Abstract

Article18 October 2007free access The structure of FtsZ filaments in vivo suggests a force-generating role in cell division Zhuo Li Zhuo Li Division of Biology, California Institute of Technology, Pasadena, CA, USA Search for more papers by this author Michael J Trimble Michael J Trimble Department of Biology, Indiana University, Bloomington, IN, USA Search for more papers by this author Yves V Brun Yves V Brun Department of Biology, Indiana University, Bloomington, IN, USA Search for more papers by this author Grant J Jensen Corresponding Author Grant J Jensen Division of Biology, California Institute of Technology, Pasadena, CA, USA Search for more papers by this author Zhuo Li Zhuo Li Division of Biology, California Institute of Technology, Pasadena, CA, USA Search for more papers by this author Michael J Trimble Michael J Trimble Department of Biology, Indiana University, Bloomington, IN, USA Search for more papers by this author Yves V Brun Yves V Brun Department of Biology, Indiana University, Bloomington, IN, USA Search for more papers by this author Grant J Jensen Corresponding Author Grant J Jensen Division of Biology, California Institute of Technology, Pasadena, CA, USA Search for more papers by this author Author Information Zhuo Li1, Michael J Trimble2, Yves V Brun2 and Grant J Jensen 1 1Division of Biology, California Institute of Technology, Pasadena, CA, USA 2Department of Biology, Indiana University, Bloomington, IN, USA *Corresponding author. Division of Biology, California Institute of Technology, 1200 East California Blvd., MC: 114-96, Pasadena, CA 91125, USA. Tel.: +1 626 395 8827; Fax: +1 626 395 5730; E-mail: [email protected] The EMBO Journal (2007)26:4694-4708https://doi.org/10.1038/sj.emboj.7601895 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info In prokaryotes, FtsZ (the filamentous temperature sensitive protein Z) is a nearly ubiquitous GTPase that localizes in a ring at the leading edge of constricting plasma membranes during cell division. Here we report electron cryotomographic reconstructions of dividing Caulobacter crescentus cells wherein individual arc-like filaments were resolved just underneath the inner membrane at constriction sites. The filaments' position, orientation, time of appearance, and resistance to A22 all suggested that they were FtsZ. Predictable changes in the number, length, and distribution of filaments in cells where the expression levels and stability of FtsZ were altered supported that conclusion. In contrast to the thick, closed-ring-like structure suggested by fluorescence light microscopy, throughout the constriction process the Z-ring was seen here to consist of just a few short (∼100 nm) filaments spaced erratically near the division site. Additional densities connecting filaments to the cell wall, occasional straight segments, and abrupt kinks were also seen. An ‘iterative pinching’ model is proposed wherein FtsZ itself generates the force that constricts the membrane in a GTP-hydrolysis-driven cycle of polymerization, membrane attachment, conformational change, depolymerization, and nucleotide exchange. Introduction All life depends on the ability of cells to divide. While many aspects of this complex process are already understood in eukaryotic cells, our knowledge of the mechanisms driving what would appear to be the more simple process of ‘binary fission’ in prokaryotes is surprisingly poor. At least 13 different proteins function cooperatively at the division site in Escherichia coli (Goehring and Beckwith, 2005; Margolin, 2005; Rothfield et al, 2005). FtsZ, the filamentous temperature sensitive protein Z, is the first to localize and assembles into a ring-like structure called the FtsZ-ring or Z-ring near the cytoplasmic membrane at midcell (Begg and Donachie, 1985; Taschner et al, 1988; Bi and Lutkenhaus, 1991). While the full role of the Z-ring is still unclear, it is required for the recruitment of the rest of the cell division machinery and might help to provide the force required to constrict the membrane. As such a vital molecule, FtsZ is present in virtually all prokaryotes and in some chloroplasts and mitochondria as well (Michie and Löwe, 2006). FtsZ has been identified as a homolog of eukaryotic tubulin based on its possession of a tubulin signature motif and its remarkable similarity to tubulin in both structure and biochemical properties (Löwe and Amos, 1998; Nogales et al, 1998a, 1998b). While FtsZ polymerizes into protofilaments, sheets, ribbons, minirings, and tubes under different in vitro experimental conditions (Mukherjee and Lutkenhaus, 1994; Erickson et al, 1996; Gonzalez et al, 2003), which of these structures are physiologically relevant is still unclear. Like tubulin, FtsZ is a GTPase. GTP-tubulin polymerizes into straight microtubules, where the GTP at the dimer interfaces is quickly hydrolyzed. Because the GDP-bound state favors a