Structure of the Major Cytosolic Glutathione S-Transferase from the Parasitic Nematode Onchocerca volvulus*

Onchocerciasis is a debilitating parasitic disease caused by the filarial worm Onchocerca volvulus. Similar to other helminth parasites, O. volvulus is capable of evading the host's immune responses by a variety of defense mechanisms, including the detoxification activities of the glutathione S-transferases (GSTs). Additionally, in response to drug treatment, helminth GSTs are highly up-regulated, making them tempting targets both for chemotherapy and for vaccine development. We analyzed the three-dimensional x-ray structure of the major cytosolic GST from O. volvulus (Ov-GST2) in complex with its natural substrate glutathione and its competitive inhibitor S-hexylglutathione at 1.5 and 1.8 Å resolution, respectively. From the perspective of the biochemical classification, the Ov-GST2 seems to be related to π-class GSTs. However, in comparison to other π-class GSTs, in particular to the host's counterpart, the Ov-GST2 reveals significant and unusual differences in the sequence and overall structure. Major differences can be found in helix α-2, an important region for substrate recognition. Moreover, the binding site for the electrophilic co-substrate is spatially increased and more solvent-accessible. These structural alterations are responsible for different substrate specificities and will form the basis of parasite-specific structure-based drug design investigations.

Onchocerciasis is a parasitic disease caused by the filarial worm Onchocerca volvulus. It is often called river blindness because of its most extreme manifestation and because the blackflies that transmit the disease breed in fast flowing waters. It is the world's second leading infectious cause of blindness. About 18 million people are infected with O. volvulus in Africa, the Arabian Peninsula, and Central and South America. Moreover, roughly 50 million individuals are at risk of acquiring the parasite. Onchocerciasis causes chronic suffering and severe disability. Furthermore, it significantly impedes the socio-economic development in affected communities (www. who.int/tdr/index.html).
Chemotherapeutic approaches to control parasite transmission and to treat onchocerciasis rely on drugs that only kill microfilariae and infectious larvae but not the adult parasites. So far this necessitates the continuous use of these microfilaricides until the adult worms die. Therefore, the development of drugs that also effectively kill the adult worms would greatly support the control and treatment of O. volvulus infections (1). Furthermore, the potential development of drug-resistant strains of the nematode also demands the identification of alternative drug candidates to control the disease (2).
O. volvulus, similar to other parasitic organisms, is capable of surviving in an immunologically competent host by employing a variety of immune evasion strategies and defense mechanisms, including the detoxification and the repair mechanisms of glutathione S-transferases (GSTs 1 ; EC 2.5.1.18). GSTs are an ancient and diverse superfamily of multifunctional proteins represented by a number of species-independent gene classes. They play prominent roles in detoxification metabolism by catalyzing the nucleophilic addition of reduced glutathione (GSH) to numerous endobiotic and xenobiotic electrophilic substrates, usually promoting their inactivation, degradation, and excretion (3). GSTs participate in the complex network of catalytic functions that protect tissue against oxidative damage by detoxifying hydroperoxides of phospholipids, fatty acids, and DNA before they become engaged in free radical propagation reactions that can ultimately lead to the destruction of intracellular macromolecules (4). Furthermore, a potential modulating activity, mediated through protein interactions, on oxidative stress responsive cascades has been proposed (5). Several additional functions attributed to GSTs include the binding, transport, and storage of hydrophobic ligands (6), the isomerization of maleylacetoacetate, and the regulation of stress kinases and apoptosis (7)(8)(9). The importance of GSTs as a promising therapeutic target is underlined by their significant role in a variety of clinically relevant processes, such as susceptibility to disease and drug resistance (10). In a parasitic context, it is especially important to consider their function in the regulation of oxidative stress response, in drug resistance, and possibly in the modulation of the host immune defense mechanisms.
