30S Subunit-Dependent Activation of the Sorangium cellulosum So ce56 Aminoglycoside Resistance-Conferring 16S rRNA Methyltransferase Kmr

Abstract

Methylation of bacterial 16S rRNA within the ribosomal decoding center confers exceptionally high resistance to aminoglycoside antibiotics. This resistance mechanism is exploited by aminoglycoside producers for self-protection while functionally equivalent methyltransferases have been acquired by human and animal pathogenic bacteria. Here, we report structural and functional analyses of the Sorangium cellulosum So ce56 aminoglycoside resistance-conferring methyltransferase Kmr. Our results demonstrate that Kmr is a 16S rRNA methyltransferase acting at residue A1408 to confer a canonical aminoglycoside resistance spectrum in Escherichia coli. Kmr possesses a class I methyltransferase core fold but with dramatic differences in the regions which augment this structure to confer substrate specificity in functionally related enzymes. Most strikingly, the region linking core β-strands 6 and 7, which forms part of the S-adenosyl-l-methionine (SAM) binding pocket and contributes to base flipping by the m¹A1408 methyltransferase NpmA, is disordered in Kmr, correlating with an exceptionally weak affinity for SAM. Kmr is unexpectedly insensitive to substitutions of residues critical for activity of other 16S rRNA (A1408) methyltransferases and also to the effects of by-product inhibition by S-adenosylhomocysteine (SAH). Collectively, our results indicate that adoption of a catalytically competent Kmr conformation and binding of the obligatory cosubstrate SAM must be induced by interaction with the 30S subunit substrate.

INTRODUCTION

Gene pools in soil and aquatic microbial communities represent largely unexplored reservoirs of antibiotic resistance (1, 2). An analysis of the resistance mechanisms present in these communities may reveal new insights into the origins, activities, and potential for mobilization of antibiotic resistance determinants to pathogenic bacterial populations. In particular, myxobacteria of the genus Sorangium have become a major source of secondary metabolites over the last several decades alongside the actinobacteria and fungi (3, 4). The model strain Sorangium cellulosum So ce56 possesses a 13.1-Mb genome with 17 gene clusters for secondary metabolites (5). Associated with this tremendous biosynthetic power is a high resistance potential against multiple antibiotics. For example, all Sorangium strains can grow in the presence of exceptionally high concentrations of kanamycin through the action of a single, specific aminoglycoside resistance-conferring 16S rRNA methyltransferase, Kmr (6).

In common with most aminoglycosides, kanamycin binds within helix (h) 44 of the 16S rRNA in the bacterial small (30S) ribosomal subunit and induces errors in mRNA decoding (7–9). Aminoglycoside-producing bacteria typically protect themselves from self-intoxication by expression of aminoglycoside resistance-conferring methyltransferases that modify the drug binding site. Two subfamilies of 16S rRNA methyltransferases introduce distinct 16S rRNA modifications at N7 of guanosine 1405 (m⁷G1405) or N1 of adenosine 1408 (m¹A1408), which confer high-level resistance to defined groups of aminoglycosides (10, 11). Aminoglycosides are polycationic compounds with a central aminocyclitol ring, most frequently 2-deoxystreptamine or streptamine, connected via glycosidic bonds to amino sugars, the distribution of which delineates three structural classes of drug. The 4,6-disubstituted 2-deoxystreptamines (4,6-DOS) include kanamycin and most clinically useful aminoglycosides such as gentamicin, tobramycin, and amikacin; the alternative 4,5-disubstituted 2-deoxystreptamines (4,5-DOS) include neomycin and paromomycin. A final group of compounds consists of those with alternative core rings or organization, such as apramycin, streptomycin, hygromycin B, and spectinomycin. The m⁷G1405 modification confers resistance to all 4,6-DOS aminoglycosides but not to other structural classes of aminoglycosides. In contrast, the m¹A1408 modification confers resistance exclusively to some members of each class of aminoglycoside, e.g., among the 4,6-DOS aminoglycosides to kanamycin but not to gentamicin.

The S. cellulosum So ce56 16S rRNA methyltransferase Kmr was found to confer a unique antibiotic resistance pattern in the sensitive bacterium Myxococcus xanthus, providing resistance against kanamycin, apramycin, gentamicin, neomycin, and tobramycin (6). However, Kmr failed to provide resistance to any of these antibiotics in Escherichia coli in further contrast to other methyltransferases, which are broadly active in diverse bacteria (6, 10, 12, 13). This distinct, extended resistance spectrum but apparent lack of function in E. coli prompted us to undertake a detailed study of Kmr in order to explore its potentially unique properties among the aminoglycoside resistance-conferring 16S rRNA methyltransferases.

In contrast to previous observations (6), using a kmr gene with codon usage optimized for expression in E. coli, we found Kmr to be fully active in E. coli with a canonical 16S rRNA (m¹A1408) methyltransferase antibiotic resistance spectrum, i.e., resistance to apramycin and kanamycin but not to gentamicin (10). Nonetheless, Kmr exhibits striking structural and mechanistic variations that point to potential fundamental differences in enzyme action on the 30S subunit among the 16S rRNA (m¹A1408) methyltransferases.