curved protofilament conformation, microtubules must be stabilized by ‘caps’ of GTP-bound dimers at their ends, and they exhibit dynamic instability: when the GTP cap is hydrolyzed, individual protofilaments quickly peel away from each other to produce curved rings (Nogales, 2001). How FtsZ assembles into polymers and the role of GTP hydrolysis is still unknown. As for tubulin, the GTP- and GDP-bound states of FtsZ seem to favor straight and curved conformations, respectively, but not exclusively (Lu et al, 2000). In vitro evidence suggests that FtsZ polymerizes in a cooperative manner (Caplan and Erickson, 2003; Chen et al, 2005), but an isodesmic mechanism that includes the cyclization of long single-stranded fibrils has also been proposed (Gonzalez et al, 2005). Our knowledge of the in vivo structure of FtsZ has advanced incrementally with improvements in imaging technology. The Z-ring was first identified by gold-labeled antibodies that were seen by electron microscopy to localize to the leading edges of the midcell invagination (Bi and Lutkenhaus, 1991). Soon afterwards, the application of fluorescent light microscopic (fLM) techniques to bacterial cells (Margolin, 1998) revealed the Z-ring's donut-like structure in 3-D (Levin and Losick, 1996; Sun and Margolin, 1998). More recently, fLM techniques of living cells have revealed dynamics: FtsZ appears to assemble into arc-like and spiral structures that move and collapse during the cell cycle (Addinall and Lutkenhaus, 1996; Ben-Yehuda and Losick, 2002; Thanedar and Margolin, 2004). Fluorescence recovery after photobleaching (FRAP) experiments show that FtsZ molecules rapidly exchange between the Z-ring and the cytoplasmic pool with a half time of ∼10 s (Anderson et al, 2004). Here, we have used the emerging technique electron cryotomography (ECT) to further advance our understanding of the structure and function of FtsZ. ECT can reveal the three-dimensional ultrastructure of small, intact cells in a life-like, ‘frozen-hydrated’ state (Lucic et al, 2005; Jensen and Briegel, 2007). ECT has already been used to visualize several bacterial cytoskeletal filaments directly in situ (Kurner et al, 2005; Komeili et al, 2006; Scheffel et al, 2006), and we recently used it to visualize a number of novel filament bundles in the widely studied model bacterium Caulobacter crescentus (Briegel et al, 2006). C. crescentus is a favorite model organism for studying bacterial cell development because its characteristic asymmetrical division yields two morphologically different progeny: a sessile, stalked cell and a motile, flagellated cell. The stalked cell can immediately induce another round of cell division by replicating chromosomal DNA and expressing cell division proteins. In contrast, the flagellated ‘swarmer’ cell must differentiate into a stalked cell by shedding its flagellum and assembling a stalk and holdfast at the newly vacated pole before it becomes replication competent. In accordance with these dramatic morphological changes, a large number of proteins experience cell cycle-dependent regulation on both the transcriptional and translational levels (Laub et al, 2000; Grunenfelder et al, 2001). The concentration of FtsZ begins at a very low number in swarmer cells, increases sharply before cell division, and then drops rapidly after division through proteolysis (Quardokus et al, 1996; Kelly et al, 1998). A process involving the protein MipZ positions the Z-ring at the midcell (Thanbichler and Shapiro, 2006). By imaging the division sites of C. crescentus at high magnification with an electron cryomicroscope, we consistently observed arc-like filaments lying underneath the inner membrane at midcell. The filaments' position, configuration, time of appearance, and resistance to A22 all suggested that they were FtsZ. The filaments' absence in an FtsZ depletion strain, higher abundance in strains overexpressing FtsZ, and dramatically higher abundance and longer lengths in a strain expressing a more stable FtsZ mutant further indicated that the filaments were in fact FtsZ. Through analysis of these filaments, the architecture of the Z-ring in C. crescentus is described and an existing model for the role of FtsZ filaments in cell division is elaborated. Results Section one: Identification of FtsZ filaments by ECT Observation of midcell arc-like filaments in predivisional wild-type C. crescentus. Predivisional NA1000 C. crescentus cells were plunge-frozen and observed in an electron cryomicroscope. Series of images were recorded of individual cells while the sample was tilted incrementally. Three-dimensional reconstructions of the cells (tomograms) were then calculated which clearly showed outer membranes, peptidoglycan layers, inner membranes, and crowded cytoplasms full of ribosome-like particles (Figure 1). Because good images of samples tilted to angles greater than ∼65° cannot yet be obtained (Iancu et