In previous studies, we have characterized three different GSTs from O. volvulus (Ov-GST1-3) that demonstrate different physiological functions within the parasite. The extracellular glycosylated Ov-GST1 is located directly at the parasitehost interface and has GSH-dependent prostaglandin D synthase activity. It therefore has the potential to participate in the modulation of immune cell functions by contributing to the production of parasite-derived prostanoids. On the basis of gene structure, amino acid sequence, and immunological and kinetic properties, it was possible to place the Ov-GST1 in close vicinity to the -class (11,12). The Ov-GST3 belongs to the -class and represents a highly stress-responsive GST that is up-regulated in response to oxidative stress, presumably providing a defense against immune-initiated peroxidation of parasite membranes (13). The Ov-GST2 is the major cytosolic GST in O. volvulus, in which the intracellular amount estimated constitutes about 0.1% of the total protein content in adult filarial worms. The general distribution of the highly abundant enzyme in all tissues and life stages of the parasite indicates its essential function. We propose that one of the biological functions of the Ov-GST2 is to neutralize, via GSH conjugation, glutathione peroxidase activity, or passive binding, the cytotoxic lipid peroxidation products arising from immune-initiated attack on parasitic membranes (14,15). Representative crystal structures of -class GSTs from different mammalian hosts, such as bovine (16), murine (17), porcine (18), and human (19 -21) are available and can be used for structural comparisons. Homology modeling and sequence alignment predicted the Ov-GST2 to be topologically related to the mammalian -class GSTs, and it was suspected that the hydrophobic substrate-binding site of the parasitic enzyme is more solventaccessible than the typical -class enzyme of the host (13). Only a few crystal structures of GSTs from endoparasites are known, namely those from Schistosoma japonica (22) and Schistosoma hematobium (23), the liver fluke Fasciola hepatica (24), and the malarial parasite Plasmodium falciparum (25,26). The essential role of Ov-GST2 in detoxification emphasizes this protein as a viable drug target against filarial parasites.

EXPERIMENTAL PROCEDURES
Heterologous expression of the Ov-GST2 in Escherichia coli, purification, and crystallization were performed as described previously (27).
Diffraction data were collected with synchrotron radiation at the consortium beamline X13 (Hamburger Synchrotron Laboratory/Deutsches Elektronen-Synchrotron). The programs DENZO and SCALE-PACK (28) were used for data analysis and reduction. Further calculations were performed applying the CCP4 program suite (29).
For crystals of the Ov-GST2⅐GSH complex, the space group was assigned to be P2 1 with cell dimensions of a ϭ 51.6, b ϭ 82.3, c ϭ 56.7 Å, and ␤ ϭ 95.9°. The asymmetric unit contains one dimer, resulting in a Matthew's coefficient of 2.49 Å 3 /Da (30) and a corresponding solvent content of 50.3%. For crystals of the Ov-GST2⅐S-hexylglutathione complex, the space group was also assigned to be P2 1 but with cell dimensions of a ϭ 51.9, b ϭ 82.0, c ϭ 107.11 Å, and ␤ ϭ 91.2°, with a c-axis almost twice as long as that for crystals of the Ov-GST2⅐GSH complex. Accordingly, the asymmetric unit contains two dimers, resulting in a Matthew's coefficient of 2.36 Å 3 /Da (30) and a corresponding solvent content of 47.5%. Statistics for the datasets are summarized in Table I.
