Escherichia coli ZipA Organizes FtsZ Polymers into Dynamic Ring-Like Protofilament Structures

ABSTRACT ZipA is an essential cell division protein in Escherichia coli. Together with FtsA, ZipA tethers dynamic polymers of FtsZ to the cytoplasmic membrane, and these polymers are required to guide synthesis of the cell division septum. This dynamic behavior of FtsZ has been reconstituted on planar lipid surfaces in vitro, visible as GTP-dependent chiral vortices several hundred nanometers in diameter, when anchored by FtsA or when fused to an artificial membrane binding domain. However, these dynamics largely vanish when ZipA is used to tether FtsZ polymers to lipids at high surface densities. This, along with some in vitro studies in solution, has led to the prevailing notion that ZipA reduces FtsZ dynamics by enhancing bundling of FtsZ filaments. Here, we show that this is not the case. When lower, more physiological levels of the soluble, cytoplasmic domain of ZipA (sZipA) were attached to lipids, FtsZ assembled into highly dynamic vortices similar to those assembled with FtsA or other membrane anchors. Notably, at either high or low surface densities, ZipA did not stimulate lateral interactions between FtsZ protofilaments. We also used E. coli mutants that are either deficient or proficient in FtsZ bundling to provide evidence that ZipA does not directly promote bundling of FtsZ filaments in vivo. Together, our results suggest that ZipA does not dampen FtsZ dynamics as previously thought, and instead may act as a passive membrane attachment for FtsZ filaments as they treadmill.

patches around the cell circumference by treadmilling. Through connections involving other divisome proteins that cross the cytoplasmic membrane, these treadmilling FtsZ protofilaments help to guide the septum synthesis machinery in concentric circles, resulting in inward growth of the septal wall until it closes and the daughter cells are separated (8,9).
Although FtsA is conserved in diverse bacterial species, ZipA is limited to gammaproteobacteria, including E. coli (10). In the absence of both FtsA and ZipA, FtsZ fails to attach to the membrane or form the protoring, demonstrating the requirement for a membrane tether (11). In the presence of only FtsA or ZipA, FtsZ filaments form a membrane-anchored ring, but septation fails to proceed (12), suggesting that the divisome is in a locked state. One major unanswered question in the field is why E. coli requires dual FtsZ membrane anchors to assemble a divisome that completes septation. Our recent study provides a potential answer by showing that FtsA exerts a specific structural and functional constraint on FtsZ protofilaments: when attached to lipid monolayers, FtsA assembles into clusters of polymeric minirings that align FtsZ polymers and inhibit their bundling (13).
In this report, we use the term "bundling" in reference to increased lateral interactions between adjacent FtsZ protofilaments, resulting in two or more polymers closely associated in parallel. The physiological role of these lateral interactions is not firmly established, but several FtsZ mutants that are defective in protofilament bundling in vitro are also defective in cell division (14)(15)(16). In addition to the intrinsic ability of FtsZ polymers to interact laterally, proteins called Zaps (ZapA, ZapC, and ZapD; FtsZ-associated proteins) help to bundle or cross-link FtsZ polymers in vitro (17,18). Inactivation of single Zap proteins is not lethal, but mutant cells lacking multiple Zap proteins have significant division defects (19)(20)(21)(22)(23). Hyperbundled mutants of FtsZ have also been isolated, and cells expressing these alleles also divide abnormally (24)(25)(26). However, one hyperbundled mutant, called FtsZ*, has gain-of-function properties (27). FtsZ*, which forms mostly double-stranded filaments in vitro, allows division of cells lacking ZipA and can resist the effects of other FtsZ inhibitors. Together, these findings suggest that lateral interactions are important for FtsZ function, but these interactions need to be balanced.
The aforementioned study (13) proposed a model in which FtsA minirings antagonize FtsZ protofilament bundling, keeping the divisome in a locked state. In this model, once the cell is ready to divide, these minirings are disrupted and are no longer a constraint for FtsZ polymer bundling. This is consistent with another model in which broken FtsA polymers start to recruit later divisome components, while FtsZ polymers become anchored to the cell membrane by ZipA (2,28). ZipA has been shown to stabilize the protoring not only by anchoring FtsZ to the membrane but also by protecting it from degradation by ClpXP protease (29)(30)(31). Whereas FtsA inhibits FtsZ polymer bundling (13), ZipA is considered an FtsA competitor for FtsZ polymers because of their common binding site at the FtsZ C terminus (32)(33)(34)(35). Thus, it is not surprising that ZipA has been suggested as a bundler of FtsZ. However, the reports on its effect on FtsZ protofilament bundling in solution are not consistent (27,(36)(37)(38)(39)(40).
