Edited* by Thomas D. Pollard, Yale University, New Haven, CT, and approved February 25, 2014 (received for review September 20, 2013)
Although the mechanisms of microtubule depolymerization are relatively well understood, those of the tubulin homologue FtsZ have been difficult to understand owing to differences in its filament architecture and dynamics compared with those of microtubules. MinC, an important negative regulator of FtsZ and a component of the Min oscillatory system in Escherichia coli, positions the Z-ring to the midcell. With single-molecule fluorescence imaging in a cell-free minimal system on supported lipid bilayers, in which a network of FtsZ bundles assemble in a chemically well-defined system, the dynamic nature of the FtsZ bundles and the mechanism of disassembly by MinC is elucidated.
In Escherichia coli, a contractile ring (Z-ring) is formed at midcell before cytokinesis. This ring consists primarily of FtsZ, a tubulin-like GTPase, that assembles into protofilaments similar to those in microtubules but different in their suprastructures. The Min proteins MinC, MinD, and MinE are determinants of Z-ring positioning in E. coli. MinD and MinE oscillate from pole to pole, and genetic and biochemical evidence concludes that MinC positions the Z-ring by coupling its assembly to the oscillations by direct inhibitory interaction. The mechanism of inhibition of FtsZ polymerization and, thus, positioning by MinC, however, is not understood completely. Our in vitro reconstitution experiments suggest that the Z-ring consists of dynamic protofilament bundles in which monomers constantly are exchanged throughout, stochastically creating protofilament ends along the length of the filament. From the coreconstitution of FtsZ with MinCDE, we propose that MinC acts on the filaments in two ways: by increasing the detachment rate of FtsZ-GDP within the filaments and by reducing the attachment rate of FtsZ monomers to filaments by occupying binding sites on the FtsZ filament lattice. Furthermore, our data show that the MinCDE system indeed is sufficient to cause spatial regulation of FtsZ, required for Z-ring positioning.
Contractile rings in cell division are known for many species, but their mechanisms of positioning and contraction rarely are understood in detail. In Escherichia coli, the contractile ring consists primarily of FtsZ protein. FtsZ, a homolog of eukaryotic tubulin, is a GTPase protein found in most eubacteria and Archaea. In bacteria, FtsZ is the first protein to localize to the midcell as part of the cytokinesis machinery (1). In vivo imaging of FtsZ shows that it forms a ring-like structure (the Z-ring) that contracts concurrently with the constriction at midcell (2⇓–4). FtsZ also displays dynamic turnover in vivo where the cytoplasmic pool exchanges with the Z-ring on a timescale of seconds (5, 6). The polymerization into a ring-like structure is facilitated by the intrinsic curve of protofilaments (7), but the correct placement of the Z-ring depends on the oscillatory system and nucleoid occlusion of the Min proteins (8).
Tubulin and FtsZ share a common ancestor and the protofilaments of FtsZ/tubulin are similar, but the large-scale filaments they form differ significantly. Despite its structural similarity to tubulin, FtsZ lacks the polypeptide loops that form the lateral contacts between the protofilaments in the microtubule lattice (9). Thus, although the longitudinal interactions between subunits in microtubules and FtsZ may be rather conserved, the lateral interactions between subunits and protofilaments are less defined. As a consequence, the morphology and dynamics of the FtsZ filaments are supposed to be significantly different from microtubules, which have a specific closed-cylinder organization. Various attempts have been made to elucidate the FtsZ polymer structures. Results from electron microscopy studies in vitro have shown a variety of structures of FtsZ polymers (10⇓⇓⇓⇓–15). Atomic force microscopy studies have revealed that FtsZ filaments continuously rearrange, break, and anneal (16). At low concentrations, FtsZ assembles into protofilaments, which are single-subunit–thick polymers (17⇓–19). They also may form nucleotide-dependent, slightly or highly curved minirings, resembling α/β/γ-tubulin (20). At higher concentrations and in the presence of crowding agents, the protofilaments display lateral interactions that organize them into sheets or bundles (11, 18). Thus, the generally accepted model of FtsZ filaments is that of single-subunit–thick protofilaments, with a mean length of 100 nm, overlapping in a staggered manner by lateral interactions (21⇓–23). The precise nature and the functional relevance of these lateral interactions constituting a bundle, however, are unclear. The role of nucleotide hydrolysis, and the resulting dynamics, also is not well understood in terms of its functional context. Most importantly, despite several proposed models, an unequivocal understanding of how exactly the reported structural elements and their dynamics support Z-ring positioning and contraction still is lacking.
The FtsZ monomer comprises two independently folded subdomains. FtsZ assembles in an orientation similar to that of tubulin in the protofilament, such that the GTP molecule binds between the N-terminus of one monomer and the C-terminus of the adjacent one (24). The nucleotide-binding pocket between the two monomers is partially exposed, raising the possibility of nucleotide exchange without subunit turnover (24). This phenomenon might well affect the dynamics of FtsZ filaments.
