Structural and Functional Analysis of SsaV Cytoplasmic Domain and Variable Linker States in the Context of the InvA-SsaV Chimeric Protein

ABSTRACT The type III secretion (T3S) injectisome is a syringe-like protein-delivery nanomachine widely utilized by Gram-negative bacteria. It can deliver effector proteins directly from bacteria into eukaryotic host cells, which is crucial for the bacterial–host interaction. Intracellular pathogen Salmonella enterica serovar Typhimurium encodes two sets of T3S injectisomes from Salmonella pathogenicity islands 1 and 2 (SPI-1 and SPI-2), which are critical for its host invasion and intracellular survival, respectively. The inner membrane export gate protein, SctV (InvA in SPI-1 and SsaV in SPI-2), is the largest component of the injectisome and is essential for assembly and function of T3SS. Here, we report the 2.11 Å cryo-EM structure of the SsaV cytoplasmic domain (SsaVC) in the context of a full-length SctV chimera consisting of the transmembrane region of InvA, the linker of SsaV (SsaVL) and SsaVC. The structural analysis shows that SsaVC exists in a semi-open state and SsaVL exhibits two major orientations, implying a highly dynamic process of SsaV for the substrate selection and secretion in a full-length context. A biochemical assay indicates that SsaVL plays an essential role in maintaining the nonameric state of SsaV. This study offers near atomic-level insights into how SsaVC and SsaVL facilitate the assembly and function of SsaV and may lead to the development of potential anti-virulence therapeutics against T3SS-mediated bacterial infection. IMPORTANCE Type III secretion system (T3SS) is a multicomponent nanomachine and a critical virulence factor for a wide range of Gram-negative bacterial pathogens. It can deliver numbers of effectors into the host cell to facilitate the bacterial host infection. Export gate protein SctV, as one of the engines of T3SS, is at the center of T3SS assembly and function. In this study, we show the high-resolution atomic structure of the cytosolic domain of SctV in the nonameric state with variable linker conformations. Our first observation of conformational changes of the linker region of SctV and the semi-open state of the cytosolic domain of SctV in the full-length context further support that the substrate selection and secretion process of SctV is highly dynamic. These findings have important implications for the development of therapeutic strategies targeting SctV to combat T3SS-mediated bacterial infection.

(SctV C ), which are connected via a ;20-40 amino acid linker, SctV L , in most bacteria using the T3SS (12). It has been reported that SctV is one of the "engines" of T3SS (25); the TM domain forms a putative proton channel, and SctV C forms a nonameric ring connecting to SctN (ATPase) through SctO (the stalk protein) to function as the F 0 F 1 -ATPase, coupling energy from ATP hydrolysis and the proton-motive force to secrete unfolded bacterial effectors into the eukaryotic host (26)(27)(28)(29). SctV C is also involved in substrate selection through recognizing gatekeeper proteins or different effector-chaperone pairs (15,(30)(31)(32)(33)(34). A previous study showed that substituting homologous TM and cytoplasmic domains between some SctV proteins caused them to retain their functions, but that SctV C controls substrate specificity (35).
Intensive structural and functional studies of T3SS have uncovered much detailed structural information and potential assembly processes of this complicated molecular machine (12,(36)(37)(38)(39)(40). However, the structure and molecular mechanism of SctV are mostly unknown due to the challenges of obtaining the fully assembled state of the full-length protein. Several studies have isolated the entire T3S injectisome and flagellar basal body for structural studies (12,(41)(42)(43). However, SctV was missing in all these trials, even though all other export apparatus components could be captured; this suggests a loose interaction between SctV and other components of the export apparatus and the basal body. The structure of SctV C has been identified as a homo-nonamer through crystal and cryo-electron microscopy (cryo-EM) structural studies, and the intermolecular polar interactions between monomers are thought to be the leading force maintaining the SctV C oligomeric state (26,(44)(45)(46). Recently, an in situ cryo-electron tomography (cryo-ET) study identified the location of InvA in the bacterial inner membrane and showed a high-order oligomeric state of the TM domain of InvA (12). More recently, Matthews-Palmer et al. and Kuhlen et al. obtained assembled full-length SctV and FlhA suitable for cryo-EM structural studies (47,48). However, due to the structural flexibility issues, both groups only determined the structure of SctV C and FlhA C , leaving the structure of linker region and the TM domain still unknown.