Recently, it has become clear that the functionalities of the protoring proteins need to be tested in a more physiological context by attaching them to a lipid surface (13,(41)(42)(43)(44)(45)(46)(47). For example, Mateos-Gil et al. (39) used atomic force microscopy to visualize FtsZ polymers bound to E. coli lipid bilayers through ZipA. These ZipA-tethered FtsZ molecules formed a dynamic two-dimensional network of curved, interconnected protofilaments that seemed to be bundled. In contrast, ZipA incorporated into phospholipid bilayer nanodiscs did not trigger significant FtsZ polymer bundling (29). Finally, Loose and Mitchison (44) reconstituted the E. coli protoring components on supported lipid bilayers and showed that FtsA organized FtsZ polymers into dynamic patterns of coordinated streams and swirling rings with preferential directions, which suggested treadmilling. Importantly, these dynamics were sharply reduced when FtsZ protofilaments were attached to the membrane by ZipA or when using artificially membrane-targeted FtsZ. Although the resulting FtsZ polymers were described as bundled, the resolution obtained by total internal reflection fluorescence microscopy (TIRFM) probably could not distinguish between single and bundled FtsZ protofilaments. More recently, it was found that artificially membrane-bound FtsZ self-organizes into similar vortices, even in the absence of FtsA (45). This effect casts doubt on the dampening effects of ZipA on FtsZ dynamics observed previously.
In this study, we revisit the effect of ZipA on FtsZ protofilaments, including its role in polymer bundling. In contrast to the prevailing model, our in vivo results show that unlike Zaps, FtsZ* or the FtsA* gain-of-function mutant (48), ZipA does not play a significant role in FtsZ protofilament bundling. We further show that as previously reported (46,49) (M. Sobrinos-Sanguino, R. Richter, and G. Rivas, unpublished data), the surface concentration of ZipA is critical in controlling the activities and interactions with FtsZ in vitro. Using a His 6 -tagged soluble variant of ZipA (sZipA) immobilized on lipids, we demonstrate that this protein organizes FtsZ into similar swirling vortices of mostly single protofilaments, a role that was previously attributed to FtsA (44). These results provide further evidence that ZipA does not inhibit FtsZ polymer dynamics at the membrane.

RESULTS
FtsA* or excess FtsZ rescues the FtsZ bundling-deficient ⌬zapA ⌬zapC double mutant. Recently, we reported that FtsZ protofilament bundling is antagonized by FtsA, both in vivo and in vitro (13). For example, FtsA overproduction reverses the toxic effects of FtsZ overbundling triggered by excess ZapA. Conversely, even a slight excess of FtsA exacerbates the already moderately filamentous cell phenotype of the ΔzapA single mutant or the more severe filamentous phenotype of ⌬zapA ⌬zapC double mutant (13), which lack one or two FtsZ bundling proteins, respectively (18). We also previously reported that the self-bundling ftsZ* mutant (encoding FtsZ with the L169R substitution [FtsZ L169R ]) could completely suppress the cell division deficiency of the ⌬zapA ⌬zapC double mutant (27). This suggested that if FtsZ protofilaments are bundled by factors independent of ZapA and ZapC, then the requirement for the latter proteins in normal cell division could be bypassed.
We first surmised that a moderate increase in intracellular FtsZ concentration could promote polymer bundling simply by molecular crowding, and this could bypass the need for ZapA and ZapC and consequently suppress the filamentation of many ⌬zapA ⌬zapC double mutant cells ( Fig. 1A and B). To produce extra FtsZ, we used the pJF119HE-FtsZ plasmid (26). As expected, FtsZ overproduction suppressed ⌬zapA ⌬zapC cell filamentation (Fig. 1C), supporting the idea that higher FtsZ protein concentration favors increased lateral interactions between protofilaments.
