Aminoglycosides are broad-spectrum antibiotics that bind to bacterial 16S rRNA. This binding causes an increase in proof-reading errors and mistranslation which eventually leads to disruption of the cell membrane and death of the microorganism. Structurally, these antibiotics consist of a 2-deoxystreptamine ring with amino sugar substitutions at either the 4,5 or 4,6 positions (Figure 1). Streptomycin is an exception to the common aminoglycoside chemical motif. These antibiotics are used to treat serious infections caused by Gram-negative bacteria.

Enzymatic modification of aminoglycoside antibiotics is the major resistance mechanism to the action of these antibiotics [1]. The aminoglycoside nucleotidyltransferase(2″)-Ia (ANT) is one of the most clinically important aminoglycoside modifying enzymes (AGMEs). This enzyme catalyzes the adenylation of aminoglycosides which contain a hydroxyl group at the 2″ position. ANT is one of only four of the over fifty known AGMEs to modify primarily at positions on ring C of aminoglycosides (Figure 1) [2]. Most AGMEs modify positions on the A or B ring.

Figure 1: Aminoglycosides Kanamycins A and B (top) and Neomycin (bottom). Substitution at the 2′-site for kanamycin B is shown in parenthesis (top). Amino groups are shown in blue with pKas indicated below. Numbering are shown in parenthesis on the side. Alternate letter-labeling of the rings are also shown.

ANT was first reported in a clinical isolate of Klebsiella pneumoniae in 1971 [3]. By 1985 ANT was the most often detected aminoglycoside modifying enzyme in resistant clinical isolates in North America and had been found in each of the major continents [4,5]. ANT continues to be one of the two most prevalent enzymatic aminoglycoside resistance mechanisms in Pseudomonas aeruginosa [6,7]. Despite its medical importance, very little is known about this enzyme. This protein is difficult to isolate in large quantities with a high level of purity [8,9]. It is also relatively unstable and has a low degree of solubility [10]. Because of these problems, earlier studies focused on kinetic assays and in vivo inhibition patterns of this enzyme [8,11-13]. More recently, analysis of thermodynamic properties of ANT–aminoglycoside complexes showed that all binding interactions were characterized by favorable enthalpy (−ΔH) and unfavorable entropy (−TΔS) [14,15], which is typical for carbohydrate-protein interactions [16-18]. This study also looked at substrate selectivity by ANT [15].

To further investigate the thermodynamics of ANT–aminoglycoside interactions the role of proton linkage and pH dependence of substrate binding was investigated. Two different calorimetric approaches were used to determine binding-linked protonation/deprotonation and the effect of pH on the association constant (K_b) of the binary ANT− aminoglycoside complexes. In the first method, enthalpy change (ΔH_obs) upon ligand (aminoglycoside) binding to the enzyme in buffers with different ionization enthalpies (ΔH_ion) was determined to derive intrinsic enthalpy change (ΔH_int) and the net number of bindingcoupled proton exchange (Δn). When pK_a shifts occur in either the protein or ligand, uptake or release of protons to the buffer will occur. This exchange will contribute to the observed ΔH and must be considered in the thermodynamic analysis of the system. The binding of aminoglycosides to AGMEs include significant contribution to the observed enthalpy of complex formation from the protonation and deprotonation of titratable groups on the enzyme and ligands [15,19,20]. In the second method, pH dependence of association was measured using a single buffer system. This buffer system encompassed the pH range of Kb values accurately measurable by ITC for ANT– aminoglycoside interactions. This paper shows that multiple groups on ligand and enzyme shift their pK_as upon formation of the binary ANT–aminoglycoside complexes.

Reagents

All materials were of the highest purity commercially available. All were purchased from Sigma-Aldrich Co. (St. Louis, MO) except for tris(2-carboxyethyl)phosphine (TCEP) purchased from Fluka (Buchs, Switzerland). ANT was prepared as described previously [10]. Activity assays of the enzyme were also performed as described previously [14].

