Crystal Structures of Aedes aegypti Alanine Glyoxylate Aminotransferase*

Mosquitoes are unique in having evolved two alanine glyoxylate aminotransferases (AGTs). One is 3-hydroxykynurenine transaminase (HKT), which is primarily responsible for catalyzing the transamination of 3-hydroxykynurenine (3-HK) to xanthurenic acid (XA). Interestingly, XA is used by malaria parasites as a chemical trigger for their development within the mosquito. This 3-HK to XA conversion is considered the major mechanism mosquitoes use to detoxify the chemically reactive and potentially toxic 3-HK. The other AGT is a typical dipteran insect AGT and is specific for converting glyoxylic acid to glycine. Here we report the 1.75Å high-resolution three-dimensional crystal structure of AGT from the mosquito Aedes aegypti (AeAGT) and structures of its complexes with reactants glyoxylic acid and alanine at 1.75 and 2.1Å resolution, respectively. This is the first time that the three-dimensional crystal structures of an AGT with its amino acceptor, glyoxylic acid, and amino donor, alanine, have been determined. The protein is dimeric and adopts the type I-fold of pyridoxal 5-phosphate (PLP)-dependent aminotransferases. The PLP co-factor is covalently bound to the active site in the crystal structure, and its binding site is similar to those of other AGTs. The comparison of the AeAGT-glyoxylic acid structure with other AGT structures revealed that these glyoxylic acid binding residues are conserved in most AGTs. Comparison of the AeAGT-alanine structure with that of the Anopheles HKT-inhibitor complex suggests that a Ser-Asn-Phe motif in the latter may be responsible for the substrate specificity of HKT enzymes for 3-HK.

thurenic acid (XA) is the major branch pathway of tryptophan metabolism (2,3). XA has been identified as a natural chemical signal produced by mosquitoes, which induces exflagellation of malaria parasites in the midgut (4,5). Recently, we reported that mosquitoes have evolved two alanine glyoxylate aminotransferases (6). One is the previously cloned Aedes aegypti 3-HK transaminase (AeHKT) (3), homologous to human AGT (hAGT), which primarily catalyzes the transamination of 3-HK to XA in mosquitoes. The other is a typical dipteran insect AGT, which primarily catalyzes the conversion of glyoxylic acid to glycine (6). A BLAST search of the genomes of other available model species (including cyanobacteria, archaea, yeast, plants, fruit fly, honeybee, frog, fish, rat, mouse, and human) revealed only one AGT in their genomes, which makes possessing two AGTs a unique property of mosquitoes. Based on both phylogenetic analysis and biochemical characterization, AeAGT is closely related to Drosophila AGT (DrAGT) and AeHKT is closely related to hAGT (6,7).
The different roles of these two aminotransferases in the mosquito may be better understood by examining their substrate specificities and temporal expression patterns in the context of the mosquito life cycle, discussed in more detail previously (6). Mosquito larvae eat protein-rich food, and a considerable portion of the tryptophan obtained from these food supplies is oxidized to 3-HK (2). In humans and other mammals, kynurenine aminotransferase I and II catalyze the transamination of kynurenine to kynurenic acid and 3-HK to XA (8 -10). Ae. aegypti kynurenine aminotransferase shows high activity toward kynurenine but no activity toward 3-HK (10 -12). In contrast, conversion of 3-HK to XA is catalyzed mainly by AeHKT in Ae. aegypti mosquitoes. Because mosquitoes cannot dispose of 3-HK through hydrolysis and subsequent oxidations like mammals, the transamination of the chemically reactive 3-HK to the relatively chemically stable XA by HKT is considered the mechanism by which mosquitoes prevent 3-HK from overaccumulating (7). Such an explanation remains a valid argument for the functional evolution of the mosquito AGTs. Interestingly, 3-HK is also the initial precursor for the production of ommochromes that are major eye pigments in mosquitoes. Compound eye development and eye pigmentation occur mainly during the pupal and early adult stages.
