MinC and FtsZ mutant analysis provides insight into MinC/MinD-mediated Z ring disassembly

The Min system negatively regulates the position of the Z ring, which serves as a scaffold for the divisome that mediates bacterial cytokinesis. In Escherichia coli, this system consists of MinC, which antagonizes assembly of the tubulin homologue FtsZ. MinC is recruited to the membrane by MinD and induced by MinE to oscillate between the cell poles. MinC is a dimer with each monomer consisting of functionally distinct MinCN and MinCC domains, both of which contact FtsZ. According to one model, MinCC/MinD binding to the FtsZ tail positions MinCN at the junction of two GDP-containing subunits in the filament, leading to filament breakage. Others posit that MinC sequesters FtsZ–GDP monomers or that MinCN caps the minus end of FtsZ polymers and that MinCC interferes with lateral interactions between FtsZ filaments. Here, we isolated minC mutations that impair MinCN function and analyzed FtsZ mutants resistant to MinC/MinD. Surprisingly, we found mutations in both minC and ftsZ that differentiate inhibition by MinC from inhibition by MinC/MinD. Analysis of these mutations suggests that inhibition of the Z ring by MinC alone is due to sequestration, whereas inhibition by MinC/MinD is not. In conclusion, our genetic and biochemical data support the model that MinC/MinD fragments FtsZ filaments

The assembly of the tubulin homologue FtsZ into the Z ring leads to the recruitment of more than two dozen proteins to form the divisome that mediates bacterial cytokinesis (1,2). Two negative regulatory systems, nucleoid occlusion and Min, cooperate to restrict Z ring assembly to midcell. In nucleoid occlusion, a site-specific DNA-binding protein SlmA blocks Z ring formation over the nucleoid (3). The binding sites are clustered around the origin of replication and chromosome segregation removes this negative signal from midcell (4). In the Min system, the concerted action of the three Min proteins prevents cell division away from midcell (5). MinC is an antagonist of FtsZ assembly (6,7) that is activated by MinD and induced to undergo pole-to-pole oscillation by MinE (8 -10). This produces a time-averaged gradient of MinC that is highest at the poles and lowest at the midcell (11).
MinC consists of two distinct functional domains, each of which contact FtsZ (12). The N-terminal domain (MinC N ) 4 interacts with FtsZ and antagonizes FtsZ polymerization in vitro (6). Overproduction of MinC N (without MinD) in vivo prevents Z ring formation leading to filamentation. In vitro MinC N is as active as full-length MinC in antagonizing FtsZ polymerization (12). The C-terminal domain (MinC C ) is responsible for dimerization and interaction with MinD, which is required for MinC C to bind to the conserved carboxyl-terminal peptide (CCTP) of FtsZ (12,13). Overexpression of MinC C in the presence of MinD competes with FtsA for binding to the CCTP leading to a block to Z ring function (13).
Mutations in ftsZ resistant to MinC/MinD alter residues located in two distinct regions, the CCTP and helix H10, and confer resistance to MinC C and MinC N , respectively (13,14). Further analysis of these mutations led to the two-pronged model for MinC/MinD inhibition of FtsZ assembly. In this model MinC C /MinD binds to a CCTP within an FtsZ filament tethered to the membrane by FtsA and ZipA, which positions MinC N at the interface between two FtsZ subunits (14). If GDP is present at the interface MinC N leads to breakage of the filament. Interestingly, the other negative spatial regulator of FtsZ assembly SlmA also uses a two-pronged mechanism (15). SlmA bound to DNA binds to the CCTP of FtsZ, which leads to a second step involving breakage of the filament. SlmA also requires that GDP be bound to the subunit interface otherwise (e.g. nonhydrolyzable GTP) it bundles the filaments (4,15).
Recently, Hernandez-Rocamora et al. (16) reported that MinC shortens FtsZ filaments by sequestering GDP-bound FtsZ subunits. Consistent with previous findings (6) MinC had a marginal effect on FtsZ filaments assembled with GMP-CPP, a slowly hydrolyzable analogue of GTP. Fluorescence anisotropy analysis further revealed that MinC and FtsZ form a 1:1 stoichiometric complex with a dissociation constant of about 10 M. Because FtsZ filaments had a narrow length distribution in the presence of MinC, they argued against a severing mechanism that should produce a random distribution.
To investigate the MinC mechanism under more physiological conditions, Arumugam et al. (17) devised reconstitution experiments in which FtsZ filaments were artificially directed to a planar membrane. A 1:1 mixture of FtsZ-YFP-MTS (the membrane targeting sequence from Escherichia coli MinD) and FtsZ, or alternatively FtsZ-MTS and FtsZ-F268C-Cy5, was added to a supported lipid bilayer. Once a stable but dynamic FtsZ bundle network emerged, enhanced green fluorescent protein-MinC was introduced to induce disassembly. Unlike FtsZ in solution, FtsZ-YFP-MTS has a low critical concentration (0.1 M) for polymerization likely due to membrane interaction. A kinetic analysis of the concentration-and time-dependent disassembly indicated that MinC decreases the FtsZ attachment rate and increases the detachment rate. Because no interaction between MinC and FtsZ F268C -Cy5 was observed in solution, they excluded a sequestration mechanism. Instead, they proposed that MinC C promotes subunit detachment by decreasing bundling and that MinC N binds to the minus end of FtsZ filaments to hinder the addition of new subunits (capping).
To gain further insight into the mechanism of FtsZ depolymerization mediated by MinC N , we isolated additional mutations in this domain and further characterized FtsZ mutants identified by Shen and Lutkenhaus (13,14) that provide resistance to MinC/MinD. Our results indicate that depolymerization of FtsZ by MinC is distinct from that caused by MinC/ MinD. Inhibition of FtsZ assembly by MinC is due to sequestration, whereas inhibition by MinC/MinD is not due to sequestration or capping the ends of FtsZ filaments. Our genetic and biochemical evidence, instead, points to the idea that MinC/MinD fragments FtsZ filaments.

