In Situ Molecular Architecture of the Salmonella Type III Secretion Machine.

STAR* Methods Contact for Reagents and Resource Sharing

Contact for reagents and resource sharing should be directed to Dr. Jorge Galan ( iorge.galan@yale.edu ) and Dr. Jun Liu ( Jun.Liu.1@uth.tmc.edu ).

Experimental Model and Subject details Bacterial strains and plasmids

All bacterial strains used in this study are derived from Salmonella enterica serovar Typhimurium strain SL1344 ( Hoiseth and Stocker, 1981 ) and are listed in resource Table. All strains were constructed by standard recombinant DNA and allelic exchange procedures as previously described ( Kaniga et al., 1994 ). Method Details Analysis of type III protein secretion function The functionality of the type III secretion system in the different S . Typhimurium strains was carried out by examining their ability to secrete type III secreted proteins to the culture supernatant. Briefly, overnight cultures of the specific strains were diluted 1/20 into LB containing 0.3M NaCl to induce the expression of SPI-1 T3SS ( Galán and Curtiss III, 1990 ). Diluted cultures were grown at 37 ° C on a rotating wheel to an OD600 of ~ 0.9 (4 to 5 hours) and then the cells were pelleted by centrifugation at 6,000 rpm. The cell pellet was resuspended in 1X SDS-running buffer at 10X concentration so that 10 μl of the resuspension equaled 100 μl of cells. The culture supernatants were filtered through a 0.45 μm syringe filter and proteins in the supernatant were recovered by trichloroacetic acid (TCA) precipitation. The protein precipitate was resuspended in 1x SDS-running buffer at 100X concentration so that 10 μl of the resuspension equaled 1 ml of culture supernatant. Ten μl of whole cell lysate sample (100 μl of cells) and 10 μl of supernatant sample (1 ml of supernatant) were run on a 10% SDS-PAGE gel for Western blot analysis with antibodies against the type III secreted proteins SipB, SipC and InvJ.

Preparation of Frozen-Hydrated Specimens

Bacterial cultures were grown overnight at 37 ° C in LB containing 0.3M NaCl and fresh cultures were prepared from a 1:100 dilution and then grown at 37 ° C to late log phase in the presence of ampicillin (200 μg/mL) and L-arabinose (0.1%) to induce the expression of regulatory protein HilA and thus increase the number of injectisomes partitioning to the minicells ( Carleton et al., 2013 ). To enrich for minicells, the culture was centrifuged at 1,000 × g for 5 min to remove bacterial cells, and the supernatant fraction was further centrifuged at 20,000 × g for 20 min to collect the minicells. The minicell-enriched preparations were then mixed with 10 nm colloidal gold particles (used as fiducial markers in image alignment) and then deposited onto freshly glow-discharged, holey carbon grids for 1 min. The grids were blotted with filter paper and rapidly frozen in liquid ethane, using a gravity-driven plunger apparatus as described previously( Hu et al., 2015 ).

Cryo-ET Data Collection and 3D Reconstructions

The frozen-hydrated specimens were imaged at −170 ° C using a Polara G2 electron microscope (FEI Company) equipped with a field emission gun and a direct detection device (Gatan K2 Summit). The microscope was operated at 300 kV with a magnification of ×15,500, resulting in an effective pixel size of 2.6 Å at the specimen level. We used SerialEM ( Mastronarde, 2005 ) to collect low-dose, single-axis tilt series with dose fractionation mode at about 5 μm defocus and a cumulative dose of ~50 e − /Å 2 distributed over 35 stacks covering an angular range of −51 ° to +51 ° with 3 ° fixed increments. Each stack contains ~8 images. To analyze over 60 TB raw data from the microscope and the direct detection device, we used Tomoauto ( Morado et al., 2016 ) to facilitate image processing: drift correction of dose-fractionated data using Motioncorr ( Li et al., 2013 ) assembly of corrected sums into tilt series, automatic fiducial seed model generation, alignment, defocus estimation, and contrast transfer function correction of tilt series using IMOD ( Kremer et al., 1996 ), and weighted back projection (WBP) reconstruction of tilt series into tomograms using Tomo3D ( Agulleiro and Fernandez, 2015 ). Each tomographic reconstruction is 3,710 × 3,838 × 1,800 voxels and ~100Gb in size. In total, 5,592 tomographic reconstructions (about 600 TB data) from 17 different strains were successfully generated and were then utilized for the subsequent sub-tomogram analysis ( Table S1 ). The original WBP tomograms were too noisy for direct visualization of cellular features. Therefore, we also used Tomo3D ( Agulleiro and Fernandez, 2015 ) to generate high contrast tomograms (618 × 639 × 300 voxels) from the binned by 6 aligned tilt series by simultaneous iterative reconstruction technique (SIRT). Some representative snapshots of the SIRT reconstructions from 16 strains are shown in Fig. S1 .