Isothermal titration calorimetry

ITC experiments were performed at 20°C using a VP-ITC microcalorimeter from Microcal, Inc. (Northampton, MA).The enzyme was exchanged into the appropriate buffer using a Sephadex G-25 column and the ligand solution was prepared using the same buffer solution. The final buffer for determination of pH dependence was 50 mM MES/Tris, 50 mM KCl and 2 mM TCEP. The buffers for determination of the net uptake of protons contained 50 mM of the indicated buffer. The following buffers with the enthalpy of ionization in kcal mol^-1 in parenthesis were used at pH 7.1: PIPES (2.68), HEPES (4.87), MOPS (5.04) and ACES (7.26), at pH 7.6: HEPES (4.87), MOPS (5.04), Bicine (6.29) and Tris (11.3), and at pH 8.1: HEPES (4.87), Bicine (6.29), TAPS (9.65) and Tris (11.3) [21]. The concentration of aminoglycoside was determined by NMR and activity assays as described previously [14]. Both enzyme and ligand solutions were degassed under vacuum for 10 min at 15°C. Titrations consisted of 29 injections of 10 μL and were separated by 240 s. Cell stirring speed was 300 rpm. Each titration contained 5–40 μM enzyme in the sample cell. The aminoglycoside concentration was 60–800 μM in the injection syringe. The standard errors represent the deviation including curve fitting errors of the three titrations. The c values (c = K_bM_t, where M_t is the concentration of macromolecule binding sites) were in the range 10–30 for experiments in the pH range 6.5–7.8. This range is ideal for accurately determining binding constants by ITC [22]. For titrations above pH 7.8 and below pH 6.5 the c values were in the range 0.4–10. The use of less than optimum c values was necessitated by the low affinity at these pH values combined with the limited solubility and strong tendency of oligomerization of this enzyme above 40 μM (~1 mg/ml) concentration. Monomeric state of the protein was confirmed by analytical size-exclusion chromatography. For all complexes observed in this study the thermodynamic parameters determined were independent of protein concentration.

The heat of dilution for each aminoglycoside was determined by titrating aminoglycoside into buffer in the absence of enzyme and subtracted prior to curve fitting. For all titrations the pH was confirmed immediately prior to the start of the experiment. All data were fit to the single-site binding model of Origin 5.0 (Microcal, Inc) to determine the binding constant (K_b), enthalpy of binding (ΔH) and stoichiometry (n). The free energy (ΔG) and entropy (ΔS) changes associated with binding were determined using the equations:

          ΔG = −RT K_b

         ΔG = ΔH −TΔS

ANT is one of the “less promiscuous” AGMEs and modifies only kanamycins and gentamicins while neomycins are competitive inhibitors of this enzyme. In this study, three different aminoglycoside antibiotics representing kanamycins and neomycins were used to determine thermodynamic properties of their complexes with ANT as a function of pH. Kanamycin A and kanamycin B were selected for this study because the thermodynamic parameters of their complexes with ANT show significant differences despite a minimal structural difference between them (-NH₂ versus –OH at the 2′-site in kanamycin B and kanamycin A respectively) (Figure 1) [15]. Neomycin B was chosen as the representative of the neomycin class because of the high affinity of this antibiotic for ANT [15].

Effect of complex formation on the ionization of functional groups: In the presence of binding-linked protonation, the observed enthalpy (ΔH_obs) includes contribution from various sources according to the equation [23]:

          ΔH_obs = ΔH_int + Δn [α ΔH_ion + (1-α)ΔH_enz ] + ΔH_bind

in which ΔH_int is the intrinsic enthalpy of binding and Δn represents the net proton transfer. ΔH_obs denotes the observed binding enthalpy of complex formation in a buffer where ΔH_ion describes the heat of ionization of the buffer. The term Δn [α ΔH_ion + (1-α)ΔH_enz ] represents the heat of ionization of groups from the ionization of buffer and the protein to maintain pH, where α represent fraction of protonation contributed by the buffer [23]. In addition, ΔH_bind represents the heat of binding of buffer to the enzyme. In the presence of sufficient salt (i.e., 50 mM KCl), ΔH_bind is assumed to be zero and the contribution from the ionization of amino acids remains the same at a given pH. Thus, by performing experiments in buffers with different heats of ionization, one can easily determine ΔH_int and Δn, which is based on the simplified version of the above equation (ΔH_obs = ΔH_int + Δn ΔH_ion). An example is shown in Figure 2 where the heat of ionization of buffers is plotted against ΔH_obs for the formation of the ANT–kanamycin B complex at pH 7.1 and 7.6. The values for Δn and were determined from the slope and the intercept respectively. However, note that ΔH_int still includes the heat of ionization of groups contributing to Δn (i.e., ΔH_int = (ΔH_int + ΔH_{functional groups} · Δn) which would represent the true ΔH_int only when Δn=0). For the buffers used in this work, a net proton uptake by enzyme–ligand complex yields a positive Δn value. Bindinglinked proton uptake was observed in the formation of the binary ANT–aminoglycoside complexes with the exception of one condition; ANT–neomycin complex yielded a negative Δn at pH 7.1 indicating a net release of protons from the complex under these conditions (i.e., lowered pK_as) (Table 1). Values of ΔH_intand Δn showed strong dependence on pH as shown in Table 1 and Figure 3. Observed positive values for Δn (Δn > 0) for the enzyme-aminoglycoside interactions suggests that several functional groups may have up shifted pK_as in the binary ANT–aminoglycoside complexes. It is also reasonable to assume that increased pK_a of amino groups in ligands contribute significantly to the observed positive value of Δn. Consistent with this, pK_as of several amino groups in enzyme-bound neomycin showed up to one pK_a unit higher shifts when bound to another aminoglycosidemodifying enzyme [24].

