Structure-Activity Relationships among the Kanamycin Aminoglycosides: Role of Ring I Hydroxyl and Amino Groups

Abstract

The kanamycins form an important subgroup of the 4,6-disubstituted 2-deoxystreptamine aminoglycoside antibiotics, comprising kanamycin A, kanamycin B, tobramycin, and dibekacin. These compounds interfere with protein synthesis by targeting the ribosomal decoding A site, and they differ in the numbers and locations of amino and hydroxy groups of the glucopyranosyl moiety (ring I). We synthesized kanamycin analogues characterized by subtle variations of the 2′ and 6′ substituents of ring I. The functional activities of the kanamycins and the synthesized analogues were investigated (i) in cell-free translation assays on wild-type and mutant bacterial ribosomes to study drug-target interaction, (ii) in MIC assays to assess antibacterial activity, and (iii) in rabbit reticulocyte translation assays to determine activity on eukaryotic ribosomes. Position 2′ forms an intramolecular H bond with O5 of ring II, helping the relative orientations of the two rings with respect to each other. This bond becomes critical for drug activity when a 6′-OH substituent is present.

INTRODUCTION

Aminoglycoside antibiotics form a large family of water-soluble, polycationic pseudo-oligosaccharides. Common to all aminoglycosides is the neamine core. The neamine core is composed of a six-member aminocyclitol (2-deoxystreptamine; ring II) glycosidically linked to a glucosaminopyranose (ring I). Additional glycosyl substituents are attached at position 5 or 6 of the 2-deoxystreptamine moiety to give rise to a variety of pseudo-oligosaccharides categorized as 4,5- or 4,6-aminoglycoside antibiotics. An important subgroup of aminoglycosides that are used clinically as broad-spectrum antibacterial agents, i.e., tobramycin, kanamycin, amikacin, dibekacin, and arbekacin, are representatives of the kanamycin group of 4,6-disubstituted 2-deoxystreptamines (38). Aminoglycoside antibiotics promote misreading and inhibit the translocation of the tRNA-mRNA complex (8, 10, 11, 31). It is mainly the neamine core that mediates sequence-specific binding to the ribosomal decoding A site (9, 13, 39). The drug binding site is composed exclusively of nucleotides in helix 44 of bacterial 16S rRNA (29). Interaction with residues G1491 and A1408 (Escherichia coli numbering is used throughout the paper) in the narrow drug binding pocket is critical (23, 24, 32, 33), as ring I of the aminoglycosides becomes properly positioned for the compounds' antiribosomal activity by stacking interaction with G1491 and the formation of hydrogen bonds with A1408 (Fig. 1).

Fig 1.

Kanamycin A binding to the bacterial A site. (A) Kanamycin A complexed to the bacterial A site. The neamine core (rings I and II) is shown in yellow, and ring III is shown in gray; 16S rRNA residues are indicated. Note the stacking interaction of the compound's ring I with rRNA residue G1491 and the two hydrogen bonds between ring I and rRNA residue A1408 (N6′-N1 A1408 and O5′-N6 A1408; indicated by the orange dashed lines). The blue dotted lines indicate the hydrogen bonds between C1409 and G1491. (B) Description of contacts between kanamycin A and specific rRNA nucleotides. W, water. The views are based on the crystal structure described previously (13).

There are diverse aminoglycoside-modifying enzymes transferring acetyl, phosphoryl, and adenylyl groups in a cofactor-dependent manner to virtually every amino or hydroxyl substituent. The 2′- and 6′-NH₂ groups of ring I are acetylated by acetyl coenzyme A-dependent N-acetyltransferases (AAC), and the 3′- and 4′-OH groups of ring I are adenylylated by the action of O-adenylyltransferases (ANT) or phosphorylated by O-phosphotransferases (APH) (28, 42). In response to these resistance mechanisms, a large variety of natural and semisynthetic kanamycin derivatives have been synthesized (1, 3, 5, 12, 14, 18, 38).

