Molecular mechanism of viomycin inhibition of peptide elongation in bacteria

Significance

Antibiotics are widely used to treat bacterial infections, but their mechanisms of action are often poorly understood. Viomycin is a tuberactinomycin antibiotic used for treating multidrug-resistant tuberculosis. Using an in vitro translation system that displays kinetic rates comparable to those in living cells we have characterized the mechanism of action of viomycin and constructed a kinetic model for how viomycin inhibits the translocation step of the peptide elongation cycle. Our results are vital for understanding the mechanism of the antimicrobial activity of viomycin and its sister drugs in living bacteria as well as resistance mechanisms against them, which can have strong implications for global health.

Abstract

Viomycin is a tuberactinomycin antibiotic essential for treating multidrug-resistant tuberculosis. It inhibits bacterial protein synthesis by blocking elongation factor G (EF-G) catalyzed translocation of messenger RNA on the ribosome. Here we have clarified the molecular aspects of viomycin inhibition of the elongating ribosome using pre-steady-state kinetics. We found that the probability of ribosome inhibition by viomycin depends on competition between viomycin and EF-G for binding to the pretranslocation ribosome, and that stable viomycin binding requires an A-site bound tRNA. Once bound, viomycin stalls the ribosome in a pretranslocation state for a minimum of ∼45 s. This stalling time increases linearly with viomycin concentration. Viomycin inhibition also promotes futile cycles of GTP hydrolysis by EF-G. Finally, we have constructed a kinetic model for viomycin inhibition of EF-G catalyzed translocation, allowing for testable predictions of tuberactinomycin action in vivo and facilitating in-depth understanding of resistance development against this important class of antibiotics.

Tuberculosis (TB) is a global threat to human health, and over 9 million people contracted the disease in 2013 (1). The emergence of strains of Mycobacterium tuberculosis, the causative agent of TB, resistant to several first- and second-line antitubercular drugs is a significant concern. The tuberactinomycin antibiotics are bacterial protein synthesis inhibitors, commonly used as second-line drugs against multidrug-resistant TB. Viomycin was the first drug of this class to be discovered (2, 3). It is a cyclic pentapeptide that contains several nonstandard amino acids (Fig. 1A) and is produced by a nonribosomal peptidyl transferase (4).

Fig. 1.

(A) Chemical structure of viomycin. (B) Overview of the bacterial ribosome with a bound viomycin molecule (red) between the two ribosomal subunits (50S in light gray and 30S in dark gray) next to the A-site tRNA (green); the P-site tRNA (orange) is also visible. The figure is based on structural data from ref. 18, and the PDB ID code is 4V7L. (C) Detailed view of the viomycin binding site showing the flipped-out state of the bases A1492 and A1493. Colors are as in B and the mRNA is visible in purple. (D) Time course of f[³H]Met-Phe-Thr tripeptide formation at varying concentrations of viomycin and 5 µM EF-G. The fast phase is due to tripeptide formation by viomycin-free ribosomes, which escape inhibition. The decrease in the amplitude of this phase with increasing viomycin concentration reflects the increasing fraction of viomycin-bound ribosomes. The overall amplitude decrease with increasing viomycin concentration is due to read through of the stop codon present after the Met-Phe-Thr reading frame. Solid lines represent fits of Eq. S1 to the data. (E) The fraction of ribosomes inhibited by viomycin at different EF-G concentrations as estimated by subtraction of the amplitude of the fast phase with viomycin from that without it, plotted as a function of the viomycin concentration. Solid lines represent fits of Eq. 2 to the data. (Inset) The same data but in the concentration range from 0–10 µM viomycin. (F) Fraction of viomycin-inhibited ribosomes at 10 µM EF-G with and without preincubation of the 70S ribosome with the drug at two viomycin concentrations. All error bars represent SEM.

Viomycin affects bacterial protein synthesis by inhibiting mRNA translocation (5) and causing misreading of the genetic code (6). The present study focuses on viomycin inhibition of translocation. During translocation the mRNA moves through the ribosome by one base triplet (codon), the peptidyl transfer RNA (tRNA) moves from the ribosomal A to P site, and the deacylated tRNA moves from the P to E site. Translocation is catalyzed by elongation factor G (EF-G) in a GTP hydrolysis-dependent manner (7) and occurs via formation of tRNA hybrid states (8), relative rotation of the ribosomal subunits (9), and movement of the L1 stalk (10). Bulk FRET, chemical footprinting, and single-molecule FRET experiments have shown that viomycin stabilizes the ribosome in a pretranslocation state with tRNAs in hybrid A/P and P/E configurations, rotated ribosomal subunits, and the L1 stalk in a closed conformation interacting with the P-site tRNA (11–16); this structure has been visualized by cryo-EM (17). However, a crystal structure of the viomycin-bound ribosome (18) and one single-molecule FRET study (19) have suggested stabilization of the tRNAs in the classical state.

Recent crystal (18, 20) and cryo-EM (17) structures of the viomycin-bound ribosome have shown that the drug binds adjacent to the ribosomal A site, in a pocket between helix 44 of the 16S rRNA in the small subunit and helix 69 of the 23S rRNA in the large subunit (Fig. 1 B and C). In these structures the monitoring bases A1492 and A1493 are in their active, flipped-out conformation, where they interact with the codon–anticodon helix in the A site (21, 22). In fact, the bound viomycin molecule sterically blocks the flipped-in conformation of these bases observed in the apo ribosome (23). This steric restriction of the monitoring bases may explain the error-inducing activity of viomycin in genetic code translation (6) and its stabilization of peptidyl tRNA in the A site (24).

In this study we used rapid kinetics methods in a state-of-the-art in vitro translation system with in vivo-like rates (25, 26) to characterize the mechanism of viomycin inhibition of the peptide elongation cycle. We constructed a quantitative model for viomycin action with precise estimates of the model parameters. This model quantifies the functional aspects of viomycin inhibition of translocation. It paves the way for deeper understanding of the basis of the antimicrobial activity of viomycin and the evolution of viomycin resistance, thereby facilitating further development of the chemically malleable tuberactinomycin class of antibiotics.