MATERIALS AND METHODS

Kmr expression and purification.

Kmr (GenBank accession no. ACB88605.1) was expressed in E. coli BL21(DE3) from a pET-44a plasmid containing an E. coli codon optimized gene obtained by chemical synthesis (GeneArt). An equivalent expression construct for KamB was described previously (14). Recombinant Kmr, NpmA, and KamB proteins were purified by heparin affinity and gel filtration chromatographies performed in 10 mM HEPES buffer (pH 7.5) and 100 mM NaCl for Kmr and as previously described for NpmA and KamB (14).

Kmr target nucleoside determination.

The site of modification by Kmr was determined by reverse transcription (RT) primer extension (15) of 16S rRNA from both in vitro and in vivo modified 30S subunits. For in vivo modified 16S rRNA, total rRNA was extracted from E. coli BL21(DE3) cells expressing recombinant Kmr as previously described (16). For in vitro modified 16S rRNA, the RNA was obtained by phenol extraction and ethanol precipitation from purified 30S subunits (E. coli MRE600) following an in vitro methylation reaction with recombinant Kmr.

For each RNA sample, 5′-end-³²P-labeled DNA oligonucleotide complementary to 16S rRNA nucleotides 1459 to 1479 was extended with avian myeloblastosis virus (AMV) reverse transcriptase (Promega). Sequencing lanes were produced using 16S rRNA purified from unmodified 30S subunits by inclusion of individual dideoxynucleotides in four separate reactions. Reverse transcription products were resolved on a 10% polyacrylamide urea denaturing gel and visualized on Typhoon Trio (GE Healthcare) with ImageQuant software.

Isothermal titration calorimetry.

Purified Kmr and KamB proteins (50 to 100 μM) were dialyzed exhaustively against 20 mM Tris buffer (pH 8.0) containing 150 mM NaCl and the final dialysis buffer used to prepare solutions of S-adenosyl-l-methionine (SAM), S-adenosylhomocysteine (SAH), and sinefungin (SFG) (1 to 2 mM). Titration experiments were performed in triplicate using an Auto-iTC₂₀₀ microcalorimeter (GE Healthcare) with 30 injections of SAM, SAH, or SFG into each protein. Data were fitted using Origin 7 software with a single binding site model to extract the binding affinity (dissociation constant [K_d]) for each protein-ligand pair.

Protein crystallization and structure determination.

Crystallization was performed by sitting-drop vapor diffusion at 4°C (apo Kmr with NaI) or 20°C (high-resolution apo form) with Kmr protein (10 mg/ml) in 20 mM HEPES buffer (pH 7.5) and 100 mM NaCl. Kmr apo form crystals were obtained in HEPES buffer (pH 7.5) containing 22% polyethylene glycol 2000 monomethylether and 200 mM potassium thiocyanate and cryoprotected by direct supplementation with 20% glycerol. Kmr-NaI crystals were grown in 100 mM morpholineethanesulfonic acid (MES) buffer (pH 6.1), containing 6% polyethylene glycol 5000 monomethylether, 8% 1-propanol, 0.5 mM SAM, and 100 mM NaI. No density was ultimately observed for SAM in the electron density maps.

X-ray diffraction data were collected on a Rigaku MicroMax 007 rotating anode generator in the Department of Biochemistry at Emory and at SER-CAT beam line ID-22 for the NaI-containing and high-resolution apo form data sets, respectively. Data were processed using HKL2000 (17). The high-resolution apo Kmr structure was determined by molecular replacement using the KamB structure (PDB accession no. 3MQ2) model in Phaser (18). The lower resolution apo Kmr structure from NaI-containing crystals was determined using single anomalous dispersion methods in the AutoSol module of Phenix (19). In both cases, two copies of Kmr were identified in the crystallographic asymmetric unit, and the final refined structures were produced by several rounds of manual rebuilding in Coot (20) and refinement in Phenix. Complete data collection, processing, and structure refinement statistics are provided in Table 3.

TABLE 3.

X-ray data collection and crystal structure determination statistics

Statistic	Results for:
	apo Kmr (PDB code 4RWZ)	apo Kmr (NaI) (PDB code 4RX1)
Unmodeled regions (residues)	185–205, 217–223 (chain A)	187–190, 218–223 (chain A)
	187–205, 215–223 (chain B)	187–202, 215–223 (chain B)
Space group	P3121	P3121
Resolution (Å)a	50–1.80 (1.86–1.80)	50–2.47 (2.56–2.47)
Cell dimensions
a, b, c (Å)	89.6, 89.6, 126.2	89.5, 89.5, 127.9
α, β, γ (°)	90, 90, 120	90, 90, 120
Wavelength (Å)	1.000	1.542
Rmergeb	0.08 (0.698)	0.098 (0.259)
I/σI	13.0 (2.5)	13.6 (12.7)
Completeness (%)	99.7 (97.5)	99.1 (90.7)
Redundancy	20.5 (10.2)	23.7 (19.9)
Figure of meritc		0.55
No. reflections (reflections used)	54,688 (51,910)	21,619 (20,514)
Rwork/Rfreed	0.20/0.23	0.21/0.25
No. of atoms	3,130	3,297
Protein	2,967	3,133
Water	163	153
B-factors
Protein	32.4	35.5
Water	36.1	32.7
Ramachandran plot
Favorable (%)	98.7	95.8
Outliers (%)	0	0
RMSe deviations
Bond lengths (Å)	0.017	0.011
Bond angles (°)	1.65	1.40

^aValues in parentheses are for the highest resolution shell.

