Evidence That Antibiotics Bind to Human Mitochondrial Ribosomal RNA Has Implications for Aminoglycoside Toxicity

Background: Aminoglycosides are toxic to humans.

Results: Aminoglycosides bind to human mitochondrial and bacterial helix69 rRNA with equal affinity, but induce the greatest conformational stability in human mitochondrial helix69.

Conclusion: The binding to human mitochondrial helix69 may contribute to toxicity.

Significance: Understanding aminoglycoside binding to human mitochondrial helix69 is crucial for the development of nontoxic aminoglycosides.

Abstract

Aminoglycosides are a well known antibiotic family used to treat bacterial infections in humans and animals, but which can be toxic. By binding to the decoding site of helix44 of the small subunit RNA of the bacterial ribosome, the aminoglycoside antibiotics inhibit protein synthesis, cause misreading, or obstruct peptidyl-tRNA translocation. Although aminoglycosides bind helix69 of the bacterial large subunit RNA as well, little is known about their interaction with the homologous human helix69. To probe the role this binding event plays in toxicity, changes to thermal stability, base stacking, and conformation upon aminoglycoside binding to the human cytoplasmic helix69 were compared with those of the human mitochondrial and Escherichia coli helix69. Surprisingly, binding of gentamicin and kanamycin A to the chemically synthesized terminal hairpins of the human cytoplasmic, human mitochondrial, and E. coli helix69 revealed similar dissociation constants (1.3–1.7 and 4.0–5.4 μm, respectively). In addition, aminoglycoside binding enhanced conformational stability of the human mitochondrial helix69 by increasing base stacking. Proton one-dimensional and two-dimensional NMR suggested significant and specific conformational changes of human mitochondrial and E. coli helix69 upon aminoglycoside binding, as compared with human cytoplasmic helix69. The conformational changes and similar aminoglycoside binding affinities observed for human mitochondrial helix69 and E. coli helix69, as well as the increase in structural stability shown for the former, suggest that this binding event is important to understanding aminoglycoside toxicity.

Introduction

Aminoglycoside antibiotics are a class of clinically relevant antibiotics used to treat aerobic Gram-negative and some anaerobic bacilli bacterial infections. Aminoglycosides prevent bacterial protein synthesis by binding to prokaryotic ribosomes with greater affinity than to eukaryotic ribosomes (1, 2). They inhibit ribosomal function through an interaction with an internal loop in helix44 (H44)² of the small ribosomal subunit decoding site (3–5). There are several classes of aminoglycosides (6). The 2-deoxystreptamine (2-DOS) containing aminoglycosides include gentamicin, kanamycin, and neomycin classes (ring II in Fig. 1) (6). These aminoglycosides are derived from paromamine (6). The binding of paromamine-derived aminoglycosides to H44 decreases the mRNA translation fidelity and inhibits translocation (4, 7, 8). Other non-paromamine-derived aminoglycosides that contain atypical ring structures include hygromycin, apramycin, and spectinomycin. These aminoglycosides are frequently used in veterinarian medicine (9–11). Although the side chains of these antibiotics can vary widely (6), apramycin and hygromycin contain 2-DOS in common (ring I in Fig. 1, C and D). Both apramycin and hygromycin B have been shown to bind H44 and disrupt translocation of mRNA and tRNA on the ribosome (12, 13). Because of these mechanisms of action, aminoglycosides are effective in treating Gram-negative bacterial infections. However, aminoglycosides are also known to cause ototoxicity and nephrotoxicity in humans (14–17).

FIGURE 1.

Structures of aminoglycoside antibiotics. A–D, the chemical structures of paromamine-derived aminoglycosides gentamicin (A) and kanamycin A (B) and non-paromamine-derived aminoglycosides hygromycin B (C), and apramycin (D).

It has been established that aminoglycosides bind, in addition to H44, the terminal hairpin of helix69 (H69) of the bacterial large ribosomal subunit (see Fig. 2) (18), which plays a critical role in ribosomal recycling and translocation (19). The H69 nucleotides to which aminoglycosides bind are highly conserved in both sequence and secondary loop structure among bacterial phylogeny (20). The terminal stem-loop of H69 is positioned to be in direct contact with aminoacyl- and peptidyl-tRNAs (A-site and P-site tRNAs respectively) (18, 21) and interacts with the penultimate stem of H44 of the small subunit, forming an intersubunit bridge. In Escherichia coli, ribosomal recycling factors (RRFs) aid in the structural rearrangements of the E. coli ribosome that enable ribosomal recycling (22). However, aminoglycoside binding suppresses the required conformational changes induced by RRFs and, thus, negates ribosomal recycling (18).

FIGURE 2.

Ribosomal RNA sequences and secondary structures. A–C, native terminal hairpin constructs of E. coli H69 (A), human cytoplasmic hcyt69 (B), and human mitochondrial hmt69 rRNAs (C) from the large subunit. D and E, E. coli H44 of 16S rRNA (black) (D) and hmt44 of 12S rRNA (black) (E) connected by non-native tetraloops (gray).