Results

The Probability That a Ribosome Becomes Inhibited by Viomycin During an Elongation Cycle Is Determined by Competition Between Viomycin and EF-G for Ribosome Binding.

To study the effects of viomycin on translating ribosomes we carried out tripeptide formation experiments at varying concentrations of the drug using a quench-flow instrument. In these experiments 70S ribosomes programmed with an mRNA encoding Met-Phe-Thr-Stop (MFT mRNA) and with P-site bound f[³H]Met-tRNA^fMet were rapidly mixed with a factor mixture containing Phe-tRNA^Phe and Thr-tRNA^Thr in ternary complex with EF-Tu·GTP, EF-G, and varying concentrations of viomycin. The reactions were stopped at different incubation times by formic acid and the relative amount of each peptide was determined by reversed-phase HPLC (RP-HPLC) with on-line radiation detection.

In the absence of viomycin (red trace in Fig. 1D) tripeptide formation displayed biphasic kinetics with a fast phase accounting for ∼85% of the total amplitude. The mean times of tripeptide formation in the fast phase were 240, 180, and 150 ms at 2.5, 5, and 10 µM EF-G, respectively. The slow phase (∼15%), probably caused by a small fraction of partially active ribosomes, had a mean time in the seconds range and was treated as a background reaction (SI Materials and Methods).

When viomycin was included in the factor mixture, so that all ribosomes were viomycin-free before mixing in the quench-flow instrument, we observed a large effect on the kinetics of tripeptide formation (orange through violet traces in Fig. 1D), whereas the rate and amplitude of dipeptide formation remained unaffected (Fig. S1). The relative amplitude of the fast phase gradually decreased from around 85% in the absence of viomycin to essentially 0 at 50 µM viomycin, whereas its mean time remained virtually unaltered, indicating that the fast phase represents tripeptide formation on viomycin-free ribosomes. A new slow phase appeared with a mean time of several tens of seconds, likely representing tripeptide formation by viomycin inhibited ribosomes after drug dissociation (Fig. 1D). The total amount of radiodetectable peptide decreased with time in a viomycin concentration-dependent manner (Fig. 1D). This behavior is an artifact, caused by viomycin-induced read-through of the stop codon following the threonine codon in the MFT mRNA in the absence of class-1 release factor. This affected the amplitude of the slow but not the fast phase (SI Materials and Methods). Therefore, the fraction of viomycin-inhibited ribosomes was estimated from the degree of reduction of the fast phase amplitude by viomycin and not from the amplitude of the slow phase.

Fig. S1.

Time courses of f[³H]Met-Phe dipeptide formation in the presence of varying concentrations of viomycin. It shows that the rate of dipeptide formation is not affected by the presence of viomycin and the average time for dipeptide formation is in all cases is 20 ms.

The fraction of viomycin-inhibited ribosomes increased with viomycin concentration in a near hyperbolic manner (Fig. 1E). For a given viomycin concentration the fraction of inhibited ribosomes was much larger at low than at high EF-G concentration (Fig. 1E). This observation suggested competition between EF-G and viomycin for binding to the pretranslocation ribosome. If EF-G bound first, the ribosome proceeded unhindered to tripeptide formation, resulting in the fast phase. Alternatively, if viomycin bound first, the ribosome stalled and the mean time of tripeptide formation was greatly prolonged, resulting in the slow phase (Fig. 1D).

When viomycin was allowed to equilibrate with the 70S initiation complex by adding it to both the ribosome and the factor mixtures, we observed tripeptide formation kinetics identical to those when the drug was included only in the factor mixture. Notably, the fraction of viomycin-inhibited ribosomes was exactly the same (Fig. 1F), meaning that preincubation of viomycin with the initiation complex did not confer any advantage to the drug in its competition with EF-G for ribosome binding. It also implies that ribosomes with an empty A-site bind only a negligible amount of viomycin at the concentrations used, indicating that stable viomycin binding requires an A-site bound tRNA.

The Time a Ribosome Remains Inhibited After Viomycin Binding Depends Only on the Viomycin Concentration.

To estimate the mean time for translocation after viomycin binding to the pretranslocation ribosome we designed a two-step tripeptide formation assay. First, viomycin-stalled pretranslocation ribosomes were formed by incubating initiated 70S ribosomes programmed with a truncated MFT mRNA and P-site bound f[³H]Met-tRNA^fMet for 5–10 s at 37 °C with Phe-tRNA^Phe–containing ternary complex and varying concentrations of viomycin. The resulting viomycin-stalled pretranslocation complexes were then mixed with EF-G, Thr-tRNA^Thr–containing ternary complex, and viomycin, leading to slow translocation of f[³H]Met-Phe-tRNA^Phe followed by rapid formation of f[³H]Met-Phe-Thr tripeptide. An mRNA truncated directly after the Thr codon was used to avoid viomycin-mediated read-through of the stop codon (Fig. S2). The reactions were stopped at different incubation times by formic acid and the relative amount of each peptide was determined as described above.

Fig. S2.

Comparison between the regular MFT mRNA and the truncated MFT mRNA in tripeptide formation in the presence of 50 µM viomycin. Whereas all ribosomes programmed with the truncated mRNA are carrying an f[³H]Met-Phe-Thr tripeptide at 400 s, fewer than 60% of the ribosomes programmed with the nontruncated mRNA are doing so. As discussed in Results, this is due to read-through of the stop codon after Thr codon by the ribosome in the presence of viomycin.

In these experiments we observed a complete absence of the fast phase of tripeptide formation (Fig. 2A), implying that virtually all ribosomes were viomycin-bound even at the lowest viomycin concentration (1 µM). In contrast, when 70S initiation complex with empty A-site and P-site bound f[³H]Met-tRNA^fMet was preequilibrated with viomycin in the previous section, the fraction of inhibited ribosomes did not change (Fig. 1F), indicating negligible drug binding. Taken together these results demonstrate that the affinity of viomycin to the ribosome greatly increases upon A-site binding of a tRNA.

Fig. 2.