Pioneering work on aminoglycosides was carried out in the mid-1970s (4, 40, 41), revealing that 2-deoxystreptamines with misreading activity on eukaryotic ribosomes share a 6′-hydroxy function in ring I as a common characteristic (40). Groundbreaking work by Benveniste and colleagues outlined key relevant structure-activity relationships among aminoglycoside antibiotics using both synthetic chemistry and enzymatic N-acetylation (4). Aminoglycosides are inactivated by N-acetylation because two effects occur in parallel: loss of a functional interaction, i.e., an important hydrogen bond provided by the charged amino group, and steric hindrance due to the introduction of the acetyl group. There is renewed interest in aminoglycosides (19), mainly as a consequence of two recent developments: (i) the availability of crystal structures revealing breathtaking details of drug-target interaction (9, 13, 39) and (ii) the advance in genetic manipulations allowing unprecedented precision in functional studies of drug-target interaction (2, 15, 36). Here, we investigated the contributions of subtle variations of ring I substituents to drug activity by focusing on kanamycin derivatives with different 2′, 3′, 4′, and 6′ substituents.

MATERIALS AND METHODS

Construction of mutant strains.

The single rRNA allelic Mycobacterium smegmatis ΔrrnB strain (SZ380) was obtained by unmarked deletion mutagenesis (21) and used for all genetic constructions; mutant strains 1408G, 1491A, and 1491C have been described previously (25). Residues 1408 and 1491 are 16S rRNA residues critical for aminoglycoside binding (24, 32, 33) and main determinants of acquired drug resistance (34, 36, 37); their phylogenetic polymorphism provides the basis for the compounds' selective mode of action (6, 20, 22).

Bacterial strains.

Clinical isolates of E. coli, Pseudomonas aeruginosa, and Staphylococcus aureus were obtained from the Diagnostic Department, Institute of Medical Microbiology, University of Zurich, Zurich, Switzerland. Recombinant E. coli strains with defined aminoglycoside resistance determinants (see Table S1 in the supplemental material) were kindly provided by P. Courvalin, Institut Pasteur, Paris, France.

MICs.

MICs were determined by broth microdilution assays in microtiter plates as described previously (23, 24).

Antibiotics.

Kanamycin A (catalog no. K4000), kanamycin B (catalog no. B5264), and tobramycin (catalog no. T1783) were obtained from Sigma; dibekacin was obtained from ANAWA Biomedical Products. 6′-Hydroxy-6′-deamino-kanamycin B (kanamycin C) (4) and 6′-hydroxy-6′-deamino-kanamycin A (6′-OH-kanamycin A) were synthesized as described in the supplemental material.

Isolation and purification of ribosomes.

Ribosomes were purified from bacterial cell pellets as described previously (7). In brief, ribosome particles were isolated by successive centrifugation and fractionated by sucrose gradient (10 to 40%) centrifugation. The 70S ribosome-enriched fraction was pelleted, resuspended in association buffer, incubated for 30 min at 4°C, dispensed into aliquots, and stored at −80°C following shock freezing in liquid nitrogen. Ribosome concentrations of 70S were determined by absorption measurements on the basis of 23 pmol ribosomes per A₂₆₀ unit. The integrity and functional activity of purified 70S ribosomes were determined by analytical ultracentrifugation and by assessing their capacities to form initiation complexes.

Cell-free luciferase translation assays.

Purified 70S hybrid ribosomes were used in translation reactions of luciferase mRNA. Firefly (F-luc) and Renilla (R-luc) luciferase mRNAs were produced using T7 RNA polymerase (Fermentas) in vitro on templates of modified plasmids pGL4.14 (firefly luciferase) and pGL4.75 (Renilla luciferase) (both Promega). In these plasmids the mammalian promoter driving transcription of luciferases was replaced by the T7 bacteriophage promoter. A typical translation reaction mixture with a total volume of 30 μl contained 0.25 μM 70S ribosomes, 4 μg F-luc mRNA, 40% (vol/vol) M. smegmatis S100 extract, 200 μM amino acid mixture, 24 units of RiboLock (Fermentas), 0.4 mg/ml of tRNAs, and energy was supplied by addition of 12 μl of commercial S30 Premix without amino acids (Promega). In addition to ribosomes, rabbit reticulocyte lysate (Promega) was used for in vitro translation of F-luc and R-luc mRNAs. A standard 30-μl reaction mixture contained 20 μl reticulocyte lysate, 4 μg F-luc and 0.4 μg R-luc mRNAs, amino acid mixture (200 μM each), and 24 units RiboLock (Fermentas). Following addition of serially diluted aminoglycosides, the reaction mixture was incubated at 37°C for 35 min, and the reaction was stopped on ice. Thirty-microliter samples of the reaction mixture were assayed for F-luc and R-luc luciferase activities using the Dual-Luciferase Reporter Assay System (Promega). Luminescence was measured using a lumino-meter (FLx800; Bio-Tek Instruments). The R-luc activity was used as an internal standard. Drug-mediated inhibition of protein synthesis is expressed as 50% inhibitory concentrations (IC₅₀s), i.e., the drug concentration that results in 50% inhibition of luciferase synthesis.