Sub-tomogram averaging and correspondence analysis

We used tomographic package I3 ( Winkler, 2007 ) for sub-tomogram analysis as described previously ( Hu et al., 2015 ). A total of 29,307 sub-tomograms of the injectisomes (400 × 400 × 400 voxels) were visually identified in the SIRT reconstructions and then extracted from 5,592 the WBP reconstructions (3,710 × 3,838 × 1,800 voxels) of the minicells. Two of the three Euler angles of each injectisome were estimated based on the orientation of each particle in the cell envelope. To accelerate image analysis, 4 × 4 × 4 binned sub-tomograms (100 × 100 × 100 voxels) were used for initial alignment and classification. The alignment proceeded iteratively with each iteration consisting of three parts in which references and classification masks are generated, sub-tomograms are aligned and classified, and finally class averages are aligned to each other. Class averages showed similar structural features: the bacterial envelope-associated needle complex, the inner membrane export apparatus, and the large cytoplasmic sorting platform. After multiple cycles of alignment and classification for 4 × 4 × 4 binned subtomograms, we used original unpinned sub-tomograms for refinement.

Fourier shell correlation

(FSC) between the two independent reconstructions was used to estimate the resolution of the averaged structures ( Fig. S1 ). The final maps have been deposited in the Electron Microscopy Data Bank (EMDB) with accession codes EMD-8544 and EMD-8545.

