Cell division in almost all bacteria is orchestrated by the essential tubulin homologue FtsZ, which assembles into a ring-like structure and acts as a scaffold for the division machinery. Division was recently validated as an important target for antibiotics by the demonstration that low-molecular-weight inhibitors of FtsZ, called benzamides, can cure mice infected with Staphylococcus aureus. In treated cells of Bacillus subtilis we show that FtsZ assembles into foci throughout the cell, including abnormal locations at the cell poles and over the nucleoid. These foci are not inactive aggregates because they remain dynamic, turning over almost as rapidly as untreated polymers. Remarkably, although division is completely blocked, the foci efficiently recruit division proteins that normally co-assemble with FtsZ. However, they show no affinity for components of the Min or Nucleoid occlusion systems. In vitro, the benzamides strongly promote the polymerization of FtsZ, into hyperstable polymers, which are highly curved. Importantly, even at low concentrations, benzamides transform the structure of the Z ring, resulting in abnormal helical cell division events. We propose that benzamides act principally by promoting an FtsZ protomer conformation that is incompatible with a higher-order level of assembly needed to make a division ring.
Cell division in nearly all bacteria is orchestrated by the essential tubulin homologue, FtsZ, which assembles at mid-cell into a contractile ring-like structure known as the Z ring (Bi and Lutkenhaus, 1991; Erickson, 1995). The Z ring acts as a scaffold for the assembly of a multi-protein complex called the divisome (Harry et al., 2006). This protein machine drives and directs the inward growth of the various cell envelope layers, resulting in the formation of two new cell poles and separation of the daughter cells. In the Gram-positive bacterium, Bacillus subtilis, the Z ring assembles early in the cell cycle (Gamba et al., 2009). Simultaneously, several non-essential accessory proteins (FtsA, ZapA, EzrA and SepF) involved in regulating the dynamics and structure of the Z ring assemble at mid-cell (Levin et al., 1999; Gueiros-Filho and Losick, 2002; Hamoen et al., 2006; Ishikawa et al., 2006; Adams and Errington, 2009). The Z ring then persists for a significant proportion of the cell cycle (∼20%), before the concerted and interdependent assembly of the membrane proteins (DivIC, DivIB, FtsL, PBP2B, FtsW) that are required for septal peptidoglycan synthesis (Gamba et al., 2009). Once assembled, the Z ring constricts, and appears to remain at the leading edge of the nascent septum throughout division.
The Z ring is not a static structure and instead exists in a constant state of flux, with individual subunits continuously exchanging between those in the Z ring and those in the cytoplasmic pool (Stricker et al., 2002; Anderson et al., 2004). This exchange is rapid, on the order of seconds, and is dependent on both the intrinsic GTPase activity of the polymer and the presence of regulatory proteins (e.g. EzrA) at the Z ring (Anderson et al., 2004). Once in place, the Z ring remains sensitive to various regulatory proteins, for example UgtP, which co-ordinates Z ring completion with nutrient availability (Weart et al., 2007).
FtsZ is a self-assembling GTPase that polymerizes in a GTP-dependent manner, through the head to tail association of individual subunits (de Boer et al., 1992; RayChaudhuri and Park, 1992; Mukherjee and Lutkenhaus, 1994; Erickson et al., 1996). The active site is formed at the subunit interface, by insertion of catalytic residues from the base of one subunit, into the nucleotide binding site of the subunit below (Scheffers et al., 2002; Oliva et al., 2004). In vitro, FtsZ assembles cooperatively into a variety of polymeric forms depending on the assembly conditions employed (Adams and Errington, 2009). However, despite intensive study, the nature of the physiologically relevant polymer remains unknown, and the ultrastructure of the Z ring in vivo is poorly understood. Recently, though, tomography of Caulobacter crescentus cells has suggested that rather than forming a continuous ring, the Z ring may be composed of short overlapping protofilaments (Li et al., 2007). Additionally, much work has been devoted to understanding the mechanism by which the Z ring constricts throughout division (Erickson, 2009). The observation of the Z ring as a discontinuous network of protofilaments in vivo led to the suggestion that it may be able to transmit force to the membrane by iterative rounds of transition between straight and curved protofilament forms (Li et al., 2007). This model is well supported by two recent groundbreaking reports. First, that membrane tethered FtsZ can itself assemble into contractile Z rings within tubular liposomes (Osawa et al., 2008), and second, that curved FtsZ protofilaments generate these forces (Osawa et al., 2009).
Bacterial cell division has long been recognized as an attractive antibacterial target (Payne et al., 2007). Its near universal conservation in bacteria and absence from humans, combined with the essential role of Z ring assembly in cell division, has resulted in FtsZ being the focus of most attention. Indeed, a diverse range of compounds have been shown to inhibit its polymerization in vitro, and some have antibacterial activity against cells in culture (Vollmer, 2006; Lock and Harry, 2008). Importantly, one such compound, PC190723, a member of the benzamide family of compounds, was shown to be efficacious in a mouse model of Staphylococcus aureus infection, thus validating bacterial cell division as a novel antibacterial target (Haydon et al., 2008). The study of resistant mutants and structural modelling revealed the likely compound binding site: a hydrophobic channel and shallow cleft on the surface of the C-terminal domain of FtsZ, lying against the central core helix (Löwe and Amos, 1998; Haydon et al., 2008). To date this region has no ascribed function, but interestingly, it is similar to the Taxol binding site found in the same region of β-tubulin (Nogales et al., 1998; Amos and Löwe, 1999). The accessibility of this cleft, which varies between organisms, appears in part to be the source of the selectivity of these compounds for Bacillus, Staphylococcus and close relatives (Haydon et al., 2008).
