Characterisation of a diazinon-metabolising glutathione S-transferase in the silkworm Bombyx mori by X-ray crystallography and genome editing analysis

Previously, we found an unclassified glutathione S-transferase 2 (bmGSTu2) in the silkworm Bombyx mori that conjugates glutathione to 1-chloro-2,4-dinitrobenzene and also metabolises diazinon, an organophosphate insecticide. Here, we provide a structural and genome-editing characterisation of the diazinon-metabolising glutathione S-transferase in B. mori. The structure of bmGSTu2 was determined at 1.68 Å by X-ray crystallography. Mutation of putative amino acid residues in the substrate-binding site showed that Pro13, Tyr107, Ile118, Phe119, and Phe211 are crucial for enzymatic function. bmGSTu2 gene disruption resulted in a decrease in median lethal dose values to an organophosphate insecticide and a decrease in acetylcholine levels in silkworms. Taken together, these results indicate that bmGSTu2 could metabolise an organophosphate insecticide. Thus, this study provides insights into the physiological role of bmGSTu2 in silkworms, detoxification of organophosphate insecticides, and drug targets for the development of a novel insecticide.

Results X-ray structural analysis of bmGSTu2. We already overexpressed recombinant bmGSTu2 in bacteria, and purified it 18 . The purified protein was crystallised in a space group of P4 1 with unit cell dimensions a = b = 86.26 Å and c = 58.77 Å. The structure was solved and the phases were refined; Table 1 includes relevant data. The final R work and R free factors were 18.7% and 22.1% for resolutions of 42.3-1.68 Å, respectively, with rootmean-square deviations for bond lengths and angles of 0.007 Å and 0.832°, respectively. The Ramachandran plot data showed that 98.74% of the main-chain dihedral angles were in the preferred regions, 0.74% in the allowed regions, and 0.50% in the outlier regions.
In the bmGSTu2 structural model, the electron densities from the following regions were poor for modelling: Ala116 to Glu123, after Phe211 of the A chain, and from Pro117 to Tyr122. On the B chain of bmGSTu2, the electron densities from Ala116 to Glu123, after Gln209 of the A chain, and from Pro117 to Tyr122 were also poor for modelling. To examine whether Pro13, Tyr107, Ile118, Phe119, and Phe211 contribute to bmGSTu2 activity, we mutated these amino acid residues to Ala. The resulting mutants were named P13A, Y107A, I118A, F119A, and F211A. After purification from Escherichia coli, we detected a single band in each final preparation upon SDS-PAGE analysis (data not shown). The specific activities of bmGSTu2 mutants were compared to those of the wild-type enzyme toward CDNB (Fig. 3A) and diazinon (Fig. 3B). For both the substrates, the activities of the mutants were decreased. The Y107A mutant resulted in the most prominent decrease in activity among all the mutants tested to date. Establishment of the mutant allele for bmgstu2. We established a mutant bmgstu2 silkworm strain to better determine its function (Fig. 4). To create this mutant strain, we used TALEN, which is a promising genome editing tool to disrupt the target gene efficiently in the silkworm and other organisms 23 . This approach involves integration of the donor plasmid into the target genome locus using the TAL-PITCh method, a TALEN-based knock-in system 24,25 . This enables the discrimination of the wild-type and mutant allele easily. In the present study, we disrupted the bmgstu2 gene and inserted the GFP sequence as a reporter (Fig. 4). TALEN vectors designed against the coding region of the bmgstu2 gene were microinjected into 300 silkworm embryos using a mixture of the donor vector (PITCh vector), resulting in a number of GFP-positive G1 individuals. The genotyping analysis Figure 2. Amino acid residues of the electron-sharing network. Carbon atoms for bmGSTu2, agGSTD3-3, and GSH are represented by green, cyan, and magenta, respectively. Atoms of oxygen, nitrogen, and sulphur are red, blue, and yellow, respectively. Amino acid residues for bmGSTu2 and agGSTd1-6 are described in green and cyan, respectively. revealed that 15 individuals showed targeted gene disruption (Fig. S1). The sequencing analysis revealed that one of these individuals showed precise integration for both 5′ and 3′ junctions (data not shown), and we utilised this mutant strain for further functional analysis of bmgstu2.
