Depletion of antibiotic targets has widely varying effects on growth

To regulate protein degradation, we used the Clp protease system that is widely conserved among bacteria. Clp consists of the highly processive ClpP protease and several accessory proteins that identify targets and help denature them before proteolysis (8). Although these factors can recognize a number of different amino acid sequences, the best characterized is SsrA, a short peptide that is added to the C terminus of incomplete proteins at stalled ribosomes (9). In Escherichia coli, SsrA-tagged proteins are recognized by two accessory factors, ClpX and SspB, and are delivered to ClpP for rapid degradation (10). Our strategy relies on appending a modified SsrA tag, masking the sequence to the target protein of interest, and inducing degradation by revealing the SsrA sequence at the C terminus using HIV-2 protease (Fig. 1A).

Although Clp and its accessory factors have been conditionally expressed in other bacteria (11, 12), the two subunits of the Clp protease and many associated factors seem to be required for normal growth in mycobacteria (6), suggesting that altering their expression level might have unwanted physiological effects. Instead, we chose to modify the target protein and use the endogenous protease. We initially confirmed that the SsrA sequence must be located at the C terminus of a protein to mediate degradation; the addition of more amino acids results in completely stable proteins (Fig. S1). Based on this observation, we developed an inducible degradation (ID) system. We engineered a C-terminal peptide, the ID tag (Fig. 1 A and B), consisting of the mycobacterial SsrA sequence (with the substitution of a single amino acid) fused to a downstream protecting peptide, which can be cleaved by a site-specific protease derived from an HIV-2 isolate. Induction of HIV-2 protease expression exposes the C terminus of the SsrA sequence and targets the protein for Clp-mediated proteolysis. We incorporated a FLAG tag (DYKDDDDK) at the C terminus of the ID tag to track uncleaved target protein and a c-myc tag (EQKLISEEDL) at the N terminus to monitor the stability of protein after cleavage.

Expression of GFP fused to the ID tag in M. smegmatis yielded fluorescent cells in the absence of HIV-2 protease (Fig. S1). In the absence of the inducer, anhydrotetracycline (ATc) (13), the presence of a plasmid encoding a tetracycline-inducible HIV-2 protease gene (codon-optimized for expression in mycobacteria) had no effect on fluorescence (Fig. S1). With added ATc, fluorescence intensity dropped to baseline after overnight incubation (Fig. 1C). Similarly, when we fused the ID tag to the Zeocin resistance protein (which mediates resistance to this bleomycin family drug by binding and neutralization) (14), we found that the Zeocin minimal inhibitory concentration (MIC) was identical in a strain that lacked HIV-2 protease and a strain containing inducible HIV-2 protease in the absence of ATc (Fig. S2). The addition of inducer resulted in dramatic target depletion, and bacteria were sensitized to Zeocin by approximately 32-fold (Fig. 1D).

The ID tag system allowed us to test the common view that antibiotics act largely by reducing the target activity below a threshold level. We chose six antibiotic target proteins: RNA polymerase β-subunit (RpoB), the target of rifampicin; gyrase A subunit (GyrA), the target of quinolones (15); alanine racemase (Alr), the target of d-cycloserine (16); dihydrofolate reductase (DHFR), the target of trimethoprim; enoyl reductase InhA, the target of isoniazid (17), and β-ketoacyl-acyl-carrier protein synthase KasA, the target of thiolactomycin (18). We used phage-mediated homologous recombination (recombineering) (19) to insert the ID tag into M. smegmatis chromosomal genes to make these diverse antibiotic targets of ID tag fusion protein. In each ID-tagged strain, we separately integrated a copy of the HIV-2 protease gene regulated by a tetracycline-dependent promoter. Without induction, these target proteins are expressed by their native promoter to execute their function, and after induction, the cleavage by HIV-2 protease will mark these proteins for degradation.