Therapeutic use of aminoglycosides can produce side effects in which mitochondrial dysfunctions are observed in human and animal renal, auditory, and muscular disorders (23–30). However, the biochemical effects of aminoglycosides on human cytoplasmic and human mitochondrial H69 have as yet to be studied. Therefore, we explored and compared the effects of aminoglycoside binding to the human cytoplasmic (hcyt69) and mitochondrial H69 (hmt69) homologs to that of the E. coli H69. The aminoglycosides exhibited similar binding affinities to hmt69, hcyt69, and E. coli H69 constructs. Interestingly, aminoglycoside binding effected larger conformational changes of hmt69 residues than hcyt69 and enhanced the conformational stability of hmt69 to a greater extent than that of hcyt69 or E. coli H69. The conformational changes and similar aminoglycoside binding affinities observed for hmt69 and E. coli H69, as well as the increase in structural stability shown for the former, suggest that this binding event is important to understanding aminoglycoside toxicity.

Experimental Procedures

Sample Preparation

RNA sequences corresponding to the H69 and H44 species (see Fig. 2) were chemically synthesized (Dharmacon). Because the presence of pseudouridine-modified residues in H69 does not considerably affect aminoglycoside binding affinity, as well as the native structure or stability, the H69 constructs did not include pseudouridine modifications (31–35). The hcyt69 construct corresponds to nucleosides 3722–3740 of human 28S rRNA, whereas the hmt69 construct corresponds to nucleosides 2894–2911 of human mitochondrial 23S rRNA and the E. coli H69 construct corresponds to nucleosides 1906–1924 of E. coli 23S rRNA. For human mitochondrial H44 (hmt44) and E. coli H44 constructs, a hyperstable RNA tetraloop was inserted between the two strands (36–39). The hmt44 construct corresponds to nucleosides 1483–1500 and 1549–1569 of human mitochondrial 12S rRNA connected by a UUCG tetraloop (40). The E. coli H44 construct corresponds to nucleosides 1400–1412 and 1488–1502 of E. coli 16S rRNA (36) connected by the tetraloop GCAA. All RNA samples were deprotected per the manufacturer's instructions, extensively dialyzed in H₂O (18 megaohms), and analyzed for purity and integrity by denaturing polyacrylamide gel electrophoresis. The concentrations were determined by UV absorbance at 260 nm at room temperature. Apramycin (Sigma-Aldrich), gentamicin mixture (C1, C1a, C2, C2a) (AMRESCO), kanamycin A (Fisher Scientific), and hygromycin B (Sigma-Aldrich) were prepared in H₂O (18 megaohms).

UV-monitored Thermal Stability Analysis

UV-monitored thermal denaturation experiments were performed as described previously (31) with the following changes. UV samples contained 0–9 μm of apramycin, gentamicin, kanamycin A, and hygromycin B. The RNA constructs (see Fig. 2) hmt69, hcyt69, E. coli H69, hmt44, and E. coli H44 were at a concentration of 1 μm in 20 mm sodium phosphate buffer at pH 6.8. Magnesium (Mg²⁺) was not included in the buffer solution because binding of Mg²⁺ ions hydrolyzes RNA backbones at high temperatures. Additionally, the presence of MgCl₂ was not found to be necessary for stability of E. coli H69 and E. coli H44 rRNA (33). Thermal denaturations and renaturations were monitored by UV absorbance at 260 nm using a Cary 100 Bio UV-visible spectrophotometer (Agilent). Data were collected at a rate of 1 °C/min from 20 to 90 °C, averaged over six cycles, and performed in triplicate. The thermodynamic parameters were derived using MeltWin (41). The binding constant (K_d) was calculated by Prism (GraphPad) using the one-site total binding equation,

where y is the total binding, [x] is the aminoglycoside concentration, and NS is the nonspecific binding proportional to the concentration of aminoglycosides.

Circular Dichroism Spectroscopy

CD spectroscopy experiments were performed in triplicate with a JASCO J815 CD spectrophotometer. Concentrations of gentamicin and kanamycin A up to 40 μm were used to bind H69 (5 μm). hmt44 and E. coli H44 (1 μm) samples were titrated with kanamycin A from 0 to 8 μm. All CD samples were prepared in 20 mm sodium phosphate buffer at pH 6.8 The CD spectra were collected at room temperature from 190 to 320 nm at a rate of 100 nm/min with a resolution of 1 nm and were averaged over six runs in triplicate. Data were analyzed using Excel (Microsoft) and Prism (GraphPad). Normalization of spectra was performed by calculating molar ellipticity Δϵ (cm²/mmol) using the formula Δϵ = θ/(32,980 × C × L × N), where θ is the CD amplitude (millidegrees), C is the RNA concentration (mol/liter), L is the path length (cm), and N is the number of nucleotides in the RNA sequence (42).

NMR Spectroscopy

¹H one-dimensional and two-dimensional NMR spectra were collected at 278 K on a Bruker Avance III 500 MHz spectrometer equipped with an ultrasensitive triple resonance cryoprobe capable of applying pulsed field gradients along the z axis. The hmt69, hcyt69, and E. coli H69 samples (300 μm) were dissolved in 10% ²H₂O, 20 mm sodium phosphate, pH 6.8. Kanamycin A was directly added to the NMR samples to a final concentration of 1.5 mm. NOESY experiments with water suppression using excitation sculpting with gradients (43) and with 200 and 400 ms mixing times were acquired for assignment of the exchangeable protons. Data were processed using TopSpin 2.1 (Bruker) and NMRPipe (44) and analyzed with TopSpin 2.1 and SPARKY 3 (45).