(A) Time course of f[³H]Met-Phe-Thr tripeptide formation at varying concentrations of viomycin and 5 µM EF-G when viomycin is prebound to the pretranslocation complex. Solid lines represent single exponential fits to the data. (B) The mean time of tripeptide formation (${\tau }_i$) on viomycin-stalled ribosomes at different EF-G concentrations plotted as a function of viomycin concentration. The solid line represents a fit of Eq. 3 to the data. (C) Average elongation cycle time (${\tau }_{avg}$) calculated from the data in Figs. 1E and 2B at different EF-G concentrations plotted as a function of the viomycin concentration. Solid lines represent fits of Eq. 4 to the data. Error bars represent SEM.

The mean time of tripeptide formation on viomycin-stalled ribosomes was estimated by fitting a single exponential function to each of the curves in Fig. 2A. The mean time increased linearly with viomycin concentration from around 45 s at 1 µM to around 120 s at 100 µM (Fig. 2B), indicating that after dissociation the drug could reassociate to the pretranslocation ribosome and prolong the stalling. In contrast to the fraction of viomycin-inhibited ribosomes, which displayed strong negative correlation with EF-G concentration (Fig. 1E), the mean times displayed no EF-G concentration dependence in the 2.5–10 µM range (Fig. 2B). This suggests that viomcyin dissociation and reassociation occur on an EF-G bound pretranslocation ribosome.

The Average Time of an Elongation Cycle Increases Nonlinearly with Viomycin Concentration.

A major determinant of the growth inhibitory effect of ribosome targeting antibiotic drugs is their ability to slow down peptide elongation. Viomycin is a “slow” inhibitor of mRNA translation that stalls the ribosome for a time, ${\tau }_i$, substantially longer than that required to carry out an uninhibited elongation cycle, ${\tau }_0$ (Fig. 2B). Therefore, its effect on the average time of one elongation cycle, ${\tau }_{avg}$, can be conveniently written as (27)

$${\tau }_{avg}={\tau }_0+P_i\cdot {\tau }_i.$$ [1]

Here, $P_i$ is the probability that the ribosome is stalled by viomycin during one elongation cycle (Fig. 1E). The average time increased nonlinearly with increasing viomycin concentration (Fig. 2C) in a manner also observed in the case of fusidic acid inhibition of elongating ribosomes (27). The average time (${\tau }_{avg}$) increased rapidly at low viomycin concentration as a consequence of the rapidly increasing probability that the ribosome becomes stalled (Fig. 1E). In contrast, at high viomycin concentration where the ribosome is virtually guaranteed to stall ${\tau }_{avg}$ increased slowly due to viomycin rebinding events (Fig. 2B). This nonlinear response of the average elongation cycle time to viomycin concentration confirms that viomycin can bind to two ribosomal states during the elongation cycle. First, the drug binds with high efficiency to a transient state, likely the EF-G–free pretranslocation ribosome (Fig. 4). This first binding event is rapidly followed by transition to a downstream stalled state, likely the EF-G–bound pretranslocation ribosome (Fig. 4). In this state, the stalling time increased linearly with viomycin concentration due to repeated slow dissociation from and fast reassociation of viomycin to the stalled ribosome.

Fig. 4.

Kinetic model for viomycin inhibition of the elongation cycle. A ternary complex binds to the ribosome and brings it to the first viomycin-sensitive state; here viomycin can bind with rate constant $k_{V1}$ or EF-G can bind with rate constant $k_G$ . The ratio of these two rate constants defines the first viomycin inhibition constant, $K_{I1}$. Binding of EF-G brings the ribosome to the second viomycin-sensitive state; here viomycin can bind with rate constant $k_{V2}$ or the ribosome can escape this state and continue with translocation with rate constant $k_{trans}$. The ratio of these two rate constants defines the second viomycin inhibition constant $K_{I2}$. Ribosomes that become viomycin-bound can still bind EF-G, which will cycle on and off these ribosomes and hydrolyze GTP with turnover rate $k_{GTP}$. Once viomycin dissociates with rate constants $q_{V1}$ or $q_{V2}$ and rebinding of the drug is avoided the ribosome will translocate, EF-G will dissociate, and the next ternary complex will bind to begin the cycle anew.

Viomycin Inhibition Leads to Multiple Rounds of Futile GTP Hydrolysis by EF-G Before Translocation.

EF-G can bind to the viomycin-stalled pretranslocation ribosome (12, 14, 17). We wished to know whether such binding leads to multiple cycles of GTP hydrolysis before successful translocation. Therefore, we designed an experiment to follow multiple turnover GTP hydrolysis by EF-G in the presence of viomycin-stalled pretranslocation ribosomes. For that, initiated 70S ribosomes programmed with MFT mRNA and P-site bound fMet-tRNA^fMet were rapidly mixed in a quench-flow instrument with a mixture of Phe-tRNA^Phe containing ternary complex, viomycin, [³H]GTP, and varying concentrations of EF-G. Viomycin-stalled pretranslocation complex was formed within 20 ms (Fig. S1). The reactions were stopped at different incubation times by formic acid and the relative amounts of [³H]GTP and [³H]GDP were determined by anion exchange chromatography with on-line radiation detection.

At all EF-G concentrations [³H]GDP accumulated linearly with time (Fig. 3A), showing that steady-state cycling of EF-G on the ribosome was established rapidly. The turnover rate of GTP hydrolysis, zero in the absence of EF-G, saturated already at the lowest EF-G concentration used (0.625 µM), as expected from the lack of an EF-G concentration dependence of the stalling time observed above (Fig. 2B). We observed a small linear increase in the GTP hydrolysis rate at higher EF-G concentrations from 3.9 s⁻¹ at 0.625 µM to 6.6 s⁻¹ at 7.5 µM EF-G (Fig. 3B), likely due to EF-G reacting with free 50S subunits. Accounting for this side reaction the maximal rate of GTP hydrolysis by EF-G on viomycin-stalled pretranslocation ribosomes, k_GTP, was estimated as 3.75 ± 0.20 s⁻¹ from the y axis intercept of the straight line in Fig. 3B. This estimate corresponds to a dwell time of 270 ± 15 ms for EF-G on the viomycin-stalled ribosome, in good agreement with previously published dwell time estimates based on single-molecule FRET experiments (12, 28).

