016/j.eng.2026.02.036

Engineering ›› 2026, Vol. 63 ›› Issue (8) :300 -315. DOI: 10.1016/j.eng.2026.02.036

Research

research-article

Mitigating the Affinity/Specificity Trade-Off: Production of Monoclonal Antibodies with High Affinity and Fine Specificity to Sulfonamides by Ligand- and Receptor-Based Rational Hapten Design

Author information +

History +

PDF (4838KB)

Abstract

The trade-off between affinity and specificity in molecular recognition elements (MREs), including antibodies, protein receptors, and aptamers, represents a well-documented challenge. Achieving MREs that combine high affinity with fine specificity is particularly difficult when targeting structurally similar analytes. To address this limitation, we propose a novel ligand- and receptor-based rational hapten design (LRRHD) strategy to regulate the generation of monoclonal antibodies (mAbs) exhibiting both high affinity and fine specificity, using sulfonamides (SAs) as model target analytes. This strategy enabled the identification of several novel haptens containing rigid spacer arms and led to the generation of mAb 10E6, which unexpectedly exhibited half-maximal inhibitory concentration values ranging from 0.23 to 20 μg∙L⁻¹ across 29 tested SAs. The molecular recognition mechanisms underlying 10E6 binding were elucidated through crystal structure determination and molecular dynamics (MD) simulations. The results revealed that the extended complementarity determining region 3 of the heavy chain (CDRH3) of 10E6 forms a broader and more flexible ligand-binding pocket, allowing accommodation of diverse SAs with high affinity. In addition, mAb 10E6 was applied in an immunoassay, demonstrating limits of detection ranging from 0.29 to 4.7 μg∙kg⁻¹ in skimmed milk and chicken samples. This study presents an effective rational hapten design strategy to mitigate the affinity/specificity trade-off and provides new insights into antibody discovery and vaccine development targeting small molecules.

Graphical abstract

Keywords

Molecular recognition elements / Antibody discovery / High affinity / Fine specificity / Rational hapten design / Molecular recognition mechanism / Crystal structure

Cite this article

Download citation ▾

Yingjie Zhang, Chenglong Li, Xiya Zhang, Changfei Duan, Qing Shen, Weilin Wu, Xuezhi Yu, Kai Wen, Jianzhong Shen, Zhanhui Wang. Mitigating the Affinity/Specificity Trade-Off: Production of Monoclonal Antibodies with High Affinity and Fine Specificity to Sulfonamides by Ligand- and Receptor-Based Rational Hapten Design. Engineering, 2026, 63 (8) : 300-315 DOI:10.1016/j.eng.2026.02.036

登录浏览全文

4963

注册一个新账户忘记密码

1. Introduction

Molecular recognition-based detection technologies have been widely applied across diverse fields, including life sciences, clinical diagnostics, food safety, and environmental monitoring [1], [2], [3]. Molecular recognition elements (MREs)—such as antibodies, protein receptors, and aptamers—are central to these technologies, as they play a critical role in determining assay sensitivity and specificity. However, a persistent challenge associated with MREs is the intrinsic trade-off between affinity and specificity toward target analytes [4], [5]. This issue becomes particularly pronounced when detecting classes of structurally related compounds, such as sulfonamides (SAs), β-lactams (β-Ls), quinolones (QNs), organophosphorus pesticides (OPs), and polycyclic aromatic hydrocarbons (PAHs). Nevertheless, simultaneous detection of analyte classes requires both high affinity and fine specificity to ensure consistent recognition within the target group and to minimize cross-reactivity (CR) with unrelated compounds, thereby preserving the analytical accuracy of the developed methods.

Despite advances in MRE-based detection technologies, current MREs remain unable to fully overcome the affinity/specificity trade-off. For example, although antibodies typically exhibit high affinity, their binding affinities toward structurally similar compounds can vary substantially [6]. In our previous work, the monoclonal antibody (mAb) 4D11 was generated and shown to recognize 26 SAs with half-maximal inhibitory concentration (IC₅₀) values ranging from 1.2 to 11 000 μg∙L⁻¹, corresponding to an affinity gap of 9500-fold [7]. Similarly, other mAbs against SAs, OPs, or PAHs exhibited IC₅₀ values at the μg∙L⁻¹ level, with affinity gaps of approximately 100-fold (Table S1 in Appendix A) [8], [9], [10]. In contrast, protein receptors and enzymes used for multi-analyte recognition in bioassays tend to recognize structurally similar analytes more uniformly, but generally with lower affinity [11]. In our previous studies, penicillin-binding proteins were successfully applied in microplate assays for β-Ls screening [12], [13]. Building on this work, dihydropteroate synthase (DHPS) was employed in biosensors for the simultaneous recognition of at least 29 SAs, yielding IC₅₀ values ranging from 4.9 to 108 μg∙L⁻¹ and an affinity gap of 22-fold [14], [15]. A similar study reported that a QN-binding protein receptor recognized eight QNs with IC₅₀ values between 27 and 58 μg∙L⁻¹, exhibiting only a two-fold affinity gap but markedly lower overall affinities compared with antibodies [16]. Furthermore, aptamers targeting SAs, QNs, and OPs generally displayed low binding affinities in the micromolar range or exhibited limited recognition spectra (Table S1) [17], [18], [19], [20]. Consequently, the affinity/specificity trade-off of MREs is widely acknowledged, and the discovery of MREs combining high affinity with fine specificity remains a formidable challenge.

Among the available MREs, antibodies—owing to their high affinity, structural diversity, and adaptability—represent the most promising candidates for overcoming the affinity/specificity trade-off. In this study, we sought to develop antibodies exhibiting both high affinity and fine specificity through a rational hapten design strategy, using SAs as target analytes. SAs are a class of synthetic antimicrobial agents with numerous variants and are widely used for prophylactic and therapeutic purposes in humans and food-producing animals. Owing to their broad-spectrum efficacy and cost-effectiveness, SAs accounted for 9.4% of total veterinary medicines consumed in food-producing animals across 31 European countries in 2022 [21]. In addition, SAs have been frequently detected in livestock products and aquatic environments at concentrations ranging from μg∙kg⁻¹ to g∙kg⁻¹ [22], [23]. Sulfadiazine (SDZ), sulfadimethoxine (SDM), sulfapyridine (SPY), sulfamethazine (SMZ), and sulfamonomethoxine (SMM) are among the most commonly detected SAs in pork, chicken, and fish samples [23], [24], [25]. Improper use of veterinary drugs, whether above or below recommended levels, can pose short- or long-term public health risks, including the emergence of resistant microorganisms, toxicity, allergic reactions, mutagenesis, teratogenicity, and carcinogenicity [26]. Accordingly, most countries—including China, the United States, and members of the European Union—have established maximum residue limits (MRLs) for SAs in food, ranging from 10 to 100 μg∙kg⁻¹ [27], [28], [29], [30], [31]. The development of SA antibodies combining high affinity with fine specificity is therefore essential for effective monitoring of multiple SA residues.

The first and most critical step in developing antibodies against small molecules is hapten design. SAs share a common p-aminobenzenesulfonamide moiety and differ at the N¹ position, where various R substituents are present (Fig. S1 in Appendix A). To date, two major classes of SA haptens have been reported (Table S2 in Appendix A). The first class exposes the characteristic structure of individual SAs to the immune system and is typically referred to as specific haptens. These haptens are directly conjugated through the aromatic primary amino group at the N⁴ position—either without a spacer arm (subclass 1), such as sulfanilamide (SN) [32], phthalylsulfathiazole (PST) [33], [34], sulfaguanidine (SG) [7], and sulfamethoxazole (SMX) [7], or with a spacer arm (subclass 2), as exemplified by S7 [34], S8 [34], SA3 [35], SA4 [35], SA5 [35], and SA6 [35] (Table S2). Although antibodies elicited by these haptens can exhibit high specificity toward their corresponding SAs, they often lack the fine specificity required for broad recognition across structurally diverse SAs [7]. The second class comprises generic haptens that expose the common aromatic amine group to the immune system while employing conjugation through the variable R group at the N¹ position. As summarized in Table S2, 21 generic SA haptens have been reported and can be categorized into three subclasses. The first subclass includes single-ring haptens lacking an R substituent, such as BS [7], [36], [37], [38], HS [7], [8], [36], [37], [39], [40], [41], SA1-CH₂- [36], and SA2 [35], which have been considered promising candidates for generating broad-specific antibodies [36]. However, previous studies demonstrated that these haptens fail to elicit antibodies combining high affinity with broad specificity [42]. The second subclass consists of two-ring haptens, including CS [7], [8], [32], [33], [36], [37], [39], [41], [43], [44], [45], [46], S9 [34], [35], SA10 [7], [8], [35], [40], S4 [33], [34], [36], [37], [45], SA1 [35], S5 [35], H1-Z [45], [8], S7-C [8], TS [7], [8], [32], [33], [34], [37], [41], [43], [47], [48], [49], and SA10-X [42]. As summarized in Table S3 in Appendix A, mAb 4E5 elicited by hapten H1-Z [45], mAb 3D1 from hapten S5 [8], and the mAb generated from hapten SBA-L [46] exhibited relatively higher affinity and broader specificity. Nevertheless, these haptens remain insufficient to fully overcome the affinity/specificity trade-off. The third subclass comprises three-ring haptens, such as PS [32], [48], [49], [50] and H2 [39], [40]. Although the benzene ring also functions as a rigid spacer arm, antibodies raised against PS [32], [48], [50] and H2 [39], [40] (Table S2) did not display the anticipated affinity and specificity. Despite extensive efforts involving diverse hapten designs, challenges such as low affinity and unexpected specificity of SA antibodies persist.