All six antibiotic targets were significantly depleted on induction of HIV-2 protease (Fig. 2A). Quantitative Western blotting revealed that incubation with inducer depleted the targets to different extents. Although RpoB levels remained at ∼20% of the levels observed without inducer, >97% of the ID-tagged Alr was degraded (Fig. 2B). Degradation of InhA, Alr, and DHFR was essentially complete within 3 h (Fig. 2C), whereas there was little depletion of RpoB or GyrA at 6 h (Fig. S3). The proteins that were less rapidly depleted could be more resistant to proteolysis or could be up-regulated in response to depletion, which might be the case for KasA (20).

If killing were proportional to the reduction in target enzyme activity mediated by these antibiotics, we expected that bacterial growth and survival would correlate with the degree of protein depletion. We measured the growth rate of each strain in the presence or absence of inducer beginning either directly (for rapidly depleted proteins) or 12 h (for slowly depleted proteins) after addition of inducer. However, we found very little correlation between protein abundance and growth. Efficient InhA depletion resulted in bacterial death, whereas >97% depletion of DHFR and Alr only modestly slowed growth. However, modest depletion of RpoB led to a dramatic growth arrest (Fig. 3).

To confirm that the ID tag mediated selective degradation (and not just, for example, removal of the tag), we used two different approaches that showed that target proteins were functionally inactivated. First, HIV-2 protease induction caused hypersensitivity to antibiotics targeting the depleted protein but not to other antibiotics (Fig. 4A). These results show that ID tag-mediated proteolysis results in functional depletion of each target and confirm that they are the targets for the cognate antibiotics. Second, we directly assayed enzyme activity for two targets, DHFR and Alr. In both cases, we found markedly reduced activity consistent with the results of immunoblotting (Fig. 4B). Thus, cells grew rapidly, despite marked depletion of these critical proteins.

Depletion of DHFR fails to phenocopy treatment with trimethoprim, a presumed inhibitor of the enzyme. This could be caused by at least three different mechanisms. Trimethoprim could have multiple targets along with DHFR, and inhibition could require those additional activities. The drug might not be a simple inhibitor and may alter enzyme activity in a manner that induces toxicity (15, 21, 22) or activates a cell death pathway (23–25). Finally, trimethoprim might simply produce a greater degree of inhibition than is seen with target depletion.

To distinguish these possibilities, we used metabolic profiling. We expected one of two patterns. If trimethoprim had additional targets or induced complex alterations in enzyme activity, we should see different changes in metabolites than those changes induced by target depletion. However, to see these effects, we needed to treat with subinhibitory concentrations of trimethoprim to avoid seeing secondary changes in metabolites associated with altered growth rate.

To perform the metabolic assays, we grew bacteria on filters to allow rapid lysis. Under these conditions, we found that depletion of DHFR was still induced by low concentrations of ATc (Fig. S5A). As seen in broth cultures, filter-grown cells exhibited only modest growth defects after DHFR depletion (Fig. S5B). We found that somewhat higher concentrations of trimethoprim were required to inhibit growth on filters than in broth. However, as in broth culture, bacteria became hypersusceptible to trimethoprim after DHFR depletion (Fig. S5B).

We examined the concentrations of a large set of identified and unidentified metabolites after DHFR depletion or treatment with trimethoprim. In agreement with previous results (26), we saw a number of metabolites whose concentration changed markedly and that are consistent with the known biochemical pathway. Homocysteine, deoxyuridine monophosphate (dUMP), serine, and aminoimidazole carboxamide ribonucleotide (AICAR) increased dramatically with either DHFR depletion or trimethoprim treatment (Fig. S5C). Higher concentrations of trimethoprim evoked considerably larger changes in the same metabolites as well as additional alterations.

We reasoned that, if DHFR is the main target of trimethoprim, than the metabolic changes induced by drug treatment should be mimicked by DHFR depletion. In addition, treatment with drug after DHFR depletion should show similar metabolic effects but at lower trimethoprim concentrations. To test this prediction, we compared the metabolic profile induced by various concentrations of trimethoprim with or without DHFR depletion. We saw changes in the concentration-dependent effects of several metabolites. In all cases, metabolic changes after DHFR depletion were similar to bacteria with WT levels of DHFR treated with subinhibitory concentrations of trimethoprim. In bacteria depleted of DHFR, trimethoprim induced changes similar to those seen in WT cells but at lower drug concentrations (representative metabolites are shown in Fig. 5B). Therefore, it seems that trimethoprim acts primarily as a DHFR inhibitor.

