Directed Evolution of a Glutathione Transferase for the Development of a Biosensor for Alachlor Determination

In the present work, DNA recombination of three homologous tau class glutathione transferases (GSTUs) allowed the creation of a library of tau class GmGSTUs. The library was activity screened for the identification of glutathione transferase (GST) variants with enhanced catalytic activity towards the herbicide alachlor (2-chloro-2′,6′-diethyl-N-(methoxymethyl)acetanilide). One enzyme variant (GmGSTsf) with improved catalytic activity and binding affinity for alachlor was identified and explored for the development of an optical biosensor for alachlor determination. Kinetics analysis and molecular modeling studies revealed a key mutation (Ile69Val) at the subunit interface (helix α3) that appeared to be responsible for the altered catalytic properties. The enzyme was immobilized directly on polyvinylidenefluoride membrane by crosslinking with glutaraldehyde and was placed on the inner surface of a plastic cuvette. The rate of pH changes observed as a result of the enzyme reaction was followed optometrically using a pH indicator. A calibration curve indicated that the linear concentration range for alachlor was 30–300 μM. The approach used in the present study can provide tools for the generation of novel enzymes for eco-efficient and environment-friendly analytical technologies. In addition, the outcome of this study gives an example for harnessing protein symmetry for enzyme design.


Introduction
Glutathione transferases (GSTs; EC. 2.5.1.18) are a widely spread family of enzymes found in both eukaryotes and prokaryotes [1,2]. They catalyze the conjugation of glutathione (GSH) with a range of hydrophobic xenobiotic compounds such as drugs, environmental pollutants, and pesticides, including chloroacetanilide herbicides [3][4][5][6][7][8][9]. Their catalytic versatility and wide-substrate specificity stem from their structural flexibility and active-site plasticity [3,[10][11][12][13]. They are homodimers or heterodimers. The isoenzyme GmGST4-4 from Glycine max is a homodimer protein with an open V-shaped configuration with 2200 Å 2 of the accessible surface buried at the interface, smaller than that of most other classes of GSTs (2800-3400 Å 2 ). Each subunit has two domains, a small α/β domain, at the N-terminus, and a large β-helical domain, at the C-terminus. The former hosts the GSH-binding site (G-site) on top of the α/β domain [3,14,15]. A hydrophobic pocket (H-site) overlaps the two domains and binds the hydrophobic substrates.
The ability of GSTs to conjugate GSH with xenobiotics has been exploited for the development of enzyme-based biosensors for the determination and monitoring of such compounds in biological and environmental samples [16][17][18][19][20][21][22]. A range of different GST isoenzymes has been used so far for the fabrication of optical, electrochemical or potentiometric biosensors for measuring pesticides, such as the herbicide atrazine [16], the insecticides DDT [17], malathion [18], dieldrin and spiromesifen [19], α-endosulfan [20], and molinate [21]. In addition, a GST-based biosensor has been developed for the determination of anticancer drugs [23].
Pesticides, although they are useful agricultural agents for fighting pests and can certainly increase food production, are also harmful to human health and environment [24][25][26]. Regulatory requirements have posed strict restrictions on maximum residue levels (MRL) that must be met before food products can enter a market [27][28][29]. As a consequence, there is a need for a robust analytical methods suitable for pesticides determination with high accuracy in food and environmental samples.
In vitro directed evolution is regarded as the most efficient method for expanding sequence diversity and creating enzymes with improved or novel properties [29,30]. Increasingly, directed evolution has been applied to create enzymes catalyzing reactions either in a more efficient way or reactions that have not been observed in nature by taking advance of the promiscuous nature of enzymes [30]. A range of different methods and techniques have been developed that have enabled the evolution of any protein [29,30], pathway [31], network [32], or entire organism of interest [33].
In the present work, a library of tau class GSTs was created using DNA shuffling. The library was activity screened, allowing the selection of an engineered enzyme (GmGSTsf) with high conjugating activity towards the herbicide alachlor. The selected enzyme was characterized, immobilized on polyvinylidenefluoride membrane, and was used for the construction of an optic biosensor for alachlor determination in water samples. Alachlor displays high toxicity to humans and animals [34,35]. Due to its extensive application in agriculture, contamination of food, water, and air has become significant and, as a consequence, adverse health effects affect all organisms [34,35].