Fig. 3.

(A) Time traces of multiple turnover GTP hydrolysis by EF-G on viomycin-stalled pretranslocation complexes at four different EF-G concentrations. A vertical offset of 20 has been added to separate the lines for clarity of presentation. (B) Rate of turnover GTP hydrolysis plotted as a function of EF-G concentration. (C) Fold change in average elongation cycle time and GTP molecules hydrolyzed per elongation cycle calculated using Eqs. 4 and 5 plotted as a function of viomycin concentration. The uninhibited elongation rate is assumed to be 20 amino acids per second and the free EF-G concentration is set to 10 µM; all other model parameters are those from the main text. All error bars represent SEM.

A Kinetic Model for Viomycin Action on Translating Ribosomes.

The simplest model for viomycin inhibition of protein synthesis that accounts for all of the in vitro results described in the present work is shown in Fig. 4. The total probability that a ribosome becomes inhibited by viomycin during an elongation cycle is a combination of the probabilities that viomycin binds either to the first or to the second sensitive state and is given by

$$P_i=1-\frac{K_{I1}\cdot \left[\text{EF}‐\text{G}\right]}{K_{I1}\cdot \left[\text{EF}‐\text{G}\right]+\left[\text{Vio}\right]}\cdot \frac{K_{I2}}{K_{I2}+\left[\text{Vio}\right]}.$$ [2]

The second term on the right side is the product of the probabilities that the ribosome escapes inhibition in the first and the second state. The inhibition constant, $K_{I1}$, is the ratio between the binding rate constants of EF-G and viomycin to the pretranslocation ribosome ($k_G$ and $k_{V1}$ in Fig. 4). The inhibition constant, $K_{I2}$, is the ratio between the rate constants for the forward step required for the ribosome to escape from the second viomycin-sensitive state and for viomycin binding to that state ($k_{trans}$ and $k_{V2}$ in Fig. 4).

The time that a ribosome remains inhibited by viomycin is given by (SI Materials and Methods)

$${\tau }_i=\frac{1}{q_{V2}}\cdot \left(1+\frac{\left[\text{Vio}\right]}{K_{I2}}\right).$$ [3]

Here $q_{V2}$ is the rate constant for viomycin dissociation from the second viomycin-sensitive state, the EF-G–bound ribosome. The average elongation cycle time can now be written in terms of the elementary rate constants of the translocation process in the presence of viomycin as

$${\tau }_{avg}={\tau }_0+P_i\cdot {\tau }_i={\tau }_0+\frac{1}{q_{V2}}\cdot \left(\frac{\left[\text{Vio}\right]}{\left[\text{Vio}\right]+K_{I1}\cdot \left[\text{EF}‐\text{G}\right]}+\frac{\left[\text{Vio}\right]}{K_{I2}}\right).$$ [4]

The parameters $K_{I1}$, $K_{I2}$, and $q_{V2}$ that define the concentration dependence of the inhibition of translocation by viomycin were estimated by fitting Eqs. 2–4 to the data in Figs. 1E and 2 B and C. The inhibition constant $K_{I1}$ is 0.55 ± 0.03, meaning that viomycin binds to the pretranslocation ribosome roughly twice as fast as EF-G. The inhibition constant $K_{I2}$ is 66 ± 5 µM. This is the viomycin concentration required to double the time a ribosome remains stalled due to successive viomycin rebinding events. The rate constant for viomycin dissociation from the EF-G–bound ribosome, $q_{V2}$, is 0.022 ± 0.0005 s⁻¹. The inverse of this rate constant estimates the residence time for viomycin on this complex as 44 ± 1 s. From Eq. 4 we determined $\mathrm{I}{\mathrm{C}}_{50}$ values for translocation inhibition by viomycin, defined as the viomycin concentration required to double the duration of an average elongation cycle. $\mathrm{I}{\mathrm{C}}_{50}$ values at 2.5, 5, and 10 µM EF-G are 5, 6, and 9 nM viomycin, respectively.

Using Eq. 4 it is possible to account for the futile cycles of GTP hydrolysis by EF-G on viomycin-stalled ribosomes. The average number of GTP molecules hydrolyzed per elongation cycle as a function of the viomycin concentration is given by

$$GT{P}_{avg}=GT{P}_0+k_{GTP}\cdot P_i\cdot {\tau }_i=2+\frac{k_{GTP}}{q_{V2}}\left(\frac{\left[\text{Vio}\right]}{\left[\text{Vio}\right]+K_{I1}\cdot \left[\text{EF}‐\text{G}\right]}+\frac{\left[\text{Vio}\right]}{K_{I2}}\right).$$ [5]

Here $GT{P}_0$ is the number of GTP molecules hydrolyzed during an uninhibited elongation cycle, that is, 2, $k_{GTP}$ is the (saturated) turnover rate of GTP hydrolysis by EF-G on the viomycin-stalled ribosome and $P_i\cdot {\tau }_i$ is the average time per elongation cycle the ribosome is viomycin-bound and stimulating futile GTP hydrolysis by EF-G. Considerably more viomycin is required to double the GTP cost of an elongation cycle than to double its duration (Fig. 3C): 15, 30, and 60 nM at 2.5, 5, and 10 µM EF-G, respectively.

Discussion

From the in vitro results presented here we constructed a simple yet powerful model for viomycin inhibition of translocation and estimated its three kinetic parameters, $K_{I1}$, $K_{I2}$, and $q_{V2}$ (Fig. 4 and Eq. 4). According to this model there are two viomycin-sensitive states in the elongation cycle. Elongating ribosomes first become viomycin-sensitive upon aminoacyl tRNA delivery to the ribosomal A site by EF-Tu. This EF-G–free pretranslocation state is the most viomycin-sensitive state. Viomycin and EF-G compete for first binding to this state such that the probability of a successful viomycin attack is 50% when the concentration of viomycin is equal to the product of the inhibition constant $K_{I1}$ and the EF-G concentration (Eq. 2). Whether or not the first state becomes viomycin-bound, EF-G binding brings the ribosome to the second viomycin-sensitive pretranslocation state, where viomycin-bound ribosomes stall for a minimum time of ∼45 s. In this state, rebinding of viomycin prolongs the stalling time, which increases linearly with viomycin concentration and is doubled when the viomycin concentration reaches $K_{I2}$, here estimated as 66 µM (Eq. 3). We note that the minimal stalling time of ∼45 s after a successful viomycin attack is equivalent to the time required to translate roughly 900 codons in rapidly growing Escherichia coli cells with an average codon translation time of 50 ms (29). This greatly exceeds the average distance of 14–26 codons between ribosomes on mRNA (30), and therefore viomycin binding to a ribosome is expected to lead to queuing of trailing ribosomes behind it.

