Structural analysis of S-ring composed of FliFG fusion proteins in marine Vibrio polar flagellar motor

We overproduced Vibrio His-FliFG fusion proteins with FliG-G214S or FliG-G215A mutation in E. coli cells. The membrane fraction containing the MS-ring formed by His-FliFG fusion proteins was treated with the detergent lauryl maltose neopentyl glycol (LMNG), and the MS-ring was precipitated by ultracentrifugation from the detergent-solubilized membrane fraction. We then tried to purify the MS-ring with the His-tag by cobalt-affinity chromatography; however, the MS-ring did not strongly bind to the cobalt column, and most of the MS-ring passed through the column due to unknown causes. The flow-through fraction was ultracentrifuged to precipitate the MS-ring, and the precipitated fraction was subjected to a size-exclusion column (Fig. S1B and C). The MS-ring was recovered near the void fractions as a very large molecular-weight complex. When the void fractions were subjected to sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), the FliFG fusion protein (~100 kDa protein band) was detected (Fig. S1D and E). These fractions were observed using electron microscopy (EM).

We observed the purified MS-rings formed by FliFG fusion proteins with FliG-G214S or FliG-G215A mutation by using EM following negative staining (Fig. S2A and B). After the samples were concentrated using an Amicon device, they were applied to an EM grid and rapidly frozen for cryo-EM observation. The MS-ring structure with the FliG-G214S mutation was observed by using cryo-EM (Fig. S2C); however, the MS-ring with the FliG-G215A mutation could not be observed, which was likely due to aggregation during the concentration process of the FliFG-G215A mutant proteins. Therefore, we performed a single-particle cryo-EM of the MS-ring with the FliG-G214S mutation for structure determination. A total of 228,461 particle images extracted from 5,962 cryo-EM micrographs were analyzed. They were separated into two classes in the 2D class averages, with them showing either 35- or 34-fold rotational symmetry on the S-ring (Fig. 1C and D; Fig. S3 and S4D). No other symmetry in the 2D class averages was observed, indicating that the purified MS-ring consisted of 35 or 34 subunits of the FliFG fusion proteins, and there was no other variation. Because the MS-ring in the native flagellar motor is thought to comprise 34 FliF subunits (19), it is inferred that the 34-fold rotational symmetry structure reflects the native S-ring structure, whereas the 35-fold rotational symmetry structure is an artificial structure induced by the overexpression of FliFG fusion proteins alone. After postprocess without symmetry correction, their resolutions of 35 and 34 subunit MS-rings were 3.76 and 4.28 Å, respectively. The MS-ring composed of FliFG fusion proteins purified in this study should include regions other than the S-ring, such as RBM1, RBM2, the two TM helices, the C-terminal region of FliF, and the C-terminally fused FliG. However, structures derived from these regions were not observed with sufficient resolution in our cryo-EM density maps. Therefore, we conducted detailed structural analysis focusing only on the S-ring. Consequently, we built the atomic model of the S-ring part (RBM3–β-collar) at 3.23 and 3.33 Å resolutions from the maps of both classes after C35 and C34 rotational symmetry corrections, respectively (Fig. 1E and F; Fig. S3 and S4). We built the structural model for the S-ring, but due to poor density, we could not build the model for the M-ring. It is noteworthy that weak densities corresponding to RBM2 were visible, similar in size to the previously reported RBM2 structures at the cross-section from side-views (Fig. 1C, D, G and H), although it was impossible to build the atomic model from our maps.

The S-ring is formed by residues 257–449 of Vibrio FliF (FliF_257–449), which corresponds to the latter half of the periplasmic region (Fig. 1B). The overall monomeric structure of our Vibrio S-ring was similar to previously reported monomeric structures of Salmonella S-rings (Fig. 2). The FliF_257–449 monomeric structure consists of two structural regions: a conserved globular RBM3 domain with an αββαβ motif (α1/β1/β6/α2/β7; residues 257–299 and 392–449), and a β-collar (residues 300–388) containing two sets of antiparallel β sheets (β2/β5 and β3/β4) with an invisible region (residues 328–366) at the top of the β-collar (Fig. 2A and B). The monomeric structures in the S-ring part from the 35- and 34-mer structures are almost identical (RMSD = 0.286 Å) (Fig. S5A and B). The monomeric structures of the Salmonella S-ring RBM3 were also highly similar to those of Vibrio; the RMSDs were less than 1.3 Å (Fig. S5C; Table S3).