Results

Thermal Characterization of H69 Hairpin Constructs

The aminoglycoside binding domain of the E. coli 23S rRNA H69 hairpin (nucleosides 1906–1924) was chemically synthesized, as were the homologous sequences for hcyt69 (nucleosides 3722–3740) and hmt69 (nucleosides 2894–2911). hmt69 and E. coli H69 constructs of native sequence include four canonical G-C base pairs in the same locations, but not orientation (Fig. 2). The hcyt69 construct includes three canonical G-C base pairs without a particular resemblance to the other two constructs. All H69 rRNAs have a wobble pair G·U located at the penultimate position of the stem (Fig. 2). hmt69 has a loop of six nucleosides, whereas the hcyt69 and E. coli H69 hairpins have loops of seven nucleosides each (Fig. 2). The three H69 hairpins behave as unimolecular species when thermally denatured and renatured under moderate ionic strength and pH (20 mm phosphate buffer, pH 6.8) (Fig. 3). The major melting transition from the helix to random coil occurred approximately between 40 and 55 °C for hcyt69, and between 48 and 65 °C for hmt69 and E. coli H69 (Fig. 3). The hmt69 and E. coli H69 demonstrated similar melting temperatures of 61 and 58 °C, respectively, whereas hcyt69 displayed the lowest T_m, 51.6 °C (Table 1). These results are expected considering that the hmt69 and E. coli H69 constructs included a greater number of G-C base pairs than the hcyt69 construct.

FIGURE 3.

Analysis of thermal stability of the H69 rRNA constructs. The denaturation and reannealing of hcyt69 (blue), hmt69 (red), and E. coli H69 (gray) hairpins were observed by absorbance at 260 nm from 5 to 90 °C in the absence of aminoglycosides. Each curve is the result of a point-by-point average of six transitions each performed in triplicate. The coefficient of variation is less than 6% for all data. The first derivative (dotted lines) of each curve shows each hairpin has one transition state.

TABLE 1.

Thermodynamic parameters of the rRNA helix69 constructs

Each thermodynamic property represents the average of three independent experiments. The coefficient variations of hmt69, hcyt69 and E. coli H69 are less than 5.15, 5.64, and 6.53% respectively for all data.

rRNA	Tm	ΔG	ΔH	ΔS
	°C	kcal/mol, 37 °C	kcal/mol	cal/(K × mol)
hmt69	58.4	−3.1	−47.9	−145
hcyt69	51.6	−2.12	−47.1	−145
E. coli H69	61.3	−4.47	−61.5	−184

Aminoglycoside Binding to hmt69, hcyt69, and E. coli H69

The ability of hmt69, hcyt69, and E. coli H69 to bind the paromamine-derived aminoglycosides, gentamicin and kanamycin A, was assessed as well as that of the non-paromamine-derived aminoglycoside, apramycin. Previous studies have shown that binding of aminoglycosides to the bacterial H44 RNA duplex increases the thermal stability of the RNA (46). Therefore, to assess aminoglycoside binding to H69, the thermal stability of the H69 constructs in the absence and in the presence of increasing concentrations of aminoglycoside was analyzed using repeated cycles of UV-monitored thermal denaturation and renaturation (Fig. 4A). The titration of hmt69, hcyt69, and E. coli H69 constructs with kanamycin A and gentamicin exhibited an increase in the T_m of the major thermal transition, indicative of aminoglycoside binding. Each binding curve derived from the change in T_m versus aminoglycoside concentration (Fig. 4, B–D) was fit to a nonlinear regression for a one-site total binding equation, and the equilibrium dissociation constant (K_d) was calculated. The resulting binding affinities of gentamicin to the hmt69, hcyt69, and E. coli H69 constructs had K_d values of 1.68, 1.69, and 1.31 μm, respectively (Table 2). The binding affinities of kanamycin A to the hmt69, hcyt69, and E. coli H69 constructs were 4.32, 5.35, and 4.02 μm, respectively (Table 3). The 3-fold greater binding affinity of gentamicin is reflective of the more rapid increase in ΔT_m during the titration as compared with that of kanamycin A (Fig. 4, B and C). Initially, we hypothesized that K_d values for E. coli H69 would be lower than the human H69 constructs because aminoglycoside antibiotics preferentially target bacterial ribosomes (2). However, the similar binding affinities to all H69 constructs may be due to aminoglycoside off-target binding and suggest a mechanism for the aminoglycoside toxicity displayed in humans.

FIGURE 4.

Aminoglycoside binding increases the thermal stability of H69. A, representative UV-monitored thermal denaturation/regeneration of hcyt69 (1 μm) titrated with kanamycin A (0–9 μm). The concentrations of kanamycin A (in μm) are indicated on the right. The RNA-kanamycin A complex was subjected to repeated thermal denaturations and renaturations. Each curve represents point-by-point averages of six total cycles of thermal denaturation and renaturation each performed in triplicate. B–D, the differences in melting temperatures (ΔT_m) of hmt69 (red), hcyt69 (blue), and E. coli H69 (gray) are plotted as a function of the concentration of gentamicin (B), kanamycin A (C), and apramycin (D). Data were fit to one-site nonlinear regression curves. Error bars represent standard error of the mean of three independent UV-thermal stability analysis experiments.