A key limitation in effective hapten design arises from conventional ligand-based strategies, in which hapten structures typically mimic only the two-dimensional (2D) or three-dimensional (3D) conformations of SAs, often neglecting receptor-based information that is critical for eliciting broadly specific antibodies [51]. Our previous study comparing the structural features of mAbs 4C7 and 4D11 with DHPS demonstrated that the intrinsic difficulty in generating broadly specific antibodies stems from the diverse orientations, stereochemical configurations, and electronic properties of the R group at the N¹ position across different SAs [6]. Therefore, incorporating detailed receptor information—including that derived from mAbs and protein receptors—into rational hapten design holds significant promise for developing more effective haptens.

In this study, we propose a novel ligand- and receptor-based rational hapten design (LRRHD) strategy, which enabled the discovery of four novel haptens and the generation of mAbs exhibiting both high affinity and fine specificity. Furthermore, the molecular recognition mechanisms of the mAb were elucidated through crystal structure determination, molecular docking, and molecular dynamics (MD) simulations. The performance of the mAb was further validated by developing a mAb-based immunoassay, demonstrating its effectiveness for detecting multiple SA residues in real samples.

2. Methods

2.1. Geometry optimization and molecular cluster of SAs

The low-energy conformations of all SAs and the haptens conjugated with lysine were optimized using density functional theory (DFT) at the def-TZVP level with the M06-2X hybrid functional, as implemented in Gaussian 09 (Gaussian, Inc., USA). Vibrational analysis at the M06-2X/def-TZVP level confirmed the absence of imaginary frequencies, indicating that all optimized structures correspond to true minima. 3D molecular structures were visualized using CYLview (Legault, Canada). Molecular alignment based on the shared p-aminobenzenesulfonamide moiety was performed in SYBYL-X (Certara, USA). Quantitative analysis of electrostatic potentials (ESPs) was performed using Multiwfn (Beijing Kein Research Center for Natural Sciences, China), and ESP surfaces were visualized with VMD (University of Illinois at Urbana-Champaign, USA).

After optimizing low-energy SA conformations, the C(aryl)-S-N-C torsion angles were calculated using GaussView. The torsion-angle dataset for the 29 SAs was then subjected to unsupervised clustering analysis to identify representative conformational patterns. K-means clustering was performed using the K-means algorithm implemented in the scikit-learn package. The optimal number of clusters (K) was selected based on the Calinski-Harabasz index evaluated over a range of K values.

2.2. Preparation and identification of haptens and conjugates

Hapten H1 was synthesized in five reaction steps via an addition reaction between SDZ and 4-pentynoic acid (Fig. S2 in Appendix A). Detailed synthesis procedures are provided in the Appendix A. ¹H NMR (400 MHz, DMSO-d₆) δ 2.50 (m, 2H, -CH₂CH₂COOH), 2.63 (t, J = 7.2 Hz, 2H, -CH₂COOH), 6.04 (s, 2H, -NH₂), 6.55 (d, J = 8.8 Hz, 2H, -CHar), 7.59 (d, J = 8.4 Hz, 2H, -CHar), 8.50 (s, 2H, -CHar), 11.59 (s, br, 1H), 12.42 (s, br, 1H). ¹³C NMR (100 MHz, DMSO-d₆) δ 15.32 (-CH₂COOH), 33.25 (-CH₂CH₂COOH), 74.73 (-CCCH₂CH₂COOH), 94.93 (-CCH₂CH₂COOH), 112.63 (-Car), 112.89 (-Car), 124.97 (-Car), 130.32 (-Car), 153.65 (-Car), 155.88 (-Car), 160.41 (-Car), 173.26 (-COOH).

Hapten H2 was synthesized de novo in nine reaction steps (Fig. S3 in Appendix A). Detailed synthesis procedures are provided in Appendix A. ¹H NMR (400 MHz, DMSO-d₆) δ 1.33-1.41 (m, 2H, -CH₂CH₂CH₂COOH), 1.52-1.59 (m, 2H, -CH₂CH₂COOH), 2.21-2.26 (m, 2H, -CH₂CH₂CH₂CH₂COOH), 2.29 (s, 6H, -CH₃), 2.44-2.48 (m, 2H, -CH₂COOH), 5.94 (s, 2H, -NH₂), 6.54 (d, J = 8.8 Hz, 2H, -CHar), 7.63 (d, J = 8.4 Hz, 2H, -CHar), 11.42 (s, br, 2H, -NH, -COOH).

Hapten H3 was synthesized in four reaction steps (Fig. S4 in Appendix A). Detailed synthesis procedures are provided in the Appendix A. ¹H NMR (400 MHz, DMSO-d₆) δ 1.47-1.48 (m, 4H, -CH₂CH₂CH₂COOH), 2.20 (t, J = 6.0 Hz, 2H, -CH₂COOH), 2.45 (t, J = 6.8 Hz, 2H, -CH₂CH₂CH₂CH₂COOH), 5.94 (s, 2H, -NH₂), 6.54 (d, J = 8.8 Hz, 2H, -CHar), 7.01 (d, J = 8.8 Hz, 1H, -CHar), 7.49-7.51 (m, 3H, -CHar), 7.94 (d, J = 1.2 Hz, 1H, -CHar), 11.40 (s, br, 2H, -NH, -COOH). ¹³C NMR (100 MHz, DMSO-d₆) δ 23.87 (-CH₂CH₂COOH), 29.91 (-CH₂CH₂CH₂COOH), 30.78 (-CH₂CH₂CH₂CH₂COOH), 33.31 (-CH₂COOH), 111.82 (-Car), 112.37 (-Car), 125.65 (-Car), 128.77 (-Car), 130.74 (-Car), 138.53 (-Car), 145.78 (-Car), 150.28 (-Car), 152.62 (-Car), 174.31 (-COOH).

Hapten H4 was synthesized in five reaction steps (Fig. S5 in Appendix A). Detailed synthesis procedures are provided in the Appendix A. ¹H NMR (400 MHz, DMSO-d₆) δ 1.42-1.51 (m, 4H, -CH₂CH₂CH₂COOH), 2.22 (t, J = 7.2 Hz, 2H, -CH₂COOH), 2.28 (s, 3H, -CH₃), 2.42 (t, J = 7.2 Hz, 2H, -CH₂CH₂CH₂CH₂COOH), 5.89 (s, 2H, -NH₂), 6.54 (d, J = 8.8 Hz, 2H, -CHar), 6.87 (d, J = 8.4 Hz, 1H, -CHar), 7.39 (d, J = 8.4 Hz, 1H, -CHar), 7.50 (d, J = 8.8 Hz, 2H, -CHar), 11.52 (s, br, 2H, -NH, -COOH). ¹³C NMR (100 MHz, DMSO-d₆) δ 19.86 (-CH₃), 24.13 (-CH₂CH₂COOH), 28.79 (-CH₂CH₂CH₂CH₂COOH), 30.23 (-CH₂CH₂CH₂COOH), 33.35 (-CH₂COOH), 109.55 (-Car), 112.27 (-Car), 126.43 (-Car), 127.39 (-Car), 128.76 (-Car), 139.71 (-Car), 150.30 (-Car), 151.81 (-Car), 152.37 (-Car), 174.32 (-COOH).

Subsequently, the N-hydroxysuccinimide (NHS) ester method was used to covalently conjugate the terminal carboxylic acid group of each hapten to carrier proteins, following our previously published protocol [52]. Hapten-keyhole limpet hemocyanin (KLH) bioconjugates were used as immunogens for antibody production, whereas hapten-bovine serum albumin (BSA) conjugates served as coating antigens for assay development. Hapten-BSA bioconjugates were characterized by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. Hapten density, defined as the number of hapten molecules per mole of carrier protein, was calculated using Eq. (1):

(1)$ \text { Conjugation ratio }=(M \text { (conjugates) }-M \text { (BSA) }) / M \text { (haptens) }$

where M (conjugates) is the measured molecular weight of the BSA-hapten conjugate, M (BSA) is the molecular weight of unmodified BSA, and M (haptens) is the molecular weight of the hapten molecule. The conjugation ratio thus represents the average number of hapten molecules covalently attached per BSA molecule.

2.3. mAbs production and characterization

Standard immunization procedures were followed as described in our previous publication and detailed in the Appendix A [52]. Briefly, blood samples were collected from each mouse on day 7 after the second and fourth immunizations via the retro-orbital venous plexus, with careful handling. Mice exhibiting high serum sensitivity (i.e., IC₅₀) and broad specificity (i.e., wide CR) toward the 29 SAs were screened by enzyme-linked immunosorbent assay (ELISA) using ten different coating antigens, as described below.

Three days before cell fusion between splenocytes and SP2/0 myeloma cells, the selected mice were boosted intraperitoneally with hapten-KLH without adjuvant. Hybridoma cell lines producing mAbs with high sensitivity and broad specificity toward the 29 SAs were isolated using conventional hybridoma technology assisted by ELISA, as described in our prior work [52]. mAbs were harvested from mouse ascites and purified by protein G affinity chromatography using an ÄKTA pure protein purification system. Antibody isotypes were determined using an isotyping ELISA kit. The IC₅₀ and CR values of the mAbs generated in this study were determined by ELISA and compared with those of previously reported antibodies and DHPS. In addition, the effects of different competitive haptens on antibody specificity were assessed by ELISA using different SA coating antigens, with mAb 10E6 used as a model system.

2.4. Fab 10E6 preparation, crystallization, structure determination, and refinement

The variable-region genes of mAb 10E6 were obtained using previously reported methods with minor modifications [53]. Briefly, total RNA was extracted from hybridoma cells and reverse-transcribed into complementary DNA (cDNA) using primers listed in Table S4 in Appendix A. The heavy-chain (VH) and light-chain (VL) variable-region genes were then amplified by polymerase chain reaction. The resulting sequences were analyzed using the abYsis web-based platform for antibody sequence analysis.