Expression, Purification, and Screening of the Wild-Type and Mutant Enzymes
GmGSTUs were expressed in Escherichia coli M15 [pREP4] and purified by affinity chromatography as described by Axarli et al. (2008) [38]. Protein purity was judged by PAGE. For activity screening of the shuffled library, transformants were grown at 37 • C in LB medium (10 mL), and the GSH/alachlor conjugation activity in crude cell free extracts was measured as described by Skopelitou and Labrou (2010) [41].

Assay of Enzyme Activity and Kinetic Analysis
Enzyme assays and steady-state kinetic measurements using CDNB and GSH were performed according to published methods [40]. Initial velocities were determined in the presence of a fixed concentration of GSH (2.5 mM) and variable concentrations of CDNB (0.015-0.150 mM). Alternatively, CDNB was used at a fixed concentration (1.5 mM) and the GSH was varied (0.05-0.5 mM). The steady-state data were analyzed by non-linear regression analysis using the GraphPad Prism Software (Version 5.03).

Molecular Modeling
The structure of the engineered GmGSTU with high activity towards alachlor (GmGSTsf) was predicted using homology modeling (SWISS-MODEL) based on the structure of the parent enzyme GmGSTU4-4 (PDB code 2vo4), determined at 1.75 Å resolution [38]. The sequence identity between the two enzymes is 95.4%, allowing the construction of a reliable model. The reliability and quality of the model was evaluated using GMQE score (0.84) [42]. The global and per-residue model quality was assessed using the QMEAN (Z-score 0.9) scoring function [43]. Moreover, PROCHECK [44] was employed to further evaluate the quality of the modelled GST. Verify3D [45] was also used to evaluate whether the model has a similar fold to known protein structures. The program PyMOL (http://www.pymol.org/) was employed for the visualization of the protein structures. Non-covalent interactions were analyzed using the Protein Contacts Atlas [46] and Arpeggio server [47].

Immobilization of Enzyme and Assessment of Enzyme Activity
Immobilization of GmGSTsf was achieved using glutaraldehyde chemistry and deposition on a polyvinylidenefluoride membrane (Durapole). The procedure was performed as follows: a polyvinylidenefluoride membrane was immersed and incubated for 30 min in potassium phosphate buffer (50 mM, pH 8.5). Bovine serum albumin (5 mg) was dissolved in potassium phosphate buffer (50 mM, pH 8.5) containing enzyme (5 U, previously dialyzed against 50 mM potassium phosphate buffer, pH 8.5). Glutaraldehyde solution was added to the mixture (final concentration 8% v/v), stirred for 15 s, and deposited on the surface of the polyvinylidenefluoride membrane. The membrane was incubated at 4 • C for 45 min and then rinsed three times with potassium phosphate buffer (50 mM, pH 7.5) to remove unreacted compounds and reaction byproducts. The same procedure was also carried out for the preparation of enzyme-free membranes (control). The membrane-immobilized enzyme was placed on the inner surface of a plastic cuvette (4 mL).
Assessment of the activity of the immobilized GmGSTsf enzyme was carried out using the CDNB/GSH or the alachlor/GSH substrate system. The enzyme reaction (37 • C) was monitored using a pH electrode. The composition of the reaction was: potassium phosphate buffer (1 mM, 10 mM NaCl pH 6.5) containing 0.5 mM CDNB or 0.5 mM alachlor (dissolved in acetone), 1 mM GSH, and 0.5 units of enzyme. Alternatively, the reaction was carried out in potassium phosphate buffer (3 mL, 50 mM, pH 6.5), containing 2.5 mM GSH, 1 mM CDNB, and immobilized enzyme. The mixture was incubated at 37 • C. The progress of the reaction was followed by measuring the absorbance at 340 nm. The enzyme-free membrane was used as a control.
Steady-state kinetics analysis of the membrane-immobilized GmGSTsf was carried out using the CDNB/GSH substrate system in potassium phosphate buffer (0.1 M, pH 6.5). The assay mixture contained 2.5 mM GSH and variable CDNB concentrations (0.05 mM-1.0 mM). The reactions were carried out at 37 • C for 30 min. Samples were taken at 5 min intervals for measuring the absorbance at 340 nm.