The structure of the Ov-GST2⅐GSH complex was determined by molecular replacement, using the program package AMoRe (31). A -class GST dimer from human placenta (Protein Data Bank entry code 1AQW) was used as a search model. All data within the resolution range of 10 -4 Å were used for the calculations, revealing the position of the dimer in the asymmetric unit. A rigid body refinement of the initial model reduced the R-factor from 45.2 to 39.7% for all data in the resolution range of 20 -3 Å The structure of the Ov-GST2⅐S-hexylglutathione complex was also solved by molecular replacement later on using the refined structure of the Ov-GST2 as a search model. Two dimers were positioned in the asymmetric unit. All non-conserved res-  idues in the search model were replaced by alanines. Insertions and deletions were not included in the first refinement cycle. Both complex structures were refined using similar protocols. The refinement was performed by program REFMAC5 (32), and the progress was monitored by the continuous decrease of the free R-value (33). Because the asymmetric unit contained two GST monomers, use was made of non-crystallographic symmetry restraints for all non-hydrogen atoms throughout the course of refinement. The CCP4 program suite (29) and the program TURBO-FRODO (34) were used for calculations and model building, respectively. The ligands were identified by difference Fourier techniques in all G-sites of the molecules in the asymmetric unit. Solvent molecules were automatically added at the end of the refinement process using the program wARP (35) and checked for chemically reasonable positions where difference density exceeded 3. The quality of the final models was verified using the program PROCHECK (36). A summary of the refinement statistics is shown in Table I.

RESULTS
The structures of Ov-GST2 in complex with the physiological substrate GSH and the competitive inhibitor S-hexylglutathione were determined independently to characterize the substrate binding site (G-site) and the electrophilic binding site (H-site) of the enzyme. Both complexes have been crystallized under identical conditions, and the space group was assigned to be P2 1 . Interestingly, although crystals of Ov-GST2 in complex with GSH reveal one dimer/asymmetric unit, crystals of Ov-GST2 in complex with S-hexylglutathione reveal two dimers/ asymmetric unit. The structure of the Ov-GST2⅐GSH complex was refined to a resolution of 1.5 Å (r ϭ 15.0%, R free ϭ 17.6%) and of the Ov-GST2⅐S-hexylglutathione complex to 1.8 Å (r ϭ 19.9%, R free ϭ 22.3%). No significant changes in the tertiary structure of both complexes were observed. The root mean square (r.m.s.) deviation of the C␣ atoms is within the range of deviations calculated between each monomer of one structure. The superimposed monomers merely show a r.m.s. deviation in the range of 0.3-0.4 Å for all C␣ atoms. The G-site seems to be relatively rigid, and the overall structure is not notably affected by different ligands. The bound ligands do not affect the crystal packing, as they are not involved in any crystal contacts. Similar effects are also observed for the human -class (19) and ␣-class (37) enzyme. The high resolution obtained facilitates the determination of solvent molecules with high accuracy. Both structures are well ordered, and the final models include all residues of the enzyme. The positions of the ligands are clearly defined by the electron density. The statistics of the model quality are summarized in Table I. The overall fold of Ov-GST2 shows that it is clearly related to -class GSTs with secondary elements given in Figs. 1 and 2. The monomers consist of two domains linked by a loop between the helices ␣-3 and ␣-4. Domain 1 (residues 1-74) is built from a four-stranded mixed ␤-sheet that is surrounded by 3 ␣-helices (␤␣␤␣␤␤␣ motif). Domain 2 (residues 82-208) is largely ␣-helical. The typical ␣-class extra helix ␣-9, the most important structural feature to distinguish between ␣and -class enzymes, is not present in this structure.
There are two residues that adopt cis-conformation. One is a characteristic proline residue (Pro 51 ), found to be highly conserved in all GSTs, which is located in the loop that connects ␣-2 to ␤-3. The other residue, Leu 50 , is a G-site residue that has two hydrogen bonds to GSH and S-hexylgutathione and also adopts a cis-configuration in both complexes.
In their biologically active form, cytosolic GSTs are dimers, and Ov-GST2 conforms to that rule. The dimeric structure is stable, with molecular recognition at the subunit interface being class-specific. Intersubunit contacts are mediated by hydrophobic contacts as well as by hydrogen bonds. Intersubunit hydrogen bonds and salt links between the monomers within 3.5 Å are: Phe 47A -Lys 127B , Arg 68A -Asp 88B , Arg 72A -Glu 81B , and Arg 72A -Thr 85B . All contacts are related by local symmetry. Particularly, the side-chain of residue Arg 68 (helix ␣-3) serves as a hydrophilic anchor and forms two hydrogen bridges to the side-chain of residue Asp 88 (helix ␣-4) from the other subunit of the homodimer. This arginine-mediated intersubunit contact is very specific for ␣and -class enzymes (38).