Coincidently, HKT activity is diminished in pupae and adults correlating with the accumulation of high levels of 3-HK in the compound eyes (13). The down-regulation of AeHKT is likely a tactic to allow 3-HK to be transported and used for eye pigmen tation. Although it seems clear that detoxification of 3-HK is a major function of AeHKT in mosquito larvae, the enzyme exhibits high AGT activity and likely plays a critical role in metabolizing glycolate-derived glyoxylate as well, especially in adults, because males eat the part of plants that is normally glycolaterich. Consequently, the interruption of AeHKT expression during the adult stages would affect the metabolism of glyoxylic acid. Therefore, a highly substrate-specific (i.e. highly active with regard to the glyoxylic acid to glycine pathway but without detectable HKT activity) and stage-specific AeAGT has evolved in mosquitoes.
Biochemical characterization shows that AeAGT is highly specific for catalyzing glyoxylic acid to glycine processing. AeHKT, on the other hand, primarily catalyzes the transamination of 3-HK to XA but also exhibits the same activities as AeAGT. To understand the structural basis underlying the stringent substrate specificity of AeAGT as compared with that of AeHKT, it is essential to determine the three-dimensional structure of AeAGT. In this

Structures of Mosquito Alanine Glyoxylate Aminotransferase
study, we determined the native three-dimensional crystal structure of AeAGT as well as the structures with its specific substrates, alanine and glyoxylic acid, respectively. This is the first time the crystal structures of the enzyme-substrate complexes for AGTs have been reported. We have also compared AeAGT complex structures with the available AGT structures and discussed the possible residues involved in substrate specificity. The results from this study, in conjunction with the results of previous studies dealing with the molecular and biochemical characterization of AeHKT (3,7), AeAGT (6), and the three-dimensional structural study of Anopheles HKT (AnHKT) (14), provide an interesting example of protein functional adaptation and evolution, in this case to meet certain physiological requirements during development.

MATERIALS AND METHODS
Overexpression and Purification of AeAGT Recombinant Protein-Fulllength AeAGT was overexpressed in a baculovirus/insect cell protein expression system (Invitrogen). The recombinant AeAGT was purified as described previously (6). The molecular weight and purity of the proteins were assessed by SDS-PAGE analysis. Protein concentrations were determined using a Bio-Rad protein assay kit using bovine serum albumin as a standard.
AeAGT Crystallization and Substrate Soaking-Initial crystallization screening was performed using Hampton Research Crystal Screens (Hampton Research, Laguna Niguel, CA) using the hanging drop vapor diffusion method. The volume of reservoir solution was 500 l, and the drop volume was 2 l, containing 1 l of protein sample and 1 l of reservoir solution. Quality AeAGT crystals were obtained in a solution containing 5 mg ml Ϫ1 protein, 12% polyethylene glycol 8000, 0.13 M magnesium acetate, 4% butanol, and 0.1 M cacodylic acid at pH 6.5. Single crystals suitable for x-ray analysis appeared within 1 to 2 weeks at 4°C. Alanine-enzyme and glyoxylic acid-enzyme complexes were obtained by soaking the crystals in 2.5 mM alanine and 1 mM glyoxylic acid, respectively, in the crystallization buffer for 1 week.
Data Collection and Processing-Individual AeAGT crystals were cryogenized using 25% glycerol in crystallization buffer as a cryoprotectant solution. Diffraction data from AeAGT crystals were collected at the Brookhaven National Synchrotron Light Source, Beamline X25 ( ϭ 0.9795 Å). Data collection was done using an ADSC Q315 charge-coupled device detector. All data were indexed and integrated using HKL software (15), and scaling and merging of diffraction data were performed using the program SCALEPACK (16). The parameters of the crystals and data collection are listed in Table 1.

