Lethal synergy involving bicyclomycin: an approach for reviving old antibiotics

Bacterial strains, culture conditions and reagents

Bacterial strains [Table 1 and Table S1 (available as Supplementary data at JAC Online)] were grown at 37°C in LB liquid medium and on LB agar¹¹ purchased from Becton, Dickinson and Company (Sparks, MD, USA). Bicyclomycin was obtained from Schering-Plough Animal Health K. K. (Tokyo, Japan), ciprofloxacin was from Bayer Healthcare (West Haven, CT, USA) and streptolydigin was from Arkady Mustaev (PHRI, Newark, NJ, USA). Other compounds, including ampicillin, chloramphenicol, rifampicin, neomycin, tobramycin, doxycycline, tetracycline and tigecycline, were from Sigma Chemical Co. (St Louis, MO, USA).

Susceptibility testing

Inhibition of growth (MIC) was measured by broth dilution. Cultures were grown to A₆₀₀ = 0.3 (about 5 × 10⁸ cfu/mL) and diluted 5000-fold such that ∼10⁵ cells were applied to liquid medium (1 mL aliquots) containing drugs at various concentrations differing by 2-fold increments. The MIC (Table 1 and Table S1) was taken as the minimal concentration that blocked visible growth following overnight incubation. Lethal activity was measured by incubating exponentially growing liquid cultures (5 × 10⁸ cfu/mL) with drugs, followed by dilution, plating and incubation (24 h at 37°C) on drug-free agar to determine the number of surviving cells. That number, normalized to an untreated control sampled at the time of drug addition, was termed the percentage input recovered. Other antimicrobials were added to cultures 10 min before the bicyclomycin to assess their effect on bicyclomycin-mediated killing. The concentrations of these agents were chosen to give no killing when used alone (generally 2 × MIC), or in some cases a concentration series was examined to show a trend.

DNA analyses

The rate of DNA synthesis was measured by incubating Escherichia coli culture aliquots in the presence of [³H]thymidine (0.1 μCi/200 μL culture) for 2 min at 37°C followed by the addition of ice-cold 10% (w/v) trichloroacetic acid to precipitate high molecular weight DNA. Precipitates were collected on filter paper discs; precipitated radioactivity was determined by scintillation spectrometry.

An empirical measure of E. coli cell lysate viscosity¹² involved treating cells with lysozyme and non-ionic detergents, dividing cell lysates into aliquots in glass tubes, heating dilutions to 80°C for 2 min to unfold the chromosomal DNA, chilling the dilutions on ice and bringing the samples to 20°C in a water bath. A 0.025 mL glass microcapillary tube was then placed in each sample, and the time required to fill the capillary, less the time for buffer alone, was taken as an empirical measure of lysate viscosity. That value was determined for several DNA concentrations for a comparison of drug treatments.

Lethal synergy involving bicyclomycin and inhibitors of gene expression

We suspected that lethal aspects of bicyclomycin might be masked by drug-induced damage repair. We first confirmed that bicyclomycin by itself shows little rapid lethal activity. When E. coli was incubated for 2 h with bicyclomycin over a wide range of drug concentrations and then scored for survivors, almost no lethal activity was observed (Figure 1a). We then combined various concentrations of bicyclomycin with bacteriostatic concentrations of inhibitors of transcription (rifampicin) or translation (tetracycline, chloramphenicol; MICs listed in Table 1). The combinations reduced the survival of E. coli by 100-fold (Figure 1a; the absence of killing or growth at zero bicyclomycin in Figure 1a illustrates the absence of lethal activity of the gene expression inhibitors also present). Thus, bicyclomycin participates in lethal synergy.

To attribute bicyclomycin activity to the inhibition of Rho, we used phage P1-mediated transduction to construct a strain (KD3686) containing a point mutation in rho that confers bicyclomycin resistance (G337S).¹³ Bicyclomycin was bacteriostatic with the parental strain (KD3505; Figure 1b, open circles) unless tetracycline was also present to elicit lethal synergy (Figure 1b, filled circles). The bicyclomycin-resistant mutant (KD3686) grew in the presence of bicyclomycin alone (Figure 1b, open squares), showing that Rho is the intracellular target of the compound. The additional presence of tetracycline blocked the growth of the mutant but did not lead to cell death (Figure 1b, filled squares). Thus, Rho is the bicyclomycin target responsible for lethal synergy.

Lethal synergy due to bicyclomycin plus inhibitors of gene expression remained unchanged once the MIC threshold was exceeded (Figure 1a). Even increasing the bicyclomycin concentration to 10 × MIC had no effect on the rate of killing, observed as survival decreasing gradually over a 3 h period (Figure 1c). Thus, the combination of bicyclomycin plus tetracycline falls into the antimicrobial class called ‘time-dependent killers’.

We also examined other clinically important species for lethal synergy. With A. baumannii, bicyclomycin alone showed a small amount of bactericidal activity, reducing the survival by 10-fold; the addition of bacteriostatic inhibitors of gene expression reduced the survival by another 20- to 100-fold (Figure 2a). K. pneumoniae, Salmonella enterica and Shigella dysenteriae behaved much like A. baumannii and E. coli (Figure 2). Thus, an enhancement of bicyclomycin-mediated lethality by gene expression inhibitors appears to be a general phenomenon among bicyclomycin-susceptible, Gram-negative pathogens.

