Molecular Mechanism of Aminoglycoside Antibiotic Kinase APH(3′)-IIIa

The aminoglycoside antibiotic kinases (APHs) constitute a clinically important group of antibiotic resistance enzymes. APHs share structural and functional homology with Ser/Thr and Tyr kinases, yet only five amino acids are invariant between the two groups of enzymes and these residues are all located within the nucleotide binding regions of the proteins. We have performed site-directed mutagenesis on all five conserved residues in the aminoglycoside kinase APH(3′)-IIIa: Lys44 and Glu60 involved in ATP capture, a putative active site base required for deprotonating the incoming aminoglycoside hydroxyl group Asp190, and the Mg2+ ligands Asn195and Glu208, which coordinate two Mg2+ ions, Mg1 and Mg2. Previous structural and mutagenesis evidence have demonstrated that Lys44 interacts directly with the phosphate groups of ATP; mutagenesis of invariant Glu60, which forms a salt bridge with the ε-amino group of Lys44, demonstrated that this residue does not play a critical role in ATP recognition or catalysis. Results of mutagenesis of Asp190 were consistent with a role in proper positioning of the aminoglycoside hydroxyl during phosphoryl transfer but not as a general base. The Mg1 and Mg2 ligand Asp208 was found to be absolutely required for enzyme activity and the Mg2 ligand Asn195 is important for Mg·ATP recognition. The mutagenesis results together with solvent isotope, solvent viscosity, and divalent cation requirements are consistent with a dissociative mechanism of phosphoryl transfer where initial substrate deprotonation is not essential for phosphate transfer and where Mg2 and Asp208 likely play a critical role in stabilization of a metaphosphate-like transition state. These results lay the foundation for the synthesis of transition state mimics that could reverse aminoglycoside antibiotic resistance in vivo.

Aminoglycoside antibiotics (see Fig. 1 below) are widely used in the treatment of bacterial infections caused by both Grampositive and Gram-negative pathogens yet the specific details of the molecular mechanisms of resistance to this class of antibiotics remain largely unknown. Aminoglycosides act first by disrupting protein expression and interfering with translation at the level of appropriate aminoacyl-tRNA recognition at the ribosomal A site; this is followed by a series of secondary effects, including membrane damage, which combine to arrest growth and kill the cell (1). The ribosomal receptor for most aminoglycoside antibiotics, including gentamicin C, kanamycin, and neomycin, is the 16 S rRNA of the 30 S subunit. The location of aminoglycoside binding site has been inferred by chemical footprinting analyses (2,3) and through the elucidation of the structure of aminoglycosides in complex with oligomers that model the 30 S rRNA binding site by NMR methods (4,5) and more recently by direct crystallographic determination of the complexes of the small ribosomal subunit with aminoglycosides (6). These studies suggest a tight complex between rings I and II of the aminoglycosides with the rRNA and highlight important interactions between conserved aminoglycoside amino and hydroxyl functional groups and the nucleic acids. Given the close fit between drug and target, it is not surprising that bacteria have evolved resistance mechanisms that decorate the aminoglycosides by covalent modification resulting in disruption of the complexes with a consequent decrease in ribosome affinity (7,8).
Aminoglycoside-modifying enzymes are widely distributed among bacterial pathogens and include O-phosphoryltransferases (kinases), N-acetyltransferases, and O-adenyltransferases. The aminoglycoside kinases (abbreviated APH) 1 include a number of enzymes with differing aminoglycoside substrate specificities and regiospecificities of phosphoryl transfer. Despite this broad tolerance for aminoglycoside substrates, all APHs share a common active site motif that resembles the active site residues found in the Ser/Thr/Tyr protein kinase superfamily (9). This similarity mirrors the threedimensional structural homology between aminoglycoside and protein kinases (10). In particular, five amino acid residues are 100% conserved between aminoglycoside and protein kinases: Lys 44 , Glu 60 , Asp 190 , Asn 195 , and Asp 208 (APH(3Ј)-IIIa numbering; see Fig. 2

below).
