Substructure-activity relationship studies on antibody recognition for phenylurea compounds using competitive immunoassay and computational chemistry

Based on the structural features of fluometuron, an immunizing hapten was synthesized and conjugated to bovine serum albumin as an immunogen to prepare a polyclonal antibody. However, the resultant antibody indicated cross-reactivity with 6 structurally similar phenylurea herbicides, with binding activities (expressed by IC50 values) ranging from 1.67 µg/L to 42.71 µg/L. All 6 phenylurea herbicides contain a common moiety and three different substitutes. To understand how these three different chemical groups affect the antibody-phenylurea recognition activity, quantum chemistry, using density function theory (DFT) at the B3LYP/6-311++ G(d,p) level of theory, was employed to optimize all phenylurea structures, followed by determination of the 3D conformations of these molecules, pharmacophore analysis, and molecular electrostatic potential (ESP) analysis. The molecular modeling results confirmed that the geometry configuration, pharmacophore features and electron distribution in the substituents were related to the antibody binding activity. Spearman correlation analysis further elucidated that the geometrical and electrostatic properties on the van der Waals (vdW) surface of the substituents played a critical role in the antibody-phenylurea recognition process.

hapten-antibody recognition mechanism in the absence of antibody physicochemical information. In particular, 2D-QSAR or 3D-QSAR methods, such as CoMSIA and CoMFA Contour, have been used in many studies to investigate the binding activities of polyclonal 14,18,19 and monoclonal 20,21 antibodies against structurally similar compounds. These methods, however, are only suitable for broad-specific antibodies that can recognize a large number of targets and are not suitable for antibodies that only recognize several compounds 22 .
In this study, a specific hapten was synthesized and conjugated to bovine serum albumin, as the immunogen, to produce a new anti-fluometuron polyclonal antibody. However, when we used the resultant antibody to develop a competitive ELISA, we encountered a general issue, as in most other immunoassays: the specific antibody could also recognize five other analogues with different activities. To explore how this antibody, elicited by one specific immunogen, could cross-react with other targets, the half maximal inhibitory concentration (IC 50 ) values, presented as the binding activity of the antibody, were calculated and analyzed. Subsequently, quantum chemistry and molecular mechanics methods were employed to explain how the substituents of the six phenylurea compounds caused the differences in binding activity. Lastly, the molecular surfaces were quantitatively analyzed to correlate the binding activities and the molecular surface descriptors of substituents in the phenylurea compounds.

Results
Antibody binding activity analysis. According to the two-dimensional structures of the phenylurea molecules ( Fig. 1), the 6 title compounds were divided into three groups. The first group consisted of fluometuron and fenuron, which have different substitutes at the R 1 position. The second group contained neburon and diuron, which have different substitutes at the R 2 position. The third group included, chlorbromuron and linuron, which have different substitutes at the R 3 position.
We performed a direct competitive ELISA to evaluate the binding activity of the resultant antibody. IC 50 and cross-reactivity (CRs) values were calculated to represent the binding activity and specificity of the antibody. These values are shown in Table 1.
The antibody with the best sensitivity to fluometuron (IC 50 = 1.67 µg/L) was used as the hapten. Although the structure of fenuron and fluometuron are identical except for the substituted group at R 1 , the IC 50 value of fenuron (11.61 µg/L) is obviously different from that of fluometuron, and the specificity dropped dramatically from 100% (for fluometuron) to 11.61% (for fenuron). This indicates that the lack of -CF 3 group at the R 1 position limited the antibody recognition of fenuron. In the second group, neburon and diuron have very different structures from fluometuron, but the antibody has the ability to sensitively recognize them with IC 50 values of 2.41 µg/L and 8.35 µg/L, respectively. The butyl group at the R 2 position in neburon is replaced by a methyl group in diuron, and this difference led to a more than 7-fold decrease in the binding activity. In the third group, the replacement of the chlorine atom by a bromine atom at the R 2 position in chlorbromuron (IC 50 = 9.77 µg/L)  increased the antibody-chlorbromuron binding activity dramatically. In addition, the comparison of the binding activity to linuron (IC 50 = 42.71 µg/L) and diuron (IC 50 = 8.35 µg/L) suggested that the different groups at the R 3 position, where a methoxy group replaced the methyl group, could cause an obvious decline in antibody activity for linuron. The ELISA results demonstrated that the differences in the hapten-antibody binding activity may be caused by the different substitutes in the otherwise structurally similar analytes. However, we cannot explain how different substitutes affect the binding ability just by examining the two-dimensional structural formulas. In the hapten-antibody recognition process, the 3D configuration of the hapten molecule should match that of the active pocket of the antibody, according to the geometric configuration matching mechanism 13,17,23 . Cross-reactions will occur if molecules have the same or similar geometric configuration as the hapten. To understand the recognition mechanism, the 3D configuration of phenylurea molecules were optimized (Fig. 2) and studied by employing quantum chemistry and molecular mechanics methods.
