The B. subtilis ndoA (toxin) gene was inactivated by homologous recombination using a plasmid-borne deletion. No growth defect was evident when mutant cells were grown in LB medium (Fig. S1), nor did the toxin deficiency affect growth inhibition caused by chloramphenicol, ciprofloxacin, kanamycin, mitomycin C, moxifloxacin, or rifampicin, as measured by minimal inhibitory concentration (MIC; Table 1). This absence of a toxin effect on antimicrobial-mediated growth inhibition allowed lethal activity to be compared under various conditions without normalizing absolute drug concentrations to MIC (such normalization is often needed to minimize contributions from factors, such as efflux and drug uptake, that could affect antimicrobial lethality through changes in intracellular drug concentration).

To determine whether an ndoA deficiency is protective or detrimental during lethal stress, we first exposed cells to antimicrobial treatment. When wild-type and ndoA-deficient cells were treated with various concentrations of kanamycin for 1 hr (Fig. 1A) or with 4 µg/ml kanamycin (10 times MIC) for various periods of time (Fig. 1B), survival of the ndoA mutant was 10- to 20-fold greater than that of wild-type cells. These data indicate that in wild-type cells the NdoA toxin contributes to the lethal activity of kanamycin. Protection from kanamycin-mediated killing by the ndoA deficiency was eliminated when a copy of ndoA was integrated into the amyE locus of the ndoA-deficient strain (Fig. 1A, open squares). Thus, the protective phenotype observed was due to the absence of the ndoA gene product rather than to cis-acting effects arising from insertion of an antibiotic marker into the ndoA locus.

To determine whether a similar phenomenon occurs with other antimicrobials, we treated wild-type and toxin (ndoA)-deficient cells with various concentrations of the fluoroquinolone moxifloxacin. The ndoA deficiency caused survival to increase by about 10-fold (Fig. 1C). The toxin deficiency also increased survival by 10- to 20-fold when B. subtilis was exposed to moxifloxacin for various times at 16-times MIC (Fig. 1D). Complementation of the protection afforded by the ndoA deficiency to the lethal action of moxifloxacin was observed when an intact copy of ndoA was integrated at the chromosomal amyE locus (Fig. 1C). The ndoA deficiency also increased survival during treatment with ciprofloxacin, oxolinic acid, and mitomycin C (Fig. S2). Thus, in wild-type cells the NdoA toxin increases the lethal action of several diverse antimicrobials.

To determine whether an ndoA deficiency also affects the bacterial response to commonly encountered environmental stressors, we examined effects of toxin deficiency on the lethal action of oxidative stress, UV irradiation, and temperature. When wild-type and toxin-deficient cells were compared for survival after a 15-min exposure to various concentrations of hydrogen peroxide, the ndoA deficiency increased survival by 5–30 fold (Fig. 2A). An even greater protective effect (up to 100-fold) was observed when cells were treated with 250 mM hydrogen peroxide for various times (Fig. 2B).

Exposure to UV irradiation at different levels revealed opposing effects of NdoA function. The mutant was hypersensitive to low levels of UV irradiation (0.14 mW/cm², Fig. 3A). But increasing the irradiance intensity to 0.4 mW/cm² showed that the mutant was less sensitive than wild-type cells at high UV exposures (Fig. 3B). At moderate irradiance intensity (0.2 mW/cm²), the ndoA mutant was 5–10 fold more sensitive than wild-type cells with short exposure times (10 to 30 sec), but the mutant became less sensitive when exposure was longer (beyond 30 sec; Fig. 3C). This cross-over from being more sensitive to less sensitive suggests that the effect of the toxin may depend on stress level, at least with some types of stress, such as UV irradiation. Interestingly, the crossover points seen in the 3 panels of Fig. 3 all occurred at about the same level of accumulated energy (e.g. 7–8 milliJoules/cm², Figure S3), which indicated that a threshold in accumulated energy rather than in irradiance intensity or exposure time determines whether the NdoA toxin exhibits a protective or detrimental effect.