Ascites containing mAb 10E6 was centrifuged at 13 670g for 10 min, and the clarified supernatant was collected. The ascites was then diluted threefold with phosphate-buffered saline (PBS) buffer (10 mmol∙L⁻¹, pH 7.4), filtered through a 0.22 μm membrane, and centrifuged again at 13 670g for 30 min. Antibody purification was performed using a protein A affinity chromatography column on an ÄKTA pure protein purification system.

Fab 10E6 fragments were generated using the Pierce™ Fab Micro Preparation Kit. Briefly, purified mAb 10E6 was digested in a papain-immobilized column. The digestion mixture was then applied to a protein A column to remove undigested immunoglobulin G (IgG) and fragment crystallizable (Fc) fragments, thereby isolating and purifying Fab 10E6 from the flow-through.

Fab 10E6 was adjusted to 20 mg∙mL⁻¹, and initial crystallization screening was performed at 293.0 K using the sitting-drop vapor diffusion method. Commercial screening kits, including Crystal Screen, Crystal Screen 2, PEGRx 1, PEGRx 2, SaltRx 1, SaltRx 2, and Index (Hampton Research, USA), were used. After initial screening, crystallization conditions were optimized using the hanging-drop vapor diffusion method. For each condition, 24-well plates were prepared with reservoir solutions. Protein and reservoir solutions were mixed at a 1:1 (v/v) ratio and placed on siliconized coverslips, which were inverted over the wells, sealed, and incubated at 293 K.

Diffraction data were collected from well-formed Fab 10E6 crystals. Crystals were harvested with a nylon loop and transferred to a cryoprotectant containing 15% (v/v) glycerol before mounting on the X-ray diffractometer. The crystal was oriented to maximize diffraction quality, and data collection parameters (including collection angles, diffraction intensity, and exposure time) were optimized based on the determined space group. Diffraction data were processed using HKL2000. Initial phases were obtained using PHASER by molecular replacement, guided by comparison with the Fab 10E6 amino acid sequence. Structural models were built and iteratively refined using COOT, REFMAC, and PHENIX. Model validation and quality assessment were performed using PROCHECK.

2.5. Molecular docking and MD simulations

Antibody-ligand complexes were generated using the CDOCKER module in Discovery Studio (DS), and the optimal binding pose for each complex was selected based on the lowest binding free energy. The selected complexes were subjected to MD simulations using GROningen MAchine for Chemical Simulations (GROMACS) with the Amber14SB force field and the transferable intermolecular potential with 3 points (TIP3P) water model. Solvated systems were energy-minimized using 5000 steps each of the steepest descent and conjugate gradient methods. A 2 ns position-restrained MD simulation was then performed under both NVT and NPT ensembles. Production MD simulations were subsequently conducted for 100 ns at 310 K and 1 atm with all restraints removed. Binding free energies were calculated using the gmx_MMPBSA toolkit based on the mmpsa.py script. Antibody-ligand complexes were visualized in DS, and corresponding 2D interaction diagrams were also generated in DS. The solvent-accessible surface area (SASA) and volume of the ligand-binding pockets of mAbs 4C7 and 10E6 were calculated using CASTpFold [54]. The normalized B-factor (B_norm) was defined as B_norm(i) = (B(i) − μ_B) σ_B, where B(i) represents the Cα B-factor of residue i, μ_B is the mean B-factor of the structure, and σ_B is the corresponding standard deviation. Residues with B_norm > 1 were considered highly flexible, whereas those with B_norm > 2 were classified as significantly flexible outliers.

2.6. Development and optimization of ELISA

Established workflows and commonly used buffer systems for indirect competitive ELISA (icELISA) targeting small molecules were adopted from previously published protocols [52]. In this study, key assay parameters were optimized using SMZ as the model analyte. These parameters included coating temperature, incubation time, pairing of the coating antigen with mAb 10E6, and the working concentrations of both antigen and antibody. Physicochemical properties of the assay buffer, including pH and ionic strength, were also systematically optimized. The effects of organic solvents, including methanol and acetonitrile, on antigen-antibody binding interactions were evaluated to determine solvent tolerance.

2.7. Sample preparation

SA-free skimmed milk and chicken samples were obtained from the National Reference Laboratory for Veterinary Drug Residues (China). For milk analysis, 1 mL of skimmed milk was diluted with 4 mL of PBS buffer (10 mmol∙L⁻¹, pH 7.4). For chicken samples, 2 g of homogenized tissue was placed in a clean centrifuge tube, followed by the addition of 10 mL PBS buffer. The mixture was vortexed for 10 min and centrifuged at 3500g for 10 min. The supernatant was collected and further diluted 1:1 with PBS before analysis. To evaluate recovery and variability of the sample preparation method, five commonly used SAs (SDZ, SMM, SMX, SMZ, and SQX) were spiked into blank milk and chicken samples at three concentration levels (low, medium, and high). The limits of detection (LODs) for each SA in milk and chicken were calculated as X₂₀ + 3SD, where X₂₀ is the mean signal from 20 blank samples and SD is the corresponding standard deviation.

3. Results

3.1. Ligand- and receptor-based rational hapten design

To address the affinity/specificity trade-off commonly encountered in MREs, we established an LRRHD strategy to design novel haptens and generate mAbs that exhibit both high affinity and fine specificity toward SAs. As illustrated in Fig. 1, the LRRHD strategy integrates ligand- and receptor-based analyses of SAs and their receptors. These analyses include ligand superposition, receptor structural analysis, receptor-based 3D quantitative structure-activity relationship (QSAR) analysis, and spacer arm analysis. Guided by these insights, novel haptens were rationally designed and further evaluated using computational chemistry approaches. Finally, the designed haptens were validated through animal immunization, mAb production, recognition mechanism analysis, and real-sample application.

Specifically, we first calculated the low-energy conformations of 29 SAs using DFT and performed structural superimposition using SN as the reference template (Fig. 2(a)). Overall, beyond differences in molecular structures, the main variations observed in the superimpositions were attributable to differences in the torsion angles of the C(aryl)-S-N-C backbone of SAs. Torsion angle-based K-means clustering analysis indicated that the 29 SAs could be classified into six structural categories (Fig. 2(b), Fig. S6 and Table S5 in Appendix A). Sulfanitran (SNT), which bears substituents at both the N¹ and N⁴ positions, was assigned to the first group and exhibited the smallest C(aryl)-S-N-C torsion angle (54.8°) (Fig. 2(c)). Most SAs clustered into the second major group, comprising 21 compounds with high conformational similarity and torsion angles ranging from 74.1° to 83.5°. For example, sulfamerazine (SMR) displayed a torsion angle of 78.3° (Fig. 2(d)). Sulfisoxazole (SIZ) (Fig. 2(e)) and sulfaphenazole (SPA) clustered into the third group, with torsion angles of 91.9° and 95.7°, respectively. As shown in Fig. 2(f), SG exhibited a torsion angle of 204.5° and was placed in the fourth group. Subsequently, sulfacetamide (SAM), sulfabenzamide (SBA) (Fig. 2(g)), and sulfamoxole (SXL) exhibited torsion angles ranging from 272.4° to 295.1°, placing them in the fifth group. Finally, SN (Fig. 2(h)), which lacks substituents at the N¹ position and therefore does not have a defined C(aryl)-S-N-C torsion angle, was categorized into the sixth group. Accordingly, selecting hapten templates from the second torsion angle group was expected to maximize the broad specificity of the resulting antibodies toward SAs.

In addition, our previous 3D-QSAR analysis indicated that bulky R groups were less favorable for binding [6], suggesting that two-ring haptens bearing benzene, pyridine, pyrimidine, or thiazole rings without additional substituents are more suitable templates for generating high-affinity, broad-specificity antibodies (Table S2). However, the 3D-QSAR analysis also showed that steric and electrostatic fields contributed nearly equally to antibody recognition. This finding indicates that R group selection should simultaneously account for electrostatic differences—such as those between pyridine and pyrimidine rings—and steric effects, including the presence or absence of small substituents such as methyl groups.

Furthermore, structural information from SA-binding receptors—including mAbs 4C7 and 4D11 and DHPS from our previous study [6]—was incorporated to guide rational hapten design. Analysis of the crystal structure of the antigen-binding fragment (Fab) complex of 4C7 with sulfathiazole (STZ) (Protein Data Bank (PDB) ID: 5CP3) revealed hydrogen bonding and hydrophobic interactions between the thiazole ring (R group) of STZ and mAb 4C7. Compared with DHPS, mAbs 4C7 and 4D11 displayed distinct dissociation and association rate constants across different SAs [6], indicating that the R group plays a major role during recognition. These observations suggest that the R group is the most critical structural feature in hapten design for determining the affinity and broad specificity of the resulting antibodies. Based on integrated ligand- and receptor-based analyses, SDZ, SMZ, and SPY—which belong to the second torsion angle class and exhibit appropriate molecular size and distinct electrostatic field distributions—were selected as primary templates for hapten design.

The spacer arm length, structure, and coupling site were also designed rationally. According to the hypothesis proposed by Muldoon et al. [44] and Wang et al. [7], the N⁴ terminus of the hapten should be exposed to the immune system to produce broad-specific antibodies against SAs. A medium-length valeric acid spacer arm has commonly been incorporated into the R group to facilitate conjugation to carrier proteins [55]. Moreover, increasing hapten rigidity has been reported to promote the initial binding and activation of B cell receptors (BCRs), thereby inducing strong antibody responses and improving antibody performance [56]. Therefore, we designed a 4-pentynoic acid group as a rigid spacer arm expected to enhance antibody performance. Based on the LRRHD strategy, four haptens were rationally designed and designated H1, H2, H3, and H4 (Fig. 3(a)). Hapten H1 was designed from the SDZ template with a rigid 4-pentynoic acid spacer arm. Hapten H2 was derived from the SMZ template and incorporated two methyl substitutions and a valeric acid spacer arm. Haptens H3 and H4 were designed from the SPY template with valeric acid spacer arms; notably, H4 included a methyl modification on the pyridine ring to increase structural diversity.

