O, CLB), which is used in human and veterinary medicine as a therapeutic drug for the pulmonary disease, is a synthetic β2-adrenoceptor agonist [1
]. However, it has been often illicitly abused as a “lean meat agent” in the feed for pig and cattle to improve growth rate, and enhance lean meat-to-fat ratio. More and more investigations have demonstrated that clenbuterol is a medium cumulative drug and residues build in animals, which can lead to symptoms such as muscle chatter, palpitation, trembling, headache, nausea, and vomiting after human consumption of meat products. It is especially harmful to patients with diseases such as hypertension, heart disease, hyperthyroidism, and prostatic hypertrophy [2
]. Although there have been no major food safety incidents under the supervision of the government in recent years, according to online public opinion surveys, the illegal use of “lean meat agent” has always been one of the food safety issues that people are most concerned about [3
Current detection methods for clenbuterol mainly include enzyme-linked immunoassay (ELISA) [4
], gas chromatography coupled with mass spectrometry (GC-MS) [5
], high performance liquid chromatography (HPLC) [6
], liquid chromatography–mass spectroscopy (LC-MS) [7
], surface molecularly imprinted polymers [8
], electrochemical analysis [9
], capillary electro-phoresis [10
], and fluorescence biosensor [11
]. ELISA is a commonly used technique in rapid detection, but there is a problem of false positives in practical use [12
]. GC-MS, HPLC and LC-MS methods can accurately detect clenbuterol, however, there are disadvantages such as long detection cycle and many operation procedures. At the same time, these methods are time consuming and labor intensive, as well as expensive equipment are needed, which are not likely to be rapid, sensitive and appropriate detection for routine monitoring [13
Compared with the above detection methods, molecular imprinting technology (MIT) enjoys a unique advantage in rapid detection. This yields MIT a wide range of applications infood analysis [14
]. MIT is a process in which a target molecule is used as a template to prepare a polymer matrix, which can selectively rebind the template molecules from a mixture of closely related compounds [17
]. Because its recognition process is similar to the relationship between enzyme and substrate, molecularly imprinted polymer (MIP) is also called "artificial antibody" [19
]. Since Mosbach and co-workers [20
] reported on the results of the preparation of MIP of theophylline in the non-covalent approach, MIT technology has gained more and more attention. Recently, with the rapid development of in silico simulation, it is a hot topic in current research to assist in the rational design of MIPs system through computational chemistry [21
] and computer simulation plays an important role in determining the optimal functional monomer, optimizing the synthesis conditions of the imprinted substance, and elucidating the mechanism of molecular imprinting recognition [22
]. For instance, based on the density functional theory (DFT), Maouche and Mazouzet al. [25
] adopted quantum chemical calculation to determine the nature of interactions between each analyte and the polypyrrole matrix and the dopamine imprinted polypyrrole sensing layer. It is worth mentioning that recently Terracina et al. [27
] developed a novel in silico method for computationally imprinting and characterizing enantioselective binding sites, which makes a new progress in elucidating the mechanism of imprinting enantioselectivity. Besides, there are also multiple experimental methods utilized to guide the rational design of MIPs. For example, pre-polymerization mixtures, changes in spectral properties, thermodynamic properties, and electrochemical parameters before and after interaction of the template with the monomer can be measured by NMR [28
], differential scanning fluorometry [29
], isothermal titration calorimetry (ITC) [30
] and conductivity measurement [31
], which enables quantitatively study of binding affinities and unravel the mechanisms underlying molecular interactions.
More recently, different approaches have been used to synthesize MIPs for CLB detection, practical application of this technique, however, is still lacking, and this study would be conducive to fill this gap in the context of the increasing attention of computational chemistry. Although we have previously reported the results of a novel molecularly imprinted sensor array for the detection of CLB and its metabolites [32
], our focus, in the past, was on whether MIP-QCM (quarzt crystal microbalance) could be established and the construction of molecular imprinting systems was sloppy. Hence, more thorough investigations into the rational design of MIPs for both CLB and its metabolites are required. In order to achieve this aim, DFT and AIM-based computational and theoretical approaches were applied in this work for describing, predicting and analyzing molecular imprinting systems. CLB, 4-hydroxymandelic acid (HMA) and 4-Aminohippuric acid (AHA) have been selected as template molecules in the past [32
]. Wherein, HMA and AHA are alternative template molecules for two metabolites. In this study, phenylephrine (PE) will also be employed as a dummy template molecule for CLB, thus the designed array will be constituted of four sensors based on four high selective molecular imprinted polymers, which can be developed into a robust and cost-effective method suitable for simultaneous detection of CLB and its metabolites.