In this study, through generating a chimeric protein consisting of the TM region of InvA and the cytoplasmic region of SsaV, we produced the full-length SctV in a highorder oligomeric state which was adequate for structural study using cryo-EM single particle analysis. Here, we present the 2.11 Å nonameric ring structure of SsaV C with a semi-open state of each monomer. Moreover, we display the cryo-EM structure of SsaV L through the single particle analysis for the first time although with low resolution. Interestingly, SsaV L exhibits two major orientations, consistent with previous reports that the conformations of the SctV linker region could be altered during the secretion cycle (49,50). Structural analysis and biochemical assays indicate that SsaV L plays an essential role in maintaining the nonameric state of SsaV. Collectively, our data provide an atomic view and mechanistic understanding of how the cytoplasmic domain and linker region of SctV facilitate its assembly and function.

RESULTS
SsaV C forms a stable homo-nonameric ring. To determine the structure of SctV, we initially tried to purify the full-length InvA protein from Salmonella SPI-1 T3SS. However, we were unable to obtain a stably-assembled InvA sample for the cryo-EM structural study, even after several rounds of high-throughput detergent screening. It has been shown that SctV C makes a significant contribution in maintaining the SctV nonamer (46). Previous structural studies suggested that InvA C tends to be a monomer in solution (51). Therefore, it may be challenging for the full-length InvA to form the stable nonamer outside of the membrane. To obtain a fully assembled SctV sample, we first tested the nonamerization ability of SsaV C , the homologous protein of InvA C from Salmonella SPI-2 T3SS. Unlike InvA C , SsaV C could maintain a very stable high oligomeric state even in the high salt concentration analyzed by the size exclusion chromatography (SEC; Fig. 1B) and dynamic light scattering (DLS; Fig. 1C). Through a cryo-EM single particle analysis approach, we reconstructed and classified two conformations of SsaV C : one 3.55 Å cryo-EM structure with double stacked nonameric rings with D9 symmetry, and the other 3.64 Å cryo-EM structure with a single nonameric ring with C9 symmetry, the particle number of which are approximately comparable ( Fig. 1D and E and Fig. S1 in the supplemental material). The double-layer ring conformation of SsaV C is considered to be an artifact, consistent with a previous study (45). Notably, most of the C9 single-layer ring of the SsaV C particles were calculated from top views, and most of the D9 double-layer ring of SsaV C particles came from side views. We speculate that this is because the top-view protein particles are more vulnerable to air-water interface damage than the side-view particles (52). Together, unlike InvA C , SsaV C could form the stable homo-nonameric ring in solution.
InvA TM -SsaV C chimeric full-length protein can assemble into a nonamer. To test if the full-length SsaV and the TM domain of InvA could form the stable nonamer in solution with SsaV C , we constructed the full-length SsaV protein and the chimeric fulllength protein InvA TM -SsaV L -SsaV C (ISS; Fig. 2A). We performed a high-throughput detergent screening to identify suitable detergents to help stabilize the assembled state of SsaV and ISS. However, very few detergents could generate protein samples of high enough quality for cryo-EM structural study. A homogeneous and oligomerized ISS protein was eventually obtained in buffer containing the detergent Glyco-diosgenin (GDN; Fig. 2B); however, SsaV (similarly to InvA) could not form the stable nonamer in this condition (Fig. 2B). Compared with full-length SsaV and InvA, GDN-solubilized ISS protein showed better SEC and cryo-EM micrograph behaviors ( Fig. 2B and C) and were suitable for cryo-EM data processing.
