Introduction

FtsZ is highly conserved among bacteria and plays an essential role in bacterial cell division1,2. During cell division, the bound nucleotide in monomer FtsZ exchanges from GDP to GTP, and then FtsZ bound to GTP assembles into a straight filament via head-to-tail associations. The FtsZ filament serves as a scaffold at the cell division position for the partners and promotes further cell division processes. FtsZ is a GTPase; thus, the upper subunit of FtsZ hydrolyzes GTP accommodated in the lower subunit. Consequently, the penultimate subunit with GDP bound in the tip of the filament disassembles accompanied by a conformational change in the subunit from the tense to relaxed conformations, in contrast, the monomer FtsZ bound with GTP attaches at the other end of the filament, promoting that the filament treadmills and then the multiple filament patches move around the cell division site3,4,5,6,7,8,9,10,11,12.

Although the above hypothesis has been proposed, details regarding the assembly cycle machinery remain unclear, even though the crystal structures of FtsZ derived from several bacterial species have been determined5,11,13,14,15,16,17,18,20. This is because no conformational change between the tense and relaxed conformations has been observed, except in staphylococcal FtsZs5,8,15,16,19,20. A principal component analysis revealed that one of the eigenvectors described the difference in conformation between the polymerized and unpolymerized conditions21. However, all conformations within the subspace of the polymerized condition have been derived from Staphylococcus alone. Therefore, it is also unclear whether the proposed conformational changes are conserved among bacteria or associated with differences between bacteria species.

FtsZ is composed of two subdomains linked by the T6 loop, H7 helix, the T7 loop, and H8 helix. The N-terminal subdomain (NTD, residues 12–173) accommodates the nucleotide, and the C-terminal subdomain acts as a GTPase-activating domain (GAD, residues 223–310). In the conformational change from the relaxed to tense conformation observed in Staphylococcus aureus FtsZ (SaFtsZ) and its mutant8,15,16, the H7 helix and T7 loop are down-shifted approximately one helical pitch, accompanied by a reorientation of the GAD (Supplementary Fig. S1). Consequently, the area of the interface between the upper and lower subunits made by the tense conformation in the filament is increased by 40% compared with the interface formed by the relaxed conformation, and the T7 loop is located deep within the lower subunit near the nucleotide. Recent cryoEM analyses revealed the three-dimensional reconstruction of the protofilaments derived from Escherichia coli FtsZ (EcFtsZ), Mycobacterium tuberculosis FtsZ (MtbFtsZ), and Klebsiella pneumoniae (KpFtsZ)8,21,22, and the tense conformation of SaFtsZ fit well into a density map of the protofilaments rather than the corresponding relaxed monomer of EcFtsZ, MtbFtsZ, or KpFtsZ. These CryoEM reconstructions predicted that the conformational change induces an assembly switch upon polymerization. However, the lower resolutions of EcFtsZ, MtbFtsZ, and KpFtsZ observed using CryoEM suggested a less strict alignment or mixture of both the tense and relaxed conformations in the straight filament as predicted in previous studies11,23. FtsZ may exhibit flexible conformations for the conformational change triggered by bound nucleotide and/or polymerization; thus, the difference in flexibility of the EcFtsZ, MtbFtsZ, and KpFtsZ protomers between the relaxed and tense conformations probably caused them to not form a filament strictly. Furthermore, the flexibility may also prevent the FtsZs from crystalizing in the tense conformation.

A molecular dynamics (MD) simulation is a physical-based theoretical method for capturing the movement of each atom in a protein or other molecular system. MD simulations can help researchers understand the mechanisms of a wide variety of biological phenomena, such as structural dynamics, perturbations, and processes of biomolecular systems. Due to rapid developments in computing power, MD simulations have become much more powerful and accessible in recent years. MD simulations can now be used to investigate the atomic dynamics of protein folding and misfolding24, elucidate the mechanisms of membrane ion channel functions25, and accurately predict the recognition and association of a protein and ligand26. Three-dimensional protein structures can be obtained with atomic-level accuracy using artificial intelligence (AI) systems (i.e., AlphaFold2 and RoseTTaFold)27,28. Although AI can be used to accurately predict a protein’s structure based on crystal structure database information, it is still difficult to reproduce the structure of protein–ligand complexes, as AlphaFold2 typically generates a model in only one functional state, which is usually the inactive state29,31,31.