Isolation of minC mutations affecting MinC N activity
Only one point mutation affecting the MinC N domain (minC G10D ) has been characterized so far (6,18). To try and gain additional insight into the mechanistic basis of FtsZ inhibition by MinC/MinD, we used PCR random mutagenesis to isolate additional mutations that inactivate minC N . This approach was used because MinC N is poorly conserved, in contrast to MinC C , which is highly conserved (Fig. 1A). Because MinC proteins from various bacterial species were aligned using Clustal W. The color code indicates the degree of conservation, with red being the most conserved and blue being the least conserved. B, MinC residues important for MinC N function. The residues corresponding to those identified in this study that define the FtsZ inhibitory region of MinC N are indicated on the T. maritima MinC structure (PDB code 1HF2).

MinC/MinD mechanism
there is no selection for such mutants it was necessary to find conditions to distinguish them from clones with no mutation as well as clones with mutations inactivating MinC C . The desired mutants should display reduced toxicity (loss of MinC N function) but localize to the Z ring in the presence of MinD (retain MinC C function and therefore some toxicity) (13,14). Mutagenized minC was cloned into pMCW26 (P lac ::gfp-minC G10D /minD) to replace minC G10D and the mutant library transformed into JS964 (⌬min::kan) and screened for clones that behave similarly to minC G10D . To do this, transformants were streaked onto plates containing 10 and 1000 M IPTG. A strain expressing WT minC (JS964/pMCW84 (P lac ::gfp-minC/minD)) is unable to form colonies at 10 M IPTG, whereas strains expressing an inactive MinC N are able to grow (JS964/pMCW26 (P lac ::gfp-minC G10D /minD). At 1000 M IPTG a strain expressing an inactive MinC N is unable to grow (due to MinC C /MinD competing with FtsA and ZipA for FtsZ binding (13)), whereas one with an inactive MinC C (unable to dimerize or bind MinD) is able to grow because such mutants are nontoxic.
Clones that passed this test were subjected to a visual screen for confirmation. Induction of GFP-MinC/MinD at 10 M IPTG produced filamentous cells with no fluorescent crossbands (Z rings destroyed), whereas induction of GFP-MinC C / MinD and GFP-MinC G10D /MinD resulted in cells that were much less filamentous and contained fluorescent cross-bands indicating Z rings were not disrupted ( Fig. 2A). Two clones grew at 10 M IPTG (with cells containing fluorescent crossbands) but did not grow at 1000 M IPTG indicating an inactive MinC N but an active MinC C . Sequencing minC from these plasmids revealed that one clone harbored a single point mutation, minC K35E . A positive charge at this position is fairly conserved among diverse MinC proteins (Fig. 1A). The other clone had two mutations, minC A39D and minC D144G . Separating these two mutations by site-directed mutagenesis revealed that minC A39D was responsible for the effect on MinC.
To further evaluate these mutations their effect on the spatial regulation of cell division was examined by introducing them into pSEB12, a mini-F plasmid containing an intact min operon (19). This plasmid complements the ⌬min phenotype of JS964 as evidenced by wildtype morphology and the absence of minicells. JS964/pSEB12 containing minC K35E or minC A39D resulted in a Min Ϫ phenotype similar to the control containing minC G10D (Fig. 2, B and C, and Fig. S1). Thus, we conclude that these mutations disrupt MinC function.

Identification of an inhibitory region within MinC N that is critical for Z ring regulation
The MinC N domain from Salmonella typhimurium (PDB code 3GHF) and E. coli (PDB code 4L1C) crystallize as dimers due to domain swapping of the first ␤ strand of each subunit ( Fig. S2) (20). We crystallized MinC 101 and also obtained a domain-swapped dimer (data not shown). We observed that MinC 101 forms a dimer in the presence of high salt, but this was prevented by the presence of FtsZ. In the structure of full-length MinC from Thermotoga maritima (PDB code 1HF2) the MinC N domain does not form a swapped dimer (21). For these reasons, it is likely that the swapped MinC N dimer observed with the enteric MinCs is a crystallographic artifact.
The two residues, Lys-35 and Ala-39, identified above were located on the EcMinC structure assuming it is not a swapped dimer (Fig. S1). This revealed that they are close together at the distal end of ␣1 with their side chains exposed to solvent. Moreover, these residues are on the same surface as Gly-10 (Fig. S2).