Further 2D classification and 3D reconstruction of ISS showed the apparent density of the SsaV C nonameric ring, which is consistent with structural features of the SsaV C domain described above (Fig. 2C and D and Fig. S2). We reconstructed a high-resolution cryo-EM structure of SsaV C with an averaged resolution of 2.11 Å (Fig. 2D and Fig. S2). Unlike SsaV C , for which we obtained high-resolution structural information, the EM density of InvA TM is blurry, and the SsaV L EM density is absent, with an approximately constant distance between the cytoplasmic domain and TM domain; this is consistent with a recent study (47). In summary, we found that the GDN-solubilized InvA TM -SsaV C chimeric full-length protein tended to assemble into a nonamer in solution.
Structure of SsaV C . Based on the 2.11 Å cryo-EM map of ISS solved above ( Fig. 2D and Fig. S2), we built an atomic model of the SsaV C nonamer by referring to the SsaV C monomer structure (PDB: 7AWA) solved recently (47) (Fig. 2E, Fig. 3A, Fig. S3, and Table S1 in the supplemental material). The high structural similarity between these two SsaV C monomer structures (Ca RMSDs: 1.3598 Å) indicates that they hold a consistent structural conformation in the context of different TM regions (Fig. S4). Consistent with other reported SctV C structures (26,44,45,47,50,53), the SsaV C monomer also has a four-subdomain (SD) structure and further assembles to a nonameric ring aligned to SD3, which is the most conserved region of SsaV C (Fig. 3A to D). The diameter of the channel at the center of the SsaV C nonamer ranges from ;54-41 Å from the cytosolic face to the TM face (Fig. 3D). This channel further connects to the channel formed by other components of the export apparatus (SctRST) and finally extends to the needle conduit with a diameter of ;25 Å (54). Through this reverse funnel-like channel, T3SS systematically unfolds effector proteins and secretes them out of the bacteria. During the effector secretion process, SsaV C was reported to involve substrate selection with the cleft between SD2 and SD4 (13,32). The dynamic conformations between SD2 and SD4 have been demonstrated through both different SsaV C structures and molecular dynamic simulation, showing that SD2 and SD4 can alternate between open and closed states hinging around the rigid SD3 (13,26,44,45,47).
Consistent with the observation above, B-factor analysis of the SsaV C structure also shows that SD3 is the most stable region of SsaV C with the more flexible SD2 and SD4 floating around it (Fig. 3E). Structural comparison between SsaV C and InvA C (51) in a closed conformation or its counterpart in flagellum, FlhA C (55), in an open conformation shows that SsaV C in a full-length context presents a semi-open conformation ( Fig. 3F and G). Recent structural study revealed that the YscV C (the homolog of SsaV from Yersinia enterocolitica) and FlhA C in a full-length context present in the open state (48). Comparison of our SsaV C structure with these two reported homologous structures also exhibits the different conformations between SD2 and SD4 ( Fig. S5 in the supplemental material). Together, these findings firstly show the dynamic conformations between SD2 and SD4 of SctV in a full-length context.
The essential roles of linker region for the structural stability and function of SsaV. To gain insights into the molecular mechanism of SsaV C nonamerization, we analyzed the electrostatic surface potential of the interfaces between SsaV C monomers within the nonamer and found that electrostatic interactions in SD3 and SD1 of SsaV C may facilitate the subunit nonamerization (Fig. 4A). Close inspection of the oligomerization interfaces revealed that four pairs of salt bridges presenting in SD3-SD3 (R534-E488, E502-R490) and SD3-SD1 (R567-E407, R563-E482) could stabilize the SsaV C nonamer ( Fig. 4B and Fig. S6A), which is consistent with the electrostatic surface potential analysis above and with previous studies (26,44,45,47). SctV L has also been reported to be required for forming the SctV C ring (45,50). In each SsaV C monomer, the hydrophobic pocket at the connection region between SD1 and SD3 is occupied by a hydrophobic peptide (M346-V347-P348-G349-A350) from the neighboring SsaV L , forming the hydrophobic interactions between two adjacent subunits ( Fig. 4C and Fig. S6B). To verify the importance of these two different intermolecular interactions for the SsaV C nonameric structure formation, we created two variants, SsaV C M4 (E407A, E482A, E488A, R490A) and SsaV C N4 (deletion of M346-V347-P348-G349), and tested the oligomerization ability of these two variants through SEC. The results showed that neither variant could oligomerize in solution (Fig. 4D), indicating that both intermolecular salt bridges and hydrophobic interactions are essential for SsaV C nonamerization.