We next extended the work to inhibitors of protein synthesis that are by themselves moderately lethal. When we assessed the killing of E. coli by neomycin and tobramycin, the addition of bacteriostatic concentrations of bicyclomycin (2 × MIC) lowered the survival: bicyclomycin decreased the dose that killed 90% of the input cells (LD₉₀) from 15 to 3.7 mg/L and from 6.6 to 2.4 mg/L for neomycin and tobramycin, respectively (Figure 3a and b). With doxycycline and tigecycline, two tetracycline-class inhibitors of protein synthesis, bicyclomycin decreased the LD₉₀ from 20 to 1.2 mg/L and from >20 to 2.7 mg/L, respectively (Figure 3c and d). Even at bacteriostatic concentrations, all four protein synthesis inhibitors exhibited lethal synergy with bicyclomycin (Figure S1, available as Supplementary data at JAC Online). Thus, the stimulating effect of bicyclomycin on inhibitors of gene expression is a general phenomenon that can be used to achieve killing at low concentrations of each drug in the combination.

Independence of bicyclomycin-mediated lethal synergy from DNA replication, replication fork restart and DNA breakage

Bicyclomycin can promote DNA breakage arising from stalled transcription complexes colliding with replication forks when measured by very sensitive PFGE and PCR assays.^7,8 Since DNA breakage can be lethal,¹⁴ we expected that the inhibition of ongoing DNA synthesis would prevent such collisions, bicyclomycin-mediated DNA damage and subsequent lethal synergy. We were surprised to find that lethal synergy due to the combination of bicyclomycin and tetracycline was unaffected by a 20 min pretreatment with hydroxyurea, which blocks DNA synthesis at the concentration examined (100 mM, 2.5 × MIC; Figure 4a). Likewise, bacteriostatic concentrations of nalidixic acid (10 mg/L, 2 × MIC), another rapid inhibitor of DNA replication, also had no effect on lethal synergy (Figure 4b).

We next asked whether lethal synergy correlates with DNA breaks in cell lysates as detected by a viscometric assay that is sensitive to DNA size.¹² No difference was observed between lysates from untreated cells and those from cells treated with either bicyclomycin alone or bicyclomycin plus tetracycline. The assay readily detected DNA breaks introduced by nalidixic acid (Figure 4c) that are known to correlate with lethality.¹² Thus, we were unable to attribute lethal synergy to chromosomal DNA breakage. The observation that bicyclomycin alone failed to reveal extensive chromosome fragmentation was consistent with its inability to rapidly kill E. coli.

We also sought indirect evidence for an association between lethal synergy and DNA damage by examining a variety of mutations known to affect the efficiency of plating on bicyclomycin-containing agar^8,10 (this assay is similar to MIC determination and is designed to measure inhibition of growth, not killing). Mutations in priA and rep, genes involved in the restarting of replication following collisions between replication forks and transcription complexes, had little effect (Figure S2, available as Supplementary data at JAC Online). In another example, the deletion of rnhA or rnhB slightly decreased rather than increased lethal synergy (Figure S2). Consequently, it is unlikely that long DNA–RNA hybrids provide a major contribution to lethal synergy, even though deficiencies in RNase H, and therefore the accumulation of DNA–RNA hybrids in stalled transcription complexes, have been implicated in replication fork blockage.⁷ Neither a deficiency in recB nor a deficiency in recA affected the rapid killing arising from the combination of bicyclomycin and tetracycline (Figure S3, available as Supplementary data at JAC Online), although the recA deficiency lowered the bicyclomycin MIC by 2-fold (Table S1). The lexA-3 mutation, which blocks the induction of the SOS response and stimulates the bacteriostatic activity of bicyclomycin,⁸ also had little effect on lethal synergy (Figure S3C). Thus, bicyclomycin-mediated lethal synergy appears to differ mechanistically from the bicyclomycin-mediated growth inhibition previously attributed to DNA-related factors.

Relationship between transcription elongation and bicyclomycin-mediated lethal synergy

Our failure to demonstrate a role for perturbations of DNA metabolism in bicyclomycin-mediated lethal synergy made it imperative to examine the interference of transcription elongation associated with the inhibition of Rho by bicyclomycin.^7,15 As a test, we treated E. coli with streptolydigin, an inhibitor of transcription elongation, 5 min before bicyclomycin and tetracycline. Streptolydigin eliminated lethal synergy (Figure 5a). We also examined the effect of an rpoB*35 RNA polymerase mutant (H1244Q). This mutant, which fails to exhibit the pausing and subsequent back-tracking characteristic of transcription elongation complexes,⁷ also failed to exhibit lethal synergy (Figure 5b). Thus, active transcription elongation and a pausing of elongation complexes is required to form the lesions necessary for bicyclomycin-mediated lethal synergy.

Pretreatment with bicyclomycin interferes with lethal synergy

We next took an untargeted approach by asking whether bicyclomycin itself could block lethal synergy if applied before tetracycline. No lethal synergy was observed when E. coli was first treated with bicyclomycin at twice the MIC for 60–90 min before tetracycline was added (Figure 6a). Since pretreatment plus an additional bicyclomycin dose administered with tetracycline also showed no lethality (Figure 6a), it is unlikely that a loss of bicyclomycin activity during the pretreatment period was the explanation for the lack of lethal synergy. Instead it appears that bicyclomycin induces one or more novel factors that prevent it from killing bacteria.

To assess the stability of the putative protective factor(s), E. coli cultures were treated with bicyclomycin for 60 min, the drug was removed by centrifugation and the cells were resuspended in drug-free LB medium. Cultures were then diluted, grown in drug-free medium for various times and tested for lethal synergy. Even after 2 h of exponential growth, subsequent treatment with tetracycline plus bicyclomycin failed to exhibit lethal synergy (Figure 6b). Thus, bicyclomycin appears to induce a stable activity that nullifies the effect of bicyclomycin-induced lesions.