Small-Angle X-Ray Scattering Analysis of the Bifunctional Antibiotic Resistance Enzyme Aminoglycoside (6′) Acetyltransferase-Ie/Aminoglycoside (2″) Phosphotransferase-Ia Reveals a Rigid Solution Structure

Abstract

Aminoglycoside (6′) acetyltransferase-Ie/aminoglycoside (2″) phosphotransferase-Ia [AAC(6′)-Ie/APH(2″)-Ia] is one of the most problematic aminoglycoside resistance factors in clinical pathogens, conferring resistance to almost every aminoglycoside antibiotic available to modern medicine. Despite 3 decades of research, our understanding of the structure of this bifunctional enzyme remains limited. We used small-angle X-ray scattering (SAXS) to model the structure of this bifunctional enzyme in solution and to study the impact of substrate binding on the enzyme. It was observed that the enzyme adopts a rigid conformation in solution, where the N-terminal AAC domain is fixed to the C-terminal APH domain and not loosely tethered. The addition of acetyl-coenzyme A, coenzyme A, GDP, guanosine 5′-[β,γ-imido]triphosphate (GMPPNP), and combinations thereof to the protein resulted in only modest changes to the radius of gyration (R_G) of the enzyme, which were not consistent with any large changes in enzyme structure upon binding. These results imply some selective advantage to the bifunctional enzyme beyond coexpression as a single polypeptide, likely linked to an improvement in enzymatic properties. We propose that the rigid structure contributes to improved electrostatic steering of aminoglycoside substrates toward the two active sites, which may provide such an advantage.

INTRODUCTION

Aminoglycosides are broad-spectrum antibiotics that are frequently used in the treatment of serious bacterial infections. Unfortunately, resistance to these antibiotics is a continual challenge for their administration. Most frequently, resistance to aminoglycosides is conferred by aminoglycoside-modifying enzymes (32, 34, 45). These transferases covalently alter aminoglycosides, rendering them unable to bind to their site of action, the bacterial ribosome. There are three classes of aminoglycoside-modifying enzyme. Aminoglycoside N-acetyltransferase (AAC) enzymes transfer an acetyl group from acetyl-coenzyme A (AcCoA) to an amino group of the aminoglycoside, producing N-acetylaminoglycoside and coenzyme A (CoASH) as products. Aminoglycoside O-phosphotransferase (APH) enzymes transfer a phosphate group from ATP or GTP in the presence of Mg²⁺ to a hydroxyl group on the aminoglycoside, yielding O-phosphoaminoglycoside and ADP or GDP. Lastly, aminoglycoside O-nucleotidyltransferase (ANT) enzymes transfer an adenylate group from ATP to the aminoglycoside, yielding O-adenylylaminoglycoside and pyrophosphate.

Monofunctional aminoglycoside-modifying enzymes are widespread, but in addition, several bifunctional aminoglycoside-modifying enzymes have been identified (49). AAC(3)-Ib/AAC(6′)-Ib′ and AAC(6′)-30/AAC(6′)-Ib′ from Pseudomonas aeruginosa, as well as ANT(3″)-Ii/AAC(6′)-IId from Serratia marcescens, have been identified in multidrug-resistant clinical isolates (7, 15, 20, 21, 29). However, by far the most widespread and best studied bifunctional resistance enzyme is AAC(6′)-Ie/APH(2″)-Ia (10). This resistance factor is found in Gram-positive organisms, including multidrug-resistant strains of Staphylococcus aureus and Enterococcus faecium, and can inactivate almost every known aminoglycoside by acetylation and/or phosphorylation (13).

Compared to the expression of independent monofunctional enzymes, it is presently unclear what selective advantage is generated by the combination of disparate activities into a single polypeptide in bifunctional enzymes. In fact, a strain of Escherichia coli has been identified that carries AAC(6′) and APH(2″) genes just 40 nucleotides apart but expressed as independent polypeptides (8). Several genetic and biochemical hypotheses have been proposed to explain the emergence and spread of bifunctional enzymes. These hypotheses are not mutually exclusive, and indeed, several may contribute to the driving force behind the evolution of bifunctional aminoglycoside resistance enzymes.

