Cell Envelope Stress Response in Cell Wall-Deficient L-Forms of Bacillus subtilis

All strains used in this study (Table 1) are derivatives of strain PDC204 and can therefore be converted from rod-shaped cells to cell wall-deficient L-forms simply through cultivation on osmotically stabilized medium. The strain PDC134 was used for quantitative real-time PCR (qRT-PCR). TMB1075 and TMB1077 were used for determination of total protein concentration. TMB1080 and TMB1082 were used for gradient plates and β-galactosidase assays. To cultivate rod-shaped cells, strains were inoculated in selective nutrient broth (NB) or streaked onto selective nutrient agar (NA) supplemented with 0.5% xylose. The concentrations of antibiotics used in this study were as follows: erythromycin, 1 μg ml⁻¹; chloramphenicol, 5 μg ml⁻¹; and spectinomycin, 100 μg ml⁻¹. L-forms were induced and propagated under nonselective conditions in the absence of xylose in NA medium or on NB plates supplemented with 20 mM MgCl₂ and 0.5 M sucrose and buffered with 20 mM maleic acid (MSM medium). All strains were incubated at 30°C. In liquid culture, only rod-shaped cells were inoculated with aeration at 200 rpm.

Plasmids pTM1 (30) and pER603 (55) (Table 1) were used to generate promoter-lacZ reporter strains. PDC204 was transformed with ScaI-linearized plasmids (pTM1 and pER603) and chromosomal DNA of strain RM29 (sacB::spec) with chloramphenicol and spectinomycin selection, resulting in the strains TMB1080 and TMB1082, respectively. Transformation and preparation of chromosomal DNA was carried out as described elsewhere (27).

Gradient plates (41, 50) were used to compare the response of promoter-lacZ fusions as well as the antibiotic resistance of rod-shaped cells and L-forms. Mid-log-phase (optical density at 600 nm [OD₆₀₀] ∼ 0.4 to 0.6) cultures of rod-shaped cells were spotted on NA supplemented with 0.5% xylose. L-form cells were spotted on NA-MSM. The plates contained gradients from 0 μg ml⁻¹ to the following maximal antibiotic concentrations: for both P_bceA and P_liaI, bacitracin at 50 μg ml⁻¹ and mersacidin at 5 μg ml⁻¹; additionally for P_liaI, vancomycin at 0.0625 μg ml⁻¹, nisin at 10 μg ml⁻¹, and daptomycin at 0.15 μg ml⁻¹. Additionally, all gradient plates were supplemented with X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside; 100 μg ml⁻¹) to visualize the promoter activity of P_liaI and P_bceA.

The strains used for gradient plate experiments are PDC204 derivatives that carry an additional deletion of the sacB gene, which encodes levansucrase (Table 1). This extracellular enzyme is expressed in the presence of sucrose and is involved in the degradation of carbohydrates (61). During this study, we observed that L-form colonies growing on agar supplemented with sucrose are not well defined and show similarities to water droplets. Since cells with this colony morphology are very hard to handle on agar plates, we decided to knock out sacB by allelic replacement with a spectinomycin resistance cassette. This deletion resulted in the formation of well-defined colonies on agar plates.

For rod-shaped cells, the assay was performed as described previously (44). For L-forms, a modified procedure was developed in the course of this study. The OD₆₀₀ of mid-logarithmic-phase L-form cells of strains TMB1080 and TMB1082 was measured, the cultures were split, and one half was treated with bacitracin (50 μg ml⁻¹) and mersacidin (5 μg ml⁻¹). The other half of the cultures served as uninduced controls. After additional incubation for 4 h at 30°C with gentle shaking, 800 μl of each sample was harvested. The cell pellets were resuspended in 800 μl working buffer (60 mM Na₂HPO₄ · 2H₂O, 40 mM NaH₂PO₄ · H₂O, 10 mM KCl, 1 mM MgSO₄, 50 mM β-mercaptoethanol). This treatment reliably disrupted the L-form cells, and the samples could directly be used for the β-galactosidase assays. Nevertheless, lysozyme (which is necessary for the lysis of rod-shaped cells) was also added in this assay in order to directly compare the results for both cell types.