In a classical -class enzyme, Tyr 49 acts as a hydrophobic "key." However, for Ov-GST2, Phe 47 (corresponds to Tyr 49 in the human -GST) (Fig. 1) acts as the hydrophobic key extending from the loop preceding ␤-3, which fits into the hydrophobic "lock" provided by the residues Gly 93 , Leu 131 , Lys 127 , and Phe 128 of the other subunit. This lock is located between helices ␣-4 and ␣-5. Comparison of the Ov-GST2 structure with the human -class enzyme yields an overall r.m.s. value on C␣ atoms of 1.5 Å (Fig. 3a), which is significantly higher than expected for members within the same class (Ͻ0.7 Å). With regard to sequence identity, this result is consistent with the low level of only 42% versus an expected sequence identity of Ͼ75%. In comparison with that, the human, pig, and mouse -class GSTs can be superpositioned with r.m.s. values of 0.5-0.6 Å. The most significant differences observed between the human -class enzyme and the Ov-GST2 occur in the regions of   (Fig. 3). Region 29 -42 includes strand ␤-2 and helix ␣-2, with a two-residue insertion (40 and 41) in the human enzyme ( Fig. 1) leading to a bent helical structure. It is highly remarkable that this region also involves a G-site residue, Phe 38 , which is replaced by Trp in the human, mouse, and pig enzyme. The largest deviation, however, is placed at the end of helix ␣-8 (residues 192-201). The sequence is not conserved, and the human enzyme reveals a distorted helix instead of a regular helix in Ov-GST2 in this region (Fig. 3).
The G-site part of the active site is formed by helix ␣-2, by residues connecting helix ␣-2 and strand ␤-3, and by a segment connecting strand ␤-4 to helix ␣-3. The overall geometry is similar but not identical to that of the human -class GST. The H-site is formed by the loop connecting strand ␤-1 and helix ␣-1, by the C-terminal part of helix ␣-4, and by residues of the C terminus following helix ␣-8. Residues in contact with GSH and the GSH part of S-hexylglutathione are Tyr 7 , Phe 8 , Leu 13 , Phe 38 , Gln 49 , Leu 50 , Ser 63 , Arg 95 , and Asp 96 (from the other monomer) as summerized in Table II. In Ov-GST2, the ␣-carbonyl group of the ␥-glutamyl moiety is aligned by two hydrogen bonds of Ser 63 and one hydrogen bond of Arg 95 (Fig. 4), which is a remarkable difference to the situation found in the human -class GST. Here, Gln 64 takes over the task of Arg 95 . The ␣-amino group interacts with Asp 96 from the other monomer, and the backbone of the ␥-glutamyl moiety is stabilized by van der Waals contacts to Leu 13 . In the human counterpart, Leu 13 is exchanged by Arg (Figs. 1 and 5). This exchange is the most drastic change, influencing the active site geometry and resulting in a reduction of the volume of the H-site binding pocket (Fig. 5). The backbone of the Cys moiety of GSH forms two hydrogen bonds with Leu 50 , a residue in cis-conformation. The Gly moiety is bound by Gln 49 , the backbone being further stabilized by van der Waals contacts to Phe 8 and Phe 38 . The third amino acid exchange in the G-site is Phe 38 against Trp in the human enzyme. In Ov-GST2, Phe 38 is only involved in van der Waals contacts, whereas in the human enzyme Trp forms a hydrogen bond to the Gly moiety. When comparing the Ov-GST2⅐GSH complex with the Ov-GST2⅐Shexylglutathione complex, there are no differences in the positions of residues participating in the G-site (Fig. 4). A similar positioning of the sulfur atom has been observed in -class enzymes from pig (38), mouse (17), and human (19 -21). It is generally accepted that GSH predominantly reacts as the anionic thiolate and that, in the -class, a network of hydrogen bonds may be responsible for the activation of the GSH sulfur. Already, it has been shown for the human -class enzyme that Tyr 7 stays in the identical position in both the apo-and the GSH-complexed structures (21). This observation supports the proposal that Tyr 7 will not be ionized in the reaction mechanism (39). This situation is also observed for the Ov-GST2 structure with Tyr 7 in hydrogen bonding distance to the sulfur atom of GSH and to the amide nitrogen of Leu 13 .
