The molecular envelope of the overall needle complex
We purified intact S. Typhimurium SPI-1 needle complexes as previously described16. Single particle reconstruction resulted in unsymmeterized maps to 7.4 Å resolution as judged by gold-standard FSC20 revealing the molecular envelope for the IM and OM rings of the basal body and the encompassed substructures with attached needle filament (Fig. 1a). Comparison to the previously determined basal body reconstruction7 reveals the general localized changes within the core secretin domains necessary to accommodate the assembled needle filament; notably the extended open orientation of the periplasmic gate (Fig. 1b). Outside of these regions, the overall structure is remarkably static between assembly states.
The structure of the 24-mer IM rings
The 24-fold features of the IM rings in the C1 reconstruction were clearly visible and further C24 symmetry averaging resulted in a reconstruction to 3.6 Å resolution (compared to 4.3 Å for the basal body IM rings; EMD-8398) allowing for refinement of the atomic structures of the periplasmic domains of PrgH171–364 and PrgK20–203 (Fig. 1c, d, Supplementary Fig. 1, Supplementary Table 1). Consistent with the conserved dimensions of the IM rings between the needle complex here and that of the previous basal body reconstructions (Fig. 1b), the refined PrgH and PrgK ring atomic models determined from each overlap closely (overall Cα RMSD 0.7 Å) suggesting an inherently highly stable structure with no significant conformational change required in this region upon needle assembly.
The export apparatus and needle filament
The low-resolution substructure density localized within the central lumen of the stacked IM PrgH and PrgK rings at the base of the needle filament has historically been referred to as the socket and cup. It is present in native and T3S incompetent mutants alike5,7, the latter of which consequently lack the rod/needle filament (themselves T3S substrates). Mutational, mass spectrometry and low resolution EM analysis have attributed this substructure density to regions of the IM associated export apparatus components SpaP, SpaQ, SpaR, and SpaS9,15,21, with knockout of one or more removing the substructure in vitro and in situ9,13. Remarkably, a recent cryo-EM structure of the isolated recombinant export apparatus complex from the bacterial flagellar system (long recognized to encode several analogous IM components to the injectisome) shows that this substructure is composed of the flagellar orthologues FliP/Q/R (25–30% sequence identity with Salmonella SPI-1 SpaP/Q/R) in their entirety despite predictions they were integral membrane proteins with 4, 2, and 6 predicted transmembrane domains, respectively. The complex adopts a helical assembly of dimensions ~90 × ~60 × ~60 Å with 5:4:1 stoichiometry (FliP/Q/R) with the single FliR bridging the IM facing FliQ tetramer and the more periplasmic facing FliP pentamer and elegantly explains previous cross-linking and mass spectrometry studies9,21,22,23. Complicated by the local symmetry mismatch (e.g., with the 24-mer IM rings which encompass the export apparatus) and potential relative conformational variation, we have been unable to resolve the high resolution structural features of this substructure in the current needle complex data or in the prior basal body reconstruction. However, whereas the latter presented a rather flat, symmetrical pore due to averaging out of the local structure (see Fig. 2b), the needle complex reconstruction here provides an inherently more detailed, asymmetric substructure than previously observed (Fig. 2a–c). Overlay of the flagellar FliP/Q/R structure matches well throughout suggesting the structure of the flagellar and non-flagellar T3SS export apparatus will be conserved, providing evidence for the precise span of the export apparatus within the assembled T3SS (Fig. 2d).