The intermolecular salt bridges are remarkably conserved in all reported SsaV C structures, consistent with a general role in maintaining the SctV C ring structure (26,44,45,47). The hydrophobic pocket at the connection region of the SD1 and SD3 is also exhibited in SsaV C homologous proteins InvA C , CdsV C , EscV C , MxiA C and FlhA C (Fig. 4E). However, the hydrophobicity of the pocket-nested peptide from the neighboring SctV L exhibits great diversity in different homologous proteins (Fig. 4F). InvA C , EscV C and FlhA C show relative hydrophobic pockets at the connection region of the SD1 and SD3. However, the interaction peptide from InvA L (VSTET) is very hydrophilic, and those from EscV L (ISPGA) and FlhA L (SLGME) are less hydrophobic than SsaV L , reducing the hydrophobic interactions between neighboring subunits. The  different strengths of intermolecular hydrophobic interactions provide possible explanations for why InvA C is unable to form the nonameric ring in solution ( Fig. 1B and C) and why the ring structures of EscV C and FlhA C are disassembled in high-concentration salt buffer (32,45). The relatively strong hydrophobic interactions via the unique amino acid sequence of SsaV L and the conserved salt bridges between adjacent subunits of SsaV C can promote formation of a high-order oligomer in different conditions, which may evolve to adapt to the unique environment of the SCV.
The SsaV linker region in chimeric ISS exhibits variable conformations. The structural and functional analysis above shows that the peptide between SsaV L and SsaV C plays an essential role in stabilizing the SsaV ring structure. The structural features of TM and linker regions of SctV were previously shown only using in situ cryo-ET method (12). To review more structural information of SsaV L and InvA TM , we further processed the cryo-EM data of the chimeric ISS in two independent strategies (Fig. S7 in the supplemental material) via which linker conformational changes of SctV were observed.
In the side views of 2D classification, clear features of the SsaV L region were captured between the SsaV C ring structure and the blurry micelle of InvA TM domain (Fig. 5A). A characteristic class of ISS map featured with visible TM region and linker region (highlighted in the red box) was obtained from 3D classifications (details in Materials and Methods and Fig. S7). Using 3D classification with this map as the reference without a local mask, the linker region was classified into variable conformations, with two major orientations, featured as left and right linkers (Fig. 5B, bottom-left panel). For independent validation of variable states of the linker region, a 3D classification was performed by skipping alignment with a local mask in the linker region, and classes of the linker region in different orientations (left and right linker) were obtained (Fig. 5B, bottom-right panel). The linker region obtained from the two separate methods both exhibit two major states of SsaV L with different orientations relative to SsaV C in the fixed position ( Fig. 5C and Fig. S7). The distance between InvA TM and SsaV C appears to vary in these two conformations at current resolution (Fig. 5C). The intriguing conformational changes of SsaV L implies that SsaV might undergo a dynamic process during the substrate selection and secretion. However, more comprehensive and accurate information on the functional mechanism of SctV will require high-resolution SctV full-length structures in different conformations.