TABLE 2.

Binding constants of gentamicin to helix69 constructs and changes in thermodynamic properties due to binding

The given values indicate the average of the three independent experiments. The coefficient variations of hmt69, hcyt69, and E. coli H69 titrated with increasing concentrations of gentamicin (1–9 μm) are less than 5.66, 5.74, and 5.45%, respectively, for all data. ΔΔG°, ΔΔT_m, ΔΔH, and ΔΔS are derived from the difference between ΔG°, ΔT_m, ΔH, and ΔS in the absence of antibiotics and the presence of antibiotic.

Helix69 rRNA	Gentamicin
	hmt69	hcyt69	E. coli H69
Kd (μm)	1.68	1.69	1.31
ΔG° (kcal/mol, 37 °C)	−4.39	−3.79	−5.90
ΔΔG° (kcal/mol, 37 °C)	−1.29	−1.67	−1.43
ΔΔTm (°C)	5.27	6.80	4.70
ΔΔH (kcal/mol)	−7.34	−11.3	−7.37
ΔΔS (cal/K × mol)	−19.5	−31.1	−19.2

TABLE 3.

Binding constants of kanamycin A to helix69 constructs and changes in thermodynamic properties due to binding

The given values indicate the average of the three independent experiments. The coefficient variations of hmt69, hcyt69, and E. coli H69 titrated with increasing concentrations of kanamycin A (1–9 μm) are less than 5.01, 5.38, and 5.01%, respectively, for all data. ΔΔG°, ΔΔT_m, ΔΔH and ΔΔS are derived from the difference between ΔG°, ΔT_m, ΔH, ΔS in the absence of antibiotics and the presence of antibiotic.

Helix69 rRNA	Kanamycin A
	hmt69	hcyt69	E. coli H69
Kd (μm)	4.32	5.35	4.02
ΔG° (kcal/mol, 37 °C)	−4.35	−3.43	−5.97
ΔΔG° (kcal/mol, 37 °C)	−1.25	−1.31	−1.50
ΔΔTm (°C)	−4.42	4.83	4.92
ΔΔH (kcal/mol)	−8.58	−10.8	−7.62
ΔΔS (cal/K × mol)	−23.6	−30.7	−19.7

The toxic mechanism of aminoglycoside binding to hmt69 may be mediated by threading intercalation of the antibiotics. The binding site of threading intercalation of a stem-binding peptide has been shown to require C-G and G-C base pairs (the two base pairs are in different orientations) directly adjacent to G·U wobble base pair, followed by a loop (47). The C-G and G-C base pairs were proposed to be crucial in providing the site of intercalation (47). hmt69 meets the requirement of the threading intercalation, and, thus, it may be the binding mechanism of the aminoglycoside antibiotics (hmt69 G2895·U2910, C2896–G2909, G2897–C2908). However, these requirements are unfulfilled in E. coli H69 (E. coli H69 G1907·U1923, C1908–G1922, C1909–G1921).

Titration of the H69 constructs with the non-paromamine aminoglycoside, apramycin, caused virtually no change in the melting temperatures of the H69 RNAs (Fig. 4D), indicating that apramycin does not bind to H69. It has been established that apramycin binds in the deep groove of H44, which is composed of noncanonical C·A and G·A base pairs (12). Because H69 constructs do not include such noncanonical base pairs, our results were not unexpected, and further analysis with apramycin was not pursued.

Structural Effects of Aminoglycoside Binding to the H69 Constructs as Observed by Circular Dichroism

Surprisingly, investigations of the thermal stability of the aminoglycoside-RNA complexes indicated that gentamicin binds the hmt69, hcyt69, and E. coli H69 constructs with similar K_d values as kanamycin A. Therefore, we investigated and compared the effects of gentamicin and kanamycin A on the conformation of all H69 constructs using CD spectropolarimetry. Changes in CD ellipticity at 200–220 and 250–290 nm are indicative of conformational changes in the ribose-phosphate backbone and the base region of RNA, respectively (48). In the CD spectra, an increasing ellipticity at base regions of H69 indicated increased base stacking, which is highly correlated to the thermal stability established by aminoglycoside binding. The 2-DOS-containing paromamine-derived aminoglycosides, such as kanamycin A and gentamicin, have been known to form a pseudo-base pair or stack with bases of H44 (49). Also, these aminoglycosides interact with phosphate groups of H44 and stabilize the position of ring I (49). Therefore, the aminoglycoside binding would likely increase the ellipticity of both the phosphate backbone and the base region of H69.

Our CD results illustrated clear distinctions of the aminoglycoside-induced structural effects in the base and the phosphate backbone regions. The three H69 constructs displayed nearly equal increases in ellipticity in the phosphate backbone regions when titrated with increasing concentrations of gentamicin (Fig. 5). However, the greatest significant increase occurred with hmt69. This suggests that although the overall binding affinity of gentamicin to all H69 constructs appears to be similar, gentamicin binding enhances the conformational stability of hmt69 more than hcyt69 or E. coli H69. Upon kanamycin A binding, hmt69 displayed a considerably larger increase in ellipticity for both the base and the phosphate backbone region of the CD spectrum than did hcyt69 and E. coli H69 (Fig. 6). However, the overall ellipticity induced by kanamycin A binding was less than that of gentamicin for all H69 constructs (Figs. 5 and 6). Because kanamycin A binding to hmt69 displayed the largest increase in ellipticity for both the phosphate backbone and the base region, off-target binding to hmt69 might play a more prominent role in kanamycin A toxicity in humans. Together, these data suggest that the binding mechanism of kanamycin A and gentamicin increases the conformational stability of hmt69 and that the resultant conformational stability is greater with hmt69 than that of E. coli H69 and hcyt69.