3. Results and Discussion
3.1. Theoretical Selection of Functional Monomer
Density functional theory (DFT) method in B3LYP level with 6-31G
(d, p) [47
] basis set has been widely applied to obtain the most stable configurations and binding energy for qualitative analysis of the hydrogen bonding-dominant weak interaction in molecular imprinting process [48
], because the property of electron cloud deformation could be effectively and accurately predicted in the modeling. However, in the simulation process, it was found that affected by the interference of some strong influence points, the change trends of △E
did not show strong consistency. Diffuse s- and p- functions for non-hydrogen atoms were then added in order to obtain a higher accuracy. It has proved that the diffuse function does make the simulation results more refined and high consistency of △E and △G
was achieved. The correlation coefficient between △E
reaches 0.955. Therefore, 6-31 + G
(d, p) [49
] was chosen as the basis set for the computational simulation.
In the meantime, it can be seen from the Figure 3
that two acidic functional monomers, AA and MAA, showed stronger binding capacity with clenbuterol and its metabolites compared to the neutral monomer AM and the basic monomer 4-VP. Considering the poor performance of the combination between AA and HMA, and MAA has been more commonly used in molecular imprinting, MAA was chosen as the functional monomer.
3.2. Theoretical Selection of Template Molecules and Determination of Functional Monomer Site of Action
Computing by molecular self-assembly, the template molecule and the selected functional monomer MAA are supposed to form a stable complex configuration. The spatial conformation of the complex, the sites of hydrogen bonding and the number of hydrogen bonds, all of these will all have a direct impact on the final imprinting effect. The construction of the initial position and conformation of the final stable complexes were calculated with the assistance of NPA charge and molecular electrostatic potential (MEP) electrostatic potential diagram in order to find out the possible coordination modes of the template compound with functional monomers.
The molecular electrostatic potential represents the attraction between the molecule and a proton, which is useful in rationalizing the interactions between molecules and molecular recognition processes. A map of the electrostatic potential onto the molecular surface of four templates and functional monomer MAA is shown in Figure 4
. According to the distribution of the electron cloud, the active sites can be directly predicted. On the basis of comprehensive consideration of the spatial conformation, the MEP map and the NPA charge of each atom were applied to analyze the active sites and to construct the template-monomer complex. As indicated in Figure 4
e, the proton donor of MAA is H12 and its proton acceptor is O10; the proton donors of CLB in Figure 4
a are H12, H13 and its proton acceptor is O34, N11, N19; while the proton donors of PE in Figure 4
b are H12, H21 and its proton acceptors are O11, O15, N20. In Figure 4
c,d, the template molecules are AHA and HMA respectively. Compared with CLB and PE, these two template molecules hold more active sites and their carboxyl and carbonyl groups have a higher reactivity. Proton donors of HMA are H12, H18, H20 and its proton acceptors are O11, O16, O17, O19; proton donors of AHA are H12, H13, H17, H24 and its proton acceptors are O15, O22, N11, N16 respectively. In addition, the NPA charges of the active sites of each template molecules are listed in the Table 2
, which is consistent with the MEP distribution analysis.
3.3. Formation of the Template-Monomer Complexes
The hydrogen bonding interactions between the active interaction sites play an important role in the formation of MIPs. Suitable molar ratio between template molecule and functional monomer will enable the prepared MIP with desired recognition property. In the present study, the different molar ratios were chosen for simulation. For each imprinting system, simulations were started from 1:1 molar ratio until the most stable conformations were reached. As the imprinting ratio increased, the binding energies were gradually reduced, and the complexes became more stable. Exceeding the upper limit of the optimum ratio, undesired hydrogen bonds could be formed via the non-characteristic bonding points between the monomers, which might lead to lower the selectivity of the synthesized polymers. The detailed binding energies changing trend is shown in Figure 5
The final configuration of the complex was depicted in the Figure 6
. The optimal conditions were obtained, which were as follows: The molar ratio of template to functional monomer for CLB and PE is 3, and the molar ratio of template to functional monomer for AHA and HMA is 5. Owing to more active molecular sites, both AHA and HMA template molecules have a larger molar ratio of template to functional monomer than CLB and PE, which requires more monomer to form a more stable conformation of the complex.