MinC/MinD mechanism
are also located on the same surface on the TmMinC structure (Fig. 1B).
To further examine the region of MinC N identified by the above mutations we used site-directed mutagenesis to alter the residues around this area. We introduced five point mutations, minC E7A , minC E7K , minC K9A , minC Q38A , and minC F42E into pMCW84 (P lac ::gfp-minC/minD) and examined their effect on the inhibitory activity of MinC. The K9A substitution was made because a positive charge at this position is conserved in all MinCs (Fig. 1A). Three of these mutations (minC E7A , minC E7K , and minC Q38A ) did not affect MinC function as Z rings were readily disrupted (only E7K and Q238A are shown). Furthermore, when these three mutations were introduced into the min operon on pSEB12 the ⌬min phenotype of JS964 was complemented ( Fig. 2, B and C, and Fig. S1). The other two mutants, however, MinC K9A and MinC F42E , behaved similarly to MinC G10D , MinC K35E , and MinC A39D . At 2 h after induction of pMCW84 (P lac ::gfp-minC/minD) with 10 M IPTG the derivatives with these mutations contained fluorescent cross-bands although some were distorted ( Fig. 2A). Consistent with this, pSEB12 carrying minC K9A , minC A39D , or minC F42E failed to complement the ⌬min phenotype (Fig. 2, B and C, and Fig. S1). Because the point mutations affecting residues on the ␣1-␤1 surface ( Fig. 1B and Fig. S2) compromise MinC N activity without affecting protein stability (Fig. S3), they define the FtsZ inhibitory region of MinC N .
Because the MinC mutants, MinC K9A , MinC K35E , MinC A39D , and MinC F42E , are impaired for function in vivo we tested their effect on FtsZ polymerization in vitro using a sedimentation assay. A MalE-MinC fusion protein prevents FtsZ sedimentation in a dose-dependent manner and this activity is due to to MinC N (6,12). The four MinC N mutants, as well as MinC G10D , had little effect on FtsZ sedimentation (Fig. 3A). Thus, the inability to inhibit FtsZ sedimentation correlates with the reduced ability of these mutants to disrupt Z rings and their inability to correct a minicell phenotype.

MinC N overexpression inhibits cell division by sequestration
MinC inhibits cell division in the absence of MinD when it is expressed ϳ40-fold over the physiological level (22), whereas MinC G10D does not, indicating the inhibitory activity is mediated by MinC N (6). Also, in the absence of MinD, MalE-MinC N is as active as MalE-MinC, whereas MalE-MinC C has no activity (12). To test the four new MinC mutants on the inhibitory activity of MinC in the absence of MinD, we used a vector expressing just MinC N (P T7 ::minC 115 -6ϫhis). This plasmid as well as derivatives with the mutations were transformed into BL21 (DE3). At 50 M IPTG, expression of minC 115 blocked colony formation (Fig. 4A). In contrast, expression of minC 115(K9A) , minC 115(G10D) , minC 115(K35E) , minC 115(A39D) , and minC 115(F42E) did not (Fig. 4, A, B, and D). Expression of these alleles from another vector (P tac ::6ϫhis-minC 115 ) in JS964 produced the same pattern (Fig. S4). Thus, mutations that reduce MinC N 's ability to prevent FtsZ sedimentation in vitro also reduce the ability of MinC N to inhibit colony formation.
Based on our previous findings (14) and this study, we assume that the inhibitory region of MinC N binds to FtsZ in a region that includes helix 10. To test how the minC mutations affect MinC N binding to FtsZ we used a biosensor assay (see "Experimental procedures"). Purified MinC 115 -His 6 and the various mutants were loaded onto Ni-NTA biosensor pins and incubated with FtsZ. Most of the mutants (MinC 115(G10D) , MinC 115(K35E) , MinC 115(A39D) , and MinC 115(F42E) ) were significantly impaired for FtsZ binding relative to MinC 115 (Fig. 5A). The binding of MinC 115(K9A) was the least affected (only reduced 2-fold) suggesting that 50% loss of binding affinity is sufficient to prevent MinC 115(K9A) from blocking FtsZ sedimentation in vitro and colony formation in vivo.
Because the binding affinity under our conditions is in the micromolar range we tested whether the interaction between MinC 115 and FtsZ could be detected by gel filtration. Because some FtsZ oligomers are present when FtsZ is purified, we used FtsZ L178E , which exists as a monomer regardless of the presence or absence of GTP and Mg 2ϩ (23). Each protein eluted as a single peak indicative of monomers (Fig. S5A), however, when the two proteins were mixed, a single peak was obtained that was shifted to a lower volume (Fig. S5B) at a position consistent with a 1:1 stoichiometric complex. The presence of the two proteins in the peak was confirmed by SDS-PAGE (Fig. S5C). In contrast, when MinC 115(G10D) -His 6 was mixed with FtsZ L178E ,