MinC, a negative regulator of FtsZ (25), directly interacts with FtsZ and is shown to couple MinDE waves to FtsZ and regulate the localization of FtsZ in vivo (26). The molecular mechanism of MinC action on FtsZ is unclear. MinC itself does not affect FtsZ GTPase activity (27, 28), nor does it have any intrinsic nucleotide hydrolysis activity (28). However, MinC disassembles FtsZ polymers only when FtsZ undergoes hydrolysis-induced turnover. MinC is a dimer and has two functional domains—MinC N-terminal (1–115) and MinC C-terminal (116–231) (29, 30), henceforth referred to as MinCN and MinCC. MinCN has been shown to prevent sedimentation of FtsZ filaments in vitro and block cell division upon overexpression (27, 30). MinCC interacts with MinD, and MinD has been shown to amplify MinC activity in vivo (25). Mutations in FtsZ rendering it resistant to depolymerization by MinCN are on the H10 helix of FtsZ (31), accessible only at protofilament minus ends. A previous model assumes that this helix is accessible only upon GTP hydrolysis, resulting in a bend between subunits, and that MinCN breaks the filament by attacking the H10 helix (31). MinCC also is reported to have inhibitory effects on the Z-ring in vivo (32, 33). A mutant version of FtsZ with the I374V mutation also has been shown to be resistant to MinCC (33). These studies suggest that MinC has multiple interactions at multiple sites on FtsZ.
A more detailed understanding of FtsZ organization and its hydrolysis-induced turnover is critical for elucidating how MinC regulates it. As GTP hydrolysis of FtsZ is a key aspect of MinC activity, the turnover of FtsZ may itself be responsible for the activity of MinC. The H10 helix, which is necessary for the action of MinC, presumably would have to be exposed stochastically along the bundle as a result of turnover. In the present study, using in vitro reconstitutions of the FtsZ assembly, which show dynamics similar to those in vivo Z-rings, we examine the precise nature of the FtsZ turnover and its role in coupling to the Min waves through MinC.
To reconstitute the dynamic FtsZ filament bundles, we studied the assembly of a membrane-binding FtsZ: FtsZ-YFP-membrane targeting sequence (MTS) assembled on supported lipid bilayers, where MTS is a membrane-targeting sequence from E. coli MinD (34). On the basis of a previous study showing that the N- and C-terminals of FtsZ can fold independently (24), we cloned and purified the N-terminal (1–196) of the FtsZ, hereafter referred to as NZ, to specifically elucidate the dynamics of protofilament “minus” ends (where the H10 helix is located) generated in the FtsZ bundles. The FtsZ bundle network is maintained by continuous addition and dissociation of monomers. Our experiments suggest that MinC reduces the addition rate of FtsZ-GTP (by occupying the monomer binding sites and thus competing with FtsZ-GTP monomers) and increases the FtsZ-GDP detachment rate. In this way, MinC shifts the steady state of the FtsZ bundle in a concentration-dependent manner, resulting in partial or full disassembly.
To visualize FtsZ assembly in real time, we used a modified version of FtsZ—FtsZ-YFP-MTS, where MTS is the membrane-targeting sequence from E. coli MinD. Because the YFP-MTS is fused to residue 366 at the C-terminus of the protein, it does not affect any other known interacting sites required for polymerization or lateral interactions. We observed the polymerization of a 1:1 mixture of FtsZ-YFP-MTS with unlabeled WT FtsZ on an E. coli polar extract lipid bilayer using total internal reflection fluorescence (TIRF) microscopy. The presence of a lipid membrane support allows protofilaments and short bundles to diffuse and interact with other protofilaments and bundles. Fig. 1A shows assembly of 0.1 µM FtsZ at 10-s intervals. Because a single MTS of the E. coli MinD shows poor affinity for the membrane, the filaments are attached to the membrane only when the FtsZ-YFP-MTS begins to polymerize in the presence of GTP. The presence of the membrane substrate lowers the critical concentrations of FtsZ-YFP-MTS for polymerization compared with FtsZ in solution (SI Appendix, Fig. S3) (35⇓⇓–38). Based on our TIRF results and previous insights from other groups, the sequence of assembly events appears to be as follows: monomers begin to polymerize into single-stranded protofilaments; the protofilaments are recruited to the membrane as the result of multiple MTS binding. The protofilaments increase in length and width by colliding and annealing with other protofilaments, resulting in filament bundles (Fig. 1 A and I, Movie S1). At higher concentrations (>0.1 µM), the filaments settle into a stable network with no free filament bundle ends. Increasing the concentrations of FtsZ results in more filament bundles or area coverage (Fig. 1C). To quantify the growth characteristics, the movies were analyzed using an ImageJ-based algorithm (SI Appendix, Fig. S4). With increasing concentrations of FtsZ, the total length of polymers in the network increases proportionally (Fig. 1B). Once the network has been formed, it remains stable in its spatial structure but dynamic in the presence of GTP (Movie S2). The apparent growth rate constant of bundles constituting the network measured from the fluorescence images is 20 µm⋅s−1⋅µM−1 per 200 µm2 of observation area or 0.1 µm−1⋅s−1⋅µM−1 (Fig.1B). Based on the measured thickness of bundles previously reported, we consider the filament bundle to be about six subunits thick (7). Using this value, the polymerization rate constant for the network is about 750 subunits per s−1⋅µM−1. As an approximation, dividing by the number of nucleation sites measured at the initial time points of the polymerization events, the rate constant lies around 7.5 s−1⋅µM−1 per nucleation site. At limiting concentrations of GTP (<50 µM), the filament disassembles. Depolymerization appears to occur reversely to polymerization, with the larger filaments breaking into smaller fragments and no indication of catastrophic events as in microtubules (Fig. 1D). We also observed that exposing the bundle network to laser light resulted in increased bundling and fixation of the network with low turnover in vitro (SI Appendix, Fig. S5). This effect might be avoided by the use of an oxygen-scavenging system, as described in Materials and Methods (39).