Genetic explanations put forth for the advantages of bifunctional resistance enzymes include the retention of two antibiotic-modifying activities in the population in the absence of a substrate for one enzymatic activity and coexpression of two enzymatic activities in response to a single antibiotic challenge (49). Genes expressing aminoglycoside-modifying enzymes are typically constitutive (46), and so we can expect that inducible expression does not play a role for bifunctional aminoglycoside-modifying enzymes. However, a mutation affecting the promoter of AAC(6′)-Ie/APH(2″)-Ia has led to overexpression of the bifunctional enzyme and increased resistance in a clinical isolate (27), demonstrating the plausibility of gene induction as a factor influencing bifunctional aminoglycoside-modifying enzymes in bacterial populations. These genetic arguments for the emergence of bifunctional aminoglycoside-modifying enzymes are reasonable given the strong selection pressures acting on antibiotic resistance genes. Additionally, genetic plasticity in populations treated with antibiotics readily favors the single-event changes that could generate a bifunctional enzyme.

In addition to genetic factors that influence bifunctional antibiotic resistance enzymes, biochemical advantages may also exist, which could resemble the advantages seen in bifunctional metabolic and signaling proteins. The main biochemical factor proposed is substrate channelling (49). This hypothesis is supported by studies that have demonstrated double modification of aminoglycosides in vitro (2, 18). If dual modification proves to be relevant in vivo, substrate channelling may improve the efficiency of this process and lead to improved antibiotic detoxification for the host bacterium. Unfortunately, little evidence exists for the presence of a biochemical advantage for bifunctional aminoglycoside resistance enzymes. Several kinetic studies of AAC(6′)-Ie/APH(2″)-Ia and its respective domains have demonstrated that the enzyme domains were active when expressed independently, and in fact, the kinetic parameters for the APH domain are effectively unchanged (4, 5, 48). Comparison of kinetic parameters for substrates for the AAC and APH domains did indicate a reduced efficiency of the AAC domain in the absence of the APH domain, but this may be a result of the particular enzyme construct studied (48). The nature of factors that have driven the spread of bifunctional resistance enzymes thus remains uncertain.

We employed small-angle X-ray scattering (SAXS) in order to probe the structure of AAC(6′)-Ie/APH(2″)-Ia. We observe that the enzyme has a bilobal structure, and rigid-body modeling of homology models of individual domains demonstrated that the AAC domain is positioned tightly adjacent to the APH domain in a rigid conformation. The addition of substrates to the enzyme demonstrated that it undergoes only subtle structural changes upon binding of substrates. These findings suggest new factors that may drive the evolutionary development of bifunctional aminoglycoside resistance enzymes.

MATERIALS AND METHODS

Protein expression and purification.

A pET-15b vector expressing N-terminal histidine-tagged AAC(6′)-Ie/APH(2″)-Ia was generously provided by G. D. Wright, McMaster University. The plasmid was transformed into Escherichia coli BL21(DE3), and expression was carried out in ZYP-5052 autoinduction medium with preculture in ZYP-0.8G, as described by studier (35). Briefly, 100 μl of stationary-phase culture was added to 2.0 ml of ZYP-0.8G and allowed to incubate, shaking, for 1 h at 37°C, prior to inoculation of 500 ml of ZYP-5052 with the entire 2.1-ml starter culture. All media were supplemented with 100 μg/ml ampicillin. The cultures were allowed to grow for 2.5 h at 37°C before the temperature was reduced to 22°C for overnight expression. Cells were harvested by centrifugation and lysed by sonication in 25 mM HEPES (pH 7.5), 1 M NaCl, 10 mM imidazole, 10 mM dl-dithiothreitol, and 2 mM phenylmethylsulfonyl fluoride. Protein was purified by nickel-affinity chromatography using Ni-nitrilotriacetic acid (NTA) Superflow resin (Qiagen) and a gradient of 10 to 250 mM imidazole to elute the protein in a single peak. The purity of the protein was verified by SDS-PAGE, and the fractions containing protein were pooled. Both acetyltransferase and phosphotransferase activities of the protein were verified using the assays described by Daigle et al. (13). Protein was concentrated using Amicon centrifugal concentrators with a 30,000-Da molecular weight cutoff (Millipore), and the buffer exchanged to 50 mM HEPES (pH 7.5), 5% glycerol, 5 mM tris(2-carboxyethyl)phosphine for all further experiments. Structure analysis and modeling of apo AAC(6′)-Ie/APH(2″)-Ia were conducted using a concentration series of 3.1 to 50 mg/ml of protein.