When B. subtilis was incubated at temperatures ranging from 50°C to 55°C for 20 min, the colony-forming capability of the ndoA mutant was 5- to 10-fold lower than that of the wild-type strain (Fig. 4A). A slightly greater hypersensitivity was observed when cells were incubated at 52°C for various times (Fig. 4B). No cross-over from the hypersensitive state to a protected state was observed for the mutant despite examination of several combinations of temperature and incubation time (not shown). Thus, wild-type toxin exerts only a protective effect with some types of stress, such as heat, unlike the dual action seen with UV irradiation and the detrimental effect shown for antimicrobials.

Since a MazF-like toxin has been implicated in myxobacterial fruiting body formation, a developmental process that leads to mother cell lysis and spore generation ^[23], we next examined the effect of the ndoA mutation on sporulation. B. subtilis cultures were incubated in Difco sporulation medium, and aliquots taken at various times were treated for 20 min at 90°C to kill vegetative cells. Spore number increased gradually as incubation time increased from 12 to 36 hr, with the wild-type culture forming 10–40 fold more spores than the ndoA mutant (Fig. 5, filled symbols). No difference in total cell number was evident during the first 36 hr of incubation (Fig. 5, open symbols). When incubation was extended to longer times, spore number stopped increasing with both wild-type and mutant cells. However, after incubation for 48 hr, the total bacterial number (CFU before heat treatment) declined (by 20–30 fold) with the ndoA mutant but not with the wild-type cells (Fig. 5, open squares). These data indicate that NdoA contributes not only to sporulation, but also to long-term survival under nutrient-limiting conditions.

In work described above, the presence of the NdoA toxin enhanced the lethal effects of kanamycin and moxifloxacin (Fig. 1). Since the lethal action of several antimicrobial agents is associated with production of reactive oxygen species ^[24], ^[25], ^[26], ^[27], we next measured peroxide accumulation associated with kanamycin treatment. As shown in Fig. 6, kanamycin exposure resulted in an NdoA-dependent surge of hydrogen peroxide; the absence of the toxin eliminated the surge (Fig. 6). Similar results were obtained with moxifloxacin-treated cells (not shown). Thus, the lethality-enhancing effects of the NdoA toxin correlate with peroxide accumulation.

The B. subtilis NdoAI-NdoA toxin-antitoxin module can either enhance or diminish the effects of lethal stress depending on the type and level of stress. The toxin increases the lethal action of antimicrobials, such as kanamycin and moxifloxacin, while it decreases killing by heat and nutrient starvation. With UV irradiation, the toxin reduces killing at low doses but increases it at high doses.

Having two effects fits data reported for the MazEF TA module of E. coli. Thus, bacteria appear to have pathways that allow cells to recover from mild stress but undergo ROS-mediated death when stress is severe.

Rifampicin, erythromycin, oxolinic acid, mitomycin C, horseradish peroxidase, 10-acetyl-3,7-dihydroxyphenoxazine (Amplex Red), and hydrogen peroxide were obtained from Sigma Chemical Co. (St. Louis, MO). Ciprofloxacin and moxifloxacin were provided by Bayer AG (West haven, CT). Kanamycin was purchased from Fisher Scientific (Fairlawn, NJ), chloramphenicol was from EMD Chemicals, Inc. (Gibbstown, NJ), and ampicillin was a product of Roche Diagnostics (Indianapolis, IN).

Strains of B. subtilis used in the work were derivatives of strain BD630 obtained from Dr. D. Dubnau, PHRI. Strain 3169 was deleted for ndoA and was constructed as described below. Strain 3322 was a derivative of strain 3169 in which a copy of native ndoA was integrated into the amyE locus as described below. B. subtilis was grown at 37°C in LB broth with shaking at 250 rpm or on LB agar ^[43] unless specified otherwise.