Excess ZipA cannot counteract cell division defects caused by deficient FtsZ bundling. As already mentioned, E. coli FtsA inhibits FtsZ polymer bundling. Our in vitro results indicate that this occurs due to the unusual miniring polymers that purified FtsA forms on lipid monolayers. In contrast, purified FtsA* forms shorter curved oligomers under similar conditions. Although the mechanism is not yet known, these FtsA* arcs no longer inhibit FtsZ polymer bundling and instead permit or promote it, consistent with our in vivo results (Fig. 1D). This was most apparent when FtsA* was combined with FtsZ* on lipid monolayers: in a striking additive effect, large sheets consisting of many laterally associated protofilaments were formed (13). Interestingly, both gain-offunction mutants that promote FtsZ bundling, FtsA* and FtsZ*, bypass the need of the third protoring component, ZipA (27,48). This, along with evidence that purified ZipA can bundle FtsZ under certain conditions, led to the hypothesis that ZipA might also trigger FtsZ bundling (54).
In this scenario, excess ZipA should be able to rescue the ⌬zapA ⌬zapC cell filamentation phenotype, similar to excess FtsZ, FtsA*, or FtsZ* (27) (Fig. 1). To test this, we first transformed the ⌬zapA ⌬zapC double deletion strain and its TB28 wild-type parental strain (55) with pKG110-ZipA, a plasmid that expresses zipA from a salicylateinduced nahG promoter and a weak ribosome binding site that keeps expression low. Notably, uninduced levels of ZipA from pKG110-ZipA did not suppress the ⌬zapA ⌬zapC filamentous phenotype (Fig. 2B). Instead, the induction of zipA expression was consistently more toxic not only for the ⌬zapA zapC double deletion strain but also for the ⌬zapA single deletion strain compared with the wild-type parent TB28 strain ( Fig. 2A). Endogenous FtsZ was produced at similar levels in both ZipA-uninduced and -induced cells, ruling out the possibility that excess ZipA could affect viability through changes in the intracellular levels of FtsZ (see Fig. S1A in the supplemental material).
Further growth until late exponential phase exacerbated the already elongated cell phenotype of the ⌬zapA ⌬zapC double mutant strain both in the absence (Fig. 2C) and presence of inducer (Fig. 2D); cells of the ΔzapA single mutant behaved similarly (not shown). These results suggest that ZipA might not be a bundler of FtsZ polymers, contrary to what we initially expected.
The region of ZipA known to interact with FtsZ polymers is the FZB (FtsZ binding) globular domain at its C-terminal end (34,37,56,57). To exclude the possibility that the toxicity of excess ZipA for ⌬zapA ⌬zapC double mutant cells was due to the accumulation of transmembrane domains at septation sites (49) or because the N-terminal transmembrane region of ZipA might affect cell division by an unknown mechanism, we separated the FZB domain from the transmembrane region. For this purpose, we used a chimeric construct containing the C-terminal part of ZipA lacking the transmembrane region (ZipA with amino acids 23 to 328 [ZipA  ]) fused to the N-terminal transmembrane domain of DjlA (DjlA 1-32 ), a protein not related to cell division (37). This hybrid membrane protein containing FZB was cloned into pKG116, a plasmid similar to pKG110 but with a stronger ribosome binding site for increased gene expression. However, similar to the intact ZipA protein, the DjlA 1-32 -ZipA 23-328 (FZB) protein was toxic and exacerbated the phenotype of ⌬zapA ⌬zapC (Fig. 2E) and ΔzapA (not shown) mutant cells. This further suggests that binding of ZipA to FtsZ polymers does not promote their bundling, or at least the type of bundling that could compensate for the lack of ZapA and ZapC (18).
To test the model further, we asked whether excess ZipA could rescue the dominant-negative effects of an FtsZ allele (FtsZ R174D ) that was reported to be defective in polymer bundling (14). Although a subsequent study suggested that FtsZR174D was capable of bundling under certain conditions (58), we recently confirmed (K. M. Schoenemann, M. Krupka, V. W. Rowlett, S. Distelhorst, B. Hu, and W. Margolin, unpublished data) that this protein is indeed more bundling defective than wild-type FtsZ, as suggested in the original report.