Strain PDC134 and liaI- and bceA-specific primers (Table 1) were used to perform qRT-PCR on total RNA, prepared from mid-logarithmic-phase rod-shaped and L-form cells after 30 min incubation in the presence of bacitracin and mersacidin, respectively (concentrations as listed above), as described previously (55). A reference value for the liaI and bceA transcription was obtained using the same strain without antibiotic treatment.

As a first test, we used antibiotic gradient diffusion plates (see Materials and Methods for details) to assess the inhibitory actions of selected cell wall antibiotics and their effects on the two biosensors, both before (rod-shaped cells) and after conversion of the strains into the L-form state. Mid-log-phase cultures of rod-shaped cells and L-forms were spotted in a line across the antibiotic gradient (Fig. 1). After incubation, the plates were visually inspected to determine the relative inhibitory effect of the tested antibiotics on the different strains, as indicated by the point of growth inhibition on the gradient. The addition of X-Gal to these plates also provided an indication for the induction of the reporter constructs in each strain by the development of a blue coloration of the patch of bacterial growth (Fig. 1). Since it is not possible to ensure an identical cell number's being initially spotted for the L-forms, this assay cannot be used as a quantitative measure of inhibitory action, at least for this cell type. Nevertheless, this easy-to-perform assay provided a first qualitative readout for the antimicrobial action of these compounds, due to the dramatic differences in sensitivities and regulatory response between the two strains.

While resistance against β-lactam antibiotics is well documented for L-forms, their response to antibiotics interfering with the membrane-anchored steps of cell wall biosynthesis, such as those used in this study, is less well understood. In the case of mersacidin and vancomycin, L-form colonies are fully viable, even at the highest antibiotic concentration applied (Fig. 1). While only moderate concentrations of vancomycin (maximum, 0.06 μg ml⁻¹) could be used in gradient plates and still allow growth of the highly sensitivity rod-shaped wild type, we verified that L-form cultures could survive in the presence of over 10 μg ml⁻¹ vancomycin, demonstrating a >100-fold-increased MIC for L-forms relative to the MIC for the wild type (data not shown). This dramatically increased resistance in the absence of a cell wall indicates that the target structure for both compounds is lacking or of no consequence to viability in L-forms, which is in good agreement with the respective MOA. In contrast, L-forms are dramatically more susceptible to daptomycin and nisin (Fig. 1), which both disrupt the integrity of the cytoplasmic membrane.

While L-forms are more resistant to bacitracin then their corresponding wild-type cells, they nevertheless retain a certain degree of susceptibility at higher antibiotic concentrations (Fig. 1). The primary inhibitory effect is a specific binding of bacitracin to undecaprenol pyrophosphate, thereby preventing its recycling to the monophosphate form, UP (66, 67). Hence, the cellular pool of UP, an essential precursor of the lipid II cycle, is rapidly depleted, thereby ultimately stopping cell wall biosynthesis. While this step is no longer essential in L-forms, our observation is in agreement with earlier reports demonstrating that bacitracin can also interfere with membrane integrity, i.e., by disrupting protoplasts (summarized in reference 46). However, the exact mechanism for this interference has not yet been elucidated.

In contrast, no significantly altered sensitivity was observed for the translational inhibitor kanamycin or the DNA-intercalating agent phleomycin (data not shown), demonstrating that the observed effects are indeed cell wall specific, as expected.

Taken together, our results demonstrate that the susceptibility of B. subtilis against antibiotics that inhibit membrane-anchored steps of cell wall biosynthesis is strongly affected. The hypersensitivity obtained in L-forms treated with nisin and daptomycin also indicates that L-forms represent a useful tool to distinguish between cell wall and membrane perturbation and to identify potential secondary MOAs. Moreover, our results suggest that both nisin and daptomycin might be potent drugs against persistent infections with pathogenic L-forms.