In order to further understand the mode of action of this compound series and to shed new light on Z ring assembly and its role in cell division, we have conducted a detailed cytological and biochemical analysis of a closely related but more potent analogue of PC190723, compound 8j [minimal inhibitory concentration (MIC) = 0.25 µg ml−1] (Fig. S1) (Haydon et al., 2010).
An immediate block in cell division
Previous work on PC190723 had established that it causes FtsZ to localize into punctate foci (Haydon et al., 2008). To study the cytological effects of these compounds in more detail a GFP fusion to FtsZ (GFP–FtsZ), expressed ectopically in a merodiploid background, was examined by epifluorescence microscopy. Prior to treatment cells displayed the characteristic localization pattern of FtsZ: mid-cell bands and dots, representing the various stages of Z ring assembly and constriction. Importantly, DMSO, into which the compound is dissolved, had no effect on FtsZ localization (Fig. 1A). However, in cells treated with compound 8j the localization of FtsZ was rapidly perturbed, with spots of fluorescence visible throughout the cell (Fig. 1B). As the cells elongated in the absence of division the number of foci continued to increase (Fig. 1C–E). DNA replication and segregation continued unaffected (Fig. 1H). Importantly though, the long filamentous cells were unstable and prone to lysis, leading to a gradual reduction in optical density (Fig. 1F), and a large reduction in viability, within a few hours (Fig. 1G).
The small foci appeared to be distributed randomly at the cell periphery and were apparently immobile, hinting that they are membrane associated. Three-dimensional reconstructions (Movie S1) showed that the foci are discrete structures, located close to the cell periphery and are probably not part of a higher-order (e.g. helical) structure. The frequent presence of foci at the cell poles and in the vicinity of the nucleoid (Fig. S2) suggests that they are insensitive to the negative spatial regulators MinC and Noc, which normally act to prevent FtsZ assembly in these regions. Intriguingly, pre-existing Z rings seemed to remain more or less intact, and persisted for approximately 60 min (Fig. 1B and C), after which they were no longer observed, suggesting that they had disassembled (Fig. 1D and E). These localization data suggest first that the compound stabilizes FtsZ polymers and prevents nascent Z ring assembly (and disassembly), and second that it works via a mechanism distinct from that of previously described chemical inhibitors of FtsZ, which resulted in rapid dissipation of the Z ring (Stokes et al., 2005).
Compound induced foci remain dynamic
To gain further insights into the nature of the compound induced FtsZ foci, and to ask whether they are inactive aggregates or if they remain dynamic, we examined the behaviour of GFP-labelled FtsZ in vivo using fluorescence recovery after photobleaching (FRAP). In agreement with previous work (Anderson et al., 2004), we found that the Z ring is a highly dynamic structure (Fig. 2A and C); if anything in our experiments recovery was even more rapid (half-time 3.7 ± 0.3 s; n = 18) than reported previously. Following growth for 1 h in the presence of 8j (2 µg ml−1), we selectively photobleached individual compound induced FtsZ foci and monitored their recovery. Surprisingly, these foci did recover (Fig. 2B and D), albeit at an approximately threefold slower rate than untreated cells (half-time 11 ± 1 s; n = 20). This was important because it showed that the foci were not simply inactive protein aggregates. Even with a 10-fold excess of the compound, the FtsZ foci remained dynamic (half-time 12 ± 1 s; n = 8). To ensure that ongoing protein synthesis did not influence these results, whole cells were bleached and their recovery was monitored over time. These cells did not exhibit any significant recovery within the time frame (60 s) of the experiment (data not shown).
Downstream components of the divisome assemble at FtsZ foci
In previous experiments examining the effects of FtsZ depletion or inhibition, other division proteins were usually found to be dispersed in either the cytoplasm or the cell membrane (as appropriate for their trans-membrane topology) (Stokes et al., 2005). Untreated (DMSO only) cells bearing fluorescent fusions to various early or late division proteins showed the expected banded patterns of localization (Fig. 3A–H, top panels). However, in compound-treated cells, bands were largely eliminated. Remarkably, for all the fusions examined, to both early and late division proteins, an irregular spotty pattern was observed, which was indistinguishable from the pattern observed for FtsZ (Fig. 3A–H, bottom panels). This suggested that the downstream division proteins are recruited to the FtsZ foci, even though the normal ring-shaped structure has not been formed. As the later assembling components are all integral membrane proteins, the ability of the FtsZ foci to recruit these proteins readily explains the observation that the foci are membrane associated.
To test whether the foci did indeed represent assemblies of the complete set of division proteins we employed three different approaches. First, we examined cells doubly labelled with FtsZ–CFP and either YFP–ZapA as a marker for the early division proteins (Fig. 3I) or YFP–DivIC as a marker for the late division proteins (Fig. 3J). In both cases the foci formed following compound treatment colocalized completely with FtsZ foci. Second, following in vivo cross-linking, we affinity-purified protein complexes from cells bearing a single, functional, histidine tagged copy of ftsZ, as described previously (Ishikawa et al., 2006). Visualization of the proteins present in each sample revealed no obvious difference between those formed in the absence or presence of compound (Fig. S3). Western blotting revealed the presence of early (EzrA–GFP; Fig. 3K) and late (PBP2B; Fig. 3L) division proteins in the pull-downs. Furthermore, if anything, in the presence of compound we reproducibly observed an increased recovery of the target protein (Fig. 3K and L).