Stability Analysis of the Free and Immobilized Enzyme
The stability upon storage of the free and membrane-immobilized GmGSTsf enzyme was studied at 4 • C in potassium phosphate buffer (1 mM potassium phosphate buffer, 10 mM NaCl, pH 7.5). Enzyme inactivation was followed by assaying the enzyme activity using the CDNB/GSH substrate system.

Colorimetric Assays for the GST-Catalyzed Alachlor Conjugation Reaction
The membrane-immobilized enzyme was placed on the inner surface of a plastic cuvette (4 mL). The GST-catalyzed alachlor conjugation reaction was monitored in potassium phosphate buffer (3 mL, 1 mM potassium phosphate buffer, 10 mM NaCl, pH 7.5) containing 2.5 mM GSH, 0.05 mM bromocresol green, and alachlor (0 mM to 0.3 mM). The reactions were incubated at 37 • C and spectra (300 to 750 nm) were recorded at time intervals (0, 10, 20 and 30 min). The absorbance change at 620 nm versus time was used for Symmetry 2021, 13, 461 4 of 15 measuring enzyme activity for each alachlor concentration. The slope of each graph was used for the construction of a calibration curve (slope versus alachlor concentration).

Determination of Alachlor in Natural Water Samples
Recovery experiments were achieved using natural tap water samples (collected from Athens water supply network), spiked with known concentrations of alachlor (0.02 mM, 0.04 mM, 0.08 mM, and 0.17 mM). A sample of zero alachlor concentration water was used as a control. The concentrations of spiked alachlor were calculated based on a calibration curve obtained as described above.

Screening of GmGSTUs Library for Identifying a Shuffled Enzyme for the Development of a Biosensor for Alachlor Determination
The isoenzymes GmGSTU2-2, GmGSTU4-4, and GmGSTU10-10 from Glycine max share 88-91% sequence identity; however, their specificity towards different electrophile xenobiotics varies significantly. Worthy of note, the specific activity of the three GmGSTUs for the herbicides atrazine or alachlor ranges between 0 to 0.76 nmol/min × mg and 8 to 166 nmol/min × mg, respectively [48,49]. Therefore, the natural catalytic promiscuity of the GmGSTUs provides a promising starting point for the development of engineering GmGSTU variants with engineered activity and/or specificity towards the pesticide alachlor.
The three isoenzymes GmGSTU2-2, GmGSTU4-4, and GmGSTU10-10 were used as parent sequences for directed evolution using the DNA shuffling method [39,40]. The GmGSTUs library was activity screened for the isolation of enzyme variants with improved catalytic activity towards the herbicide alachlor. One enzyme variant (GmGSTsf) that displayed the highest specific activity (Figure 1) was identified and selected for further characterization. This variant was sequenced ( Figure 2) and showed that it has been derived from GmGSTU4-4 with segments mainly from GmGSTU2-2 and a few from GmGSTU10-10. The sequence of GmGSTsf possesses eight mutations (see Table 1), a three amino acid deletion, and two-point mutations compared to the parent GmGSTU4-4 enzymes ( Figure 2). The GmGSTsf variant was expressed in E. coli, purified and subjected to steadystate kinetic analysis ( Table 2). The results confirmed that the enzyme displays improved k cat value towards alachlor and CDNB, the common substrate for assaying GSTs. Although the enzyme showed slightly increased K m value for both electrophile substrates, its specificity constants (k cat /K m ) appeared augmented for both substrates, suggesting that the engineered GmGSTsf enzyme exhibits improved specificity towards CDNB and alachlor compared to the parent enzyme GmGSTU4-4. Table 1. Summary of mutations and deletions found in GmGSTsf sequence (a derivative of GmGSTU4-4), in comparison with the parent enzymes GmGSTU2-2 and GmGSTU10-10. The other amino acid sequence is identical to that of GmGSTU4-4. Numbering refers to the parent enzyme GmGSTU4-4.