MinC/MinD mechanism
the proteins eluted in discrete peaks that were identical to their respective monomeric peaks (data not shown), indicating that it had markedly reduced affinity for FtsZ L178E .
Because MinC 115 and FtsZ L178E form a complex that can be detected by gel filtration in the same buffer we used for the in vitro FtsZ sedimentation experiments, we reasoned that MinC N blocks FtsZ polymerization by sequestering FtsZ monomers. This is consistent with the report by Hernandez-Rocamora et al. (16) that MinC shortens FtsZ filaments by forming a 1:1 complex. However, in our studies, the apparent K d values were in the range of 1ϳ6 M, whereas Hernandez-Rocamora et al. (16) determined a value of ϳ10 M. Examination of these studies revealed different buffer conditions were used in the polymerization assays so we employed the biosensor assay to assess the effects of buffer on the interaction. An increase in pH clearly weakened the interaction of MinC 115 -His 6 with FtsZ L178E (Fig. S6A). In agreement with Hernandez-Rocamora et al. (16) increasing the salt concentration also reduced the binding (Fig. S6B). In contrast, varying the salt concentration had no effect on FtsZ binding to His 6 -ZipA ( Fig. S6C). At low ionic strength (50 mM KCl) and pH 6.5 the apparent dissociation constant for the MinC-FtsZ interaction was comparable with one that we previously reported (ϳ1 M) (Fig. S7, A and B), however, it shifted to ϳ14 M at pH 7.5 (Fig. S7, C and D). These results indicate a significant electrostatic contribution to the interaction between MinC N and FtsZ. Consistent with the biosensor assay and Hernandez-Rocamora et al. (16), MinC did not block FtsZ sedimentation in a high salt buffer (500 mM KCl) (Fig. 3B). In contrast, MalE-SulA prevented FtsZ polymerization regardless of the ionic strength.
FtsZ is estimated to be in ϳ10-fold excess over MinC in vivo (24). Hence, a ϳ40-fold overexpression of MinC in the absence of MinD will likely result in substantial sequestration of FtsZ even if biochemical parameters such as pH and salt concentrations are taken into consideration. Also, the FtsZ level only has to be decreased by ϳ30% to inhibit division (25). Therefore, our results indicate that inhibition of FtsZ sedimentation by MinC in vitro correlates with inhibition of cell division by overexpression of MinC in the absence of MinD in vivo. In both cases inhibition of FtsZ assembly is caused by MinC N binding to free FtsZ subunits, thereby reducing the FtsZ subunit pool available for polymerization.

MinC/MinD mechanism Genetic evidence for an electrostatic contribution to the inhibitory activity of MinC N
The above results suggest that MinC N binding to FtsZ involves an electrostatic component. Consistent with this, mutations that reduced the activity of MinC N either removed a positively charged residue (K9A or K35E) or introduced a negatively charged residue (G10D, A39D, and F42E). Mutations that did not reduce MinC activity either increased the positive charge (E7K), did not affect the charge (Q34A), or removed a negative charge (E7A). One possibility is that a positive electrostatic potential in the inhibitory region might be critical for MinC function, so we examined several mutations in more detail.
Interestingly, JS964 (min::kan) cells containing pSEB12-E7K (P BAD ::minC E7K /minD/minE) were somewhat longer than those expressing the wildtype min operon (Fig. 2B and Fig. S1). Because no minicells were detected (Fig. 2C), minC E7K might represent a more active form of MinC due to an effect on the electrostatics. To test this possibility, we introduced minC E7K into pQE80L-E (P tac ::6xhis-minC 115 ) and expressed it in JS964. Indeed, at 100 M IPTG MinC 115(E7K) suppressed cell division and colony formation more effectively than MinC 115 (Fig. S4). Reexamination of inhibition by GFP-MinC E7K /MinD revealed that it was also more active than the wildtype (data not shown).
Because we suspected that electrostatics played a role in the MinC mechanism, we wondered what effect the E7K mutation would have on those minC mutations that introduce a negatively charged residue in the FtsZ inhibitory region. Interestingly, pSEB12 containing minC E7K/K35E /minD/minE complemented the ⌬min phenotype of JS964 indicating E7K was an intragenic suppressor of K35E (Fig. 2, B and C, and Fig. S1). In addition, induction of GFP-MinC E7K/K35E /MinD with 10 M IPTG led to a loss of cross-bands (JS964/pMCW84 -7/35 (P lac ::gfp-minC E7K/K35E /minD), suggesting that the double mutant is at least as active as wildtype MinC ( Fig. 2A). In addition, adding E7K to minC K35E , minC A39D , and minC F42E , as well as minC G10D , in the context of the min operon, restored MinC activity as the Min Ϫ phenotype was complemented (Fig. 2, B and C, and Fig. S1). E7K also rescued the inhibitory activity of these mutations when expressed from pSEB104CD (P ara ::minC/minD) (Fig. 6). Thus, those mutants that have lost inhibitory activity due to introduction of a negative charge in MinC N were rescued by the addition of the E7K mutation.
In light of the above findings, we speculated that the loss of activity due to the minC G10D mutations could be due to the introduction of the negatively charged aspartate residue. Therefore, lysine was substituted for glycine. JS964/pSEB12 containing minC G10K displayed a wildtype phenotype (Fig. 2, B  and C, and Fig. S1). Consistent with minC G10K behaving like WT MinC, we did not detect any cross-bands when GFP-MinC G10K /MinD was induced with 10 M IPTG in JS964/ pMCW84 -10 (P lac ::gfp-minC G10K /minD) (Fig. 2A). These results suggest that it is not the size of the residue at this position but the charge that results in the loss of the activity. As controls, we altered charged residues just outside of the putative FtsZ inhibitory region of MinC N . Mutations minC K44A and minC G72E (Fig. 1) did not affect MinC activity as pSEB12 con-taining these mutations complemented the Min Ϫ phenotype (Fig. 2, B and C). In summary, the loss of activity of MinC due to addition of a negative charge in the inhibitory region can be recovered by increasing the positive charge nearby.

