High-resolution crystal structures of Escherichia coli FtsZ bound to GDP and GTP

Escherichia coli has served as the main model system for examining cell division in bacteria; however, the structure of the central cytokinesis protein FtsZ has not been determined to date. Here, high-resolution structures of E. coli FtsZ in GDP-bound and GTP-bound states are described.

Treadmilling is now established as the primary mechanism for FtsZ assembly dynamics.
Apart from treadmilling at steady state, FtsZ assembly in vitro exhibits two aspects of cooperative assembly: an unfavorable nucleation step and a critical concentration. However, this raises the question of how a single-stranded protofilament with only one type of bond can exhibit cooperative assembly. Several groups have suggested that this could be achieved if FtsZ had two conformations, one with a high and one with a low affinity for making longitudinal bonds (Michie & Lö we, 2006;Dajkovic & Lutkenhaus, 2006;Dajkovic et al., 2008;Huecas et al., 2008;Miraldi et al., 2008). To explain the cooperativity, the model proposes that FtsZ monomers are highly favored to be in the low-affinity conformation, whereas assembly into a protofilament switches the subunit to favor the high-affinity conformation (Michie & Lö we, 2006;Dajkovic & Lutkenhaus, 2006;Dajkovic et al., 2008;Huecas et al., 2008;Miraldi et al., 2008).
Prior to 2012, numerous FtsZ crystal structures were solved from different bacterial species in GTP-bound, GDP-bound or apo states (Lö we & Amos, 1998;Leung et al., 2000;Cordell et al., 2003;Oliva et al., 2004Oliva et al., , 2007. These structures all had the same basic structure: two globular subdomains connected by a long helix, H7. The N-terminal subdomain contains the binding site for GTP, and the C-terminal subdomain has, by analogy with tubulin protofilaments, key catalytic residues that contact the GTP in the subunit below to initiate GTP hydrolysis (Nogales et al., 1998;Erickson, 1998;Scheffers et al., 2002). Importantly, these structures were mostly of FtsZ monomers.
A major advance in our understanding of FtsZ structure and polymerization was achieved when two laboratories independently crystallized FtsZ from Staphylococcus aureus (SaFtsZ) and found that it crystallized in the form of long, straight protofilaments (Matsui et al., 2012;Elsen et al., 2012). The FtsZ in these protofilaments showed a striking conformational change relative to the previous monomeric forms, even the closely related Bacillus subtilis FtsZ (BsFtsZ) (Oliva et al., 2007;Matsui et al., 2012;Elsen et al., 2012). As noted, the globular domain of both FtsZ and tubulin is divided into two subdomains. The conformational change from BsFtsZ to SaFtsZ involved a large rotation (25-28 ) of the C-terminal subdomain relative to the N-terminal subdomain. Accompanying this rotation, the helix H7 shifted downwards by a turn (Matsui et al., 2012;Elsen et al., 2012). This new structure of SaFtsZ assembled in straight protofilaments with a large interface and was an excellent candidate for the high-affinity, protofilament-forming conformation, while the monomeric BsFtsZ represented the low-affinity conformation. The two conformations of FtsZ have been designated T (tense) for the high-affinity state and R (relaxed) for the low-affinity state (Matsui et al., 2014;Fujita et al., 2017). Alternatively, the two conformations have been designated open and closed, referring to the cleft between the two subdomains (Wagstaff et al., 2017). We will adopt the R and T nomenclature here.