To test this idea, we constructed a strain with two plasmids: pDSW210F-ZipA-GFP and either pKG110-FtsZ or pKG110-FtsZ R174D . We did so that expression of ZipA-GFP (ZipA fused to green fluorescent protein [GFP]) is controlled by isopropyl-␤-D-1thiogalactopyranoside (IPTG) and expression of the FtsZ derivatives is controlled by sodium salicylate. The ZipA-GFP is functional and can complement a zipA1(Ts) mutant (59). Expression of FtsZ R174D at any level above 1 M sodium salicylate was strongly dominant negative (Fig. S2), in contrast to FtsZ, which allowed viability even at 2.5 M (and higher [not shown]). Notably, ZipA, whether uninduced or induced with IPTG, was unable to counteract the dominant-negative effects of FtsZ R174D , consistent with the idea that ZipA does not promote FtsZ bundling (Fig. S2). This is in sharp contrast with hyperbundled FtsZ*, which is able to suppress the dominant-negative effects of FtsZ R174D (Schoenemann et al., unpublished). Interestingly, the toxicity of ZipA at IPTG concentrations above 50 M was antagonized by extra FtsZ. One possible explanation

E. coli ZipA Organizes FtsZ Polymers into Structures
® is that increased FtsZ bundling triggered by its higher intracellular levels (Fig. 1C) counteracts the negative effects of excess ZipA (Fig. S2).
FtsA* and FtsZ* confer at least 10-fold resistance to excess ZipA. In our recent studies, we demonstrated that FtsZ* has an intrinsic capacity to bundle compared with wild-type FtsZ (27), whereas FtsA* can promote bundling of wild-type FtsZ protofilaments (13). Moreover, both gain-of-function mutants correct the defective division phenotype of ⌬zapA ⌬zapC underbundled mutants (Fig. 1). If ZipA acts to bundle FtsZ polymers, its excess in an ftsZ* or ftsA* background should result in overbundling and be toxic for the cells by inhibiting cell division, as previously reported (27).
To test this idea, we transformed E. coli strains WM1659 and WM4915, which replace the native chromosomal ftsA or ftsZ with ftsA* or ftsZ* alleles, respectively, with pKG110-ZipA in the WM1074 (MG1655) strain background. We found that ZipA overproduction from the nahG promoter was toxic at 5 M sodium salicylate in the wild-type parent strain and became more toxic with 10 M inducer (Fig. 3). In contrast, the presence of ftsA* in strain WM1659 conferred full resistance against excess ZipA (Fig. 3), consistent with the original report (48). The effects of ftsZ* in strain WM4915 were more modest, but nonetheless resulted in at least a 10-fold increase in resistance at 5 M inducer (Fig. 3). The effects of ftsA*, ftsZ*, or ZipA levels on viability were not due to changes in FtsZ levels, as these remained unchanged in the various conditions (Fig. S1B). The ability of alleles that promote FtsZ protofilament bundling to antagonize ZipA toxicity instead of exacerbate it is yet another argument against the idea that ZipA is a bundler of FtsZ.
The zipA1(Ts) strain is also inviable at 37°C, and some factors can suppress the thermosensitivity of zipA1 at these lower temperatures, including inactivation of certain amino acid biosynthesis genes (59). This suggests that the ZipA1 protein is partially active at 37°C, although not sufficient to sustain viability. To give the Zap proteins the best chance of suppressing zipA1, we tested whether the Zap proteins might be able to partially compensate for a partially defective ZipA at this less stringent temperature. We found that neither ZapD nor ZapA were able to suppress zipA1 thermosensitivity at 37°C, but ZapC was (Fig. 4B). We also noticed a weak synergistic effect upon coexpression of both ZapA and ZapC, where there was a limited level of viability even at 42°C (Fig. 4C). Moreover, gene pairs zapA plus zapC, or zapA plus zapD, also conferred partial suppression of zipA1 thermosensitivity at 37°C (Fig. 4B). These results indicate that the Zap proteins and ZipA may have weak overlapping roles in FtsZ protofilament bundling, perhaps by enhancing the stability of the protoring and its tethering to the membrane and to the nucleoid (60). Low-surface-density ZipA organizes FtsZ into circular protofilament structures on lipid monolayers. So far, our in vivo data presented here are not consistent with the previous data that suggested ZipA is a major enhancer of FtsZ protofilament bundling. This prompted us to test whether ZipA had any effect on FtsZ bundling in an in vitro membrane system. For this, we examined the properties of FtsZ polymers on lipid monolayers. To date, this assay has been mainly used to visualize oligomeric structures of FtsZ protofilaments along with their FtsA membrane tethers by electron microscopy (13,35,43,61,62). Whereas FtsA has a short C-terminal amphipathic helix that acts as a membrane anchor (11,61,63), ZipA has a short N-terminal periplasmic region followed by a transmembrane domain (36,37,64). Consequently, full-length ZipA could not be used in our assay.