DISCUSSION
As one of the engines of T3SS, SctV plays essential roles in effector selection and secretion. Due to the challenges in obtaining well-assembled SctV outside of the bacterial inner membrane, the structure and functional mechanism of SctV have been largely undefined. Through constructing the InvA TM -SsaV C chimeric protein, we produced a homogenous high-order oligomeric SctV protein for cryo-EM structural study. However, we could only determine the high-resolution structure of nonameric SsaV C and show a low-resolution map of SsaV L . A few of features of InvA TM can also be observed ( Fig. 5C and Fig. S7) but detailed structural information was lacked even after intense data processing. It might be resulted from the high flexibility of linker region, otherwise, InvA TM might be intrinsically unstable or even unable to form the nonamer when it is extracted from the bacterial membrane, perhaps resulting from losing structural support from other partners of the T3SS. Further efforts may be needed to reconstitute SctV into the lipid membrane, perhaps with other potential interaction partners, to force SctV TM to form the stable nonamer.
It has been hypothesized that T3SS is energized by the ATPase-dependent ATP hydrolysis coupled with the proton-motive force (PMF) to secrete unfolded effectors, which is executed by the complex of export gate protein (SctV), center stalk protein (SctO), and ATPase (SctN) with a rotary catalytic mechanism on ATP hydrolysis consistent with the evolutionarily related F 0 F 1 -ATPase (25)(26)(27)(28)(56)(57)(58)(59)(60). In F 0 F 1 -ATPase, the membrane-embedded c ring of F 0 and hydrophilic ATPase F 1 are two rotatory motors connected via the g-subunit to translocate protons and generate a difference in potential by hydrolyzing ATP (56,60). However, how the PMF and ATP hydrolysis coupling for the substrate secretion of T3SS is largely unknown. In our study, the 3D classification of SsaV L shows two major orientations of SsaV L , implying that SsaV L could be very dynamic during its functioning. The dynamic SctV L , SctV C , and SctO might form a bridge to coordinate the coupling between the SctV TM (PMF) and SctN (ATP hydrolysis) to facilitate the substrate secretion of T3SS in an efficient manner (Fig. 6A). The dynamic conformation of SctV L may also provide a structural explanation for the resent model for the action of FlhA in flagellar export that the FlhA C /SctV C need to move to the FlhA TM /SctV TM back and forth during the secretion cycle (49). Furthermore, considering the constructional, compositional and functional similarity between T3SS core engine system (SctVON) and F 0 F 1 -ATPase (Fig. 6B), as well as our observation of highly dynamic loop region of SctV in the full-length context, suggests an interesting hypothesis that T3SS may share a conserved rotary catalytic mechanism (SctV and SctN as two rotatory motors coupled via the SctO) with F 0 F 1 -ATPase to generate energy for protein secretion. However, this rotating model during SctV functioning is highly speculative. Further high-resolution full-length SctV structures and functional assays are imperative to fully dissect the molecular mechanism of how T3SS is energized for protein selection and secretion.
SctV C was reported to involve substrate selection and secretion through the intramolecular cleft formed by SD2 and SD4 and the intermolecular cleft formed by two neighboring SD4s (13,32). The SD2-SD4 cleft is dynamic and ranges from open to closed to selectively bind and release effector-chaperon pairs (13). The SD4-SD4 cleft has been shown to interact with central stalk protein SctO to facilitate the connection between SctV and ATPase SctN (26,61). Due to the essential function of SctV in the T3SS secretion process, blockage of these two vital clefts of SctV should significantly diminish the function of T3SS and the virulence of the T3SS-employing bacterial pathogens (59). Therefore, SctV C could be considered a potential novel target for developing anti-virulence drugs to some antibiotic resistant Gram-negative bacterial pathogens.
In this paper, we present near atomic-level insights into the assembly and functional mechanism of SctV C and report variable states of SctV linker region. This study sheds light on important but heretofore poorly understood aspects of the remarkably complex biology of T3SS export gate protein SctV and thus has important implications In F 0 F 1 -ATPase, the membrane-embedded c ring of F 0 and hydrophilic ATPase F 1 are two rotatory motors connected via the g-subunit to translocate protons and generate a difference in potential by hydrolyzing ATP. In T3SS, the ATPase complex (SctN) is associated with the export gate (SctV) through interaction with the central stalk (SctO). The ATPase complex (SctN) generates energy by ATP hydrolysis in a rotary catalytic mechanism, which may drive effectors unfolding and secretion coupling with PMF generated by SctV in a cooperative manner.
for the development of therapeutic strategies targeting SctV C to combat T3SS-mediated bacterial infection.