FIGURE 5.

Gentamicin exerts differential structural effects on H69 constructs. A–C, CD spectra are the result of 5 μm of E. coli H69 (A), hmt69 (B), and hcyt69 titrated with gentamicin (0–40 μm) (C). The RNA-to-aminoglycoside molar ratios are indicated in panel A (bottom right). The dotted lines are aligned at 274 nm. Each curve represents a point-by-point of six repeats in triplicate. The standard error is less than 5 cm²/mmol for all data between 200 and 320 nm.

FIGURE 6.

Kanamycin A produces the greatest structural effects to the hmt69 construct. A–C, CD spectra are the result of 5 μm of E. coli H69 (A), hmt69 (B), and hcyt69 (C) titrated with kanamycin A (0–40 μm). The RNA-to-aminoglycoside molar ratios are indicated in panel A (bottom right). The dotted lines are aligned at 274 nm. Each curve represents a point-by-point average of six repeats in triplicate. The standard deviation is less than 4.7 cm²/mmol for all data between 200 and 320 nm.

Nucleosides Affected by Kanamycin A Binding as Observed by NMR

Our CD results indicated that kanamycin A and gentamicin binding increase the conformational stability of the hmt69 construct to a greater extent than that of hcyt69 or E. coli H69. Because kanamycin A binding to hmt69 indicated the largest increase in CD ellipticity for both the phosphate backbone and the base region, we further investigated kanamycin A binding to the hmt69, hcyt69, and E. coli H69 constructs using NMR. In this analysis, each RNA construct was titrated with kanamycin A to observe changes in the resonances of base-paired imino protons of the H69 stem. It is known that aminoglycosides bind the stem region, and not the loop of the H69 hairpin (18). These resonances are highly sensitive to changes in environment and dynamics. RNA base-paired imino resonances occupy a region of the NMR proton spectrum far downfield of other resonances exhibited by RNA or aminoglycosides, allowing them to be monitored unambiguously.

The base-paired imino resonances of the hcyt69, hmt69, and E. coli H69 constructs were assigned using ¹H one-dimensional and two-dimensional ¹H-¹H NOESY NMR experiments. In the one-dimensional ¹H NMR spectra of hcyt69 and hmt69, seven imino proton resonances were observed between 10.0 and 14.0 ppm for the expected six base pairs of the stems; the E. coli H69 construct displayed five resonances in that region (see Fig. 9, A and B, D and E, and G and H). The noncanonical G·U base pair has two imino protons (G3723·U3739, G2895·U2910 and G1907·U1924 for the hcyt69, hmt69, and E. coli H69 constructs, respectively; see Fig. 9, A, D, and G) with characteristic upfield shifts, from which the remaining imino proton peaks were sequentially and unambiguously assigned using the cross-peaks present in the same spectral region of the ¹H-¹H NOESY (see Fig. 9, C, F, and I). To distinguish between the G·U resonances, we noted that in other occurrences of G·U pairs in standard A-form RNA geometry, the distance between the G or U imino protons and the imino proton of the guanosine nucleoside in the adjacent base pair on the loop side (5′ to the G and 3′ to the U) is shorter between the cross-strand than the intrastrand imino protons (50). Correspondingly, the weaker of the two cross-peaks between the guanosine and the G·U imino protons was assigned to the resonance situated on the same strand as the adjacent guanosine. We also note that spin diffusion was observed at long mixing times in certain resonances in the E. coli H69 sample and may have affected apparent peak volumes, and therefore G·U assignments, in that sample or others. Neither of the weak resonances at 12.4 and 12.9 ppm for hcyt69, or at 12.4 and 13.3 ppm for hmt69, were observable in the ¹H-¹H NOESY spectra, a common occurrence for fast exchanging imino protons in terminal nucleosides at the ends of stems (50); they are nonspecifically attributed to G3722 and U3727 and to G2911 and U2899, respectively.

FIGURE 9.

Kanamycin A binding induces larger conformational changes to hmt69 than to hcyt69 and E. coli H69. A, D, and G, sequence and predicted secondary structure of hcyt69, hmt69, and E. coli H69, respectively. Nucleotides G and U that have base-paired imino protons are shown in black. Nucleotides C and A that do not have imino protons and the unpaired loop nucleotides that are not visible in this spectrum are shown in gray. B, E, and H, detection of the base-paired imino protons of hcyt69, hmt69 and E. coli H69 before (black) and after (blue, red, and gray) the addition of kanamycin A in a 1:5 RNA:aminoglycoside ratio. Peaks with an asterisk were too weak to unambiguously assign, but correspond to G3722/U3727 in hcty69 and U2899/G2911 in hmt69. C, F, and I, ¹H-¹H NOESY spectra of the imino regions of hcyt69, hmt69, and E. coli H69 displaying the sequential NOE connectivities between the imino protons that afforded unambiguous assignment of five of the seven base-paired protons in each construct.