The present study assessed the differences in flexibility between bacterial FtsZs in several conformational states using MD simulations of the monomer FtsZ within the tense and relaxed conformations bound to GTP and GDP, respectively. Although both conformations of SaFtsZ were determined by X-ray crystallography16, the tense conformation of Bacillus subtilis FtsZ (BsFtsZ) has not been observed. A tense model of BsFtsZ was generated from SaFtsZ as a template using SWISS-MODEL32, and then all FtsZ structures were equilibrated in aqueous solution, and their stability and fluctuation were estimated.

Results

Evaluation of equilibration of the system

All systems were equilibrated during 200-ns MD simulations (Fig. 1). In the relaxed conformation of SaFtsZ, the root mean square deviation (RMSD) values of SaFtsZ bound to both GTP and GDP were stable, at 1.5 ± 0.1 Å and 1.6 ± 0.1 Å after 50 ns, respectively (Fig. 1a). These systems equilibrated rapidly. Rapid equilibration was also observed in the relaxed conformation of BsFtsZ bound to GTP and GDP, with RMSD values of 1.2 ± 0.1 Å and 1.3 ± 0.1 Å after 50 ns, respectively (Fig. 1b). In contrast to the relaxed conformation, the tense conformation of SaFtsZ showed a different equilibration pattern in binding of GTP and GDP (Fig. 1c). For the SaFtsZ tense conformation bound to GTP (tense SaFtsZ-GTP), the RMSD value reached ~ 1.5 Å at 20 ns and remained stable for the remainder of the simulation. The RMSD value of the tense conformation bound to GDP (tense SaFtsZ-GDP) gradually reached a plateau, with an average value during the equilibration of 2.7 Å. The tense conformation of BsFtsZ exhibited a similar pattern (Fig. 1d). Tense BsFtsZ showed different trajectories in binding GTP versus GDP, with average RMSD values for these two systems of 1.9 ± 0.1 Å over the period of 10–200 ns and 3.1 ± 0.1 Å over the period of 70–200 ns, respectively. These results suggest that the relaxed conformation in the crystal structure was already thermodynamically stable, regardless of bacterial species and bound nucleotide. In contrast, the tense conformations of both FtsZs may form a higher energetic state. Furthermore, the tense conformation of FtsZ-GDPs derived from both species exhibited a change in the stable conformation of approximately 3 Å-difference relative to the initial state during the 200-ns MD simulations.

Figure 1
figure 1

RMSD trajectories during 200-ns MD simulations. (a,b) MD simulations of relaxed molecules derived from (a) SaFtsZ (5H5G molecule B16) and (b) BsFtsZ (2VXY14)[], respectively. Relaxed conformations bound to GTP and GDP are drawn as orange and blue lines, respectively. (c,d) MD simulations of tense molecules derived from (c) SaFtsZ (5H5G molecule A16) and (d) SWISS-MODEL of BsFtsZ templated from 5H5G molecule A16. Tense conformations bound to GTP and GDP are depicted as red and green lines, respectively.