The existence of two viomycin-sensitive states during the elongation cycle, one before and one after binding of EF-G to the ribosome, is supported by two recent structures of the viomycin-bound pretranslocation ribosome. One is a 3.3-Å crystal structure (18) containing three tRNAs and no EF-G, and the other is a 7-Å cryo-EM structure (17) containing two tRNAs and EF-G. In the former structure the ribosome is observed in the nonrotated state and the bound tRNAs are in the classical A, P, and E sites, whereas in the latter structure the ribosome is in the rotated state and the tRNAs are in the hybrid A/P and P/E states. Because FRET experiments suggested that viomycin binding induces rotation of the ribosomal subunits (11), the functional significance of the nonrotated viomycin-bound structure is unclear. Further structural work is required to understand the significance of different viomycin-bound states of the ribosome and also to clarify how the drug affects large-scale ribosome dynamics during translocation.

Our conclusion that viomycin binds tightly to the ribosome only after a tRNA has been delivered to the A site by EF-Tu also finds support in the crystal structure of the viomycin-bound ribosome (18). Here viomycin occupies the space between H69 and h44 that is vacated by the monitoring bases A1492 and A1493 when they engage with the codon–anticodon minihelix during mRNA decoding (21, 22, 31). This suggests that A-site binding of a tRNA liberates the viomycin binding site, which would otherwise be occupied by the monitoring bases. Further evidence that A-site binding of tRNA increases viomycin affinity to the ribosome comes from data presented in a recent single-molecule study (13). From their data we estimate a dissociation constant for viomycin binding to 70S ribosomes with an empty A site as large as 20 µM. In contrast, we observe that when viomycin is equilibrated with ribosomes with peptidyl tRNA in the A site, drug binding is fully saturated already at 1 µM, implying a dissociation constant much smaller than 1 µM.

The free concentration of viomycin required to significantly reduce the average peptide elongation rate is very low, in the nanomolar range under in vivo-like conditions (Fig. 3C). In addition to cell permeability factors, drug efflux pumps and drug degradation pathways the sensitivity of bacteria to viomycin will depend on the concentration of intracellular ribosomes and their sensitivity to the drug, as determined by the numerical values of the three parameters, $K_{I1}$, $K_{I2}$, and $q_{V2}$. Here it is relevant to consider the clinical drug target M. tuberculosis, which is known to be significantly more sensitive to viomycin and its sister drug capreomycin than E. coli (32). Slow-growing mycobacteria such as M. tuberculosis maintain a much lower intracellular ribosome concentration than the fast-growing E. coli (33). This by itself would increase the drug susceptibility of M. tuberculosis because at any intracellular drug concentration a larger fraction of the ribosomes would be disabled. However, the drug susceptibility of individual ribosomes does play a major role, as evidenced by multiple resistance mutations present either in ribosomal RNA (34) or in the rRNA methylase TlyA (32, 35). In our model such resistance mutations would lead to changes in the numerical values of the parameters $K_{I1}$, $K_{I2}$, and $q_{V2}$. Destabilization of drug binding on the ribosome would lead to a larger dissociation rate constant, $q_{V2}$, whereas decreased drug binding rate or decreased lifetimes of the two drug sensitive states would be reflected by larger $K_{I1}$ and $K_{I2}$ constants. Other members of the tuberactinomycin family such as capreomycin and enviomycin share the same binding site on the bacterial ribosome and have effects on bacterial cells similar to those of viomycin. Thus, our model likely describes translocation inhibition by the entire tuberactinomycin class of antibiotics, and differences in drug efficacy are likely caused by different values of the three parameters rather than different mechanisms.

Our results show that viomycin inhibition confers an extra energy cost due to the futile cycling of EF-G on viomycin-stalled ribosomes. However, much higher concentrations of viomycin are required to increase the GTP cost of an elongation cycle than are required to reduce the elongation rate (Fig. 3C). The in vivo implication of this extra energy loss by viomycin-bound ribosomes is hard to estimate, but it could be significant under energy-limited conditions.

In summary, we have characterized the mechanism of viomycin inhibition of the elongation step of bacterial protein synthesis. The kinetic model we present here along with estimates of its three key parameters provides a quantitative basis for understanding the antimicrobial activity of viomycin, also applicable in vivo. We are optimistic that our model is general enough to be instrumental also for characterization of other tuberactinomycin antibiotics and resistance mechanisms that have evolved also against these viomycin-related drugs. Finally, our model will aid characterization of more effective elongation inhibitors, which is one of the main goals of global antibiotics research.

Materials and Methods

Reagents and Buffers.