Site-directed mutagenesis and affinity labeling studies on the broad spectrum aminoglycoside kinase APH(3Ј)-IIIa found in Gram-positive pathogens such as staphylococci, streptococci, and enterococci, have supported the importance of several of the conserved residues. For example, Lys 44 was implicated in ATP binding in APH(3Ј)-IIIa and is homologous to an invariant Lys in protein kinases with a similar role (10,11). Asp 190 aligns with an Asp residue widely described in the protein kinase literature as an active site base necessary for deprotonating the incoming nucleophilic hydroxyl, and indeed mutation of Asp 190 to Ala in APH(3Ј)-IIIa has a major impact on catalysis (10). These studies along with the demonstration that APHs can phosphorylate some protein and peptide substrates (12) support the functional link between aminoglycoside and protein kinases. APH(3Ј)-IIIa has also been determined to regiospecifically phosphorylate 4,6-disubstituted deoxystreptamine aminoglycosides such as kanamycin and amikacin at the 3Ј-hydroxyl exclusively and at the 3Ј-and 5Љ-hydroxyls of 4,5-disubstituted deoxystreptamine aminoglycosides such as neomycin and ribostamycin ( Fig. 1) (13) with ADP release governing the maximal rate in the steady state (14).
The determination of the structures of APH(3Ј)-IIIa in complex with various substrates and inhibitors together with the revelation of the structural similarity with Ser/Thr/Tyr protein kinases has immensely aided our understanding of the mechanism of this enzyme. Guided by studies on the bettercharacterized protein kinases, we have examined the roles of conserved residues lining the active site of this enzyme and characterized their contribution to catalysis. Site-directed Mutagenesis-The Glu 60 3 Ala, Asp 190 3 Asn, and Asp 190 3 Glu site mutants were prepared by the "Megaprimer" methodology (15) using primers described in Table I. Briefly, the primers pGlu60Ala, pAsp190Ala, and pAsp190Glu were appropriately combined with primers p3Ј-PCR or p5Ј-PCR complimentary to ends of aph(3Ј)-IIIa to generate the megaprimer by PCR using the plasmid pETSACG1 as template, and this primer was used to amplify the complete gene containing the desired mutation. The remaining mutants, Asn 195 3 Ala, Asp 208 3 Asn, and Asp 208 3 Glu were prepared by the QuikChange mutagenesis protocol (Stratagene, La Jolla, CA) using primers pAsn195Ala, pAsp208Asn, and pAsp208Glu (Table I) and their reverse complementary oligonucleotides. All the mutant genes were identified by DNA sequencing, and subsequently sequenced in their entirety at the Central Facility of the Institute for Molecular Biology and Biotechnology, McMaster University, to ensure that undesired mutations had not been incorporated during PCR reactions.

Chemicals
Purification of Mutant APH(3Ј)-IIIa Proteins-APH(3Ј)-IIIa was overexpressed from Escherichia coli BL21(DE3)/pETSACG1 and purified as previously described (16,17). The purification of mutant APH(3Ј)-IIIa proteins followed essentially the same procedure with minor modification in some cases such as the addition of a gel filtration step using Superdex 200 (HR 10/30) (Amersham Pharmacia Biotech), to generate pure proteins. Purity of APH(3Ј)-IIIa was assessed by monitoring the purification using SDS-polyacrylamide gel electrophoresis and the concentration of protein determined by Bradford assay (18).
Protease Susceptibility of APH(3Ј)-IIIa-A solution consisting of purified APH(3Ј)-IIIa (12 g) in 50 mM HEPES, pH 7.5, 40 mM KCl, 10 mM MgCl 2 in a final volume of 30 l was incubated with 0.06 g of subtilisin on ice for up to 60 min. The reaction was quenched by the addition of 1 mM phenylmethylsulfonyl fluoride, and the products were separated by SDS-polyacrylamide gel electrophoresis. Matrix-assisted laser desorption time-of-flight mass spectrometry (MALDI-TOF) on the digestion products was performed at the Harvard Microchemistry Center (Boston, MA).
Kinetic Assay-The kinetic assay used to monitor APH(3Ј) activity has previously been described (16). The assay measures the production of ADP generated upon aminoglycoside phosphorylation and couples that production to the oxidation of ␤-NADH using the enzymes pyruvate kinase and lactate dehydrogenase. The rate of ADP production was determined by monitoring the decrease in absorbance at 340 nm. Initial rates were fit by nonlinear least squares methods to Eq. 1, if substrate inhibition was detected, using Grafit version 3.0 (19). ATP was fixed at 1 mM, and kanamycin A was fixed at 100 M, when measuring the steady-state kinetic parameters for aminoglycosides substrates and ATP, respectively.