Therefore, we decided to use an N-terminally truncated ZipA (soluble ZipA [sZipA]) replacing the first 25 amino acids with an N-terminal His 6 tag (65). To attach sZipA to the lipid monolayer, input lipids were supplemented with a nickel-chelating lipid, 1,2-dioleoyl-sn-glycero-3-[(N-(5-amino-1-carboxypentyl)iminodiacetic acid) succinyl] (nickel salt) (DGS-NTA), that anchors the His 6 tag, thus mimicking the membrane topology of the full-length protein (44,49,65). The density of sZipA on the lipid monolayer surface was tuned by controlling the amount of NTA lipids added, as these two values are linearly proportional (M. Sobrinos-Sanguino, R. Richter, and G. Rivas, unpublished data). Importantly, we lowered the surface density of ZipA compared with previous studies (39, 44) by using 0.5 to 1% of NTA lipids instead of 10%, which more closely mimic physiologically relevant levels of ZipA. 0.5% NTA corresponds to a surface density of ~2,000 ZipA molecules per m 2 . Unperturbed E. coli cells contain~1,500 ZipA molecules per cell (66), which corresponds to around 400 molecules per m 2 assuming a uniform distribution. If 30% of these ZipA molecules are in a midcell ring that comprises 5 to 10% of the cell length, the estimated protein concentration in the ring would bẽ 2,000 molecules per m 2 , which is the low surface density we used. As expected, when FtsZ was added without sZipA and examined by negative stain transmission electron microscopy (EM), FtsZ polymers were scattered sparsely on the lipid monolayer, consistent with the requirement for a membrane anchor such as FtsA or ZipA (1). This residual binding was likely a result of random association of the FtsZ from the added solution onto the grid (Fig. S3B). However, when FtsZ was added to monolayers coated with low-density sZipA, we observed extensive FtsZ protofilament patterns. Most notably, these patterns differed depending on the concentration of NTA lipids, which in turn dictated the concentration of ZipA on the monolayer. For example, when FtsZ (1 to 5 M) was polymerized with nonhydrolyzable GTP {guanosine-5=-[(␣,␤)-methyleno]triphosphate (sodium salt) (GMPcPP)} on monolayers seeded with low-density ZipA (0.5% NTA lipids out of the total input lipids), it became strikingly organized into circular structures of mostly single protofilaments in a repetitive pattern (Fig. 5A). These circular structures contained an average of nine filaments. The external diameter was 279 Ϯ 50 nm, with a lumen, lacking filaments, of~100 nm in diameter. The lateral separation between the filaments was 10 Ϯ 4 nm. The filaments that were closest together mostly appeared as double filaments, but the filaments were very loose and noncontinuous (more than 70 structures were measured).
In the presence of GTP, which should support GTPase activity and filament treadmilling, the ring-like structures contained a smaller number of filaments (6 Ϯ 2) but were larger than the structures formed in GMPcPP, with an external diameter of 400 Ϯ 80 nm and a lumen 190 Ϯ 20 nm in diameter. The GTP-FtsZ filaments appeared more separated than those formed with GMPcPP, as the average separation was 20 Ϯ 9 nm (more than 50 structures were measured) (Fig. 5B). For both GTP and GMPcPP ring-like structures, the spacing measurements were compatible with the FtsZ filament arrangement found in the presence of FtsA minirings (13). To assess the effects of different lipids on these structures, we made lipid monolayers with 1,2-dioleoyl-sn-glycero-3phosphocholine (DOPC). Similar ring-like structures containing FtsZ were observed with GTP (Fig. S4).  Next, we asked whether increasing surface density of sZipA might affect the ring-like structures of FtsZ polymers. We saw no difference between monolayers containing 0.5% versus 1% NTA lipids (not shown). We then significantly increased the surface concentration of sZipA on monolayers by increasing the NTA concentration to 10%, mimicking ZipA overproduction in vivo. Whereas no oligomeric structures were detected with sZipA alone (Fig. S3A), when FtsZ was added to the sZipA at this high surface density, polymers were strikingly aligned into parallel tracks of long, straight protofilaments spaced~20 nm apart, and the formation of ring-like swirls observed at lower ZipA densities was abolished. Even at this high density of ZipA, most of FtsZ protofilaments remained unbundled (Fig. 5C). Whether FtsZ formed straight alignments or swirls was independent of FtsZ concentrations added to the reaction mixtures within the physiological range of 1.5 to 5 M (Fig. 5 and data not shown) (66).