al, 2005), the resolution of the reconstructions was anisotropic and the boundaries of objects were blurred in the direction parallel to the beam. Thus the membranes at the top and bottom of the cell were not resolved, and any filaments perpendicular to the beam would also have been difficult to discern. Figure 1.Arc-like filaments at the division site in C. crescentus NA1000. (A) Low magnification cryo-EM view of a C. crescentus NA1000 cell plunge-frozen on an EM grid after acquisition of the tilt-series. The darker gray elongated object is the cell, suspended in vitreous ice across a circular hole in the carbon support film. The dashed box shows the region segmented in (B). ST: stalked pole; SW: pole opposite the stalked pole; IC: inner curvature; OC: outer curvature. Scale bar 500 nm. (B) 3-D segmentation of the inner membrane (blue), outer membrane (yellow), and arc-like filaments (red) in the reconstruction from the same cell. Because the image contains 3-D perspective effects, no scale bar is included. (C) 8 nm slice through the reconstruction parallel to the grid, showing the filaments in cross-section (small dark dots near the center of the circles) next to the membrane as well as the orientations of the other slices shown in panels (D–F) (dashed lines). (D–F) 5.4 nm slices through the reconstruction, showing that while filaments and other small proteins all appeared as small dots in cross-section, filaments were recognized by their continuous, elongated shape in ‘side’ views such as these. Unfortunately, space does not allow all relevant slices to be shown. Thus here and in other figures, the locations and extents of all visible filaments were manually traced and presented as colorized, segmented models (like panel B), which are necessarily interpretations but do allow 3-D structures to be shown in 2-D figures. The colors and labels are consistent within all the figures: IM: inner membrane; OM: outer membrane; SL: surface layer; F: arc-like (FtsZ) filaments. The scale bar in (D) is 100 nm and serves for panels (C-F). Download figure Download PowerPoint Close examination of the 3-D tomograms of predivisional cells revealed arc-like filaments lying just inside the inner membranes on one or both sides of the midcell constriction approximately perpendicular to the long axis of the cell (Figure 1B–F). Such filaments were seen in seven reconstructed predivisional cells, always just inside the constriction point (two out of seven cells had filaments just on one side, the other five cells had filaments on both sides). In the particular cell shown in Figure 1, the two filaments seen along the inner curvature of the cell were not strictly parallel (Figure 1E), but two other filaments on the outer curvature were (Figure 1F). The filaments were between 80 and 160 nm long, covered arc-lengths of 0.45–0.85 radians, and were ∼5 nm in diameter. Appearance, configuration, and disassembly of the filaments during cell division. In ECT, the cells being imaged are frozen and cannot grow, and the high-energy electrons used to produce the images gradually destroy biomolecular structures, so it is not possible to gather data from the same cell repeatedly as it progresses through division. Instead, pseudo-time-courses must be built up by arranging reconstructions of different cells imaged at various stages of the process in relative order. During the course of these experiments, a total of 15 stalked NA1000 cells were reconstructed. Morphological features like the presence or absence of a flagellum, the length-to-width ratio of the cell body, and the degree of membrane constriction were used to place these cells in relative order approximating the progression of C. crescentus through cell division (Figure 2). Of the four stalked cells that were not yet visibly constricted, only the one with the longest length-to-width ratio (5.6) had midcell arc-like filaments (Figure 2B), consistent with the known timing of FtsZ localization to the midcell (Quardokus et al, 2001). Figure 2.Filament dynamics during cell division. Each row represents a different stage of cell division, arranged in approximate sequence from an unconstricted stalked state (A) until after fission of the inner membrane has formed two separate cytoplasmic compartments (G). Within each row, the first four columns show different depictions of a single cell reconstructed by ECT and the last column shows an fLM image of a different cell taken at an apparently analogous stage. (A–G) Low magnification cryo-EM images. Stalked poles are shown with arrows. Scale bars 500 nm. (A1–G1) 3-D segmentations of the division sites. (A2–G2) ‘Face-on’ views from the cytoplasm of the ‘left’ side of the cell wall. (A3–G3) ‘Face-on’ views of the ‘right’ side of the cell wall, again from the cytoplasm. (A4–G4) fLM images of NA1000 cells expressing FtsZ-YFP. Scale bar 2 μm. In D1–D2, an additional cytoplasmic filament bundle (Briegel et al, 2006) was surface-rendered in orange. In G1–G3, no arc-like filaments are observed; instead, the most filamentous densities are surface-rendered in orange for comparison. Note that the cells shown in rows C and E are the same as those shown in Figures 1 and 5AB, respectively, and that supporting unsegmented tomographic slices are shown in Supplementary Figures S1–S6. Download figure Download PowerPoint There were 10 stalked cells (the seven mentioned earlier, plus three additional A22-treated cells that will be discussed later) that were visibly constricted to different degrees but whose inner membrane had not yet separated. Midcell arc-like filaments were seen in every case, but there was no discernable evolution in their number, position, or configuration (Figure 2C–F). Most of the filaments were between 40 and 120 nm, and the longest was ∼160 nm. Although the ‘missing wedge’ in ECT smears details in the direction of the electron beam, the arc-like filaments were nevertheless clearly short chords and arcs rather than complete rings. In this respect, the two membranes served as omnipresent internal controls: they are consistently blurred at the same positions around the top and the bottom of the cell, but the arc-like filaments frequently ended nearer the center, before the missing wedge would have obscured them (see also Figures 7B, G and 10G). The numbers of filaments on each side were not consistent or equal. In the cell shown in Figure 2D, for example, there were four filaments of different length on the right-hand side but only one very short one on the left, but in the cell shown in Figure 2C, there were two filaments on each side. More pronounced asymmetries were also seen: in one case, there were nine filaments on one side and none on the other (data not shown). No regular bundling pattern, spacing, or relationship was observed between neighboring filaments. In one cell, an additional long filament bundle parallel to the long axis of the cell (orange filaments in Figure 2D) was observed in the cytoplasm near the midcell arc-like filaments that was similar to the ‘cytoplasmic bundle’ identified in earlier work (Briegel et al, 2006). In the last (15th) cell of the sequence, the outer membrane was deeply constricted and the inner membrane had separated to form two distinct cytoplasmic compartments (Figure 2G), as has been observed previously (Judd et al, 2005). The diameter across the constriction in the outer membrane was 154 nm. In this cell, no clearly arc-like filaments were observed at the division site. While there were a number of elongated densities (orange in Figure 2G), they were significantly shorter than the midcell arc-like filaments and their distances from the membrane were more variable. fLM images of cells at apparently similar stages of division were collected for comparison (right column of Figure 2), showing that the midcell arc-like filaments assemble at the same time as Z-rings localize to the midcell (in very late phase unconstricted stalked cells), and both structures persist throughout constriction. Following separation of the inner membranes, FtsZ-YFP remained localized to the new poles (Thanbichler and Shapiro, 2006), but the midcell arc-like filaments were no longer identifiable. Filament abundance and length in mutants with different levels of FtsZ expression and activity. Because the midcell arc-like filaments' position, orientation, and dynamics all suggested that they were FtsZ, the following further experiments were carried out to test that hypothesis. Since effective labeling methods for ECT of intact cells have not yet been developed, instead we looked at cells expressing different levels and forms of FtsZ to see if the number and/or length of the putative FtsZ filaments changed correspondingly. First, NA1000 cells overexpressing FtsZ were imaged (Quardokus et al, 2001). Presumably because the stoichiometry of FtsZ to other components of the divisome is critical, overexpression of FtsZ blocks cell division and produces extended constricted sites containing multiple FtsZ foci (Figure 3A). Tomograms of these cells exhibited arc-like filaments with the same diameter, curvature, distance from the membrane, and orientation with respect to the long axis of the cell as seen previously in dividing NA1000 cells, but the filaments were more numerous (14 arc-like filaments in ∼1.3 × 0.6 × 0.3 μm3) and were collected in bunches at the most constricted points (Figure 3C–E), just as would be expected for FtsZ in this overexpression strain. Figure 3.Increase in the abundance of arc-like filaments in cells overexpressing FtsZ. (A) Differential interference contrast (DIC) and fLM overlay image of cells overexpressing (wild-type) FtsZ, showing that FtsZ-YFP accumulates in the most constricted points. Scale bar 4 μm. (B) Low magnification cryo-EM view of a similar cell plunge-frozen in vitreous ice across