Metal Ion Dependence-The metal ion (M 2ϩ ) dependence of both initial rates and the steady-state kinetic parameters of APH(3Ј)-IIIIa were examined. To measure the effect on initial rates, the concentrations of ATP (1 mM) and kanamycin (100 M) were fixed, and the concentration of MgCl 2 varied between 0.1 and 10 mM. To ensure that the coupled assay system was active at such low concentrations of Mg 2ϩ , ADP was added and the rate of reaction was determined. Similarly, for studies using Mn 2ϩ as the divalent metal, conditions were established where the rate of ADP turnover by pyruvate kinase was not rate-limiting. At all concentrations of M 2ϩ tested, the rate of ADP turnover was faster than the rate of aminoglycoside phosphorylation. The initial rates were fit to either Eq. 1 or Eq. 2, depending on whether or not Mg 2ϩ inhibition was noted. The best fits of the data were obtained when the initial rates were plotted against the concentration of free Mg 2ϩ and not when plotted against the concentration of total Mg 2ϩ . The concentration of free Mg 2ϩ was calculated using the parameters described previously (20).
Measurement of Solvent Isotope and Solvent Viscosity Effects for APH(3Ј)-IIIa-The solvent isotope and solvent viscosity effects on the Asn 195 3 Ala and the Glu 60 3 Ala were determined by the methods previously described for WT APH(3Ј)-IIIa (14). For solvent isotope studies the H 2 O content was Յ3.5% (v/v), and for solvent viscosity studies glycerol was used as the microviscogen unless otherwise noted. Briefly, glycerol was added to the coupled enzyme assay buffer to concentrations from 0 to 30% (w/v), and the steady-state kinetic parameters for ATP were determined. The concentration of kanamycin A was fixed at 125 M. Assays were carried out in duplicate at 37°C. The relative viscosity of the glycerol-containing assay buffer was determined in quadruplicate using an Ostwald viscometer, as described in Ref. (14). Macroviscogen controls were performed in all cases using PEG 8000.

Determination of the Minimal Inhibitory Concentration (MIC) of Kanamycin A-The minimal inhibitory concentration (MIC) is defined as the lowest concentration of kanamycin A required to completely inhibit
growth of E. coli BL21(DE3)/pETSACG1. To determine MICs, cultures were streaked out over LB agar with ampicillin (50 g/ml) as the selection marker and grown overnight at 37°C. From this, single colonies were selected and restreaked on LB agar with ampicillin and grown at 37°C for 18 -24 h. A few colonies were picked and suspended in sterile 0.85% (w/v) NaCl to achieve an OD 625 of 0.08 -0.1. These suspensions were then diluted to 1:10, and 5 l of the diluted suspensions were added to 95 l of Mueller-Hinton broth in sterile microtiter plate wells containing serial dilutions of kanamycin A (0 -512 g/ml). The final dilution corresponds to ϳ5 ϫ 10 5 colony forming units/ml. A sterile lid was placed over the microtiter plate, and it was allowed to incubate at 37°C for 16 -20 h before being visually inspected for bacterial growth.

Protease Susceptibility of APH(3Ј)-IIIa
Because protease sensitivity can be a sensitive marker of changes in protein structure, we examined the susceptibility of APH(3Ј)-IIIa to various proteases in the presence and absence of substrates in an effort to ensure that changes in the activity of the site mutants were not the result of significant changes in protein conformation. Exposure of the WT enzyme to subtilisin results in the cleavage of the enzyme at positions His 78 and Glu 157 , which are in exposed loop regions (Fig. 3B), as determined by MALDI-TOF mass spectrometry, and addition of tobramycin completely protected against proteolysis whereas addition of ATP alone gave only modest protection (Fig. 3A). All of the site mutants described in this work showed protease sensitivity that was indistinguishable from the effects on the WT enzyme.

Analysis of APH(3Ј)-IIIa Mutants
Glu 60 -Glu 60 is completely conserved in both protein and aminoglycoside kinases, forming a salt bridge with Lys 44 (Fig.  2). The Glu 60 3 Ala mutant nonetheless did not show significant changes in either k cat or k cat/ K m with respect to the WT enzyme (Table II). The only significant effect was a 5-fold increase in neomycin B k cat /K m for the 4,6-disubstituted deoxystreptamine-containing neomycin B. In vivo kanamycin resistance data parallel the little change in in vitro effects as assessed by MIC studies in liquid culture (Table III).