Polymerization and depolymerization of an FtsZ network. (A) Polymerization of 0.1 µM FtsZ on a supported lipid bilayer, with frames taken every 10 s. (B) Growth characteristics of the FtsZ network. A higher concentration of FtsZ results in an increased net length of filaments (increased area coverage). (C) Increased concentrations of FtsZ result in increased area coverage/net length of FtsZ filaments. (D) Depolymerization of FtsZ filaments is qualitatively the reverse of assembly. (E) FRAP of an FtsZ filament network assembled at 0.2 µM. The spatial pattern before and after fluorescence recovery remains unchanged. (F) Kymograph of FRAP along a filament, showing uniform recovery at optical resolution limits. (G) Fluorescence recovery curves after photobleaching of FtsZ-YFP-MTS with GTP and GMPCPP. (H) A temporal overlay of two-color TIRF imaging shows NZ interacting throughout the length of the filament bundle. The overlay image shows FtsZ in blue, NZ in green, and colocalization in white pixels. (I) Schematic representing a polar FtsZ filament and a filament bundle. Scale bar: A and D, 1 µm; C and E, 10 µm; I, 5 µm.
The FtsZ bundle network appears stable once assembled; however, fluorescence recovery after photobleaching (FRAP) demonstrates that the filament bundles undergo hydrolysis-dependent turnover (Fig. 1 E–G). The observed recovery is a result of exchange occurring between the filament bundle and solution. The filament bundles recover all along the length of the preexisting filaments (Fig. 1 E and F, Movie S2). The half-time of recovery in the presence of GTP is about 10 s. The off rates obtained from a binding model (40) are about 0.1 s−1 (Fig. 1G). At optical resolution scales, the recovery appears uniform (Fig. 1F, Movie S2). It cannot be determined whether individual subunits or larger fragments dissociate. However, considering that a protofilament will have substantial lateral interaction with the rest of the bundle, it would be energetically unfavorable to dissociate as a large fragment. It is more realistic to assume that upon hydrolysis, individual subunits or smaller filament fragments dissociate.
Dissociation of subunits or small protofilament fragments and association of new filaments should stochastically generate protofilament ends in the bundle. To verify this, we used the N-terminal domain (1–196) of FtsZ to study the distribution of protofilament ends along the FtsZ bundles. As NZ lacks the C-terminus half of FtsZ to enable it to polymerize further, it acts as a minus end-capping unit for the FtsZ protofilaments.
We probed the interaction of very low concentration of NZ-cyanine (Cy)5 (10–25 nM) with an FtsZ network (Fig. 1H), which enables us to image single NZ-Cy5 molecules binding to filament minus ends (Movie S3). By recording sufficiently long time-lapse images and temporally overlaying them, we found that NZ binds homogeneously along the length of the bundle (Fig. 1H; SI Appendix, Fig. S6). This confirms that the protofilament ends to which NZ may attach constantly are generated throughout the bundle as a result of monomers moving out of the lattice as a result of GTP hydrolysis. At sufficiently high concentrations, NZ disassembles the FtsZ network, as described further in this manuscript.
The distributions of single-molecule residence times of FtsZ and NZ (Fig. 2 A and B), as well as other data presented below, were analyzed in terms of a model of dynamic turnover of the FtsZ network incorporating these three essential steps (for more details, see SI Appendix, Text S3):
i) FtsZ-GTP binds to the filament bundle with the rate ka.
ii) Once within the filament, FtsZ hydrolyzes GTP with the rate kh.
iii) The GDP form of FtsZ subsequently dissociates from the filament with the rate kd.
Residence time distributions and interactions of MinC with FtsZ. Single-molecule residence time distributions of (A) FtsZ-Cy5, (B) NZ, and (C) EGFP-MinC dimer on FtsZ filament bundles. (D) FtsZ bundles can be depolymerized stepwise by addition of MinC, and the MinC-induced depolymerization rates are increased in the presence of MinD. (E) Twenty-nanometer resolution localization of EGFP-MinC on an FtsZ bundle network. (F) Montage (Left) and kymograph (Right) of a depolymerizing FtsZ filament, showing uniform loss of intensity along the length. (G) Filaments showing fragmentation events while being depolymerized by MinC. (H) Depolymerization of FtsZ polymers upon addition of MinC. Scale bar: E, F, and H, 5 µm.
The time evolution of the concentrations of the GTP and GDP forms of FtsZ in the filament, CT and CD, respectively, then is described by the following differential equations:
The rates ka, kh, and kd determine the steady-state concentrations of the GTP and GDP forms of FtsZ in the filament: cT(ss) = ka/kh and cD(ss) = ka/kd. The fits of distributions of FtsZ and NZ to Eqs. S6 and S7 (SI Appendix, Text S3), respectively, yield two characteristic times: the hydrolysis time and the time between hydrolysis and detachment. The hydrolysis time (τh = 1/kh), τh = 7.6 and 6.6 s for FtsZ and NZ (FtsZ + NZ), respectively, agree well with the known values. The time between hydrolysis and detachment, τd = 1.5 and 1.6 s for FtsZ and NZ, respectively, are shorter than those determined from the MinC experiments described below. The likely reason is that the fits to the MinC or NZ experiments are affected by the approximations in the model, as explained in SI Appendix, Text S3.