Although wild-type SaFtsZ assembles preferentially into protofilaments with subunits in the T conformation, several mutants of SaFtsZ have been found assembled into protofilaments with subunits in the R conformation (Matsui et al., 2014;Wagstaff et al., 2017). Interestingly, Fujita and coworkers obtained a crystal of wild-type SaFtsZ containing subunits in both R and T states (Fujita et al., 2017). Apparently, there is only a small free-energy difference between the R and T conformations when SaFtsZ subunits are assembled into protofilaments under crystallization conditions. Strikingly, although E. coli has provided the most detailed understanding of the machinery and mechanism of bacterial cell division, the structure of the E. coli FtsZ (EcFtsZ) protein has remained elusive. A likely problem is the tendency of EcFtsZ to aggregate under non-assembly conditions (Mukherjee et al., 1993). Du and coworkers provided a potentially important advance when they explored several EcFtsZ mutants that had previously been identified as defective in assembly (Du et al., 2015;Li et al., 2013). Du and coworkers focused on one mutant, L178E, which they confirmed to be assembly-incompetent in the presence of GTP. They also found that EcFtsZ(L178E) lost its tendency to aggregate when bound to GDP, as determined by sizeexclusion chromatography (Du et al., 2015). This suggested that this mutant would be a good candidate for crystallography, and indeed they reported that 'FtsZ-L178E is selfinteracting at higher concentrations since it assembled into protofilaments when crystallized' (Du et al., 2015). However, these crystals were apparently not suitable for effective structure determination; the structure was never described nor was it deposited in the Protein Data Bank (Joe Lutkenhaus, personal communication). Thus, we decided to address this gap and repeated crystallographic screens in order to obtain better crystals. Here, we report the 1.35 and 1.40 Å resolution structures of E. coli FtsZ(L178E) bound to GDP and GTP, respectively.

Production and purification of E. coli FtsZ(L178E)
The pET-11 vector encoding the E. coli K-12 ftsZ gene was originally obtained from Bramhill and Thompson (Bramhill & Thompson, 1994). Truncation and site-directed mutants were generated from the vector and were also expressed in pET-11. The E. coli FtsZ truncation mutant used in this study encodes residues 10-316 of the protein. Mutants, ftsZ(L178E), ftsZ(L178E/M206E) and ftsZ(L178E/L272E), were also generated using QuikChange (Agilent) in the context of the truncated protein. E. coli BL21(DE3) cells were transformed with these expression vectors. Cells expressing each mutant protein were grown to an OD 600 of $1.2 and induced by adding 0.5 mM isopropyl -d-1-thiogalatopyranoside for 3 h at 37 C. After induction, the cells were centrifuged for 45 min at 15 000g. The resultant cell pellets were frozen at À80 C until use. Frozen cells were suspended in resuspension buffer consisting of 50 mM Tris-HCl pH 7.9, 100 mM NaCl, 2 mM phenylmethylsulfonyl fluoride, and 0.75 mg ml À1 lysozyme was added. After 30 min on ice the cells were refrozen, thawed and lysed by sonication. The cells were sonicated eight times for 30 s on ice. The lysate was centrifuged at 80 000g for 30 min and the supernatant was recovered. The soluble bacterial proteins were first precipitated with 50% saturated ammonium sulfate. The resuspended proteins were then dialyzed against a buffer composed of 50 mM Tris-HCl pH 7.9, 50 mM KCl, 10%(v/v) glycerol, 1 mM EDTA. The proteins were purified by chromatography on a Resource Q anionexchange column (GE Healthcare) with a linear gradient from 50 to 500 mM KCl. Fractions containing protein were identified by SDS-PAGE analysis, and those containing protein were pooled and the protein was concentrated using a Microcon 30 kDa cutoff concentrator (Millipore). The FtsZ protein concentrations were determined as described previously using bicinchoninic acid colorimetric assays (Lu et al., 1998).