The needle filament observed in our reconstruction forms a continuation of the export apparatus with the internal channel diameter the same as the export apparatus atrium (Fig. 1b, Fig. 2c). The needle (PrgI) is proposed to be initiated by a homologous protein historically termed the rod (PrgJ)—both detected by MS analysis here (Supplementary Table 2) but absent, as designed, from the purified secretion incompetent basal body7. Although the general dimensions and span of the filament are clearly observed, like with the export apparatus and the issues with localized symmetry mismatch with surrounding components, we were unable to sufficiently resolve the high-resolution structural features within the assembled needle complex to allow model building and refinement. To further study the atomic details of the needle therefore, we isolated native filaments from assembled needle complexes and subjected them to cryo-EM analysis resulting in a reconstruction with helical averaging to 3.3 Å resolution with excellent density for side chains throughout (Supplementary Fig. 2, Fig. 3c, Supplementary Table 1) allowing us to build a de novo model of PrgI lacking only the first two N-terminal residues (Figs. 1c–d, 3). The PrgI monomer adopts a helix-turn-helix motif that polymerizes via an extensive network of interactions along both helices with surrounding subunits (Fig. 3a, b). The structure supports the solid-state NMR model of PrgI presented by Loquet et al.18 (Cα RMSD 1.5 Å over 78 residues), which was fundamentally different from prior low resolution cryo-EM derived models24,25; however, some notable differences in the monomeric structure and helical packing, as well as the definition of the majority of side chain positions extend our understanding of the intra-subunit interactions and the internal channel passaged by effectors. The kinked loop in the N-terminal helix (residues 20–23) is more extended and forms a tighter interface with the N-terminus of monomer i + 6 while the C-terminal Arg80 has its side chain and carboxylate flipped, altering the interactions with monomers i −1, i –5, and i – 6 (Fig. 3d, e). The refined polymer forms a helical assembly with approximately 11 subunits per two turns but a minor difference in the helical rise compared to the solid-state NMR model (helical rise 4.33 Å vs. ~4.2 Å, respectively) creates a small but significantly accumulating shift in subunit packing (Fig. 3a, b). Along with the specific structural changes at the C-terminus, this creates a right handed spiraled groove with polar and hydrophobic character extending along the needle lumen which was not observed in the NMR model, with a cluster of conserved residues around the C-terminus defining the raised edges of the groove (Fig. 3f–h). Taking into account the resolved side-chains here (Fig. 3b, c), the axial diameter is considerably less than previously measured in lower resolution reconstructions24,26,27 (~15 Å vs. ~25 Å, only very closely accommodating a single helix for example) suggesting unfolded or partially unfolded secreted effectors would necessarily track along the deeper and longer helical path during passage (Fig. 3f–h). Indeed, previous scanning alanine mutations in the Shigella orthologue MxiH showed that mutation of conserved charged residues that here define the raised groove (equivalent to PrgI Lys66, Lys69, Asp70, Arg80; Fig. 3e–g) had differential effects on needle polymerization and length, secretion, and regulation28 suggesting a specific contribution of the interior needle channel surface in the hierarchical secretion of T3S effectors as opposed to merely providing a channel for passive diffusion. The helical nature of the secretion channel also perhaps raises the intriguing possibility that an as yet undefined rotational force, perhaps linked to energy from the proton motive force required for secretion29,30, might contribute to effector secretion. Fitting of the cryo-EM derived model into the C1 needle complex map shows a good agreement (Chimera correlation coefficient = 0.95) between the wider part of the filament density (Figs. 1c, 2c). This needle placement suggested by our data here, anchored deep into the secretin lumen in keeping with its needed stability within the relatively severe mechanical environment of the infected cell, is consistent with a recently proposed role of the protein PrgJ as a relatively small bridge or adaptor between the IM export apparatus and the needle filament22 rather than an extended rod as its historical name unfortunately had predicted (here accounting for the narrowing region of density at the base of the filament (Fig. 2c)). Taken together, these structures now demonstrate that the SpaP/Q/R export apparatus complex is positioned to provide a structural foundation for the assembly of the needle filament with the helical, pseudo-hexameric arrangement of the central 5× SpaP and 1× SpaR subunits an ideal template for the downstream T3S dependent polymerization of the helical filament, with several (likely hexameric) copies of PrgJ (predicted stoichiometry 3–6 by mass spectrometric analysis22) putatively assembling onto the SpaR/P platform (cross-linking demonstrates close interaction21) and in turn providing a helical template to initiate the helical needle polymerization.
The open state of the OM secretin gate
In these needle complex samples, the strong 24-mer symmetry of the IM rings dominates the 2D class averages, a consequence of which is that the region of the C1 needle complex reconstruction corresponding to the ~1 MDa OM secretin component is somewhat smeared out and lacked obvious symmetric or high-resolution features. We therefore employed a partial signal subtraction and masked focussed local refinement procedure31 with the obvious 15-fold symmetry resulting used subsequently in averaging to produce reconstructions to 4.1 Å resolution for the entire secretin and 3.9 Å resolution for the core secretin structure encompassing the N3, secretin and S domains (Fig. 1c, d, Supplementary Fig. 3). The density in the core region (InvG176–557) showed defined secondary structure with the majority of side chains clearly visible (Supplementary Fig. 3h) allowing for unambiguous building and refinement of the structure including significant additionally resolved and structurally changed regions from the previously determined closed form7. Although the more peripheral N-terminal N0 and N1 domains were of lower local resolution (Supplementary Fig. 3g), the map was of sufficient quality to reveal secondary structure elements for the N1 domain in particular allowing us to position 15 copies of our prior InvG34–173 crystal structure for that domain (PDB 4G085) (Supplementary Fig. 5e) further expanding the defined model of the injectisome. The final secretin model spanning all domains (residues 34–557, N3 domain loop 228–251 disordered; Fig. 4, Supplementary Table 1) was refined with density restrained symmetry refinement in Rosetta and notably reveals the dramatic conformational changes involved in secretin gate opening during needle assembly, a fundamental and functionally critical aspect of this giant gated pore family.