3D Visualization and Molecular Modeling

We used IMOD to visualize the maps and to generate 3D surface rendering of Salmonella minicells and UCSF Chimera ( Pettersen et al., 2004 ) ( http://www.rbvi.ucsf.edu/chimera ) to visualize sub-tomogram averages in 3D and molecular modeling. To better visualize the tag densities, we used the difference maps between the injectisome structure with specific tags and the wild type injectisome structure, and then used UCSF Chimera for segmentation and surface rendering. The EM map of the purified NC from Salmonella (EMD-1875) was fitted into our intact injectisome map using the function “fit in map” in UCSF Chimera ( Pettersen et al., 2004 ). We built the initial model based on the following refined structures from S. Typhimurium: InvG N (the amino-terminal domain of InvG: PDB-3J1V) ( Bergeron et al., 2013 ), PrgH C (the carboxyterminal domain of PrgH: PDB-3J1X) ( Bergeron et al., 2013 ), and PrgH N (the amino-terminal domain of PrgH: PDB-3J1W) ( Bergeron et al., 2013 ). A large remodeling was required to refine PrgH C . InvA C (the carboxy-terminal domain of InvA: PDB-2×4a) ( Worrall et al., 2010 ) was used to build the nonameric ring based on the homologous structure from Shigella flexneri MxiA C (PDB-4A5P) ( Abrusci et al., 2013 ). Structures of the flagellar ATPase complex FliI–FliH (PDB-5B0O) ( Imada et al., 2016 ) and FliJ (PDB-3AJW) ( Ibuki et al., 2011 ) were used to build the model of the InvC–InvI–OrgB complex. Modeling of the InvA nonameric ring We built the InvA C nonameric ring using the MxiA C nonameric ring as a template ( Fig. S3 ), and then we fitted the modeled structure into our intact injectisome map using the function “fit in map” in UCSF Chimera. Comparison of the protein densities in sub-tomograms of minicells obtained from the wild type strain with those obtained from a strain expressing a GFP-tagged InvA identified a new density at the bottom of the toroidal-shape density corresponding to the protein tag confirming the localization of InvA within this structure ( Fig. 3 and Fig. S3 ). Furthermore, since the GFP tag was placed at the carboxy-terminus of InvA, the location of the extra density confirms the orientation of the InvA C nonameric ring, which guided the placement of the atomic structure into the cryo-ET map ( Fig. S3 ). Modeling of InvG, PrgK and PrgH The protein density map of the purified S. Typhimurium NC (EMD-1875) ( Schraidt and Marlovits, 2011 ) was fitted into our intact injectisome map using the function “fit in map” in UCSF Chimera. The major structural scaffold of the NC base is comprised of 15 copies of InvG, which form the neck and outer rings, and 24 copies each of PrgH and PrgK, which are arranged in a concentric fashion and form the inner rings. Secondary structure prediction analysis indicates that PrgH contains a transmembrane domain (from amino acid 142 to 162), which separates the protein into two soluble domains, the amino-terminal domain located in the cytoplasm and the carboxy-terminal domain located in the periplasm ( Fig. S6 ). We built the initial model of the NC base using the map of the purified S. Typhimurium NC (EMD-1875), and the atomic structures of InvG (PDB-3J1V), the carboxy-terminal domain of PrgH (PDB-3J1X), and the amino-termimal domain of PrgH (PDB-3J1W) ( Bergeron et al., 2013 ). Modeling of InvC and OrgB Comparison of tomograms obtained from a S. Typhimurium Δ invC mutant with those of wild type identified a missing density located within the central nave-like hub that connects the six spokes of the wheel-like structure that caps the sorting platform on the cytoplasmic side ( Fig. 4 ). This density most likely corresponds to InvC. To confirm the location and orientation of InvC, we imaged a S. Typhimurium strain expressing InvC tagged at its carboxy-terminus by GFP, which when compared with wild type showed an additional density (presumably corresponding to the GFP tag) on the membrane-facing side of nave-like hub. This observation suggests that the carboxy-terminus of InvC faces the export apparatus components such as the carboxy-terminus of InvA ( Fig. 4 ). In the flagellar ATPase complex, FliH C2 (a homolog of OrgB) shows an unusually asymmetric homodimeric structure that binds to the amino-terminal region of the ATPase FliI (a homolog of InvC) ( Imada et al., 2016 ). A hexameric ring model of the FliH C2 -FliI complex has been built previously using the hexamer model of V-type ATPase as a template ( Imada et al., 2016 ). Using this model as well as the partial structure of the OrgB-SpaO complex ( Notti et al., 2015 ) we built a model in which the hexameric InvC ATPase fits the nave-like hub density we observed in our sub-tomogram averages, and the structure of OrgB C2 fits well into the spokes of the wheel-like structure that cradles the InvC hexamer. The model places the Cterminal domain of InvC facing the export apparatus, and the carboxy-sand amino-terminal domains of OrgB interacting with InvC and SpaO, respectively. The model provides further support to the proposed location of OrgB and InvC. Modeling of InvI InvI is a small coiled-coil protein similar to the F1-γ subunit and flagellar protein FliJ ( Ibuki et al., 2011 ). FliJ binds in the central pore of the FliI 6 ring (the homolog of InvC) to form the FliI 6 FliJ complex, which resembles the F1-α3β3γ complex. To gain insight into the potential location of InvI we compared sub-tomogram averages obtained from a strain that expresses InvI tagged at its amino-terminus by GFP with those of wild type. After density subtraction we detected an additional density, presumably corresponding to GFP, located at the center of the export-apparatus-facing side of the proposed InvC hexameric ring.

Quantification and Statistical Analysis

Quantification and statistical analyses are integral parts of the algorithms and software used in our high throughput cryo-electron tomography pipeline. In particular, massive data enabled us to use multivariate statistical analysis and classification (which are implemented in tomographic package i3) for processing and interpretation of the sub-tomograms of injectisomes extracted from Salmonella minicell reconstructions.

Data and Software Availability

All software used in this study have been extensively described in previous publications from our and other laboratories. See the Methods Details section for citations to the original publications. All data are available upon request. The final sub-tomogram averages have been deposited in the Electron Microscopy Data Bank (EMDB) with the accession codes EMD-8544 and EMD-8545.