All experiments were performed at 37 °C in Hepes-polymix buffer [95 mM KCl, 5 mM NH₄Cl, 0.5 mM CaCl₂, 8 mM putrescine, 1 mM spermidine, 5 mM potassium phosphate, 5 mM Mg(OAc)₂, 1 mM dithioerythritol, and 5 mM Hepes, pH 7.5]. All reaction mixes except those for the GTP hydrolysis experiments contained 1 mM ATP, 1 mM GTP, 10 mM phosphoenolpyruvate (PEP), 1 µg/mL pyruvate kinase (PK), and 0.1 µg/mL myokinase (MK). In the GTP hydrolysis experiments the reaction mixes contained 1.5 mM ATP, 0.5 mM [³H]GTP, and 10 mM PEP but no PK or MK. His-tagged initiation factors IF1, IF2, and IF3; elongation factors EF-G, EF-Tu, and EF-Ts; and aminoacyl tRNA-synthetases (ThrRS and PheRS) were purified using nickel-affinity chromatography (HiTrap; GE Healthcare). All protein concentrations were determined using the Bradford assay. Ribosomes (E. coli MRE600) and f[³H]Met-tRNA^fMet were prepared according to ref. 36, and ribosome concentration was determined spectrophotometrically. XR7 mRNA with the coding sequence Met-Phe-Thr-Stop (AUG-UUU-ACG-AUU) and the truncated mRNA coding MFT (AUG-UUU-ACC) were was prepared according to ref. 27 and the truncated mRNA was purchased from IBA. Bulk tRNA was prepared from E. coli MRE600 according to ref. 37. tRNA^Phe was overexpressed in E. coli MRE600 from a plasmid with a T7 promoter and prepared from these cells as in ref. 37. [³H]Met and [³H]GTP were from Perkin-Elmer, viomycin was from USP, and all other chemicals were from either Merck or Sigma-Aldrich.

Quench-Flow Tripeptide Formation Experiments.

Two mixtures were prepared. The initiation mixture contained 70S ribosomes (0.3 µM), IF1, IF2, and IF3 (1 µM each), f[³H]Met-tRNA^fMet (1.5 µM), and MFT mRNA (0.7 µM). The elongation mixture contained EF-Tu (12 µM), EF-Ts (2 µM), EF-G (10, 20, or 40 µM), threonine (200 µM), phenylalanine (200 µM), ThrRS (0.5 µM), PheRS (0.5 µM), bulk tRNA (150 µM, of which tRNA^Thr1 and tRNA^Thr2 made up 4 µM and tRNA^Phe 1.7 µM), and 2.3 µM of overexpressed tRNA^Phe. Viomycin (1–100 µM) was added either to the elongation mixture or to both mixes as indicated. After 15-min incubation at 37 °C equal volumes of the two mixes were rapidly mixed and the reaction quenched at different time points with formic acid (17% final concentration) using a quench-flow instrument (RQF-3; KinTek Corp.). After quenching the samples were centrifuged at 20,800 × g and the supernatant discarded. The pellet was resuspended in 165 µL 0.5 M KOH to cleave the peptides from the tRNA. After 10 min 13 µL of 100% formic acid was added, the samples were centrifuged at 20,800 × g, and the radioactive peptides in the supernatant were analyzed by RP-HPLC using a H₂O/MeOH/trifluoroacetic acid (62/38/0.1 by volume) mobile phase and a C-18 column (Merck) with on-line scintillation counting (β-RAM model 3; IN/US Systems) to quantify the relative amounts of f[³H]Met, f[³H]Met-Phe, and f[³H]Met-Phe-Thr peptides.

Sequential Tripeptide Formation Experiments.

Three mixtures were prepared. The initiation mixture contained ribosomes (0.9 µM), IF1, IF2, and IF3 (1 µM each), f[³H]Met-tRNA^fMet (0.8 µM), and truncated MFT mRNA (1 µM). Two elongation mixes were prepared. The first contained phenylalanine (200 µM), PheRS (0.5 µM), tRNA^Phe (4 µM), EF-Tu (5 µM), EF-Ts (2 µM), and viomycin (2–200 µM). The second contained threonine (200 µM), ThrRS (0.5 µM), EF-Tu (8 µM), EF-Ts (2 µM), bulk tRNA (225 µM, of which tRNA^Thr1 and tRNA^Thr2 made up 6 µM), viomycin (1–100 µM), and EF-G (15–60 µM). All three mixes where incubated for 15 min at 37 °C. During the experiment, one volume of the initiation mixture was mixed with one volume of the first elongation mixture, and the resulting mixture was incubated for 5–10 s and then one volume of the second elongation mixture was added. The reaction was quenched at different time points after the addition of the second elongation mixture using formic acid (17% final). The samples were treated identically to the quench-flow samples above.

GTP Hydrolysis Experiments.

Two mixtures were prepared. The initiation mixture contained 70S ribosomes (0.5 µM), fMet-tRNA^fMet (0.8 µM), and MFT mRNA (0.8 µM). The elongation mixture contained phenylalanine (200 µM), PheRS (0.5 µM), overexpressed tRNA^Phe (4 µM), EF-Tu (8 µM), EF-Ts (1 µM), EF-G (1–40 µM), and viomycin (400 µM). Both reaction mixes contained 0.5 mM [³H]GTP, 1.5 mM ATP, and 10 mM PEP. After 15-min incubation at 37 °C equal volumes of the two mixes were rapidly mixed and the reaction quenched at different time points with formic acid (17% final concentration) using the quench-flow instrument. The acid-quenched samples were centrifuged at 20,800 × g. The supernatant containing the [³H]GTP and [³H]GDP was analyzed by anion exchange chromatography with on-line scintillation counting (β-RAM model 3; IN/US Systems). A Mono-Q GL column (GE Healthcare) was used and the mobile phase was a multistep gradient of 0–2 M NaCl in 20 mM Tris (pH 7.5).

Data Analysis and Curve Fitting.

All curve fitting was done in MATLAB R2014b (MathWorks) using the Leonardt–Marquardt algorithm as implemented in the curve-fitting toolbox. Detailed descriptions of curve-fitting procedures and derivations of the equations in the main text can be found in SI Materials and Methods.

SI Materials and Methods

mRNA Sequences.

The nucleotide sequences of the two model mRNA molecules used in this study are given below; the start codons of the short ORFs are bolded.

MFT.

• 5′-GAAUUCGGGCCCUUGUUAACAAUUAAGGAGGUAUACUAUGUUUACGUAAUUGCAGAAAAAAAAAAAAAAAAAAAAA-3′

Truncated MFT.

• 5′-UAACAAUAAGGAGGUAUUAAAUGUUUACG-3′

Data Analysis in the Quench-Flow Tripeptide Formation Experiments.