Full size image

Flexibility of the nucleotide-bound formations

The average structure during the last 50 ns of the equilibrated state was generated, and the flexibility of each residue during that period was evaluated based on the root mean square fluctuation (RMSF) value (Supplementary Figs. S2–S3). Several specific regions fluctuated in relaxed SaFtsZ bound to either GTP or GDP (Supplementary Fig. S2a). Higher fluctuations occurred in the H3 helix (Ser85–Asp97) in the NTD and H8 helix (Asp209–Ser219) in the linker region of relaxed SaFtsZ-GDP. In contrast, T6 (Asp174–Ala182) and T7 (Ser204–Leu207) in the linker region, and T10 (Gly267–Leu270) and T11 (Asn299–Glu305) in the GAD exhibited greater flexibility when binding GTP (Supplementary Fig. S2a,b). Although the differences in flexibility between bound nucleotides in tense SaFtsZs were lower than those in relaxed SaFtsZs (Supplementary Fig. S1a,c), several residues in tense SaFtsZ also exhibited different dynamics depending on the bound nucleotide. The RMSF value of Ser204 at the T7 loop (Leu201–Leu209) in tense SaFtsZ-GTP was 3.8 times higher than that in SaFtsZ-GDP. In contrast, the RMSF value of Asn220 at the T8 loop (Asn220–Ser223) in tense SaFtsZ-GTP was 3.7 times lower than that of SaFtsZ-GDP (Supplementary Fig. S2c,d).

Loops T6 and T7, which are the flanks of the H7 helix, showed higher deviations in relaxed and tense SaFtsZ-GTP compared with the GDP-bound states. Ser204 of the T7 loop in tense SaFtsZ-GTP exhibited a higher RMSF value of 3.3 Å compared with the relaxed state value of 1.4 Å. The RMSF values of Gly33 at H1 helix (Gly20-HIS33) and Asn35 at the T1 loop (Gly34–Glu39) in tense SaFtsZ-GTP were > 2.0 Å; Ser269 in the T10 loop was only observed in the relaxed conformation and exhibited an RMSF value of 2.7 Å.

A comparison of the relaxed conformations of SaFtsZ and BsFtsZ indicated that relaxed BsFtsZ bound to either nucleotide was more stable than those of SaFtsZ (Supplementary Fig. S2, S3). Only Lys175 in the T6 loop markedly fluctuated in relaxed BsFtsZ bound to GDP and GTP, with RMSF values of 2.2 Å and 1.5 Å, respectively. However, the RMSF value of Lys175 in the T6 loop of relaxed SaFtsZ-GDP was lower than that of the molecule bound to GTP (1.5 versus 2.0 Å, respectively). Furthermore, relaxed SaFtsZ exhibited several fluctuating loops, including the T6 loop, and these loops showed higher deviations in the GTP-bound state, whereas no similar behavior was exhibited by relaxed BsFtsZ-GTP.

A comparison of the tense conformations of SaFtsZ and BsFtsZ showed that BsFtsZ formed a more stable conformation than SaFtsZ under the same bound nucleotide condition. The T1 and T7 loops in tense SaFtsZ-GTP were flexible, whereas no marked fluctuation in those loops was observed in tense BsFtsZ-GTP. However, the RMSF value of Lys175 of the T6 loop in tense BsFtsZ-GTP, at 2.6 Å, was similar to the values of that in the tense SaFtsZ bound to GTP and GDP, at 2.8 Å and 2.4 Å, respectively. A significantly large fluctuation at Glu139 of the T5 loop (Arg133–Arg141) was found only in tense BsFtsZ bound to GDP and GTP, with RMSF values of 2.2 Å and 3.5 Å, respectively.

Structural comparisons of MD simulation models with reference structures

To better understand the conformational differences, MD-simulated structures were compared with reference structures. First, the structural features of the tense and relaxed conformations against the crystal structures deposited previously in the PDB database5,14,16,17,33 were identified based on inter-subdomain distances using the relative distance between the centers in each subdomain (Supplementary Table S2). The inter-subdomain distances between the centers were 28.4 ± 0.1 Å and 26.2 ± 0.5 Å in the tense and relaxed conformations, respectively. Figure 2 shows the trajectories in each conformation of SaFtsZ and BsFtsZ. The relaxed conformations of both bacterial FtsZs bound to GDP and GTP remained at a stable distance, as observed in the crystal structures of the relaxed FtsZs, at 26.0 ± 0.4 Å and 26.4 ± 0.3 Å in SaFtsZ and at 26.2 ± 0.3 Å and 26.3 ± 0.3 Å in BsFtsZ, respectively. The inter-subdomain distance of tense BsFtsZ, regardless of bound nucleotide, exhibited an average value of approximately 28 Å. In contrast to tense BsFtsZ, although tense SaFtsZ-GTP exhibited a stable average distance similar to that calculated from the tense conformation in the crystal structure, tense SaFtsZ-GDP exhibited a shift in the distance, which was similar to the average value of the relaxed conformation, at 26.6 ± 0.3 Å for the last 50 ns. These data suggested that both FtsZs assumed a different conformation accompanied by a difference in stability, especially in the tense conformation bound to GDP.