Substrate dialysis.

To study the effect of substrate binding on AAC(6′)-Ie/APH(2″)-Ia, 10-mg/ml solutions of the enzyme were dialyzed against combinations of the following substrates, all obtained from Sigma-Aldrich: CoASH, AcCoA, GDP, guanosine 5′-[β,γ-imido]triphosphate (GMPPNP, a nonhydrolysable analogue of GTP), and kanamycin A (Kan). Samples of 10 mg/ml protein (0.175 mM) were dialyzed in buffer containing the various substrates (CoASH, AcCoA, GDP, and GMPPNP at 1 mM each and Kan at 2 mM) alone or in combination using Spectra-Por membrane with a 15,000-Da molecular weight cutoff. A concentration of 2 mM magnesium chloride was also included in any dialysis including GDP or GMPPNP. Dialysis was carried out over 24 to 48 h at 4°C prior to SAXS analysis. Background scattering from the dialysis buffer was recorded for each dialysate and subtracted from its respective protein solution.

SAXS data collection.

All SAXS data were recorded using the SAXSess mc² nanostructure analysis system with charge-coupled-device (CCD) detection (Anton Paar GmbH). All samples were measured at 4°C. Data were collected in 30-min exposures for the 50-mg/ml and 25-mg/ml protein samples, 2.5 h for the 12.5-mg/ml sample, 8 h for the 6.3-mg/ml sample, and 10 h for the 3.1-mg/ml sample. Thirty-minute exposures were collected for all 10-mg/ml-substrate dialyzed samples. All scattering curves were checked for radiation damage before averaging (none was found).

Data processing and modeling.

Buffer subtraction and Porod correction were conducted using SAXSQuant (Anton Paar). Guinier analysis for all apo and substrate-bound data sets was conducted using the AutoRG function of PRIMUS (22). The concentration series of apo AAC(6′)-Ie/APH(2″)-Ia scattering data shows the same features at all concentrations at high angles and improved signal-to-noise at higher concentrations but some interparticle interference at low scattering angles (Fig. 1A). To eliminate this effect in the scattering, PRIMUS was used to merge high- and low-concentration data sets to obtain concentration-independent scattering of AAC(6′)-Ie/APH(2″)-Ia (Fig. 1B). Scattering data up to a momentum transfer of 0.6 Å⁻¹ were used for subsequent analysis and molecular modeling. The concentration-independent scattering curve was fit using the indirect Fourier transform program GNOM (37). The processed concentration-independent scattering data were used for 50 independent modeling cycles with the program GASBOR (36). These models were subsequently superimposed and averaged using the DAMAVER package (42). Crystal structures of related aminoglycoside resistance enzymes AAC(6′)-Ib [PDB ID: 2PRB, 22% identity and 50% similarity to the AAC domain of AAC(6′)-Ie/APH(2″)-Ia, E-value of 4 × 10⁻¹⁹] (28, 41) and APH(2″)-IIa [PDB ID: 3R6Z, 28% identity and 54% similarity to the APH domain of AAC(6′)-Ie/APH(2″)-Ia, E-value of 5 × 10⁻³³] (47) were used as templates for the construction of homology models of the AAC and APH domains of AAC(6′)-Ie/APH(2″)-Ia using MODELLER (33). The theoretical scattering from these models was calculated using CRYSOL (38). The homology models of AAC and APH domains of AAC(6′)-Ie/APH(2″)-Ia were modeled against the processed scattering data using the program BUNCH (31), with a 17-residue linker spanning residues 172 to 188. The ab initio and rigid-body models were superimposed using SUPCOMB (23). Electrostatic potential was calculated using APBS (3).

Fig 1.