Crystallization of E. coli FtsZ(L178E)-GDP and E. coli FtsZ(L178E)-GTP complexes
Full-length, truncated and truncated mutant FtsZ proteins were used in crystallization trials. The EcFtsZ(L178E) protein truncated to the globular domain from residues 10 to 316 produced well diffracting crystals and thus was used for structural studies [and will be referred to here as EcFtsZ(L178E)]. To obtain crystals of EcFtsZ(L178E) with GDP, the protein was concentrated to 37 mg ml À1 and GDP was added to the solution, resulting in a final concentration of 1 mM. The mixture was centrifuged for 3 s at 20 000g and then used in hanging-drop vapor-diffusion crystallization screens. The screens employed were Wizard Classic 1-4 (Rigaku), Wizard Cryo 1 and Cryo 2 (Rigaku), PEGRx 1 and PEGRx 2 (Hampton Research), Crystal Screen (Hampton Research), SaltRx 1 and SaltRx 2 (Hampton Research). The screens were all performed at room temperature using 24-well plates. In the screens, 1 ml protein solution was mixed with 1 ml cystallization reagent and placed over the reservoir. The most promising crystals were obtained with Wizard Classic 3 condition 35. This condition was further optimized and the best crystals were obtained by combining 5 ml EcFtsZ(L178E)-GDP complex with 5 ml crystallization solution consisting of 50 mM sodium cacodylate pH 6.5, 11%(w/v) PEG 8000, 150 mM calcium acetate, 20%(v/v) glycerol and hanging the drop over a reservoir containing 750 ml crystallization solution (Table 1). The crystals grew overnight and reached a maximum size of 0.3 Â 0.5 Â 0.8 mm over 1-1.5 weeks. The EcFtsZ(L178E)-GTP crystals were obtained similarly using protein at 30 mg ml À1 with a final concentration of 1 mM GTP (Table 1). These crystals were isomorphous to the EcFtsZ(L678E)-GDP crystals. The EcFtsZ(L178E)-GTP complex was screened (at room temperature) using all of the commercial screens used for the EcFtsZ(L178E)-GDP complex. The best crystals of the EcFtsZ(L178E)-GTP complex were also obtained with Wizard Classic 3 condition 35. Optimal crystals were obtained by mixing the complex in a 1:1 ratio with a reservoir consisting of 13%(w/v) PEG 8000, 50 mM sodium cacodylate pH 6.5, 150 mM calcium acetate, 20%(v/v) glycerol and hanging the drop over a reservoir containing 750 ml crystallization reagent ( Table 1). The crystals grew overnight and reached similar sizes to the EcFtsZ(L178E)-GDP crystals in 1-2 weeks. Both crystals were cryopreserved directly from the crystallization solution, which acted as a cryosolvent. Data were collected on beamline 8.3.1 at the Advanced Light Source (ALS), USA. The crystals were looped and placed directly in the liquidnitrogen stream in the hutch and data were collected.
2.3. Crystal structure determination of the E. coli FtsZ(L178E)-GDP and E. coli FtsZ(L178E)-GTP complexes X-ray intensity diffraction data for the EcFtsZ(L178E)-GDP and EcFtsZ(L178E)-GTP complexes were collected on beamline 8.3.1 at the ALS. The data were processed using MOSFLM and scaled with SCALA (Leslie, 2006;Potterton et al., 2003; Table 2). Both the EcFtsZ(L178E)-GDP and EcFtsZ(L178E)-GTP crystals belonged to the monoclinic space group P2 1 . Molecular replacement (MR) using the Pseudomonas aeruginosa FtsZ structure (PDB entry 2vaw; Oliva et al., 2007) without the GDP and waters as a model was successful in solving the EcFtsZ(L178E)-GDP structure using MOLREP and Phaser (Vagin & Teplyakov, 2010;McCoy et al., 2007). Both produced clear solutions [with a log likelihood gain (LLG) of 309.67 in Phaser]. There is one EcFtsZ(L178E) protein molecule in the crystallographic asymmetric unit. After an initial round of refinement in Phenix (Liebschner et al., 2019) the residues that were not identical between P. aeruginosa and E. coli were replaced by their E. coli counterparts. This structure was then subjected to multiple iterative cycles of rebuilding in O and refinement in Phenix   Tables 2 and 3 for data-collection and refinement statistics.