The entrance of the H-site can be described as a pocket with side walls formed by Phe 8 and Tyr 108 . The bottom is formed by Lys 35 , Ile 10 , and Gly 204 and the deep end by Leu 13 , Thr 102 , and Val 99 (Fig. 5b). The S-hexyl-moiety of S-hexylglutathione is sandwiched between the aromatic rings of Phe 8 and Tyr 108 and is kept in place by Ile 10 and Gly 204 . Phe 8 , Ile 10 , and Gly 204 are in van der Waals contacts within the range of 3.7-3.8 Å (Figs. 4 and 5). Tyr 108 has been shown to enhance Michael additions of GSH to electrophilic substrates (21,40). This residue is strictly conserved in all -class GSTs and hydrogen bonding interaction between the hydroxyl group of Tyr 108 , and the amide nitrogen of Gly 204 has also been observed in mouse, pig, and human -class structures (17,21,38). Even though mammalian -class GSTs share high levels of sequence identity, they can display large variations in their specific activities toward various substrates (41,42). These variations can be attributed to rather small differences in the H-sites of each enzyme. Direct comparison of Ov-GST2 with the host's counterpart reveals that only Phe 8 , Tyr 106 , and Gly 204 are conserved; all other H-site residues are exchanged, resulting in a different binding environment for potential co-substrates (Figs. 1 and 5). As already mentioned, the most crucial difference is probably the exchange of Leu 13 against Arg in the human enzyme. In the human enzyme, Arg 13 and Glu 97 form two hydrogen bridges. Because of the Leu-Arg exchange, more space of ϳ40 Å 3 is available for an electrophilic co-substrate in the region between Leu 13 , Thr 102 , Val 99 , and Arg 95 (Fig. 5, a and b). Moreover, the exchange of Thr 102 against Ile increases the H-site by 20 Å 3 once again.

DISCUSSION
The Ov-GST2 is structurally related to the -class GSTs. This GST class plays an important part in the detoxification of mutagens and toxins and is implicated in neoplastic development and drug resistance. High levels of -class GST expression have been shown to be directly associated with tumor drug resistance; however, the precise role played by the enzyme remains ill defined (10,(43)(44)(45)(46). One possibility is direct detoxification of the chemotherapy agent; alternatively, but not to the contrary, it may be the ability to interact with c-Jun Nterminal kinase (JNK) through direct protein-protein interaction that protects the cell against toxic metabolites. Here, the -class GST acts as a direct endogenous inhibitory regulator but also has a role in regulating the constitutive expression of specific phase II detoxification enzymes and antioxidant proteins that are downstream molecular targets of the JNK signaling pathway (10). Moreover, activation of the antioxidant enzyme 1-CYS peroxiredoxin by glutathionylation is mediated by the heterodimerization with -class GST (47). Interestingly, in response to drug treatment or exposure to pro-oxidants, helminth GSTs are highly up-regulated. Furthermore, overexpression of GSTs has been correlated to the ability of the helminth to detoxify immune-initiated cytotoxic products and has been associated with drug resistance (48 -50). Moreover, GST levels have been shown to increase in parasitic helminths during chronic infection, emphasizing the reliance of the parasite on this detoxification system (51,52).