(a) Small-angle X-ray scattering data measured from solutions of AAC(6′)-Ie/APH(2″)-Ia. Data were collected using the Anton Paar SAXSess mc² small-angle X-ray scattering system. Data are displaced vertically to illustrate features. Concentrations of protein solutions are indicated. q, momentum transfer; I(q), intensity of scattered X-rays. (b) Merged concentration-independent scattering from AAC(6′)-Ie/APH(2″)-Ia, fit using GNOM, and Kratky transformation of data (inset). (c) Pair-distribution function generated from GNOM fit of AAC(6′)-Ie/APH(2″)-Ia scattering data.

RESULTS

SAXS analysis of AAC(6′)-Ie/APH(2″)-Ia.

The radius of gyration (R_G) calculated from the concentration-independent AAC(6′)-Ie/APH(2″)-Ia scattering data was 32 Å, with a calculated radius of homogeneous sphere (R_H) of 42 Å, consistent with dynamic light scattering measurements of the pure protein (data not shown). The Kratky plot of the merged and desmeared scattering curve is indicative of a particle that is fully folded in solution (Fig. 1B). The Guinier plot for this curve is linear for (q × R_G) < 1.3. The pair-distribution plot obtained from GNOM for the scattering by AAC(6′)-Ie/APH(2″)-Ia shows a distribution characteristic of “dumbbell”-shaped particles (Fig. 1C). The maximum dimension (D_max) recovered from this analysis is 100 Å, which is notably smaller than the sum of the D_max values calculated for the individual domains (52 Å for AAC and 73 Å for APH). The linearity of the Guinier plot combined with this dimension excludes the possibility that the protein exists as a dimer and, thus, supports the idea that the bifunctional protein is monomeric in solution, in agreement with past reports of gel filtration data for AAC(6′)-Ie/APH(2″)-Ia (5). A D_max of 100 Å greatly restricts the possible conformations of the protein in solution and, in fact, implies that the AAC and APH are rigid in position and orientation with respect to each other. Flexibility would increase the mean distance between the domains, which would certainly generate a D_max much larger than 100 Å.

Solution structure of AAC(6′)-Ie/APH(2″)-Ia.

The DAMAVER-averaged GASBOR ab initio model of AAC(6′)-Ie/APH(2″)-Ia shows a distinct bilobal appearance, with one lobe notably larger than the other (Fig. 2A). Rigid-body modeling of AAC(6′)-Ie/APH(2″)-Ia against the scattering data places the AAC(6′) domain close to the APH(2″) domain (Fig. 2B), with a distance of 44 Å between their centers of mass. The ab initio and rigid-body models of AAC(6′)-Ie/APH(2″)-Ia superimpose well, with a normalized spatial discrepancy (NSD) value of 0.86. Unfortunately, at the resolution and information content available from SAXS, the absolute orientation of the AAC domain cannot be uniquely established. However, its location with respect to the APH domain remains consistent throughout repeated modeling cycles, and geometrical requirements of the linker limit the possible orientations the AAC domain can adopt, which has led us to select one representative model for display.

Fig 2.

SAXS models of full-length AAC(6′)-Ie/APH(2″)-Ia. (a) Ab initio model of AAC(6′)-Ie/APH(2″)-Ia generated by constructing 50 independent dummy residue models using GASBOR and averaging with DAMAVER. (b) Superposition of ab initio and example rigid-body model of AAC(6′)-Ie/APH(2″)-Ia. Rigid-body model was constructed from AAC (cyan) and APH (magenta) homology models connected with a 17-residue dummy atom chain (orange) and modeled against AAC(6′)-Ie/APH(2″)-Ia SAXS data by BUNCH. Images were generated using PyMol (the PyMOL molecular graphics system, version 1.3; Schrödinger, LLC).

Effect of substrate binding on SAXS scattering of AAC(6′)-Ie/APH(2″)-Ia.

