Structures of the Type IX Secretion/Gliding Motility Motor from across the Phylum Bacteroidetes.

Gliding motility using cell surface adhesins, and export of proteins by the type IX secretion system (T9SS) are two phylum-specific features of the Bacteroidetes. Both of these processes are energized by the GldLM motor complex, which transduces the proton motive force at the inner membrane into mechanical work at the outer membrane. We previously used cryo-electron microscopy to solve the structure of the GldLM motor core from Flavobacterium johnsoniae at 3.9-Å resolution (R. Hennell James, J. C. Deme, A. Kjaer, F. Alcock, et al., Nat Microbiol 6:221-233, 2021, https://dx.doi.org/10.1038/s41564-020-00823-6). Here, we present structures of homologous complexes from a range of pathogenic and environmental Bacteroidetes species at up to 3.0-Å resolution. These structures show that the architecture of the GldLM motor core is conserved across the Bacteroidetes phylum, although there are species-specific differences at the N terminus of GldL. The resolution improvements reveal a cage-like structure that ties together the membrane-proximal cytoplasmic region of GldL and influences gliding function. These findings add detail to our structural understanding of bacterial ion-driven motors that drive the T9SS and gliding motility. IMPORTANCE Many bacteria in the Bacteroidetes phylum use the type IX secretion system to secrete proteins across their outer membrane. Most of these bacteria can also glide across surfaces using adhesin proteins that are propelled across the cell surface. Both secretion and gliding motility are driven by the GldLM protein complex, which forms a nanoscale electrochemical motor. We used cryo-electron microscopy to study the structure of the GldLM protein complex from different species, including the human pathogens Porphyromonas gingivalis and Capnocytophaga canimorsus. The organization of the motor is conserved across species, but we find species-specific structural differences and resolve motor features at higher resolution. This work improves our understanding of the type IX secretion system, which is a virulence determinant in human and animal diseases.

INTRODUCTION

The type IX secretion system (T9SS) is a protein export system found exclusively in the Bacteroidetes phylum of Gram-negative bacteria (1, 2). Substrates of the T9SS are transported to the periplasm by the Sec system, following which a conserved C-terminal domain (CTD) directs export through an outer membrane T9SS translocon (Fig. 1a) (2, 3). In most cases, the CTD is then removed and the substrate protein is either released into the environment or anchored to the outer membrane as a lipoprotein (4). The human oral pathogen Porphyromonas gingivalis uses the T9SS to secrete gingipain proteases and other virulence factors to evade the host immune system (5). The T9SS has also been identified as essential to the virulence of several economically relevant fish and poultry pathogens (6–8). In commensal and environmental Bacteroidetes species, the T9SS is characteristically used to secrete enzymes that enable the organisms to utilize complex polysaccharides as a food source (1, 9, 10). Many Bacteroidetes species with a T9SS also exhibit gliding motility, in which cells travel rapidly across surfaces (2). This motility depends on the movement of cell surface adhesin molecules that are secreted to the cell surface by the T9SS (Fig. 1a).

FIG 1

Role and phylogenetic diversity of the GldLM motor complex. (a) Cartoon illustrating the involvement of the GldLM motor complex in theT9SS and gliding motility. The GldLM motor converts electrochemical potential energy from the proton-motive force across the inner membrane (IM) into mechanical work that the periplasmic portion of GldM transfers across the periplasm to the outer membrane (OM). This mechanical energy is used to drive gliding adhesin movement (left) and protein transport through the T9SS (right). Coupling between these processes and GldM is thought to be mediated by a GldKN lipoprotein complex. (b) Cartoon representation of the structure of the F. johnsoniae GldLM′ complex solved previously (11) (PDB no. 6SY8 and EMDB no. EMD-10893). (Left) Whole structure. The five GldL chains are colored salmon, blue, green, teal, and tan and the two GldM chains are colored dark gray and white. (Right) Individual GldL and GldM chains are shown and rainbow colored from the N terminus (blue) to the C terminus (red). The most N-terminal (N′) and C-terminal (C′) modeled residues of each chain are marked with a sphere. (c, d) Maximum-likelihood phylogenetic tree of GldL (c) and GldM (d) sequences in the Bacteroidetes phylum. Branches are colored by taxonomic order and the positions of proteins for which structures were determined are indicated. (e) Increased resolution of the new T9SS/gliding motor complex structures shows improved side chain density. Chain GldLc is shown for each species with EM density displayed at the same contour level.