^bR_merge = Σhkl Σi|Ii (hkl) − 〈I(hkl)〉|/Σhkl Σi Ii(hkl).

^cFigure of merit (FoM), m = cos(α − α_best).

^dR_work = Σhkl|F_o (hkl) − F_c (hkl)|/Σhkl|F_o (hkl), where F_o and F_c are observed and calculated structure factors, respectively. R_free applies to the 5% of reflections chosen at random to constitute the test set.

^eRMS, root mean square.

Homology modeling of Kmr was performed with the Swiss-Model server (21) using KamB (PDB accession no. 3MQ2, chain A) as the template. Modeled coordinates corresponding to residues of the disordered β6/7 linker (amino acids 183 to 208) were manually inserted into the high-resolution apo Kmr crystal structure. The resulting hybrid model was subjected to 10 cycles of energy minimization in Refmac (22) to reduce clashes between the modeled and experimentally determined structural components.

All figures showing crystallographic or modeled molecular structures were prepared using PyMOL (23).

Site-directed mutagenesis and Kmr functional assays.

Site-directed mutagenesis was performed by a whole-plasmid PCR approach using either the QuikChange Lightning kit (Agilent Technologies) or in a two-step protocol involving initial standard PCR to generate “megaprimers” for use in the second round of plasmid amplification (24).

The MICs of kanamycin, apramycin, and gentamicin for wild-type Kmr and of kanamycin for Kmr mutants were determined in liquid culture as previously described (14, 25). The MIC was defined as the lowest concentration of antibiotic resulting in no growth (culture A₆₀₀ of <0.05 above background).

In vitro competition methylation assays were performed by modification of a previously described approach (14, 26). The reaction mixtures contained 32 pmol of E. coli (MRE600) 30S subunits, 32 pmol of purified Kmr, KamB, or NpmA, 0.7 μCi (13.2 mCi/mmol) of S-adenosyl-l-[methyl-³H]-methionine ([³H]SAM; GE Healthcare), and methylation buffer (50 mM HEPES-KOH [pH 7.5], 10 mM MgCl₂, 50 mM NH₄Cl, and 5 mM β-mercaptoethanol) adjusted to a 40-μl final volume. Each reaction mixture was sampled at 0, 3, 10, and 40 min by removing 5 μl and adding it directly to 50 μl of ice-cold 5% trichloroacetic acid. Quenched samples were filtered under vacuum through GF/C filter discs (Whatman) and washed three times with 100 μl of ice-cold 5% trichloroacetic acid and twice with 100 μl of ethyl alcohol. ³H incorporation into 30S was quantified by liquid scintillation counting.

Protein structure accession numbers.

Crystallographic coordinates have been deposited in the RCSB PDB under accession numbers 4RWZ and 4RX1 for the apo and iodide-bound structures, respectively.

RESULTS

Kmr is a 16S rRNA (m¹A1408) methyltransferase active in E. coli.

The S. cellulosum So ce56 Kmr gene contains nine proline (CCC) codons that are rare in E. coli. To test whether poor heterologous expression was the cause of the apparent lack of Kmr activity in E. coli (6), we cloned a synthetic gene with E. coli codon optimization into a pET44a-derived vector that we have used previously for studies of other aminoglycoside resistance-conferring methyltransferases (14). This recombinant Kmr protein, with no additional tags at either terminus, was robustly expressed (data not shown) and provided high-level resistance to kanamycin and apramycin but not to gentamicin in E. coli BL21(DE3) (Table 1). This resistance phenotype corresponds to that of other 16S rRNA (m¹A1408) aminoglycoside resistance-conferring methyltransferases (Table 1) (10, 27) and suggests that Kmr likely also methylates 16S rRNA on A1408.

TABLE 1.

Aminoglycoside resistance profiles conferred by S. cellulosum So ce56 Kmr and the 16S rRNA (m¹A1408) methyltransferase KamB

Antibiotic	Antibiotic MIC (μg/ml) for:
	No plasmid	Empty pET44a	pET-Kmr	pET-KamB
Kanamycin	<5	<5	>1,000	>1,000
Gentamicin	<5	<5	20	10
Apramycin	15	15	>1,000	>1,000

Methylation at A1408 was confirmed by reverse transcription (RT) primer extension analysis of 16S rRNA modified both in vivo and in vitro by Kmr (Fig. 1). In both cases, a clear new stop in the RT reaction, absent in the unmodified control, is observed at position C1409, corresponding to modification by Kmr at A1408. Thus, in addition to the strains previously shown to be protected from the effects of aminoglycosides (6), our results demonstrate equivalent activity in E. coli and functionally confirm Kmr as an aminoglycoside resistance-conferring 16S rRNA (m¹A1408) methyltransferase.