A slight effect of interparticle interactions was observed in the solutions containing substrate dialysates due to the concentration of protein used (10 mg/ml). However, because the protein concentration was identical between samples, a direct comparison between apo and substrate-incubated enzyme could be made. Scattering data obtained from AAC(6′)-Ie/APH(2″)-Ia in the presence of AcCoA, CoASH, GDP, GMPPNP, and Kan did not produce obvious changes to the overall scattering pattern. Nevertheless, Guinier analysis of scattering from solutions of AAC(6′)-Ie/APH(2″)-Ia in the presence of AcCoA, CoASH, GDP, or GMPPNP resulted in a smaller R_G than the protein alone at this concentration, although at 0.8 to 1.3 Å, the difference is small (Table 1). The addition of Kan had the opposite effect, generating a larger apparent R_G, as well as a marked departure from linearity in the Guinier plot (not shown). This suggests that some degree of aggregation occurs when Kan is added, which agrees with casual observations that the addition of Kan to some preparations of AAC(6′)-Ie/APH(2″)-Ia can result in a turbid solution. Interestingly, incubation with Kan, as well as CoASH or GMPPNP, but not GDP greatly ameliorated this effect (Table 1).

Table 1.

Radius of gyration calculated from SAXS profiles of 10-mg/ml solutions of AAC(6′)-Ie/APH(2″)-Ia incubated in the presence of substrate(s)^a

Solution	AAC cosubstrate	APH cosubstrateb	Aminoglycoside	RG (Å [±SD])c
AAC(6′)-Ie/APH(2″)-Ia alone				30.2 ± 0.2
AAC(6′)-Ie/APH(2″)-Ia plus substrate(s)	CoASH			29.4 ± 0.2
	AcCoA			29.0 ± 0.2
		GDP		29.0 ± 0.2
		GMPPNP		29.4 ± 0.2
	CoASH	GDP		29.1 ± 0.4
	AcCoA	GDP		28.6 ± 0.3
			Kan	33.9 ± 0.2d
	CoASH		Kan	30.6 ± 0.2
		GDP	Kan	33.3 ± 0.2d
		GMPPNP	Kan	30.5 ± 0.2
	CoASH	GMPPNP	Kan	29.4 ± 0.2
	CoASH	GDP	Kan	30.1 ± 0.2

^aSolutions of 10 mg/ml of AAC(6′)-Ie/APH(2″)-Ia were prepared and dialyzed in the presence of 1 mM AcCoA, CoASH, GDP, or GMPPNP or 2 mM Kan and combinations thereof. Dialysis was carried out for 24 to 48 h prior to SAXS data collection. Corresponding buffer measurements were recorded for every substrate combination and subtracted from the corresponding protein solution.

^bAll incubations with GDP or GMPPNP also contained 2 mM magnesium chloride.

^cR_G for each data set was determined with the AutoRG function of PRIMUS.

^dInspection of the Guinier plot revealed that aggregation occurred in these samples. Preparation of enzyme-substrate solutions using these substrates was also seen to generate turbid solutions.

DISCUSSION

Rigid structure of AAC(6′)-Ie/APH(2″)-Ia.

The arrangement of domains in AAC(6′)-Ie/APH(2″)-Ia forms an elongated but compact structure. As aminoglycoside phosphotransferases such as the APH domain of AAC(6′)-Ie/APH(2″)-Ia share structural homology with eukaryotic protein kinases (19), it is interesting to compare this structure to eukaryotic kinase structures. Indeed, Src (9), Abl (30), and Fes (16) kinases have been crystallized with regulatory SH2 or SH3 domains in a location adjacent to the N-terminal lobe of the kinase domain, similar to the AAC domain of AAC(6′)-Ie/APH(2″)-Ia. These SH2 and SH3 domains fulfill a different functional role in their respective kinases and interact more transiently than the enzymatic AAC domain in AAC(6′)-Ie/APH(2″)-Ia, but the architectural similarity is noteworthy.

The compact nature of the SAXS-derived model for AAC(6′)-Ie/APH(2″)-Ia is consistent with previous biochemical studies. For example, the results reported by Boehr et al. (5) suggest that an overlapping region of polypeptide between AAC and APH domains is required for both enzyme activities, which can be readily rationalized by our low-resolution model. Disruption of the interface between the two domains, as it is located close to both active sites, will undoubtedly negatively affect both enzyme activities. The model is also consistent with kinetics experiments on the AAC and APH domains that could not find evidence of cooperativity between the two domains but did demonstrate that the addition of substrate for one domain protects the activity of the other from thermal denaturation (5, 25). A lack of cooperativity is consistent with a rigid structure for this enzyme, and the tight association of domains would suggest that stabilization of one domain would in turn stabilize the other against denaturation.