The T9SS and the gliding motility apparatus share a motor complex that uses the proton motive force (PMF) across the inner membrane to drive both protein transport and gliding adhesin movement at the outer membrane (11–13). The motor complex is formed from the integral inner membrane proteins GldL and GldM (11, 14). GldL has two transmembrane helices (TMHs) and a cytoplasmic domain. GldM has one TMH and a large periplasmic region, which crystal structures have shown forms an extended dimer of four domains (D1 to D4) (11, 15, 16). The periplasmic region of GldM is long enough to span the periplasm to contact the outer membrane components of the T9SS and gliding motility apparatus.

We previously solved the structure of the core of the GldLM motor complex from the gliding bacterium Flavobacterium johnsoniae (11). This structure contains the full transmembrane region of the motor complex, but the cytoplasmic domain of GldL was not visible, and the periplasmic region of GldM had been genetically truncated after the first (D1) domain, forming a construct termed FjoGldLM′ (11). The structure reveals that the T9SS/gliding motor is a GldL5GldM2 heteroheptamer in which the 10 GldL TMHs surround the two GldM TMHs (Fig. 1b). The symmetry mismatch between the total number of GldL and GldM TMHs results in an inherently asymmetric relationship between the two types of subunit around the GldL ring. Amino acid substitutions showed that several conserved protonatable residues in the transmembrane helices of GldL and GldM are important for T9SS and gliding motility function. These residues are likely to be involved in coupling proton flow across the inner membrane to mechanical motions in the motor. Based on the organization of the GldLM transmembrane helices and on the structural homology of the transmembrane part of GldLM to the ion-driven motor complexes that drive bacterial flagella (17, 18), we proposed that GldLM forms a rotary motor in which the GldM subunits rotate within the ring of GldL helices (11, 17). The periplasmic domain of GldM is then envisaged to transmit this rotary motion across the periplasm to the outer membrane components of the T9SS and gliding motility systems.

The structure of the FjoGldLM′ complex was determined to a resolution (3.9 Å) at which only limited information can be inferred about the position of mechanistically important amino acid side chains. In addition, the protein was captured in a single conformational state, providing only one snapshot of the mechanism of the motor complex. Here, we have sought to overcome these limitations in the structural characterization of the GldLM motor by determining the structures of the motor complexes from a phylogenetically diverse range of organisms. We anticipated that some of these complexes would allow structure determination at improved resolution and in alternative conformational states. Here, we present structures of the T9SS/gliding motor core from the human pathogens P. gingivalis and Capnocytophaga canimorsus and the environmental bacteria Schleiferia thermophila and Sphingobacterium wenxiniae.

RESULTS AND DISCUSSION

The architecture of GldLM′ complexes is conserved across the Bacteroidetes phylum.

To structurally survey the diversity of gliding motility/T9SS motor complexes, we selected proteins from a range of Bacteroidetes bacteria for recombinant expression in Escherichia coli. These proteins were chosen to maximize phylogenetic spread, to include proteins from organisms growing at a range of temperatures and from different environments (marine, fresh water, terrestrial, and commensals/pathogens), and to include examples both from gliding bacteria and from nongliding bacteria with a T9SS. All constructs included the full-length GldL homologue. For the GldM homologue, we trialed constructs that included different numbers of periplasmic domains. However, as with the earlier F. johnsoniae GldLM′ structure, we were only able to obtain structures by cryo-electron microscopy (cryo-EM) from complexes in which GldM was truncated after the first periplasmic D1 domain (GldM′), with the exception of one construct in which GldM was truncated after the D2 domain (GldM″). However, even in the latter case, this effectively produced a GldLM′ structure, as the D2 domain was not resolved, as discussed below.

Following screening for expression, purification, and cryo-freezing, we determined structures for the T9SS/gliding motor complexes of P. gingivalis (PgiPorLM′), Capnocytophaga canimorsus (CcaGldLM″peri and CcaGldLM″TMH), Schleiferia thermophila (SthGldLM′), and Sphingobacterium wenxiniae (SweGldLM′) (Fig. 1 and Table 1; see also Fig. S1 to S6 in the supplemental material). P. gingivalis is a nongliding oral pathogen, the gliding bacterium C. canimorsus is a dog commensal and opportunistic human pathogen, S. thermophila is a thermophile isolated from a hot spring (optimum growth temperature, 50°C), and S. wenxiniae was isolated from a wastewater treatment plant. Note that in P. gingivalis the motor proteins are termed PorLM rather than GldLM. Figure 1c and d show the phylogenetic positions of these motor proteins within the diversity of GldL and GldM proteins. The GldL and GldM sequence similarity matches the bacterial phylogenetic tree in most cases, indicating that GldL has been predominantly vertically rather than horizontally transmitted, in agreement with an earlier analysis (19). The resolutions of the new structures ranged from 3.0 to 3.9 Å, with the higher-resolution structures allowing for more confident positioning of side chains than our previous 3.9-Å resolution FjoGldLM′ structure (Fig. 1e).