Since classical disk diffusion assays (10, 64) could not be successfully adapted for L-forms (data not shown), we instead modified the gradient plates described above for this purpose. Addition of the chromogenic β-galactosidase substrate X-Gal allowed direct assessment of the response of the LiaRS- and BceRS-dependent promoters in both the rod-shaped and L-form states as a function of the antibiotic concentration. In analogy to the disk diffusion assay, the strongest response of our wild-type biosensors was observed in colonies spotted at the highest tolerated concentrations (Fig. 1). Unfortunately, the intensity of the blue color was significantly weaker on gradient plates than in disk diffusion assays. Moreover, the strength of coloration does not directly correlate with induction strength but is instead reporter strain dependent, as noted in previous studies (11, 40, 64). Nevertheless, we observed a clear blue color for the L-forms containing P_bceA in the presence of suitable amounts of both bacitracin and mersacidin (Fig. 1A), which are known inducers of the Bce system (64). The induction of the P_liaI-derived reporter, which already provides a much weaker readout in disk diffusion assays (our unpublished data), is harder to detect on gradient plates. Nevertheless, a reproducible formation of light blue colonies at the highest tolerated concentrations can be observed for rod-shaped wild-type colonies in the presence of four of the five known inducers tested (Fig. 1B). In the case of daptomycin, the lack of a signal correlates with a lack of inhibition of rod-shaped cells even at the highest concentrations, indicative of insufficient antibiotic concentration. However, higher concentrations could not be used, since they completely prevented the growth of the corresponding L-form colonies (data not shown).

The BceR-dependent reporter strain TMB1082 showed comparable promoter activities in both wild-type and the corresponding L-form colonies in response to bacitracin and mersacidin (Fig. 1A). Interestingly, the minimal inducing concentrations for the two cell types were identical, irrespective of the significant differences in sensitivity, as described above. In contrast, no induction could be observed in L-form colonies of the LiaRS-dependent biosensor TMB1080 (Fig. 1B).

Our results therefore demonstrate that gradient plate experiments using the established biosensors P_bceA and P_liaI can be applied as a first comparative and qualitative screen for antibiotic induction, both in rod-shaped wild-type and in L-form colonies. Importantly, these initial results indicate that the response of the P_bceA-derived reporter strain is unaffected by the absence of a cell wall and therefore the damage caused by these antibiotics, indicative of a more direct sensing of the two compounds. In contrast, the P_liaI-based reporter no longer seems to respond to its known inducers in the L-forms, which can be interpreted as a more indirect, damage-induced sensing of antibiotic action. However, the low sensitivity in case of P_liaI of the assay and the narrow range of concentrations that can be tested are significant limitations of this method. Hence, a more sensitive and quantitative measure is necessary to support these initial observations.

B. subtilis strains carrying fusions of reporter genes to P_bceA and P_liaI have proven to be valuable biosensors for the study of antibiotics that interfere with the membrane-anchored steps of cell wall biosynthesis (8, 40, 64). The gradient plate experiments described above indicated that the classical lacZ-based reporter gene fusions can also be used in L-forms (Fig. 1). We therefore next aimed at establishing the classical β-galactosidase assay for liquid cultures of B. subtilis L forms to further develop and evaluate the use of promoter-based biosensors in this genetic background. This type of assay has traditionally been normalized by expressing the enzyme activities relative to cell density (44). Hence, we first verified that measuring the OD₆₀₀ is also a reliable measure for liquid L-form cultures by demonstrating a linear correlation of total protein concentration with optical density (data not shown).