Mechanism of Z ring disassembly by MinC/MinD on the membrane is distinct from sequestration
Although our study of MinC N indicated inhibition by sequestration, the level of MinC is about 10-fold lower than FtsZ (22,24) suggesting a sequestration mechanism is unlikely for MinC/MinD. We therefore tested whether E7K, which rescues minC mutations in the context of MinC/MinD, also rescues in the context of minC 115 . To do this E7K was combined with each of the minC mutations and expressed in BL21(DE3)/pET2␣-MinC 115 (P T7 ::minC 115 -6xhis). In contrast to what was observed with MinC/MinD (Fig. 6) most mutations were not rescued (Fig. 4, A-C). Only minC 115(K9A) was rescued (Fig. 4B). Similar results were obtained when minC E7K was added to minC mutations and then expressed in JS964/pQE80L-MinC 115 (P tac ::6xhis-minC 115 ) (Fig. S4). Because E7K failed to rescue the in vivo activity of these mutants we suspected it would not rescue the binding to FtsZ. Consistent with the in vivo results, E7K rescued K9A binding

MinC/MinD mechanism
but not that of the other mutants (Fig. 5, B and C). In summary, we find that E7K rescues mutations that introduce a negative charge in the MinC N domain when present in the MinC/MinD context but it does not rescue these mutations in the context of just MinC N , either in vivo or when testing the interaction with FtsZ in vitro. With MinC N we see very good correlation between the affinity for FtsZ in vitro and inhibition of Z ring formation in vivo indicating inhibition of FtsZ is by sequestration. However, our data strongly indicate that MinC/MinD antagonizes Z rings through a distinct mechanism.

A minC mutation that differentially affects inhibition by MinC N versus MinC/MinD
To further explore the difference between inhibition by MinC/MinD and MinC N we substituted the conserved lysine residue at position 9 with arginine (K9R). A lysine is present in most MinCs and therefore might be critical for MinC function (Fig. 1A). Importantly, this substitution replaces this highly conserved lysine with a bulky arginine residue while maintaining the positive charge. MinC 115(K9R) inhibited cell division, albeit a little less effectively than MinC 115 (Fig. 4D). This was also observed in JS964/pQE80L-MinC 115 (Fig. S4). Consistent with the K9R substitution not affecting inhibitory activity, the binding of MinC 115(K9R) to FtsZ was unaffected (Fig. 5A). Surprisingly, however, when MinC K9R was expressed along with MinD (JS964/pSEB104C K9R D [P BAD ::minC K9R /minD]) inhibition was not observed (Fig. 6B). It was not only defective in inhibiting colony formation but it also failed to cause cell filamentation (Fig. S8). Also, JS964/pSEB104C K9R DE (P BAD ::minC K9R /minD/minE) failed to complement a ⌬min phenotype (data not shown). These results show that the K9R mutation eliminates MinC activity in the context of MinC/ MinD but has little effect in the context of MinC N . Thus, MinC K9R in the cytoplasm functions as an FtsZ inhibitor, consistent with a sequestration mechanism, whereas MinC K9R / MinD on the membrane is inactive even though it should localize to the Z ring and bind to the ends of FtsZ polymers. Thus, the K9R mutation differentially affects the activity of MinC depending on the context and indicates distinct mechanisms.

Characterization of FtsZ mutants resistant to MinC in vivo
The differential effect of the K9R mutation on inhibition by MinC N (cytoplasmic MinC) versus MinC/MinD (on the membrane) indicates distinct mechanisms. To gain additional support for this we tested ftsZ mutations that we previously isolated as resistant to MinC/MinD for resistance to MinC 115 . Consistent with our previous reports, S7/pBang112 (P con ::ftsZ) carrying ftsZ mutations that alter residues in helix 10 provided resistance to MinC/MinD (Fig. S9). However, resistance to MinC 115 (P tac ::minC 115 ) was more varied (Fig. 7A). FtsZ R271G , FtsZ N280D , and FtsZ L270V exhibited resistance to MinC 115 , whereas FtsZ R271C and FtsZ I294T exhibited much less resistance. Strikingly, FtsZ E276D , although resistant to MinC/MinD, was completely susceptible to MinC 115 (Fig. 7A).
To determine whether resistance to MinC 115 correlated with FtsZ binding the interaction of MinC 115 -His 6 with the various FtsZ mutants was examined using the biosensor assay. Two mutants were at the extreme. FtsZ R271G , which showed the most resistance to MinC 115 , displayed no binding (Fig. 7B). In contrast, FtsZ E276D , which was as sensitive to MinC 115 as WT FtsZ, bound to MinC 115 as well as wildtype FtsZ. The two mutants that showed some resistance showed intermediate binding (FtsZ N280D and FtsZ I294T ). The interaction of FtsZ E276D (and the lack of interaction of FtsZ R271G ) with MinC 115 was verified by gel filtration (data not shown). Because the proximal region of helix 10 (Leu-270 and Arg-271) seemed important for resistance to MinC/MinD and MinC 115 , as well as binding to MinC 115 , we made the ftsZ L272E mutation. This mutation, like ftsZ R271G , eliminated MinC 115 binding (Fig. 7B). Resistance to MinC could not be tested, however, due to the inability of L272E to complement in vivo (data not shown). Nonetheless, this result supports the idea that binding of MinC 115 to FtsZ involves helix 10 and the degree of resistance to MinC 115 correlates with a reduction in binding.
Because FtsZ E276D is resistant to MinC/MinD, expression of GFP-MinC/MinD in a strain carrying FtsZ E276D should not cause filamentation and cells should contain fluorescent crossbands. To confirm this, pBang85 (P lac ::gfp-minC/minD) was introduced into S7/pBang112-E276D. For controls pBang85 was also introduced into S7/pBang112-WT, S7/pBang112-E276D/I374V, and S7/pBang112-I374V. When GFP-MinC/ MinD was induced with 10 M IPTG for 30 min in the strain with WT FtsZ, cross-bands were rarely observed and cells were filamentous (Fig. 8). In contrast, cells were not that filamentous and cross-bands were observed in cells containing FtsZ E276D , consistent with resistance to MinC/MinD (Fig. 8). The I374V mutation provides some resistance to MinC/MinD because MinC C /MinD does not localize to Z I374V rings (13). Consistent with this, cells were not filamentous and the fluorescence was at the membrane in strains S7/pBang112-E276D/I374V and