TABLE 1

Cryo-EM data collection, refinement, and validation statistics for the PgiPorLM′C, SthGldLM′, CcaGldLM″peri, GldLM″TMH, and SweGldLM′ structures

Statistic	P. gingivalis PorLM′ (PDB no. 7SAT, EMDB no. EMD-24956)	S. thermophila GldLM′ (PDB no. 7SAU, EMDB no. EMD-24957)	C. canimorsus GldLM″peri (PDB no. 7SB2, EMDB no. EMD-24961)	C. canimorsus GldLM″TMH (PDB no. 7SAZ, EMDB no. EMD-24959)	S. wenxiniae GldLM′ (PDB no. 7SAX, EMDB no. EMD-24958)
Data collection and processing
Magnification (×)	81,000	105,000	105,000	105,000	105,000
Voltage (kV)	300	300	300	300	300
Electron exposure (e− Å−2)	55.6	62.4 (without fOM), 61.2 (with fOM)	59.1	59.1	56.9
Defocus range (μm)	1.0–3.0	1.0–3.0	1.0–3.0	1.0–3.0	1.0–3.0
Pixel size (Å)	0.832	0.832	0.832	0.832	0.832
Symmetry imposed	C1	C1	C1	C1	C1
Initial particle images (no.)	8,205,503	13,743,455	9,197,926	9,197,926	7,167,266
Final particle images (no.)	649,359	394,678	595,559	77,223	111,727
Map resolution (Å)	3.9	3.0	3.4	3.0	3.0
FSC threshold	0.143	0.143	0.143	0.143	0.143
Map resolution range (Å)	3.7–4.9	2.8–3.7	3.3–6.0	2.8–4.3	3.0–4.3
Refinement
Initial model used (PDB code)	None	None	None	None	None
Model resolution (Å)	3.9	3.0	3.4	3.0	3.0
FSC threshold	0.143	0.143	0.143	0.143	0.143
Model resolution range (Å)	3.7–4.9	2.8–3.7	3.3–6.0	2.8–4.3	3.0–4.3
Map sharpening B factor (Å2)	−200	−93	−122	−58	−83
Model composition
No. of nonhydrogen atoms	6,581	6,317	4,728	5,708	5,301
No. of protein residues	822	791	600	722	689
Ligands	0	0	0	0	0
B factors (Å2)
Protein	72.16	51.61	93.02	55.39	63.20
Ligand	NAb
RMSa deviation
Bond length (Å)	0.004	0.005	0.005	0.005	0.009
Bond angle (°)	0.747	0.600	0.781	0.675	0.700
Validation
MolProbity score	2.16	1.68	2.39	1.82	1.50
Clashscore	15.42	5.98	22.84	9.65	5.09
Poor rotamers (%)	0.28	0.15	0.00	0.00	0.00
Ramachandran plot
Favored (%)	92.57	94.85	90.34	95.48	96.44
Allowed (%)	7.43	5.15	9.66	4.38	3.56
Disallowed (%)	0.00	0.00	0.00	0.14	0.00

RMS, root mean square.

NA, not applicable.

Purification of GldLM′′, PorLM′, and GldLM′ complexes. (a) Size exclusion chromatography traces for the indicated protein complexes. Proteins were analyzed using a Superose 6 10/300 Increase column (Cytiva). The peaks used to prepare cryo-electron microscopy (cryo-EM) grids are indicated with arrowheads. (b to e) SDS-PAGE gels for the fractions used to make cryo-EM grids for the indicated protein complexes. Download FIG S1, TIF file, 1.9 MB.

The overall architecture of the FjoGldLM′ complex is conserved in the four new motor complex structures, with five copies of GldL surrounding two copies of GldM (Fig. 2a to c). As in the FjoGldLM′ structure, only the transmembrane helices and periplasmic loop of GldL were fully resolved, with almost all of the C-terminal cytoplasmic domain not seen in the structures (11). The periplasmic D1 domains of the GldM dimer were visible in all structures. The precise angles between the two copies of the D1 domain varied from 30° to 45° between structures. However, in all cases the D1 pairs adopted the splayed arrangement seen in the previous FjoGldLM′ structure. This splayed arrangement contrasts with the closed arrangement of the D1 pair seen in crystal structures of the isolated periplasmic domain of GldM/PorM (15, 16), but is consistent with the low-resolution cryo-EM structure of full-length PgiPorLM (11).