Kanamycin A was added to each of the H69 NMR samples to a final concentration of five molar equivalents, as this is the molar ratio of aminoglycoside to RNA at which the maximum spectral changes occurred in both CD and thermal stability experiments. All peaks exhibited line broadening to various degrees. For all three constructs, a subset of chemical shifts showed significant deviations after kanamycin A titration. The resonances of G3723 and G3726 of hcyt69 shifted downfield by ∼0.1 ppm, whereas U3739, G3724, and U3737 showed slighter downfield chemical shift differences. In the hmt69 construct, the observed line broadening and chemical shift differences were even more significant, with the resonances for G2898 and G2897 shifting by ∼0.2 ppm downfield and those of G2909, U2910, and G2895 shifting by ∼0.1 ppm downfield. Similarly, for E. coli H69, the G1910 resonance shifted by ∼0.2 ppm downfield, and the G1907 and U1923 resonances shifted by ∼0.1 ppm downfield and upfield, respectively (Table 4).

TABLE 4.

Imino proton chemical shift changes of hcyt69 and hmt69 with kanamycin A

Nucleoside	Free H69 (δ1 , ppm)	Bound H69 (δ2 , ppm)	Δδ |(δ2-δ1)|
hcyt69
G3722/U3727	12.95	12.99	0.04
G3723	10.63	10.70	0.07
G3724	12.12	12.14	0.02
G3726	13.18	13.28	0.10
U3727/G3722	12.44	Not detected	Not detected
U3737	13.71	13.74	0.03
U3739	11.62	11.66	0.04
hmt69
G2895	11.04	11.08	0.04
G2897	12.05	12.25	0.20
G2898	13.03	13.19	0.16
U2899/G2894	13.27	13.31	0.04
G2909	12.73	12.81	0.08
U2910	11.38	11.50	0.12
G2911/U2899	12.44	12.51	0.07
E. coli H69
U1923	11.87	11.75	0.12
G1910	12.51	12.68	0.17
G1921	12.16	12.14	0.02
G1922	13.17	13.19	0.02
G1907	11.45	11.52	0.07

UV Thermal Analysis of Kanamycin A Binding to hmt44 and E. coli H44

To validate our approach in analyzing the binding of aminoglycosides to H69 constructs, we conducted UV-monitored thermal analyses of kanamycin A binding to the hmt44 and E. coli H44. Unlike H69 constructs and in contrast to a previously published study (46), hmt44 and E. coli H44 displayed two thermal transition states. Our H44 constructs were designed to be larger than those in other studies because of the propensity of aminoglycosides to bind stem regions. The lower temperature and smaller thermal transition are probably due to the presence of an additional internal loop in the constructs used in this study (Fig. 2). In titrating the RNA with increasing concentrations of kanamycin A, the T_m of the E. coli H44 construct increased, suggesting kanamycin A binding to E. coli H44. The binding of kanamycin A to the bacterial H44 and hmt44 exhibited K_d values of 2.46 ± 2.89 and 2.99 ± 1.92 μm, respectively. This compares favorably to a binding study using fluorescence anisotropy that demonstrated that the binding affinity of kanamycin B to bacterial H44 was K_d = 2.14 ± 0.29 μm (51). With the binding of concentrations greater than 3 μm kanamycin A, the E. coli H44 construct no longer exhibited the low temperature transition, and thus, reflected the one-site binding (see Fig. 8) observed by others (46). The T_m of hmt44 also increased due to kanamycin A binding, yet the increase in T_m was marginal as compared with E. coli H44. These results validate our approach of utilizing UV analysis to understand aminoglycoside binding and further indicate that aminoglycosides may prefer to bind to bacterial H44 than hmt44.

FIGURE 8.

Kanamycin A binding increases the thermal stability of E. coli H44 and hmt44. A and B, the thermal stabilities of kanamycin A (0–9 μm) bound 1 μm of E. coli H44 (A), and hmt44 (B) hairpins were observed through their UV absorbance at 260 nm and their denaturation and reannealing (from 5 to 90 °C). The concentrations of kanamycin A (in μm) are indicated on the right in panels A and B. Each curve is the result of a point-by-point average of six transitions performed in triplicate. The coefficient of variation is less than 7.6% for E. coli H44 and 6% for hmt44 data. C and D, the first derivatives of the E. coli H44 (C) and hmt44 (D) curves show that the constructs have two transition states.

Structural Effects of Kanamycin A Binding to hmt44 and E. coli H44 Observed by CD Analysis

Given the understanding that the prokaryotic H44 is the primary target of aminoglycosides, we investigated the conformational effects of kanamycin A binding to E. coli H44 and hmt44 using CD, as we did for H69. CD spectral results determined that kanamycin A binding increased the ellipticity of E. coli H44 more than that of hmt44 in both the phosphate backbone and the base regions of the CD spectra (Fig. 7). Together with the UV-monitored thermal stability measurements (Fig. 8), this suggests that kanamycin A binding produces a conformational stability in the E. coli H44. This enhanced structural stability generated by aminoglycoside binding was similarly observed with hmt69. CD spectral analysis with H44 constructs confirmed that one of the salient effects of aminoglycoside binding to E. coli H44 is an increase of conformational stability mediated by increased base stacking.