FIG 1.

S. cellulosum So ce56 Kmr is an A1408 rRNA methyltransferase. (A) Structure of the 30S subunit, highlighting the location of A1408 (black sphere) and its sequence and secondary structural context (boxed, right) within helix 44 (h44). Additional 30S subunit features noted are h44 (semitransparent surface), head (h), body (b), platform (p), and spur (s). (B) A reverse transcription (RT) reaction demonstrating that A1408 is the modification target of Kmr. m¹A1408 modification produces a pronounced stop in the RT reaction at the preceding nucleotide (C1409). Lanes are unmodified 16S rRNA (lane 1), Kmr in vitro modified 30S (lane 2), and 30S from cells expressing Kmr (lane 3). The sequencing lanes (lanes C, T, A, and G) correspond to RT of unmodified 16S rRNA with reaction mixtures containing the complementary dideoxynucleotide.

Kmr has remarkably low affinity for its cosubstrate (SAM).

The pathogen-derived m¹A1408 methyltransferase NpmA binds its cosubstrate, S-adenosyl-l-methionine (SAM), approximately 30-fold more weakly than the reaction by-product S-adenosylhomocysteine (SAH) (Table 2) (28). Using isothermal titration calorimetry (ITC), we determined similar values for KamB, the orthologous enzyme from the tobramycin producer Streptoalloteichus tenebrarius (Fig. 2A and Table 2). Additionally, we measured KamB affinity for the methyltransferase inhibitor sinefungin (SFG) and found that it bound with an affinity intermediate between those for SAM and SAH (Table 2). Compared to NpmA and KamB, Kmr has a substantially reduced affinity for SAH (50- and 90-fold, respectively). Additionally, Kmr exhibits similar affinities for SAH and SFG, unlike KamB, where the former is bound ∼4.5-fold more tightly. Most remarkably, however, we found that Kmr has no measurable affinity for SAM (Fig. 2B and Table 2). These results suggest that the relative SAM/SAH affinities observed with other m¹A1408 methyltransferases, i.e., low SAM and high SAH affinity, are maintained in Kmr but with the result that the K_d for the Kmr-SAM interaction is reduced beyond the limit of detection in our experiments (more than mM range). Thus, despite its ability to provide high-level resistance in bacteria and efficient methylation of 30S subunits in vitro, Kmr in isolation is effectively unable to bind its essential cosubstrate at biologically relevant concentrations.

TABLE 2.

Kmr and other m¹A1408 methyltransferase-ligand affinities

Enzyme	Affinity, Kd (μM)
	SAM	SAH	SFG
Kmr	No binding	53	42
KamB	65	1.0	9.4
NpmAa	20	0.6	NDb

^aData for NpmA are from reference 28.

^bND, not determined.

FIG 2.

Kmr has no measurable affinity for its cosubstrate SAM. (A) Representative isothermal titration calorimetry (ITC) experiments for analysis of KamB binding to SAM (left) and SAH (right). Extracted binding affinities are shown in Table 2. (B) As for panel A but for Kmr. No binding is detected for Kmr with SAM.

X-ray crystal structures of Kmr.

To provide a structural framework for understanding the unusual properties of Kmr, we determined the 1.8-Å resolution X-ray crystal structure of apo Kmr by molecular replacement using the structure of KamB (PDB accession no. 3MQ2) as a search model (Table 3). Unambiguous electron density allowed modeling of all amino acids in both copies of Kmr in the crystallographic asymmetric unit with the exception of the region connecting β-strands 6 and 7 (“β6/7 linker”). As a result, amino acids 186 (in chain A) or 188 (in chain B) through 205 were excluded from the final model (Fig. 3A and B). In correlation with the weak affinity observed for SAM and SAH, extensive efforts to obtain a cocrystal structure of Kmr with either of these ligands were unsuccessful. However, these experiments produced an iodide derivative that allowed calculation of experimentally phased maps and a second structure of apo Kmr at 2.5 Å (Table 3). The structure built using these maps included part of the β6/7 linker (residues 191 to 205) in one chain (Fig. 3C) but was otherwise essentially the same as the higher resolution structure.

FIG 3.

X-ray crystal structures of S. cellulosum So ce56 Kmr. (A) High-resolution (1.8 Å) X-ray crystal structure of apo Kmr. The core β strands (1 to 7; blue) and surrounding α helices (green) of the conserved class I methyltransferase fold, and extended sequences that augment this structure within the A1408 methyltransferase subfamily (yellow) are highlighted. The short N-terminal extension forms a β-hairpin (β-N1/β-N2) and the β5/6 linker forms an extended unstructured loop. No electron density was observed for residues of the β6/7 linker, which are therefore not modeled. The approximate location of the SAM-binding pocket is also indicated (dashed box). (B) A second view of apo Kmr rotated 90° around the horizontal axis. The approximate expected location of the disordered β6/7 linker, based on the structures of KamB and NpmA, is indicated by a dashed line. (C) The structure of apo Kmr from NaI-derived crystals. The structure is shown in the same orientation as in panel B. The lower resolution (2.5 Å) maps derived from experimental phases allowed part of the β6/7 linker to be modeled with a single short helical structure.