Because of the large difference in timescales between tripeptide formation on viomycin-free and viomycin-bound ribosomes the time curves could be treated as the sum of two parallel reactions, as in ref. 26. In the absence of viomycin tripeptide formation can be described as a process of three sequential irreversible steps with time constants ${\tau }_1$, ${\tau }_2$, and ${\tau }_3$. In our experiments there is also an additional slow phase present even in the absence of viomycin; the origin of this phase is unknown. It is treated as a background reaction involving a fourth sequential step with time constant ${\tau }_x$ occurring on a fraction of the ribosomes.

In the presence of viomycin a fraction of the ribosomes, denoted $P_i$, will become viomycin-bound. On these ribosomes tripeptide formation is much slower than on viomycin-free ribosomes and is treated as a single-step reaction with time constant ${\tau }_i$. Further, in the presence of viomycin there is read-through of the stop codon present at the end of our short three-codon ORF; the rate of this read-through is low and viomycin concentration-dependent. Because of this the total amount of tripeptide shrinks with time as longer peptides are formed. This means that the time constant ${\tau }_i$ could not be accurately estimated from these experiments.

Accounting for all these factors, fitting functions for tripeptide formation in the presence of viomycin can be set up as

$$\begin{array}{l}tri{p}_0\left(t\right)=A_1\cdot f\left({\tau }_1,{\tau }_2,{\tau }_3\right)+A_2\cdot f\left({\tau }_1,{\tau }_2,{\tau }_3,{\tau }_x\right)+bg\\ tri{p}_v\left(t\right)=A_1\cdot \omega \cdot \left(\left(1-P_i\right)\cdot \left(f\left({\tau }_1,{\tau }_2,{\tau }_3\right)+\frac{A_2}{A_1}\cdot f\left({\tau }_1,{\tau }_2,{\tau }_3{\tau }_x\right)\right)+P_i\cdot \left(f\left({\tau }_i\right)+\frac{A_2}{A_1}\cdot f\left({\tau }_i,{\tau }_x\right)\right)\right)+bg.\end{array}$$ [S1]

Here the functions $f\left(\tau \right)$ are the solutions of differential equation systems describing the accumulation of the end product of reaction schemes of sequential irreversible steps, one step per time constant indicated. $A_1$ and $A_2$ are the amplitudes of rapid and slow tripeptide formation in the absence of viomycin. $P_i$ is the viomycin concentration-dependent probability that a ribosome is bound by viomycin rather than translocating and $\omega$ is the final remaining tripeptide amplitude taking read-through of the stop codon into account. By fitting several curves obtained in the presence and absence of viomycin in parallel the parameter $P_i$ and the total average time for tripeptide formation on viomycin-free ribosomes ${\tau }_0={\tau }_1+{\tau }_2+{\tau }_3$ could be estimated with high precision. Note that the individual average times ${\tau }_1$, ${\tau }_2$, and ${\tau }_3$ cannot be separately determined.

Determination of the Slow Time, Viomycin Dissociation Rate Constant and the Second Inhibition Constant from Sequential Tripeptide Formation Experiments.

When initiated ribosomes are mixed with viomycin and ternary complexes containing a tRNA cognate for the A-site codon viomycin-stalled pretranslocation complexes are rapidly formed. Due to the very tight binding between viomycin and the pretranslocation complex all ribosomes will be viomycin-bound if viomycin is supplied in excess. When these ribosomes are then mixed with EF-G and ternary complexes containing tRNA cognate for the second codon a quasi-equilibrium is rapidly established between EF-G–bound and EF-G–free viomycin-stalled ribosomes. Once viomycin dissociates from any of these complexes one of two things can happen: either viomycin can rebind to the ribosome and maintain the stalling or the A-site tRNA can be translocated by EF-G and a tripeptide formed. This process can be described by a system of coupled differential equations:

$$\{\begin{array}{l}\frac{d{R}_{PRE}^{\mathrm{0,0}}}{dt}=-\left(k_v^0\cdot \left[V\right]+k_G\cdot \left[G\right]\right)\cdot R_{PRE}^{\mathrm{0,0}}+q_v^0\cdot R_{PRE}^{Vio,0}+q_G\cdot R_{PRE}^{0,G}\\ \frac{d{R}_{PRE}^{Vio,0}}{dt}=-\left(q_v^0+k_G\cdot \left[G\right]\right)\cdot R_{PRE}^{Vio,0}+k_v^0\cdot \left[V\right]\cdot R_{PRE}^{\mathrm{0,0}}+q_G\cdot R_{PRE}^{Vio,G}\\ \frac{d{R}_{PRE}^{0,G}}{dt}=-\left(q_G+k_v^G\cdot \left[V\right]\right)\cdot R_{PRE}^{0,G}+k_G\cdot \left[G\right]\cdot R_{PRE}^{\mathrm{0,0}}+q_v^G\cdot R_{PRE}^{Vio,G}\\ \frac{d{R}_{PRE}^{Vio,G}}{dt}=-\left(q_G+q_v^G\right)\cdot R_{PRE}^{Vio,G}+k_G\cdot \left[G\right]\cdot R_{PRE}^{Vio,0}+k_v^G\cdot \left[V\right]\cdot R_{PRE}^{0,G}\\ \frac{d{R}_{POST}}{dt}=k_t\cdot R_{PRE.}^{0,G}\end{array}$$ [S2]

Here $R_{\mathrm{PRE}}^{\mathrm{0,0}}$ is the pretranslocation ribosome free from both viomycin and EF-G, $R_{PRE}^{Vio,0}$ is the pretranslocation ribosome bound by viomycin but free from EF-G, $R_{PRE}^{0,G}$ is the pretranslocation ribosome free from viomycin but bound by EF-G, $R_{PRE}^{Vio,G}$ is the pretranslocation ribosome bound by both viomycin and EF-G, and $R_{POST}$ is the first state along the translocation pathway where the ribosome is no longer sensitive to viomycin inhibition. $k_v^0$ and $k_v^G$ are, respectively, the binding rate constants for viomycin to the EF-G–free and the EF-G–bound ribosome. $q_V^0$ and $q_v^G$ are, respectively, the dissociation rate constants for viomycin from the EF-G–free and the EF-G–bound ribosome. $k_G$ and $q_G$ are the association rate and dissociation rate constants for EF-G to the pretranslocation ribosome. Finally, $k_t$ is the rate constant by which the ribosome leaves the second viomycin-sensitive pretranslocation state for a downstream state that is no longer sensitive to viomycin inhibition.