Conformational analysis and phenylurea molecules alignment. As seen from the optimized structures in Fig. 2, all molecules have similar bond lengths, bond angles and dihedral angles from atom a to atom j. Thus, they are well superposed based on their common structures, and the difference in the geometric configuration is mainly located in the different substitutes. Moreover, we can see that molecules in each group have a high degree of alignment based on their common structures (Fig. 3). In the first group, the fluometuron molecule has a very high fitting degree with fenuron based on their common structure (atoms a-j, R 2 and R 3 ), showing an alignment RMS value of 0.0190. Even so, the -CF 3 group at the R 1 position in fluometuron caused the carbon atoms on the phenyl ring to slightly bend out of the phenyl ring plane. This change contributes to an almost 7-fold gap in the IC 50 value determined from the geometric configuration. We speculate that the antibody raised by fluometuron contained a -CF 3 complementary cavity, and this cavity could spatially recognize certain bulky groups, such as -CF 3 .
For neburon, the existence of the bulky butyl group at the R 3 position also dramatically changed the general conformational structure that was exposed to the antibody, especially the dihedral angle of C 10 -C 5 -C 12 -C 13 (−164.4°). This value is obviously smaller than those of the other 5 molecules (are all more than 174.0°), and the steric hindrance of the butyl group caused the whole neburon molecule be poorly aligned with the diuron molecule (the alignment RMS value was only 0.9346). Furthermore, neburon shows a better binding ability than diuron. For the third group, the alkoxy group at the R 3 position in the chlorbromuron and linuron molecules also affects the spatial conformation of the methyl group at the R 4 position. Chlorbromuron has a high degree of alignment with linuron, and the alignment RMS value is 0.0106. However, the replacement of the chlorine atom by a bromine atom at the R 2 position resulted in the bond angles of C 1 -C 13 -C 11 (122.2°) and C 2 -C 11 -C 13 (121.9°) in chlorbromuron being larger than those in linuron. This scenario contributes to an almost 4-fold activity gap in the binding activity. The geometric configuration analysis shows that certain bulky chemical group at the R 1 and R 3 positions play a positive role the antibody-phenylurea binding process. Moreover, the antibody binding sites can also accurately recognize variations of the common structures of these molecules, resulting in differences in binding activity. Pharmacophore model analysis. Fig. 4 shows the pharmacophore models of the phenylurea molecules.
The pharmacophore is an ensemble of steric and electronic features of the molecule, such as hydrophobic centroids, hydrogen-bond acceptors or donors and aromatic rings, which are necessary for molecular recognition of a ligand by a biological macromolecule 24,25 . The pharmacophore model of each phenylurea molecule had an aromatic ring feature in the phenyl ring area, a hydrophobic centroid at the R 4 position and a hydrogen-bond acceptor at the carbonyl group position. The -CF 3 group at the R 1 position in fluometuron serves as a hydrophobic centroid. Simultaneously, the chemical features of neburon indicate that it cannot align well with diuron, though  their only difference is located at the R 3 group; the hydrophobicity of the butyl group may influence the steric configuration of the whole neburon molecule, resulting in a higher activity (IC 50 = 2.41 µg/L) than that of diuron. The chlorbromuron molecule has the same pharmacophore features as linuron but with a bromine atom, and all chlorine atoms serve as hydrophobic centroids. However, we cannot offer a reason for the different activities based simply on pharmacophore features. Except for the R 3 group, linuron has the same structure as neburon, but the oxygen atom in the R 3 group in the linuron molecule serves as a hydrogen-bond donor. This led to a decrease in activity, as suggested by the pharmacophore features.
There are some limits in exploring the activity differences according to the pharmacophore analysis. For example, the bromine atom in chlorbromuron and the chlorine atom in linuron serve as hydrophobic centroids, yet we cannot distinguish the differences between them. Furthermore, the hydrogen-bonding and hydrophobic interactions were essentially caused by the electron distribution on the molecular surface 26 . In this way, ESP is also an important and indispensable parameter in the study of the hapten-antibody recognition mechanism.