Third, to investigate the rapid perturbation of FtsZ localization we used a cell wall-less derivative of B. subtilis, known as an L-form (Leaver et al., 2009). These cells divide independently of FtsZ via an unknown mechanism. However, ftsZ is still present and expressed, although the protein does not form Z rings (Leaver et al., 2009). Using this Z ring-free system we visualized GFP–FtsZ, which was expressed ectopically from its native promoter. We found that in these cells, FtsZ was, on the whole, diffuse and few large structures were observed (Fig. S4A). However, when compound was added the localization of FtsZ in the same cells changed rapidly, with the formation of multiple discrete foci visible within 1 min (Fig. S4A). Moreover, the late division protein DivIC behaved in an identical manner (Fig. S4B). As expected, the addition of DMSO had no effect on protein localization. These experiments confirmed that the effect of the compound is rapid and support the idea that FtsZ foci can recruit downstream division proteins even when it has not formed a ring structure.
Spatial regulatory proteins do not form compound induced foci
It was curious that the compound induced FtsZ foci occupied locations (cell poles and over the nucleoid) from which Z ring assembly is normally excluded (by the Min and Noc systems respectively). We therefore examined the localization of these factors in compound-treated cells. The various components of the Min system generally localize at mid-cell, arriving late, after the cell has become committed to or has actually initiated division, and at mature cell poles (Marston et al., 1998; Gregory et al., 2008; Bramkamp and van Baarle, 2009; Lenarcic et al., 2009). In untreated cells, GFP fusions to MinC, MinD and DivIVA all showed the expected patterns of localization: mid-cell and poles (Fig. 4A, C and E). In contrast to the multiple dispersed foci seen for the proteins of the division machinery, in the elongated treated cells, all three proteins were mainly located at the cell poles (Fig. 4B, D and F).
Noc protein has a quite different localization pattern, mainly coinciding with the nucleoid but with spots at the periphery of the cell overlying the nucleoid (Wu et al., 2009) (Fig. 4G). Again, treatment with compound had no obvious effect on the pattern of Noc–YFP localization (Fig. 4H). Moreover, in the affinity pull-down assays using FtsZ–His we did not detect Noc in either the absence or the presence of the compound (Fig. 3M).
Foci do not possess detectable cell wall synthetic activity
That the compound induced foci contain the majority of cell division proteins normally assembled at mid-cell raised the possibility that they may also possess cell wall synthetic activity. To address this question we stained cells using fluorescent vancomycin (Van-FL), which is thought to label the sites of nascent peptidoglycan synthesis (Daniel and Errington, 2003). In B. subtilis, this results in a helical pattern of staining in the cylindrical part of the cell, as well as intense staining at sites of septal biogenesis (Fig. S5A). Importantly, the cell poles do not stain using this technique and are thought to be inert (Daniel and Errington, 2003). Following growth with compound, live cells were stained with Van-FL and found to exhibit the normal helical pattern of staining in the cylindrical part of the cell (Fig. S5B). And, as expected, given the cell division block, no septal-like staining was observed. However, no extraneous sites of peptidoglycan synthesis were observed and crucially, the cell poles, at which foci are readily found, appeared to remain inert (Fig. S5B). Additionally, thin section and scanning electron microscopy of fixed cells showed no evidence of inappropriate cell wall synthesis and the cell wall appeared unperturbed (Fig. S5C–F). These results indicate that the compound induced foci of division proteins do not posses significant cell wall synthetic activity.
Compound 8j enhances FtsZ polymerization in vitro
To investigate the effect of 8j on the assembly of FtsZ, we purified native, full-length, B. subtilis FtsZ. Andreu et al. recently presented a comprehensive analysis of the effects of PC190723 on FtsZ assembly in vitro (Andreu et al., 2010). As expected, given the chemical similarity between the two compounds, the effects of 8j on FtsZ assembly presented here are in good agreement with those of Andreu et al.
Using ultracentrifugation, FtsZ sediments efficiently, in a GTP-dependent manner (Fig. 5A). However, when 8j was included in the assembly reaction we observed a potent, dose-dependent enhancement of sedimentation (Fig. 5A). Maximal stimulation of sedimentation was achieved at 20 µg ml−1, with almost all FtsZ recovered in the pellet fraction. Importantly, the ability of the compound to enhance the sedimentation of FtsZ was reproducible under more physiological conditions, at both low (Fig. 5B) and high (Fig. 5C) salt concentrations. In all cases the effects of compound 8j were strictly GTP dependent (Fig. 5A–C). DMSO had no appreciable effect on sedimentation under any of the conditions employed.
To quantify the observed enhancement of assembly and to gain real-time data on the dynamics of the polymerization reaction, we employed 90° angle light scattering. Under conditions of reversible assembly, FtsZ assembled in a GTP-dependent manner, reaching a quasi-steady state after approximately 5 min. As reported previously, as the polymers hydrolyse their nucleotide, subunits are released and continue to recycle with fresh GTP from solution until it is exhausted. At this point subunit release predominates and the polymers disassemble, returning the light scattering signal to baseline within 15 min (Fig. 5D).