The single molecules of FtsZ as well as NZ remained stable within their dwell time in the filament, suggesting that the filaments do not show sliding as has been proposed but, rather, may show treadmilling-like behavior. In a bundle, however, owing to different degrees of lateral interactions (SI Appendix, Fig. S2), the treadmilling behavior may not be completely pure, as has been reported for FtsZ recruited to the membrane by FtsA (41), but may be more complex and dependent on the number of lateral interactions.
We also measured the rate of nucleotide exchange into the FtsZ filaments to consider any stabilization effects of fresh GTP exchanged into a GDP-bound subunit. We find that the nucleotide exchange (rate 0.2 min−1) is very slow compared with the turnover (SI Appendix, Fig. S7 and Text S2) and, thus, may be ignored.
To measure the dynamics of MinC on the FtsZ filament bundles, we used EGFP-MinC. MinC and EGFP-MinC were purified as dimers by size-exclusion chromatography (SI Appendix, Fig. S8). To assess the dimerization state of EGFP-MinC at single-molecule level low concentrations, we compared the particle brightness of EGFP-MinC with that of monomeric EGFP by fluorescence intensity distribution analysis (FIDA) (42) and single-molecule imaging (SI Appendix, Figs. S8 and S9). The experiments indicate that even at nanomolar concentrations, EGFP-MinC exists as a dimer.
To perform two-color imaging of FtsZ and EGFP-MinC, we replaced FtsZ-YFP-MTS with FtsZ-MTS and WT FtsZ with cysteine-mutant FtsZ-F268C, which can be labeled with Cy5. FtsZ-F268C shows assembly identical to that of WT FtsZ and WT-like turnover when copolymerized (7, 18) and has its C-terminal intact, which is required for interaction with MinCC (33). It also was shown previously to be functional in vivo (43). Single-molecule tracking and localization of EGFP-MinC dimers on an FtsZ bundle network of FtsZ-MTS and FtsZ-F268C-Cy5 did not show any movement of EGFP-MinC along the FtsZ filament. This is in agreement with the single-molecule studies of FtsZ or NZ molecules not showing any movement, as described earlier in this manuscript. EGFP-MinC binds uniformly with the same affinity throughout the bundle, showing no preferential binding at any particular region (Fig. 2E). The residence time distribution of the EGFP-MinC dimers, fitted to the same model as the residence time distribution of NZ (Eq. S7) yields two characteristic times, τh = 11.9 and τd = 4.2 s (Fig. 2C), both longer than the corresponding times of FtsZ and NZ. This suggests that MinC interacts with more than one FtsZ monomer within the bundle, and can stay within the bundle even after the FtsZ dissociates.
Next, we explored the activity of WT MinC on FtsZ bundles assembled on lipid bilayers. WT MinC disassembled the FtsZ-YFP-MTS–WT FtsZ filament network on bilayers (Fig. 2 F–H). The FtsZ-YFP-MTS assembled with guanylyl-(α,β)-methylene-diphosphonate (GMPCPP) was resistant to disassembly by MinC, as expected (SI Appendix, Fig. S10). The process of depolymerization shows two main characteristics: an overall decrease in the bundle intensity (Fig. 2F) and fragmentation of bundles over the course of depolymerization (Fig. 2G, Movie S4). The MinC dimers depolymerize FtsZ in a concentration-dependent manner in steady state, with complete depolymerization observed with an FtsZ-to-MinC ratio of about 1:1.
The time course of depolymerization at different MinC concentrations was analyzed with the model assuming a time delay between hydrolysis and detachment of FtsZ (SI Appendix, Text S3, Eqs. S1 and S2 and fitting function Eq. S5). The obtained hydrolysis rate kh = 0.128 s−1 agrees well with the previously published values (18, 44, 45), and the attachment rates decrease and detachment rates increase with the MinC concentration (Fig. 3F; SI Appendix, Fig. S11). This suggests that MinC promotes depolymerization of FtsZ bundles by enhancing the detachment rate of FtsZ-GDP (without affecting the spontaneous hydrolysis rate) and also reduces the FtsZ-GTP attachment rate by partially blocking the binding sites.
MinC and NZ disassemble FtsZ polymers with different dynamics. (A) Depolymerization dynamics of FtsZ by MinC fit to the model described in this paper. FtsZ was assembled at a concentration of 0.8 µM. Fit parameters are summarized in Fig. S2B. (B) FCCS studies of MinC and NZ interaction with FtsZ in the solution above the FtsZ network; for details, see the main text and SI Appendix, Text S1. (C) Montage showing depolymerization of FtsZ filaments on addition of NZ. The characteristics are similar to those of MinC-induced depolymerization: an overall decrease in filament intensity as well as breakage of the filaments. (D) Depolymerization dynamics of FtsZ bundles assembled at a concentration of 0.8 µM by NZ. The fits are described in SI Appendix, Text S3; Eq. S5 was used for fitting. (E) Initial rates of depolymerization of FtsZ with MinC, NZ, and MinD + MinC. (F) The experimental dependence of the rates ka and kd on MinC concentration. The experimental values are in blue and the fit is in red. Scale bars: C, 10 µm.