FIG 2

GldLM′ has conserved architecture across the Bacteroidetes phylum. (a) EM density maps for GldLM″TMH/GldLM′/PorLM′ complexes from the indicated species at high (colored by protein chain) and low (transparent) contour. The structure of F. johnsoniae GldLM′ was solved previously (11). The resolution of the structures is indicated above the panels. (b) Cartoon representations of the structures with chains colored as in panel a. (c) Slab through the protein density from panel a viewed from the cytoplasm and sliced approximately half-way through the membrane region. (d) The new GldLM″TMH/GldLM′/PorLM′ complex structures (colored as indicated) overlaid on F. johnsoniae GldLM′ (orange). GldM″TMH/GldM′/PorM′ subunits are shown in a darker shade than the GldL/PorL subunits. (e) Conservation of residues that are functionally important in F. johnsoniae GldLM (11) (top left panel) in other GldLM″TMH/GldLM/PorLM complexes. The proposed intersubunit salt bridge is between the labeled Glu residue in the salmon GldL chain and the labeled Arg residue in the white GldM chain. For clarity chains GldLD, GldLE and GldLG are hidden for each structure.

Structural data were obtained from a C. canimorsus construct (CcaGldLM″), which includes both the D1 and D2 periplasmic domains of GldM. During three-dimensional (3D) classification, classes could be identified by either the D2 domain or the transmembrane portion of the complex being well-resolved (the latter corresponding to the GldLM′ complex structures determined from other organisms), but never both (Fig. S2 and S3 and Fig. S7a and b). This suggests that the CcaGldLM″ complex did not adopt a conformation in which both the transmembrane helices and D2 domains are simultaneously ordered. In the CcaGldLM″ structure with the ordered transmembrane helices (CcaGldLM″TMH), the D1 domains are splayed as in the other GldLM′/PorLM′ structures and in the low-resolution structure of the complete PorLM complex (Fig. S7a and d, 11). In contrast, in the CcaGldLM″ structure where the transmembrane domain is unresolved (CcaGldLM″peri), the D1 domains adopt a parallel orientation (Fig. S7b, c, and e) similar to the isolated FjoGldM periplasmic domain crystal structure (15), even though no crystal contacts are present. These observations suggest that splaying of the D1 domains is the most stable arrangement of the D1 domains in the intact motor complex, but leaves open the possibility that a parallel arrangement of the D1 domains might occur transiently during operation of the motor.

Data processing workflow for the CcaGldLM″peri map. (a) Example micrograph for the CcaGldLM′′ sample collected at approximately 2.6 μm defocus. Bar, 500 Å. (b) Representative 2D class averages used to produce the initial model. Bar, 100 Å. (c) Representative 2D class averages used to produce the CcaGldLM″peri map. Bar, 100 Å. (d) Data processing workflow for the CcaGldLM″peri map. The handedness of the maps was flipped between the left and right columns once helices were visible. (e) Local resolution estimates (in Å) for the sharpened CcaGldLM″peri map. (f) Fourier shell correlation (FSC) plot for the CcaGldLM″peri map. Resolution at the gold-standard cutoff (FSC = 0.143) is indicated by the dashed line. Curves are colored as follows: red, phase-randomized; green, unmasked; blue, masked; black, modulation transfer function (MTF) corrected. Download FIG S2, TIF file, 1.6 MB.

Data processing workflow for the CcaGldLM″TMH map. (a) Representative 2D class averages used to produce the CcaGldLM″TMH map. Bar,100 Å. (b) Data processing workflow for the CcaGldLM″TMH map. (c) Local resolution estimates (in Å) for the sharpened CcaGldLM″TMH map. (d) FSC plot for the CcaGldLM″TMH map. Resolution at the gold-standard cutoff (FSC = 0.143) is indicated by the dashed line. Curves are colored as follows: red, phase randomized; green, unmasked; blue, masked; black, MTF corrected. Download FIG S3, TIF file, 2.9 MB.