Molecular electrostatic potential analysis. The ESP has a unique role in the prediction and analysis of molecular recognition and is often helpful in demonstrating non-covalent molecular interaction properties 27,28 . By employing the ESP surface, we can determine the spatial regions in the molecular structure at which the molecular electrostatic potential is negative or positive. This helps visualize charged regions of a molecule and is qualitatively useful in the analysis of electrostatic interactions between the antibody and the title compound 13,29 .
Three-dimensional plots of the molecular electrostatic potentials of the phenylurea compounds are illustrated in Fig. 5. The ESP surface depicts the phenylurea compound size, configuration, and charge density. The ESP color code ranges from −40.24 to +50.90 kcal/mol. In addition, the minima and maxima of the ESP on the vdW surface were analyzed by Multiwfn and mapped onto the surface of the molecules. All the phenylurea compounds have areas of positive potential at the front of the hydrogen atoms arising from the lower electronegativity of these atoms. This is comparable to carbon or nitrogen atoms, and the most positive potential area (show in dark blue) is located on the surface of the hydrogen atoms associated with nitrogen atoms, where the global maximum point is located. These positive regions can be easily attracted by the negative potential regions of the antibody binding sites. The most negative ESP region (shown in dark red), with a global minimum lower than −32.6 kcal/ mol, is located at the front of the carbonyl oxygen atoms. This large negative value is due to the lone pair of electrons on oxygen, which means that this region can be easily attracted by the region of positive potential. The ESP over the phenyl ring carbons is moderately negative, and two minima with similar values are located above and below the ring, reflecting the abundant π electrons of the phenyl ring. As the R 1 and R 2 substituents are all There are no electron-withdrawing groups associated with the phenyl ring in fenuron, so more electrons are around its phenyl ring than around those of the other five phenylurea molecules. In addition, the minimum values in this area are −19.06 and −19.36 kcal/mol. For fluometuron, the -CF 3 group, an electron-withdrawing group, attract some electrons around the phenyl ring, resulting in a moderate electron density and three minimum points in front of the R 1 group, and this -CF 3 group also led a decrease in the electron density and minimum ESP value around the phenyl ring. We further speculated that the electron-withdrawing group at the R 1 position could change the electron distribution around the phenyl ring and the R 1 region, and this was reflected in the higher selectivity for the antibody by electrostatic interaction.
Molecules in the second group have almost identical surface charges, except for in the region of the R 3 group. As the electron-withdrawing effect is caused by chlorine atoms, there is a high electron density area and two ESP minima between the R 1 and R 2 groups. However, the four-carbon chain in neburon changes the electron distribution around the R 3 position, resulting in a lower electron density and several lower ESP maxima (+15.40 to +21.50 kcal/mol) compared with diuron (electron density of +22.95 and the ESP maxima of +27.74 kcal/mol and +27.96 kcal/mol). This indicates that the alkyl chain could scatter the electron distribution of the R 3 position and increase the antibody-phenylurea binding activity.
The alkoxy group in linuron reduced the binding activity, possibly because the oxygen atom attracted electrons around it and shared the electrons around the carbonyl group, resulting in a moderate electron density and a minima (−12.57 kcal/mol) around the alkoxy group. Compared with linuron, chlorbromuron also has an alkoxy group at the R 3 position. However, the chlorine atom at the R 2 position in linuron has a stronger electron-withdrawing ability than the bromine atom at R 2 in chlorbromuron. Thus, chlorbromuron has a lower electron density and smaller minima in this area, and the change in the charge distribution may result in the difference in activity.
Quantitative analysis of the molecular surface. Weak interactions in biochemical systems can be well predicted and explained by analyzing the ESP descriptors on the molecular surface 30,31 . The quantitative analysis of the molecular surfaces in this study is summarized in Table S1. The Spearman correlation analysis between the molecular fragment descriptors and antibody activity (IC 50 values) is summarized in Table 2. We can see that the parameters -including S + , P, A, Va + and B of the R 1 group; B of the R 2 group; and S − , P, I, Va, Va − and B of the R 3 group -have a positive relationship with the antibody activity, while S and S − of the R 1 group have a negative relationship with antibody activity. Among these parameters, B of the R 1 group has a key relationship with the antibody activity coefficient, which is significant at the 0.01 level (2-tailed). S + , S − , P, A and Va + of the R 1 group also have a crucial relationship with the antibody activity coefficient. All are significant at the 0.05 level. The results showed that the antibody binding activity is linked to geometrical and electrostatic descriptors on the surfaces of the phenylurea molecules, particularly the substituent groups. Specifically, the R 1 and R 3 groups played a key role in the antibody-phenylurea binding process.