Ser2Gln
Spontaneous mutation Glu4Gln Spontaneous mutation Tyr32Ser GmGSTU2-2 Ile69Val GmGSTU2-2 Asn149Asp GmGSTU2-2 Val158Ile GmGSTU2-2 Tyr161Asp GmGSTU2-2 Deletion (166AlaTyrGlu168) GmGSTU2-2 Thr172Ser GmGSTU2-2 or GmGSTU10-10 Ile183Val GmGSTU2-2 or GmGSTU10-10 Met210Val GmGSTU10-10   Kinetic inhibition constant (K i ) and half-maximal inhibitory concentration (IC 50 ) were also measured to evaluate the effect of mutations on the affinity of the enzyme for alachlor. Since alachlor is a substrate for the enzyme, the inactive hydrolyzed analogue (nonsubstrate) of alachlor (hAlachlor; 2-hydroxyl-2 ,6 -diethyl-N-(methoxymethyl)acetanilide), was employed in the study and the results are shown in Figure 3 and listed in Table 3. The determination of K i and IC 50 values for the GmGSTU4-4 and GmGSTsf variants provides a direct estimation of the binding affinity of hAlachlor towards the enzyme. The analysis showed that the GmGSTsf variant displayed about 17-fold lower K i and 4-fold lower IC 50 values compared to the parent enzyme GmGSTU4-4, suggesting that the affinity of the variant has been improved significantly.   [50]. GmGSTU4-4 numbering and secondary structure (helices and arrows) are shown above the alignment. Beta turns are marked with TT. Conserved areas are shown shaded. The symbol ( ) above sequences indicates important amino acid residues. NCBI accession numbers for sequences are in parentheses: GmGSTU4-4 (AAC18566), GmGSTU2-2 (CAA71784), and GmGSTU10-10 (AAG34800.1). Table 3. Kinetic inhibition constant (K i ) and half maximal inhibitory concentration (IC 50 ) of hAlachlor towards the GmGSTU4-4 and GmGSTsf enzymes.

Molecular Modeling
Homology modeling was employed to predict the structure of the GmGSTsf variant and put the results of the kinetic analysis in a structural context. Figure 4a shows the position of amino acid residues that have been mutated during the DNA shuffling protocol in the GmGSTsf variant. Most of the mutated amino acids are located at regions that appear to be insignificant in terms of catalysis or binding affinity, because they are distant from the active site and are exposed to the solvent without participating, directly or indirectly, in key molecular interactions. The only exception is the Ile69Val mutation, which seems to have a profound effect on the structure and catalysis. The amino acid at position 69 is located in the short helix α3 (residues 66-78), to which previous studies have attributed important functional roles in substrate binding and catalysis. Helix α3 provides two key amino acid residues, Glu66 and Ser67, that directly participate in GSH binding. In particular, the side chain of Glu66 forms a strong ionic interaction with the amino group of glutamyl residue of GSH. The side chain of Ser67 is hydrogen bonded with γ-carboxylate group of GSH. Moreover, the side chains of Glu66 and Ser67 are involved in the formation of the electron-sharing network, a key determinant in the catalytic mechanism of GmGSTU4-4 [49] and other GSTs [51]. In the GmGSTU4-4, the total number of non-covalent interactions that Ile69 makes are significantly larger, compared to that of Val69 in the GmGSTsf variant (Table 4). Therefore, the Ile69Val mutation seems to affect the structural integrity of helix α3, causing, as a consequence, structural rearrangements to Glu66 and Ser67.