FtsZ swirls formed at low ZipA densities are highly dynamic and driven by GTP hydrolysis. The ring-like structures formed by FtsZ polymers on low-density sZipA resembled the dynamic vortices formed either by FtsZ bound to membrane-attached FtsA or FtsZ fused to yellow fluorescent protein (FtsZ-YFP) and a membrane targeting sequence (FtsZ-YFP-mts) on supported lipid bilayers (44,45). This prompted us to analyze the dynamics of fluorescently labeled FtsZ protofilaments on bilayers containing 0.5% NTA (low-density sZipA) using confocal microscopy. When GTP was added to trigger FtsZ polymerization, we observed swirling vortices with a chiral clockwise rotation, similar to those from the aforementioned reports ( Fig. 6A; see Movie S1 in the supplemental material). A consistently negative slope of the kymographs (Fig. S5) confirmed the directionality of the rotation within the rings. The estimated rotational speed within these structures was~1.8 m min Ϫ1 . Similar, but markedly less dynamic, swirling rings were observed in the presence of GMPcPP (Fig. 6B; Movie S2). The estimated speed was 0.3 m min Ϫ1 , consistent with the idea that vortex formation is driven by GTP hydrolysis (45).
To visualize the structures of these vortices in more detail, we used superresolution microscopy (stimulated emission depletion [STED] microscopy). These structures were sharper than those imaged by standard confocal microscopy, and their size was similar to the size of the lipid monolayer-attached swirls observed previously by electron microscopy (Fig. 6C). We also used total internal reflection fluorescence microscopy (TIRFM) to visualize the FtsZ swirls at low ZipA surface density. We confirmed that these swirls formed both in GTP and GMPcPP (data not shown). However, the TIRFM approach, which is highly sensitive to the distance of the fluorophore from the surface, was hampered by significant image fluctuation, most probably due to the movement of the unstructured domain of ZipA. This precluded a more precise analysis of the FtsZ swirls by TIRFM.

DISCUSSION
Here, we provide in vivo and in vitro evidence that ZipA does not inhibit FtsZ treadmilling dynamics, unlike what was suggested previously (44). Instead, when FtsZ

E. coli ZipA Organizes FtsZ Polymers into Structures
® protofilaments are tethered to lipids by sZipA at levels that probably more closely mimic physiological conditions, they align and curve to form dynamic swirls that are very similar to those observed previously by FtsA-mediated tethering to lipids (44), fusion to an artificial membrane tether (45), direct adsorption to a mica surface (67,68), or when subjected to crowding agents (69). These swirls, whose dynamics depend on GTP hydrolysis, likely represent treadmilling FtsZ polymers that comprise the FtsZ ring in vivo (8,9). When we tested sZipA at an artificially high density on the lipid surface by applying a high (10%) concentration of NTA lipids, FtsZ protofilaments aligned into large, straight, apparently static structures that are micrometers long. This observation is consistent with a previous study using high surface concentrations of sZipA, which concluded that ZipA curtails FtsZ dynamics (44). Therefore, we propose that lower surface ZipA densities are necessary to allow FtsZ protofilaments the needed flexibility for their characteristic dynamic movement along the membrane, which is crucial for guiding septum synthesis (8,9).
Despite previous reports that ZipA bundles FtsZ when in solution, including stabilizing highly curved or circular forms of FtsZ polymers (40), here we present several lines of evidence that ZipA does not directly bundle FtsZ protofilaments at lower, probably more physiological, densities on lipid surfaces or in E. coli cells. When attached to a lipid monolayer at these densities, sZipA efficiently tethers and aligns FtsZ protofilaments, but close lateral associations were uncommon. Even at high surface densities of sZipA that promoted extensive and relatively static FtsZ filament alignments, most protofilaments remain apart, indicating that ZipA does not directly bundle FtsZ like FtsA* does (13).
Furthermore, if ZipA actually stimulates FtsZ protofilament bundling, then it might be expected to replace the bundling functions of Zap proteins in cells. Instead, and in contrast to FtsA*, excess ZipA failed to rescue the cell division deficiency of ΔzapA or ⌬zapA ⌬zapC mutant cells. ZipA also failed to counteract the dominant-negative phenotype of the likely bundling-defective FtsZ R174D . In another test of ZipA's bundling ability in vivo, it was predicted that excess ZipA might be more toxic in a bundlingproficient ftsA* or ftsZ* strain background compared with a normal background, due to FtsZ overbundling. Instead, the ftsA* or ftsZ* allele actually antagonized the toxicity of excess ZipA by at least 10-fold, suggesting again that ZipA is not acting significantly to bundle FtsZ. However, it is also possible that FtsA* and FtsZ* may have already maximally bundled the FtsZ in the cell, leaving no room for additional bundling by ZipA if it were to occur. The mechanism by which ftsA* or ftsZ* suppresses ZipA toxicity cannot yet be ascertained, as it is not yet known why excess ZipA is toxic.