The lack of any large change to the R_G of AAC(6′)-Ie/APH(2″)-Ia upon substrate binding suggests that the protein does not experience major interdomain conformation changes during its catalytic cycle. Local conformational changes restricted to the AAC or APH domain can easily explain the small (0.8 to 1.3 Å) change in R_G seen when binding CoASH, AcCoA, GDP, or GMPPNP. Overall, it appears that AAC(6′)-Ie/APH(2″)-Ia maintains a rigid conformation while binding substrate. This behavior is similar to what has been observed for APH(3′)-IIIa, another clinically important aminoglycoside resistance enzyme (6), but is in sharp contrast to what is seen for APH(9)-Ia, an enzyme whose role in antibiotic resistance is debatable (17). We have previously suggested that the absence of major conformational changes may signify that the aminoglycoside resistance enzyme is optimized for rapid drug detoxification and, thus, is an important feature of dedicated resistance enzymes (6). This would clearly be consistent with the existence of AAC(6′)-Ie/APH(2″)-Ia in numerous clinical isolates (10). It should be noted that while the addition of Kan to the protein resulted in an increase in measured R_G for the bifunctional enzyme, the Guinier analysis of this scattering profile (and the Kan-plus-GDP profile) indicated that some aggregation occurred in these samples, so the apparent R_G value may be considered erroneously high. Whether this effect is due to a specific interaction of Kan with AAC(6′)-Ie/APH(2″)-Ia or some nonspecific effect is unclear, but the negation of this effect by the addition of CoASH or GMPPNP but not GDP is an intriguing observation.

Implications of rigid structure of AAC(6′)-Ie/APH(2″)-Ia.

Our observation of the gross features of this enzyme, with AAC and APH domains tightly associated, though consistent with previous biochemical data is actually surprising. The simple genetic fusion of two independent enzyme domains would be expected to yield loosely tethered domains connected by a flexible linker, as the termini of most proteins are typically more flexible than the core. Our observation of tight association of the domains in AAC(6′)-Ie/APH(2″)-Ia strongly suggests that in addition to fusion of the two enzymes, further change with some adaptive advantage must have occurred to generate a tight interface between the domains. This observation influences our interpretation of the hypotheses that have been proposed to explain the emergence and spread of bifunctional antibiotic resistance enzymes.

Genetic explanations for bifunctionality do not require AAC(6′)-Ie/APH(2″)-Ia to adopt a rigid structure in solution, as the maintenance of two activities in the population or inducible expression of two activities can easily also be accomplished in an enzyme possessing two loosely tethered domains with a flexible linker. Therefore, genetic explanations alone are insufficient to explain the structure we observe for this enzyme. Of the various biochemical hypotheses that may be surmised for bifunctional enzymes, substrate channelling has been suggested (49), and while it can be argued that efficient dual modification could provide an evolutionary advantage, the observation that phosphorylation alone typically reduces binding to the same degree as double modification suggests that dual modification may not be relevant in vivo (24). The level of detail available from our SAXS-derived model of AAC(6′)-Ie/APH(2″)-Ia is insufficient to prove or disprove the existence of a substrate channelling path, and this remains an interesting possibility. However, analysis of the structure suggests an additional explanation for the emergence and spread of bifunctional aminoglycoside-modifying enzymes.

Aminoglycosides are polycationic, and because of this, aminoglycoside-modifying enzymes almost invariably have low pI and a negatively charged antibiotic-binding site (40). The negatively charged binding site is thought to steer the aminoglycosides toward the active site of the enzyme in a manner similar to although more complex than that observed for acetylcholinesterase (39) and other enzymes, such as β-lactamases (43). Similar to other aminoglycoside-modifying enzymes, homology modeling of the AAC and APH domains of AAC(6′)-Ie/APH(2″)-Ia reveal pronounced negatively charged patches at the location of the two aminoglycoside-binding sites. The electrostatic potential surrounding the bifunctional enzyme will influence binding of positively charged substrate to both active sites, and this potential in three dimensions will be a function of the relative position and orientation of the two individual domains. Thus, the relative domain orientations are important for the functioning of this enzyme.