Misreading was assessed in a gain-of-function assay as described previously (16, 27). We introduced Arg245 (the CGC near-cognate codon) into the firefly luciferase protein to replace residue His245 (the CAC codon). Arg245 F-luc mRNA and wild-type (wt) F-luc mRNA were used in in vitro translation reactions in addition to R-luc mRNA as an internal control.

RESULTS

The kanamycins show close structural similarities; chemical differences are restricted to positions 2′, 3′, 4′, and 6′ of ring I (hydroxy, amino, or deoxy substituents; Table 1 summarizes the relevant chemical differences).

Table 1.

General structures of kanamycins and ring I substituents

^a R, residue.

^b The charge for rings II and III is 3 for all compounds.

Interactions with wild-type and mutant bacterial ribosomes.

We investigated drug-target interactions in cell-free translation assays using wild-type and mutant bacterial ribosomes. The mutant ribosomes were chosen to sample the relevant rRNA nucleotides interacting with ring I, i.e., A1408G, G1491A and G1491C (Fig. 1). Kanamycins with a 6′-NH₂ group show comparable activities on bacterial wt and mutant ribosomes. The 6′-NH₂ kanamycins are slightly affected by a purine replacement at 16S rRNA residue 1491 (G→A; relative difference from the wild type, 3- to 10-fold) but more affected by a purine-to-pyrimidine switch (G→C; relative difference from the wild type, 30- to 100-fold). In contrast, an A→G purine replacement at residue 1408 results in a >1,000-fold loss in activity for kanamycins with a 6′-NH₂ group (Table 2).

Table 2.

Activities against wild-type and mutant bacterial ribosomes

Drug	IC50 (μg/ml)
	Wt	G1491A	G1491C	A1408G
Kanamycin B	0.02	0.07	0.7	50.2
Kanamycin A	0.03	0.1	1.7	150.5
Tobramycin	0.01	0.1	1.0	43.2
Dibekacin	0.02	0.2	1.3	32.3
Kanamycin C	0.3	2.1	27.0	4.5
6′-OH-kanamycin A	12.5	73.5	>500.0	350.1

Replacement of the 6′-NH₂ by a 6′-OH group has a profound effect on drug activity. It decreases the activity of kanamycin B on wild-type ribosomes by 15-fold with corresponding losses in activity for both G1491A and G1491C mutant ribosomes (relative differences between 1491 mutant ribosomes and wild-type ribosomes, 7- and 90-fold, respectively). However, the activity for A1408G ribosomes is less affected (relative differences between A1408G mutant ribosomes and wild-type ribosomes, 15-fold for 6′-OH versus >1,000-fold for 6′-NH₂). A 6′-NH₂-to-6′-OH change virtually abolishes the activity of kanamycin A on wild-type bacterial ribosomes (>400-fold decrease) with corresponding losses of activity for G1491 mutant ribosomes (relative difference between G1491A mutant ribosomes and wild-type ribosomes, 6-fold, and G1491C mutant ribosomes, 40-fold). As with kanamycin C, a 6′-OH substituent in kanamycin A to a large extent retains the compound's relative activity for A1408G ribosomes compared to a 6′-NH₂ group (relative differences between A1408G mutant ribosomes and wild-type ribosomes, approximately 30-fold for the 6′-OH derivative versus >1,000-fold for 6′-NH₂ kanamycin A).

Antibacterial activity.