We then assayed the effects of increasing concentrations of compound. Low levels of compound (≤ 0.2 µg ml−1) resulted in a modest but reproducible increase in polymerization, as well as delayed disassembly (Fig. 5D). Interestingly, although polymer disassembly was evident, the signal did not completely return to the baseline, suggesting that some polymers persist after bulk disassembly. Higher concentrations of compound (> 0.2 µg ml−1), however, had a dramatic effect on the assembly reaction (Fig. 5D). First, the extent of polymerization was markedly increased, with a maximal signal of ∼3× that of the control reaction. Furthermore, disassembly did not occur within the timescale of the assay. This, taken together with the delayed disassembly at low concentrations of compound, suggested that the polymers were stabilized.
When the relative stability of these polymers was assayed over a longer period, we found that they were indeed hyperstable and exhibited a dose-dependent increase in stability (Fig. S6). Furthermore, the addition of excess GDP to polymers pre-assembled with compound did not stimulate their complete disassembly, reinforcing the notion that the polymers are more stable (Fig. S7). To rule out the possibility that aggregation was the source of the sustained signal we confirmed that discrete polymers were still extant using electron microscopy (results below and data not shown). Finally, the timing of compound addition to the assembly reaction was not important, and polymerization was stimulated when compound was added either before or after GTP (Fig. S8).
Compound 8j inhibits FtsZ GTPase activity
In agreement with previous work showing that PC190723 inhibits the GTPase activity of FtsZ (Haydon et al., 2008; Andreu et al., 2010), we found that 8j inhibited the GTPase activity of B. subtilis FtsZ in a dose-dependent manner (Fig. 5E). However, while inhibition was clearly evident, it was incomplete, with a maximal sixfold reduction in GTPase activity achieved at the highest concentration of compound tested (Fig. 5E). A competition assay, with increasing amounts of GTP, revealed that the observed inhibition was not competitive (data not shown), consistent with the genetic and molecular docking evidence that the compound binding site is different from that of the nucleotide (Haydon et al., 2008). These data indicate that the suppression of FtsZ dynamics is likely due to the inhibition of GTPase activity. However, they do not discriminate between direct or indirect effects on GTPase activity (see below).
Polymers assembled with compound 8j are highly curved
Previously, factors known to modulate the assembly of FtsZ have often been shown to alter the morphology of the assembled polymer. The regulatory protein, ZapA, the divalent cation, Ca2+, and the chemical inhibitor, Zantrin Z3, are three such examples (Yu and Margolin, 1997; Gueiros-Filho and Losick, 2002; Margalit et al., 2004). All enhance the polymerization of FtsZ in vitro, stimulating the assembly of FtsZ into multi-stranded polymer bundles. To examine whether the ability of compound 8j to stimulate and stabilize FtsZ polymerization was accompanied by changes in polymer morphology, we examined the assembled polymer by transmission electron microscopy.
Under physiologically relevant conditions, FtsZ readily assembles into single-stranded protofilaments, which are on the whole straight, although a gentle degree of curvature is often observed (Fig. 6A). Additionally, as previously reported, DMSO has no effect on polymer morphology (not shown). However, in the presence of compound, polymer morphology was clearly perturbed. While protofilaments remained single stranded and were not seen to form any higher-order structure such as bundles, the polymers were strikingly curved (Fig. 6B). Similar effects on protofilament geometry were also observed in reactions assembled at lower pH (data not shown). However, in contrast to the large bundles observed by Andreu et al., the protofilaments remained mainly single stranded and despite exhaustive investigation, no such structures were found. This may be due to differences in the assembly conditions (see Experimental procedures) but, nevertheless, suggests that the reduced dynamics and lower GTPase activity that we observe here is not caused by protofilament bundling.
In an attempt to counteract the effects of 8j on protofilament morphology we co-assembled FtsZ with the known bundling agent CaCl2. In the absence of compound, Ca2+ stimulated the formation of multi-stranded polymer bundles, as expected (Fig. 6C). However, in the presence of compound, the protofilaments remained, on the whole, single stranded and curved. Remarkably though, these curved protofilaments now frequently formed arcs and ring-like structures (Fig. 6D), as well as extended spiral structures (Fig. 6E).
FtsZ polymer morphology is altered in vivo
The data so far indicate that benzamides enhance both the assembly and stability of FtsZ polymers, and also alter the morphology of the polymers in vitro. However, the moderate reduction in FtsZ dynamics observed in compound-treated cells seemed insufficient to explain the failure to assemble the Z ring. Structure activity relationship analyses have shown that the benzamide moiety of the compound series is the principal determinant of its antibacterial activity (Czaplewski et al., 2009; Haydon et al., 2010). 3-Methoxybenzamide (3-MBA) was itself (Fig. S1) previously shown to target FtsZ and possess weak but on target antibacterial activity, causing filamentation in B. subtilis (Ohashi et al., 1999). However, its effect on FtsZ localization was not reported. To further understand the mechanism(s) by which benzamides prevent the assembly of the Z ring, and to dissect the effects of increased polymer stability from those of altered polymer morphology, we re-examined the effects of 3-MBA in detail.
As expected, following growth with 10 mM 3-MBA, cells were clearly longer and cell division was inhibited. In place of the sharp Z rings (bands) normally observed at mid-cell (Fig. 7A), cells frequently had a short spiral or doublet structure (Fig. 7B). The regular spacing of these structures and their absence from the cell poles indicates that they remain sensitive to spatial regulation by the Min and Noc systems. Furthermore, when growth with 3-MBA was continued, the compact FtsZ structures unravelled, forming extended helical structures (Fig. 7C). Some of these structures then appeared to generate abnormal, ‘twisted’ cell division events. This unexpected morphological effect was not dependent on the presence of the GFP–FtsZ fusion as it also occurred in wild-type cells (Fig. 7D). Based on these results and the altered FtsZ polymer morphology that we observed in vitro, we wondered whether 3-MBA and other benzamides act by the same mechanism, but, differ only in potency. If so, low concentrations of 8j might generate a 3-MBA-like phenotype.