To confirm that MinC does not interact with FtsZ monomers in solution, as was reported previously (46), we performed fluorescence cross-correlation spectroscopy (FCCS) (47). EGFP-MinC does not show any cross-correlation with FtsZ-F268C-Cy5, confirming that MinC and FtsZ do not interact in solution (Fig. 3B). We therefore can rule out a model in which MinC depolymerizes FtsZ by sequestering monomers in the solution. On the other hand, NZ-Cy5 showed cross-correlation with FtsZ-YFP-MTS, suggesting that NZ might sequester FtsZ monomers in solution (Fig. 3B).
Adding NZ at different concentrations causes depolymerization of the FtsZ-YFP-MTS bundle network (Fig. 3 C and D). In contrast to MinC, different concentrations of NZ do not strongly affect the initial rates of depolymerization, suggesting a minor or no effect on the hydrolysis and detachment rates, kh and kd (Fig. 3 D and E; SI Appendix, Eq. S5). The attachment rates ka decreased with higher NZ concentration, causing partial disassembly of the network as a result of incomplete turnover. Taken together with the presence of NZ binding to FtsZ monomer in the solution, the mechanism of disassembly by NZ is a combination of sequestration of monomers in the solution by NZ and capping of filament ends, as seen by single-molecule imaging. Consistent with the sequestration mechanism, the critical concentration for FtsZ polymerization also increased with NZ concentration, whereas for MinC, it did not change the critical concentrations (SI Appendix, Figure S3).
Our finding that disassembly by MinC depends on concentration differently than disassembly by NZ suggests that MinC antagonizes FtsZ by a more complex mechanism. Thus our results, combined with the previous observations described below, form the basis for a new model (Fig. 4).
Schematic showing the FtsZ filament bundles and their interaction with NZ and MinC. FtsZ bundles assemble by (i) longitudinal annealing and (ii) lateral interactions. In steady state, the FtsZ bundles constantly exchange subunits with the solution as a result of GTP hydrolysis. This results in stochastic exposing of the (−) ends. About 50% of the subunits in the bundle lattice are bound to GDP. NZ binds to the (−) ends of the filaments. NZ can sequester monomers in the solution (iii), and it can cap the filament (−) ends in the bundles (iv). MinC also binds to the (−) end of the FtsZ filaments through the N-terminal, capping and preventing annealing (v). MinC also interacts with the C-terminal of FtsZ through its C-terminal. MinC can bind to exposed (−) ends in the filaments caused by a leaving FtsZ subunit. Once bound to FtsZ, MinC remains bound until the subunit leaves, frustrating lateral interactions with incoming FtsZ subunits or FtsZ fragments (vi). The net action of MinC also results in a release of FtsZ-GDP subunits trapped in the filament lattice, resulting in a concentration-dependent increase in the rate of depolymerization.
MinC interacts with FtsZ filaments using its N- and C-terminal domains independently (28). The MinCN terminal interacts with the H10 helix of the FtsZ subunits and MinCC has been shown to inhibit lateral interactions (31, 33). GTPase activity of FtsZ is necessary for the action of MinC on the polymers. A reduction in or an absence of GTPase activity of FtsZ renders the filaments resistant to depolymerization by MinC both in vitro (28) and in vivo (31). The above-described FtsZ turnover constantly generates filament ends along the length of an FtsZ filament bundle. FtsZ polymers contain a substantial amount of subunits in the GDP form (48). MinC is a strong dimer (29, 30). It interacts uniformly throughout the length of the FtsZ bundle, and its turnover on FtsZ bundles is coupled to the turnover of FtsZ, as observed by single-molecule studies of MinC on FtsZ filaments.
MinC dimers can bind to the exposed H10 helices through the MinC N-terminal. However, in contrast to the previous model in which the H10 helix is exposed only upon GTP hydrolysis and subsequently breaks the filaments (31), we show that the minus ends are stochastically exposed in the bundle as a result of turnover. Once bound to the minus end, MinC can be released only when the terminal FtsZ subunit leaves the filament. The second interaction by MinC is that of the MinC C-terminal binding to the C-terminal of the FtsZ (33). The overall effect of MinC binding on the FtsZ bundles is a decrease in longitudinal and lateral attachment rates of fresh FtsZ monomers and short filaments, as MinC blocks the binding sites.
At higher concentrations, MinC disassembles the network more effectively than expected by a mechanism that does not change the hydrolysis, or detachment rates, as exemplified by NZ-induced depolymerization (Fig. 4). Because the GTPase activity of FtsZ is not affected at all by MinC, we hypothesize that GDP-bound subunits may be released faster from the filaments upon MinC binding to the filament, thereby lowering lateral interactions owing to decreased replenishment of FtsZ monomers in the bundle and MinC occupying interacting sites on the filament. Therefore, the model for the depolymerization kinetics must incorporate a lag time between GTP hydrolysis by an FtsZ subunit and its release from the filament (see SI Appendix, Text S3 for the detailed formulation of the model). Interaction with MinC then affects the average duration of the lag between hydrolysis and detachment. Our model, in which the filaments contain FtsZ-GDP subunits in the lattice and there is a lag time in their release after hydrolysis, agrees with the depolymerization kinetics observed in the presence of MinC. About 50% of FtsZ subunits in the filaments previously were shown to be GDP bound (48). From the analysis of our depolymerization kinetics, we obtain a value of about 53% (SI Appendix, Fig. S2B). MinC results in depolymerization and a concentration-dependent release of trapped GDP-bound subunits from the filaments. With increasing concentrations of MinC, at a new steady state of the FtsZ bundles, the GDP-bound fraction decreases, as thinner filaments have a decreased ability to trap GDP-bound subunits, and the kd increases as a result (Fig. 4; SI Appendix, Text S3 and Fig. S11).