E. coli FtsZ(L178E)-GDP and E. coli FtsZ(L178E)-GTP structure determination
The EcFtsZ(L178E) protein, which previous studies had indicated to be a monomer in the presence of GDP or GTP (Du et al., 2015), produced well diffracting crystals in the presence of GDP and, when optimized, permitted us to collect data to 1.35 Å resolution. As the P. aeruginosa FtsZ protein is the most similar in amino-acid sequence to EcFtsZ (overall identity of 67% over the conserved core of our truncated protein), we used the P. aeruginosa FtsZ-GDP structure (PDB entry 2vaw; Oliva et al., 2007) as a search model for molecular replacement. One clear solution was obtained, consistent with the presence of one EcFtsZ(L178E) molecule in the crystallographic asymmetric unit. Clear electron density was observed in the structure for GDP (Figs. 1 and 2a). After the addition of the GDP and water molecules and further refinement, the structure converged to final R cryst and R free values of 18.6% and 20.4%, respectively, to 1.35 Å resolution.
The GTPase activity of FtsZ is largely dependent on polymerization, as the GTP-binding pocket at the 'top' of a subunit must make contact with the catalytic T7 loop residues at the Overall structure of the EcFtsZ(L178E)-GDP complex. The protein is shown as a ribbon diagram with the N-terminal subdomain (residues 9-178) colored yellow, the connecting H7 helix colored red and the C-terminal subdomain (residues 201-216) colored cyan. S and H indicate strands and helices, respectively. The bound GDP is shown as sticks. Also shown as sticks is residue Glu178, the site of the mutation. The structure is labeled according to the common nomenclature for the tubulin/FtsZ family of GTPases (Nogales et al., 1998). All ribbon diagrams were made using PyMOL (DeLano, 2002).  bottom of an adjacent subunit in the protofilament to hydrolyze the bound GTP (Erickson, 1998;Scheffers et al., 2002). Because EcFtsZ(L178E) has been shown to be deficient in GTPase activity owing to its inability to form filaments (Du et al., 2015), we considered that it was likely that it would crystallize with GTP. Crystals were successfully obtained of the EcFtsZ(L178E)-GTP complex that were isomorphous with the EcFtsZ(L178E)-GDP crystals. The EcFtsZ(L178)-GTP structure was solved and refined to final R cryst and R free values of 18.5% and 20.5%, respectively, to 1.40 Å resolution. Notably, the electron density for the -phosphate moiety of the GTP in the EcFtsZ(L178E)-GTP structure was weak (Fig. 2b); the average B factor for the atoms of the -phosphate is 63.1 Å 2 compared with 19.2 Å 2 for the remaining atoms of the GTP. These data suggest that this mutant may retain residual GTPase activity or that the -phosphate is disordered in the structure.

E. coli FtsZ(L178E) crystallizes in the form of straight protofilaments
In spite of evidence that the L178E mutation in EcFtsZ inhibits assembly, the crystals showed that EcFtsZ(L178E) assembled into straight protofilaments, as noted in the previous study of Du et al. (2015). However, comparison of the EcFtsZ(L178E) filament with the SaFtsZ filament reveals that the EcFtsZ filament is much looser (see below for the buried surface areas). While the residues in the EcFtsZ(L178E) loops from residues 63 to 70 and 168 to 178 make somewhat similar interactions with neighboring subunits to those observed in the SaFtsZ filament, they are more surface-exposed in the EcFtsZ(L178E) crystal filament and their relatively high B factors (the average B factors for residues 63-70 and residues 168-178 are 56.9 and 45.5 Å 2 , respectively, compared with the overall average protein B factor of 23.9 Å 2 ) indicate they are somewhat flexible. In addition, the electron density for these residues is sparse. Presumably, as the residue corresponding to Leu178 in SaFtsZ is wedged into a hydrophobic pocket of an adjacent subunit in the SaFtsZ filament, the similar burial of the Glu178 side chain into the hydrophobic pocket in the EcFtsZ(L178E) filament is predicted to be unfavorable. As a result, the EcFtsZ(L178E) protofilaments have a more open subunit interface. In this filament, Leu68 from the