Experimental Model and Subject details Bacterial strains and plasmids

All bacterial strains used in this study are derived from Salmonella enterica serovar Typhimurium strain SL1344 ( Hoiseth and Stocker, 1981 ) and are listed in resource Table. All strains were constructed by standard recombinant DNA and allelic exchange procedures as previously described ( Kaniga et al., 1994 ).

Method Details Analysis of type III protein secretion function The functionality of the type III secretion system in the different S . Typhimurium strains was carried out by examining their ability to secrete type III secreted proteins to the culture supernatant. Briefly, overnight cultures of the specific strains were diluted 1/20 into LB containing 0.3M NaCl to induce the expression of SPI-1 T3SS ( Galán and Curtiss III, 1990 ). Diluted cultures were grown at 37 ° C on a rotating wheel to an OD600 of ~ 0.9 (4 to 5 hours) and then the cells were pelleted by centrifugation at 6,000 rpm. The cell pellet was resuspended in 1X SDS-running buffer at 10X concentration so that 10 μl of the resuspension equaled 100 μl of cells. The culture supernatants were filtered through a 0.45 μm syringe filter and proteins in the supernatant were recovered by trichloroacetic acid (TCA) precipitation. The protein precipitate was resuspended in 1x SDS-running buffer at 100X concentration so that 10 μl of the resuspension equaled 1 ml of culture supernatant. Ten μl of whole cell lysate sample (100 μl of cells) and 10 μl of supernatant sample (1 ml of supernatant) were run on a 10% SDS-PAGE gel for Western blot analysis with antibodies against the type III secreted proteins SipB, SipC and InvJ.

Preparation of Frozen-Hydrated Specimens

Bacterial cultures were grown overnight at 37 ° C in LB containing 0.3M NaCl and fresh cultures were prepared from a 1:100 dilution and then grown at 37 ° C to late log phase in the presence of ampicillin (200 μg/mL) and L-arabinose (0.1%) to induce the expression of regulatory protein HilA and thus increase the number of injectisomes partitioning to the minicells ( Carleton et al., 2013 ). To enrich for minicells, the culture was centrifuged at 1,000 × g for 5 min to remove bacterial cells, and the supernatant fraction was further centrifuged at 20,000 × g for 20 min to collect the minicells. The minicell-enriched preparations were then mixed with 10 nm colloidal gold particles (used as fiducial markers in image alignment) and then deposited onto freshly glow-discharged, holey carbon grids for 1 min. The grids were blotted with filter paper and rapidly frozen in liquid ethane, using a gravity-driven plunger apparatus as described previously( Hu et al., 2015 ).

Cryo-ET Data Collection and 3D Reconstructions

The frozen-hydrated specimens were imaged at −170 ° C using a Polara G2 electron microscope (FEI Company) equipped with a field emission gun and a direct detection device (Gatan K2 Summit). The microscope was operated at 300 kV with a magnification of ×15,500, resulting in an effective pixel size of 2.6 Å at the specimen level. We used SerialEM ( Mastronarde, 2005 ) to collect low-dose, single-axis tilt series with dose fractionation mode at about 5 μm defocus and a cumulative dose of ~50 e − /Å 2 distributed over 35 stacks covering an angular range of −51 ° to +51 ° with 3 ° fixed increments. Each stack contains ~8 images. To analyze over 60 TB raw data from the microscope and the direct detection device, we used Tomoauto ( Morado et al., 2016 ) to facilitate image processing: drift correction of dose-fractionated data using Motioncorr ( Li et al., 2013 ) assembly of corrected sums into tilt series, automatic fiducial seed model generation, alignment, defocus estimation, and contrast transfer function correction of tilt series using IMOD ( Kremer et al., 1996 ), and weighted back projection (WBP) reconstruction of tilt series into tomograms using Tomo3D ( Agulleiro and Fernandez, 2015 ). Each tomographic reconstruction is 3,710 × 3,838 × 1,800 voxels and ~100Gb in size. In total, 5,592 tomographic reconstructions (about 600 TB data) from 17 different strains were successfully generated and were then utilized for the subsequent sub-tomogram analysis ( Table S1 ). The original WBP tomograms were too noisy for direct visualization of cellular features. Therefore, we also used Tomo3D ( Agulleiro and Fernandez, 2015 ) to generate high contrast tomograms (618 × 639 × 300 voxels) from the binned by 6 aligned tilt series by simultaneous iterative reconstruction technique (SIRT). Some representative snapshots of the SIRT reconstructions from 16 strains are shown in Fig. S1 .