Figure 2: Change in the observed enthalpy of the binary enzyme–kanamycin B complex at pH 7.1 (○) and at pH 7.6 (■) as a function of the heat of ionization of different buffers.

Table 1: pH-dependent protonation in the binary ANT-Aminoglycoside complexes

ANT-aminoglycoside association as a function of pH

The second calorimetric method involves determination of the association constant (K_b) for the formation of binary enzyme– aminoglycoside complexes as a function of pH. A typical titration data demonstrating the effect of pH on the binding affinity of neomycin to ANT is shown in Figure 3 (please note the differences in y-axis scale). Effect of pH on the binding constant (K_b) of ANT–aminoglycoside complexes was determined by performing binding experiments in the pH range of 6.3 to 8.1. In general, all three aminoglycosides showed an increased affinity with increasing pH up to pH ~6.8. A plateau was observed at neutral pH region and the affinity started to decline above pH 7.5. Data for the binding of kanamycins and neomycin are shown in Figures 4, and 5 respectively.

Figure 3: Titration of ANT with neomycin at pH 7.4 (left panel) and pH 8.2 (right panel). Thermograms (upper panels) are shown with the fitted data (lower panels).

Binding of kanamycin A vs. kanamycin B to ANT

Kanamycin A and B are 4.6-disubstituted aminoglycosides which are identical except for the substituent at the 2′ position. Kanamycin B contains a 2′-NH₂ while kanamycin A possesses a 2′-OH. The presence of an amino instead of a hydroxyl at this position results in a fivefold increase in affinity for kanamycin B relative to kanamycin A [15]. Binding of kanamycin A and B to ANT leads to a similar binding linked protonation profile. In both cases Δn increases from 7.1 to 7.6 then decreases from 7.6 to 8.1 (Table 1, Figure 4). At pH 7.6 and 8.1 the difference in Δn for the two complexes is within error suggesting that the 2′ position does not significantly affect Δn at these pH values. At pH 7.1 kanamycin B has a Δn value that is 0.3 less than the value for kanamycin A. If the binding interactions and energetic contributions remain consistent at the other positions in the two complexes, the difference must be attributable to the 2′-NH₂ vs. 2′-OH. Crystal structures of aminoglycosides bound to AGMEs show the predominant interactions involve the amino groups on the substrate and aspartyl and to a lesser extent glutamyl side chains on the enzyme [25-27]. In structures of AGMEs which show a strong preference for kanamycin B relative to kanamycin A (similar to ANT) the 2′ position interacts with an acidic residue [25,27,28]. The most plausible explanation is that the contributions from the other positions on kanamycin result in a larger positive Δn for complex formation with ANT. The active site aspartate or glutamate that interacts with the 2′ position may have an up shifted pK_a due to multiple acidic residues in the active site. The binding of kanamycin B with a positively charged amino group at the 2′ position will result in a larger decrease in pK_a for the carboxyl group on the enzyme due to the favorability of a charge-charge interaction [29,30]. This effect would be magnified because the binding of aminoglycoside would be removing the carboxyl group from solvent [29]. If the hydroxyl group of kanamycin A forms a hydrogen bond with the same residue, it would also decrease the pK_a of the side chain carboxyl, but to a lesser extent [29,31]. Although less likely, the difference in Δn at pH 7.1 could be due to a decrease in pK_a of the 2′-NH₂ of kanamycin B upon binding to ANT and an increase in affinity of 2′-NH₂ compared to 2′-NH₃ + for the enzyme, which also fits the data. In most cases, though, protonated amino groups lead to higher affinity for AGMEs [32]. Additionally, increases in pK_a values for amino groups that are directly involved in binding to AGMEs have been observed previously [21].