MinC/MinD mechanism
S7/pBang112-I374V (Fig. 8) and no cross-bands were observed (unable to localize to Z I374V rings). We also analyzed MinC K9R as it should localize to Z ring without disrupting them. We transformed S7/pBang112-WT with pBang85 carrying gfp-minC K9R /minD. Following induction, cross-bands were observed and cells were not filamentous consistent with the mutation inactivating MinC N in the MinC/MinD context (Fig.  8). Thus, GFP-MinC K9R /MinD behaves similarly to GFP-MinC G10D /MinD and the other MinC mutants we isolated (Fig.  2). Also, FtsZ E276D is resistant to MinC/MinD even though it is sensitive to MinC 115 in vivo and binds MinC 115 in vitro. These results support the notion that the mechanisms of inhibition by MinC in the cytoplasm and MinC/MinD on the membrane are different.

Characterization of FtsZ mutants resistant to MinC in vitro
As stated earlier, the prevention of FtsZ sedimentation by MinC in vitro correlates with the ability of overexpressed MinC 115 to inhibit cell division and colony formation in vivo. This suggests that the polymerization of FtsZ E276D should be susceptible to MinC 115 . To test this FtsZ was incubated with MinC 115 -His 6 in the presence of GTP and filaments recovered by sedimentation. Although FtsZ E276D polymerization was a little less efficient than wildtype FtsZ, it was indeed susceptible to MinC 115 (Fig. 9A). This is consistent with MinC 115 inhibiting FtsZ E276D by sequestration. On the other hand, the sedimentation of FtsZ R271G , which is resistant to MinC 115 in vivo, was largely unaffected by MinC 115 . Also, MinC 115(K9R) antagonized FtsZ polymerization as effectively as MinC 115 , whereas MinC 115(G10D) did not (Fig. 9B). These results further demonstrate that the inhibition of division by overexpression of MinC in the absence of MinD correlates with the ability of MinC to bind FtsZ and prevent FtsZ sedimentation in vitro. However, as detailed above this ability is separable from the mechanism used by MinC/MinD to inhibit division.

Discussion
Our earlier studies (13,14) led to a model for the mechanism of MinC/MinD in antagonizing Z ring assembly. In this model MinC C /MinD captures the CCTP of FtsZ in a filament positioning the MinC N domain to interact with the H10 helix of FtsZ at the interface of two FtsZ subunits, and if GTP has been hydrolyzed, to cause filament breakage. This model highlights synergistic cooperation between MinC C and MinC N , and also explains why FtsZ mutants defective in GTPase activity are resistant to MinC/MinD (6). However, due to a paucity of direct evidence that MinC/MinD severs FtsZ filaments, questions remain about the mechanism. In this study, we sought to advance our understanding of the mechanism employed by MinC by isolating and characterizing mutations in the MinC N domain defective in blocking Z ring assembly. Our study revealed that residues important for MinC N function are clustered on the MinC structure, which we designated the FtsZ inhibitory region, and verified that residues located at the proximal end of helix H10 of FtsZ are required for MinC activity. Also, our examination of MinC E7K , MinC K9R , and FtsZ E276D clearly distinguished inhibition by MinC (in the absence of MinD) from inhibition by MinC/MinD. The results reveal that MinC inhibits by sequestration, whereas MinC/MinD, the physiologically relevant mechanism, prevents FtsZ polymerization through a distinct mechanism not involving sequestration or capping suggesting the mechanism is severing.
Cell division inhibition by MinC/MinD has largely been considered to be similar to the inhibition observed by overexpression of MinC, which correlates with the prevention of FtsZ sedimentation in vitro (6). MinD activation of MinC was largely thought to be due to recruiting MinC to the membrane and increasing its affinity for the CCTP of FtsZ (13,26). Many MinC mutants isolated here in the context of MinC/MinD (K9A, K35E, A39D, and F42E) were defective in inhibiting cell division when overexpressed in the absence of MinD. In addition, they were compromised for binding to FtsZ and were unable to prevent FtsZ sedimentation in vitro. Consistent with this, size  MinC/MinD mechanism exclusion chromatography revealed that MinC interacts with FtsZ through its MinC N with a 1:1 stoichiometry, whereas the MinC mutants failed to form a complex. Even though this interaction is quite strong at low pH and low ionic strength (K d ϳ 1 M) it is weaker under physiological conditions. However, we see a good correlation between binding in vitro and inhibition of Z ring formation in vivo indicating that sequestration is the likely mechanism. However, inhibition of FtsZ by MinC/ MinD is unlikely to be due to sequestration because FtsZ is in 10-fold excess over MinC. Although capping remained a possibility, the CCTPs internal to the filament would compete with the terminal subunit CCTP for MinC/MinD, making capping less likely.