A quantitative analysis of the formation of hydrogen bond networks is illustrated in Table 3
, all formed hydrogen bond lengths were scaled from 1.60243 to 2.47659 Å, just between the range of the general O–H single bond and van der Waals radius. The mean hydrogen bond length of the imprinted molecule complexes was 2.26305, 2.06338, 1.76894, and 2.01016 Å for the template CLB, PE, AHA, and HMA, respectively. It can be seen that molecular imprinted complex constructed from template AHA has the lowest binding energy, which is not only related to the high molar ratio of template to functional monomer, but also to the existence of formed multiple hydrogen bonds in the complex. Linked by the double H bond, two adjacent molecules are approximately coplanar, which interaction pattern can improve the stability of the complex. Similarly, for molecularly imprinted complex formed with template CLB, the number of hydrogen bonds generated is relatively small and the bond length is relatively long due to less active sites on the guest molecules and relatively weaker activity as well.
3.4. AIM Topology Analysis
Koch et al. [50
] and Lipkowski et al. [51
] proposed eight general topological criteria for existence of HB interactions, of which the electron density of ρ(r) (density of all electrons) and ▽2
ρ(r) (Laplacian of electron density) should be within the ranges of 0.002~0.035 a.u. and 0.024~0.139 a.u., respectively. With respect to the electron density characteristics obtained for the complexes studied, Rozas et al. [52
] suggest that these criteria can be used to characterize HBs, i.e., when▽2
ρ(r) > 0, H(r) > 0, the formation of the electrostatic interaction between the molecules formed weak hydrogen bonds; when▽2
ρ(r) > 0 and H
(r) < 0, there is a moderate hydrogen bond between molecules; when▽2
ρ(r) < 0, H
(r) < 0, there is a strong interaction between molecules, and most of them are covalent. The resulting calculated properties of the electron density ρ(r) are shown in Table 3
. The topological analysis of the V
(r) (electrostatic potential) is the resultant at each point r, which is the net electrostatic effect produced at the point r by both the electrons and nuclei of the molecule. A positive (negative) value reveals that the electrostatic potential at r is dominated by the charge of the nucleus (electron). The results of the weak interaction between templates and monomers analyzed by Multiwfn are also shown in Table 4
For adduct CLB + 3MAA, the maximum value of ρ(r) at BCP in hydrogen bond is 0.043535397 a.u. and the minimum is 0.01507693 a.u. The range of ▽2ρ(r) is within 0.038173768–0.127935635 a.u. The ρ(r) at BCP-O35 exceeds 0.035 a.u., suggesting that the bond formed at the BCP has a strong large hydrogen bond and a significantly shorter bond length (1.69824 Å). The average hydrogen bond energy of CLB + 3 MAA is 4.89 Kcal/mol.
For adduct PE + 3MAA, the maximum value of ρ(r) at the BCP in the hydrogen bond is 0.03996 a.u. and the minimum is 0.00975 a.u. The range of ▽2ρ(r) value is within 0.03516–0.10997 a.u. The ρ(r) of BCP at O23 is over 0.035 a.u., indicating that the hydrogen bond strength is stronger than other hydrogen bonds and the hydrogen bond length is also short (1.75467 Å). The average hydrogen bond energy of PE +3 MAA is 5.89 Kcal/mol, which is a reasonable value as well.
The calculated results show that for the two adducts of AHA + 5MAA and HMA + 5MAA, there are 4 and 5 BCPs with large values of ρ(r) respectively. These sites are formed by the template molecule with a carboxyl group, a carbonyl group and an alcoholic hydroxyl group bonded to the carboxyl group of the monomer MAA. Due to the simultaneous donation of double hydrogen bonds, it can facilitate the tight association of the auxiliary and substrate, and thus appears to be a particularly effective method for polymer synthesis.
3.5. Theoretical Selection of Crosslinker
In order to make the imprinting process more effective, it is hoped that the functional residues derived from the functional monomers can be uniformly distributed in the entire cross-linked networks. The main function of the crosslinking agent in the entire imprinting process is to copolymerize with the functional monomer and to fix the three-dimensional structure of the monomer-template complex in space. In prepolymerization and copolymerization, the crosslinking agent may complex with both the functional monomer and the template molecule via hydrogen bonding or electrostatic interactions. The formation of such interference complexes is difficult to experimentally define. With the help of computational simulation and calculation, some of the undesirable factors could be avoided. As a result (Figure 7
), for CLB, PE, and HMA templates, when EDGMA was used as the crosslinking agent, the interaction between the template molecule and the crosslinker was minimal, that is, the crosslinker has minimal interference with the imprinting process. Compared with crosslinker TRIM or PETRA, EDGMA can reduce unwanted interaction energy by about 3% to 7% for CLB, PE and HMA. For AHA, EDGMA was slightly less effective than TRIM, but it was not much different. Both EDGMA and TRIM had significantly better results than PETRA. EDGMA is a widely used crosslinker with a chemical structure similar to that of MAA. When the random copolymerization of MAA and EDGMA occurs, the product obtained by the crosslinking agent EDGMA and the functional monomer MAA can form a uniform distribution of carboxylic acid groups. In addition, given the limited advantages of TRIM over EDGMA in AHA composites and consistency in practice, the cross-linker was chosen as EDGMA.