Next, we investigated if β-galactosidase assays can be applied to L-forms to quantitatively monitor gene expression in these cell wall-deficient bacteria. For this purpose, we measured the promoter activity in strains TMB1080 (P_liaI) and TMB1082 (P_bceA) in both rod-shaped cells and their corresponding L-forms. As inducers, only bacitracin and mersacidin were chosen for these in-depth studies, since they are the only two compounds that activate both systems in wild-type cells (64). Samples of mid-logarithmic-phase cultures were split into three aliquots: one was an uninduced reference sample, and the other two were induced, one with bacitracin (50 μg ml⁻¹) and one with mersacidin (5 μg ml⁻¹). Because of the significantly reduced growth (and hence also protein production) rate of L-forms, it was critical to adjust the incubation time postinduction to 4 h (instead of the usual 30 min for wild-type cultures). This incubation time did not affect the basal promoter activity but resulted in significantly increased activities after antibiotic induction (data not shown). After 30 min or 4 h, the cells of wild-type or L-form cultures, respectively, were harvested and subjected to β-galactosidase assays as described in Materials and Methods. The results are shown in Fig. 2A. Rod-shaped wild type cells containing either P_bceA-lacZ (TMB1082) or P_liaI-lacZ (TMB1080) responded strongly to the presence of bacitracin and mersacidin, as expected (Fig. 2A) (64). In the corresponding L forms, the overall background activity was higher than in the wild type for both biosensors (Fig. 2A). No induction of P_liaI activity was observed in strain TMB1080 after 4 h of incubation in the presence of bacitracin and mersacidin. In contrast, TMB1082 (P_bceA-lacZ) showed five- and sevenfold-increased β-galactosidase activities for bacitracin and mersacidin, respectively (Fig. 2A). These data are in good agreement with the results obtained with gradient plates (Fig. 1). Inhibitors of translation did not induce either of the biosensors, as expected, thereby demonstrating the specificity of the responses in both rods and L-forms (data not shown).

To verify our data, we next analyzed the transcription of liaI and bceA directly by quantitative real-time RT-PCR, using RNA prepared from wild-type and L-form cultures of strain PDC134 after treatment of the cells for 30 min with bacitracin. RNA from an uninduced culture was prepared in parallel as a reference. Bacitracin strongly induced the expression of both genes in wild-type cells (>150- to 200-fold relative to the basal expression level) (Fig. 2B), as expected. Moreover, a clear induction of bceA expression was also observed in the L-form sample (∼40-fold). In contrast, no significant induction of liaI can be detected in L-forms containing P_liaI-lacZ. For TMB1080, we also tested the activity of P_liaI in the presence of other known inducers, such as vancomycin and nisin. Again, P_liaI responded to both antibiotics in the wild type, as described previously (40), while no induction of the liaI promoter was observed in the derived L-forms (data not shown). The lower induction ratio for bceA in the L-form samples was due to an overall higher basal signal, as was observed by β-galactosidase assay (Fig. 2A). The overall higher dynamic range determined by quantitative real-time RT-PCR is commonly observed, due to the higher sensitivity of this assay. Thus, the quantitative real-time RT-PCR results are in perfect agreement with the observations from both gradient plates and liquid β-galactosidase assays.

The hypersensitivity of L-forms to nisin seems somewhat surprising. Despite its known pore-forming bactericidal activity, nisin is known to require the pyrophosphate moiety of lipid II as a docking molecule to initiate efficient pore formation at low antibiotic concentrations (5, 7, 35), a molecule that is absent in our L-forms. On the other hand, lipid II-independent pore formation has also been observed at higher nisin concentrations and the presence of anionic lipids in the target membrane (5, 6, 22). Moreover, it was recently shown that nisin can also bind the precursor of wall teichoic acid biosynthesis (47). This molecule is also loaded onto the lipid carrier undecaprenol phosphate, formerly termed lipid IV, and flipped to the outer surface of the cytoplasmic membrane. While formation of wall teichoic acid cannot occur in the absence of peptidoglycan, a specific target molecule of nisin is therefore still present in our murE deletion strain.