Sub-tomogram averaging and correspondence analysis

We used tomographic package I3 ( Winkler, 2007 ) for sub-tomogram analysis as described previously ( Hu et al., 2015 ). A total of 29,307 sub-tomograms of the injectisomes (400 × 400 × 400 voxels) were visually identified in the SIRT reconstructions and then extracted from 5,592 the WBP reconstructions (3,710 × 3,838 × 1,800 voxels) of the minicells. Two of the three Euler angles of each injectisome were estimated based on the orientation of each particle in the cell envelope. To accelerate image analysis, 4 × 4 × 4 binned sub-tomograms (100 × 100 × 100 voxels) were used for initial alignment and classification. The alignment proceeded iteratively with each iteration consisting of three parts in which references and classification masks are generated, sub-tomograms are aligned and classified, and finally class averages are aligned to each other. Class averages showed similar structural features: the bacterial envelope-associated needle complex, the inner membrane export apparatus, and the large cytoplasmic sorting platform. After multiple cycles of alignment and classification for 4 × 4 × 4 binned subtomograms, we used original unpinned sub-tomograms for refinement.

Fourier shell correlation

(FSC) between the two independent reconstructions was used to estimate the resolution of the averaged structures ( Fig. S1 ). The final maps have been deposited in the Electron Microscopy Data Bank (EMDB) with accession codes EMD-8544 and EMD-8545.