To test this we examined the localization of FtsZ, as well the ability of cells to divide at subinhibitory concentrations of 8j. As described above, at 2 µg ml−1 cells are highly filamentous and FtsZ assembles into discrete foci (but not rings) throughout the cell (Fig. 1B–E). However, at 0.2 µg ml−1 FtsZ foci were largely absent. FtsZ instead assembled into fairly well-spaced spirals, bands and doublets (Fig. 7E). Occasionally, a spiral FtsZ structure was observed that coincided with an abnormal cell division event. However, the cells still became filamentous and subsequently lysed (Fig. 7F and Fig. S9). At a 10-fold lower concentration, which is well below the MIC, cells were no longer filamentous. However, cell division was still perturbed as cells were clearly longer than untreated cells (Fig. 7G–H and Table 1). Importantly, when cells did divide, this occurred almost exclusively in a helical manner (Fig. 7G–J), apparently mediated by an extended helical Z ring (Fig. 7G), consistent with the ability of Z assemblies to determine the shape of the division septum (Addinall and Lutkenhaus, 1996). The increased length of these cells is likely caused by the delayed constriction and completion of helical division septa. Since our in vitro analysis indicated that at these concentrations the effect of the compound on the dynamics and GTPase activity of FtsZ polymers is limited, it is possible that the compound mediated alteration of polymer morphology plays a major role in preventing Z ring assembly in vivo (see Discussion).
Table 1. Effects of compound 8j on FtsZ assembly and cell division.
[Compound 8j]µg ml−1
Approximate molar ratio assuming an intracellular [FtsZ] of 6.6 µM, based on an average of 5000 molecules per cell (Feucht et al., 2001) and a cell volume of 1.26 µm3.
Determined after growth for 2 h at 37°C and expressed as mean length ± standard error of the mean.
Abnormal cell division in compound treated S. aureus
To investigate whether the abnormal cell division events in B. subtilis were relevant to the growth and division of a pathogenic organism, we treated S. aureus with a dilution series of 8j. As expected, at a high concentration of 8j (2 µg ml−1) cells became much larger than those in the untreated controls (Fig. 8A–C and N), forming smooth ‘balloons’ typical of a complete block in cell division (Fig. 8D–F) (Pinho and Errington, 2003; Haydon et al., 2008). However, at a 10-fold lower concentration of 8j (0.2 µg ml−1), although the cells again became enlarged, most were morphologically abnormal, frequently with unusual cell division events that appeared to have occurred in a twisted or helical manner (Fig. 8G–I and O). At an 8j concentration of 0.1 µg ml−1 the cells continued to divide aberrantly, although in an apparently more productive manner, as they remained comparable in terms of both size and growth rate to those of the untreated controls (Fig. 8J–L and M). These results further support the idea that 8j and related benzamides act by altering FtsZ polymer morphology in vivo. Moreover, they indicate that, even at low concentration, 8j compromises the fitness of S. aureus (Fig. 8M).
Effects of the benzamides in vivo and in vitro
We have characterized in detail the effects of the benzamide inhibitors of cell division in vivo and in vitro. Table 1 provides a summary and overview of the results of the various experiments. At high concentrations of compound (e.g. 2–20 µg ml−1 8j), FtsZ had a punctate appearance with foci at essentially random locations throughout the cell. These foci remained dynamic but with a significantly reduced rate of turnover. In vitro experiments established that under these conditions, FtsZ polymers are indeed more stable, and that the inhibition of GTPase activity is the likely cause of this effect. Nevertheless, FtsZ mutants with severely impaired GTPase activity, which exhibit reduced turnover in vivo, retain the ability to assemble Z rings and support cell division (Mukherjee et al., 2001; Stricker et al., 2002; Anderson et al., 2004). Therefore, the inhibition of GTPase activity and the reduced dynamics of compound treated FtsZ are not sufficient to explain its redistribution into foci. Moreover, while the polymers assembled with 8j in vitro were hyperstable, in the cell FtsZ does not assemble in isolation. A diverse complement of regulatory proteins co-assembles with FtsZ, which may explain the discrepancy between the cellular and in vitro dynamics we observed. Finally, the FtsZ foci remained in a state capable of recruiting most if not all of the key proteins of the division machine, indicating that the failure to assemble the Z ring is not caused by the loss of a known protein–protein interaction.
Strikingly, in addition to its effects on stability, FtsZ polymers assembled in the presence of 8j adopt a highly curved morphology. It seems likely that the highly curved protofilaments will be affected in bundling or higher-order assembly. The foci seen in vivo would then likely represent the failure of these curved protofilaments to correctly propagate around the inner circumference of the cell, instead forming discrete clusters held in place on either only a portion of their curvature, or else by forming ring-like assemblies that now sit flat on the surface of the membrane. In either case, this would prevent the assembly of a productive structure. The idea that polymer morphology is the basis of the inhibition of Z ring assembly is lent further support by our finding that 3-MBA itself alters the structure of the Z ring. Indeed, a recent report identified an essential cell division protein in C. crescentus, FzlA, that appears to modulate Z ring assembly by regulating the curvature of FtsZ polymers (Goley et al., 2010). Intriguingly, the overproduction of FzlA also caused FtsZ to localize as discrete foci.