To evaluate the conformational behavior of haptens when conjugated to carrier proteins, we constructed and simulated the lowest-energy conformers of hapten-lysine (LYS) conjugates, because lysine residues represent the primary conjugation sites on carrier proteins [57]. As shown in Fig. 3(b), the spacer arms and carboxyl groups of H1-LYS, H2-LYS, H3-LYS, and H4-LYS did not fully overlap; nevertheless, their overall orientations were similar. The C(aryl)-S-N-C torsion angles of these hapten-LYS conjugates ranged from 74.1° to 83.5° (Table S5), showing no notable deviation from those of their parent drugs (SDZ, SMZ, and SPY). In addition, atom 2 in H1-LYS and H2-LYS was a nitrogen (N) atom carrying a negative charge (Fig. 3(c), Table S6 in Appendix A), whereas in H3-LYS and H4-LYS, atom 2 was a carbon (C) atom, resulting in a lower absolute atomic charge than in H1-LYS and H2-LYS. These results indicate that the designed haptens possess distinct electrostatic properties, which may contribute to the induction of different antibody responses toward SAs.

Comparative analysis of atoms 3, 5, and 6 among the four hapten-LYS conjugates further indicated that methyl substitution on the R group altered both steric and electrostatic fields. Fig. 3(d) shows the molecular ESP distributions of these conjugates. Notably, surface minima of −8.27 and −12.69 kcal∙mol⁻¹ were observed near the spacer arm of H1-LYS, whereas positive surface maxima were observed in the other conjugates. When methyl groups were introduced on the pyrimidine ring, the ESP near the ring decreased from 7.05 kcal∙mol⁻¹ in H1-LYS to −0.88 and −4.32 kcal∙mol⁻¹ in H2-LYS. A similar trend was observed for the pyridine-based conjugates, with ESP values decreasing from −7.44 kcal∙mol⁻¹ in H3-LYS to −9.43 kcal∙mol⁻¹ in H4-LYS. These results indicate that differences between pyridine and pyrimidine rings, as well as the presence or absence of methyl substituents, substantially influenced the ESP of the conjugates. Collectively, the optimized conformations, charges from ESPs using a grid-based method (CHELPG), and ESP analyses confirm the successful design of the novel haptens from structural and electronic perspectives.

3.2. Deliberate synthesis of novel haptens and hapten-carrier conjugates

In this study, hapten H1 was synthesized via an addition reaction between SDZ and 4-pentynoic acid (Fig. S2). Hapten H2 was synthesized de novo by first preparing 4,6-dimethylpyrimidin-2-amine, followed by conjugation with a spacer arm to obtain ethyl 5-(2-amino-4,6-dimethylpyrimidin-5-yl)pentanoate. After an addition reaction between N-acetylsulfanilyl chloride and ethyl 5-(2-amino-4,6-dimethylpyrimidin-5-yl)pentanoate, the target product was obtained through a total of nine steps (Fig. S3). For the synthesis of the two-ring haptens containing the common paminobenzenesulfonamide moiety, H3 (Fig. S4) and H4 (Fig. S5), the protocols were similar to those used for hapten S4 [33] (Table S2), with two additional steps. The chemical structures of the four haptens were confirmed by ¹H and ¹³C nuclear magnetic resonance spectra and are presented in Figs. S7-S10 in Appendix A.

The haptens were then conjugated to KLH and BSA using the NHS ester method, and the hapten-BSA conjugates were analyzed by matrix-assisted laser desorption/ionization time-of-flight mass spectrometry. The results showed an obvious mass shift of the BSA peak after conjugation with haptens, confirming successful coupling to the carrier proteins (Fig. S11 in Appendix A). According to Eq. (1), the conjugation ratios of H1-BSA, H2-BSA, H3-BSA, H4-BSA, BS-BSA, HS-BSA, SA10-BSA, TS-BSA, CS-BSA, and SA10-X-BSA were 13.1, 12.0, 11.5, 13.0, 14.3, 13.8, 12.3, 11.8, 11.5, and 12.2, respectively (Table S7 in Appendix A). The chemical structures of the SA haptens BS, HS, SA10, TS, CS, and SA10-X were prepared previously and are listed in Table S2.

3.3. Characterization of antisera and mAbs to SAs

To evaluate the designed haptens and generate antibodies against SAs, the H1-KLH, H2-KLH, H3-KLH, and H4-KLH conjugates were used for immunization, following the detailed procedures described in the Appendix A. The performance of antisera collected after the third immunization was evaluated by ELISA. Inhibition rates were calculated for 29 SAs using concentrations of 100 μg∙L⁻¹ for single-ring SAs and 50 μg∙L⁻¹ for two-ring SAs; higher inhibition rates indicate higher antibody affinity toward SAs. Initially, ten coating antigens (hapten structures listed in Fig. S1) were used in ELISA to evaluate antisera from mice. Among these coating antigens, SA10-BSA for antisera elicited by SAPM-KLH, BS-BSA for antisera elicited by SADMPM-KLH, TS-BSA for antisera elicited by SAPY-KLH, and BS-BSA for antisera elicited by SAMPY-KLH were selected as optimal coating antigens because the highest inhibition rates were obtained with these competitive antigens.

As shown in Fig. 4(a), antisera derived from hapten H1 exhibited higher inhibition rates against 14 SAs, with mean inhibition values exceeding 50%. These SAs included two-ring SAs containing a five-membered ring (e.g., STZ and SXL); SAs containing a six-membered pyridine ring (e.g., SPY); SAs containing a six-membered pyridazine ring, including sulfachloropyridazine (SCP), sulfamethoxypyridazine (SMP), and sulfaethoxypyridazine (SEP); SAs containing a quinoxaline ring (e.g., sulfaquinoxaline (SQX)); and SAs containing a six-membered pyrimidine ring, including SMM, sulfisomidine (SIM), SDM, SDZ, sulfamethoxydiazine (SMD), SMR, SMZ, and sulfabromomethazine (SBM). Antisera derived from hapten H2 showed a recognition profile broadly similar to that of H1-derived antisera but exhibited overall lower inhibition rates, with ten SAs showing mean inhibition values above 50%. Compared with H2, hapten H1—containing a rigid 4-pentynoic acid spacer arm and lacking methyl substitution on the pyrimidine ring—may therefore enhance antibody generation in vivo. Antisera elicited by hapten H3, which contains a pyridine ring, were highly specific to SPY, SMP, and SEP, with only three SAs showing mean inhibition rates above 50%. In contrast, antisera from hapten H4 showed improved recognition of ten SAs (STZ, SPY, SCP, SMP, SEP, SQX, SMM, SDM, SDZ, and SMR), with inhibition rates exceeding 50%. These results indicate that methyl substitution on the pyridine ring in H4 is a key structural determinant enhancing recognition of SQX, SMM, SDM, and SMR.

However, all antisera showed relatively weak recognition of single-ring SAs (e.g., SN, SAM, and SG); two-ring SAs containing a five-membered ring (e.g., SPA, PST, and SIZ); and two-ring SAs containing a six-membered ring (e.g., SBA, SNT, sulfasalazine (SSA), sulfachlorpyrazine (SCY), sulfalene (SLE), and sulfadoxine (SDM′)). This observation supports our hypothesis that greater similarity in spatial orientation between SAs and the designed haptens improves antibody recognition. As shown in Fig. 2 and Table S5, the C(aryl)-S-N-C torsion angles of SN, SAM, SG, SPA, SIZ, SBA, and SNT differ markedly from those of the H1-LYS, H2-LYS, H3-LYS, and H4-LYS conjugates. Consequently, antisera elicited by immunogens H1-KLH, H2-KLH, H3-KLH, and H4-KLH displayed relatively poor recognition of these SAs.

These results demonstrate that the structure of the immunizing hapten is closely associated with the recognition capability of the resulting antibody [56]. Haptens containing a pyrimidine ring conferred higher affinity toward pyrimidine-containing SAs than toward other SAs, whereas pyridine-containing haptens showed enhanced affinity for pyridine-containing SAs. Substituents on the six-membered heterocyclic ring of haptens also influenced antibody recognition. For example, the methyl group in H4 improved recognition of SQX and certain SA analogs. Most importantly, hapten H1, which lacks methyl substitution on the pyrimidine ring but incorporates a rigid 4-pentynoic acid spacer arm, markedly enhanced antibody affinity toward most SAs, including the single-ring SAs examined in this study, yielding higher inhibition rates than those induced by the other three haptens. These findings indicate that introducing a rigid 4-pentynoic acid spacer arm in H1 plays a critical role in promoting effective antibody induction in vivo.

Within each immunization group, the mouse showing the highest inhibition rates against the greatest number of individual SAs was selected for cell fusion. Five hybridoma clones were obtained for mAb production: clones 10E6 and 7C3 from immunogen H1-KLH, clone 4F5 from immunogen H2-KLH, clone 5F7 from immunogen H3-KLH, and clone B2 from immunogen H4-KLH. The resulting mAbs were classified as IgG2b(κ) for 10E6 and 7C3, IgG1(κ) for 4F5 and 5F7, and IgG2a(λ) for B2.