Data processing workflow for the PgiPorLM′ map. (a) Example micrograph for the initial PgiPorLM′ dataset collected with a K2 detector at approximately 1.8 μm defocus. Bar, 500 Å. (b) Representative 2D class averages from the K2 dataset used for initial processing. Bar, 100 Å. (c) Example micrograph from the second PgiPorLM′ dataset collected with a K3 detector at approximately 1.8 μm defocus. Bar, 500 Å. (d) Representative 2D class averages from the K3 dataset used for final processing. Bar, 100 Å. (e) Data processing workflow for the PgiPorLM′ map. Scale bar on 2D class averages, 100 Å. (f) Local resolution estimates (in Å) for the sharpened PgiPorLM′ map. (g) Fourier shell correlation (FSC) plot for the PgiPorLM′ map. Resolution at the gold-standard cutoff (FSC = 0.143) is indicated by the dashed line. Curves are colored as follows: red, phase randomized; green, unmasked; blue, masked; black, MTF corrected. Download FIG S4, TIF file, 1.7 MB.

Data processing workflow for the SthGldLM′ map. (a) Example micrograph for the initial SthGldLM′ dataset collected with a K2 detector at a defocus of approximately 2.7 μm. (b) Example micrograph for the SthGldLM′ plus fluorinated octyl maltoside (fOM) dataset collected with a K3 detector at a defocus of approximately 1.7 μm. (c) Example micrograph for the SthGldLM′ dataset collected with a K3 detector at defocus of approximately 1.7 μm. (d) Representative 2D class averages for the K2 dataset used in initial processing. (e) Data processing workflow for the initial dataset collected with a K2 detector. (f) Representative 2D class averages for the SthGldLM′ plus fOM (left) and SthGldLM′ only (right) K3 datasets. (g) Data processing workflow for the combined datasets collected with a K3 detector. In the bottom left panel, each cylinder represents a view orientation, and the height of the cylinder corresponds to the number of views of that orientation. (h) Local resolution estimates (in Å) for the sharpened SthGldLM′ map. (i) FSC plot for the SthGldLM′ map. The resolution at the gold-standard cutoff (FSC = 0.143) is indicated by the dashed line. Curves are colored as follows: red, phase randomized; green, unmasked; blue, masked; black, MTF corrected. Scale bar for micrographs, 500 Å. Scale bar for 2D class averages, 100 Å. Download FIG S5, PDF file, 0.08 MB.

Data processing workflow for the SweGldLM′ map. (a) Example micrograph for SweGldLM′ dataset collected with K3 detector at defocus of approximately 1.5 μm. Bar, 500 Å. (b) Representative 2D class averages for initial particle selection. Bar, 100 Å. (c) Data processing workflow for initial particle selection. (d) Representative 2D class averages for second, side view-focused particle selection. Bar, 100 Å. (e) Data processing workflow for second particle selection. (f) Local resolution estimates (in Å) for the sharpened SweGldLM′ map. (g) Fourier shell correlation (FSC) plot for the SweGldLM′ map. The resolution at the gold-standard cutoff (FSC = 0.143) is indicated by the dashed line. Curves are colored as follow: red, phase randomized; green, unmasked; blue, masked; black, MTF corrected. Download FIG S6, PDF file, 0.06 MB.

Comparison of CcaGldLM″TMH and CcaGldLM″peri structures. (a) CcaGldLM″TMH EM density map. GldL is colored light blue and GldM″TMH is colored dark blue. (b) CcaGldLM″peri EM density map. (c) Protein model of CcaGldLM″peri. It was not possible to build protein into the transmembrane density. (d) Overlay of the CcaGldLM″TMH and CcaGldLM″peri structures showing the different relative orientations of the GldM D1 domains. Proteins are colored as in panels a and c. (e) Overlay of CcaGldLM″peri (yellow) and FjoGldMperi (PDB no. 6EY4) (orange) structures, showing the similarity in GldM D1 domain orientations. All maps and models in this figure were aligned to the second helix of CcaGldLM″peri chain A, indicated with a dashed line in panel c. Download FIG S7, TIF file, 2.1 MB.

Mutagenesis was previously used to identify residues within the transmembrane domain of the F. johnsoniae GldLM motor that are important for function (11). These residues are well conserved in the new motor structures, with the exception that FjoGldL Tyr13 is replaced by isoleucine in S. wenxiniae GldLM. The side chain positions of these residues can be assigned with more confidence due to the improved resolution of the CcaGldLM″TMH, SthGldLM′, and SweGldLM′ structures. Notably, the orientation of these side chains is similar between the different motor structures where the resolution of the structures allows this judgement (Fig. 2e). The side chain position is least well defined for the FjoGldL Glu49 equivalent, which we previously proposed forms a salt bridge between one copy of GldL and Arg9 in one of the two FjoGldM molecules (11). The poor definition of this residue is not surprising, as glutamate sidechains are susceptible to damage by the electron beam during cryo-EM experiments, meaning that they are often not visible in EM density maps and cannot be accurately modeled (20). Nevertheless, in all the motor protein structures, the side chain of the FjoGldM Arg9 equivalent is oriented toward the FjoGldL Glu49 equivalent, suggesting that a salt bridge interaction between these residues is a conserved feature of the GldLM/PorLM complex.