Figure 4: Effect of pH on the binding constant (Kb) of kanamycin A (panel A, filled circles) and kanamycin B (panel B, filled circles) to ANT. Determined Δn (Δ) is shown with the overall charge on aminoglycosides ([NH₃⁺]- open squares) and the total charge ([NH₃⁺]+Δn, open diamonds) at three different pH.

The affinity as a function of pH curves for kanamycin A and B binding to ANT show a similar profile from pH 6.4 to 8.1 (Figure 4A and B). At all pH values kanamycin B has a higher affinity for ANT than kanamycin A. Accurate binding data could not be obtained for kanamycin A below pH 6.4 or above 8.1 due to weak binding and the solubility limits of the enzyme. As shown in Figure 4A and B, while the charge ([NH₃⁺]) on the amino groups of kanamycin A, which was determined based on pK_as of kanamycin A in solution declines above pH 7.1, an increase in Δn compensates for this loss and the apparent total charge ([NH₃⁺] + Δn) remains unchanged until pH 7.6. Above pH 7.6, both Δn and ([NH₃⁺] + Δn) start to decline. The loss of positive charge on kanamycin parallels the loss of affinity in the pH range 7.6 to 8.1 supporting the conclusion that positively charged amino groups of kanamycin interact with negatively charged amino acids in the active site. Thus loss of such favorable interactions leads to lower affinity. However, the low pH side of the curve clearly shows that this is not the only interaction that affects the affinity of kanamycin A to the enzyme. Deprotonation of a group or groups on the protein, ligand or both decreases the affinity of kanamycin to the protein above pH 6.8. Consistent with the most probable explanation for the binding linked protonation data at pH 7.1, deprotonation of the functional group of an acidic residue (or residues) with an up shifted pK_a is likely. Deprotonation that may occur on the aminoglycoside, however, cannot be excluded. Only the 3-amino group of the kanamycin A would be significantly affected in this pH regime (if it remains unaltered or changed only by a small amount upon binding to ANT). If this were the case, the deprotonated form of the 3-NH₂ makes more favorable contacts with ANT.

Binding of kanamycin vs. neomycin to ANT

Neomycin is a 4,5-disubstituted aminoglycoside that is a competitive inhibitor of ANT. The pattern of binding linked protonation for this aminoglycoside binding to ANT differs from both kanamycin A and B. In the ANT–neomycin complex, Δn increased continuously with increasing pH (Table 1). At pH 7.1 Δn is slightly negative indicating a small net release of protons from the complex. This value is approximately 0.4 units less than for the ANT–kanamycin B complex. The positions on the A and B ring which are critical for binding ANT are identical in neomycin and kanamycin B [15]. Therefore the difference in binding linked protonation is a result of lowered pK_as on either neomycin or ANT caused by interactions between the C or D ring of the antibiotic with the enzyme. A similar Δn is observed for all three antibiotics at pH 7.6. Unlike the kanamycins, however, net proton uptake continues to increase for neomycin from 7.6 to 8.1. This difference may be attributed to the fact that amino groups of neomycin have higher pK_a values free in solution and they become even higher when bound to an AGME [24]. Thus, even the highest pH used in this work may be still well below the pK_as of its amino groups which causes continued proton uptake.