The affinity and specificity of the five mAbs toward SAs were evaluated by ELISA. As shown in Table 1 [7], [8], [15], [41], [45], [46], mAbs 10E6 and 7C3 elicited by hapten H1 exhibited the highest affinity toward two-ring SAs containing a pyrimidine ring, with IC₅₀ values as low as the sub-parts per billion range (0.20-0.94 μg∙L⁻¹) for SDZ, SMD, SMR, SMZ, and SBM, all of which belong to the second torsion-angle cluster (Table S5). This result is consistent with our previous finding that greater similarity in conformational and electrostatic characteristics between SAs and haptens is associated with stronger antibody affinity. Nevertheless, for SNT in the first group, PST and SSA in the second group, and SPA in the third group, bulky substituents at the N¹ or N⁴ position resulted in lower affinity, although IC₅₀ values still remained around 20 μg∙L⁻¹. Likewise, for single-ring SAs such as SN, SG (fourth group), and SAM (fifth group), the molecular shapes did not align well with an antibody binding pocket generated against two-ring haptens, resulting in IC₅₀ values also around 20 μg∙L⁻¹. For the remaining SAs, IC₅₀ values ranged from 0.58 to 8.8 μg∙L⁻¹. Notably, despite the different C(aryl)-S-N-C torsion angles of SIZ (third group), SBA, and SXL (fifth group) relative to hapten H1, SIZ (IC₅₀, 2.5 μg∙L⁻¹), SBA (IC₅₀, 0.78 μg∙L⁻¹), and SXL (IC₅₀, 0.61 μg∙L⁻¹) still bound tightly to 10E6. This finding suggests that SA conformations can adjust to fit the antibody binding site during recognition.

Compared with DHPS (Table 1), mAbs 10E6 and 7C3 showed comparable specificity but exhibited a one- to two-order-of-magnitude increase in affinity. Moreover, relative to other mAbs—including 4F5, 5F7, and B2 (elicited by alternative haptens) and previously reported antibodies—mAbs 10E6 and 7C3 demonstrated superior affinity and broadened specificity, recognizing all 29 SAs at sensitivities meeting the 100 μg∙kg⁻¹ requirement (Table 1). In contrast, mAbs 4F5, 5F7, and B2 derived from haptens H2, H3, and H4 recognized only 16, 8, and 16 SAs, respectively, with IC₅₀ values below 100 μg∙kg⁻¹. These results further support that introducing a rigid 4-pentynoic acid spacer arm in H1 plays a critical role in promoting effective antibody induction in vivo. Accordingly, mAbs 10E6 and 7C3 effectively mitigate the affinity/specificity trade-off, highlighting the value of the LRRHD-guided rational hapten design strategy proposed here.

The influence of different competitive haptens on the specificity of mAb 10E6 was further evaluated by ELISA against 11 representative SAs: SN (single-ring); STZ and SMX (five-membered ring); SPY (pyridine ring); SMP (pyridazine ring); SQX (quinoxaline ring); and SDM′, SDZ, SMD, SMR, and SMZ (pyrimidine ring). The working concentrations of coating antigen and mAb 10E6 were optimized by checkerboard assays, with B₀ of approximately 1.5. IC₅₀ values for the 11 SAs were obtained from calibration curves, and CR values were calculated as the ratio of the IC₅₀ for SQX to that for each analyte. Because of low binding affinity, the optimal concentration of mAb 10E6 exceeded 1.7 mg∙L⁻¹ when paired with BS-BSA, HS-BSA, or SA10-X-BSA, making these pairs unsuitable for assessing the influence of the corresponding competitive haptens on antibody specificity.

As shown in Fig. 4(b), the CR values of mAb 10E6 followed a similar trend across different coating antigens (i.e., different competitive haptens). Specifically, mAb 10E6 showed the strongest recognition of SQX, SDM′, SDZ, SMD, SMR, and SMZ, with CR values ranging from 80.7% to 1930.8%. Moderate binding was observed for STZ, SMX, SPY, and SMP, with CR values between 4.0% and 198.5%, whereas SN showed poor recognition, with CR values below 7.8%. This consistency indicates that the coating antigen has minimal influence on antibody specificity toward SAs, implying that the competitive hapten does not substantially alter antibody specificity in ELISA. However, when SA10-BSA and H2-BSA were used as coating antigens, lower IC₅₀ values were obtained for the tested SAs, suggesting that although competitive haptens may have limited effects on specificity, they can markedly influence ELISA sensitivity.

In our previous study, the use of a site-heterologous hapten (competitive hapten), rather than a homologous hapten or a spacer arm-length heterologous hapten, significantly improved assay sensitivity [40]. Consistent with this observation, varying the coating antigens produced different IC₅₀ values (Fig. 4(b)). The highest sensitivity was achieved when SA10-BSA (1.30 mg∙L⁻¹) was combined with mAb 10E6 (340 μg∙L⁻¹) in ELISA. Similarly, in the mAb 3D1-based assay, HS-ovalbumin (OVA) was identified as the optimal coating antigen, further supporting that heterologous haptens are generally more effective than homologous haptens in competitive ELISA [8].

3.4. Molecular recognition mechanism analysis of mAb 10E6

To elucidate the recognition mechanism underlying the high affinity and fine specificity of mAb 10E6 toward SAs, we determined the crystal structure of 10E6 and analyzed its interactions with SAs. mAb 10E6 was first purified from mouse ascites and digested with papain in the solid phase. The resulting Fab fragments (Fab 10E6) were further purified by size-exclusion chromatography (Fig. S12 in Appendix A). Crystals of Fab 10E6 were obtained using the hanging-drop vapor diffusion method under conditions of 400 mmol∙L⁻¹ diammonium phosphate (pH 7.4) and 10% (w/v) PEG 3350, identified from a PEG/ion screen (Fig. S13 in Appendix A). We then attempted to determine the crystal structures of Fab 10E6 in complex with SDZ and SMZ using both soaking and co-crystallization approaches. However, these attempts were unsuccessful, likely because ligand binding interfered with crystal formation and disrupted the crystal lattice. Apo-Fab 10E6 crystals belonged to the P1211 space group and diffracted to 2.5 Å resolution (PDB ID: 9VJK). A molecular replacement solution was obtained, and the structure was refined to a working R-factor/free R-factor (R_work/R_free) of 0.22/0.26 through iterative model building and refinement (Table S8 in Appendix A). As shown in Fig. 5(a), the asymmetric unit contained one copy of Fab 10E6, comprising four typical domains: VH, heavy-chain constant region 1 (CH1), VL, and light-chain constant region (CL). The ligand-binding pocket was formed mainly by complementarity determining region 3 of the heavy chain (CDRH3), complementarity determining region 2 of the antibody heavy chain (CDRH2), and complementarity determining region 3 of the antibody light chain (CDRL3).

To investigate why mAb 10E6 exhibits higher affinity and broader specificity toward SAs, we first compared the variable fragment (Fv) sequences of 4C7, 4D11, and 10E6. As shown in Fig. 5(b) and Table S9 in Appendix A, the VH domain of 10E6 shares 46.7% and 49.5% identity with 4C7 and 4D11, respectively, and 61.5% and 68.6% similarity. The VL domain of 10E6 shares 53.5% and 53.1% identity with 4C7 and 4D11, respectively, and 69.6% and 69.4% similarity. We next compared CDR lengths and found that CDRH3 of 10E6 contains 12 amino acid residues, substantially longer than the 4 residues in 4C7 and the ten residues in 4D11 (Kabat numbering). In contrast, CDRL1 of 10E6 comprises 11 residues, shorter than the 16 residues in 4C7 and the 15 residues in 4D11 (Fig. 5(b)). These differences in CDR length may underlie the distinct affinities and recognition ranges of these antibodies.

We then compared the crystal structures of 10E6 and 4C7 by 3D structural superimposition. As shown in Fig. 5(c), the overall root-mean-square deviation (RMSD) was 1.6 Å. Consistent with the sequence alignment, the main structural differences were localized to the CDRH3 and CDRL1 regions. Notably, CDRL1 of 4C7 is positioned far from the binding pocket and does not participate in ligand binding in the 4C7-STZ complex [6]. Similarly, CDRL1 of 10E6 is located away from the ligand-binding pocket (Fig. 5(a)). These observations support the hypothesis that the extended CDRH3 loop of 10E6 contributes to its enhanced affinity and broadened recognition spectrum.

We next compared the SASA and volume of the ligand-binding pockets of mAbs 4C7 and 10E6 using CASTpFold [54]. As shown in Fig. S14 and Table S10 in Appendix A, the SASA and volume of the ligand-binding pocket of apo-10E6 were 186.3 Å² and 86.8 Å³, respectively, both substantially larger than those of ligand-bound 4C7 (PDB: 5CP3), which were 97.5 Å² and 47.0 Å³. Although the predicted SASA and volume of apo-4C7 (PDB: 5CP7) were 181.9 Å² and 89.8 Å³, respectively, the linker channel between the first and second cavities is too narrow to accommodate an intact ligand molecule, as confirmed by the crystal structure of the ligand-bound 4C7 complex. Therefore, the effective surface area and volume of the 4C7 ligand-binding pocket are better approximated by the values observed in the ligand-bound complex. These analyses indicate that mAb 10E6 possesses a broader ligand-binding pocket. To compare structural flexibility, B_norm were calculated for Cα atoms and analyzed at the residue level (Table S11 in Appendix A). The apo-Fv structure of 10E6 exhibited a similar proportion of residues with B_norm > 1 compared with apo-Fv 4C7 (15.8% vs 16.0%). However, 10E6 showed a markedly higher proportion of residues with B_norm > 2 than apo-Fv 4C7 (3.9% vs 1.3%), and this proportion was also higher than that observed for ligand-bound Fv 4C7 (2.2%). This difference was particularly pronounced in the CDRH3 region, a major component of the ligand-binding pocket. In this region, 10E6 contained 38.5% highly flexible residues (B_norm > 1) and 23.1% highly flexible residues (B_norm > 2), largely attributable to its extended CDRH3 loop. Collectively, these results indicate a rightward shift in the B_norm distribution for 10E6, reflecting increased global flexibility and enhanced binding-pocket flexibility relative to mAb 4C7. Accordingly, the extended CDRH3 loop in 10E6 appears to create a larger and more flexible ligand-binding pocket, which may underpin its higher binding affinity and broader recognition capability.