3D Visualization and Molecular Modeling

We used IMOD to visualize the maps and to generate 3D surface rendering of Salmonella minicells and UCSF Chimera ( Pettersen et al., 2004 ) ( http://www.rbvi.ucsf.edu/chimera ) to visualize sub-tomogram averages in 3D and molecular modeling. To better visualize the tag densities, we used the difference maps between the injectisome structure with specific tags and the wild type injectisome structure, and then used UCSF Chimera for segmentation and surface rendering. The EM map of the purified NC from Salmonella (EMD-1875) was fitted into our intact injectisome map using the function “fit in map” in UCSF Chimera ( Pettersen et al., 2004 ). We built the initial model based on the following refined structures from S. Typhimurium: InvG N (the amino-terminal domain of InvG: PDB-3J1V) ( Bergeron et al., 2013 ), PrgH C (the carboxyterminal domain of PrgH: PDB-3J1X) ( Bergeron et al., 2013 ), and PrgH N (the amino-terminal domain of PrgH: PDB-3J1W) ( Bergeron et al., 2013 ). A large remodeling was required to refine PrgH C . InvA C (the carboxy-terminal domain of InvA: PDB-2×4a) ( Worrall et al., 2010 ) was used to build the nonameric ring based on the homologous structure from Shigella flexneri MxiA C (PDB-4A5P) ( Abrusci et al., 2013 ). Structures of the flagellar ATPase complex FliI–FliH (PDB-5B0O) ( Imada et al., 2016 ) and FliJ (PDB-3AJW) ( Ibuki et al., 2011 ) were used to build the model of the InvC–InvI–OrgB complex. Modeling of the InvA nonameric ring We built the InvA C nonameric ring using the MxiA C nonameric ring as a template ( Fig. S3 ), and then we fitted the modeled structure into our intact injectisome map using the function “fit in map” in UCSF Chimera. Comparison of the protein densities in sub-tomograms of minicells obtained from the wild type strain with those obtained from a strain expressing a GFP-tagged InvA identified a new density at the bottom of the toroidal-shape density corresponding to the protein tag confirming the localization of InvA within this structure ( Fig. 3 and Fig. S3 ). Furthermore, since the GFP tag was placed at the carboxy-terminus of InvA, the location of the extra density confirms the orientation of the InvA C nonameric ring, which guided the placement of the atomic structure into the cryo-ET map ( Fig. S3 ). Modeling of InvG, PrgK and PrgH The protein density map of the purified S. Typhimurium NC (EMD-1875) ( Schraidt and Marlovits, 2011 ) was fitted into our intact injectisome map using the function “fit in map” in UCSF Chimera. The major structural scaffold of the NC base is comprised of 15 copies of InvG, which form the neck and outer rings, and 24 copies each of PrgH and PrgK, which are arranged in a concentric fashion and form the inner rings. Secondary structure prediction analysis indicates that PrgH contains a transmembrane domain (from amino acid 142 to 162), which separates the protein into two soluble domains, the amino-terminal domain located in the cytoplasm and the carboxy-terminal domain located in the periplasm ( Fig. S6 ). We built the initial model of the NC base using the map of the purified S. Typhimurium NC (EMD-1875), and the atomic structures of InvG (PDB-3J1V), the carboxy-terminal domain of PrgH (PDB-3J1X), and the amino-termimal domain of PrgH (PDB-3J1W) ( Bergeron et al., 2013 ). Modeling of InvC and OrgB Comparison of tomograms obtained from a S. Typhimurium Δ invC mutant with those of wild type identified a missing density located within the central nave-like hub that connects the six spokes of the wheel-like structure that caps the sorting platform on the cytoplasmic side ( Fig. 4 ). This density most likely corresponds to InvC. To confirm the location and orientation of InvC, we imaged a S. Typhimurium strain expressing InvC tagged at its carboxy-terminus by GFP, which when compared with wild type showed an additional density (presumably corresponding to the GFP tag) on the membrane-facing side of nave-like hub. This observation suggests that the carboxy-terminus of InvC faces the export apparatus components such as the carboxy-terminus of InvA ( Fig. 4 ). In the flagellar ATPase complex, FliH C2 (a homolog of OrgB) shows an unusually asymmetric homodimeric structure that binds to the amino-terminal region of the ATPase FliI (a homolog of InvC) ( Imada et al., 2016 ). A hexameric ring model of the FliH C2 -FliI complex has been built previously using the hexamer model of V-type ATPase as a template ( Imada et al., 2016 ). Using this model as well as the partial structure of the OrgB-SpaO complex ( Notti et al., 2015 ) we built a model in which the hexameric InvC ATPase fits the nave-like hub density we observed in our sub-tomogram averages, and the structure of OrgB C2 fits well into the spokes of the wheel-like structure that cradles the InvC hexamer. The model places the Cterminal domain of InvC facing the export apparatus, and the carboxy-sand amino-terminal domains of OrgB interacting with InvC and SpaO, respectively. The model provides further support to the proposed location of OrgB and InvC. Modeling of InvI InvI is a small coiled-coil protein similar to the F1-γ subunit and flagellar protein FliJ ( Ibuki et al., 2011 ). FliJ binds in the central pore of the FliI 6 ring (the homolog of InvC) to form the FliI 6 FliJ complex, which resembles the F1-α3β3γ complex. To gain insight into the potential location of InvI we compared sub-tomogram averages obtained from a strain that expresses InvI tagged at its amino-terminus by GFP with those of wild type. After density subtraction we detected an additional density, presumably corresponding to GFP, located at the center of the export-apparatus-facing side of the proposed InvC hexameric ring.

Supplementary Material 1 Supplementary Figure S1 related to Fig. 1. Gallery of 2D snapshots from representative tomographic reconstructions of different S. Typhimurium strains. For each strain, a tomographic slice shows an overall image of a minicell in the left panel, and a corresponding zoom-in view of the injectisomes embedded in the cell envelope in the right panel. Noticeably, some cells have injectisomes with the needle substructure while others do not. The resolution of the average structure of the intact T3SS injectisome in situ is 17 Å as estimated by Fourier shell correlation (0.5 cutoff). Supplementary Figure S2 related to Fig. 3 and 4.