loop connecting S3 and H3 is inserted into a hydrophobic pocket composed of Ala11, the C atom of Asp96, Ile200 and Phe210 of the adjacent subunit (Fig. 3b). Leu68 has an elevated average B factor (48.6 Å 2 ) compared with the overall B factor of 23.9 Å 2 for the protein, suggesting that the contact might be weak. In the EcFtsZ(L178E) loop connecting H6 and H7, which contains the L178E mutation, Arg174 makes electrostatic interactions with Glu233 and stacks with the Arg271 side chain from the adjacent subunit, and Leu172 from the loop makes weak van der Waals interactions with the side chain of Leu272 from the neighboring subunit (Fig. 3b). Next to this loop, Phe137 also makes contacts with Leu272 and also Met206 and Phe275 from the adjacent subunit (Fig. 3b). Other cross-contacts include electrostatic interactions between Lys141 and Arg142 and the neighboring subunit residues Asp209 and Asp212, respectively. Consistent with the finding that the EcFtsZ(L178E) filament has fewer contacts between subunits, the buried surface area (BSA) between subunits in the interface of the EcFtsZ(L178E) protofilament is only 720 Å 2 , compared with the BSA of $1200 Å 2 observed in the snug interface between subunits in the SaFtsZ filament (Krissinel & Henrick, 2007;Matsui et al., 2012;Fig. 3a). This is similar to the BSA of 740-810 Å 2 that Matsui and coworkers observed for SaFtsZ mutants that polymerized in the R form (Matsui et al., 2014). The helices, strands and loops are colored cyan, red and magenta, respectively. The blue mesh corresponds to the F o À F c OMIT map (in which the GDP was removed) contoured at 3.5 around the GDP. (b) Close-up of the GTP in the EcFtsZ(L178E)-GTP complex with the protein secondary-structural elements colored as in Fig. 2(a). The F o À F c map is shown as a blue mesh and is contoured at 2.9.
Further supporting the notion that the EcFtsZ(L178E) filament is not the form found in the Z-ring, the longitudinal spacing of subunits in the crystallized protofilament is 47.0 Å , which is significantly larger than the value of 42-43 Å measured for EcFtsZ protofilaments by negative-stain EM (Erickson et al., 1996), the value of 44.0 Å observed in the SaFtZ protofilaments in crystals (Matsui et al., 2012) and the value of $40 Å measured by cryo-EM under native-like conditions (Szwedziak et al., 2014). Also in this filament, the T7 loops are not properly positioned to interact with the GTP of the subunit below (Fig. 3a). However, we note that this filament could represent an intermediate in the assembly or disassembly of functional filaments.
3.3. E. coli FtsZ(L178E) structures are in the R conformation DALI structural homology searches revealed that the FtsZ structures with the strongest similarity to EcFtsZ(L178E) are the P. aeruginosa FtsZ (PDB entry 1ofu; Cordell et al., 2003) and S. aureus FtsZ (PDB entry 3wgl; Matsui et al., 2014) structures. PDB entry 1ofu is the structure of P. aeruginosa FtsZ bound to SulA (Cordell et al., 2003), and 293 of the C atoms in the FtsZ subunit of this structure could be superimposed onto the corresponding atoms in the EcFtsZ(L178E) structure with an r.m.s.d. of 1.1 Å (Fig. 4a). The structure with PDB code 3wgl contains an S. aureus mutant in which the SGEVN sequence of the T7 loop was replaced by a shorter GAN sequence (Matsui et al., 2014). This FtsZ mutant was crystallized with GDP to produce the structure deposited as PDB entry 3wgl. Despite the fact that S. aureus FtsZ shows 14% less sequence identity to the E. coli protein compared with P. aeruginosa FtsZ, PDB entry 3wgl shows essentially the same degree of structural homology; 289 C atoms of the 3wgl S. aureus FtsZ structure could be superimposed onto the corresponding C atoms of the EcFtsZ(L178E) structure with an r.m.s.d. of 1.1 Å . Importantly, both PDB entries 1ofu and 3wgl are in the R conformation, supporting the adoption of the R state by EcFtsZ(L178E) in the crystal. Indeed, superimposition of the EcFtsZ(L178E) structure onto the T-state conformation of FtsZ (PDB entry 3voa; Matsui et al., 2012) resulted in a large r.m.s.d. of 2.3 Å for 269 corresponding C atoms (Fig. 4b).