To further clarify the recognition mechanism of 10E6, we performed molecular docking followed by MD simulations of Fv 10E6 bound to six representative SAs—SNT, SMR, SIZ, SG, SBA, and SN—each representing a distinct torsion-angle group. The RMSD values of amino acid backbone atoms remained stable throughout the simulations, indicating that reasonable complex structures were obtained (Fig. S15 in Appendix A). As depicted in Fig. 6(a), the loops that form the binding pocket of 10E6—CDRH3, CDRH2, and CDRL3—underwent conformational fluctuations during the 100 ns MD simulations, resulting in minor structural changes in the ligand-binding pocket of Fv 10E6 (Fig. S16 in Appendix A). Nevertheless, the hydrophobicity of the binding pocket remained largely unchanged, consistent with the stable SASA of the entire protein (Fig. S17 in Appendix A). In all simulations, the N⁴ terminus of the SAs inserted into the cavity of 10E6, providing a key structural basis for the broad-specific recognition exhibited by 10E6 (Fig. S16).

We then analyzed the non-bonding interactions between 10E6 and SAs (Fig. 6(b)). The results indicated that hydrophobic interactions (including π-π stacking, π-alkyl, and π-sulfur interactions) and van der Waals forces were major contributors to stabilizing the ligand-antibody complexes. Two-ring SAs such as SMR and SBA formed stronger hydrophobic interactions, consistent with their higher binding affinities. Ligand-binding residues from the heavy chain, particularly within CDRH3, contributed predominantly to SA recognition by 10E6. Specifically, Trp103 in CDRH3 formed a key hydrogen bond with an oxygen atom of the sulfonamide group. Hydrogen-bond monitoring showed that this interaction persisted for nearly the entire simulation period and was consistently observed across all 10E6-SA complexes (Fig. S18 in Appendix A). In addition, His99 (CDRH3), Tyr59 (CDRH2), and Phe94 (CDRL3) formed critical π-π stacking and hydrophobic interactions with the benzene ring and the secondary heterocyclic ring of the SAs. Moreover, the average binding free energies of 10E6 with SNT, SMR, SIZ, SG, SBA, and SN ranged from −23.2 to −35.0 kcal∙mol⁻¹ (Table S12 in Appendix A), with SMR showing the most favorable binding free energy, whereas SN and SG showed the least favorable binding free energies. These trends agreed with the CR results obtained by ELISA (Table 1). In summary, the extended CDRH3 loop of 10E6 forms a broader and more flexible ligand-binding pocket that can accommodate diverse SA structures and establish more ligand interactions, thereby supporting its high affinity and broad-specific recognition.

3.5. Application of mAb 10E6-based icELISA in real sample determination

To evaluate the practical performance of mAb 10E6, we developed a mAb 10E6-based icELISA for detecting multiple SA residues in real samples. As shown in Fig. 4(b), the highest sensitivity was achieved when the coating antigen SA10-BSA was paired with mAb 10E6 in ELISA. We then investigated the effects of physicochemical parameters on the performance of this antigen-antibody pair. The parameters B₀, IC₅₀ of SMZ, and the IC₅₀/B₀ ratio were used as key indicators. As shown in Fig. 7(a), PBS buffers with different NaCl concentrations (34, 68, 136, 272, and 544 mmol∙L⁻¹) were evaluated. Among these conditions, 136 mmol∙L⁻¹ NaCl produced an optimal B₀ value of approximately 1.5 and the lowest IC₅₀/B₀ ratio, indicating the best assay performance. Therefore, a buffer composed of 136 mmol∙L⁻¹ NaCl, 8 mmol∙L⁻¹ Na₂HPO₄, 2 mmol∙L⁻¹ KH₂PO₄, and 2.6 mmol∙L⁻¹ KCl (i.e., standard PBS) was selected for subsequent experiments. For pH optimization, values from 5.0 to 8.0 were tested. As shown in Fig. 7(b), pH 7.4 yielded the lowest IC₅₀/B₀ ratio.

The influence of organic solvent concentration on assay performance was assessed by dissolving SMZ standards in 0, 1%, 2%, 5%, 10%, or 20% organic solvent. As shown in Figs. 7(c) and (d), competition between SMZ and the competitive hapten SA10 (structure shown in Fig. S1) for binding to mAb 10E6 was notably disrupted at 10% acetonitrile. These results indicate that the ELISA tolerates up to 10% acetonitrile and up to 20% methanol without compromising assay performance.

The matrix effect in skimmed milk was evaluated using 2.5-, five-, and ten-fold dilutions in PBS. As shown in Fig. 7(e), a simple five-fold dilution was sufficient to effectively eliminate the matrix effect. For chicken samples, homogenized tissue was extracted with PBS at a 1:5 (w/v) ratio, and the undiluted extract as well as two- and four-fold dilutions were tested. As shown in Fig. 7(f), the matrix effect in chicken samples was effectively mitigated by a two-fold dilution with PBS. The resulting calibration curves for SMZ in PBS and in the processed extracts were nearly identical, confirming effective matrix elimination and reliable assay performance in complex matrices.

To assess the accuracy and precision of the developed icELISA, five SAs commonly used in veterinary practice—SMX, SDZ, SMZ, SMM, and SQX—were spiked into blank skimmed milk and chicken samples. Results are presented as recovery (%) and coefficient of variation (CV, %). As shown in Tables S13 and S14 in Appendix A, recoveries in skimmed milk ranged from 63.2% to 107.1%, with CVs below 19.4%. In chicken samples, recoveries ranged from 62.7% to 110.5%, also with CVs below 19.4%, demonstrating the suitability of the developed icELISA for detecting multiple SA residues in both matrices. The LODs for SMX, SQX, SMM, SMZ, and SDZ in skimmed milk were 1.5, 1.2, 1.1, 1.0, and 0.29 μg∙L⁻¹, respectively. In chicken, the corresponding LODs were 4.7, 1.9, 2.8, 1.3, and 0.81 μg∙kg⁻¹, respectively. In conclusion, mAb 10E6, which exhibits both high affinity and fine specificity, was successfully applied for monitoring multiple SA residues in real samples.

4. Discussion

The discovery of MREs is persistently constrained by the intrinsic trade-off between affinity and specificity. In this study, SAs were selected as model analytes to mitigate this affinity/specificity trade-off through LRRHD and to generate mAbs with both high affinity and fine specificity. For small molecules such as SAs, hapten design is a decisive step in producing antibodies that achieve high affinity together with the desired specificity. Using a traditional hapten design strategy based on 2D structural analysis, Muldoon et al. [44] designed the generic hapten SUL; however, the resulting mAb, Sulfa-1, did not achieve the expected specificity (Table S3). Subsequent conformational optimization analysis showed that the minimum-energy conformation of SUL differed markedly from those of cross-reactive SAs, suggesting that mAb Sulfa-1 may have recognized a different structural conformation of the hapten. Building on conformational optimization and atomic charge distribution analyses, Adrian et al. [35] designed hapten SA1, noting that many SAs contain an electron-withdrawing atom adjacent to the carbon in the R groups relative to the N¹ position. Computational evaluations of dihedral angles, charge distributions, and pKa values indicated minimal geometric and electrostatic variation near the sulfonamide bond across different SA structures. The resulting polyclonal antibody (pAb), As155, exhibited CR with 10 SAs, each with IC₅₀ values below 100 μg∙L⁻¹ (Table S3). Building on these insights, we designed hapten SA10-X, which generated a pAb reactive with 18 SAs, also with IC₅₀ values below 100 μg∙L⁻¹ [42]. Further 3D-QSAR analysis suggested that incorporating one or two nitrogen atoms into the second aromatic ring, together with methyl or methoxy substitutions, may improve future hapten designs.

Despite substantial progress in ligand-based hapten design, generating SA antibodies with both high affinity and fine specificity remains challenging, largely because of the inherent affinity/specificity trade-off. Ligand-based hapten design is often affected by molecular input bias and provides limited insight into antigen-antibody interactions; these limitations are particularly evident when developing haptens for broad-spectrum antibody recognition, where epitope formation is difficult to predict. In contrast, receptor- or structure-based drug design (SBDD) has been widely and successfully applied in the pharmaceutical industry, highlighting the value of integrating molecular interaction information into rational design strategies [58], [59]. Incorporating receptor information into hapten design can reduce randomness, strengthen mechanistic rationality, and increase the likelihood of generating target-specific antibodies [51]. In our previous work, structural insights into SA-binding receptors—including mAbs 4C7 and 4D11 as well as DHPS—were established [6]. Therefore, in the present study, we integrated ligand-based analyses with receptor-based structural analyses of SA-binding antibodies and protein receptors and established the LRRHD strategy, as illustrated in Fig. 1. From the ligand-based analysis, the C(aryl)-S-N-C torsion angle emerged as a key indicator for hapten design, and haptens derived from the second torsion-angle group were expected to maximize the broad specificity of the resulting antibodies toward SAs. From the receptor-based analysis, we determined that the R group of the hapten is a critical determinant of both affinity and broad specificity in the generated antibodies. Guided by these integrated ligand- and receptor-based insights, four novel haptens were rationally designed.