FIGURE 7.

Kanamycin A produces greater structural changes in the E. coli H44 construct consistent with stabilization. A and B, CD spectra of 1 μm of E. coli H44 (A) and hmt44 (B) are graphed as a function of the molar ratio of RNA to kanamycin A concentration (0–8 μm), excluding 1:0.4. The RNA-to-aminoglycoside molar ratios are indicated in panel A (bottom right). The dotted lines are aligned at 274 nm. Each curve represents a two-point-by-two-point average of six repeats in triplicate. The standard deviation is less than 5.5 cm²/mmol for all data between 200 and 320 nm.

Discussion

The main target of aminoglycosides is the prokaryotic H44 rRNA of the small subunit (1, 2). H69 rRNA of the E. coli large subunit was identified as a secondary binding site (18). We hypothesized that the off-target binding to hcyt69 and hmt69 may be important for understanding toxic side effects of paromamine-derived aminoglycosides. UV analysis indicated that aminoglycoside binding affinities to hmt69 and E. coli H69 were similar. NMR analysis also demonstrated that the most significantly perturbed chemical shifts of hmt69 upon the binding occurred in the same respective regions in E. coli H69 and to a comparable degree. CD analysis demonstrated that among H69 constructs, hmt69 exhibited the greatest conformational stability upon the binding. These results support the concept that hmt69 binding of aminoglycoside antibiotics may be important to understand toxicity in humans.

As mitochondria display similarities with bacteria in their protein expression machinery (52, 53), we hypothesized that the mechanism of aminoglycoside binding to the mitochondrial and bacterial ribosomes may be similar. Components of the mitochondrial translational machinery resemble those employed by bacteria, but are distinct from their cytoplasmic counterparts (54). For example, an RRF homolog is present in eukaryotic mitochondria, but absent in the cytoplasm (55–57). Bacterial RRF aids structural arrangements of H69 to enable ribosomal recycling (18); the absence of RRF is lethal (58, 59). In E. coli H69, gentamicin has been shown to bind the H69-RRF complex and suppress structural rearrangement, resulting in ribosomal recycling inhibition (18). The bacterial RRF has a highly conserved region that is known to interact with the H69 loop (nucleosides 1915 and 1916) and the stem (nucleosides 1908–1910) of E. coli H69 (22, 60). NMR results presented here revealed that aminoglycosides interact with the H69 stem nucleosides 2898 and 2897 of hmt69, which correspond to nucleosides 1909 and 1910 of E. coli H69 (Fig. 9). Therefore, the RRF binding site would likely overlap with the aminoglycoside binding site in mitochondria and impair the effects of RRF. This similarity exhibited by hmt69 and E. coli H69 may indicate that the aminoglycoside off-target binding to hmt69 may also inhibit ribosomal recycling and contribute to toxicity in humans.

The structure and components of the large subunit of human mitochondrial ribosomes, however, also display unique characteristics that differ from those of the bacterial or the cytoplasmic ribosomes (61). Although the protein composition of the large subunit of human mitochondrial ribosomes is substantially increased, the length of human mitochondrial rRNA in the large subunit is half that of the bacterial ribosomes (61). Human mitochondrial tRNA also has reduced or completely absent canonical D- and/or T-loops that form the tRNA elbow (61). In the P-site of the human mitochondrial large subunit, additional elbow-stabilizing elements such as uL5, possibly the homolog of the bacterial protein L5, and helix84 (H84) are deleted (61). However, in the P-site of the bacterial large subunit, the minor groove of the tRNA D stem rests on the minor groove of bacterial H69 (62). Deletions of rRNA domains observed in the human mitochondrial ribosome large subunit may indicate that the large subunit could be more susceptible to aminoglycosides, indirectly affecting misreading and thus causing toxicity; alternatively, fewer binding sites could be protective to mitochondrial translation.

The similarity of binding affinities of aminoglycosides to hmt69 and bacterial H69 with comparable secondary and tertiary structures (Figs. 2 and 10) supports the identification of hmt69 as a likely off-target binding site of aminoglycosides. However, the mechanism of aminoglycoside binding to hmt69 may differ from that of E. coli H69. Aminoglycoside binding to hmt69 resulted in higher conformational stability as suggested by increased base stacking. A peptide that binds to RNA stems by threading intercalation recognized different orientations of two C-G base pairs adjacent to a G·U base pair, and the binding was accompanied by a higher binding affinity (47). hmt69 meets the required C-G base pair orientation for threading intercalation. However, E. coli H69 does not meet the requirement. The crystal structure of E. coli H69 in the presence of gentamicin demonstrated that threading intercalation does not occur when binding to E. coli H69 (18). Because UV analysis indicated similar antibiotic binding affinities for hmt69 and E. coli H69 under equilibrium conditions, an investigation of aminoglycoside on- and off-rates may provide further insights into the possible threading intercalation to hmt69. NMR results showed that one of the foremost chemical shift perturbations in hmt69 by aminoglycoside binding occurs with a guanosine residue, G2897, a residue from the required C-G base pairs for threading intercalation (Table 4). However, NMR analysis also demonstrated that the chemical shift perturbation of residues from the necessary C-G base pairs is not observed in E. coli H69 (Table 4). This suggests that the binding of aminoglycosides to hmt69 may be mediated by threading intercalation, and thus, differ from that of E. coli H69. The binding mechanism to hmt69 may be important to our understanding the toxicity of aminoglycosides.