Molecular Modeling
Homology modeling was employed to predict the structure of the GmGSTsf variant and put the results of the kinetic analysis in a structural context. Figure 4a shows the position of amino acid residues that have been mutated during the DNA shuffling protocol in the GmGSTsf variant. Most of the mutated amino acids are located at regions that appear to be insignificant in terms of catalysis or binding affinity, because they are distant from the active site and are exposed to the solvent without participating, directly or indi- formation of the electron-sharing network, a key determinant in the catalytic mechanism of GmGSTU4-4 [49] and other GSTs [51]. In the GmGSTU4-4, the total number of noncovalent interactions that Ile69 makes are significantly larger, compared to that of Val69 in the GmGSTsf variant (Table 4). Therefore, the Ile69Val mutation seems to affect the structural integrity of helix α3, causing, as a consequence, structural rearrangements to Glu66 and Ser67.   (Table 1) are shown as stick representations and labeled; (b) Interactions of Ile69 in the structure of GmGSTU4-4 (PDB code: 2vo4). The bound inhibitor S-nitrobenzyl-glutathione (GTB) is shown as stick representation and colored according to the atom type. The amino acid residues Glu66 and Ser67 that contribute to the formation of the electron-sharing network are shown as stick representations and labeled. The second subunit in the dimer is coloured in grey; (c) Asteroid plot of interactions of Ile69 in the parent enzyme GmGSTU4-4 (PDB code: 2vo4). The inner ring shows direct individual contacts and the outer ring indirect individual contacts. The size of each circle is proportional to the total number of contacts involved. The diagrams were created by Protein Contact Atlas [46]; (d) Chord plot of the interactions between helix α3 (brown) and helix α6 (i, green) or helix α4 (ii, yellow-brown). Table 4. Comparison of non-covalent interactions involved with amino acid at position 69 in GmGSTU4-4 (Ile69) and GmGSTsf (Val69). The analysis was carried out using the Arpeggio web server [47]. The monomers of the GmGSTU4-4 form a two-fold symmetry. In each monomer, the N-terminal domain of the one subunit is in contact with the C-terminal domain of the second subunit, and vice versa. The loop α2-β2, the strand β3 and the helix α3 of the N-terminal domain interact with the helices α4 and α5 of the C-terminal domain. The helix α3 and α4 provides residues for the symmetric arrangement of the two subunits and form the lock-and-key motif. Besides, the amino acid at position 69 is part of the lock-and-key motif [38,49] that contributes to helix α4 stabilization and structural integrity. Worthy of note, helix α4 provides key amino acid residues, such as Tyr107 and Trp114, that affect product release and are the main determinants of both k cat and k cat /K m regulation [40].

Type of Interaction
The electron-sharing network has been established to play an important role in the catalysis of GSTs. It allows the glutamyl γ-carboxylate of GSH to accept the H + from the sulfhydryl group of glutathione and function as a base [40,49,51]. A water molecule is involved in the transfer of the H + from the sulfhydryl group to the γ-glutamyl carboxylate, affecting the rate of catalysis. In the GmGSTU4-4, the electron-sharing network is composed of the strictly conserved residues Arg18, Glu66, Ser67, and Asp103. We can speculate that conformation changes caused by the Ile69Val mutation may affect either the ionization of GSH or the rate of proton transfer from -SH to γ-glutamyl carboxylate, leading to enhanced reaction rate.

Enzyme Immobilization
The best suited enzyme for the development of an analytical biosensor should display high activity to place the threshold of sensitivity/detection as low as possible. Therefore, the improved catalytic and alachlor-binding capabilities of GmGSTsf variant prompted us to exploit it for developing a GST biosensor for the direct monitoring of alachlor in water samples.
For the construction of the biosensor, the GmGSTsf variant was immobilized directly on a polyvinylidenefluoride membrane by crosslinking with glutaraldehyde [16,52]. The functionalized membrane was placed on the inner surface of a plastic cuvette (Figure 5a). Glutaraldehyde was chosen as a cross-linking agent to generate a three-dimensional network consisting of GmGSTsf interconnected through its exposed amino groups with bovine albumin (BSA). BSA was used to provide mechanical and physicochemical stabilization of the system. Figure 5b shows the reaction progress of the catalytic reactions (CDNB and alachlor as substrates) using the membrane-immobilized GmGSTsf. The curves were obtained by measuring the rate of proton release (HCl) from the conjugation reactions (Scheme 1) using a pH electrode. The immobilized enzyme was subjected to kinetics analysis to determine whether its kinetic properties were affected by the immobilization reaction. The results are shown in Figure 5c. The enzyme displayed a Km 0.52 ± 0.0711 mM for CDNB, very close to that obtained with the free enzyme ( Table 2), suggesting that the immobilized enzyme had retained its kinetic features.  The immobilized enzyme was subjected to kinetics analysis to determine whether its kinetic properties were affected by the immobilization reaction. The results are shown in Figure 5c. The enzyme displayed a K m 0.52 ± 0.0711 mM for CDNB, very close to that obtained with the free enzyme ( Table 2), suggesting that the immobilized enzyme had retained its kinetic features.