Even though the overall structure of Ov-GST2 reveals major characteristic points of similarity toward the -class GSTs, there are some crucial points of differences. Although the Gsite is structurally very similar to the human counterpart, its H-site differs significantly, explaining the observed differences in the specificity toward the same substrate. In previous ex- periments, a series of ␤-carbonyl-substituted GSH conjugates was evaluated as potential inhibitors of Ov-GST2. Their specificity for the parasite-derived protein was assessed through comparison with their inhibition of human -class GST. Selective inhibition of Ov-GST2 has been demonstrated at low micromolar concentrations of ␤-carbonyl GSH conjugates. Selectivity for Ov-GST2 over human -class of Ͼ25-fold was generated through appropriate substitutions of the conjugate moiety. Aryl and long chain or bulky alkyl groups enhance Ov-GST2 selectivity (13). This is in accordance with the expectations of a more open active site in parasitic enzymes (53). Our x-ray structures prove that the H-site of Ov-GST2 is spatially increased in comparison to the host enzyme, and this is also in good agreement with the results discussed above. Only three residues of the H-site are conserved in comparison with the human counterpart, and we can use these structural differences to design specific inhibitors.
The human -class GST recognizes GSH by an induced-fit mechanism (54) and in the apoenzyme helix ␣-2 was especially identified as being rather flexible (55). Because we did not investigate the apo-form of the Ov-GST2, we cannot confirm this finding or refute it. However, it is noteworthy that the identified -class key residues for this induced-fit mechanism, Tyr 49 and Cys 47 , are not conserved in Ov-GST2 and are replaced by Phe (Fig. 1).
Oxidation of the human -class GST leads to the formation of an intramonomer disulfide bond between Cys 47 and Cys 101 , with subsequent inactivation of the enzyme (56 -58). Both sulfur atoms of these residues show a distance of Ͼ18 Å in the active enzyme, again indicating the high flexibility of helix ␣-2.
In Ov-GST2, however, the second Cys (Cys 101 ) residue is replaced by Val, making it debatable whether Ov-GST2 can be inactivated by oxidation at all. An exchange of Cys 47 to Ala made the human enzyme more resistant toward irreversible inactivation by 0.1 mM N-ethylmaleimide. Furthermore, the mutant Ala 47 was also more resistant to inactivation by the physiological disulfides, cystamine or cystine, which cause mixed disulfide and/or intra-or intersubunit disulfide bond formation (58).
The C-terminal part of helix ␣-8 (residues 192-201) is a region demonstrating large structural deviations between the human and parasitic GST. Helix ␣-8 is neighboring the loop that connects strand ␤-1 and helix ␣-1, and therefore it is in close distance to Val 10 and Arg 13 of the active site. Nevertheless, the distance to the H-site is Ͼ15 Å, and the biological meaning of this structural difference remains unclear.
Tyr 7 , one of the G-site residues, seems to be protonated, accepting a hydrogen bond from the main chain amide group of Leu 13 . Tyr 7 does not appear to play a role as a general base in the enzymatic reaction. Similar conclusions have been reported for other mammalian -class GSTs (19,21,59).
In the absence of a protective vaccine, control of onchocerciasis relies on measures directed against the vector and on chemotherapy. The only available drug for mass treatment of onchocerciasis is ivermectin (60). 2 Because control of the disease is at present largely dependent on ivermectin and will probably be so in the near future, it is a worrying possibility that resistance may be emerging in O. volvulus. Ivermectin resistance emerged relatively quickly in helminths of veterinary concern (62)(63)(64). Recent open case-controlled investigations, carried out in two onchocerciasis-endemic foci in Ghana, confirm the existence of populations of adult female O. volvulus that respond poorly to repeated doses of ivermectin (65,66). As there are no practical alternative strategies for the mass control of O. volvulus, there is an urgent need for the development of new drugs, notably drugs that kill or sterilize adult worms.
Our structural analysis of the GST2 from O. volvulus shows that it is a suitable drug target for intervention in filarial infections, with the potential advantage of not compromising the human host. Knowing the three-dimensional shape and chemical nature of the Ov-GST2 binding pocket will aid in the discovery and design of novel anti-filarial compounds. We suggest that neutralization of the Ov-GST2 may tip the molecular balance in favor of the host during this characteristically chronic nematode infection.