The concentration of MinC in E. coli cells is about 400 molecules per cell, compared with an FtsZ concentration of 15,000 molecules per cell. Overexpression of MinC might inhibit cell division independent of the Min waves (25). In our in vitro experiments, complete depolymerization occurred at a ratio of about 0.8:1.2 of MinC to FtsZ. The lower stoichiometry of MinC in vivo with respect to FtsZ must be explained. We argue that because MinC works in concert with the MinDE waves, a high local concentration of MinC should be sufficient to disassemble FtsZ polymers locally. To check this, we reconstituted the MinDE waves together with FtsZ and MinC (Fig. 5 A–C). At MinC:FtsZ ratios of 1:5, FtsZ filaments are localized complementary to the MinDE proteins comprising the waves (Movie S5). This effect requires the presence of MinC. In the absence of MinC, FtsZ bundles remain unaffected by MinDE waves (Fig. 5A, Right). The filaments show normal turnover activity, as observed without MinDE waves, and remain stable during their lifetime at a particular spot (Fig. 5B). With MinC regulated by MinDE waves, a lower overall concentration ratio of MinC to FtsZ compared with that required for complete depolymerization is sufficient to disassemble FtsZ bundles locally. The presence of other destabilizing factors, such as FtsA (41), also may reduce the effective MinC concentration required to destabilize FtsZ filaments.
Spatial regulation of FtsZ by MinCDE waves. (A) FtsZ is depolymerized by MinCDE waves. In the absence of MinC, FtsZ is not spatially regulated, Concentrations used were 1 µM MinD, 1.5 µM MinE, 0.5 µM MinC, and 2 µM FtsZ. MinE was doped with 20 mol% MinE-Cy5 and FtsZ with 50 mol% FtsZ-YFP-MTS. (B) The filaments show a reaction-dominant recovery on photobleaching (note the sharp boundaries throughout the recovery) as on supported lipid bilayers. (C) Montage showing depolymerization and repolymerization cycles at a fixed spot in the sample with time. The white arrow points in the direction of the Min waves. Scale bars: A–C, 25 µm.
The high-resolution visualization of FtsZ polymerization and depolymerization on a lipid membrane allowed us to directly observe its assembly/disassembly characteristics and its spatial regulation by the Min proteins. In particular, we could assemble dynamic FtsZ filament bundles that anneal and branch, resulting in a dynamic network on supported bilayers. By using a minus end-capping fragment, NZ, we can show that the protofilament ends are generated, and the proteins may be exchanged, constantly throughout the length of the filament bundles. This specific dynamic turnover, quite distinct from the dynamics observed in microtubules, constitutes the basis for Z-ring positioning by the Min protein machinery.
The Z-ring presumably consists laterally of 6–10 protofilaments (7, 21, 49, 50). FtsZ first polymerizes into single-stranded short protofilaments and subsequently interacts laterally and longitudinally to settle into longer, staggered bundles. The exchange of subunits is occurring not only from the ends, as for microtubules, but also from within the filament bundles. We cannot, however, distinguish whether individual monomeric FtsZ subunits or polymeric fragments dissociate. Considering that longer protofilaments will have substantial lateral interactions and that the FtsZ hydrolysis-induced dissociation events take place independent of the position of monomers in the protofilament, it is realistic to assume that most dissociation is monomeric FtsZ. This argues for the following characteristics concerning the dynamics of FtsZ filament bundles: Once the protofilaments form a bundle, individual monomers or small protofilament fragments can dissociate and reassemble back to the bundle, resulting in a constant turnover. Recent in vivo studies also showed that the FtsZ counteroscillates to the Min waves in complex with bundling proteins such as ZipA or ZapA even at the early stages of ring assembly, suggesting that the FtsZ filaments already may be bundled (51). The extent of coupling between hydrolysis and depolymerization has been difficult to study experimentally. A recent study shows that hydrolysis events occur randomly and independently of each other all along the protofilament (52). The study also emphasizes that nucleotide exchange is an important factor in determining depolymerization dynamics (14, 52), suggesting that FtsZ-GDP at the interface of a second FtsZ must be maintaining its contacts, and that FtsZ-GDP resides in the lattice for a substantial time.
A general understanding linking morphology and dynamics of evolutionary related polymers still is lacking. Although tubulin and FtsZ resemble each other structurally, the morphology and dynamics of FtsZ filament bundles clearly differ from those of microtubules. Microtubules show a more hierarchical assembly, based on heterodimeric protein assembly into sheets and folding into a closed cylindrical geometry (53). This geometry allows the microtubules to exhibit dynamic instability, resulting from a GTP cap that holds the GDP-bound protofilament lattice against its preferred outward curvature. Although the precise interactions that lead to this very specific assembly are not understood, it is evident that FtsZ filaments show vastly different dynamics in that they exchange subunits throughout the filament bundles. An intuitive explanation for the different suprastructures of FtsZ and tubulin might be the different nature of lateral interactions. Tubulin assembles into a sheet that transitions into a cylindrical geometry (53, 54), whereas FtsZ assembles from single-stranded protofilaments, which come together in a staggered manner into a filament bundle. Each subunit in the filament bundle retains its ability to hydrolyze GTP and exchange into the solution, although with different rates depending on the longitudinal and lateral contacts they form. This essentially results in different dynamics. Phenomena such as dynamic instability and treadmilling are less probable with a filament organization like the staggered bundle, in which the lateral interactions strongly affect the dynamics, rendering it more isotropic.