Median lethal dose (LD50) and acetylcholine (ACh) measurements. We measured the LD50 levels for diazinon, a widely used organophosphate insecticide, in the mutant and control B. mori strains. The LD50 value in the mutant was decreased to 70% of that in the control strain of B. mori (Fig. 5). Acetylcholine (ACh) levels were also estimated by using acetylcholinesterase (AChE) and choline oxidase. Binding of organophosphate insecticide to AChE results in inactivation of AChE, which indicates that AChE is unable to metabolise ACh. Organophosphate insecticides exert their toxicity by allowing ACh to overact at its receptors in the central and peripheral nervous systems. Notably, ACh levels in the control strain were decreased to 70% of that in the knock-in strain (Fig. 6).

Discussion
In our previous study, we identified and characterised bmGSTu2 18 . We also found that bmGSTu2 was capable of metabolising diazinon. Mutation of putative amino acid residues in the G-site showed that Ile54, Glu66, Ser67, and Asn68 are crucial for enzymatic function.
In the present study, we focused on the crystal structure of bmGSTu2 to examine amino acid residues that contribute to the conjugation reaction of diazinon on GSH, as well as the physiological role of bmGSTu2 by genome editing analysis using TALEN. Additionally, we solved the crystal structure of bmGSTu2. Superposition of the structure of bmGSTu2 onto that of nlGSTD demonstrated that the α-helices and β-strands of bmGSTu2 were conserved across the structures (Fig. 1A,B).
Although amino acid residues in the G-site are conserved in B. mori GSTs 11,[26][27][28][29][30] , the sequence diversity of the H-site is attributed to substrate specificity 31 ; furthermore, this diversity may be the reason why substrate specificities of B. mori GSTs are varied. Our mutagenesis experiments indicate that the residues Pro13, Tyr107, Ile118, Phe119, and Phe211 in bmGSTu2 play important roles in its enzymatic functions. Among all the mutants tested, the Y107A mutation showed the most apparent decrease in activity. Another critical residue, Tyr107, is located in the H-site and is highly conserved among delta-class GSTs, whereas the Phe residue is conserved instead of Tyr among epsilon-class GSTs 32 . Similar results were obtained for A. dirus delta-class GST4-4 (adGST4-4). Tyr111 of adGSTD4-4, the corresponding residue to Tyr107 of bmGSTu2, was reported to contribute to substrate-binding and stabilisation of GSH 33 . Among the GSTs, the H-site of delta-class A. gambiae GSTs (adGSTd1-6) was the most similar. The H-site of agGSTd1-6 (PDB ID: 1PN9) is composed of residues that are mostly hydrophobic: Leu6, Ala10, Pro11, Leu33, Met34, Tyr105, Phe108, Tyr113, Ile116, Phe117, Phe203, and Phe207 34 . In the bmGSTu2 sequence, 5 of the 12 residues (Pro13, Tyr107, Ile118, Phe119, and Phe211) are equivalent to the above agGST1-6 residues, and the remainder (Val8, Phe35, and Ile110) exhibit similarity to those in agGSTd1-6.
Other sites proposed to be requisite for bmGSTu2 function are the lock-and-key motif, the small hydrophobic core, and an ionic bridge 21,22,35,36 . The lock-and-key motif is important for stabilising hydrophobic interactions of GST monomers 22,27 . An intersubunit motif (lock-and-key) was found in the dimer interface of bmGSTD, a delta-class GST 27 , wherein the 'key' residues (Glu66, Arg68, and Tyr100) are introduced into a hydrophobic Figure 5. The effect of diazinon on LD50 values. Day-1 fifth-instar larvae of the mutant (homozygote) and control (heterozygote) were exposed to diazinon solutions via direct contact with the larval abdomen. At 24 h post-treatment (on day 2), LD50 values were determined. Relationships between two variables were examined by one-way ANOVA, with a significance level at P < 0.05. Figure 6. The effect of diazinon on ACh content. Crude extracts were prepared from the whole bodies of mutant (homozygote) and control (heterozygote). ACh production was determined by a colorimetric assay described in the Methods section. The colorimetric intensity was normalised to body weight, in grams. The intensities are expressed as the means of triplicate experiments. The significance of differences between each group was calculated based on one-way ANOVA analysis (P < 0.05). region (Ala69, Leu99, Tyr100, and Ile103), the 'lock' of the neighbouring subunit. Given that bmGSTu2 contains Glu66, Asn68, Ala69, Leu99, Cys100, and Lue103 in its amino acid sequence, it may potentially conserve the lock-and-key system. A small hydrophobic core and an ionic bridge (Leu6, Thr31, Leu33, Ala35, Glu37, Lys40, and Glu42) contributing to the stabilisation of the α helix were reported in adGSTD4-4 33 . The corresponding residues in bmGSTu2 are Val8, Val33, Phe35, Ala37, Glu39, Thr42, and Asp44. The electron-sharing network can be divided into type I and II classes 21,22 . The type I electron-sharing network is exemplified by delta-, theta-, omega-, and tau-class GSTs, which have an acidic amino acid residue at position 64, whereas type II network GSTs (alpha, mu, and pi, and sigma classes) contain a polar amino acid residue (glutamate) at this site that is capable of interacting with the γ-glutamyl portion of GSH. We demonstrated that Gln66 is conserved in the bmGSTu2 sequence, which is a characteristic of type II networks.