Kanamycin A, kanamycin B, tobramycin, and dibekacin show comparable antibacterial activities in MIC assays (see Table S1 in the supplemental material). As expected, a 3′-deoxy substituent (tobramycin and dibekacin) escapes modification by APH (3′), as does a 4′-deoxy substituent (dibekacin) by ANT (4′). The MIC activities of the various compounds for clinical isolates of E. coli, S. aureus, and P. aeruginosa are quite heterogeneous, reflecting the presence of various resistance determinants in these strains. In contrast to tobramycin and dibekacin, kanamycin A and B are rather ineffective against P. aeruginosa, due to the presence of APH (3′)-IIb (17). Corresponding to the IC₅₀s, kanamycin C and, in particular, 6′-OH-kanamycin A show poor antibacterial activity.

Activity on eukaryotic ribosomes.

To investigate the activities of kanamycins on higher eukaryotic ribosomes, we used rabbit reticulocyte ribosomes. We assessed drug-mediated inhibition of protein synthesis by determining IC₅₀ values in translation of luciferase mRNA (Fig. 2). Kanamycin B, kanamycin A, tobramycin, dibekacin, and kanamycin C showed similar levels of inhibition of eukaryotic protein synthesis. Replacing the 6′-amino substituent with a 6′-hydroxy substituent greatly diminished the activity of kanamycin A.

Fig 2.

Activities of kanamycins on rabbit reticulocyte ribosomes. (Top) Inhibition of luciferase mRNA translation and IC₅₀s (μM) (inset). Conc., concentration. (Bottom) Drug-induced misreading.

To assess aminoglycoside-induced misreading, we adopted a gain-of-function assay. Mutation of amino acid position 245 (CAC to CGC) in the active site of luciferase results in loss of enzymatic activity, with enzymatic function being restored by misreading (16, 27, 35). The kanamycins with a 6′-NH₂ substituent show little, if any, misreading, while the two with a 6′-OH substituent show pronounced misreading (Fig. 2).

DISCUSSION

We have previously used in vivo MIC assays of isogenic mutants to study the roles of individual contacts in aminoglycoside-rRNA interactions (23, 24, 32, 33). These studies have provided important insights (21). However, whole-cell in vivo MIC assays are subject to a number of nonribosomal variables, such as compound permeability, enzymatic drug modification, and off-target effects with an impact on test results that is difficult to predict. To avoid confounding nonribosomal effects and to study the molecular parameters of drug-target interaction more directly, here, we used an in vitro translation system with purified isolated ribosomes.

Replacing the 6′-NH₂ with an OH substituent in ring I of the neamine core significantly decreases the inhibitory activities of kanamycins on bacterial wild-type ribosomes (e.g., IC₅₀ of kanamycin B, 0.02 μg/ml, versus IC₅₀ of kanamycin C, 0.3 μg/ml) (Table 2). This is in contrast to the 4,5-aminoglycosides neomycin and paromomycin. Neomycin and paromomycin show a “neamine” core identical to those of kanamycin B and kanamycin C, respectively, i.e., a 6′-NH₂ group in neamine and kanamycin B versus a 6′-OH group in paromamine and kanamycin C. However, the replacement of the 6′-NH₂ substituent in neomycin with a 6′-OH group, resulting in paromomycin, has little effect on inhibitory activity on bacterial wt ribosomes (IC₅₀ of neomycin, 0.01 μg/ml, versus IC₅₀ of paromomycin, 0.02 μg/ml) (22). The different effects of a 6′-NH₂-to-6′-OH change for 4,6- versus 4,5-aminoglycosides cannot be sufficiently explained at the structural level, as the neamine core of all the aminoglycoside antibiotics binds to the small ribosomal subunit A site in a virtually identical manner (13). Apparently, molecular interactions outside the neamine core can influence small changes within the neamine core. As a crystal structure provides a snapshot rather than a dynamic interaction, we cannot exclude the possibility that the observed crystal structures of the aminoglycosides in complex with their ribosomal target may not reflect the functionally most relevant binding mode.