C-terminal subdomain structural differences revealed in filament compared with nonfilament FtsZ
Our analyses show that the EcFtsZ(L178E) structure adopts the R state and as such has a closed cleft between its N-terminal and C-terminal subdomains (Wagstaff et al., 2017). When we compared the structure of the N-terminal subdomain of the EcFtsZ(L178E) structure with the N-terminal subdomain of both R-state and T-state FtsZ structures, we found they are all very similar; they all overlay with r.m.s.d.s of <1 Å (Fig. 4c). Interestingly, although the C-terminal subdomain of EcFtsZ(L178E) is structurally similar to those of other R-state structures (r.m.s.d.s of between 0.8 and 1.2 Å ; Fig. 4d), T-state C-terminal subdomain structures showed significant structural differences from the EcFtsZ(L178E) C-terminal subdomain and thus other R-state C-terminal subdomains, with r.m.s.d.s of >2.1 Å for these overlays (Fig. 4e). A main difference in these C-terminal subdomain structures appears to be in the region encompassing the T7 loop and helix H8. The relevance of this structural difference The EcFtsZ(L178E) filament. (a) Side-by-side comparison (after aligning the middle subunit) of the EcFtsZ(L178E) filament (red) generated by its crystal structure packing and the filament observed in the T-state SaFtsZ structure (PDB entry 3voa; yellow). The position of the T7 loop in each subunit is labeled, showing that this key catalytic loop is not properly positioned in the EcFtsZ filament (next to the GTP) as in the T-state filament. (b) Close-up view of the intersubunit contacts in the EcFtsZ(L178E) filament. between the C-terminal subdomains of the T and R states of FtsZ, if any, remains to be determined.

Conclusions
Studies in the last several decades using E. coli have revealed the machinery of the cell-division divisome and its association with the FtsZ Z-ring. A plethora of regulatory proteins that interact with E. coli FtsZ to affect its ability to form a Z-ring and its localization in the cell have also been discovered (Ortiz et al., 2016). However, to date a structure of EcFtsZ has not been reported. Here, we describe high-resolution structures of the EcFtsZ(L178E) protein in complex with GDP and GTP. These structures also allow the interpretation of multiple mutations that have been constructed and tested for their effects on cell division in this protein (Lu et al., 2001;Stricker & Erickson, 2003;Addinall et al., 2005;Redick et al., 2005;Osawa et al., 2008;Haeusser et al., 2015;Má rquez et al., 2017) and can serve as the foundation for the design of future celldivision experiments in E. coli to assess the role of specific residues in the function of FtsZ as well as its interactions with regulatory proteins.  The EcFtsZ(L178E) structure adopts the R state. (a) Superimposition of the EcFtsZ(L178E) structure (red) onto the R-state S. aureus FtsZ structure (PDB entry 1ofu; cyan). (b) Overlay of the EcFtsZ(L178E) structure (red) onto the T-state S. aureus FtsZ structure (PDB entry 3voa; green). (c) Overlay of the EcFtsZ(L178E) N-terminal subdomain (red) onto the N-terminal subdomain of the T-state structure (PDB entry 3voa; yellow), showing that the N-terminal subdomains adopt the same conformations in R and T states. (d) Superimposition of the C-terminal subdomain of EcFtsZ(L178E) (red) onto the C-terminal subdomain of the R-state FtsZ structures (PDB entry 3wgl, magenta; PDB entry 1ofu, cyan), showing they have the same general fold (r.m.s.d.s of 1.1 and 0.8 Å , respectively). (e) Superimposition of the EcFtsZ(L178E) C-terminal subdomain (red) onto the C-terminal subdomains of the T-state FtsZ structures (PDB entry 4dxd, cyan; PDB entry 3voa, magenta), showing that the EcFtsZ (R-state) C-terminal subdomain has structural differences compared with T-state C-terminal subdomains (both r.m.s.d.s are >2 Å ).