an EM grid after acquisition of the tilt-series. The gray box shows the region segmented in panel (E). For scale, the circular hole in the carbon support film on the EM grid has a diameter of 2 μm. (C) 8 nm slice through the tomogram parallel to the grid. Scale bar 100 nm (for panels C and D). (D) 5 nm slice through the tomogram along the dash line in (C). The white arrow points to the arc-like filaments. (E) 3-D segmentation. Download figure Download PowerPoint Next, cells overexpressing a hyperstable mutant form of FtsZ (FtsZG109S) were imaged. This mutation is located close to the nucleotide-binding pocket and corresponds to the classical mutant FtsZ84 in E. coli, which has been shown to slow both GTP-binding and hydrolysis activity. These changes lead to an ∼3-fold longer half-time of fluorescence recovery in FRAP experiments, indicating that the filaments it forms are less dynamic/more stable (Anderson et al, 2004). In C. crescentus, overexpression of FtsZG109S blocks cell division and produces long, deeply constricted regions containing large amounts of FtsZ, including extended patches (Wang et al, 2001) (Figure 4A). Tomograms of such deeply constricted regions revealed large numbers of tightly packed arc-like filaments (Figure 4B–E). In the ∼400 × 400 × 900 nm3 volume shown in Figure 4E, for instance, ∼50 arc-like filaments can be seen on both sides of the cell. Again these filaments were indistinguishable from those seen earlier, except that some filaments were much longer than any seen previously and they were much more numerous so that entire patches of the cell wall were covered and neighboring filaments were frequently seen aggregated into bundles of 4–5 filaments. Since many of these filaments extended across the top or bottom of the cells where they were difficult to resolve, their full lengths could not be determined. Nevertheless, because the extended division sites were thinner (∼300 nm diameter) than those in NA1000 cells, better resolutions were achieved in the reconstructions and at least at the thinnest point (boxed region in Figure 4E), it was almost possible to follow these longer mutant filaments around the entire circumference of the cell. Here the filaments appeared to form helices that wrapped all the way around the inner membrane (Figure 9C). Figure 4.Dramatic increase in the abundance and length of filaments in a mutant overexpressing FtsZG109S. (A) fLM image of cells overexpressing FtsZG109S, showing that FtsZ-YFP accumulates in the extended, constricted division sites. Scale bar 4 μm. (B) Low magnification cryo-EM image of a cell overexpressing FtsZG109S. Scale bar 500 nm. (C) 8 nm slice through the 3-D reconstruction parallel to the grid, showing tens of filaments in cross-section as small black dots close to the membrane. The dashed lines show the position of the slice shown in panel (D). Scale bar 100 nm (for panels C and D). (D) 5.4 nm ‘glancing’ slice through the 3-D reconstruction showing tens of filaments just underneath the inner membrane. (E) 3-D segmentation with another cytoplasmic filament bundle colored orange. The box identifies a region referred to in Figure 9C. Download figure Download PowerPoint As a final, ‘negative’ control, strain YB1585 was imaged, wherein FtsZ can be conditionally depleted (Wang et al, 2001). When FtsZ is depleted, these cells elongate without visible constrictions (Supplementary Figure S8). While it is unclear where to look for filaments in these cells, tomograms of random samples throughout the cell body were collected, and no arc-like filaments were seen (Supplementary Figure S9). Persistence of midcell arc-like filaments in the presence of the MreB-inhibitor A22. The correlation in the length, number, and distribution patterns of the filaments with the known FtsZ expression levels, stability, and fluorescence localization patterns in these three mutant strains strongly supported the proposition that the filaments were in fact FtsZ, rather than for instance some other, unknown filament that might also be found at the division site. Nevertheless there is at least one other filament, composed of the actin-like MreB, that is thought to be present in a ring at the midcell of predivisional cells. By fLM, MreB usually appears to form helical structures along the cell wall that span most of the long axis of the cell body, but these have been observed to condense into a ring-like structure when C. crescentus divides (Figge et al, 2004; Gitai et al, 2004). A22 was identified as a small molecule that diffuses rapidly through the cell membrane and depolymerizes MreB (Gitai et al, 2005). To rule out the possibility that the midcell arc-like filaments were MreB, two types of cells that exhibited these filaments (NA1000 cells and cells overexpressing FtsZG109S) were treated with A22 and imaged by ECT as before. Three A22-treated dividing NA1000 cells were reconstructed, and all three