Mutations differentially affecting MinC versus MinC/MinD
Three different mutations clearly differentiate inhibition by MinC from that of MinC/MinD. The K9R mutation in the minC 115 background inhibited cell division and colony formation and as expected, the MinC K9R mutant bound to FtsZ in a biosensor assay and prevented FtsZ polymerization in vitro. Surprisingly, however, MinC K9R did not inhibit cell division nor induce cell filamentation in the presence of MinD. In addition, fluorescence microscopy revealed that Z rings were not disrupted by MinC K9R /MinD and instead were decorated with GFP fusions to MinC K9R /MinD as expected for loss of MinC N function. These results indicate that MinC/MinD inhibits cell division via a mechanism distinct from FtsZ sequestration. Because MinC K9R can still bind FtsZ, and therefore, should bind to helix H10 of the terminal subunit in an FtsZ filament, a capping mechanism can also be ruled out.
This difference between inhibition by MinC and MinC/ MinD is also supported by the FtsZ E276D mutant. A strain carrying the ftsZ E276D allele was sensitive to MinC 115 overexpression. Consistent with this, FtsZ E276D bound MinC 115 as well as WT FtsZ and was unable to polymerize in the presence of MinC 115 in vitro. Nonetheless, FtsZ E276D was resistant to MinC/MinD in vivo and Z E276D rings were decorated with GFP-MinC/MinD and not readily disrupted. Study of the E7K mutation also highlighted the difference between inhibition by MinC and MinC/MinD. The E7K mutation produced a more active MinC and suppressed minC mutations that decreased the positive electrostatic potential. Although this mutation suppressed the minC mutations in the context of MinC/MinD it had little effect on these mutations in the context of MinC 115 .

Implications of the mutations on mechanism
Besides further confirming differences between the mechanisms of the two forms of MinC the mutations lead to speculation about the mechanism of MinC/MinD. In SaFtsZ filaments produced with PC190723, the loop and proximal region of H10 constituting the FtsZ intersubunit interface is only partially Figure 10. The FtsZ residues important for MinC activity are located at the interface of the subunits. A, three subunits of a Staphylococcus aureus FtsZ filament assembled with PC197023 (PDB code 3VOB) (40). The residues important for MinC/MinD activity are indicated (colored red and green). The C-terminal tail of FtsZ emerges from the magenta colored residue and is indicated by a dotted line with the CCTP at the end. B, model for the interaction of MinC/MinD with an FtsZ filament. A MinD dimer is bound to the membrane and interacts with a MinC dimer through the MinC C domain, which enhances its affinity for FtsZ's CCTP. This interaction puts the MinC N domains in position to interact with the negatively charged FtsZ filament, which is tethered to the membrane by FtsA and ZipA. It is possible that the Lys-9 residue of MinC interacts with the Glu-276 residue of FtsZ to weaken the interaction between adjacent FtsZ subunits, which helps to sever the filament. It is also possible that another MinD dimer is bound to the other MinC C domain in the MinC dimer (43).
MinC/MinD mechanism exposed to solvent (Fig. 10A). Thus, we presume that the proximal (residues Leu-270 and Arg-271) and the central region of H10 (residues Glu-276 and Asn-280) become more exposed to solvent once GTP hydrolysis occurs. Loss of the T7 loop contacts may weaken the subunit-subunit interaction and expose the proximal end of helix 10 (41, 42). Although the mechanistic details are beyond the scope of this work, we would suggest that once MinC/MinD is bound to a CCTP in an FtsZ filament the MinC N domain interacts with the filament through an electrostatic interaction (Fig. 10B). Mutations that introduce a negative charge or reduce the positive charge in MinC N (K35E, A39D, G10D, F42E, and K9A) are unable to interact with the filament, but this deficiency can be restored by addition of the E7K mutation. In contrast, MinC K9R retains the positive charge but is still inactive in the context of MinC/MinD suggesting that a positive charge is not sufficient and that a specific steric interaction is required. A specific steric interaction is also suggested by the failure of the FtsZ E276D mutant to respond to MinC/ MinD. This suggests that once the MinC N domain interacts with the FtsZ filament the Lys-9 residue is required in some way to break the filament, possibly by interacting with residue Glu-276 of FtsZ (Fig. 10B). However, the observation that E7K suppresses K9A would argue that a lysine at position 7 can also accomplish this task.

Comparison of MinC/MinD with other inhibitors
Both SlmA and MinC/MinD are positioned in the cell and antagonize Z ring formation where they take up residence; MinC/MinD on the membrane with the highest concentration at the poles, whereas SlmA is bound to DNA around the origin of replication (3,4,11). In both cases binding to their respective surface potentiates their activity. SlmA bound to its DNA-binding site is activated to disassemble FtsZ filaments with little activity in the absence of DNA (4,27). The DNA binding results in a conformational change in SlmA that leads to increased affinity for the CCTP of FtsZ (15,28). Upon binding the CCTP, SlmA severs FtsZ filaments provided GTP has been hydrolyzed. The precise mechanism is not clear, but it requires a positive charge at position 190 at the midpoint of the long helix connecting the two globular domains of FtsZ (15). How SlmA bound to the CCTP takes advantage of this positive charge to elicit filament breakage is not clear.
Although MinC prevents FtsZ assembly in vitro, this activity is pH-and salt-sensitive (16), and as shown here separable from the mechanism employed by MinC/MinD. In vivo the activity of MinC is activated 40-fold by MinD (26), but this enhancement has not been observed in vitro. We have been unable to see any enhancement of MinC's activity by MinD in vitro despite some effort. Perhaps, MinD has to be bound to the membrane to effectively activate MinC. We know that the conformation of MinD is affected by membrane binding because it can only be induced to undergo ATP hydrolysis by MinE when membrane bound (29). Also, we have only observed interaction between MinC/MinD and the CCTP of FtsZ when MinD/MinC are on the membrane (13,30). Arumugam et al. (17) observed some enhancement in their reconstituted system but not to the degree observed in vivo. This may be related to the FtsZ bundles observed in their system that may deviate from FtsZ organization in the Z ring.
Interestingly, many inhibitors of FtsZ interact with the region around the proximal end of H10. Although this does not include SlmA, it does include MinC, SulA (31) as well as MciZ from Bacillus subtilis (32). SulA inhibits by sequestration (33,34), whereas MciZ inhibits by capping. One of the main differences between the two is that MciZ has a much higher affinity for FtsZ, about 0.15 M, in contrast to about 1 M for SulA. This is likely related to their physiological role. Although inhibition by SulA needs to be rapidly reversible (35), the MciZ inhibition occurs in the dying mother cell during sporulation and does not have to be reversible (36).