These results suggest that ZipA is not a significant bundling factor for FtsZ or at least that its mechanism of action is distinct from that of Zaps, FtsA*, and FtsZ*. Nevertheless, extra ZapC could rescue the thermosensitivity of a zipA1 mutant at 37°C, and even at the most stringent temperature of 42°C, a combination of ZapA and ZapC was able to rescue growth somewhat. One explanation for this is that cross-linking of FtsZ polymers by extra ZapA/ZapC generally promotes FtsZ protofilament alignment in parallel superstructures (i.e., swirls) that mimic the swirls assembled by ZipA, thus stabilizing the protoring. Because it is not clear what functions of the mutant ZipA1 protein are compromised at less stringent nonpermissive temperatures, it is difficult to know what ZapC is rescuing at 37°C that it cannot rescue at 42°C. This brings up a broader question: why is ZipA essential for divisome function if it performs what seems to be a very similar function as FtsA? Both promote FtsZ protofilament alignment without permitting bundling in vitro, and their in vivo phenotypes are consistent with this, so why are both necessary in vivo? For example, when ZipA is inactivated, even in the presence of FtsA, recruitment of downstream divisome proteins is blocked, implicating ZipA in that essential function (70). We favor the idea that ZipA has additional roles in later divisome function that are distinct from those of FtsA. Furthermore, the ability of certain mutants such as FtsA* and FtsZ* mutants to bypass ZipA may not be due solely to restoration of FtsZ bundling. For example, FtsA* likely recruits downstream divisome proteins more effectively than FtsA, and it can accelerate cell division (28,51,71). It remains to be seen what these other activities of ZipA are and how they differ from the activities of FtsA. It was previously suggested that the ability of ZipA to form homodimers via its N-terminal domain might enhance FtsZ protofilament bundling (72). Although our lipid monolayer assays probably did not permit homodimerization of sZipA given that the native N terminus is missing, our genetic data using native ZipA suggest that its homodimerization does not significantly promote FtsZ bundling in vivo.
Another important question is how the FtsZ protofilaments become aligned as they self-assemble on lipids along with their membrane tethers. The study of plant microtubules may provide clues. During growth of the cortical microtubule array in plant cells, microtubules align with each other in a self-reinforcing mechanism. When the plus end of a microtubule meets another microtubule at an angle of less than 40°, the first polymer's plus end changes direction and ends up parallel with the encountered polymer. When faced with another microtubule at angles greater than 40°, the plus end is more likely to disassemble (catastrophe), thus selecting against crossovers and reinforcing parallel alignments (73,74). Such behavior, coupled with the tendency of intrinsically curved FtsZ protofilaments to adopt the intermediate curved conformation (67,75,76), could explain how the swirls become established and self-perpetuate. These curved groups of FtsZ polymers may be important to generate bending forces at the membrane (42,75). It is possible that highly curved FtsZ also has a role in this activity, given that FtsZ minirings with a diameter of only~25 nm can assemble on lipid monolayers (40). Although a specific type of membrane tether is not required for the generation of swirls (45), our data from this study and from our recent report (13) indicate that both FtsA and ZipA maintain FtsZ protofilaments in an aligned but mostly unbundled state. Yet the gain-of-function properties of FtsA* and FtsZ* and their ability to specifically promote FtsZ lateral interactions suggest that progression of the divisome requires a set of factors that ultimately switch FtsZ protofilaments to a bundled form. The ability of FtsA* and FtsZ* to bypass ZipA suggests that ZipA itself may be one of these factors but that it does not necessarily act directly on FtsZ.
Strains, plasmids, and cell culture. All E. coli strains and plasmids used in this study are listed in Table 1. Cells were grown in Luria-Bertani (LB) medium at 30°C, 37°C, or 42°C (as indicated) supplemented with the appropriate antibiotics (ampicillin 50 g ml Ϫ1 , chloramphenicol 15 g ml Ϫ1 , or tetracycline 10 g ml Ϫ1 ) and gene expression inducers IPTG (isopropyl-␤-D-1-thiogalactopyranoside) and sodium salicylate.