Discussion
In summary, ELISA analysis confirmed that the obtained antibody could recognize six structurally similar analogs. By analyzing the geometrical conformation and physicochemical properties of the optimized structures, we speculated that the antibody binding pocket could recognize slight differences in the geometric conformation of the molecules. In addition, the negative electrostatic potential of the hydrophobic -CF 3 group at the R 1 position and the hydrophobic butyl group at the R 3 position also contributed to the molecule-antibody binding activity. Meanwhile, the alkoxy group, which served as hydrogen-bond donor, at the R 3 position decreased the binding activity. Through Spearman correlation analysis between the molecular descriptors and antibody activity, we found that S + , S − , P, A, Va + and B of the R 1 group had key relationships with the antibody activity. The results indicated that the geometrical and electrostatic properties on the vdW surface of the R groups, especially the R 1 and R 3 groups, played a key role in the antibody-phenylurea binding activity. This investigation demonstrated that we could explore the antibody recognition mechanism using computational chemistry techniques, despite lacking structural information of the antibody. This study also draws attention to the need to design new haptens to produce specific or broad-spectrum antibodies in the future.

Methods
Antibody preparation. In this study, a hapten with a three-carbon spacer arm was synthesized, and then, an immunogen-and enzyme-labeled antigen was obtained by conjugating the hapten to BSA and HRP via the active ester method. The hapten-BSA conjugates were used as the immunogen to immunize two 3-month-old male New Zealand white rabbits (numbers I and II). After the initial immunization and four additional immunizations, the rabbits were sacrificed, and the sera were obtained by centrifugation. Aliquots of the sera were stored at 4 °C in 50% saturated ammonium sulfate. Subsequently, the antibodies were purified by Protein A-Sepharose 4B immunoaffinity chromatography.
Direct competitive ELISA protocol. The resulting antibodies were evaluated by a conventional direct competitive ELISA method. First, 0.25 µg/well of antibody was coated on a polystyrene 96-well plate using 100 µL/ well of coating buffer (0.05 mol/L NaHCO 3 , pH 9.6). After overnight incubation at 4 °C, the excess binding sites were blocked by 200 µL of a blocking solution at 37 °C for 1 h. Then, a mixture of the standard solution and hapten-HRP solution were simultaneously added to each well. After 1 h of competing reaction at 37 °C, substrates A and B were added sequentially for coloration. Finally, the enzymatic reaction was stopped by adding 2 mol/L H 2 SO 4 . The absorbance was determined at a primary wavelength of 450 nm, with a reference wavelength of 650 nm, using a microplate reader. After each step, 250 µL of PBST (10 mM PBS, pH 7.4, 0.05% Tween-20) was used to wash the plate. IC 50 and cross-reactivity (CR) values were calculated to evaluate the activity of the resulting antibody 13,18,32 . Calculation of the antibody binding activity. Inhibition ratios and IC 50 values were calculated. The inhibition ratio was defined by the following equation: For this equation, A Control represents the absorbance value of the control group and A Inhibit and A Blank represent the absorbance values of the inhibited group and blank group, respectively. The IC 50 (µg/L) value was the analyte concentration when the inhibition ratio was 50%.

Geometry optimization and moleculars alignment. All structures were built in Gaussian 09
ES64L-G09RevE.01 (Gaussian, Wallingford, CT, USA) according to the phenylurea configurations in the Pubchem database. Then, DFT calculations with the B3LYP functional and 6-311++ G(d,p) basis set were performed to optimize the molecules using the Gaussian 09 package 13,33 . There were no imaginary frequencies for all structures, ensuring the local minimum was found. The Molecule Overlay dialog in Discovery Studio 2016 (Accelrys Software, Inc., San Diego, CA, USA) was used to conduct the three-dimensional molecular alignment. The alignment root-mean-square (RMS) value was introduced to measure the degree of molecular superposition in each group.
Pharmacophore model analysis. The Auto Pharmacophore Generation protocol in Discovery Studio 2016 was employed in this study. This protocol generated ten pharmacophore models for each molecule, and we chose the best one for analysis.
Molecular electrostatic potential analysis. The Gaussian 09 and Gaussian View 5 packages were employed to conduct the molecular electrostatic potential (ESP) analysis of the title compounds. Minima and maxima of the ESP were generated in the Multiwfn software package and mapped onto the vdW surface.
Quantitative analysis of the molecular surface. By defining the specific chemical groups (R 1 , R 2 and R 3 ) in the Multiwfn package, descriptors were calculated. Spearman correlation via SPSS software served to detect relationships between molecular descriptors of the molecular fragments of phenylurea and the recognition ability (IC 50 ) of a polyclonal antibody.