Insensitivity to spatial regulation
In stark contrast, the various proteins involved in regulation of the site of FtsZ ring assembly and division, the Min and Noc systems, appeared not to interact significantly with the compound generated foci. This strongly suggests that the crucial interactions between the regulatory proteins and the division machine are unable to take place while FtsZ is in the highly curved state. Recently, the curved form of FtsZ has been implicated in playing a role during Z ring constriction. Moreover, previous work has shown that the Z ring becomes refractory to inhibition by the Min system after division initiates (Marston et al., 1998; Gregory et al., 2008). It is tempting to speculate that the curved protofilament form induced by compound 8j may be related to these observations.
Recent work with DivIVA protein, which is required for recruitment of the Min system to division sites and cell poles (Edwards and Errington, 1997), has shown that its cellular localization is directed to sites of highly concave membrane curvature (Lenarcic et al., 2009), which occur at the invaginating division septum and at completed cell poles. The fact that DivIVA did not associate at all with the FtsZ foci in inhibited cells is consistent with the foci being incapable of significant membrane invagination, even though the various division proteins are all present.
Unexpected subinhibitory effects of the benzamides
The fully inhibitory state discussed so far occurred at a concentration of compound roughly stoichiometric with that of FtsZ (Table 1). In principle, this could correspond to a state in which the compound is occupying its putative binding site in most FtsZ molecules. However, significant effects on cells were observed at much (c. 100-fold) lower concentrations of compound. Given the nature of the binding site and the chemistry of the benzamides, it seems unlikely that these effects are due to covalent attachment of the compound (Haydon et al., 2008; Czaplewski et al., 2009; Haydon et al., 2010). Furthermore, inhibition of cell division by the compounds is readily reversible (D.W. Adams, unpublished), so compound binding is probably reversible. Therefore, the compounds can somehow ‘poison’ the normal process of Z ring assembly by acting at only a small number of FtsZ subunits. Andreu et al. proposed that compounds such as PC190723 should bind preferentially to FtsZ polymers rather than monomers. Indeed, given that our in vivo analysis indicates that even very low concentrations of 8j can affect Z ring assembly, an especially tight mode of binding seems possible. However, as the representative binding constants are currently unknown, we cannot rule out that the effects we observe at low compound concentrations could also be caused by the reversible binding of 8j with equally fast on- and off-rates. Further work will be necessary to distinguish between these possible modes of binding.
The mechanisms of Z ring formation and constriction remain poorly understood. Our finding that low concentrations of benzamides result in distorted division structures lends support to the idea that there are at least two key steps in division: first, the assembly of a ring structure at mid-cell, perpendicular to the long axis of the cell; second, constriction along the path defined by the Z ring. In the presence of low concentrations of benzamides, Z ring formation was prevented and the major Z assemblies took on an extended helical shape with varying degrees of pitch. It is well established that mutations in ftsZ and other division factors can result in the formation of helical Z structures and division events (Addinall and Lutkenhaus, 1996; Feucht and Errington, 2005). However, in our experiments with both B. subtilis and S. aureus, the benzamides induced such events in wild-type cells. Nevertheless, the ability of these helical Z assemblies to support division, albeit of a morphologically abnormal kind, shows that the assembled division machine retained the ability to direct membrane invagination and cell wall synthesis. This family of compounds should provide a useful new tool with which to dissect the different steps in cell division.
Implications for Z ring assembly
Recent ground-breaking experiments from the Erickson lab have shown that under appropriate conditions FtsZ is capable of assembling into ring structures in the absence of other division factors (Osawa et al., 2008). This behaviour probably requires two important properties in the protofilaments. First, lateral association (e.g. bundling), which is required to cohere individual protofilaments into higher-order structures. Even if protofilaments are long enough to span the circumference of the cell, lateral interaction would be required to close and stabilize a ring. The observation of ring fusion in the lipid tube experiments as well as bundling in many other in vitro experiments supports the existence of such lateral interactions. Second, a tendency for the protofilaments to orientate in the direction of highest membrane curvature, which in a rod-shaped cell is perpendicular to the long axis of the cylinder. The effect of low concentrations of benzamides on Z ring closure presumably reflects an effect on one or both of the above activities. It is plausible that compound binding fixes the FtsZ subunit in a geometry that kinks the protofilament and impairs its ability to interact correctly with adjacent protofilaments and/or to find the perpendicular path around the cell. Gündoğdu et al. recently reported that SepF, which appears to share an overlapping role with FtsA (Ishikawa et al., 2006), assembles into regularly sized rings in vitro. These rings were shown to align FtsZ protofilaments and together form highly organized tubules (Gündoğdu et al., 2011). Given that the FtsZ protofilaments that make up these tubules are highly straight, one would anticipate that the highly curved FtsZ protofilaments that form in the presence of 8j might disrupt these structures. Finally, helical structures of FtsZ have previously been observed in vegetative cells of B. subtilis (Peters et al., 2007) and Escherichia coli (Thanedar and Margolin, 2004). It has been suggested that the rearrangement and subsequent collapse of these helical intermediates may facilitate Z ring assembly. Moreover, the switch from medial to polar Z rings at the onset of sporulation in B. subtilis is mediated by a spiral-like intermediate of FtsZ (Ben-Yehuda and Losick, 2002). It seems equally plausible that in addition to the effects described above, the alteration of FtsZ polymer morphology by 8j could compromise this process.