To determine the physiological role of bmGSTu2 in B. mori, we successfully constructed a mutant of this gene using a genome-editing approach. We then exposed silkworm larva to diazinon and measured the resulting LD50 values and ACh levels in vivo. The decreased LD50 value observed for the mutant after diazinon exposure indicates that bmGSTu2 is involved in diazinon tolerance in vivo. The organophosphate insecticide diazinon is a specific inhibitor of AChE, which is a common neurotoxicity biomarker. Once diazinon binds to AChE irreversibly, AChE is unable to metabolise ACh. We found that the control strain preferentially contains ACh, compared to that in the mutant. This result indicates that the reduced ACh levels in the mutant silkworm strains after diazinon exposure did not occur through AChE inhibition. However, we did not observe a complete disappearance of ACh in the knock-in strain. Thus, there might be other detoxification enzymes responsible for diazinon degradation. This may also be the reason why the LD50 value was not decreased to more than 70%. Cytochrome P450 and esterase, for example, are major detoxification enzymes that are able to degrade organophosphorus insecticides 37 . The silkworm genome reveals the existence of these enzymes in B. mori. To understand their involvement in the insecticide detoxification system in this species, it may be useful to compare their detailed properties, such as expression rates, activities, substrate specificities, and resistance spectra. Investigations along these lines are underway in our laboratories.
In summary, we provide evidence that bmGSTu2 contributes to diazinon tolerance in B. mori. We identified amino acid residues of bmGSTu2 that play important roles in catalysis. We are currently attempting co-crystallisation of bmGSTu2 with diazinon or a suitable substrate analogue conjugate to aid in the determination of amino acid residues involved in bmGSTu2 catalysis. The existence of a bmGSTu2 homologue in other agricultural pests must be determined to understand their importance for other insecticide detoxification system. Together, our findings may facilitate the development of more effective and safe insecticides.

Methods
Crystallisation and preparation of heavy atom derivative. Recombinant bmGSTu2 was purified as described previously 38  Data collection and structural determination. For data collection, crystals were selected using a cryoloop and flash frozen with liquid nitrogen. X-ray diffraction data collections were performed using synchrotron radiation on a SPring-8 beamline BL44XU 39 with λ = 0.9000 Å for the native data set and with λ = 1.0070 Å (Hg peak) for the heavy atom data set, in a nitrogen vapour stream at 100 K. Data sets were integrated and scaled using the program DENZO and SCALEPACK as implemented in the HKL2000 program package 40 . Phasing was performed by the single-wavelength anomalous diffraction (SAD) method using the Hg-derivative in the program SHELX C, D, and E 41 as implemented in HKL2MAP 42 ; the initial model was constructed using the program ARP/wARP 43 . The model constructed using the Hg-derivative data was used as a search model for a molecular replacement method with the program Phenix Phaser-MR 44 against the native data. The program ARP/wARP was used for further model building. After manual adjustment using the program Coot 45 , refinement was carried out using the program phenix.refine 46 . Site-directed mutagenesis. Site-directed mutagenesis of bmGSTu2 were performed using the Quick-Change Site-Directed Mutagenesis kit (Stratagene Corp., La Jolla, CA), based on the manufacturer's instructions. The mutagenesis primers were as follows: 5′-ATCGGAATTAATTCGGATCCGAATTCATG-GTTCTAAAATTATATGCCGTTTCTGATGGCCCTGCTTCGCTCTCTGTGAGGCAAGCC −3′ for P13A, 5′-GTGTAGGCTGAGATATTAGCAGCATAGCTTGATAAGTTAAAAC-3′ and 5′-GTTTTAACTT