As expected, Kmr possesses a class I SAM-dependent methyltransferase fold (29) with α-helices flanking a central seven-stranded β-sheet core with a characteristic 3↑2↑1↑·4↑5↑7↓6↑ topology (Fig. 3). Kmr and other m¹A1408 methyltransferases have three extended sequences that augment this core fold: a short extension at the N terminus and two larger insertions between the last three β-strands (β5/6 and β6/7 linkers). In common with the structures of KamB and NpmA (25, 28), the additional N-terminal sequence forms a β-hairpin structure that extends the central β-sheet core. In contrast, the β5/6 linker appears highly variable in structure between these m¹A1408 methyltransferases (Fig. 3 and 4). The Kmr β5/6 linker forms an extended, unstructured loop, while the equivalent region is much more compact but still largely unstructured in KamB and forms two distinct helical conformations in NpmA (25) (Fig. 4).

FIG 4.

Structural variation in 16S rRNA (m¹A1408) methyltransferase regions proposed to mediate interaction with the 30S subunit. Comparison of the extended β5/6 (top row) and β6/7 (bottom row) linkers in the known m¹A1408 methyltransferase structures Kmr (this study), KamB (PDB accession no. 3MQ2), and NpmA (PDB accession no. 3MTE). The β5/6 linker is the most variable in structure, adopting four distinct conformations, including the two observed for chain A (A) and chain B (B) in the crystal of NpmA. The β6/7 linker contains residues critical for enzyme activity and is essentially identical in KamB and NpmA. In Kmr, the loop adopts a unique structure with a partial (as shown for the NaI-derived crystal) or complete disorder of this region. Bound SAM is shown for KamB and NpmA as sticks.

The most striking difference between Kmr and the other m¹A1408 methyltransferases is observed in the β6/7 linker. In both NpmA and KamB, this extended sequence forms two short helices connected by a loop that caps the SAM-binding pocket (Fig. 4). In Kmr, this sequence is either partially or fully disordered in the NaI-derived and high-resolution apo crystal structures, respectively. In the former structure, a helical segment of the β6/7 linker is observed, but this single helix is oriented differently from NpmA and KamB (Fig. 4). While formation of the “closed” β6/7 linker conformation observed for KamB and NpmA is precluded for Kmr by crystal packing contacts, its disorder in the crystal, correlating with dramatically reduced SAM and SAH binding affinity, most likely reflects a highly dynamic nature of the apo Kmr β6/7 linker.

Modeling reveals potential for “closure” of the Kmr β6/7 linker.

The observation of a dynamic β6/7 linker was unexpected as structural insertions in this region of class I methyltransferases are often determinants of substrate specificity (29), and critical roles for several residues in the β6/7 linker have been proposed from studies of the m¹A1408 methyltransferases KamB and NpmA (25, 28, 30). To assess the potential for the Kmr β6/7 linker to form a closed structure similar to that exhibited by KamB and NpmA and thus position the functionally critical residues similarly, we used Swiss-Model to generate a homology model of Kmr. Using the KamB structure (PDB accession number 3MQ2) as a template, a complete homology model was generated and then the β6/7 linker model subsequently extracted and appended to our high-resolution Kmr structure. This chimeric high-resolution crystal structure-β6/7 linker model was subjected to 10 cycles of energy minimization. The resulting model (Fig. 5) demonstrates the potential closure of the Kmr β6/7 linker, repositioning of residues T192 and W194 to complete the SAM binding pocket and fulfill the latter residue's putative role in substrate positioning, as observed in the NpmA-30S complex (30).

FIG 5.

Model of the Kmr β6/7 linker structure and mutagenesis targets. The high-resolution structure of Kmr is shown (colored as in Fig. 3) appended with the homology model for the β6/7 linker (shown with sketch-style rendering). The cosubstrate SAM (spheres) is shown positioned by alignment with Kmr of the NpmA-SAM structure (PDB accession no. 3MTE). The conserved GxGxG motif that forms the base of the SAM-binding pocket is also highlighted. Residues mutated to assess their potential roles in Kmr activity are shown as sticks.

Kmr possesses a canonical SAM binding pocket but with reduced sensitivity to mutation.

Despite its remarkably low affinity for SAM and SAH, our structure and modeling show that Kmr retains all of the critical elements of the SAM binding pocket through its common class I methyltransferase core fold as well as those specific to the 16S rRNA (m¹A1408)-modifying enzymes within the β6/7 linker. The base of the SAM binding pocket in Kmr is formed by a conserved GxGxG motif (Kmr amino acids ³²GTGDG³⁶) which spans core strand β1 and the following loop (Fig. 5 and 6). Adjacent to this motif is a highly conserved aspartic acid residue (D30) that forms hydrogen bonds to the SAM amino group via two water molecules in cocrystal structures of other class I methyltransferases, including KamB and NpmA (25, 28). Kmr also possesses the highly conserved β2 aspartic acid (D55) that forms hydrogen bonds with both hydroxyl groups of the SAM ribose moiety. Using measurements of kanamycin MICs for E. coli expressing mutant proteins, we found that individual substitution of each conserved aspartic acid residue had a markedly different effect: while D30A effectively eliminated Kmr activity, the D55A substitution had no effect on the resistance conferred (Table 4). These findings are distinct from those for KamB, where both changes inactivated the enzyme (25), but comparable to those for NpmA, where D30A eliminated activity but D55A only partially reduced the MIC and in vitro methylation activity despite ablating SAM and SAH binding in vitro (28).