We compared the Vibrio S-ring FliF 34-mer structure with three previously reported Salmonella S-ring FliF 34-mer structures (Fig. 3) where a notable difference is the tilt angle of the RBM3 domain. When measuring the α1 helix angle from the side-view of the ring, the RBM3 outer part is tilted downwards by 8° in the Vibrio S-ring, whereas it is tilted upwards by 6° or 8° or is nearly horizontal in the Salmonella S-ring (Fig. 3B). When comparing the β2/β5 sheet angles, they all exhibited a similar angle (55°) when viewed from the outside of the ring, whereas the angle is smallest in the Vibrio S-ring (58°) compared with the Salmonella S-ring (64°–68°) when viewed from the side (Fig. 3B). The relative angle variation between the RBM3 and the β2/β5 sheet from the side-view indicates that the linker connecting them is structurally flexible. As for the β3/β4 sheet, all exhibit similar angles of ~85°–89° when viewed from the side (Fig. 3B). Residues 309–318 connecting β2 and β3 form a protruding triangular loop (β2–β3 loop) in the Vibrio S-ring, which is shorter than the loop (281–294) in the Salmonella S-ring (Fig. 3C and D; Fig. S6A). The relative angles between the β2/β5 sheet and the β3/β4 sheet are observed to be wider in the Vibrio S-ring (28°) compared with the Salmonella S-rings (19°–25°) (Fig. 3B). This difference arises from variances in the shape of the β2–β3 loop in Vibrio and Salmonella.

When focusing on the subunit interface of the RBM3 during ring formation, interaction was facilitated by two sites: an electrostatic interaction between E270 and K276 and a hydrophobic interaction between A399, V429, and L447 in the Vibrio S-ring (Fig. 4A and B). These two interaction sites are conserved in the Salmonella S-ring, with an electrostatic interaction between E242 and R248 and hydrophobic interactions among V390, L415, and V433. In contrast, the Salmonella S-ring has additional hydrophobic interactions among L235, F237, and V241, which are not conserved in Vibrio (Fig. 4C and D), indicating that the interaction between the adjacent RBM3 regions in the S-ring is weaker in Vibrio than in Salmonella.

The internal surface of the S-ring is known to interact with the rod and export gate. Here, we found that the inner surface electrostatic distributions of the S-ring are different between Vibrio and Salmonella (Fig. 5A and B).

In Salmonella, the S-ring internal surface directly interacts with specific regions of FliE and FlgB. Some residues at the inner upper part of the β-collar in FliF (R294/S295/Q297/N365/E367) interact with a flexible loop in the D_C domain in FlgB (residues 58–82), and some residues around the base of the β-collar in FliF (E276/T278/E280/R370/I372/H374) interact with the N-terminal α-helix in FliE (Fig. 5C and D; Fig. S6 and S7) (21, 22, 26, 27). Most of the S-ring residues that interact with FliE or FlgB were conserved in our Vibrio S-ring structure, except for Q297 in Salmonella (Y321 in Vibrio) and E280 and R370 in Salmonella (K308 and T379 in Vibrio) (Fig. 5D and E; Fig. S6A).

The bacterial strains and plasmids used here are listed in Table S1. E. coli cells were cultured in LB medium (1% [wt/vol] bactotryptone, 0.5% [wt/vo;] yeast extract, and 0.5% [wt/vol] NaCl). Ampicillin was added at a final concentration of 100 µg/mL. Point mutations (FliG-G214S or FliG-G215A) in plasmid pRO301 were introduced using the QuikChange site-directed mutagenesis method (Agilent).

An overnight culture of E. coli cells containing plasmid (pRO302 or pRO303) was inoculated at 1/50 dilution into 50 mL of LB medium containing ampicillin and grown at 37°C for 4 h. The 20-mL culture was inoculated into 2 L of LB medium containing ampicillin and grown at 37°C for 3–4 h. When the optical density at 660  nm reached 0.4–0.5, isopropyl-β-D-thiogalactopyranoside was added at 0.5 mM, the culture was cooled on ice for 20 min, and then incubated at 16°C for 20 h. The cells were collected by low-speed centrifugation and the pellet was stored at −20°C, if necessary. The pellet was resuspended in 25 mL of TK buffer (20 mM Tris-HCl [pH 8.0], 200 mM KCl) or TN buffer (20 mM Tris-HCl [pH 8.0], 200 mM NaCl) containing 1mM EDTA.

The suspension was placed in a conical tube and sonicated three times at a power of 6 and 50% duty ratio for 1 min. After low-speed centrifugation, the supernatant was recovered, and the precipitated cells were suspended in 25 mL of TK or TN buffer containing 1 mM EDTA and sonicated again under the same conditions as before (repeated twice in total). A 1/500 vol of 1 M MgCl₂ was added to the pooled supernatants and centrifuged at low speed. The resulting supernatant was ultracentrifuged at 150,000 × g for 60 min. The precipitates were suspended in 20 mL of TK or TN buffer, and 2 mL of 10% (wt/vol) LMNG was added and incubated at 37°C for 30 min. After low-speed centrifugation, the supernatant was ultracentrifuged at 150,000 × g for 60 min. The precipitate was suspended in 1 mL of TK or TN buffer containing 0.01% (wt/vol) LMNG. The suspension was applied to a cobalt column, and the flow-through fraction was ultracentrifuged and suspended in 1 mL of TK or TN buffer containing 0.01% (wt/vol) LMNG. The suspended solutions were subjected to the size-exclusion column (Superose 6 10/300, GE healthcare) equilibrated with TK100L buffer (20 mM Tris-HCl [pH 8.0], 100 mM KCl, 0.0025% [wt/vol] LNMG) or TN100L buffer (20 mM Tris-HCl [pH 8.0], 100 mM NaCl, 0.0025% [wt/vol] LNMG).