If we assume that EF-G binding and dissociation is much faster than viomycin dissociation, which is reasonable due to the very long inhibition times at low viomycin concentrations and the very slow increase in inhibition time with viomycin, we can replace the rate constants $k_v^0$,$k_v^G$,$q_v^0$,$q_v^G$, and $k_t$ with their weighted averages over EF-G–bound and EF-G–free ribosomes:

$$\begin{array}{l}{k}_v=\frac{k_v^G\left[G\right]+k_v^0{K}_G}{\left[G\right]+K_G}\\ q_v=\frac{q_v^G\left[G\right]+q_v^0{K}_G}{\left[G\right]+K_G}\\ k_t=\frac{k_t^G\left[G\right]}{\left[G\right]+K_G}.\end{array}$$ [S3]

This yields a simplified system of three differential equations:

$$\{\begin{array}{l}\frac{d{R}_{PRE}^0}{dt}=-\left(k_v\cdot \left[V\right]+k_t\right)\cdot R_{PRE}^0+q_v\cdot R_{PRE}^{Vio}\\ \frac{d{R}_{PRE}^{Vio}}{dt}=-q_v\cdot R_{PRE}^{Vio}+k_v\cdot R_{PRE}^0\\ \frac{d{R}_{POST}}{dt}=k_t\cdot R_{PRE.}^0\end{array}$$ [S4]

Here $R_{PRE}^0$ is the ribosome free of viomycin and either bound by or free of EF-G. $R_{PRE}^{Vio}$ is the ribosome bound by viomycin and either bound by or free of EF-G. $R_{POST}$ is as above the first viomycin-insensitive state along the translocation pathway.

Integrating these differential equations from 0 to infinite time and taking the appropriate initial conditions [$R_{PRE}^0\left(0\right)=0$,$R_{PRE}^{Vio}\left(0\right)=1$ and $R_{POST}\left(0\right)=0$] into account yields a system of algebraic equations:

$$\{\begin{array}{l}0=-\left(k_v\left[V\right]+k_t\right)\cdot {\tau }_{PRE}^0+q_v\cdot {\tau }_{PRE}^{Vio}\\ -1=-q_v\cdot {\tau }_{PRE}^{Vio}+k_v\cdot {\tau }_{PRE}^0\\ 1=k_t\cdot {\tau }_{PRE.}^0\end{array}$$ [S5]

Here ${\tau }_{PRE}^0$ and ${\tau }_{PRE}^{Vio}$ are the average life times of their respective complexes and ${\tau }_{tot}={\tau }_{PRE}^0+{\tau }_{PRE}^{Vio}$ is the average time required for a viomycin-bound ribosome to translocate and reach the $R_{POST}$ state. Solving for the average times, using the substitution above and assuming that $q_v^0=q_v^G$ and $k_v^0=k_v^G$ yields

$$\begin{array}{l}{\tau }_{PRE}^0=\frac{\left[G\right]+K_G}{\left[G\right]\cdot k_t}\\ {\tau }_{PRE}^{Vio}=\frac{1}{q_v}\cdot \left(1+\frac{k_v\cdot \left[V\right]}{k_t}+\frac{k_v\cdot \left[V\right]\cdot K_G}{k_t\cdot \left[G\right]}\right).\end{array}$$ [S6]

At EF-G concentrations much larger than $K_G$, as is the case in our experiments and likely also in the living cell, this gives

$${\tau }_{PRE}^{Vio}={\tau }_i=\frac{1}{q_v}\cdot \left(1+\frac{\left[V\right]}{K_{I2}}\right).$$ [S7]

This is the time a ribosome, once bound by viomycin, stays inhibited before it successfully translocates. Linear fitting of this expression to the average time plots in Fig. 2 yields precise estimates of $q_v$ and $K_{I2}$.

Multiple Turnover GTP Hydrolysis on Viomycin-Stalled Ribosomes.

The measurements of the GTPase activity of EF-G cycling on viomycin-stalled PRE complexes is complicated by the instability of this complex, even in the presence of viomycin. The steady-state rate of GTP hydrolysis taking the decay of this complex by peptidyl-tRNA drop off from the A site and translocation of peptidyl tRNA to the P site into account can be written as

$$\frac{dGDP}{dt}=k_{PRE}^{Vio}\cdot e^{-\left(k_d+k_{trans}\right)t}+k_{TER}\cdot \frac{k_d}{k_d+k_{trans}}\cdot \left(1-e^{-\left(k_d+k_{trans}\right)t}\right)+k_{POST}\cdot \frac{k_{trans}}{k_{trans}+k_d}\cdot \left(1-e^{-\left(k_d+k_{trans}\right)t}\right).$$ [S8]

Here $k_{PRE}^{Vio}$ is the rate of multiple turnover GTP hydrolysis on viomycin-stalled pretranslocation complexes, $k_{TER}$ is the multiple turnover rate of GTP hydrolysis on ribosomes where the peptidyl tRNA has dropped off, leaving an empty A site and a deacylated tRNA in the P site, $k_{POST}$ is the multiple turnover rate of GTP hydrolysis on posttranslocation ribosomes, $k_d$ is the rate of peptidyl tRNA drop off from the A site, and $k_{trans}$ is the rate of translocation. At the high viomycin concentration used in these experiments (200 µM) both $k_d$ and $k_{trans}$ are very small compared with $k_{PRE,VIO}$ and therefore at small t the term$e^{-\left(k_d+k_{trans}\right)t}\approx 1$; this gives the simpler differential equation

$$\frac{dGDP}{dt}=k_{PRE}^{Vio},$$ [S9]

which readily solves to

$$GDP\left(t\right)=k_{PRE}^{Vio}\cdot t.$$ [S10]

Thus, the initial, linear phase of GDP accumulation in these experiments directly gives the rate of GTP hydrolysis by EF-G on the viomycin-stalled PRE complex.