possessed arc-like filaments that were indistinguishable from those seen in the nontreated cells (Figure 5A and B). Similarly, two A22-treated cells overexpressing FtsZG109S were reconstructed which also both still exhibited large numbers of densely packed, unusually long filaments (Figure 5C and D). Supporting fLM studies were carried out in both cases to confirm that the A22 treatment did in fact delocalize the MreB foci as expected but had no perceptible effect on FtsZ (Supplementary Figure S10). Figure 5.Arc-like filaments persist in the presence of A22. Low magnification cryo-EM views (A, C) and corresponding segmented 3-D reconstructions (B, D) of a NA1000 cell (A, B) and a cell overexpressing FtsZG109S (C, D), both after treatment with A22, again showing arc-like filaments. Scale bar in (A) 500 nm. For scale in (C), the circular hole in the carbon support film on the EM grid has a diameter of 1.3 μm. See also supporting fLM studies in Supplementary Figure S10. Download figure Download PowerPoint Taken together, these results provide compelling evidence that the arc-like filaments observed at the division site were in fact FtsZ: (1) they appeared at the leading edge of the constriction where FtsZ is known to localize; (2) their width, position inside the membrane, and orientation were as expected for FtsZ filaments; (3) the timing of their appearance and their persistence during cell division followed FtsZ's known dynamic pattern; (4) they persisted in the presence of A22, which depolymerizes the other filament (MreB) thought to exist at midcell in dividing cells; (5) their abundance correlated with increases and decreases in the expression of FtsZ; (6) their numbers and length were both dramatically increased in a strain overexpressing a more stable, mutant form of FtsZ; and (7) the details of their distribution pattern (bunches versus extended patches) in two different overexpression strains matched what had been observed for FtsZ by fLM. Section two: Characterization and measurement of FtsZ filaments In total, 248 putative FtsZ filaments were seen (36 filaments in 11 NA1000 cells, 30 filaments in two cells overexpressing wild-type FtsZ, and 182 filaments in three cells overexpressing FtsZG109S). Distance from and connections to the inner membrane. FtsZ filaments were close to but not lying directly against the inner membrane. To measure their average distance from the membrane, a ‘profile’ of all the densities seen as a function of distance both above and below the inner membrane was calculated in one of the constricted regions of a cell overexpressing FtsZG109S (Figure 6). Outside the inner membrane, the expected peaks for the peptidoglycan layer, outer membrane, and surface layer were observed. Inside the inner membrane, an additional peak was resolved 16 nm (peak to peak) into the cytoplasm corresponding to the accumulated FtsZ filaments. As a control, a similar calculation on an unconstricted region of an NA1000 cell where no filaments were observed showed all the same peaks except the cytoplasmic peak attributed to FtsZ. Figure 6.FtsZ filaments are on average 16 nm away from the membrane. Density profiles through the cell wall of a cell overexpressing FtsZG109S in the region of an extended division site containing FtsZ filaments (lower curve) and a random, unconstricted region of one NA1000 cell body not containing FtsZ as a control (top curve). The horizontal axis is the distance either outside (negative numbers) or inside (positive numbers) the inner membrane. The surface layer (SL), outer membrane (OM), peptidoglycan (PG), inner membrane (IM), and FtsZ filament layers can be resolved. Download figure Download PowerPoint Most of the FtsZ filaments showed no apparent connection to the inner membrane (one example from a cell overexpressing FtsZG109S is shown in Figure 7A). In a minority of cases (7, 4, and 23 filaments analyzed in NA1000 cells, cells overexpressing wild-type FtsZ, and cells overexpressing FtsZG109S, respectively), however, mid-filament ‘bridge-like’ densities connected the filaments to the inner membrane (Figure 7B and C) and/or the peptidoglycan (Figure 7D). In seven additional cases (all from cells overexpressing FtsZG109S), the filaments appeared to originate (or terminate) in the inner membrane (Figure 7E and F). In some rare instances, the inner membrane and the filament seemed to deviate from smooth arcs together, as if their separation was rigidly controlled (Figure 7G), but in some other rare cases, filaments were seen unusually close to the membrane (Figure 7H). Unfortunately, the identities of the bridging macromolecules remain unclear, some of them may have been present for reasons other than to fulfill a specific bridging function, and it is possible that additional bridges existed which were too small to resolve by these methods. Figure 7.Connections between FtsZ filaments and the cell wall. (A–H) 6.7 nm