Elution fractions containing the MS-ring composed of FliFG fusion proteins with the FliG-G214S or FliG-G215A mutations were concentrated five-fold using an Amicon Ultra 100 K device (Merck Millipore). The G214S and G215A mutants were diluted 50- and 100-fold, respectively, in TN100L buffer. A 5-µL solution was applied to a glow-discharged continuous carbon grid. Excess solution was removed using filter paper, and the sample was subsequently stained on a carbon grid with 2% (wt/vol) ammonium molybdate. Images were recorded using a H-7650 transmission electron microscope (Hitachi) operated at 80 kV and equipped with a FastScan-F114 CCD camera (TVIPS) at a nominal magnification of 40,000×.

A concentrated MS-ring sample composed of the FliFG fusion protein with the FliG-G214S mutation was applied to a Quantifoil holey carbon grid (R1.2/1.3 Cu 300 mesh, Quantifoil Micro Tools GmbH) with glow-discharge treatment on one side of the grid. The grids were placed in liquid ethane cooled with liquid nitrogen for rapid freezing using a Vitrobot Mark IV (Thermo Fisher Scientific) with a blotting time of 7 s at 4°C and 100% humidity. The data were collected on Titan Krios electron microscope (FEI) equipped with a thermal field-emission electron gun operated at 300 kV. Image data sets of 6,372 micrographs collected by a Titan Krios microscope were automatically recorded on a K3 direct electron detector camera (Gatan) at a nominal magnification of 64,000× corresponding to a pixel size of 1.14 Å with a defocus range from −0.7 to −1.7 µm, using the SerialEM software. Micrographs were taken after a total exposure time of 7.329 s, and an electron dose of 0.78 electrons/Ų per frame. The 64 micrograph frames were recorded at a rate of 0.115 s/frame. The data collection and image analysis are presented in Fig. S3 and Table S2.

A total of 6,372 micrographs were motion corrected, and their contrast transfer function (CTF) values were estimated using the CryoSPARC package. A total of 500 micrographs were used for the initial particle picking by using a blob picker. A total of 125,759 particles were automatically extracted from the micrographs, and 2D classification was performed thrice to remove false particles. In total, 15,833 particles selected from the 2D classification were used as input templates for the template picker. A total of 2,059,366 particles were automatically extracted from the micrographs, and 2D classification was performed once to remove false particles. The top or bottom views of the 2D classification averages with 35- or 34-fold rotational symmetries were observed, and all true particles (860,145 particles) were combined for the subsequent 3D classification. Ab initio reconstruction was performed to generate two initial 3D models using 512,737 particles selected from the 2D classification. Heterogeneous refinement with C35 symmetry was then performed using two of the three initial models to generate a 3D model with C35 symmetry. Heterogenous refinement was performed with 860,145 particles by using two of three 3D models with C35 symmetry and 2D classification to remove false particles. The remaining 510,049 particles underwent heterogeneous refinement applied individually with C1, C34, and C35 symmetries to generate 3D models. The remaining 320,933 particles from heterogeneous refinement without symmetry (C1) underwent further heterogeneous refinement using the C34 and C35 models. Of these, 261,524 and 55,606 particles remained in classes with C35 and C34 symmetries, respectively. The particles belonging to each class underwent homogenous refinement with C35 and C34 symmetries to generate models for the final heterogeneous refinement to remove false particles. The remaining 183,915 and 43,546 particles with C35 and C34 symmetries, respectively, underwent homogeneous symmetry refinement and Global CTF refinement. Subsequently, the particles with C35 symmetry underwent homogenous refinement, and a map was obtained at 3.23-Å resolution (Class 1; C35). To relax the symmetry, local refinement was performed, and a map was obtained at 3.76-Å resolution (Class 1; C1). Particles with C34 symmetry underwent homogeneous refinement, and a map was obtained at 3.33-Å resolution (Class 2; C34). Moreover, homogenous refinement without symmetry and homogenous refinement were performed, and a map was obtained at 4.28-Å resolution (Class 2; C1). The local resolution of the 3D volumes with C34 and C35 symmetries was estimated using the CryoSPARC package at 0.143 of its Fourier shell correlation threshold.

The atomic model of the S-ring was constructed from the maps of Class 1 (C35) and Class 2 (C34) using Coot (29) and refined using Phenix (30). The initial models were built from the monomeric RBM3–β-collar structure of Vibrio FliF produced by SWISS-MODEL (swissmodel.expasy.org) based on the RBM3–β-collar structure from Salmonella (PDB ID: 8T8P). A summary of the model refinement is presented in Table S2. Structural comparisons and analyses were performed using PyMOL (Schrödinger) and Chimera (31).

HS-AFM imaging was performed using a laboratory-built HS-AFM operated in the tapping mode, as previously described (24). The MS-rings composed of FliFG fusion proteins were deposited on a bare mica substrate for HS-AFM imaging. After 5 min of incubation, the residual proteins were washed off using observation TN100L buffer. HS-AFM imaging was performed in the sample solution at room temperature.