Our finding that replacement of the 6′-NH₂ with OH does not affect the inhibitory effects of kanamycin B on higher eukaryotic ribosomes contrasts with early studies that demonstrated that kanamycin C, but not kanamycin B, potently inhibits protein synthesis in wheat germ and in the lower eukaryote Tetrahymena thermophila (30, 40). However, both wheat germ and T. thermophila lack the characteristic C-A disruption of the 1409-1491 base pairing that is typical for higher eukaryotes and instead show a bacterial 1409C-1491G base pair interaction, resulting in a drug binding pocket similar to that of bacterial A1408G ribosomes (see Fig. S1 in the supplemental material; compound activities on bacterial A1408G ribosomes are shown in Table 2). In the case of A1408G ribosomes, a 6′-NH₂-to-OH change profoundly affects drug-ribosome interactions, readily explaining the previous observations in wheat germ and Tetrahymena. At the structural level and compared to A1408 wild-type ribosomes, a 6′-NH₂ interaction with A1408G mutant ribosomes is compromised both by loss of hydrogen bonding and by steric repulsion. In contrast, a 6′-OH group could still accept a hydrogen bond from N1 and/or N2 of 1408G and would not be subject to steric repulsion (26, 33). In higher eukaryotes, however, the antiribosomal activity of aminoglycosides is determined not only by residue 1408, but also by disruption of 1409-1491 base pairing and is thus the result of a combined effect (22).

The hydroxyl groups at positions 3′ and 4′ of ring I form hydrogen bonds to the phosphate groups of the two bulged adenine bases A1492 and A1493, and the 2′-amino group is involved in water-mediated hydrogen bonds to ring II and with the phosphate group of A1492 (13). Tobramycin (3′-deoxy), dibekacin (3′,4′-dideoxy), and kanamycin A (2′-hydroxy) show activities on wild-type and mutant bacterial ribosomes similar to that of kanamycin B (Table 2). Apparently, the contribution of hydrogen bonding from the substituents at position 2′, 3′, or 4′ to overall drug binding is limited as long as a 6′-NH₂ is present. In the presence of a 6′-OH substituent, however, the 2′ group (NH₂ versus OH) becomes critical for drug activity. Despite hydrogen bonds provided by the 4′-OH and 3′-OH substituents, the replacement of 6′-NH₂ by OH in kanamycin A virtually abolishes antiribosomal activity (compare 6′-OH-kanamycin A with kanamycin C). This surprising finding cannot be rationalized satisfactorily on the basis of contacts based on available X-ray crystallography data. Apparently, the intramolecular H bond between substituents C2′ and O5 of ring II, which supports the relative orientations of the two rings with respect to each other (13), becomes critical in the presence of a 6′-OH. In addition, 6′-OH-kanamycin A is the compound with the smallest charge, since ring I is neutral (Table 1). In the crystal structure complex, ring I is intercalated within RNA base pairs and in close contact with the negatively charged phosphate groups that are turned inward (and not outward as in regular helices). Other known intercalators (e.g., ethidium bromide and proflavine) are charged, and one can hypothesize that a positive charge contributes to the stabilization of an intercalating group within a nucleic acid helix.

The recent developments in genetic manipulation of ribosomal nucleic acids and the crystal structure analyses of aminoglycoside-ribosome complexes with resolution at the atomic level have provided important insights. However, we are still far away from being able to predict the effects of small structural changes on drug-target interactions, not least because subtle energetics that cannot be easily deduced from structures at the current resolution may play a critical role in function. As in any molecular complex, the strength of binding results from the free energy of complex formation, which is the sum of all types of interactions between the molecular groups constituting the binding site, including van der Waals contacts and charge-charge and charge-dipole interactions that are not necessarily apparent in a crystal structure. The network of interactions between the aminoglycoside and the rRNA A site is built upon key interactions, e.g., pseudo-base pair formation between A1408 and ring I, in particular the substituent at position 6′, and the constraints of a ring I-G1491 stacking interaction, most prominently for aminoglycosides with a 6′-OH (9, 13, 32, 33). All these energy contributions are hardly simply additive, and some may be synergistic. Thus, the intricate interplay between the various interactions and the resulting complexity renders design-driven chemical synthesis difficult. Combining chemical compound modifications with mutant ribosomes in functional studies of drug-ribosome interaction provides a wealth of information and will help to dissect the complex architecture of small-molecule–RNA interaction.