In our study, we found that depletion of drug targets results in varying effects. The cidal effects of reducing the level of InhA protein seen, albeit more slowly, with an inducible promoter substitution (17) suggest that, although isoniazid may target multiple enzymes (27–29), inhibition of InhA alone could account for antibiotic activity. The rapid cell death seen with InhA depletion and the growth cessation with even only mild depletion of RpoB suggest that these biochemical processes are hypersensitive to inhibition. In contrast, targets like DHFR and alanine racemase are present in considerable excess. Thus, drugs that target these enzymes have special properties. Trimethoprim, for example, induces high levels of inhibition, either because the compound is particularly potent or it achieves high local concentrations at its target.

Although many proteins are essential for bacterial growth and survival, these potential antibiotic targets vary widely in how much inhibition is required to block cellular replication or induce death. This is true even for the targets of known drugs. Proteins such as InhA or RpoB, whose activity needs only limited inhibition to block bacterial growth, are particularly attractive as targets for new drugs. In other cases, the target/compound pair has unique properties. For example, fluoroquinolone antibiotics not only inhibit DNA gyrase but also prevent religation of cleaved DNA, leading to double-stranded breaks (30). Similarly, antibiotics that affect protein synthesis can kill cells by blocking protein secretion (22). Such unique interactions can be critical for drug action but cannot easily be modeled in mutant bacterial strains. However, the genetic strategy described here is a useful approach to identifying targets that could be very effective targets for compounds that are pure inhibitors.

M. smegmatis strains were grown at 37 °C with shaking in Middlebrook 7H9 (Difco) supplemented with 0.2% glycerol, 0.05% Tween-80, 0.5% BSA, 0.2% dextrose, and 0.085% sodium chloride. Mycobacterial transformation was done as described (31), and transformants were grown on LB agar (Difco).

The final concentrations of antibiotics used for E. coli were 100 μg/mL for ampicillin, 150 μg/mL for hygromycin, and 50 μg/mL for kanamycin. For M. smegmatis, concentrations were 50 μg/mL for hygromycin and 20 μg/mL for kanamycin. ATc was added to cultures at a final concentration of 50 ng/mL unless otherwise stated.

The plasmids and primers used in this study are listed in Tables S1–S3.

Pfu Polymerase (Stratagene) or Phusion polymerase (New England Biolabs) was used for all PCR. Primers (Invitrogen), restriction enzymes (New England Biolabs), sequencing (GENEWIZ), synthetic genes (GenScript), anti–c-myc antibody (used as 1:10,000 dilution; Novus), and reagents for Western blots (ThermoScientific) were obtained from commercial sources. Anti-ClpP antibody was a gift from Olga Kandror and Alfred Goldberg.

Substrates were designed to have at least 120 bp chromosomal homology in each flank. For GyrA and RpoB, long primer pairs (gyrA-mycF1, gyrA-mycF2, gyrA-hygR1, gyrA-hygR2, rpoB-mycF1, rpoB-mycF2, rpoB-hygR1, and rpoB-hygR2) were used to generate substrate PCR products using two rounds of PCR from template plasmid pBSK::zeo-ID-hyg. For Alr, DHFR, InhA, and KasA, primer pairs (dfrAF, dfrAR, alrF, alrR, inhAF, inhAR, kasAF, and kasAR) were used to generate substrate PCR products from template plasmids pUC57::alr-ID-hyg, pUC57::dfrA-ID-hyg, pUC57::inhA-ID-hyg, and pUC57::kasA-ID-hyg, respectively.