MATERIALS AND METHODS
Expression and purification of SsaV C and InvA TM -SsaV C chimera. The DNA for SsaV C (encoding residues 346-682) was cloned into pET15b (Novagen, Gibbstown, NJ) with a thrombin-cleavable His 6 tag at the N-terminus. To generate the InvA TM -SsaV C chimera, residues 316-685 of InvA were replaced by residues 326-682 of SsaV through Gibson assembly (62). All primers used in this study are listed in Table S2 in the supplemental material and all constructs were checked by DNA sequencing. An N-terminal Strep-tag and a SUMO protein in tandem were fused with InvA TM -SsaV C chimera. Overexpression in Escherichia coli BL21 was induced overnight with 0.2 mM isopropyl-b-d-thiogalactopyranoside (IPTG) at 22°C when OD 600 reached 0.8 (for InvA TM -SsaV C chimera, OD 600 = 1.2).
For purification of SsaV C , culture was harvested by centrifugation at 5,050 Â g for 15 min at 4°C and resuspended in lysis buffer containing 20 mM Tris-HCl (pH 8.0) and 150 mM NaCl and disrupted through a high-pressure homogenizer. After centrifugation at 17,000 Â g for 50 min at 4°C, the supernatant was applied into Ni-NTA resin (Qiagen, Hilden, Germany) and washed three times with 10 ml lysis buffer plus 20 mM imidazole. The protein was then eluted with lysis buffer with 300 mM imidazole, and further purified through an anion-exchange column (Hitrap Q, GE Healthcare, Fairfield, CT). Peak fractions were pooled and concentrated using 10 kDa cutoff Centricon filters (Millipore, Boston, MA) and subjected to size exclusion chromatography (Superdex 200 Increase 10/300 GL, GE Healthcare) in the lysis buffer. Finally, peak fractions were collected and concentrated to 5 mg/ml by 100 kDa cutoff Centricon filters for cryo-EM analysis.
To purify the InvA TM -SsaV C chimera protein, 6 L of culture was collected, resuspended in lysis buffer, and disrupted through a high-pressure homogenizer. Insoluble fractions were removed by centrifugation at 20,000 Â g for 20 min, while the supernatant was further ultracentrifuged at 150,000 Â g for 1 h. The pellet (containing the membrane fraction) was resuspended in lysis buffer supplemented with 1% (wt/vol%) GDN and incubated at 4°C overnight. After centrifugation at 150,000 Â g for 30 min, the supernatant was applied into Strep-Tactin Beads (Smart-Lifesciences, Changzhou, China) by gravity and washed with buffer W (lysis buffer plus 0.004% [wt/vol%] GDN). The target protein was eluted with buffer W after the SUMO tag was cleaved on the beads. The eluent was concentrated with a 100 kDa cutoff Centricon filter and further purified through size exclusion chromatography (Superose 6 Increase, GE Healthcare) in buffer W. SsaV and InvA were purified using the same procedure. For detergent screening, detergents were changed from membrane extraction to Superose 6 column in the purification. For cryo-EM analysis, peak fractions were concentrated to ;10 mg/ml using 100 kDa cutoff Centricon filters.