Figure 2
figure 2

Trajectories of 200-ns MD simulation inter-subdomain reorientation. (a,b) Inter-subdomain distances of (a) relaxed and (b) tense conformations of SaFtsZ. (c,d) Inter-subdomain distances of (c) relaxed and (d) tense conformations of BsFtsZ. Inter-subdomain distances derived from tense (PDB 5H5G mol A16) and relaxed (PDB 5H5G mol B16) forms are shown as dotted lines, with values of 28.4 and 26.2 Å, respectively. (e,f) Inter-subdomain angles of (e) relaxed and (f) tense conformations of SaFtsZ. (g,h) Inter-subdomain angles of (g) relaxed and (h) tense conformations of BsFtsZ. Relaxed conformations bound to GTP and GDP are drawn as orange and blue lines, respectively. Tense conformations bound to GTP and GDP are depicted as red and green lines, respectively.

Full size image

In contrast to the inter-subdomain distances, no rotational axis in all average structures against the corresponding reference structure was detected using DynDom34, suggesting that the relative orientation between subdomains as seen in the difference between relaxed and tense conformations in the previous investigation5,8,16,19,20 was not observed. Furthermore, the conformations of NTD and GAD in the average structures during the last 50-ns simulations were superimposed well with those in the crystal structures, respectively (Fig. 3 and Supplementary Fig. S4), suggesting that there was no conformational change in each subdomain. However, the whole conformations of all simulated structures seemed to be changed (Supplementary movies S1–S8). Then, the vector angles between the GAD and NTD of the MD simulations were also evaluated against the reference conformation (Fig. 2e–h, Supplementary Tables S3–S4). In the 200-ns MD simulation, relaxed SaFtsZ-GTP and tense SaFtsZ bound to both nucleotides exhibited lower relative vector angles compared with the initial conformations, at 2.8 ± 1.1°, 2.7 ± 1.6°, and 3.6 ± 1.6°, respectively (Supplementary Table S3). In contrast to those conformations, the relative vector angle of relaxed SaFtsZ-GDP compared with the initial model (PDB 5H5G, mol B16) was 6.8 ± 1.8°. However, the vector between the subdomains in relaxed SaFtsZ-GDP formed angles with the subdomains of BsFtsZ (PDB 2VXY14) and Methanocaldococcus jannaschii FtsZ (MjFtsZ) (PDB 1W5933) of 2.0 ± 1.0° and 2.1 ± 1.0°, respectively (Supplementary Table S3), suggesting that in the MD simulation, relaxed SaFtsZ-GDP equilibrated to assume a similar conformation to the crystal structures of BsFtsZ and MjFtsZ.

Figure 3
figure 3

Average structure during last 50 ns of MD simulations compared with the reference structure. Average structure of (a) relaxed SaFtsZ bound to GDP compared with the crystal structure (PDB 2VXY14, BsFtsZ). (b) Relaxed SaFtsZ bound to GTP compared with the crystal structure (PDB 5H5G mol B16). Tense SaFtsZ bound to (c) GDP and (d) GTP compared with the crystal structure (PDB 5H5G mol A16). Relaxed BsFtsZ bound to (e) GDP and (f) GTP compared with the crystal structure (PDB 2VXY14). Tense BsFtsZ bound to (g) GDP and (h) GTP compared with the SWISS-MODEL structure. The average structures of MD simulations started from the relaxed and tense forms, and the reference structures are depicted as blue, orange, and gray cartoons, respectively. Conformational changes from the crystal or model structure to the average structures indicated in panels (a) through (h) are illustrated on the initial structure (gray color) as vectors. Coordinate differences > 2.0 Å are drawn as red arrows.