To prepare competent cells for recombineering, a stationary culture of M. smegmatis with pNit::ET (GenBank no. GU459073) was diluted to 1:100, and the inducer isovaleronitrile was added to a final concentration of 10 μM. The culture was grown overnight to OD₆₀₀ = 0.8–1.0, and competent cells were prepared (31). Electroporation was performed as previously described (31). Positive clones (by Western blot) were plated on LB agar containing 10% sucrose to counterselect against the recombinase plasmid and then, were scored for growth on kanamycin-containing medium to confirm the loss of pNit::ET.

Bacterial strains were grown to stationary phase and then back-diluted to 1:100 in fresh medium; 150 μL culture were taken, and fluorescence was measured at 485/520 nm by Fluoroskan ascent FL (ThermoScientific). Results represent the median ± SD of triplicate samples.

We used a modification of the method described by Franzblau et al. (32). Briefly, Zeocin was serially diluted into Middlebrook 7H9 broth. Bacterial cultures were grown to midlog phase (OD_600nm = 0.5) and diluted to a calculated OD₆₀₀ = 0.001; 90 μL culture were added to each well. The leftmost column of the 96-well plate contained no antibiotic. Plates were sealed with breathable film (VWR) and incubated at 37 °C for 24 h. Alamar Blue (Biosource) reagent was added to each well, and plates were reincubated for an additional 12 h at 37 °C. The MIC was defined as the lowest drug concentration that prevented a color change from blue to pink.

Total protein lysates were prepared from equivalent cell numbers using bead beating, and Western blot analysis was performed as previously described (33). For Fig. 2, bacterial cultures were grown to stationary phase, diluted 1:20 with fresh medium with or without ATc, and then cultured for 12–15 h. For Fig. 3, cultures for Western blots were grown in parallel to those used for cfu determination as described in Fig. 3. Detection was performed using SuperSignal West Femto Chemiluminescent Substrate (ThermoScientific) according the manufacturer's protocol. In all cases, blots were stripped and reprobed with anti-ClpP to ensure equivalent loading of samples.

Strains were grown at 37 °C up to log phase (OD₆₀₀ ∼ 0.4). Cultures were diluted 1:100 into fresh medium with or without ATc. Strains with knockdowns of GyrA, KasA, and RpoB were preincubated with ATc for 12 h; the remainders were not preincubated. cfu was determined by plating and colony counting. Results represent mean ± SD for triplicate samples.

Bacterial cultures were grown to stationary phase, diluted 1:50 to fresh medium with or without ATc, and cultured for 12–15 h. On harvest, cell numbers were normalized using OD₆₀₀ determination. Cells were centrifuged and washed with 1 mL 1× DHFR assay (for DHFR) buffer or 100 mM Tricine/NaOH (pH 8.5; for alanine racemase assay), and then, they were resuspended in 300 μL same buffer. Cells were lysed by bead beating three times for 45 s. Protein concentrations were measured using the Bradford assay (Quick Start; Bio-Rad) to normalize the amount for enzyme activity assay. DHFR assay was performed using the Dihydrofolate Reductase Assay Kit (Sigma) according the manufacturer's protocol. Alanine racemase activity was measured spectrophotometrically at room temperature as described previously (34). Results represent mean ± SD for triplicate samples.

Bacterial cultures were first grown to OD = 1.0 and then inoculated onto 0.22-μm nitrocellulose filters under vacuum filtration. Filters with bacteria were then placed atop 7H10 agar with supplements and suitable supplements, such as ATc (50 ng/mL) or varying concentrations of trimethoprim, and allowed to grow at 37 °C. We determined cfu by scraping bacteria from the filter 24 h after inoculation followed by resuspension in 7H9 medium before quantitative culture. To determine metabolite concentrations, we first sequenced by plunging filters into acetonitrile/methanol/H₂O (40:40:20) precooled to −40 °C and then disrupted cells using bead beating with 0.1 mm Zirconia beads in a Precellys tissue homogenizer. Lysates were clarified by centrifugation and then filtered across a 0.22-mm filter. To normalize bacterial biomass of individual samples, we determined the residual protein content of metabolite extracts by Bradford assay. Liquid chromatography-MS for lysates was performed as previously described (35).