Among the mAbs produced using these four haptens, mAbs 10E6 and 7C3 derived from hapten H1 showed ultralow IC₅₀ values and unexpectedly broad specificity toward SAs (Table 1 and Table S3). In particular, mAb 10E6 exhibited outstanding performance, expanding the recognition spectrum to 29 SAs while maintaining the affinity gap within 88-fold. This represents a substantial improvement over previously developed antibodies, including 4D11 [7], 4E5 [45], Chen [41], 3D1 [8], and those reported by Chen et al. [41] and Li et al. [46]. Compared with DHPS, mAb 10E6 showed a > 25-fold improvement in the mean IC₅₀ for SAs, although its affinity gap remained larger than that of DHPS. For mAb 4F5, our results are consistent with those of Chen et al. [8], who reported that antibodies generated by haptens H2 and S7-C (Table S2), both containing a dimethylpyrimidine moiety, displayed high specificity toward SMZ. Although hapten S7-C contained a 4-pentenoic acid spacer arm, antibody specificity was not substantially affected, suggesting that the characteristic epitope of the hapten plays a dominant role in determining antibody specificity, whereas the spacer arm primarily functions to present the antigenic determinant to the immune system. When comparing mAb 5F7 with other antibodies, such as pAb S4-II produced against hapten S4 [34] (Table S2), the affinities measured against SMP (IC₅₀, 0.68 μg∙L⁻¹), sulfamethizole (SMT) (IC₅₀, 0.86 μg∙L⁻¹), STZ (IC₅₀, 1.7 μg∙L⁻¹), and SPY (IC₅₀, 2.9 μg∙L⁻¹) were comparable to those observed for mAb 5F7. Similarly, for pAb As155 generated from hapten SA1 [35], the IC₅₀ values for STZ, SPY, and SEP were 1.3, 2.6, and 5.0 μg∙L⁻¹, respectively (Table S3). However, for SIM, SMR, and SMZ—compounds bearing one or two methyl substituents on the pyrimidine ring—the IC₅₀ values ranged from 1.8 to 3.7 μg∙L⁻¹ (Table S3). This discrepancy is likely attributable to the heterogeneity of pAb As155, resulting in differences in recognition behavior between pAb As155 and mAb 5F7.

CDRH3 plays a central role in determining antibody-antigen affinity and specificity, and its length is a key factor influencing binding characteristics [60], [61]. Compared with the SA-binding antibody 4C7 [6], mAb 10E6 contains a substantially longer CDRH3 (12 amino acid residues). Structural superposition suggests that this extended CDRH3 contributes to a broader and more flexible binding pocket [62], enabling accommodation of structurally diverse SAs and allowing formation of additional ligand interactions—features that collectively support its high affinity and broad specificity.

The CR of an antibody with ligands is closely associated with molecular structure [6], [42]. In this work, three SAs with N⁴-position modifications—PST, SNT, and SSA (Fig. S1)—showed higher IC₅₀ values of 20, 14, and 17 μg∙L⁻¹, respectively (Table 1). This is most likely because mAb 10E6 recognizes epitopes near the N⁴ position of SAs, and modification at this site interferes with antibody binding (Fig. 6(b)), leading to elevated IC₅₀ values. In addition, single-ring SAs (SN, SG, and SAM) also exhibited increased IC₅₀ values of 20, 20, and 9.6 μg∙L⁻¹, respectively, primarily because they lack key interactions present in double-ring SAs (Fig. 6(b)), thereby reducing binding affinity. For the remaining SAs, IC₅₀ values ranged from 0.23 to 10 μg∙L⁻¹ and showed a significant positive correlation with their torsion angles (P = 0.0033), indicating that larger torsion angles correspond to lower antibody affinity (Fig. S19 in Appendix A). Collectively, these results demonstrate that the CR profile of mAb 10E6 across SAs is strongly correlated with molecular structure.

The key steps in hapten design involve selecting an appropriate core epitope and spacer arm. Here, to achieve high affinity and fine specificity toward SAs, the hapten core epitope was designed based on the predominant conformational cluster (Group 2; C(aryl)-S-N-C torsion angles) identified using the LRRHD strategy. However, this approach inevitably produced uneven CR toward different SAs among the resulting antibodies. This outcome reflects an inherent theoretical limitation: designing a hapten to mimic the dominant conformational population will necessarily compromise binding affinity toward analytes with distinct conformational preferences. This limitation helps explain why uniformly high specificity across a class of structurally related analytes is difficult to achieve. Accordingly, rational hapten design can mitigate the affinity/specificity trade-off to some extent, but it cannot fully eliminate it. Further studies are therefore needed to clarify how hapten conformation influences immune responses and to elucidate the molecular mechanisms by which antibodies discriminate among structurally diverse analytes.

In this study, we proposed an LRRHD strategy to mitigate the affinity/specificity trade-off commonly observed in MREs. Four haptens—H1, H2, H3, and H4—were rationally designed and synthesized. mAb 10E6, derived from hapten H1 containing a novel rigid 4-pentynoic acid spacer arm, exhibited unexpectedly high affinity and fine specificity, largely attributable to its extended CDRH3, which creates a broader and more flexible binding pocket capable of recognizing diverse SAs. Based on mAb 10E6, an icELISA was developed for monitoring multiple SA residues in real samples.

References

Publishing order | Descend order by publishing year | Descend order by cited within

[1]	Krämer J, Kang R, Grimm LM, De Cola L, Picchetti P, Biedermann F. Molecular probes, chemosensors, and nanosensors for optical detection of biorelevant molecules and ions in aqueous media and biofluids. Chem Rev 2022; 122 (3):3459-636.

[2]	Seitz WR, Grenier CJ, Csoros JR, Yang R, Ren T. Molecular recognition: perspective and a new approach. Sensors 2021; 21(8):2757.

[3]	Jayan H, Pu H, Sun DW. Recent development in rapid detection techniques for microorganism activities in food matrices using bio-recognition: a review. Trends Food Sci Technol 2020; 95:233-46.

[4]	Ausserwöger H, Krainer G, Welsh TJ, Thorsteinson N, de Csilléry E, Sneideris T, et al. Surface patches induce nonspeciﬁc binding and phase separation of antibodies. Proc Natl Acad Sci USA 2023; 120(15):e2210332120.

[5]	Rabia LA, Desai AA, Jhajj HS, Tessier PM. Understanding and overcoming trade-offs between antibody afﬁnity, speciﬁcity, stability and solubility. Biochem Eng J 2018; 137:365-74.

[6]	Li C, Liang X, Wen K, Li Y, Zhang X, Ma M, et al. Class-speciﬁc monoclonal antibodies and dihydropteroate synthase in bioassays used for the detection of sulfonamides: structural insights into recognition diversity. Anal Chem 2019; 91(3):2392-400.

[7]	Wang Z, Beier RC, Sheng Y, Zhang S, Jiang W, Wang Z, et al. Monoclonal antibodies with group speciﬁcity toward sulfonamides: selection of hapten and antibody selectivity. Anal Bioanal Chem 2013; 405(12):4027-37.

[8]	Chen Y, Guo L, Liu L, Song S, Kuang H, Xu C. Ultrasensitive immunochromatographic strip for fast screening of 27 sulfonamides in honey and pork liver samples based on a monoclonal antibody. J Agric Food Chem 2017; 65(37):8248-55.

[9]	Zhao F, Tian Y, Wang H, Liu J, Han X, Yang Z. Development of a biotinylated broad-speciﬁcity single-chain variable fragment antibody and a sensitive immunoassay for detection of organophosphorus pesticides. Anal Bioanal Chem 2016; 408(23):6423-30.

[10]	Li X, Kaattari SL, Vogelbein MA, Vadas GG, Unger MA. A highly sensitive monoclonal antibody based biosensor for quantifying 3-5 ring polycyclic aromatic hydrocarbons (PAHs) in aqueous environmental samples. Sens Biosensing Res 2016; 7:115-20.

[11]	Ahmed S, Ning J, Cheng G, Ahmad I, Li J, Mingyue L, et al. Receptor-based screening assays for the detection of antibiotics residues-a review. Talanta 2017; 166:176-86.

[12]	Zeng K, Zhang J, Wang Y, Wang ZH, Zhang SX, Wu CM, et al. Development of a rapid multi-residue assay for detecting b -lactams using penicillin binding protein 2x*. Biomed Environ Sci 2013; 26(2):100-9.

[13]	Zhang J, Wang Z, Wen K, Liang X, Shen J. Penicillin-binding protein 3 of Streptococcus pneumoniae and its application in screening of b-lactams in milk. Anal Biochem 2013; 442(2):158-65.

[14]	Wang Z, Liang X, Wen K, Zhang S, Li C, Shen J. A highly sensitive and class-speciﬁc fluorescence polarisation assay for sulphonamides based on dihydropteroate synthase. Biosens Bioelectron 2015; 70:1-4.

[15]	Liang X, Li C, Zhu J, Song X, Yu W, Zhang J, et al. Dihydropteroate synthase based sensor for screening multi-sulfonamides residue and its comparison with broad-speciﬁc antibody based immunoassay by molecular modeling analysis. Anal Chim Acta 2019; 1050:139-45.

[16]	Jin Q, Fan Y, He T, Peng J, Liu J, Wang J. Fluorescence polarization assay based on a new recognition motif QepA for the one-step detection of fluoroquinolones in eggs. J Agric Food Chem 2023; 71(49):19749-59.

[17]	Zhang L, Wu S, Liu J, Ping M, Yang W, Fu F. Isolation of aptamers with excellent cross-reactivity and speciﬁcity to sulfonamides towards a ratiometric fluorescent aptasensor for the detection of nine sulfonamides in seafood. Talanta 2024; 277:126380.

[18]	Li X, Yang Z, Waniss M, Liu X, Wang X, Xu Z, et al. Multiplexed SELEX for sulfonamide antibiotics yielding a group-speciﬁc DNA aptamer for biosensors. Anal Chem 2023; 95(44):16366-73.

[19]	Yang W, Luo D, Zheng S, Zhang Y, Wang Z, Fu F. Screening of cross-reactive aptamers for the detection of 24 quinolones by using a Liebig’s law-guided parallel-series strategy. Anal Chem 2024; 96(21):8576-85.

[20]	Wang L, Liu X, Zhang Q, Zhang C, Liu Y, Tu K, et al. Selection of DNA aptamers that bind to four organophosphorus pesticides. Biotechnol Lett 2012; 34 (5):869-74.

[21]	European Medicines Agency. Sales of veterinary antimicrobial agents in 31 European countries in 2022: trends from 2010 to 2022:thirteenth ESVAC report. Report. Luxembourg: Publications Ofﬁce of the European Union; 2023.