Asn 195 -Asn 195 is also absolutely conserved among the APHs, and the protein kinases and is involved in coordinating Mg2 (see Fig. 2 for Mg nomenclature). Mutation of this residue to Ala resulted in no changes in k cat but had a greater than 5-fold increase in k cat/ K m for ATP (Table II). Furthermore, although k cat /K m for aminoglycosides was largely unaffected, greater than 5-fold increases were recorded for amikacin, isepamicin, and butirosin, aminoglycosides substituted at N1. The kanamycin MIC was also reduced to 64 g/ml, correlating the important role in enzyme activity with the biologically relevant phenotype of drug resistance (Table III).
Asp 190 -Asp 190 is equivalent to the invariant protein kinase Asp in Hanks' consensus sequence VI that has been ascribed the role of active site base. The Asp 190 3 Ala mutant of APH(3Ј)-IIIa has previously been shown to be over 500-fold less active than the WT enzyme (10). To further explore the role of Asp 190 in the catalytic mechanism of APH(3Ј)-IIIa, Asp 190 3 Glu and Asp 190 3 Asn mutants were generated. Both mutants were severely impaired in phosphoryl transfer capacity and in most cases only estimates of k cat were possible (Table II). The importance of Ala 190 was also supported by a greater than 32-fold decrease in MIC (Table III). Asp 208 -The crystal structure of APH(3Ј)-IIIa⅐ADP demonstrates that Asp 208 is a ligand for both Mg1, providing two of the six possible Mg1 coordination positions, and Mg2, providing one coordination ligand (10), and we have previously shown that Asp 208 is critical to the protein kinase activity of APH(3Ј)-IIIa (12); an Asp 208 3 Ala mutant is at least 1800-fold less active than the WT enzyme. We therefore generated the Asp 208 3 Asn and the Asp 208 3 Glu mutants to further probe the role of this residue in enzyme activity. Both mutants lacked any detectable activity above the assay background, underscoring the critical nature of this residue to catalysis. Assessment of the biological activity of the Asp 208 mutants was also supportive of an essential role in catalysis with kanamycin A MIC levels indistinguishable from controls containing no APH(3Ј)-IIIa (Table III).

Metal Ion Dependence of WT and Mutant APH(3Ј)-IIIa Proteins
Mg 2ϩ is a required element for APH(3Ј)-IIIa rate enhancement, and other metal ions, for example Mn 2ϩ , do not support catalysis at a concentration of 10 mM (16). The fact that a number of invariant active site residues are also metal ligands suggested that the active site mutants generated in this study might have altered metal ion dependences. As a first test of this possibility, we examined the ability of the Asn 195 3 Ala mutant to use Mn 2ϩ as the divalent cation (Table IV). Unlike the WT enzyme, the Asn 195 3 Ala mutant was quite active in 15 mM MnCl 2 (Fig. 4).
The choice of divalent cation also had a significant impact on the rate of reaction for the Asn 195 3 Ala mutant. The k cat measured for the Mn 2ϩ -catalyzed phosphorylation of kanamycin A was decreased 4.6-fold (Table IV). There is a corresponding decrease in the k cat determined when ATP is the variable substrate. A comparison of the rates of reaction determined at 15 mM MnCl 2 to those determined at 10 mM MgCl 2 is reasonable, because the rate of the Mn 2ϩ catalyzed reaction varies little between 5 and 15 mM MnCl 2 .  We next examined the dependence of initial rates of the WT and Asn 195 3 Ala enzymes at fixed concentrations of ATP and kanamycin, on both Mg 2ϩ and Mn 2ϩ ion concentration (Fig. 4). The results obtained for the WT enzyme clearly indicate that increasing concentrations of both these metal ions are in fact inhibitory, with the effect being more pronounced when Mn 2ϩ is used as the divalent cation. On the other hand, the lack of inhibition at higher concentrations of divalent cation is striking compared with the Asn 195 3 Ala mutant.
The observation that the Asn 195 3 Ala mutant is not significantly inhibited by high concentrations of Mg 2ϩ prompted an investigation of the Mg 2ϩ dependences of the other active site mutants examined in this study. The results from this analysis are presented graphically in Fig. 5 and quantitatively in Table  V.