FIGURE 10.

Superimposition of the stems of E. coli H69 and hmt69. A, a superposition of the stems of the hmt69 from the single-particle cryo-electron microscopy structure (3.4 Å resolution based on the gold-standard Fourier shell correlation = 0.143 criterion, red) (61) and the E. coli H69 from the x-ray crystal structure (3.5 Å resolution, green) (72) in a licorice representation. The residues of the stem region of hmt69 are indicated. B, an enlarged view of the highlighted region in A in a ribbon representation.

The majority of adverse effects reported during treatment with gentamicin and kanamycin A, both paromamine-derived aminoglycosides, are similar to the symptoms exhibited in mitochondrial dysfunctions and include renal, auditory, and muscular disorders in both humans and animals (23–29). More acute symptoms are associated with the treatment with gentamicin (27, 30, 63–68). It is possible that aminoglycoside-induced mitochondrial dysfunctions in humans may arise from off-target binding to human mitochondrial RNA, such as hmt69. UV observation of aminoglycoside binding to H69 indicated that gentamicin displays a higher binding affinity to H69 constructs than kanamycin A. Based on our CD results, it appears that gentamicin binding to H69 constructs also perturbs the RNA conformation more than kanamycin A. The more severe side effects observed with gentamicin therapy may be associated with the higher binding affinity and nonspecific structural effects observed when binding to H69 of the human mitochondrial and cytoplasmic ribosomes. To further understand the mechanism of aminoglycoside toxicity, it would be interesting to investigate aminoglycoside binding to hmt69 in vivo and observe the resultant toxic effects to mitochondria in cells.

Although hmt69 appears to be the target of clinically relevant paromamine-derived aminoglycosides, our results indicate that apramycin and hygromycin B, non-paromamine aminoglycosides, do not bind hmt69 or E. coli H69 (data not shown). Hygromycin B binding to H69 was not observed, although previous studies demonstrated its affinity for H44 of the small subunit of prokaryotic ribosomes (4, 69). Although only hygromycin B and apramycin were tested, these results indicate that non-paromamine aminoglycosides do not bind hmt69 or E. coli H69. However, it would be interesting to determine whether any aminoglycosides, especially non-paromamine-derived aminoglycosides, bind to H69 of non-human mammals, because they are typically used in veterinary medicine (10, 70).

Extensive efforts have been made to understand aminoglycoside selectivity for the bacterial ribosome. It is established that rings I and II of the paromamine-derived aminoglycosides are important aminoglycoside domains for binding to bacterial rRNAs (71). Ring I forms a pseudo-base pair with nucleoside A1408 of H44 of the bacterial small subunit 16S rRNA (47), and ring II provides an additional anchor for some aminoglycosides (71). Modification of ring I has been found to decrease off-site binding to human cytoplasmic and mitochondrial ribosomes without reducing the antibacterial activity against the bacterial ribosome (49). The selectivity of paromamine-derived aminoglycosides depends on the structural context of nucleosides 1408 (A for bacterial and mitochondrial and G for cytoplasmic) and 1491 (G for bacterial, C for mitochondrial, and A for cytoplasmic) (49).

One of the most prevalent toxic side effects of aminoglycosides, ototoxicity, is linked to A1555G and C1949U mutations of eukaryotic mitochondrial ribosomes (49). NMR results indicated that analogous residues of E. coli H69 and hmt69 experienced the most significant chemical shift perturbation upon aminoglycoside binding with one important exception. The one exception, G2897 of hmt69, also displayed a comparable chemical shift perturbation upon binding, yet the respective residue of bacterial H69 was not affected by aminoglycoside binding. Presently, the role of G2897 of hmt69 in the binding of antibiotic has yet to be revealed.

The data presented here provide a greater understanding about the interaction of aminoglycosides with the E. coli and human cytoplasmic and mitochondrial H69 domains of the ribosome and offer further insight for the development of novel aminoglycoside antibiotics that lack toxicity to humans without compromising antibacterial activity. We propose that off-target aminoglycoside binding to hmt69 results in mitochondrial dysfunctions and contributes to side effects displayed in human patients receiving aminoglycoside antibiotic therapy. The conformational changes and similar binding affinities demonstrated for hmt69 and E. coli H69, in addition to the increased structural stability of hmt69 upon binding, support the hypothesis that aminoglycoside binding to hmt69 is important to our understanding of aminoglycoside toxicity.

Author Contributions

S. H. designed the study, performed the experimental research, conducted analysis of the data, and wrote the paper. K. M. F. and P. F. A. designed the study, conducted analysis of the data, and wrote the paper. K. D. F. performed the experimental research and conducted analysis of the data. K. A. H. and K. L. S. performed the experimental research, conducted analysis of the data, and wrote the paper.

Note Added in Proof

Fig. 5A and Fig. 5B were reversed in the version of this article that was published on June 9, 2015 as a Paper in Press. This error has been corrected. The change does not affect the interpretation of the results or the conclusions.