Type 9 secretion system structures reveal a new protein transport mechanism

Abstract

The type 9 secretion system (T9SS) is the protein export pathway of bacteria of the Gram-negative Fibrobacteres-Chlorobi-Bacteroidetes superphylum and is an essential determinant of pathogenicity in severe periodontal disease. The central element of the T9SS is a so-far uncharacterized protein-conducting translocon located in the bacterial outer membrane. Here, using cryo-electron microscopy, we provide structural evidence that the translocon is the T9SS protein SprA. SprA forms an extremely large (36-strand) single polypeptide transmembrane β-barrel. The barrel pore is capped on the extracellular end, but has a lateral opening to the external membrane surface. Structures of SprA bound to different components of the T9SS show that partner proteins control access to the lateral opening and to the periplasmic end of the pore. Our results identify a protein transporter with a distinctive architecture that uses an alternating access mechanism in which the two ends of the protein-conducting channel are open at different times.

Extended Data Fig. 1. Phenotypic analysis of strains expressing Halotag and Twin-Strep-SprA fusion proteins.

a,b, Immunoblot analysis of (a) HaloTag-SprA and (b) Twin-Strep-SprA expression in whole cell lysates. Similar data were obtained for three biological repeats. c, Immunoblot detection of the T9SS-dependent chitinase ChiA levels in culture supernatants. Similar data were obtained for three biological repeats. d, T9SS-dependent spreading (gliding) morphology of colonies on agar. Scale bar, 5 mm. Similar data were obtained for three biological repeats. e, Peptide mass spectrometry of the three highest molecular mass bands in Fig. 1d. Intensity values are normalized to the most abundant SprA peptide detected for each band. Peptide numbering is from the N-terminus of the native SprA precursor sequence. For immunoblot source data see Supplementary Fig. 1.

Extended Data Fig. 2. Experimental quality and resolution estimation of SprA complexes.

a, Representative micrograph of SprA complexes. b, Selected reference-free 2D class averages. c, Gold-standard FSC curve of the final map calculated using a soft-edged mask. d, Local resolution estimates of the final maps. e, Representative density for SprA in the PorV complex (left) and Plug complex (right).

Extended Data Fig. 3. Structural analysis of the SprA complex components.

a, Structural alignment of SprA-bound proteins against the homology models used in initial sequence docking. b, Access routes to the SprA pore viewed (Top) from the periplasm or (Bottom) towards the lateral opening. In the PorV complex two loops involved in coordinating PorV occlude the lateral opening. The step domain is poorly ordered in the Plug complex. For clarity, PorV, Plug, and PPI are not shown. c, The PorV barrel is tilted relative to the SprA barrel. Aromatic residues on the surface of PorV are shown in green spacefill.

Extended Data Fig. 4. Phenotypes of SprA partner deletion strains.

a, Immunoblot analysis of Twin-Strep-SprA levels in whole cell lysates. Similar data were obtained for three biological repeats. For immunoblot source data see Supplementary Fig. 1. b, Quantification by liquid chromatography–mass spectrometry of T9SS substrates detected in cell culture supernatants according to CTD type. A detection threshold of >1% of the protein abundance in the sprA⁺ parental strain was applied. c, d, Heat map representation of secreted T9SS substrates from b with either (c) type A CTDs or (d) type B CTDs. Protein intensities for each protein are normalised to the level detected in the sprA⁺ parental strain. e, Measurement of vancomycin inhibition zones in a disc diffusion assay. The mean radius of inhibition (red bar) is measured from the disc edge. Error bars represent 1SD (n=4 for ΔporV and Δplug; n=5 for other strains) and statistical significance is indicated above each measurement set from a one-way ANOVA with post-hoc Dunnett's test using sprA⁺ as control group (n.s. = not significant, **** = p < 0.0001). Other comparisons (bracketed) use two tailed unpaired t tests *, p = 0.0134; ****, p < 0.0001). A control for the DMSO solvent used to dissolve the vancomycin is shown. a-e, All sprA⁺ strains are expressing the Twin-Strep-SprA fusion.

Extended Data Fig. 5. Single-particle cryo-EM image processing workflow for the SprA complexes.

Cryo-EM data sets of SprA complexes in LMNG, in the presence or absence of fluorinated detergents, were combined following 2D classification and subjected to 3D classification against a low resolution model generated from the fluorinated octyl-maltoside dataset. Particle images corresponding to the PorV complex or Plug complex were then independently refined. A soft mask, generated from these maps, was then used to perform masked refinement against the same particle images, resulting in global map resolutions of 3.5 Å for the PorV complex and 3.7 Å for the Plug complex.

Fig. 1. Characterization of SprA.

a, Localization of fluorophore-labelled SprA in F. johnsoniae cells. Exemplar data from experiments repeated on three different cultures with similar results. b, c, Population distributions of (b) SprA foci per cell and (c) SprA molecules per focus (from 119 photobleaching traces). d, Coomassie-stained SDS-PAGE gel of affinity-purified SprA with band identities from peptide mass spectrometry. Fjoh_0645 is a biotin-containing contaminant protein that binds to the streptavidin affinity matrix. Similar data were obtained for three independent preparations. For gel source data see Supplementary Fig. 1. e, EM density of the SprA complexes. Membrane position inferred from the location of the detergent micelles. f, Domain organisation of SprA. g, Overall structures of the SprA complexes. The same SprA domain colouring scheme is applied in e-g except in g far right panel (SprA-blue, PorV/Plug-grey).

Fig. 2. Structural features of SprA extracellular domains.

a, Structures of the SprA step, cap domains, and inserts from the PorV complex. b, Locations of the cap and step domains in the PorV complex. c, Insert 1 is inserted within SprA β-strand 28. d, The PPI active site contains a proline-containing SprA surface loop (red).

Fig. 3. Structural analysis of the SprA translocon complexes.

a, SprA barrel morphologies in the PorV (left) and Plug complexes (right) viewed from the periplasm and with cap domains cut away for clarity (as shown in insert). PorV and Plug, grey. b, Structural comparison of PorV with FadL. c, Surface conservation of SprA (left and centre, shown in the context of the PorV complex) and Plug (right, shown in the Plug complex) as determined by Consurf. d, Structural comparison of SprA with other large β-barrel bacterial outer membrane proteins.

Fig. 4. Biological consequences of removing SprA partner proteins

a-b, Secretome analysis of culture supernatants. a, Coomassie-stained SDS-PAGE gel. The major T9SS-secreted chitinase ChiA is highlighted (green boxing). Similar data were obtained for three biological repeats. For gel source data see Supplementary Fig. 1. b, T9SS substrates detected by proteomics at >1% of the protein abundance in strain sprA⁺, for which 30 Type A (TIGR04183 family) and 5 Type B (TIGR04131 family) CTD-dependent proteins were detected. c, Spreading (gliding) morphology of colonies on agar. Similar data were obtained for three biological repeats. d, Vancomycin sensitivity by disc diffusion assay. Mean radius of inhibition (mm) ± S.D. measured from the disc edge (n=4 for ΔporV and Δplug; n=5 otherwise), statistical significance by one-way ANOVA with post-hoc Dunnett's test using sprA⁺ as control group, n.s. = not significant. a-d, sprA⁺ strains express the Twin-Strep-SprA fusion. e, Model for the T9SS translocon mechanism.