Full size image

The relaxed BsFtsZ-GDP, relaxed BsFtsZ-GTP, and tense BsFtsZ-GTP exhibited lower vector angles compared with the references, at 2.0 ± 1.0°, 2.4 ± 1.2°, and 2.1 ± 1.2°, respectively (Supplementary Table S4). Only tense BsFtsZ-GDP exhibited a vector angle > 5.7 ± 2.1° during the entire MD simulation. Thus, only tense BsFtsZ-GDP assumed a conformation with a relative orientation much different from that of previous structures in the PDB.

The conformational changes relative to the reference structures are summarized in Fig. 3 and Supplementary Fig. S5. In the average conformation during the last 50 ns of the MD simulation, the conformation of relaxed SaFtsZ-GDP was similar to that of BsFtsZ (PDB 2VXY14), except for slight conformational differences in the T10 and T11 loops in the GAD and in the T7 loop (Fig. 3a). No significant conformational changes were observed between relaxed SaFtsZ-GTP and the reference structure (SaFtsZ, PDB 5H5G mol A16), even though there was a small difference in the rotational angle of 2.8° between both conformations over 2-Å conformational shifts at positions far from the NTD (Fig. 3b and Supplementary Table S3).

Figure 4 shows a map of the kernel density estimation, which is the trajectory of the RMSD differences for each SaFtsZ conformation at each 1-ns step against the relaxed and tense forms. The MD simulation revealed that the relaxed SaFtsZ-GDP exhibited a 1–1.5 Å difference with approximately 0.5-Å deviations from the reference of the relaxed conformation (Fig. 4a), suggesting that the conformation of relaxed SaFtsZ-GDP fluctuates in the relaxed conformation by over 2-Å difference compared with the tense conformation. The relaxed SaFtsZ-GTP converged with the static conformation with a higher kernel density > 10 during the 200-ns MD simulation (Fig. 4b). These results suggest that the relaxed conformation of SaFtsZ, regardless of bound nucleotide at the equilibrium state, is similar to the relaxed conformation previously observed. In the tense conformation, SaFtsZ-GDP exhibited a conformational shift in the GAD (Fig. 3c), but the deviations with respect to the references of the relaxed and the tense conformations were almost all distributed within 1 Å and converged at a density value of 6.4 (Fig. 4c). The H9 and H10 helices of the GAD in the tense SaFtsZ-GTP rotated and moved closer to the H7 helix and T7 loop (Fig. 3d), indicating that the tense SaFtsZ-GTP shifted to a different conformation not previously deposited in the PDB. The conformational difference could also be explained by the finding that the trajectory was distributed far from the tense and the relaxed conformation matrices (Fig. 4d).

Figure 4
figure 4

Differences in RMSD values of MD simulations with the crystal structure. Trajectories of (a) SaFtsZ relaxed-GDP compared with the relaxed-form (PDB 2VXY14) and tense-form (PDB 5H5G mol A16). (b) SaFtsZ relaxed-GTP compared with relaxed-form (PDB 5H5G mol B16) and tense-form (PDB 5H5G mol A16). (c) SaFtsZ tense-GDP compared with relaxed-form (PDB 5H5G mol B16) and tense-form (PDB 5H5G mol A16). (d) SaFtsZ tense-GTP compared with relaxed-form (PDB 5H5G mol B16) and tense-form (PDB 5H5G mol A16). (e) BsFtsZ relaxed-GDP compared with relaxed-form (PDB 2VXY14) and tense-form (PDB 5H5G mol A16). (f) BsFtsZ relaxed-GTP compared with relaxed-form (PDB 2VXY) and tense-form (PDB 5H5G mol A16). (g) BsFtsZ tense-GDP compared with relaxed-form (PDB 2VXY14) and tense-form (PDB 5H5G mol A16). (h) BsFtsZ tense-GDP compared with relaxed-form (PDB 2VXY14) and tense-form (PDB 5H5G mol A16). Each 1-ns step is depicted as a gray dot. The kernel density estimation for each conformation is shown as a contour map, with the contour level of each panel also shown.