Cryo-EM data acquisition. Aliquots of 4 ml concentrated samples were applied to glow-discharged holey carbon-coated grids (Quantifoil Au R1.2/1.3, 200 mesh, Beijing Zhongjingkeyi Technology, Beijing, China). Grids were blotted for 3.5 s at 8°C with 100% humidity and frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific, Waltham, MA). Grids were transferred to a Titan Krios (Thermo Fisher Scientific) operating at 300 kV and equipped with Gatan K3 Summit detector (Pleasanton, CA) and a GIF Quantum energy filter (slit width 20 eV). Micrographs were recorded in the super-resolution mode with a nominal magnification of 105,000x, resulting in a calibrated pixel size of 0.422 Å. Each stack of 32 frames was exposed for 2.13 s with an exposing time of 0.067 s per frame. The total dose was ; 50 e-/Å 2 for each stack. AutoEMation (63) was used for the fully automated data collection. All 32 frames in each stack were aligned and summed using the whole-image motion correction program MotionCor2 (64) and binned to a pixel size of 0.8433 Å. The defocus value of each image was set to 20.8 mm to 21.5 mm and determined in cryoSPARC (65).
EM data processing for SsaV C . The data acquisition of SsaV C is described as above Cryo-EM data acquisition section (named as data set 1). Out of 4,020 micrographs, 1,987,239 particles were automatically picked by cryoSPARC. After two rounds of 2D classification using cryoSPARC, a small subset of good particles was selected to generate the initial model; 584,301 good particles after 2D classification were used for 3D classification with C9 symmetry using cryoSPARC. Double-layer class (111,741 particles) and single-layer class (90,916 particles) were classified and further processed using Non-Uniform refinement, with D9 and C9 symmetry, respectively, resulting in double-layer map at 3.55 Å and single-layer map at 3.64 Å. A flowchart showing the data processing is shown in Fig. S1 in the supplemental material.
EM data processing for InvA TM -SsaV C chimeric protein (ISS). (i) Data processing for SsaV C of ISS. The data acquisition of chimeric ISS is described as above Cryo-EM data acquisition section (named as data set 2). Out of 11,550 micrographs, 3,673,478 particles were automatically picked by cryoSPARC. After two rounds of 2D classification using cryoSPARC, a small subset of good particles was selected to generate the initial model: 1,225,081 good particles after 2D classification were used for 3D classification with C9 symmetry using cryoSPARC, and 734,284 good particles from the 3D classification were processed with further nonuniform refinement and local CTF refinement with C9 symmetry, resulting in the SsaV C EM map with an averaged resolution at 2.11 Å. Features of the TM region and linker region of ISS are invisible in this map at current resolution. A flowchart showing the data processing is shown in Fig. S2 in the supplemental material.
(ii) Data processing for Initial model of InvA TM . For checking the sample quality of the InvA TM -SsaV C chimeric protein, 168 micrograph stacks were recorded using Talos Arctica (Thermo Fisher Scientific) at 200 kV equipped with a K2 detector (Pleasanton, CA), with motion correction using with MotionCor2 and CTF estimation using cryoSPARC, resulting in a calibrated pixel size of 1.17 Å. Out of 168 micrographs, 52,849 particles were automatically picked by cryoSPARC. After two rounds of 2D classification using RELION3.0 (66), a small subset of good particles was selected to generate the initial model; 44,442 good particles after 2D classification were used for 3D classification with C9 symmetry using RELION3.0. A representative InvA TM map with featured TM region was obtained from 3D classification (Fig. S7A, red box, Model 1, 14,370 particles). A flowchart showing the data processing is shown in Fig. S7A. (iii) Data processing for InvA TM and SsaV L without local mask. For the InvA TM -SsaV C chimeric protein described as above (data set 2), 375,338 particles in side views after 2D classifications were selected to process with 3D classification using Model 1 obtained above as the reference with C9 symmetry using RELION3.0, resulting in a full-length EM map class of InvA TM -SsaV C chimeric protein with features of InvA TM , SsaV L , and SsaV C (Fig. S7B, red box, Model 2, 28,922 particles). After two rounds of 3D classification, ISS EM maps with different linker states (Left', Left, Right, Right') were obtained. Representative EM maps (Left linker, 13.99 Å and Right linker, 13.49 Å) were obtained after refinement using RELION3.0. Reported resolutions were calculated on the basis of the FSC 0.143 criterion. A flowchart showing the data processing is shown in Fig. S7B in the supplemental material. (iv) Data processing for SsaV L with local mask. For further estimating linker states of SsaV, 546,372 particles after 2D classification described above (Data set 2) were used to run a 3D classification by skipping alignment, using Model 2 as the reference, with a local mask in the linker region and C9 symmetry using RELION3.0. Two distinguished classes featured with left linker (31,427 particles) and right linker (43,387 particles) were obtained from the 3D classification. A flowchart showing the data processing is shown in Fig. S7C in the supplemental material.