What is most evident from Fig. 5 is the extent to which the WT enzyme is inhibited by high concentrations of Mg 2ϩ (K i ϭ 0.57 Ϯ 0.23 mM), in comparison to the active site mutants described here. The Mg 2ϩ dependence of the Phe 264 3 Ala mutant (17) was included as a negative control to show that a mutation far from the catalytic center of APH(3Ј)-IIIa does not have as great an impact on the Mg 2ϩ dependence as do those residues that make direct or indirect contact with either Mg1 or Mg2.

Solvent Isotope and Solvent Viscosity Effects
The dramatic change in the Mn 2ϩ dependence of the Asn 195 3 Ala mutant, suggested that the rate-limiting step for this mutant was no longer exclusively of ADP release as it is with the WT enzyme (14). Therefore, to examine whether or not proton abstraction was now contributing to the rate, we measured solvent isotope effects (SIEs) for the Asn 195 3 Ala mutant by performing kinetic experiments in D 2 O, in the presence of either 10 mM MgCl 2 or 15 mM MnCl 2 (Table VI). The SIEs on both k cat and k cat /K m were decreased compared with WT, and there was a significant inverse effect on k cat /K m (Table VI).
In a further effort to compare the rate-limiting step of the Asn 195 3 Ala mutant to that of the WT enzyme, we examined the solvent viscosity effect (SVE 2 ), using the microviscogen glycerol. Macroviscogen controls using PEG 8000 were also performed and showed no effect as previously reported for the WT enzyme (14). The SVE of APH(3Ј)-IIIa, on both k cat and k cat /K m , is ϳ1 (14), whereas the SVE on k cat(ATP) of the Asn 195 3 Ala mutant, is 0.20 Ϯ 0.06.
The SIE and the SVE for the Glu 60 3 Ala mutant were also determined. The SIE is identical to the effect on the WT enzyme (Table VI). On the other hand, the SVE of this mutant on k cat(ATP) is quite large (2.89 Ϯ 0.12), suggesting that a number of diffusion controlled steps affect the measured rate of the reaction catalyzed by the Glu 60 3 Ala mutant. The possibility that this result was due to direct inhibition by glycerol was explored by performing the assay using sucrose as microviscogen and here again the SVE was greater than 1 (1.4 Ϯ 0.2).

DISCUSSION
Elucidation of the mechanism of phosphoryl transfer catalyzed by aminoglycoside kinases and the contribution of specific active site residues is central to efforts to understand antibiotic resistance at the molecular level. The remarkable three-dimensional similarity between Ser/Thr/Tyr protein kinases and APH(3Ј)-IIIa suggested the possibility that these enzymes may catalyze phosphoryl transfer in a conserved fashion, and, indeed, of the five invariant amino acids between protein and aminoglycoside kinases, all are located within the active site. In an effort to understand the contribution of APH(3Ј)-IIIa active site residues to rate enhancement, we have analyzed the effect of mutagenesis of Glu 60 , Asp 190 , Asn 195 , and Asp 208 , which represent four of the active site amino acids conserved between protein and aminoglycoside kinases. The effects of mutation of the invariant Lys and its role in ATP binding have been reported (10).
Phosphoryl transfer can occur through two extreme mechanisms in which nucleophilic attack either precedes (associative mechanism) or occurs after (dissociative mechanism) leaving group bond breakage (Scheme 1). The very practical importance of the determination of the mechanism of phosphoryl transfer in aminoglycoside kinase enzymes lies in developing a molecular understanding of transition state structure, which has utility in the design of specific inhibitors. This requires distinguishing between a minimal reaction coordinate distance with overall charge of Ϫ3 for a fully associative mechanism and larger reaction coordinate distance with overall charge of Ϫ1 for a full metaphosphate-generating dissociative mechanism. However, as has been previously noted (21), a continuum of mechanisms is possible and most enzymes likely do not adopt either fully associative or dissociative mechanisms but stabilize transition states in which bond breakage or formation predominate but are not complete. One of the goals of these mutagenesis studies was to provide supportive evidence to describe the phosphoryl transfer mechanism of APH(3Ј)-IIIa.