Full size image

In BsFtsZ, the average structure of the relaxed conformation bound to GDP or GTP within the last 50 ns of the MD simulation was very close to that of the reference structure of the relaxed conformation, with slight differences in some loop regions (Fig. 3e,f). The deviations in relaxed BsFtsZ-GDP relative to the reference structure converged; in contrast, the conformation of relaxed BsFtsZ-GTP exhibited a greater deviation relative to the reference of the tense conformation (Fig. 4e,f). In the tense conformation, a larger fluctuation was found for the T6 loop of BsFtsZ-GDP. Moreover, the T10 and T11 loops in the GAD moved closer to the NTD, which was accompanied by a change in the relative orientation of the GAD (Fig. 3g). Therefore, the conformation of tense BsFtsZ-GDP exhibited deviations distributed over 1.5 Å relative to the reference of the relaxed conformation (Fig. 4g). Only a small difference in the loop regions was observed in the conformation of tense BsFtsZ-GTP (Fig. 3h), and consequently, the conformation converged well with the reference, with a density value of 7.2, which was two times higher than that of tense BsFtsZ-GDP (Fig. 4g,h).

Discussion

The conformations of relaxed-GDP, relaxed-GTP, and tense-GTP of both FtsZs rapidly equilibrated, with an RMSD value of < 2.0 Å relative to the initial structure (Fig. 1). On the other hand, the conformation of the tense-GDP in both FtsZs exhibited a 3.0-Å conformational change and then equilibrated later than the others, suggesting that both FtsZs share an equilibration pattern for all conformations (Fig. 1). However, the distance between the centers of the NTD and GAD in the tense SaFtsZ-GDP was shifted from 28 Å observed in the tense form to 26 Å in the relaxed form (Fig. 2b). In contrast to SaFtsZ, the center of the NTD of the tense BsFtsZ-GDP maintained a 28-Å distance from that in the GAD (Fig. 2d), but the vector between the centers formed an angle of 5.7 ± 2.1° relative to the reference (Fig. 2h and Supplementary Table S4), suggesting that the conformation converged with the structure during the MD simulation, accompanied by a previously unreported different relative orientation between the subdomains.

Each conformation of BsFtsZ exhibited less fluctuation than the corresponding conformation of SaFtsZ in the last 50 ns of the simulation (Supplementary Figs. S2 and S3). SaFtsZ had some fluctuating regions, such as the T6 and T7 loops connected by the H7 helix, and the T1 and T8 loops close to the H7 helix, depending on the conformational state with bound nucleotide. In particular, dynamic changes in the T1, T7, and T8 loops were specifically observed in the conformations of the tense SaFtsZ. In contrast, the T3, T5, and T6 loops located near the nucleotide-binding site exhibited greater fluctuations in the tense BsFtsZ. Furthermore, the T6 loop also exhibited greater flexibility in the relaxed BaFtsZ-GTP and -GDP. These results suggest that SaFtsZ and BsFtsZ exhibit different dynamics and that the more-flexible regions are located around the H7 helix in SaFtsZ and nucleotide-binding site in BsFtsZ, respectively.

In the relaxed conformation, the trajectories of the conformational differences versus the reference structures suggested that the relaxed conformation fell into a narrow range of deviation relative to the reference of the relaxed structure but was widely distributed due to a larger difference relative to the reference of the tense structure (Fig. 4). Relaxed SaFtsZ-GTP and relaxed BsFtsZ-GDP converged with the highest densities, suggesting that FtsZ in the relaxed state exhibits a different stable condition depending on the bacterial species and bound nucleotide.