The membrane-proximal region of GldL forms a cage-like structure.

The majority of the cytoplasmic region of GldL was not visible in any of our structures. However, the part that is immediately adjacent to the inner face of the cytoplasmic membrane was much better resolved in the PgiPorLM′, SthGldLM′, and SweGldLM′ structures than in the previously determined FjoGldLM′ structure. It can now be seen that extended coils from the C-terminal end of TMH2 form a cage-like structure below the detergent micelle (Fig. 3a). In SweGldLM′, the coil was only fully resolved for chain GldLC, which is involved in the putative salt bridge with chain GldMB. This coil is braced by the N terminus of the adjacent chain GldLD, an interaction not seen in the other structures. The cage structure is resolved best in the SthGldLM′ complex, revealing that the constituent coils are held together by a network of hydrogen bonds and hydrophobic packing interactions between aspartate, tryptophan, and tyrosine residues (Fig. 3c and d). The cage structure exhibits high sequence conservation, suggesting that it is of structural and/or functional importance (Fig. 3b and e). The cytoplasmic interactions between the GldL chains in the cage structure could help coordinate movements between subunits that do not contact each other in the TMH bundle. The surface of the cage is acidic (Fig. 3f) creating a region of negative charge that may assist in the release of protons flowing through the transmembrane part of the motor complex. Alternatively, it may help maintain separation between the cytoplasmic domain of GldL and the negatively charged phospholipid head groups of the cytoplasmic membrane.

FIG 3

The membrane-proximal part of the GldL cytoplasmic domain forms a cage. (a) Overlay of EM density and the built model for the cytoplasmic region of each GldL structure. EM density is shown at the same contour level for all species. Side chains are shown for Chain C in the cage region. The most N-terminal residue modeled for Chain D is indicated with an arrowhead to highlight the bracing interaction between Chains C and D of SweGldLM′. (b) Sequence alignment of the cage region for the five GldL/PorL sequences. Residues that could be modeled for each structure are highlighted in green. Residues constituting the cage region are boxed in red. (c, d) Interaction network at the base of the cage-like structure in SthGldLM′. The view direction is parallel to the membrane (c) or from within the TMH bundle (d). For clarity chains D and F are hidden in panel c. Hydrogen bonds are shown as yellow dashes. Selected side chains are displayed for the same residues in each chain and are labeled on (c) Chain C (salmon) or (d) Chain F (teal). For other residues, backbone atoms are shown if they form hydrogen bonds. (e) Sequence conservation analysis of the cage region of S. thermophila GldL using the program Consurf (40, 41). (f) Coulombic surface potential representation for the cage region of S. thermophila GldL. The view direction is from within the TMH bundle (left) or parallel to the membrane (right). For clarity, the first TMH and N-terminal residues of GldL are hidden. (g) Effects of modifications in the GldL cytoplasmic cage on F. johnsoniae gliding motility (spreading) on plates. The region of the F. johnsoniae sequence enclosed by the red box in panel b (residues E64-L74) was either substituted with GSSGSSGSSGS (gldL) or deleted (gldLΔ). The results are representative of three independent experiments.

We investigated the importance of the cage structure to motor function through mutagenesis of the chromosomal gldL gene in the genetically tractable organism F. johnsoniae. Mutant cells in which the cage was completely deleted (removal of residues 64 to 74; gldLΔ) had a small but reproducible gliding defect as measured by colony spreading on agar plates, while cells in which the cage sequence was replaced by a GSS repeat linker of the same length (gldL) showed no gliding defect (Fig. 3g). The fact that changing the cage sequence to a GSS repeat had no effect on motility on agar indicates that the length of this cage region, rather than its precise sequence or structure, is most important to motor function.