FIG 6.

Sequence and structural alignment of the aminoglycoside resistance-conferring 16S rRNA (m¹A1408) methyltransferases. The sequence alignment of Kmr, KamB, and NpmA, highlighting amino acid conservation (boxed residues) (absolute and functional conservation denoted by the black and white backgrounds, respectively), the conserved class I methyltransferase GXGXG motif (red), and the β5/6 and β6/7 linker regions (gold), is shown. Secondary structural assignments for Kmr and NpmA are shown above and below the sequences, respectively. Residues mutated to assess their role in SAM binding (orange), 16S rRNA interaction (purple), and A1408 positioning (green) are indicated; residues where a single substitution inactivates enzymatic activity in antibiotic MIC assays are shown in solid font, those with no or minimal impact are in outline font (also see Table 4).

TABLE 4.

Kanamycin MICs for Kmr and other A1408 methyltransferase mutants

Proposed function	Kmr		KamBa		NpmAb
	Substitution	MIC (μg/ml)	Substitution	MIC (μg/ml)	Substitution	MIC (μg/ml)
SAM binding	D30A	20–50	D30A	20	D30A	8
	D55A	1,000	D55A	10	D55A	512
	T192A	>1,000	T191A	10	S195A	1,024
16S rRNA binding			R195A	400	K199A	1,024
	H197A	>1,000	R196A	10	R200A	1,024
			R201A	10	R205A	1,024
	R204A	>1,000	R203A	800	R207A	16c
	R205A	>1,000
A1408 positioning/catalysis	W107F	1,000	W105F	10
	W107A	10	W105A	10	W107A	8
	W107A/F144A	20
	F144A	>1,000
	F144A/W194A	20–50
	W194A	400–800	W193A	10	W197A	8
	W194F	800	W193F	10

^aData for KamB are from reference 25.

^bData for NpmA are from reference 28.

^cMIC for NpmA-R207A is from this study.

Both NpmA and KamB additionally interact with the carboxylate group of SAM/SAH through residue T191/S195 (KamB/NpmA) located in the β6/7 linker. Our model of the Kmr β6/7 linker indicates that the equivalent residue, T192, might be repositioned to have similar interactions upon ordering of this loop (Fig. 5 and 6). However, a T192A substitution did not reduce the activity in the MIC assay (Table 4), again in contrast to KamB but in common with an equivalent substitution made in NpmA. We conclude from these published results and our observations with Kmr that defects in SAM cosubstrate binding must be compensated for during enzyme interaction with its 30S substrate.

Kmr is insensitive to substitution of additional conserved residues that are functionally critical in other 16S rRNA (m¹A1408) methyltransferases.

We next examined the effects of substitutions at residues around the putative Kmr active site, including the β6/7 linker. The KamB β6/7 linker contains four partially conserved arginine residues that are proposed to play roles in the 16S rRNA interaction; individual substitution with alanine at these residues produced effects from modest kanamycin MIC reduction to complete loss of enzyme activity (25). However, substitution of the equivalent or immediately adjacent residues, H197A, R204A and R205A, in the Kmr β6/7 linker failed to affect enzyme activity (Table 4). These findings mirrored the effects of equivalent changes in NpmA, except at R207 (Kmr R205) where substitution with alanine ablates activity. Our data thus suggest that Kmr lacks a residue equivalent to NpmA residue R207, which plays a critical role in A1408 base flipping by NpmA (30).

The Kmr β6/7 linker additionally contains one of two tryptophan residues that are universally conserved among known and putative 16S rRNA (m¹A1408) methyltransferases. As revealed by the NpmA-30S complex structure, these residues act in concert to sequester the target A1408 nucleoside in the enzyme active site. Substitution of either residue with phenylalanine in KamB or with alanine in KamB and NpmA, completely ablated enzyme activity (25, 28). In contrast, the equivalent substitutions of W107 and W194 in Kmr had an unexpected range of effects on enzyme function (Table 4). At W107, substitution with phenylalanine (W107F) produced a mutant enzyme with a kanamycin MIC indistinguishable from that of wild-type Kmr, whereas a W107A substitution completely ablated Kmr activity. Thus, Kmr W107 plays an equivalent role in A1408 positioning but, unlike in KamB, a phenylalanine is sufficient to fulfill the requirement of an aromatic side chain for stacking with the target adenine base. In contrast, substitution of the second tryptophan, W194, with either alanine or phenylalanine had a similar, only very modest, effect on Kmr activity (Table 4). We hypothesized that the reduced sensitivity to substitution of W194 in Kmr, a critical residue in other m¹A1408 subfamily members, might result from functional compensation by an additional adjacent aromatic residue. Due to the unique and extended structure of the β5/6 linker in Kmr compared to those of NpmA and KamB, residue F144 comes into close proximity to W107 and, in our modeled structure for the Kmr β6/7 linker, W194 (Fig. 5). While a single substitution of F144A had no impact on Kmr activity, double substitution with either W107A or W194A resulted in an inactive enzyme (Table 4). In the former case, this was the anticipated result as W107A alone ablates activity. However, the complete loss of activity in the F144A/W194A double mutant indicates that F144 may indeed compensate for functionality in the absence of an aromatic residue at 194 in the W194A Kmr mutant.