The role of the compound binding site?
The likely compound binding site has no known function and varies greatly between organisms, although some residues are broadly conserved (Haydon et al., 2008). Interestingly, Taxol binds to a site in the equivalent region of β-tubulin (Nogales et al., 1998). Indeed, the effects of 8j and Taxol initially appear to be quite similar, as both stabilize their respective polymeric targets. However, while Taxol allows the polymerization of GDP-tubulin (Diaz and Andreu, 1993), 8j acts in a strictly GTP-dependent manner. Nevertheless, given the common ancestry of tubulin and FtsZ (Erickson, 1995; Erickson, 1997), the functional analogies are intriguing and worthy of further investigation.
We have shown that binding 8j to this site regulates the assembly of FtsZ. However, does the site have a physiological role, whether as a binding site for a regulatory partner, or to accommodate a conformational switch (Erickson, 2009; Martin-Galiano et al., 2010) involved in the normal assembly mechanism? Further work will be necessary to determine any potential role of the binding site in vivo. Intriguingly, though, a substitution (S245F) located on the edge of the putative compound binding site, at the rear of the hydrophobic channel, is lethal in E. coli and causes FtsZ to localize into discrete, membrane-associated foci (Stricker and Erickson, 2003). The equivalent mutation is also lethal in B. subtilis (D.W. Adams, unpublished), suggesting that the cleft may play a role in Z ring assembly in divergent bacteria.
Bacterial strains and plasmids
Bacillus subtilis strains used in this study are listed in Table S1, together with the plasmids used and their construction. S. aureus strain RN4220 is a restriction-deficient mutant of 8325-4 (R. Novick).
Cells were made competent for transformation with DNA as previously described (Hamoen et al., 2002). Solid medium used for growing B. subtilis was nutrient agar (Oxoid) and liquid media were either PAB (Oxoid Antibiotic Medium No. 3) or Luria–Bertani broth (LB). Solid medium used for growing S. aureus was tryptic soy agar (Bacto) and liquid medium was tryptic soy broth (TSB). Where necessary antibiotics and additional supplements were added as follows: chloramphenicol (5 µg ml−1), erythromycin (1–10 µg ml−1), kanamycin (5 µg ml−1), phleomycin (1 µg ml−1), spectinomycin (50 µg ml−1), tetracycline (12 µg ml−1), xylose (0.05–0.5%) and IPTG (0.05–0.2 mM).
Strain DWA2 was constructed by transforming competent cells of strain DWA1 with the replacing cassette plasmid pVK73 (Chary et al., 1997), selecting for neomycin resistance, and screening the transformants for the loss of spectinomycin resistance.
Strains 168 and RN4220 were grown at 37°C in PAB and TSB, respectively, to an optical density of 0.4–0.6. The culture was then diluted to an optical density of 0.1 and mixed with an equal volume of pre-warmed medium containing either DMSO or compound at twice the desired final concentration. The culture was then transferred to a 96-well microtitre plate. The plate was incubated overnight at 37°C in a microtitre plate reader (FluoStar Galaxy, BMG Lab Technologies) and the optical density at 600 nm was followed over time. Readings were taken every 6 min, with 5 min shaking prior to each reading. Growth curves were plotted as the average of six replicates.
Cell viability assays were performed in a microtitre plate format as described above. During growth in the absence and presence of compound, samples were removed every 60 min, serially diluted and plated on nutrient agar. Plates were incubated at 37°C overnight (∼16 h). Viability was then determined by counting the number of colony-forming units.
Compound 8j was supplied by Biota Europe. 3-MBA was purchased from Sigma-Aldrich. Compounds were dissolved in DMSO to create stock solutions and stored at −20°C.
Exponentially growing cultures were mixed with an equal volume of pre-warmed media containing either 8j or 3-MBA at twice the desired final concentration, as indicated in the text. In all cases, the final concentration of DMSO was adjusted to 1% in both the absence and presence of the chemical inhibitors.
Cells containing GFP/YFP/CFP fusions were grown at 30°C. Cells were mounted on 1.2% agarose pads, covered with a No. 1 glass coverslip and observed. Cell membranes were stained by mixing 30 µl of culture with 1 µl of FM5-95 (200 µg ml−1; Invitrogen). Nucleoids were stained with either 4,6-diamidino-2-phenylindole (DAPI; Sigma), by mixing 10 µl of culture with 1 µl of DAPI (1 µg ml−1 in 50% glycerol) or by adding SYTO Blue 42 (Invitrogen) to a final concentration of 0.5 µM. Images were acquired with a Sony Cool-Snap HQ cooled CCD camera (Roper Scientific) attached to either a Zeiss Axiovert 200 M microscope or a Deltavision RT (Applied precision). Digital images were acquired and analysed with METAMORPH software. Manipulation was limited to adjusting the brightness and contrast to give optimal prints.
Fluorescence recovery after photobleaching
Strain 2020 was grown in LB at 30°C, in the absence (1% DMSO) and presence of 8j, and mounted on 1.2% agarose pads in the same medium. To induce the expression of GFP–FtsZ, xylose was included in the media at a final concentration of 0.05%. Cells were imaged and FRAP performed using a Nikon Eclipse Ti inverted microscope system coupled to a laser source, equipped with a Yokogawa spinning disc (CSU22) and Cool-Snap HQ2 camera (Photometrics). All images and FRAP measurements were made at room temperature (24–26°C).