Daptomycin has not been reported to require a component of the cell wall biosynthesis machinery for its inhibitory action (57, 62, 68). But both the exact MOA and hence the reason for the hypersusceptibility of B. subtilis L-forms is less clear. Our results seem to argue that the local daptomycin concentration on the outer surface of the cytoplasmic membrane is significantly increased in the absence of a cell wall to allow disruption of the cytoplasmic membrane even at very low external antibiotic concentrations. We therefore postulate that the Gram-positive cell wall seems to function as a molecular sieve that keeps the local daptomycin concentration low at its site of action.

An alternative explanation for the dramatic difference in daptomycin sensitivity between rod-shaped cells and L-forms could be a difference in membrane composition between the two cell types. This idea is based on three independent observations. First, phosphatidylglycerol levels in membranes have been identified as important determinants of the daptomycin susceptibility of B. subtilis (26). Second, it was shown previously that removing the cell wall of S. aureus in order to generate protoplasts significantly alters the membrane composition (77). Third, a very recent report on B. subtilis L-forms—based on the same strain that was used in this study—showed that a number of membrane lipids and their precursors, such as cardiolipin, phosphatidylethanolamine, unsaturated fatty acids, and lysyl-phosphatidylglycerol, are not essential for proliferation of cells in the L-form state. On the other hand, membrane fluidity determined by the incorporation of branched-chain fatty acids does play a crucial role in this process (42).

In order to determine the reason for the observed hypersensitivity, it might be worthwhile for a future study to determine the membrane composition of L-forms and their respective wild types. Both potential mechanisms described for daptomycin might of course also contribute to the observed hypersensitivity against nisin.

Daptomycin is the only compound for which no clear experimental evidence has so far been provided that links its activity to interference with cell wall biosynthesis. While such an activity has repeatedly been discussed (12, 48), our results can be viewed as the first, albeit indirect, evidence for such a MOA. It is generally accepted that the primary bactericidal activity of daptomycin is its interference with membrane integrity (12, 26, 48). Accordingly, L-forms show a significantly increased susceptibility in the absence of a protecting cell wall (Fig. 1B). Strong induction of liaIH expression by daptomycin, through activation of the LiaRS-dependent promoter P_liaI, was well documented by two independent studies (25, 73). But despite maintaining its biological potency, daptomycin fails to induce the P_liaI-lacZ biosensor in L-forms (Fig. 1B and data not shown). Therefore, the primary killing mechanism of daptomycin does not act as the stimulus sensed by the LiaRS system. Instead, this again requires an intact cell wall biosynthesis machinery. This observation strongly suggests that daptomycin—in addition to its primary MOA—also somehow interferes with (most likely some membrane-anchored step of) cell wall biosynthesis. This hypothesis is strongly supported by a recent report demonstrating that daptomycin distorts the cytoplasmic membrane in B. subtilis. This results in mislocalization of membrane-anchored proteins of the cell wall biosynthesis machinery, thereby also affecting cell wall biosynthesis and integrity (54). These data not only provide a convincing explanation for the observed induction of the cell envelope stress response in B. subtilis described above but also highlight the potential and predictive power of using the P_liaI- and P_bceA-derived biosensors in both the wild-type and L-forms of B. subtilis.