Glu 60 -In a ternary structure model of APH(3Ј) in complex with ATP and the aminoglycoside ribostamycin (17), the carboxylate of Glu 60 is positioned at least 10 Å away from the phosphate-accepting hydroxyl group of the aminoglycoside. Glu 60 is however interacting with the ATP ligand Lys 44 in an arrangement that is conserved in protein kinases (Fig. 2). Therefore, it was predicted that mutation of this residue to Ala was unlikely to exert a direct effect upon aminoglycoside affinity, but that ATP binding could be affected. However, this mutant was not dramatically impaired in substrate affinity, catalysis, or MIC. The SVE on k cat was significantly larger than that observed for WT enzyme (ϳ1 (14)), which can be interpreted as an impact of the viscogens glycerol and sucrose on changes in protein conformation since these are not inhibitors of APH(3Ј)-IIIa. Thus the role of Glu 60 is complex, not limited to positioning of Lys 44 for optimal ATP binding and/or ADP release. The suggestion that changes in protein conformation are important to the role of Glu 60 finds a protein kinase parallel with the x-ray structure of the Tyr kinase Hck where the region of the protein surrounding the invariant Glu undergoes substantial reorganization upon activation of the tyrosine kinase activity of this enzyme (22).
Asn 195 -The results obtained for the Asn 195 3 Ala mutant are consistent with a role for this residue in ATP binding, as K m(ATP) is increased over 5-fold. Asn 195 interacts with ATP through Mg2, which suggests that the decrease in affinity reflects a non-optimally coordinated metal ion. Productive capture (k cat /K m (23)) for most aminoglycoside substrates was unaffected with the exception of amikacin and isepamicin, both 4,6-disubstituted aminocyclitol aminoglycosides that are alkylated at N1. The molecular basis for this difference is not readily evident from either the available crystal structure data or modeling studies of ternary complexes, and further analysis awaits additional structural determination but may reflect the predicted altered binding of these N1-substituted compounds (17).
The values of k cat measured for both ATP and a number of aminoglycoside substrates did not vary significantly for the Asn 195 3 Ala mutant. An SVE of 0.20 for this enzyme is consistent with the suggestion that in this mutant k cat is no longer dominated by diffusion-controlled ADP release (14), indicating that k cat is underestimated with respect to WT APH(3Ј)-IIIa and is dominated by a viscosity-independent step, likely phosphoryl transfer. There was a slight reduction of the SIE on k cat rather than an increase compared with WT (Table  VI), suggesting that proton transfer (i.e. base catalysis) does not contribute to the maximal rate, consistent with a dissociative mechanism of phosphoryl transfer and inconsistent with the traditionally ascribed role of the conserved catalytic Asp in aminoglycoside and protein kinases (Asp 190 ) as a general base. At the same time the SIE on k cat /K m was dramatic and inverse for both ATP and kanamycin. An inverse effect on productive substrate capture has been noted for other enzymes, including alcohol dehydrogenase (24) and ␤-lactamase (25) (see Ref. 24 for more examples). Such inverse SIEs have been interpreted as being the result of the increased viscosity of D 2 O solutions (26), the dissociation of an exchangeable hydrogen from a thiol group, which has a fractionation factor of less than 1 (27), medium effects where SIEs result from changes in solvation of solutes in D 2 O versus H 2 O (28), the dissociation of a metalchelated water molecule (27), or the result of restrictions on the torsional movement of the affected bond (29). Although the Asn 195 3 Ala mutant did show an inverse SVE, the levels at the concentration of glucose matching the viscosity of D 2 O (ϳ9%) were not sufficient to match the observed inverse SIE. There are no active site thiols in APH(3Ј)-IIIa, eliminating the possibility of contribution of exchange of a Cys SH group. Medium effects are expected to be global and common to all the studies performed here on WT and mutant enzymes and are not likely to be the source of the inverse effect in the Asn 195 3 Ala mutant alone. Asn 195 is an Mg2 ligand, and therefore, loss of the amide-Mg bond will result in altered metal binding at the active site and constellation of requisite ligands. Therefore, the inverse SIE could in fact reflect a dissociation of a metal-OD Ϫ interaction in this mutant. Similarly, is it not possible to rule out that the inverse effect is due to restricted torsional effects