GFP–FtsZ was imaged through a 100×/1.40 numerical aperture Plan Apochromat objective lens using a 491 nm laser and the focus maintained throughout the experiment using the Perfect Focus System (Nikon). Individual Z rings or FtsZ foci were bleached using a 100 ms pulse of a 405 nm laser. Pre- and post-bleach images were acquired using 500 ms exposure times. Recovery was monitored at 2 s intervals for 1 min, and for each time point the background intensity of an unbleached region subtracted. The subtracted data were then normalized to the average pre-bleach intensity and plotted against time. Recovery half-times were calculated using non-linear regression by fitting the data to a one phase association curve using GraphPad Prism (GraphPad Software, San Diego, CA, USA). The recovery half-time was then calculated as ln(2)/K.
Protein complex purification
FtsZ–His12 protein complexes were purified from B. subtilis as described (Ishikawa et al., 2006) with the following modifications: strains were grown in 200 ml of LB at 37°C until the OD600 reached 0.4–0.6, at this stage cells were then treated with either DMSO (1% final) or 8j (2 µg ml−1 final) for 1 h. Purified protein complexes were then analysed by silver staining using the silver stain plus kit (Bio-Rad) and by Western blotting with antibodies raised against GFP (laboratory stock), PBP2B (Daniel et al., 2000) and Noc (L.J. Wu, unpublished).
Protein expression and purification
Untagged, full-length FtsZ from B. subtilis was purified as follows. A fresh transformant of E. coli strain BL21 (DE3), harbouring plasmids pCXZ and pBS58, was grown in 2× TY containing ampicillin (100 µg ml−1), spectinomycin (50 µg ml−1) and glucose (0.4%). An overnight culture grown at 30°C was diluted 1/100 in fresh media and grown at 37°C to an OD600 of 0.4–0.7, at which point IPTG was added to a final concentration of 0.5 mM to induce protein expression. After 4 h, cells were harvested by centrifugation (5000 r.p.m.; 10 min; 4°C), washed once with ice-cold PBS containing 1 mM PMSF and snap-frozen in liquid nitrogen. Cell pellets were stored at −80°C overnight. FtsZ was then purified as previously described (Oliva et al., 2007) except that the gel filtration step was omitted. Protein concentration was determined using the BCA Protein Assay Kit (Thermo Scientific).
FtsZ (10 µM) was allowed to equilibrate for 5 min at room temperature in either MES buffer (50 mM MES pH 6.5; 50 mM KCl; 10 mM MgCl2) or Tris buffer (50 mM Tris pH 7.4; 50 or 300 mM KCl; 10 mM MgCl2). Compounds or DMSO was then added to the desired concentration before polymerization was initiated by the addition of GTP to 1 mM. Where indicated, GDP was included at a final concentration of 1 mM. The reaction was then allowed to proceed for 10 min at 30°C, before being subjected to ultracentrifugation (80 000 r.p.m.; 10 min; 25°C). The supernatants were carefully removed and the protein pellets resuspended in an equal volume of SDS-PAGE sample buffer. Equal volumes of the various fractions were then analysed by SDS-PAGE.
90° angle light scattering
Light scattering experiments were performed using a Cary Eclipse fluorescence spectrophotometer (Varian). Excitation and emission wavelengths were set to 350 nm, with slit widths of 2.5 nm. The photomultiplier was set to 600 V. Nucleotide-free polymerization reactions were assembled as described above in a Quartz SUPRASIL® cuvette (Hellma), with a 1 cm light path. Baseline data were collected for 2 min before the addition of GTP to 100 µM. For GDP disassembly experiments, GDP was added to 1 mM. Throughout the experiment data were collected every 100 ms and samples maintained at 30°C. For ease of interpretation the data were then smoothed by plotting the moving average intensity over time.
Measurement of GTPase activity
The GTPase activity of FtsZ was measured as the release of phosphate using the malachite green assay (Stokes et al., 2005). FtsZ (10 µM) was assembled in MES buffer, in the absence and presence of compound 8j, as described above. Polymerization was initiated by the addition of GTP to 1 mM and the samples incubated at 30°C for 30 min.
Negative stain electron microscopy
FtsZ (10 µM) was assembled in Tris buffer (50 mM Tris pH 7.4; 300 mM KCl; 10 mM MgCl2), in the absence and presence of compound 8j. Where indicated, CaCl2 was included at a final concentration of 20 mM. Polymerization was initiated by the addition of GTP to a final concentration of 1 mM and the reactions incubated at 30°C for 10 min. Two microlitres of the polymerization reaction was then applied to a glow-discharged, carbon-coated, copper grid (Newcastle Biomedical EM Unit) and blotted dry. The grid was then washed with 100 µl of a 2% uranyl-acetate solution, and again blotted dry. Grids were observed using a Philips CM100 Compustage Transmission Electron Microscope (FEI) with an AMT CCD camera (Deben).
We thank members of the Centre for Bacterial Cell Biology for stimulating discussions. We are grateful to Yoshikazu Kawai, Pamela Gamba, Patricia Domínguez-Cuevas and Richard Daniel for the kind gift of strains and acknowledge the technical expertise of Nada Pavlendová, Tracey Davey, Vivian Thompson and Ian Selmes. D.W.A. is supported by a BBSRC CASE studentship with Biota Europe.