Depolymerization of FtsZ by MinC appears to be a complex process. It requires the GTPase activity of FtsZ and the turnover resulting from GTP hydrolysis, but MinC does not directly modulate the GTPase activity in any way (55). Our model is based on the interactions between FtsZ and MinC reported by Shen and Lutkenhaus (31), who propose that MinC first binds to FtsZ filament using its C-terminus and that MinCN attacks the H10 helix and “breaks” the longitudinal bond. We propose that MinC exploits the GTPase-induced turnover activity of FtsZ itself to disassemble the filaments. A concern proposed by Shen and Lutkenhaus regarding the MinCN attack on the H10 helix and its breaking of the longitudinal bond is that the H10 helix between FtsZ dimers normally is not solvent exposed and therefore prevents MinCN interaction. The authors argue that this helix is solvent exposed upon hydrolysis. Our experiments refine this model. Owing to constant hydrolysis and turnover in the filament bundle, which results in monomers leaving the filament lattice, the H10 helix may be exposed at the protofilament minus ends in the bundle, as shown by the NZ dynamics on FtsZ bundles. MinC readily binds to these filament ends in the bundles through the MinCN, coupling its activity to FtsZ turnover. By binding to the exposed H10 helix in the minus ends through its N-terminus and presumably to the FtsZ C-terminus by its C-terminus, MinC blocks new subunit additions. We also can account for the increase in depolymerization rates with increasing concentrations of MinC by considering the release of FtsZ-GDP trapped in the filament lattice. This assumption that FtsZ-GDP is trapped in the filament also was hinted at in previous studies (48). FtsA, another dimeric protein that does not affect FtsZ GTPase activity and has no ATPase activity of its own, but depolymerizes FtsZ bundles (41), may have a similar mechanism—by blocking bundling and occupying turnover sites.
As shown by FRAP, the filament bundle network formed on the supported lipid bilayers show complete recovery, with a half-time of 10 s, which is comparable to in vivo turnover rates of 8 s (6). Toroids of FtsZ-GFP bundles formed in yeast upon overexpression turn over in about 11 s (56). Turnover of filaments measured in vitro is about 3.5–7 s (48). This suggests that the lateral interactions in these bundles are weak and probably different in nature from those created by divalent ions or bundling proteins, which decrease GTPase activity and turnover. At high concentrations of MinC, when most of the FtsZ bundles are depolymerized, new filaments formed do not bundle, as MinCC inhibits bundling (28), but they still undergo hydrolysis and turnover at the new steady state. This results in the GTPase activity not being affected by MinC.
How is filament-based inhibition preferable for the Min system compared with sequestration of FtsZ monomers? As stated above, the concentration of MinC in cells is far below the concentration of FtsZ. Interestingly, from our previous work on reconstituted Min protein waves on supported membranes (57, 58), it appears the concentration of all Min proteins at the membrane surface, and particularly in the trailing edge of the wave, is significant. Acting on the FtsZ filaments recruited to the membrane, possibly in concert with other destabilizing factors, such as the membrane adaptor FtsA (41), will significantly lower the MinC concentration finally required for positioning the Z-ring.
The E. coli FtsZ and Min proteins exemplify a self-organized system in which energy is consumed independently by two different components: the MinD–MinE system, in which MinD hydrolyses ATP, and the FtsZ system, whose dynamics are based on GTP hydrolysis. MinC links the two systems by virtue of its binding to MinD and its depolymerizing activity on FtsZ polymers. The in vitro experiment combining all four components—MinCDE and FtsZ—directly confirms that the Min proteins can spatially regulate and eventually position FtsZ polymers, as has been shown on the basis of genetic and biochemical experiments. Very recently, it was shown that FtsA, which has been considered mainly a membrane anchor for FtsZ, shows a very similar coupling to FtsZ dynamics, destabilizing the filaments and initiating treadmilling, which may lead to collective longitudinal movement, i.e., spiraling rings (41). Taken together, the dynamics of FtsZ we have addressed here, albeit quite different from actin and tubulin polymers, seem to provide the basis for many different ways to spatially regulate and position the Z-ring. They also may offer important clues to address the still open questions regarding the molecular mechanism of ring constriction.
E. coli FtsZ-YFP-MTS, FtsZ-F268C, and FtsZ were purified as described elsewhere (34). Briefly, FtsZ-YFP-MTS was expressed from pET11b vector in BL21 cells. Cells were lysed by sonication, and the FtsZ-YFP-MTS was precipitated from the supernatant by 40% ammonium sulfate; resuspended, dialyzed, and further purified by using Resource Q column (Amersham Biosciences); desalted; and stored in aliquots. To assemble FtsZ polymers, HMKKG buffer (Hepes 50 mM, magnesium acetate 5 mM, potassium acetate 300 mM, potassium chloride 50 mM, and 10% glucose) was used. The supported lipid bilayers were prepared with the same buffer and the required amount of protein and GTP was diluted into the solution. FtsZ-YFP-MTS and WT FtsZ at 1:1 were polymerized at a final concentration of 0.1 µM and 500 µM GTP. NZ, MinC, and EGFP-MinC were purified by overexpression from a pET28a vector in BL21 cells, and purified by using Ni-nitrilotriacetic acid (NTA) columns as previously described (58). The purified proteins were confirmed by SDS/PAGE. For single-molecule imaging of EGFP-MinC dimers, FtsZ-F268C-MTS labeled with Cy5 was used.