P. gingivalis PorL has a N-terminal helix that is absent in other structurally characterized GldL/PorL proteins.

The P. gingivalis PorL protein has a N-terminal extension relative to the other motor proteins that we have structurally characterized (Fig. 4a). This extension forms a helix that is not present in the other structures (Fig. 4b). The helix points away from the transmembrane GldL helix bundle at an angle of approximately 100° from TMH1 (Fig. 4c and d) and approximately tangential to the circumference of the bundle (Fig. 4b). The helix lies against the curved surface of the detergent micelle (Fig. 4e) suggesting that in vivo it is likely to lie along the membrane surface with the positively charged N terminus (Fig. 4b and c) interacting with the negatively charged phospholipid head groups. This helix may, therefore, play a role in stabilizing the position of the PorLM complex in the cytoplasmic membrane. The presence of this N-terminal helix in PgiPorLM′ and the bracing interaction seen between the N terminus of GldLD and the cytoplasmic region of GldLC in the SweGldLM′ structure noted in the last section suggests a role for the N terminus of GldL in species-specific functional tuning of the motor complex.

FIG 4

P. gingivalis PorL has a N-terminal membrane surface-associated helix. (a) Sequence alignment of the N-terminal regions of the structurally characterized GldL homologues. The first transmembrane helix of each sequence is highlighted in blue. The additional N-terminal helix of PgiPorL is highlighted in red. (b) View from the cytoplasm of P. gingivalis PorLM′ in cartoon (left) and coulombic surface (right) representation. For clarity PorM′ is hidden in the coulombic view. (c) Coulombic surface (left) and cartoon (right) representations of P. gingivalis PorL Chain C viewed from within the membrane. (d) Side view of PgiPorLM′ in cartoon representation. (e) Side view of the PgiPorLM′ complex model overlaid with the EM density map displayed at low contour level. The approximate boundary of the detergent micelle is marked with a dashed line. (b to d) An asterisk (*) indicates the N-terminal helix, and C′ indicates the most C-terminal modeled residue.

Conclusion.

This work, together with our previous study (11), provides the structures of the transmembrane cores of five T9SS/gliding motor complexes from species across three orders of the Bacteroidetes. These structures show that the architecture of the GldLM motor complex is well conserved and imply that the mechanism by which the motor converts proton flow to mechanical movement is the same across the Bacteroidetes phylum. The yield of the recombinant S. wenxiniae GldLM′ complex is much higher than that of the previously purified FjoGldLM′ protein (Fig. S1) (11), which should expedite future in vitro mechanistic studies of the T9SS/gliding motor. Future work should also explore how GldL and GldM interact with other components of the T9SS and gliding motility machinery to convert motor motions into the useful work of protein translocation and adhesin propulsion.

MATERIALS AND METHODS

Bioinformatics analysis.

A phylogenetic tree of GldL sequences was generated as follows. GldL sequences were obtained by BLAST searches against the UniRef90 database using the F. johnsoniae GldL, P. gingivalis PorL, and S. wenxiniae GldL sequences as queries (21, 22). Each sequence in the UniRef90 database represents a cluster of sequences with more than 90% identity to the representative sequence. Searching against the UniRef90 database reduces the number of highly similar sequences in the results compared to searching against an unfiltered database. Sequences duplicated between the three BLAST searches, sequences representing clusters where the lowest common taxon was higher than order (and thus likely incorrectly phylogenetically assigned), and sequences from organisms outside the Bacteroidetes phylum were removed. The sequences were then aligned using Clustal Omega (23). A phylogenetic tree was inferred using the maximum likelihood function of the MEGA X program with a Jones-Taylor-Thornton (JTT) matrix-based model with default settings (24, 25). A phylogenetic tree of GldM sequences was generated using the same approach.

Bacterial strains and growth conditions.

Strains and plasmids used in this study are listed in Table S1 in the supplemental material. For cloning procedures, E. coli cells were routinely grown in Luria-Bertani (LB) medium (26) at 37°C with shaking. F. johnsoniae cells were routinely grown in Casitone yeast extract (CYE) medium (27) at 30°C with shaking. PY2 medium (28) was used to assess motility on agar plates. When required, kanamycin was added to LB medium at 30 μg · mL−1 or to Terrific Broth (TB) medium at 50 μg · mL−1. When required, erythromycin was added at 100 μg · mL−1.

Bacterial strains and plasmids used in this study. Download Table S1, DOCX file, 0.02 MB.

Genetic constructs.

Primers used in this work are described in Table S2. All plasmid constructs were verified by sequencing.

Oligonucleotides used in this study. Download Table S2, DOCX file, 0.02 MB.

GldL and C-terminally truncated and twin-Strep-tagged GldM proteins (GldM′-TS/GldM″-TS) were expressed from vectors derived from the plasmid pT12 (29) under the control of rhamnose-inducible promoters.