In a bifunctional enzyme with the two domains tethered by a flexible linker, conformational sampling will generate an inconsistent electrostatic potential field around the enzyme, in which attractive and repulsive interactions frequently cancel out. In the most extreme case, one enzyme domain may spend a considerable amount of time interacting with the other in a nonproductive orientation, reducing the availability of the enzyme active site for catalysis. This configuration is clearly detrimental for conferring resistance. We propose that to produce a consistently constructive electrostatic field, AAC(6′)-Ie/APH(2″)-Ia has evolved to form a rigid structure with tight interactions between the two domains. SAXS does not allow us to identify the precise rotational orientation of the AAC domain, but our analysis is most consistent with an orientation that places its active site near the APH active site and allows both aminoglycoside-binding sites to lie on the same face of the protein (Fig. 3). In this manner, the electrostatic fields of the individual domains generate a region of negative potential that guides and orients aminoglycoside substrates to the two active sites. Should this prove to be the case, a form of channeling is possible in which substrate that is not productively modified by one enzymatic domain can diffuse to and be modified by the other domain, instead of leaving the enzyme unmodified. The bifunctional enzyme can thus provide a biochemical advantage over individual and separate monofunctional enzymes by generating a larger region of negative electrostatic potential. A constructive electrostatic field such as this could be a conserved feature that improves the substrate binding to bifunctional aminoglycoside-modifying enzymes.

Fig 3.

Solvent-accessible surface potential of rigid-body model of AAC(6′)-Ie/APH(2″)-Ia, calculated using APBS. The surface potential ranges from −5 to +5 kT/e (red to blue, respectively). Locations of AAC and APH aminoglycoside binding sites in this model are indicated.

It is now well established that many antibiotic resistance elements have a prehistoric origin (11). Indeed, the rigid structure of AAC(6′)-Ie/APH(2″)-Ia suggests that it existed prior to human antibiotic administration. The resistance factor later identified as AAC(6′)-Ie/APH(2″)-Ia was first observed in clinical strains in 1977 (14, 26). This finding was less than 30 years following the identification of its first potential substrate, neomycin, in 1949 (44). It is unlikely that the adaptive change necessary to generate this rigid bifunctional enzyme could have occurred in this time, suggesting that, like the bifunctional β-lactamase bla_LRA-13 (1), AAC(6′)-Ie/APH(2″)-Ia existed in the environment, in the antibiotic resistome, prior to human overproduction and use of antibiotics (12, 46).

The other bifunctional aminoglycoside-modifying enzymes AAC(3)-Ib/AAC(6′)-Ib′, AAC(6′)-30/AAC(6′)-Ib′, and ANT(3″)-Ii/AAC(6′)-IId may be considered more recent gene fusions, as the catalytic domains in these bifunctional enzymes are almost identical to those of their monofunctional counterparts. It is unclear whether these enzymes also adopt rigid conformations in solution or exhibit a loosely tethered behavior. It is possible that AAC(6′)-Ie/APH(2″)-Ia is unique in possessing a rigid structure, which has been selected to optimize enzymatic properties. However, if electrostatic steering indeed plays a role in the binding of substrates to bifunctional aminoglycoside-modifying enzymes, these enzymes may also exhibit rigid structures, though we would not expect these enzymes to be catalytically optimized. Indeed, studies on catalysis by AAC(3)-Ib/AAC(6′)-Ib′ demonstrated that fusion actually attenuates the acetyltransferase activity of the AAC(6′)-Ib′ domain (21).

Our model of AAC(6′)-Ie/APH(2″)-Ia gives us a first look at the solution structure and domain arrangement of this enzyme. The rigid conformation observed for the protein is surprising but consistent with existing biochemical data gathered for this enzyme. Furthermore, the observed conformation has sparked speculation on the role of electrostatic steering in bifunctional aminoglycoside-modifying enzymes. It will be informative to examine other bifunctional aminoglycoside resistance enzymes to identify whether those also possess a rigid structure or whether this feature is unique to AAC(6′)-Ie/APH(2″)-Ia. AAC(6′)-Ie/APH(2″)-Ia remains the most prevalent resistance enzyme to aminoglycosides in Gram-positive bacteria, and a thorough structural understanding of this enzyme is imperative to better understand and develop strategies to counter the resistance conferred by this enzyme.