upon productive substrate capture. The large SIE in the presence of Mn 2ϩ determined for the Asn 195 3 Ala mutant only in k cat /K m for ATP again likely reflects the alteration in optimal metal ligand availability and indicated that, for the larger manganese, there is a significant barrier to productive ATP capture in D 2 O, quite the opposite of the situation with magnesium where this rate is enhanced in D 2 O. The molecular Asp 190 -We have previously described the importance of Asp 190 to the catalytic mechanism of APH(3Ј)-IIIa and the Asp 190 3 Ala mutant is ϳ550-fold less active than the WT enzyme (10). The corresponding mutation in yeast protein kinase A (PKA) results in a similar 340-fold decrease in k cat (30). Although in both cases the effect on the chemical step is likely to be larger, given that both APH(3Ј)-IIIa and PKA k cat are limited by the diffusion-controlled release of ADP, and that the rate of the PKA-catalyzed chemical step is ϳ24-fold greater than product release (31). These results are strengthened by structural evidence placing the carboxylate group of this residue close to the predicted path of phosphate transfer. Together, these results have contributed to the suggestion that this invariant residue acts as an active site base, deprotonating the incoming hydroxyl group and thus increasing its nucleophilicity. The suggestion that this Asp residue is an active site base is however challenged by the magnitude of rate reduction upon mutation to Ala, which is not generally consistent with a critical role in substrate deprotonation. Supporting evidence has also arisen in the protein kinase field where alternate substrate studies using the protein Tyr kinase CsK revealed that fluorinated Tyr analogues, which are deprotonated at assay pH, are not improved substrates for the enzyme over protonated analogues, and Brnstead values are not consistent with an associative mechanism where base catalysis would be predicted to play a central role (32,33).
To further characterize the importance of Asp 190 in aminoglycoside antibiotic phosphorylation, we generated the Asp 190 3 Glu and Asp 190 3 Asn mutants, where Glu conserves charge but not position, whereas the Asn mutant is isosteric but without appreciable charge. The Asp 190 3 Glu mutant was markedly impaired in its ability to phosphorylate a variety of aminoglycoside substrates with a resulting 1100-fold decrease in k cat /K m . Comparison of the catalytic rates of the Asp 190 3 Ala to Asp 190 3 Glu mutants indicates that positioning of the carboxylate contributes at least two orders of magnitude to rate enhancement. A similar amount of catalytic power has been attributed to the rate enhancement afforded by the correct positioning of His 122 in the catalytic mechanism of nucleoside diphosphate kinase (34).
The Asp 190 3 Asn mutant was significantly impaired in catalytic capacity but maintained demonstrable phosphoryl transfer activity above the Asp 190 3 Ala mutant. The pK a of the Asn amide is less than 0 and is therefore unlikely to act as a general base. These results are therefore not consistent with a role for Asp 190 as a general base. Rather, the carboxylate may participate in orienting the incoming aminoglycoside hydroxyl for optimal attack upon the ␥-phosphate of ATP, but that phosphate transfer likely precedes substrate deprotonation, which is supportive of a dissociative phosphoryl transfer mechanism.
Asp 208 -Asp 208 is both an Mg1 and Mg2 ligand (Fig. 2), and we have previously reported the importance of Asp 208 to the protein kinase activity of APH(3Ј)-IIIa (12). The Asp 208 3 Ala mutant lacks any detectable aminoglycoside kinase activity, thereby indicating that this residue is absolutely critical for the catalytic mechanism of APH(3Ј)-IIIa. This result parallels those obtained when the corresponding residue of yeast PKA was mutated to Ala resulting in inviable cells (30). Mutation of APH(3Ј)-IIIa Asp 208 to Glu or Asn generated mutants without detectable kinase activity. Because the isosteric Asn is also a potential Mg 2ϩ ligand, these results suggest that charge neutralization of Mg is important for generating and or stabilizing the transition state, although Mg 2ϩ ligation by Asn will admittedly likely be different than Asp. Thus Asp 208 and Mg 2ϩ (likely Mg1), and not Asp 190 , are critical factors required for transition state formation. Given that a role for Asp 190 as a catalytic base is not required for APH(3Ј)-IIIa catalysis argues against an associative transition state and therefore favors a dissociative mechanism with a metaphosphate-like intermediate.