[22]	Duan W, Cui H, Jia X, Huang X. Occurrence and ecotoxicity of sulfonamides in the aquatic environment: a review. Sci Total Environ 2022; 820:153178.

[23]	Zuo X, Han A. The residues and risk assessment of sulfonamides in animal products. J Food Qual 2021; 2021:5597755.

[24]	Odundo F, Ngigi A, Magu M. Sulfonamides and b-lactam antibiotic residues and human health risk assessment in commercial chicken meat sold in Nairobi City, Kenya. Heliyon 2023; 9(8):e18810.

[25]	Chang CP, Hou PH, Yang WC, Wu CF, Chang CC, Tsai MY, et al. Analytical detection of sulfonamides and organophosphorus insecticide residues in ﬁsh in Taiwan. Molecules 2020; 25(7):1501.

[26]	Mesﬁn YM, Mitiku BA, Tamrat AH. Veterinary drug residues in food products of animal origin and their public health consequences: a review. Vet Med Sci 2024; 10(6):e70049.

[27]	The Ministry of Agriculture and Rural Affairs of the People’s Republic of China. GB 31650-2019: National food safety standard-maximum residue limits for veterinary drugs in foods. Chinese standard. Beijing: Standards Press of China; 2019. Chinese.

[28]	The Ministry of Agriculture and Rural Affairs of the People’s Republic of China. GB 31650.1-2022: National food safety standard-maximum residue limits for 41 veterinary drugs in foods. Chinese standard. Beijing: Standards Press of China; 2022. Chinese.

[29]	The Japan Food Chemical Research Foundation. Residue limits of agricultural chemicals. Japanese standard. Osaka: The Japan Food Chemical Research Foundation; 2006.

[30]

Commission Regulation (EU) No 37/2010 of 22 December 2009 on pharmacologically active substances and their classiﬁcation regarding maximum residue limits in foodstuffs of animal origin (text with EEA relevance) (revoked) [Internet]. London: legislation. gov.uk; undated [cited 2024 Apr 4]. Available from: https://www.legislation.gov.uk/eur/2010/37.

[31]	Part 556-tolerances for residues of new animal drugs in food[Internet]. Washington, DC: Code of Federal Regulations; 2019 Jul 11 [cited 2024 Apr 4]. Available from: https://www.ecfr.gov/current/title-21/part-556.

[32]	Cliquet P, Cox E, Haasnoot W, Schacht E, Goddeeris BM. Generation of group-speciﬁc antibodies against sulfonamides. J Agric Food Chem 2003; 51 (20):5835-42.

[33]	Pastor-Navarro N, García-Bover C, Maquieira A, Puchades R. Speciﬁc polyclonal-based immunoassays for sulfathiazole. Anal Bioanal Chem 2004; 379(7-8):1088-99.

[34]	Pastornavarro N, Gallego-Iglesias E, Maquieira A, Puchades R. Development of a group-speciﬁc immunoassay for sulfonamides. Application to bee honey analysis. Talanta 2007; 71(2):923-33.

[35]	Adrian J, Font H, Diserens JM, Sánchez-Baeza F, Marco MP. Generation of broad speciﬁcity antibodies for sulfonamide antibiotics and development of an enzyme-linked immunosorbent assay (ELISA) for the analysis of milk samples. J Agric Food Chem 2009; 57(2):385-94.

[36]	Franek M, Diblikova I, Cernoch I, Vass M, Hruska K. Broad-speciﬁcity immunoassays for sulfonamide detection: immunochemical strategy for generic antibodies and competitors. Anal Chem 2006; 78(5):1559-67.

[37]	Zhang H, Duan Z, Wang L, Zhang Y, Wang S. Hapten synthesis and development of polyclonal antibody-based multi-sulfonamide immunoassays. J Agric Food Chem 2006; 54(13):4499-505.

[38]	Ermolenko D, Eremin S, Martianov A, Zherdev A, Dzantiev B. A new generic enzyme immunoassay for sulfonamides. Anal Lett 2007; 40(6):1047-62.

[39]	Li JS, Li XW, Yuan JX, Wang X. Determination of sulfonamides in swine meat by immunoafﬁnity chromatography. J AOAC Int 2000; 83(4):830-6.

[40]	Wang Z, Li Y, Liang X, Zhang S, Shi W, Shen J. Forcing immunoassay for sulfonamides to higher sensitivity and broader detection spectrum by site heterologous hapten inducing afﬁnity improvement. Anal Methods 2013; 5 (24):6990.

[41]	Chen Y, Liu L, Xu L, Song S, Kuang H, Cui G, et al. Gold immunochromatographic sensor for the rapid detection of twenty-six sulfonamides in foods. Nano Res 2017; 10(8):2833-44.

[42]	Li C, Luo X, Li Y, Yang H, Liang X, Wen K, et al. A class-selec tive immunoassay for sulfonamides residue detection in milk using a superior p olyclonal antibody with broad speciﬁcity and highly uniform afﬁnity. M olecules 2019; 24(3):443.

[43]	Sheth HB, Sporns P. Development of a single ELISA for detection of sulfonamides. J Agric Food Chem 1991; 39(9):1696-700.

[44]	Muldoon MT, Font IA, Beier RC, Holtzapple CK, Young CR, Stanker LH. Development of a cross-reactive monoclonal antibody to sulfonamide antibiotics: evidence for structural conformation-selective hapten recognition. Food Agric Immunol 1999; 11(2):117-34.

[45]	Zhou Q, Peng D, Wang Y, Pan Y, Wan D, Zhang X, et al. A novel hapten and monoclonal-based enzyme-linked immunosorbent assay for sulfonamides in edible animal tissues. Food Chem 2014; 154:52-62.

[46]	Li ZB, Cui PL, Liu J, Liu JX, Wang JP. Production of generic monoclonal antibody and development of chemiluminescence immunoassay for determination of 32 sulfonamides in chicken muscle. Food Chem 2020; 311: 125966.

[47]	Cliquet P, Cox E, Haasnoot W, Schacht E, Goddeeris BM. Extraction procedure for sulfachloropyridazine in porcine tissues and detection in a sulfonamide-speciﬁc enzyme-linked immunosorbent assay (ELISA). Anal Chim Acta 2003; 494(1-2):21-8.

[48]	Haasnoot W, Cazemier G, Pre JD, Kemmers-Voncken A, Bienenmann-Ploum M, Verheijen R. Sulphonamide antibodies: from speciﬁc polyclonals to generic monoclonals. Food Agric Immunol 2000; 12(1):15-30.

[49]	Haasnoot W, Pre JD, Cazemier G, Kemmers-Voncken A, Verheijen R, Jansen BJM. Monoclonal antibodies against a sulfathiazole derivative for the immunochemical detection of sulfonamides. Food Agric Immunol 2000; 12 (2):127-38.

[50]	Assil HI, Sheth H, Sporns P. An ELISA for sulfonamide detection using afﬁnity-puriﬁed polyclonal antibodies. Food Res Int 1992; 25(5):343-53.

[51]	Li P, Wu W, Li Y, Qin Z, Zhang Y, Duan C, et al. Antibody recognition proﬁle-aided hapten design to modulate antibody generation with anticipated performance for immunoassay development. Anal Chem 2024; 96 (50):19869-79.

[52]	Li H, Ma S, Zhang X, Li C, Dong B, Mujtaba MG, et al. Generic hapten synthesis, broad-speciﬁcity monoclonal antibodies preparation, and ultrasensitive ELISA for ﬁve antibacterial synergists in chicken and milk. J Agric Food Chem 2018; 66(42):11170-9.

[53]	Zhang Y, Mi J, Wu W, Fei J, Lv B, Yu X, et al. Investigation of antibody tolerance in methanol for analytical purposes: methanol effect patterns and molecular mechanisms. Adv Sci 2024; 11(31):e2402050.

[54]	Ye B, Tian W, Wang B, Liang J. CASTpFold: computed atlas of surface topography of the universe of protein Folds. Nucleic Acids Res 2024; 52(W1): W194-9.

[55]	Bai Y, Liu R, Dou L, Wu W, Yu W, Wen K, et al. The influence of hapten spacer arm length on antibody response and immunoassay development. Anal Chim Acta 2023; 1239:340699.

[56]	Zhang Y, Wu W, Li Q, Zhou P, Wen K, Shen J, et al. The hapten rigidity improves antibody performances in immunoassay for rifamycins: immunovalidation and molecular mechanism. J Hazard Mater 2024; 469:133977.

[57]	Wu W, Li C, Liu D, Ji J, Zhu J, Lu H, et al. Ultrasensitive antibody production strategy based on hapten property for simultaneous immunoassay. Food Chem 2022; 395:133565.

[58]	Wang X, Song K, Li L, Chen L. Structure-based drug design strategies and challenges. Curr Top Med Chem 2018; 18(12):998-1006.

[59]	Sharma V, Wakode S, Kumar H. Structure- and ligand-based drug design:concepts, approaches, and challenges. In: Sharma N, Ojha H, Goyal RK,editors. Chemoinformatics and bioinformatics in the pharmaceutical sciences. New York City: Academic Press; 2021. p. 27-53.

[60]	Zhao G, Richaud AD, Williamson RT, Feig M, Roche SP. De novo synthesis and structural elucidation of CDR-H3 loop mimics. ACS Chem Biol 2024; 19 (7):1583-92.

[61]	Barrios Y, Jirholt P, Ohlin M. Length of the antibody heavy chain complementarity determining region 3 as a speciﬁcity-determining factor. J Mol Recognit 2004; 17(4):332-8.

[62]	Jeliazkov JR, Sljoka A, Kuroda D, Tsuchimura N, Katoh N, Tsum oto K, et al. Repertoire analysis of antibody CDR-H 3 loops suggests afﬁnity ma turation does not typically result in rigidiﬁcation. Front Immunol 2018; 9:413.