3.6. Theoretical Selection of Solvent
In the molecularly imprinted polymer synthesis process, the solvent is also known as "porogen" in addition to playing the role of dissolving the polymerization reagents. This is because the solvent can provide a porous structure for the imprinted molecular polymer to increase the speed of bonding the template molecule during recognition. For non-covalent imprinting, the choice of solvent has a direct impact on the formation of non-covalent adducts between the monomer and the template and its imprinting effect.
As can be seen from Figure 8
, CHF and THF display lower solvation energy overall and therefore their imprint effect is limited, while MeOH show the strongest effect with each template molecule. The strong interaction between the template molecule and the solvent shields the molecular interaction sites and weakens the interaction with the MAA, making the molecular recognition effect of the imprinted polymer relatively poor. Previously, ACN was used in our laboratory and the solvation energy of the template molecules was relatively high. If CHF and THF were used instead, the specificity of the imprinted polymer could be further improved. However, in practice, it is also necessary to consider the solubility of the template molecule and the functional monomer in the solvent and the porosity of the solvent. Although there are still some realistic factors that need to be coordinated, this calculation result can provide an idea for the direction of improvement.
3.7. Selectivity Simulation
From the results of selective simulation, each single imprinted polymer shows the strongest binding energy for its corresponding target molecule, which reflects the selectivity of molecularly imprinted polymer. In the Figure 9
, the binding energy of two guest molecules, AHA and HMA, is greater than that of CLB and PE, because the molar ratio of template to functional monomer of CLB and PE is 3, while for AHA and HMAs, the ratio is 5. This is consistent with the judgment in the previous subsection that both AHA and HMA molecules have more active sites.
As indicated in Figure 9
, each designed imprinted system exhibits a higher specificity for its target molecule, however, it is worthy to be pointed out that in practical applications there are rather few false positive responses depending on the individual molecularly imprinted polymer. Because even a single molecularly imprinted sensor produces a higher response signal, this does not prove to be caused by the target molecule to be captured. In terms of our approach, simultaneous detection of CLB and its metabolites may reduce false positives, because the data measured from the sample is no longer a single dimension of CLB content, but rather the simultaneous detection of CLB together with its metabolites in the pig urine sample, which results in an increase in the data dimension. The data will be then processed by statistical or intelligent algorithms, so that the analysis results for the sample to be tested are more accurate. The selectivity simulation described in this article makes some predictions on the possible situations and provides the forecast and theoretical support for the interpretation of the actual detection.
3.8. Experimental Verification
Finally, on the basis of the optimal experimental conditions obtained by the simulation, MIPs were synthesized on gold electrode surface of the QCM sensors. To demonstrate the applicability of this method, the fully integrated sensor array was applied to detect CLB and its two metabolites. Standard curves for CLB sensor, AHA sensor, and HMA sensors were plotted and illustrated in Figure 10
. The R2
values of the three sensors reached 0.9935, 0.9927, and 0.9918, respectively. The 1 × 10−8
M Clenbuterol solution was equivalent to a mass concentration of about 3 μg/L, and the selectivity of the sensor was also investigated. As can be seen from the Figure 11
, each sensor could specifically recognize the target molecules and the respective signal responses toward the target solutions were 3 times larger than the response values toward non-target solutions (For CLB-MIP-QCM, AHA, and HMA ethanol solutions are non-target solutions; for AHA-MIP-QCM, CLB and HMA ethanol solutions are non-target solutions; for HMA-MIP-QCM, CLB and AHA ethanol solutions are non-target solutions). The experimental results therefore verified the theoretical predictions and found to be in good agreement revealing specific affinity of the prepared MIPs to each target analytes.
However, it is worth noting that, with regard to the real sample determination under the harsh environmental conditions, unknown factors, which may influence the combining capability of the MIPs, should be involved to modify the simulation model and to improve its applicability. In addition, in the future, we are also interested in expanding our screening approach to more complex analyte mixtures (i.e., CLB and its metabolite analytes present in swine urine samples simultaneously). Nevertheless, the method of computationally imprinting enantioselective binding sites allows for greater understanding of the mechanisms underlying MIP binding, and the proposed method provides a promising platform for fabricating simple, fast, and economical sensing system to detect trace amounts of contaminants in food samples.