Suicide vectors to genetically modify F. johnsoniae were produced using the vectors pRHJ012 (11) and pYT354 (30), then introduced into the F. johnsoniae ΔgldL strain Fl_082 (11) using Escherichia coli strain S17-1 (31) as previously described (11).

Full details of all genetic constructs are given in Text S1.

Supplemental methods. Download Text S1, DOCX file, 0.04 MB.

Purification of protein complexes.

Briefly, protein complexes were overexpressed in BL21(DE3) cells and then extracted from cell membranes using lauryl maltose neopentyl glycol (LMNG; Anatrace). Protein complexes were then affinity purified using Strep-Tactin XT resin (IBA) and then further purified using size exclusion chromatography. Full details of the purification scheme are given in Text S1.

Typical yields per liter of cell culture were as follows: 20 μg (CcaGldLM″); 50 μg (PgiPorLM′); 6 μg (SthGldLM′); 350 μg (SweGldLM′).

Cryo-EM sample preparation and imaging.

Aliquots (4 μL) of purified samples at an A280 of 1 were applied onto glow-discharged holey carbon-coated grids (Quantifoil 300 mesh, Au R1.2/1.3), then adsorbed for 10 s, blotted for 2 s at 100% humidity at 4°C, and plunge-frozen in liquid ethane using a Vitrobot Mark IV (FEI). To prepare samples with fluorinated octyl maltoside (fOM; Anatrace), proteins were concentrated to an A280 of 3 and 13.5 μL was mixed with 1.5 μL of 7 mM fOM in buffer W (100 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM EDTA) plus 0.01% LMNG. All samples were centrifuged at 18,400 × g for 10 min at 4°C immediately before grid preparation.

Data were collected using a Titan Krios G3 instrument (FEI) operated at 300 kV and fitted with either a GIF energy filter (Gatan) and a K2 Summit detector (Gatan) or a BioQuantum imaging filter (Gatan) and a K3 direct detection camera (Gatan). Full details of the data collection strategy are given in Text S1.

Cryo-EM data processing.

Motion correction, dose weighting, contrast transfer function determination, particle picking, and initial particle extraction were performed using SIMPLE 3.0 (32). Gold-standard Fourier shell correlations (FSC) using the 0.143 criterion, and local resolution estimations were calculated within RELION 3.1 (33).

In general, extracted particles were subjected to reference-free two-dimensional (2D) classification in SIMPLE, followed by 3D classification in RELION using either the previously solved FjoGldLM′ map (11) or another map produced in this study as a reference. Classes with clear secondary structure detail were selected and used for 3D autorefinement. Successive rounds of Bayesian particle polishing, 3D classification, and 3D autorefinement in RELION were used to generate the final maps for each data set. A full description of the data processing strategy for each data set is given in Text S1.

Model building and refinement.

The Phyre2 server was used to generate homology models for each new sequence from the structure of FjoGldLM′ (PDB no. 6SY8) using one-to-one threading (34). These models were rigid-body fitted into the cryo-EM volume using Coot, and residues were built de novo or removed as necessary in Coot (35). Rebuilding in globally sharpened and local-resolution-filtered maps was combined with real-space refinement in Phenix using secondary structure, rotamer, and Ramachandran restraints to give the final models described in Table 1 (36, 37). Validation was done in Molprobity (38). Structures were analyzed using ChimeraX (39), PyMOL 2.3.3 (Schrodinger), and the Consurf Server (40, 41).

Measurement of gliding motility on agar.

Strains were grown overnight in PY2 medium, washed once in PY2 medium, then resuspended in PY2 medium to an optical density at 600 nm (OD600) of 0.1. A 2-μL sample was then spotted onto PY2 agar plates. Plates were incubated at 25°C for 48 h before imaging with a Zeiss Axio Zoom MRm charge-coupled device (CCD) camera and Zeiss software (ZenPro 2012 v. 1.1.1.0).

Data availability.

The cryo-EM volumes and atomic coordinates presented in this paper have been deposited in the Electron Microscopy Data Bank (EMDB) and the Protein Data Bank (PDB), respectively, with the following accession codes: CcaGldLM″TMH (EMDB no. EMD-24959 and PDB no. 7SAZ), CcaGldLM″peri (EMDB no. EMD-24961 and PDB no. 7SB2), PgiPorLM′ (EMDB no. EMD-24956 and PDB no. 7SAT), SthGldLM′ (EMDB no. EMD-24957 and PDB no. 7SAU), and SweGldLM′ (EMDB no. EMD-24958 and PDB no. 7SAX).