Structures of Mosquito Alanine Glyoxylate Aminotransferase
Structure Determination-The structure of AeAGT was determined by the molecular replacement method. The initial search model was built with the homology module in Insight II (Accelrys) using hAGT (Protein Data Bank code 1h0c) as the template structure (17). The program Molrep (18) in the CCP4 suite was employed to calculate both the cross-rotation and translation of the homology model. The initial model was sub-jected to iterative cycles of crystallographic refinement with Refmac 5.2 (19) and graphic sessions for model building using the program O (20). Solvent molecules were automatically added and refined with ARP/wARP (21) together with Refmac 5.2. For the AeAGT structure, a careful inspection of the electron density in the enzyme active site was performed when the R-factor dropped to a value of 0.20 at 1.75 Å. This revealed an electron density compatible with one butanol molecule, which was subsequently manually modeled based on both the 2F o Ϫ F c and F o Ϫ F c electron density maps. The AeAGT-alanine and AeAGT-glyoxylic acid complex structures were refined by using the final coordinates of the AeAGT native structure as the starting model from which all solvent and butanol molecules were removed. The substrate molecules were modeled only when the R-factor dropped to a value of 0.2 at full resolution for the AeAGT-alanine structure and a value of 0.2 at 1.75 Å resolution for the AeAGT-glyoxylic acid structure.
Structure Analysis-Superposition of structures was done using Lsqkab (22) in the CCP4 suite. Protein and substrate interaction was analyzed using Ligplot (23). The figures were generated using Pymol (24) and Ligplot (23).

RESULTS AND DISCUSSION
Overall Structure-The structure of AeAGT was determined by molecular replacement and refined to 1.75 Å resolution. The final model contains 385 residues, yielding a crystallographic R-value of 17.3% and an R free value of 20.3% with ideal geometry (Table 1). There are two molecules in an asymmetric unit that form the homodimer. The residues of the two subunits in AeAGT are numbered 1A to 385A for chain A and 1B to 385B for chain B. The refinement results are summarized in Table 1.
All residues are in favorable regions of the Ramachandran plot as defined with PROCHECK (25). The last eight residues (386 -393) were not defined in the 2F o Ϫ F c electron density map and are absent from the final model. The protein architecture revealed by AeAGT consists of the prototypical fold of aminotransferase subgroup I (26,27), characterized by an N-terminal arm, which consists of a random coiled stretch made of residues 1-18, a small C-terminal domain (residues 280 -385), and a large N-terminal domain (residues 19 -279) (Fig. 1), which is similar to the reported AGT structures from other species (14, 17, 28, 29). As  Active Site of AeAGT-Residual electron density clearly revealed the presence of covalently bound pyridoxal 5-phosphate (PLP) in the cleft situated at the interface of the subunits in the biological dimer ( Figs. 1 and 2A). This cleft also corresponds to the cofactor binding sites in the related enzymes and therefore defines the active site of AeAGT. The C4A atom of PLP is covalently attached to the NZ of Lys 206A through the formation of an internal Schiff base, the internal aldimine giving rise to residue LLPA, represented as sticks in Fig. 2A. The PLP-pyridoxal ring is stacked between residues Trp 105A and Val 182A by hydrophobic interactions, and the C2A atom of PLP exhibits hydrophobic interaction with the Thr 151A side chain. The side chains of Ser 155A and Asp 180A are hydrogen-bonded to O-3 and N-1 of the pyridoxal, respectively. The phosphate moiety of the PLP is anchored by polar interactions with the peptide amide groups of residues Gly 79A and His 80A as well as by the side chains of Ser 78A , His 80A , and Gln 205A . The binding site of the PLP-phosphate is completed by hydrogen bonds with the side chains of Tyr 257B and Tyr 260B from the second subunit in the biological dimer (Fig. 2B). The N-terminal extension makes a turn around residues Gly 24A , Pro 25A , and Gly 26A exposing the Pro 25A carbonyl group toward the active site. This conformation allows for polar interaction with the NZ of Lys 206A and could also interact with the amino group from the pyridoxamine intermediate during the enzymatic reaction.