The tense conformation exhibited different features among bacterial species. Quite marked differences were observed in the tense-GDP form. SaFtsZ-GDP exhibited a converged conformation with a density value of 6.4; in contrast, BsFtsZ-GDP was distributed widely and exhibited highest dynamics, with a value of < 3.15. In the tense-GTP form, however, the conformational differences of SaFtsZ and BsFtsZ with the reference of the relaxed conformation was approximately 2.0 and 2.5 Å with approximately 1.0-Å deviations, respectively, and BsFtsZ exhibited a lower difference with the reference of the tense conformation, with a higher convergence than SaFtsZ. Furthermore, the B-factors, which indicate the thermodynamical fluctuation, of atoms in the nucleotide bound to SaFtsZ during the last 50 ns revealed that GTP showed higher fluctuation than GDP, and GDP bound to tense SaFtsZ was more stable than others (Fig. 5a–d). In contrast to the fluctuation of the nucleotide, relaxed SaFtsZ-GTP exhibited lower fluctuation than relaxed SaFtsZ-GDP but tense SaFtsZ-GTP was slightly more unstable than tense SaFtsZ-GDP (Fig. 4), suggesting that higher fluctuation of the nucleotide influenced the lower conformational deviations in the relaxed SaFtsZ but the higher conformational deviations in the tense SaFtsZ conformation. However, the effect of the nucleotide stability in BsFtsZ was opposite to that in SaFtsZ, suggesting that the bound nucleotide affected the different stabilization in the relax- and tense-conformations of FtsZ between Staphylococcus aureus and Bacillus subtilis. These MD simulation results suggested that different features of both FtsZs affect the stability of their conformations.

Figure 5
figure 5

B-factor representations of atoms in the nucleotide during the last 50-ns simulation. The fluctuation behaviors of atoms in the nucleotide bound to (a) relaxed SaFtsZ-GDP, (b) relaxed SaFtsZ-GTP, (c) tense SaFtsZ-GDP, (d) tense SaFtsZ-GTP, (e) relaxed BsFtsZ-GDP, (f) relaxed BsFtsZ-GTP, (g) tense BsFtsZ-GDP, and (h) tense BsFtsZ-GTP were shown. FtsZ molecule and nucleotide were depicted by gray cartoon and sticks with the continuous scaled color of B-factor (higher; red, lower; blue).

Full size image

X-ray crystal structures of FtsZs derived from several bacteria have been reported over the last 2 decades, when the first conformation of FtsZ was reported4, and all conformations reported previously were in the relaxed state before the conformation of SaFtsZ was reported by independent groups5,19,20. The results of biochemical studies and cryoEM structural analyses of the blobbed resolution predicted that FtsZs derived from some bacteria probably assume the tense conformation8,21, 22,35,36, and indeed, FtsZs derived from only Staphylococcus aureus5,8,11,15,16,19,20,37,38,39 and Staphylococcus epidermidis (PDB 4M8I) gave crystal structures in the tense state8,21,23. The current models of treadmilling are based on the assumption that FtsZ monomers are predominantly in the relaxed conformation and that a switch to the tense conformation is caused by assembly into a protofilament8,9,10. Previous investigations also indicated that the filament may promote not only the stability of the tense conformation8,21 but also the transition to the relaxed conformation23. The results of MD simulations in this study suggested distinct dynamic properties of monomeric SaFtsZ and BsFtsZ. The relaxed BsFtsZ-GDP was identified as the most stable conformation among all conformations of BsFtsZ in the solution, consistent with the models proposed by the treadmilling8,9,10. On the other hand, the tense SaFtsZ-GDP surprisingly exhibited more stability than the relaxed SaFtsZ-GDP. The monomer structure of the relaxed SaFtsZ in the MD simulation was generated from the relaxed conformations of SaFtsZ longitudinally assembled in the crystal. The influence of the longitudinal interaction likely persisted, especially as only the T7 loop in the relaxed SaFtsZ adopts a markedly distinct conformation16. This caused the relaxed SaFtsZ model to settle into a higher energy state, conceivably a local minimum, compared to the monomeric state, despite undergoing energy minimization and equilibration. To realize a stable conformation of the relaxed SaFtsZ as calculated by the MD simulation, it is presumed that extending the time of an equilibration MD run or the crystal structure of the relaxed SaFtsZ with the conserved conformation of the T7 loop as a monomer state could be necessary.