Model building and structure refinement. The 2.11 Å reconstruction map was used for model building. The starting model of SsaV C based on the structure of SsaV C (PDB:7AWA) was manually built in Coot (67), followed by refinement against the corresponding maps in PHENIX (68) with secondary structure and geometry restraints. Statistics of 3D reconstruction and model refinement are summarized in Table S1 in the supplemental material. Structural figures were made using PyMOL v.2.3.2 (69) and UCSF ChimeraX v.1.1 (70). Analysis of sequence conservation was determined by the ConSurf server (71) according to sequence alignment using ClustalW (72). The pore diameter diagram calculated using the Hole program (73) in Coot. Phyre2 (74) was used to model the protein structures.
DLS measurement. DLS measurements were carried out using cuvette-based systems on a DynaPro NanoStar (WYATT, Santa Barbara, CA). Purified proteins were diluted to 0.5 mg/ml in lysis buffer containing 20 mM Tris-HCl (pH 8.0) and 150 mM NaCl. After centrifugation at 17,000 Â g for 5 min, aliquots of 8 ml of samples were analyzed at 25°C.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 1.7 MB.

ACKNOWLEDGMENTS
We thank Jorge Galan (Yale University) and Yongjian Huang (University of California, Berkeley) for constructive and helpful discussions and Yigong Shi (Westlake University) for his kind support on cryo-EM resource and computational resource during this project. We thank Xiaoju Li from Core facilities for life and environmental sciences, Shandong University and Kang Li from the cryo-EM facility for Marine Biology at QNLM for helping with cryo-EM sample screening. We thank Fan Yang (Tsinghua University) for technical support during EM image acquisition. We thank the Tsinghua University Branch of China National Center for Protein Sciences (Beijing) and Fudan University for providing the cryo-EM facility support. We thank the computational facility support on the cluster of Bio-Computing Platform (Tsinghua University Branch of China National Center for Protein Sciences Beijing), Center of Cryo-Electron Microscopy (Fudan University), and HPC Cloud Platform (Shandong University).
The atomic coordinates and EM map have been deposited in the Protein Data Bank (www.rcsb.org) and Electron Microscopy Data Bank (www.ebi.ac.uk/pdbe/emdb/) with the accession codes 7FEB,7FEC,7FED and EMD-31551, EMD-31552, EMD-31553, respectively. Materials are available from the corresponding authors on request.
X.G. and Y.Z. conceived the project. X.G., J.X., and Y.Z. designed the experiment. J.X. performed the experiments of purification with the help of J.W. and all assays. Y.Z. and J.X. performed cryo-EM sample preparation and data collection. Y.Z. performed cryo-EM data processing. J.X. performed structural model building. All authors contributed to data analysis. X.G., Y.Z., and J.X. wrote the manuscript. This work was funded by the National Key R&D Program of China (2018YFE0113000 to X.G.), National Natural Science Foundation of China (31770143 and 32122007 to X.G.), the Major Basic Program of Natural Science Foundation of Shandong Province (ZR2019ZD21 to X.G.), the Youth Interdisciplinary Innovative Research Group of Shandong University (2020QNQT009 to X.G.), the Taishan Young Scholars Program (tsqn20161005 to X.G.), Shanghai Rising-Star Program (21QA1401200 to Y.Z.), Original Research Program from Fudan University (IDH1340064/009 to Y.Z.), and startup funds from Fudan University (JIH1340063 to Y.Z.).
We declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.