From these mutagenesis studies, we conclude that while Kmr retains the key elements of the SAM binding pocket and has the potential to form a canonical m¹A1408 methyltransferase structure through closure of the disordered β6/7 linker, it is almost entirely insensitive to changes in conserved residues that inactivate other members of this enzyme subfamily.

30S subunit-dependent methyltransferase activation.

The findings described thus far have demonstrated that while Kmr is an active 16S rRNA (m¹A1408) methyltransferase in vitro and in vivo, the enzyme has no measurable affinity for SAM and a region (the β6/7 linker) expected to be essential for multiple aspects of enzyme function displays an apparent structural disorder. These observations led us to ask whether Kmr interaction with the 30S subunit substrate might be the key event that regulates its enzymatic activity.

Using an in vitro methylation assay with only purified 30S subunits, Kmr and [³H]SAM, we found Kmr to be active and comparable to NpmA and KamB (Fig. 7A to C, compare black curves). We next tested the ability of Kmr and the other enzymes to methylate 30S subunits in in vitro competition methylation assays that included a range of concentrations of the reaction by-product (SAH) at 1- to 1,000-fold excess over SAM. Kmr methylation activity was largely unaffected at up to 10-fold excess SAH, and even at the highest concentration (1,000-fold excess SAH), measurable methylation activity was observed (Fig. 7A). In contrast, equivalent experiments with KamB demonstrated substantially stronger product inhibition of methylation: methylation activity was substantially reduced by only 10-fold excess SAH and was at near background levels in the presence of 100-fold excess SAH and above (Fig. 7B and D). NpmA showed an intermediate profile compared to those of Kmr and KamB, with an insensitivity to lower relative concentrations of SAH but with a greater reduction in activity at 100- and 1,000-fold excess SAH (Fig. 7C and D).

FIG 7.

30S in vitro competition methylation assays with Kmr, KamB, and NpmA. (A) Methylation of 30S by Kmr measured using tritium incorporation from [³H]SAM in the absence and presence of 1, 10, 100, and 1,000-fold excess of SAH (see color key). Kmr shows robust methylation of the 30S substrate even in the presence of a large excess of product. (B) As panel A, but for KamB in the presence of SAH. (C) As panel A, but for NpmA in the presence of SAH. (D) Comparison of relative methylation at the 10-min time point for each enzyme in the presence of excess SAH (solid lines, data from panels A to C) or the general methyltransferase inhibitor sinefungin (SFG; dashed lines, data from panels E and F). (E) As panel A, but for Kmr with SFG. (F) As panel A, but for KamB with SFG.

Competition methylation experiments were also performed using the general methyltransferase inhibitor SFG. In contrast to experiments with SAH, the activities of both Kmr and KamB were similarly affected: little effect on the activity of either enzyme was observed at 10-fold excess SFG, while 100- and 1,000-fold excess SFG caused strong and almost complete inhibition, respectively (Fig. 7E and F). Thus, for both Kmr and KamB, SFG is an equally effective competitive inhibitor, whereas Kmr but not KamB is unusually resistant to the effect of product inhibition by SAH at higher relative concentrations (shown by the gray shaded area in Fig. 7D).

Collectively, these results suggest that the 30S subunit plays a direct role in forming the Kmr SAM-binding pocket and additionally influences the binding affinity of SAM relative to that of SAH.

DISCUSSION

Bacterial resistance to aminoglycosides is widespread. While aminoglycoside-modifying enzymes remain the predominant resistance mechanism among human and animal bacterial pathogens, aminoglycoside resistance-conferring 16S rRNA methyltransferases are increasingly being identified in clinical and veterinary isolates (11, 31–33). The ease of their spread among pathogens within mobile genetic elements and the broad, high-level resistance conferred by these rRNA modification enzymes make them a profound threat to the future usefulness of aminoglycosides (34, 35). The recent reevaluation of potential applications of aminoglycosides in the clinic, particularly against multidrug-resistant organisms (36, 37), thus warrants a more complete understanding of the origins and modes of action of these emerging resistance determinants.