MinC/MinD mechanism Microscopy
To visualize the location of GFP-MinC in cells grown in liquid media, JS964 containing pMCW84 (P lac ::gfp-minC G10D /minD) and various derivatives with mutations were grown overnight at 37°C to A 600 ϳ0.02 at which point they were induced with 100 M IPTG. The cells were analyzed by fluorescence microscopy 0.5-2 h later. At the same time point, the cell morphology was assessed by phase-contrast microscopy. The toxicity of the MinC mutants was assessed by the degree of filamentation and lysis during growth. To determine whether minC mutations affected normal cell division, they were introduced into the mini-F plasmid pSEB12 containing the complete min operon. The resultant plasmids and the wildtype plasmid were transformed into JS964 (⌬min) and selected on LB-medium plates containing chloramphenicol. The morphology of cells was examined during exponential growth. The frequency of cells with polar septa was determined with at least 200 cells scored per experiment. The localization of GFP-MinC/MinD and its variants expressed with wildtype or mutants FtsZ were recorded according to the procedure described previously.

Site-directed and random mutagenesis
Various minC mutations such as K9A and G10D were introduced into pHJZ108, pSEB12, pET21␣C N , pSEB104CD, and pSEB104CDE using a QuikChange II site-directed mutagenesis kit according to the manufacturer's instruction (Agilent Technologies). The ftsZ mutations such as E276D and N280D were introduced into ftsZ on pBang112 using the same method. Random mutagenesis of the minC gene was performed by PCR with GeneMorph II (Stratagene). The amplification reaction mixture had a volume of 100 l and used 1 g of pSEB104CD plasmid template (115 ng of minC target sequence). The amplified product was digested with XbaI/XmaI and ligated into the same site of pMCW26. The resulting plasmid pool was introduced into JS964 by electroporation and selection for spectinomycin resistance on LB plates. The colonies were then patched in a grid pattern onto spectinomycin plates that contain either 10 or 1000 M IPTG. The colonies were screened as described in the main text.

Biosensor assays
The Blitz system (Fortebio) was used to examine the protein-protein interaction and was previously described in detail. MinC 115 -His 6 and various mutants in 1ϫ Pol buffer (25 mM HEPES-NaOH, pH 6.8, 50 mM KCl, 10 mM MgCl 2 ) were incubated with Ni-NTA biosensors pre-equilibrated with 1ϫ Pol buffer. Once MinC 115 or its variants were stably immobilized on the biosensor, FtsZ or FtsZ mutants in the same buffer were incubated with the MinC-coated biosensors for 2 min. Kinetics of MinC 115 and FtsZ bindings were monitored. Data analysis was carried out using GraphPad Prism (GraphPad Software) as described (23).

Gel filtration
MinC 115 -His 6 was incubated with FtsZ and FtsZ mutants in Pol buffer (50 mM HEPES-NaOH, pH 6.8, 50 mM KCl, 10 mM MgCl 2 ) for 5 min at 25°C. A 500-l volume at a final concentration of 3 g/ml for FtsZ and 1 g/ml for MinC 115 were thereafter applied to Superdex 200 coupled to AKA-fast protein liquid chromatography (GE Healthcare) with a flow rate of 0.4 ml/min. UV absorbance at 280 nm was measured.

FtsZ sedimentation
The sedimentation of FtsZ was measured as described previously (6). In addition to MalE-MinC, we also used His 6 -MinC and MinC 115 -His 6 to exclude the effects of the MalE tag. Their biochemical activity was identical to MalE-MinC. For sedimentation assays, FtsZ at the indicated concentrations was incubated in Pol buffer (50 mM Mes-NaOH, pH 6.5 (or 50 mM HEPES-NaOH at pH 6.8), 50 mM KCl, 10 mM MgCl 2 ) prior to mixing with MinC or SulA. The reactions were incubated at room temperature for 10 min with addition of GDP or GTP (2.5 mM) and centrifuged at 175,000 ϫ g for 15 min in TLA 100.2 rotor (Beckman). Pellets and supernatants were analyzed on 12.5% PAGE and stained with Coomassie Blue.

Western blot analysis
JS964 (⌬min) containing pSEB12 or its derivatives with various minC alleles were grown in LB medium with chloramphenicol to exponential phase. At A 600 0.3-0.6, cells from a 1-ml volume of culture was centrifuged and resuspended in 100 l of SDS-loading buffer, boiled for 5 min, and subjected to SDS-12.5% PAGE. Proteins were electrophoretically transferred to a nitrocellulose membrane, and MinC was detected by using a rabbit antiserum raised against MalE-MinC as the primary antibody and goat alkaline phosphatase-conjugated anti-rabbit immunoglobulin (Bio-Rad) as the secondary antibody.