Plasmid pBS-NdoA-Ery, which contains an erythromycin (ery)-resistance cassette bounded by homologous regions upstream and downstream of ndoA, was used to inactivate the toxin gene (ndoA) by homologous recombination. The plasmid was obtained from Ciaran Condon ^[21] and was introduced into wild-type B. subtilis strain BD630 by bacterial transformation to generate the ndoA mutant (strain 3169). The deficiency of the targeted ndoA gene in the mutant was confirmed by PCR-mediated amplification of nucleotides between the upstream gene alr and the ery cassette, using primers Alr-F ([gene: 5′-CAT GGA TGA GAT TGC AGG AAG-3′]) and Ery-R ([gene: 5′-GAG GTG TAG CAT GTC TCA TTG-3′]) followed by gel electrophoresis analysis of product size and by nucleotide sequence determination as in ^[21].

Growth inhibition (MIC) was determined by broth dilution using a series of tubes, each containing about 10⁵ bacteria in 1 ml of LB liquid medium supplemented with antimicrobials at various concentrations that differed by 2-fold increments. The lowest drug concentration that prevented visible growth after overnight incubation was taken as MIC.

Rapid killing of bacterial cells was measured with cultures that had been grown overnight (about 17 hrs) and then diluted by 50-fold before treatment with various stressors for various times. After treatment, cells were diluted with LB medium and applied to LB agar plates lacking stressor. Colonies were counted after overnight incubation at 37°C. Untreated cells, taken at the time stress was applied, served as a time-zero control; survival was calculated relative to CFU at time zero.

The ndoAI-ndoA operon containing a 484-bp upstream promoter region was amplified by PCR and cloned into plasmid pDR66 ^[44] between the Bam HI and Eco RV restriction sites. This strategy allows removal of the P_spac promoter from pDR66 and placement of a native copy of the ndoAI-ndoA operon between two DNA segments of the amyE locus. A single copy of ndoAI-ndoA was then integrated into the chromosomal amyE locus via homologous recombination. Confirmation of correct integration of ndoA into the amyE locus of the ndoA mutant (strain 3169) was achieved by PCR (using one primer derived from the pDR66 vector and another primer from the end of ndoA) and nucleotide sequence determination. The newly constructed strain (3322), along with the wild-type (BD630) and the ndoA mutant (strain 3169), were examined for susceptibility to the lethal action of kanamycin and moxifloxicin.

Overnight cultures were diluted by 10¹ to 10⁶-fold with fresh LB medium, and 3 aliquots of 10 µl for each dilution were spotted on LB agar plates. The plates were exposed in the dark to a UV (254 nm) lamp (model G18T8, General Electric Co, Cleveland, Ohio) at irradiance intensity of 0.14, 0.20, or 0.40 mW/cm² for various times (10 to 60 sec). The plates were then incubated overnight at 37°C in the dark. Survival was determined relative to CFU measured without UV treatment.

A single B. subtilis colony grown for 24-hr was inoculated into Difco sporulation medium ^[45] and incubated at 37°C with shaking at 300 rpm for various times. Aliquots were removed periodically, and half of each sample was incubated at 90°C for 20 min to kill vegetative cells. Samples were diluted and plated on LB agar for determination of spore number. The other half of each sample was diluted and plated without heating for determination of total bacterial counts. Sporulation frequency was taken as the ratio of CFU after heat treatment to CFU before heat treatment.

Bacteria were treated with 4 µg/ml (10 MIC) kanamycin for various times after which cells were harvested by centrifugation at 10,000× g for 5 min. Cells were resuspended in 0.9% NaCl, and centrifugation was repeated. Hydrogen peroxide concentration was measured as described ^[46] by horseradish peroxidase- (4 units) and H₂O₂-mediated oxidation of 50 µM 10-acetyl-3,7-dihydroxyphenoxazine (Amplex Red) to the highly fluorescent compound resorufin ^[47]. Fluorescence was measured using a Spectramax M2 microplate reader (Molecular Devices, Sunnyvale, CA) at 587 nm with excitation at 563 nm.