MD simulations also suggested that the monomeric tense FtsZ derived from both bacteria did not spontaneously show the conformational transition from the tense to the relaxed conformations in the solution. Therefore, the transition to the relaxed conformation may be regulated by the protofilament instead of the spontaneous conformational change in the solution. By resolving the crystal structure of the relaxed SaFtsZ at the monomeric state and the detailed mechanisms in which residues or regions contribute to the stability of these conformations in the future, a greater understanding of the machinery that regulates conformational changes and filament treadmilling will be deeply understood.

Methods

MD simulations

MD simulations were performed using GROMACS 2016.340, as described in previous studies41,42. SaFtsZ from 5H5G molecules A and B16 and BsFtsZ from 2VXY molecule A14 were used for this study. The model structure of BsFtsZ was built with SWISS-MODEL32 using PDB 3VOA derived from SaFtsZ5 as a template. The metal ion coordinated in the PDB was removed as followed by previous MD simulation23. For each simulated system, the FtsZ molecule was also placed at the center of a rectangular periodic boundary box filled with water molecules. The size of the box used for the FtsZ molecule is summarized in Supplementary Table 1. The protonation states of residues and the hydrogen coordinates of FtsZ were determined using MOE (MOLSIS Inc., Tokyo, Japan)24. For simulation of the FtsZ-GTP complex, the GTP molecule was superposed with GDP bound to FtsZ, and the coordinates of the GDP molecule were replaced with those of GTP. The N- and C-termini of the FtsZ molecules were capped with an acetyl group and N-methyl group, respectively, using MOE 2020.09 (MOLSIS Inc.). Instead of N-methyl capping, a negative point charge was added to the C-terminus of FtsZ derived from PDB 5H5G molecule A. Sodium ions were added to neutralize the entire system as followed by the previous MD simulation of MtbFtsZ23. AMBER ff99SB-ILDN energy parameters43 were used for FtsZ and the sodium ion, and the TIP4P-Ew model44 was used for water molecules. The general Amber force field45 was used for GDP and GTP. The partial atomic charges of GDP and GTP were calculated using the restrained electrostatic potential46 methodology based on DFT calculations (B3LYP/6-31G*) using the Gaussian 09 revision E01 program package (Gaussian, Inc., Wallingford CT, 2016)47. After running the steepest descent energy minimization, relaxations were run at 250 K and 310 K for 1 ns each, with positional restraints of 1000 kJ mol−1 nm−2 for all non-hydrogen atoms of FtsZ and the nucleotides. Subsequently, the system was pre-equilibrated at 310 K for 1 ns and 1 bar using a Berendsen barostat48. An equilibration MD simulation without positional restraints was performed at 310 K and 1 bar for 200 ns. The temperature and pressure were maintained using a velocity rescaling thermostat and Parrinello-Rahman barostat with relaxation times of 0.2 and 5.0 ps, respectively41. All bonds connected to hydrogen atoms were constrained using the LINCS49 algorithm, and the time step was set to 2 fs. Long-range Coulomb interactions were calculated using the smooth particle mesh Ewald50 method, with a grid spacing of 0.30 nm. The real space cut-off for both the Coulomb and van der Waals interactions was 1.2 nm.

The geometric center of NTD and GAD were obtained from the average coordinates of all Cα atoms in the range of residue 12–173 and 223–310, respectively (Supplementary Fig. S6). The inter-subdomain distance between the geometric centers of each subdomain was calculated each time step. The vector angle was calculated from the dot product between the inter-subdomain vector at each step and the vector at 0 ns (Supplementary Fig. S7). The thermodynamical fluctuation of the atoms in the nucleotide while the last 50-ns simulation was calculated as B-factor51.