Another difference between neomycin and the kanamycins is that ΔHint continued to become more negative with increasing pH. In addition, overall change in ΔH_int was much larger than those observed for the complexes of kanamycin A and kanamycin B between pH 7.1 and 8.1. The large increase of ΔH_int in the ANT–neomycin complex can be accounted for by the ionization of amino groups as reflected in the increasing Δn as pH increases. As indicated earlier, ΔH_int may contain contributions from the heat of ionization of titratable groups unless Δn is zero. Intrinsic enthalpy change (ΔH_int) determined at a pH where Δn ≈ 0, subtracted from values obtained at regimes where Δn ≠ 0 and the result divided by Δn provides insight into the contribution of ΔH_enz to the observed ΔH_int. This calculation yields 8.9 and 10.8 kcal mol-1 at pH 7.6 and 8.1 for neomycin. It appears that most of the increase in ΔH_int can be attributed to binding-induced pK_a shifts of amino groups which typically have a heat of ionization of 9-12 kcal mol^-1 [33,34]. A similar calculation with formation of the ANT–kanamycin B complex results in values of 8.3 and 8.5 kcal mol^-1 at pH 7.6 and 8.1 respectively. These values are too small for the contribution of amines only and too large for contribution of imidazole (~6 kcal mol^-1) indicating that multiple groups with different heats of ionization contribute to ΔH_int. Note that the heat of ionization of carboxyl groups (0.5^-1 kcal mol^-1) would not greatly impact the observed enthalpy change unless several of them contribute to ΔH_int. These calculations suggest that either different titratable groups are contributing to ΔH_int for kanamycins and neomycin or the same groups contribute differentially due to differences in magnitude of pK_a shifts between the two types of antibiotics.

The log Kb versus pH curves for the formation of binary enzyme– aminoglycoside complexes shows that neomycin has a higher affinity for ANT over the entire pH range. At the lower pH range (6.1-7.0) the neomycin and kanamycin B curves are nearly identical (except the neomycin is shifted up at each point). This supports the concept that protonation of the same functional group(s) on the enzyme are responsible for the decreased affinity at pH values less than 6.7. At higher pH values the differences between ANT–neomycin and ANT–kanamycin complexes are more apparent. The affinity for both kanamycin A and B drops sharply above pH 8.2 while only a small decrease in K_b is observed with neomycin (Figures 4 and 5, note the K_b values are presented on a log scale). For neomycin the association constant decreases only from 3.7 x 10⁵ M^-1 to 1.4 x 10⁵ M^-1. For kanamycin B there is a tenfold decrease in affinity over the same pH range. At pH 8.2, K_b = 1.3 x 10⁵ M^-1. At pH 8.4 the association constant decreases to 1.2 x 10⁴ M^-1. The difference in the K_b vs. pH profiles at higher pH values can be explained by the continued increase in Δn for neomycin which leads to an increase in. total charge ([NH₃ ⁺] + Δn) (Figure 5). This increase is in contrast to kanamycin A and B which decline in total charge above pH 7.6. This is consistent with the presence of groups with up shifted pK_as in enzyme–neomycin complex. However the fact that affinity decreases as total charge increases shows that the charge of the amino groups of neomycin cannot solely account for the change in affinity at higher pH. The decrease observed in affinity must be due to loss of other favorable interactions because these groups would still be highly charged.

Figure 5: Effect of pH on the binding constant (Kb) of neomycin (filled symbols) to ANT. Determined Δn (Δ) is shown with the overall charge on aminoglycosides ([NH₃⁺]- open squares) and the total charge ([NH₃⁺]+Δn, open diamonds) at three different pH.

In a system involving multiple ionizable groups on both substrate and enzyme such as aminoglycosides binding to AGMEs, the observed global thermodynamic parameters will reflect contributions from several functional groups. In this work, two important contributors to the pH dependent parameters are identified. First, the importance of the interaction between the 2′ position of the aminoglycoside with the enzyme was determined by observed differences in Δn and K_b between kanamycin A and B. Second, the higher affinity for neomycin compared to kanamycins can be partially attributed to a larger up shift in pK_as of amino groups on neomycin. These two factors do not fully account for the pH dependent properties of the ANT–aminoglycoside complexes. The net effect of protonation and deprotonation of other substrate amino groups and enzyme functional groups could not be parsed into individual contributions to the total thermodynamic profile. This difficulty highlights problems with breaking down observed global thermodynamic parameters into individual components, even in simple 1:1 enzyme–substrate complexes. Thus, attribution of changes in these parameters to specific sites should be made carefully since variable and opposing effects may contribute to the same parameter. This is especially true for thermodynamic studies of systems that may include ionizable groups and enzymes that may bind structurally different ligands. Almost all AGMEs are capable of binding structurally different aminoglycosides. In this work properties that contribute to substrate selectivity were shown by determining binding linked protonation and affinity as a function of pH. Structural and further thermodynamic analysis will be necessary for a more detailed analysis of the factors that contribute to the substrate selectivity and promiscuity of ANT.

This research was partly supported by a Grant from the National Science Foundation (MCB 0842743 to EHS).