High-resolution structures of aminoglycoside resistance m⁷G1405 and m¹A1408 16S rRNA methyltransferases from aminoglycoside producers and pathogenic bacteria revealed the common structural scaffolds conserved within each methyltransferase subfamily (25, 28, 38, 39). However, the accompanying functional analyses of the 16S rRNA (m¹A1408) methyltransferases KamB and NpmA highlighted significant differences in the relative contributions of conserved residues to enzyme activity (25, 28). Such structural conservation but with an apparent lack of common function, even among highly conserved residues, potentially complicates inferences about m¹A1408 methyltransferase evolution, mechanisms of action, and acquisition routes in pathogens. To broaden our understanding of these resistance determinants, we undertook detailed structural and functional analyses of the methyltransferase Kmr from S. cellulosum So ce56, a Gram-negative bacterium evolutionarily distant from those harboring KamB or NpmA.

Our results demonstrate that Kmr possesses a class I SAM-dependent methyltransferase fold and is a bona fide aminoglycoside resistance-conferring enzyme acting at 16S rRNA nucleoside A1408 to confer a canonical aminoglycoside resistance spectrum in E. coli. However, the observation of β6/7 linker disorder in Kmr was unexpected as insertions between β-strands 6 and 7 in class I methyltransferases are often key determinants of substrate specificity (29). Kmr β6/7 linker disorder functionally manifests itself in a dramatically reduced affinity for both SAM and SAH compared to other 16S rRNA (m¹A1408) methyltransferases. With a single exception (27), the 16S rRNA (m¹A1408) methyltransferases bind the methylation reaction by-product SAH with higher affinity than their obligatory cosubstrate SAM (typically high nanomolar to low micromolar range). Kmr may therefore display an extreme example of the relative SAM/SAH affinities that are a characteristic, although not universal, hallmark of the 16S rRNA (m¹A1408) methyltransferases. For the majority of these enzymes, the relative cellular concentrations of SAM and SAH (10 to 100:1) should allow for a ready exchange of SAH with SAM to perform multiple rounds of 30S modification. In contrast, Kmr appears to be dependent upon its 30S substrate in order to form a structure capable of binding SAM.

Collectively, our results indicate that the Kmr interaction with the 30S subunit is the key event that drives enzyme activation, presumably by inducing a functional remodeling of the β6/7 linker and/or increasing cosubstrate affinity. While conserved residues in the SAM binding pocket (e.g., both D55 and T192) may have interactions with SAM identical to those for other m¹A1408 methyltransferases, Kmr appears to have dispensed entirely with their requirement to bind SAM. In addition to defining one end of the SAM binding pocket, in NpmA the β6/7 linker contains several residues critical for precisely positioning the target nucleotide: residues R207 and W197 directly stabilize the “flipped” conformation of A1408 and its sequestration in the enzyme active site, respectively (30). The KamB β6/7 linker residues R196 and R201 were found to be indispensable for its activity, indicating that one or both of these residues in concert may fulfill the role of NpmA R207. In contrast, our mutational analysis of Kmr at each equivalent residue and adjacent potential alternatives failed to identify any critical positively charged residue that might control base flipping of A1408. In Kmr, multiple residues may act in concert to promote the conformational changes necessary for catalysis in both substrate and enzyme, in a process strictly dependent upon binding to the 30S subunit. Interestingly, while this dependence is inherent in Kmr, it can also be induced by mutation in NpmA, suggesting that the underlying mechanism is common to both enzymes. In contrast, KamB has an absolute requirement for the functional interactions of these residues with SAM, irrespective of the presence of the 30S subunit substrate. Like KamB, the m¹A1408 methyltransferase CmnU from the Gram-positive capreomycin producer Saccharothrix mutabilis subsp. capreolus (40) is also highly sensitive to substitution of conserved residues important for interaction with SAM or the 30S subunit (P. M. Desai, S. M. Prezioso, N. Zelinskaya, and G. L. Conn, unpublished data). The possibility that different molecular mechanisms underpin 30S recognition and modification by Kmr/NpmA and KamB is further supported by the differences we identified in product inhibition by SAH for these enzymes. Finally, in addition to the potential for functional compensation we identified for residue F144, further unique roles for the unusually extended β5/6 linker structure in Kmr might also be anticipated since this region is the site of the most significant conformational change in NpmA upon 30S binding (30). However, complete understanding of the novel features of Kmr-substrate recognition, including the molecular basis for Kmr dependence upon 30S binding to form a catalytically productive complex with SAM, will require a structure of the 30S-Kmr complex.

Based upon these observations, it is tempting to speculate that there exists an intrinsic functional difference in m¹A1408 methyltransferases from Gram-negative and Gram-positive bacteria, further delineating this family of antibiotic resistance enzymes. An implication of this idea is the potentially polyphyletic origin of m¹A1408 methyltransferases, with at least two distinct groups residing in predominantly drug-producing Gram-positive and pathogenic or other Gram-negative bacteria. Alternatively, the molecular phenomena revealed by our studies of Kmr may reflect a broad spectrum of dependencies upon conserved residues within the m¹A1408 methyltransferase family and also their 30S substrate, which are built upon an otherwise common structural framework and mechanism of action. Understanding the common and unique features of the aminoglycoside resistance-conferring methyltransferase function has significant implications for potential future design of effective strategies to block the action of these enzymes.