Substrate Recognition and Catalysis-Enzyme activity assays of AeAGT revealed that alanine acts efficiently as an amino group donor substrate during the transamination reaction, yielding pyruvate, and glyoxylic acid acts efficiently as an amino acceptor, yielding glycine (6). The enzyme-catalyzed reaction is shown in Scheme 1.
To identify the structural determinants of AeAGT involved in substrate recognition and binding, we determined the three-dimensional crystal structures of the enzyme in complex with alanine (Fig. 3) and glyoxylic acid (Fig. 4). Inspection of the AeAGT-alanine complex crystal structure reveals that the substrate lies near the internal Schiff base between the PLP cofactor and the NZ of Lys 206A . However, PLP is still in its internal aldimine form (Fig. 3). Several residues, structurally conserved throughout the aminotransferases described thus far, define the substrate binding site and contact the alanine molecule. In particular, Arg 356A forms a strong salt bridge with the alanine carboxylate group at a distance of 2.5 Å. A hydrogen-bonding network, involving Ser 155A and Gly 345A backbone oxygen atoms, locks the Arg 356A side chain in the optimal conformation for alanine binding. The anchoring of the alanine carboxylic moiety by Arg 356A in the AeAGT-alanine complex (Fig. 3) results in the reactive ␣amino group being placed into the ideal position (the nitrogen atom is above the PLP C4A reactive center at a distance of 3.7 Å) for attacking the PLP C4A atom and forming the external aldimine. The alanine is also recognized by a number of interactions with protein residues, including Tyr 257B , His 45B , Pro 25A , Ser 155A , and Leu 347A (Fig. 3). Gly 26A is located at the turning point of the loop that dives into and partially plugs the enzyme active site, thus shielding the substrate binding pocket from the bulk solvent. It is interesting that alanine was found only in one subunit of the dimer. Alanine is a primary amino acid substrate for most of AGTs, but its K m is much higher than glyoxylic acid,  (subunit B). The glyoxylic acid substrate (GLV) and the protein residues within 4 Å distance of glyoxylic acid are shown. The 2F o Ϫ F c electron density map covering the glyoxylic acid and LLPA is shown contoured at the 1.8 sigma level. This panel was generated using Pymol. B, schematic view of the glyoxylic acid binding site. The key interactions involving glyoxylic acid binding are indicated together with the distances for hydrogen bonds or salt bridges observed in the AeAGT-glyoxylic acid structure. This panel was generated using Ligplot. C, the omit map calculated without the glyoxylic acid substrate is shown as an F o Ϫ F c electron density map contoured at the 2 sigma level. Subunits A and B are colored in deep teal and violet, respectively. and therefore this might be a reason that the alanine concentration in the soaking solution was not high enough for the amino acid to get into the active site of the enzyme in all of the molecules of the crystal. A similar behavior was observed in the complex of Paracoccus denitrificans aromatic amino acid aminotransferase with 3-phenylpropionate (30).