ATCAAGCTATGCTGCTAATATCTCAGCCTACAC-3′ (antisense) for Y107A, 5′-GTGCGTTCGTAGTC AAAGAAAGCAGGTGCCAT TGTGTAGGCTG-3′ and 5′ -CAGCCTACACAATGGCACCTGC TTTCTTTGACTACGAACGCAC-3′ for I118A, 5′-TGGAGTGCGTTCGTAGTCAAAAGCTATAGGTGCC ATTGTGTAGG-3′ and 5′-CCTACACAATGGCACCTATAGCTTTTGACTACGAACGCACTCCA-3′ for F119A, and 5′-ATAGGGTGTATAGGATGGTTGAGGTGAGTCAAGTCTGGAGGGTTAGCGGC AGCATGTTGGATCTCTTTCATGG-3′ and 5′-ATCTCAGTGGTGGTGGTGGTGGTGCTCGAGTTAAT TTTTGATTTTTCGGATAGGGTGTATAGGATGGTT-3′ for F211A. An expression plasmid for preparation of the recombinant bmgstu2 was used as a template, and each mutagenesis was confirmed by DNA sequencing of full-length mutated cDNA. Measurements of enzyme activity. GST was assayed spectrophotometrically using 1-chloro-2,4-dinitrobenzene (CDNB) and 5 mM GSH as standard substrates 47 . GST activity was expressed as mol CDNB conjugated with GSH per min per mg of protein. The ability of bmGSTu2 to metabolise diazinon was determined by HPLC according to previous reports. The eluate was monitored at 246 nm for the detection of metabolites. Specific activity toward diazinon was determined on the basis of the corresponding peak area identified per mg of protein.

Construction of TALEN and PITCh vectors. TALEN vectors were constructed as described by Takasu et al.
(2014, 2016) 48,49 . The target site was selected within the coding region of the bmgstu2 gene, and the sequence around the target site was determined in the w1-pnd strain. TALEN was assembled using the Golden Gate TALEN and TAL Effector kit (Addgene, Cambridge, USA) and the TALEN backbone vector pBlue-TAL 23 . The mRNA was in vitro synthesised using the mMESSAGE mMACHINE T7 kit (Ambion, Carlsbad, USA). PITCh vector construction was carried out following the method described in Tsubota and Sezutsu (2017) 50 . Inverse PCR was carried out using the primers 5′-TGAGGCAAGCCTTGAAGTACCATTTGAACT AGCACCACCTGTTCCTGTAG-3′ and 5′-CGCGACAAGGCAGAGAGCGATGGAGGGCCACTCGAA TTAGATCTTTGG-3′ against the pBachsp90GFP-3xP3DsRed plasmid 24,51 . The PCR product was self-ligated, and the inserted sequence was checked using Applied Biosystems 3130xl (Life Technologies) after cycle sequencing with BigDye Terminator V3.1 (Life Technologies, Carlsbad, USA).

Microinjection.
Microinjection was carried out following the method described in Tamura et al. 52 . The TALEN mRNAs and PITCh vector were injected into w1-pnd embryos at the syncytial preblastderm stage. The TALEN mRNA concentration was 125 ng/µL each, and the PITCh vector concentration was 500 ng/µL.
Tissue dissection. Each tissue was isolated on ice from day-3 fifth-instar larvae and kept at -80 °C until use.
Day-1 fifth-instar larvae were exposed to various concentrations of diazinon solutions via direct contact with the larval abdomen. At 24 h post-treatment (on day 2), LD50 values were recorded. For genome-editing experiments, the w1-pnd (non-diapausing) strain was used to establish the mutant strains. The established strains were crossed with the w-c (diapausing) strain to maintain the stocks. These strains were reared using an artificial diet (Nihon Nosan Kogyo, Yokohama, Japan) at 25 °C under a 12-h light/dark photoperiod. Acetylcholine (ACh) measurements. The ACh content of silkworms was measured using a colorimetric acetylcholine assay kit (Cell Biolabs Inc., San Diego, CA, USA). Briefly, whole bodies of the silkworms were homogenised in chloroform/methanol (2:1, v/v). After centrifugation, the lower phase was collected and dried completely. The resulting extract was dissolved in chloroform/methanol/water (86:14:1, v/v/v) and used as crude extract. Acetylcholinesterase and choline oxidase were added to detect acetylcholine. Acetylcholine content was measured at a wavelength of 540 nm after incubation and estimated as the arbitrary colorimetric units normalised to the milligrams of body weight used.