Thermal Stability of the Immobilized Enzyme
To evaluate the storage stability of the immobilized enzyme, heat inactivation studies were accomplished ( Figure 6). The immobilized enzyme, along with the free enzyme, was incubated at 4 • C in potassium phosphate buffer (1 mM, pH 7.5) and subsequently assayed for residual activity. The free enzyme was inactivated almost completely in 15 days at 4 • C. However, the immobilized enzyme exhibited significantly higher stability and lost less than 30% of its initial activity after 15 days of incubation. These findings indicate that the immobilization of GmGSTsf on polyvinylidenefluoride membrane significantly improved the enzyme operational stability upon storage at 4 • C, an important consideration that determines the practical viability of the biosensor.

The Basis of an Optometric Biosensor for Alachlor Determination
The assay was based on the activity of GmGSTsf to catalyze the reaction of GSH with alachlor with the simultaneous production of H + . The amount of produced H + is proportional to the concentration of alachlor (Scheme 1). Therefore, the concentration of alachlor can be determined indirectly by colorimetric measurements of pH alterations in the presence of a pH indicator. The proposed optometric biosensor makes use of the chromogenic alterations of the bromocresol green in order to transduce the catalytic reaction into an optical signal. Optometric biosensors show several advantages compared to the common chromatographic methods, such as cost effectiveness, direct determination, and easy applicability in routine analysis [52,53]. Figure 7a shows typical spectra obtained during the reaction and Figure 7b illustrates the rate of absorbance changes at 620 nm. The slope was calculated for a 30 min reaction after initiation of the enzymatic reaction. The phosphate concentration in the buffer solution was 1 mM. Lower signals were obtained for higher buffer capacities, for a given substrate concentration; however, larger buffer concentrations (>2 mM) compromised sensor response. Figure 7c shows the calibration curve obtained using different concentrations of alachlor. It can be seen that the rate of absorbance change at 620 nm is linearly related to alachlor concentration within the concentration range of 0-300 µM.
Natural water samples were used for recovery experiments. The samples were spiked with known concentrations of alachlor and the amounts of alachlor were determined using the standard curve. The results are shown in Table 5 and Figure 7d. The study revealed that alachlor recoveries ranged between 92 and 112%, with mean value = 99.05 (N = 6), SD = 8.247 and Std. Error = 3.367. Correlation analysis between the added and found alachlor concentrations is shown in Figure 7d   The absorbance change at 620 nm was calculated as a difference between reference signal (control, spontaneous reaction in the absence of enzyme) and reaction signal; (d) Recovery experiments using natural water samples spiked with known amounts of alachlor. The concentrations of spiked alachlor (added alachlor) were calculated based on the standard curve (found alachlor).

Conclusions
In the present study we report on a protein engineering approach for the creation of a mutant glutathione transferase with improved catalytic performance toward the alachlor/GSH conjugation reaction. We showed that the symmetric structure of GmGSTU4-4 can be efficiently manipulated through in vitro directed evolution, making new enzymes with altered catalytic activity and specificity. The engineered variant was exploited for the development of an optometric GST-based biosensor for alachlor determination in water samples. The advantages of this biosensor include low cost, real-time detection, and a wide linear range.