Glyoxylic acid is the most physiologically relevant amino acceptor substrate for AGTs, including hAGT, AeAGT, AeHKT, and DrAGT (6,7,31). Detoxifying glyoxylic acid is the major function of AGTs in mammals, including human beings. It has been proposed that peroxisomal AGT is responsible for detoxifying glycolate-derived glyoxylic acid, whereas mitochondrial AGT plays a major role in detoxifying hydroxyproline-derived glyoxylic acid (32)(33)(34)(35)(36)(37). In plants, AGT has been known to be involved in the photorespiratory glyoxylic acid cycle within peroxisomes (38,39). In yeast, disruption of AGT can lead to glycine auxotrophy (40). The importance of hepatic AGT in minimizing endogenous oxalate production in some mammals has been clearly shown by the autosomal recessive disorder of glyoxylic acid metabolism primary hyperoxaluria type 1, a potentially lethal condition in which AGT deficiency leads to excessive oxalate synthesis and excretion and the deposition of insoluble calcium oxalate in the kidney (41,42). Because there have not been any reports regarding the three-dimensional crystal structure of an AGT-glyoxylic acid complex, the structure reported here provides the essential basis for understanding the generic interaction of glyoxylic acid with AGTs. Structurally, glyoxylic acid lies near the internal aldimine, and the C-2 atom of glyoxylic acid is above the PLP C4A reactive center at a distance of 3.0 Å. As in the alanine complex, the carboxylate group of glyoxylic acid forms a tight ion pair with Arg 356A at a distance of 2.9 Å, and Ser 155A and Gly 345A backbone oxygen  atoms fix the Arg 356A side chain in the optimal conformation for glyoxylic acid binding. One oxygen atom of the glycine carboxylate group also forms a hydrogen bond with the side chain of Ser 155A . Other residues around the glyoxylic acid molecule within 4 Å are Tyr 57B , His 45B , Arg 25A , Ser 155A , LLPA, and Thr 260B , and in particular Tyr 257B , Pro 25A , and LLPA, which establish hydrophobic interactions with glyoxylic acid. Finally, Ser 155A OG is bound to the O-2 atom of glyoxylic acid (Fig. 4B).
Comparison with Other AGT Structures with Regard to Substrate Specificity-In principle, differences in enzyme function should be reflected in both the primary and three-dimensional structures. Based on the available three-dimensional crystal structures of AGTs, hAGT (17), yeast AGT (28), Nostoc AGT (29), AnHKT (14), and the one reported herein, all residues involved in PLP binding are conserved within the five structures (not shown). AnHKT shares 73% amino acid sequence identity with AeHKT, and an AnHKT-inhibitor complex structure with an inhibitor that is structurally similar to 3-HK and kynurenine has been determined (14). Both AnHKT and AeHKT exhibit high activity toward 3-HK and kynurenine (3,7,43), whereas AeAGT displays no activity toward 3-HK and kynurenine (6). On superposing the AeAGT-alanine complex onto the AnHKT-inhibitor structure (Fig. 5A), we identified the residues within 4 Å of both ligands and found that most residues are conserved. An exception was the Ser 43 -Asn 44 -Phe 45 triad of AnHKT that is changed to Gly 44A -His 45A -Leu 46A in AeAGT (Fig. 5B). The amino acid sequence alignment shows that AeHKT and AnHKT share the same sequence in this area (boxed in Fig. 5C). The results suggest that the Ser-Asn-Phe motif is involved in 3-HK and kynurenine substrate recognition for mosquito HKTs. A future mutation experiment may confirm this hypothesis. Glyoxylic acid is the most suitable substrate for all AGTs, and therefore its binding site should be conserved in AGTs. A three-dimensional crystal structure of hAGT-amino-oxyacetic acid (AGT inhibitor) is available for comparison (17). On superposing hAGT structure onto the AeAGT-glyoxylic acid complex structure (Fig. 6A), we identified the residues within 4 Å of both ligands and found that all the residues (Ser, Pro, Arg, PLP, and Tyr (in the other unit)) were the same in both structures (Fig. 6B). The sequence comparison showed that Anopheles AGT, AeHKT, and AnHKT also shared the same residues for glyoxylic acid binding and that yeast AGT and Nostoc AGT shared similar residues, with Ser substituted by Thr.
In summary, evolution has led to the development of two AGTs, AGT and HKT, in the mosquito. This duplication occurred before the mosquito/fruit fly divergence. The primary function of the mosquito AGT and HKT enzymes is quite different. The three-dimensional structures of AeAGT and its complexes with alanine and glyoxylic acid obtained in this study, in comparison with the three-dimensional structure of the AnHKT, provide some reasonable explanation regarding the structural basis underlying the substrate specificity between mosquito AGT and HKT. As this is the first report of the structures of an AGT with its native substrates, it may also help in understanding the catalyzing mechanism of AGTs, particularly hAGT, which is related to a lethal condition in humans.