Analysis of type

III secretion function in the different S. Typhimurium strains used in this study. Cultured supernatants and cell lysates of the indicated strains were analyzed by Western immunoblotting for the presence of the indicated type III secreted proteins. Supplementary Figure S3 related to Fig. 3. Localization of InvA in the intact injectisome and in situ model of the InvA C nonameric ring in intact injectisomes.. ( A ) A central section of the wild-type injectisome structure. The densities underneath the inner membrane (IM) are highlighted by cyan and purple arrows. ( B ) Injectisome structure obtained from a S. Typhimurium strain expressing a sfGFP-tagged version of InvA shows extra densities (indicated with green arrows), which are clearly visible in the difference map ( C ). The extra densities most likely correspond to the sfGFP tags. ( D ) Injectisome structure obtained from a S. Typhimurium mutant strain lacking spaO (a central component of the sorting platform) and invA; the large portion of the cytoplasmic complex and the densities shown by purple arrows in panel ( A ) are absent. In all strains, the densities highlighted by cyan arrows (most likely corresponding to IR2) remain. ( E ) The model of the InvA C nonameric ring viewed in two orientations. ( F ) Fitting of the ring model into the toroidal-shaped density (assigned to InvA) located underneath the inner membrane. Modeling was carried out with the “fit in map” function of UCSF Chimera as indicated in the Methods. Views from the top (top panel) and from the bottom (lower panel) are shown and the first and last residues of InvA C are indicated in blue and red, respectively. ( G ) Side view of the injectisome depicting the location of InvA C nonameric ring. Supplementary Figure S4 related to Fig. 1 and 2. Comparison of injectisome and flagellar structures . Comparison of the Salmonella T3SS structure reported here (A–D) with the Shigella T3SS structure (E–H) and the Salmonella flagellar motor (I and J). Note that the Salmonella T3SS structure shows significantly more details particularly underneath the inner membrane (IM) revealing the six-patch organization of the IR2 (D), which is not visible in the Shigella structure (H). The T3SS sorting platform appears strikingly different from the flagellar C ring (I,J). Supplementary Figure S5 related to Fig. 4 and 5. Localization of key components (OrgB and SpaO) of the sorting platform and the cytoplasmic domain of PrgH in the intact injectisome. ( A and B ) Cross-sections of the sorting platforms derived from the wild-type, the mutant with GFP tags on SpaO or OrgB, and the difference map, respectively. The extra densities in the tagged proteins are indicated with yellow arrows. ( C ) A central section of the wild-type injectisome structure. The densities underneath the inner membrane (IM) are highlighted by cyan arrows. ( D ) With the GFP tag, extra densities (shown in green arrows) are evident. ( E ) In the spaO deletion mutant, the cytoplasmic sorting platform is absent. ( F ) Without SpaO, the GFP tags (green arrows) are also linked to the densities (shown in cyan arrows). Supplementary Figure S6 related to Fig. 5. Remodeling of the PrgH C ring upon assembly of the sorting platform. Secondary structure prediction analysis indicates that PrgH contains a transmembrane domain (from amino acid 142 to 162) connecting two soluble domains: the N-sterminal cytoplasmic domain (PrgH N ) and the C-terminal periplasmic domain (PrgH C ) (top panel). An initial model of PrgH was built based on the atomic structures of its soluble domains and the cryoEM map of the purified needle complex as indicated in Materials and Methods. We fitted the protein density map of the purified needle complex and its associated atomic model into the intact injectisome in situ map ( 1 ), which shows that PrgH N ring would be completely embedded in the inner membrane, a location incompatible with the topology of this component of the needle complex. We therefore shifted the PrgH N ring as a rigid body ~6 nm to relocate it immediately underneath the inner membrane ( 2 ). Although the 24-unit ring model does not fit into the six patches observed in the cryo-ET map of the in situ structure, it accommodates well a four-unit PrgH N module into each one of the six patches totaling 24 PrgH N subunits.

Supplementary Table

S1 related to Fig. 1, 3, 4, and 5 . Number of tomograms of minicells and sub-tomograms of injectisomes processed in this study Movie S1 related to Fig. 1 . Cryo-electron tomography of Salmonella minicells reveals intact injectisomes in situ. Movie S2 realted to Fig. 3 and 4 . Localization of key components of the type III protein secretion injectisome by integrating specific protein tags and high-throughput cryo-selectron tomography. The position of the protein tags are depicted in green and the location of the carboxy-terminus of InvA is depicted in red. Movie S3 related to Fig. 5 and 6 . Molecular architecture of the entire type III protein secretion injectisome in situ depicting the remodeling of the secretion machine during its assembly. 2 3 4