Overnight cell cultures were diluted 1:100 in the appropriate medium and grown until the cultures reached an optical density at 600 nm (OD 600 ) of 0.2, followed by their back-dilution 1:4. After the second dilution, cells were cultured to an OD 600 of 0.2 and spotted on plates at 1ϫ, 0.1ϫ, 0.01ϫ, 0.001ϫ, and 0.0001ϫ dilutions. For differential interference contrast (DIC) microscopy, the cells were further cultured in the presence of inducers, maintained in exponential phase, harvested 2 h after induction, and fixed with 1% formaldehyde.
Protein purification and labeling. E. coli FtsZ was purified by the Ca 2ϩ -induced precipitation method (78). The soluble mutant of ZipA lacking the transmembrane region (sZipA) was isolated as described previously (65). FtsZ and ZipA were labeled with Alexa Fluor probes (1:10 molar ratio). FtsZ was labeled under conditions that promote protein polymerization to ensure minimal interference of the dye with FtsZ assembly as described previously (79).
Lipid monolayer assay. Lipid monolayers were prepared as described previously (13,61). Briefly, 0.2 g of E. coli polar lipid extract supplemented with 0.5 to 10% of NTA lipids when needed were floated on Z-buffer (50 mM Tris-HCl [pH 7.5], 300 mM KCl, 5 mM MgCl 2 ) using a custom-made Teflon block (80) E. coli ZipA Organizes FtsZ Polymers into Structures ® and placed in a humid chamber for 1 h to evaporate the chloroform. Electron microscopy grids were then placed on the top of each well followed by sequential additions and incubations of 1 M sZipA (1 h), 0.5 to 5 M FtsZ (15 min), and 2 mM GTP or 0.5 mM GMPcPP (5 min). The grids were then removed followed by negative staining with uranyl acetate as described previously (27) and imaged with a JEOL 1230 electron microscope operated at 100 kV coupled with a TVIPS TemCam-F416 complementary metal oxide semiconductor (CMOS) camera. FtsZ protofilament spacing was measured using the Plot Profile tool in ImageJ (81).
Self-organization assays on supported lipid bilayers. Lipid bilayers were formed by fusion of small unilamellar vesicles (SUVs) mediated by CaCl 2 (82). Lipids (polar extract phospholipids from E. coli or DOPC) with or without NTA at 0.5 to 1% (wt/wt) ratios, were prepared by drying a proper amount of the lipid stock solution under a nitrogen stream and kept under vacuum for at least 2 h to remove organic solvent traces. The dried lipid film was dissolved in supported lipid bilayer (SLB) buffer (50 mM Tris-HCl [pH 7.5], 150 mM KCl) to a final 4-g/liter concentration, resulting in a solution containing multilamellar vesicles (MLVs). After 10-min sonication of MLVs, small unilamellar vesicles were obtained. A 1-mg/ml suspension of SUVs was added to a hand-operated chamber (a plastic ring attached on a clean glass coverslip using UV-curable glue [Norland optical adhesive 63]). SLBs were obtained by the addition of 2 mM CaCl 2 and incubated at 37°C for 20 min. Samples were washed with prewarmed SLB buffer to remove nonfused vesicles.
Confocal images were collected with a Leica TCS SP5 AOBS inverted confocal microscope with a 63ϫ immersion objective (numerical aperture [NA] of 1.4 to 0.6/Oil HCX PL APO, Lbd.Bl.) and confocal multispectral Leica TCS SP8 system with a 3ϫ STED (stimulation emission depletion) module for superresolution (Leica, Mannheim, Germany). Total internal reflection fluorescence microscopy (TIRFM) experiments were performed on a Leica DMi8 S wide-field epifluorescence microscope. Images were acquired every 0.3 s with Hamamatsu Flash 4 scientific CMOS (sCMOS) digital camera.
For self-organization assays, SLB buffer was replaced by Z-buffer prior to protein addition. The final volume of the assay mixtures was 100 l. First, 0.5 M Alexa Fluor 647-labeled sZipA was added on top of the lipid bilayer with a given amount of NTA lipids. Once the fluorescent sZipA was visualized as attached to the lipid bilayer, 1 M FtsZ-Alexa Fluor 488 was added, followed by the addition of 2 mM GTP (or 0.5 mM of GMPcPP) to induce FtsZ polymerization.
Data availability. We declare that all data supporting the findings of this study are available within the article or supplemental material or from the authors upon request.