Taken together, our results reveal a fundamental difference in the mechanism of sensing cell wall antibiotics by two well-established biosensors, which are derived from the cell envelope stress-responsive promoters P_liaI and P_bceA of B. subtilis (40, 64). These new findings are summarized in a graphic model (Fig. 3). In rod-shaped wild-type cells that contain an intact cell wall, all five antibiotics exhibit their respective inhibitory activity, which directly or indirectly also interferes with the lipid II cycle of cell wall synthesis. In the case of mersacidin and bacitracin, this cycle is directly affected (3, 45) while vancomycin prevents the release of undecaprenol pyrophosphate by blocking the incorporation of new cell wall material into the existing peptidoglycan network (13, 63). Nisin uses lipid II as a docking molecule and thereby also interferes with this molecule (3). Daptomycin, on the other hand, primarily affects the integrity of the cell membrane and hence does not require cell wall biosynthesis for exhibiting its primary inhibitory function (57, 62, 68). All five antibiotics studied are known inducers of P_liaI activity in wild-type cells (40, 64). In L-forms, mersacidin and vancomycin completely lose their inhibitory activity, while only a reduced activity is seen for bacitracin. In contrast, nisin and daptomycin—which both perturb the cytoplasmic membrane—show a strongly increased potency (Fig. 1). Nevertheless, none of these five antibiotics is able to induce P_liaI activity in L-forms (Fig. 1B and data not shown). Hence, the compounds themselves cannot represent the true stimuli sensed by the LiaRS system. Instead, we conclude that this TCS requires the presence of a functional cell wall biosynthesis machinery to sense the five antibiotics addressed in this study. Our collective data strongly suggest that the true stimulus sensed by the LiaRS system is some aspect of interference with cell wall biosynthesis by those antibiotics. LiaRS is therefore a damage-sensing signal-transducing system. This hypothesis is supported by the wide and varying range of cell wall antibiotics sensed by the homologous systems in other members of the Firmicutes, such as VraSR of Staphylococcus aureus and CesSR of Lactococcus lactis. In addition to some of the compounds used in this study, these systems also respond to β-lactams or cationic antimicrobial peptides of the immune system (32, 38, 52). This rather general induction mechanism of LiaRS-like TCS is also reflected by their biological function: especially in Gram-positive cocci, these systems represent the primary but also rather general cell envelope stress response. They control large regulons that seem to provide general layers of protection in the presence of adverse conditions affecting the cell envelope (29, 59).

In contrast, the BceRS-dependent promoter P_bceA responds to its known inducers irrespective of the presence or absence of a cell wall, despite the fact that mersacidin and bacitracin have no and very little inhibitory activity in L-forms, respectively. The promoter responds in a concentration-dependent manner, and the regulatory sensitivity is unaffected by the lack of a cell wall in the L-forms and the resulting differences in resistance (Fig. 1A). This strongly argues for a direct drug-sensing mechanism of stimulus perception. This observation is in good agreement with the known mechanism of signal transduction for Bce-like detoxification modules. In contrast to classical TCS, the histidine kinase BceS and its homologs do not function as sensor proteins themselves. Instead, they recruit the transporter BceAB for this purpose (17, 23, 55). Since these ABC transporters therefore act as both sensors and highly efficient resistance determinants against peptide antibiotics (17, 23, 28, 64), such a direct sensing of the respective compound to be transported makes perfect sense.

The goal of this study was to establish β-galactosidase assays for the analysis of lacZ-based biosensors in B. subtilis L-forms and to evaluate such biosensors for cell wall antibiotics in the genetic background of strain PDC134. In this strain, the murE operon is under the transcriptional control of a xylose-dependent promoter. Using an osmotically stabilizing medium allows the conversion of rod-shaped (wild-type) cells into L-forms simply by omitting xylose and thereby depleting the cells of essential steps in cell wall biosynthesis (34).

This strain therefore allows a direct comparison of the response of cell wall-containing rod-shaped cells with that of cell wall-deficient L-forms when both types of cells are exposed to antibiotics that interfere with cell wall biosynthesis and the integrity of the cell envelope. By establishing β-galactosidase assays using these L-forms, it is now possible to analyze both the physiological response of the cells and the transcriptional activation of biosensors in parallel and directly compare the results obtained with wild-type and L-forms. This approach allows an unprecedented insight into antibiotic action and sensing at the same time, as demonstrated for a number of antibiotics interfering with the membrane-anchored steps of cell wall biosynthesis or membrane integrity. We believe that the PDC134-derived biosensors for cell wall antibiotics constitute a very useful addition to the toolbox currently available both for initial screening and subsequent MOA studies of novel antibiotics that target the cell envelope.

Moreover, our study highlights the potential of using daptomycin or type A lantibiotics such as nisin to combat persistent infections by bacterial L-forms, which have repeatedly been reported (43) and which are insensitive to many classical antibiotics such as the β-lactams or vancomycin.