His-MinD and His-MinE were purified and fluorescently labeled as previously described (57, 58). EGFP-MinC was purified as previously described (58) with an extra size-exclusion chromatography step.
PCR with primers GGT CGC TAG CAT GGA ACT TAC CAA TGA CG and ATT ACG ATC GGC GAT ACC TTG CAC AGC was used to amplify the N-terminal sequence of FtsZ from pET11b FtsZ-YFP-MTS corresponding to 1–196 amino acids (NZ). The fragment was digested with NheI and ligated into similarly treated pET28a to obtain a plasmid containing NZ. The resulting ORF encoded for a fusion protein of NZ linked to the N-terminal hexahistidine tag by a short linker and the sequence for thrombin cleavage site and a T7 tag. NZ then was overexpressed from pET28a NZ using a similar protocol for the Min proteins.
For reconstitution of MinCDE waves and FtsZ filaments, a Tris buffer was used (50 mM Tris, pH 7.5; 150 mM KCl; 7.5 mM MgCl2). The total FtsZ concentration used for FRAP, single-molecule, and depolymerization experiments was 0.2 µM. Assays were performed by adding FtsZ at different concentrations for polymerization or by adding NZ or MinC for depolymerization immediately after initiating image acquisition on the TIRF microscope. For reconstitution of the waves, total concentrations of the proteins used were 1 µM MinD, 1.5 µM MinE, 0.2 µM MinC, and 1 µM FtsZ. MinE was doped with 20 mol% MinE-Cy5 and FtsZ with 50 mol% FtsZ-YFP-MTS. GTP and ATP were at 500 µM each. All buffers in the experimental chamber were supplemented with an oxygen-scavenging system consisting of 75 U⋅mL−1 glucose oxidase, 1,500 U⋅mL−1 catalase, 0.25 wt/vol β-d-glucose, and 1 mM Trolox just before the experiments.
Small unilamellar vesicles (SUVs) were prepared by sonification of E. coli lipid extract 4 mg/mL (Avanti Lipids) with 0.1 mol% di-alkyl indocarbocyanine (DiI) in FtsZ polymerization buffer at room temperature. The suspension was diluted to 0.5 mg/mL, added to the substrates, and warmed to 37 °C. Adding CaCl2 to 2.5 mM induced fusion of the SUVs on mica, leading to bilayer formation. The sample was rinsed with 2 mL polymerization buffer to remove unfused SUVs.
Fluorescence confocal imaging, photobleaching, and FCCS were performed with a Zeiss LSM 780 with a Zeiss 40× 1.2 N.A. objective. YFP was excited with 488 nm and fluorescence detected through a 505–550-nm emission filter. Membrane-labeling dye DiD was excited with 633 nm and detected through an LP 650 filter. FRAP was performed using the same system.
Total internal reflection fluorescence microscopy for single molecules and polymerization assays were performed using a laboratory-built objective-based TIRF system with a Zeiss 100× 1.45 N.A. or an Olympus 60× 1.45 N.A. objective. A magnifier lens was inserted between the exit port and an Andor iXon EMCCD camera to oversample in space, resulting in each pixel being equal to 50 nm in real space. Total internal reflection (TIR) was created by moving a beam focused at the back focal plane of the objective away from the principal axis. Two-color images were obtained by splitting the emission signal in front of the camera.
Filament lengths were quantified using the Track Skeleton plugin for ImageJ. Briefly, it tracks the length, number of junctions, branches, and end points in a binary image generated from a fluorescence image. FRAP intensities were obtained from ImageJ, and the values were plotted and fitted with a binding model using Origin (OriginLab). FIDA was performed on MATLAB according to previously established methods (42).
Detailed experimental methods may be found in SI Appendix, Text S1.
The authors thank Harold Erickson for the FtsZ plasmids and Martin Loose for the EGFP-MinC constructs. The authors thank Daniel J. White (Max Planck Institute for Molecular Cell Biology and Genetics) for help with the ImageJ plugins. We are indebted to William Margolin, David Drechsel, Satyaki Prasad, Angika Basant, Ariadna Martos, and Katja Zieske for their critical reading of the manuscript. S.A. is the recipient of a Dresden International Graduate School for Biomedicine and Bioengineering Fellowship; P.S. acknowledges a Human Frontier Science Program and Leibniz grant.
↵1Present address: Unité Mixte de Recherche 168, Membrane and Cellular Functions, Institut Curie, 75005 Paris Cedex 05, France.
Author contributions: S.A. and P.S. designed research; S.A. and Z.P. performed research; S.A. and Z.P. contributed new reagents/analytic tools; S.A. and Z.P. analyzed data; and S.A. and P.S. wrote the paper.
The authors declare no conflict of interest.
↵*This Direct Submission article had a prearranged editor.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1317764111/-/DCSupplemental.