Roles of M 2ϩ -In an effort to better understand the role that metal ions play in the catalytic mechanism of APH(3Ј)-IIIa, we examined the concentration dependence of free Mg 2ϩ on the initial rates of reaction, and as well on the steady-state kinetic parameters. APH(3Ј)-IIIa-catalyzed aminoglycoside phosphorylation was both activated and inhibited by increased Mg 2ϩ concentration, demonstrating potent substrate inhibition with an optimal concentration of ϳ1 mM free Mg 2ϩ , which resulted in a roughly 2-fold decrease in k cat from 1-10 mM. The Glu 60 3 Ala mutant was less affected by Mg 2ϩ concentration, but required 3-fold more metal ion for maximal activity. Although the altered metal ion dependence of this mutant enzyme may reflect a change in the rate-limiting step, it is more likely that the loss of the indirect interactions that the Glu 60 carboxylate maintains with Mg1 is responsible for these changes. Such an interaction occurs between the carboxylate of Glu 60 and the ⑀-amino group of Lys 44 (10). Lys 44 interacts with the nonbridging oxygens of the ␣and ␤-phosphates, and is important for nucleotide binding (10). Thus, this interaction would clearly be important for holding ADP in place after phosphoryl transfer has occurred. Therefore, the loss of binding energy albeit indirect, would be expected to affect the inhibition caused by increasing concentrations of magnesium. If this is the case, then it indicates that increases in solvent viscosity would inhibit a conformational change that is required for phosphate transfer. Because the presence of Glu 60 normally prevents this dramatic solvent viscosity effect, the pocket, created by the removal of this functional group, could become filled with the viscogen, which would prevent the diffusion-controlled movement of residues lining this region of the active site.
Mg 2ϩ inhibition is virtually non-existent in the Asn 195 3 Ala mutant. This result, in combination with the effects noted on the affinity of ATP for this mutant enzyme, indicates that Mg2 is required for the optimal binding of ATP. However, neither Mg2 nor Asn 195 is essential for catalysis. This is remarkable SCHEME 1.

Potential phosphate transfer mechanisms catalyzed by APH(3)-IIIa.
when one considers that the loss of the Asp 208 carboxylate, and in effect the coordination site(s) on Mg1, results in the complete loss of any detectable activity. These results taken together suggest a role for Mg2 in facilitating nucleotide triphosphate binding and that Asn 195 plays a critical role in making ADP release the rate-limiting step for APH(3Ј)-IIIa.
The observation that the Asp 190 3 Glu mutant has an altered Mg 2ϩ dependence suggests that this residue may also play a role in Mg 2ϩ binding. In the APH(3Ј)-IIIa⅐ADP crystal structure the carboxylate of Asp 190 is 3.9 Å from Mg1 at its closest approach. Although this distance is long in the available ground state structures of APH(3Ј)-IIIa, it is possible that Mg1 could move closer to this functional group in the transition state. Thus the change in the apparent dissociation constant for the metal ion with this mutant enzyme may reflect the presence of the carboxylate of this mutant in a suboptimal orientation.
Conclusions-The kinetic analyses of the mutant APH(3Ј)-IIIa enzymes described herein help explain the contribution of each particular amino acid residue to the enhancement of phosphoryl transfer. We conclude that substrate deprotonation is not important for the catalytic activity of APH(3Ј)-IIIa and that Asp 190 serves to orient the aminoglycoside hydroxyl for productive phosphate transfer. Furthermore, Asp 190 may also play a role in metal binding and this property is likely important for transition state formation. We also conclude that the position and charge of Asp 208 are critical to the catalytic mechanism of this enzyme. These results, combined with solvent isotope and solvent viscosity effects, favor a dissociative-like transition state. These results also attempt to explain the roles of the two magnesium ions that are required for catalysis to occur. Mg2 is important for nucleotide binding, whereas Mg1 is the catalytic magnesium ion that directs phosphoryl transfer and could thereby aid the formation of a meta-phosphate-like intermediate.
The results also describe the importance of a number of residues that are absolutely conserved among APHs and protein kinases. The results described herein confirm that these residues are not only conserved at the primary and tertiary structure levels but are also functionally conserved. This finding has profound implications for the evolution of enzymes when one considers that sequence conservation among the APH and protein kinase families of enzymes is less than 5%, and that only five absolutely conserved residues are shared between these two superfamilies of enzymes. Interestingly, these residues all make up the catalytic core of the enzyme and are involved in nucleotide binding, metal binding, or the promotion of phosphate transfer. Thus it appears likely that these two families of enzymes evolved from a common progenitor; i.e. as these enzymes evolved, the architecture of the active site was maintained, while residues responsible for substrate binding were replaced to fulfill the unique substrate binding requirements of each individual enzyme.