Overlay of the needle complex and prior closed basal body reconstructions illustrates that the primary sites of conformational change upon needle assembly are in the conserved core secretin region: the N3 ring domain—the tightly oligomerized foundation upon which the secretin assembles7—and the inner β-barrel and membrane associating lip of the central secretin domain, all of which have reoriented to accommodate the presence of the needle filament (Fig. 1b). The N-terminal N0 and N1 domains, outer β-barrel and C-terminal surface exposed S domain conversely show very little structural difference between assembly states; as with the static IM ring components they appear to provide a highly stable scaffold or shell to buttress the inner conformational changes required for needle passage. The most dramatic conformational change upon needle assembly is in the inner β-barrel. In the previously observed closed state7, the extended β-hairpin formed by strands 4 and 5 (numbering consistent with the closed InvG structure7, here termed GATE1 based on mutational analysis in secretin pIV32) is kinked and twisted around residues Asn386 and conserved Gly407, with the spoke like radial projections afforded by the pentadecameric symmetry forming a collective barrier across the ~75 Å secretin lumen that is permeable only to small molecules32,33,34. In the T2SS, mutation of the Asn386 equivalent in PulD (Gly458) increased antibiotic sensitivity35 and accordingly resulted in a partial opening of the gate in the cryo-EM structure of GspD (Gly453)36. In the fully open state defined here, the GATE1 hairpin is now unkinked, with the residues at the tip having moved position by a remarkable ~40 Å (measured from the Gly395 Cα) (Fig. 5). The open position is stabilized by intimate packing against the outer β-barrel of itself and neighboring monomer i + 1 (monomers defined clockwise looking down from the OM), and extending vertically toward the lip region of monomer i + 2 (Figs. 4c, 5b–d) with a dramatic increase in buried surface area between the inner and outer β-barrels (2111 Å2 per monomer compared to 1153 Å2 in the closed state) (Fig. 5b–e). Side chain density is defined for the large majority of residues along the length of both GATE1 strands with the exception of 5 residues (Glu396-Ala400) at the tip. Although no longer kinked, the twist in GATE1 is maintained in the extended state such that strand 4 is oriented closer to the outer β-barrel forming multiple specific interactions. Of note, Lys392 and Ile394 at the C-terminus interact with residues on the outer β-barrel and lip of monomers i + 1 and i + 2 (Fig. 5b–d). In our previous secretion assays, mutation of either Lys392 or Ile394, as well as Gly395, which forms a tight turn in both closed and open gate conformations, significantly impacted secretion7. The inner β-barrel strands 6 and 7 (termed GATE2 here) are straightened around Gly430 and Gly451 compared to the closed form. More significantly, the GATE2 loop (residues Asp433-Gly451) undergoes a 180° rotation, effectively repositioning the regions oriented above the GATE1 hairpin in the closed state against the outer β-barrel in the open state (Fig. 5). These changes in GATE1 and GATE2 of the inner β-barrel result in an increase in the secretin pore diameter from ~15 Å in the closed state to a remarkable ~75 Å in the open state (Figs. 4b, 5a).
The secretin outermost lip constitutes a kinked 45-stranded β-barrel, capping the top of the 60-stranded outer β-barrel wall7 (Fig. 4). With one strand less per monomer, the well-ordered lip was observed in the basal body closed reconstruction to collapse inwards to structurally accommodate the reduced strand count. Decorated on the exterior by an amphipathic helical loop (the AHL)—the most conserved sequence across the secretin family7—this region was proposed on structural features to be central to membrane association, BAM-independent insertion and OM span7. In the needle assembled open state here, the lip has been pushed up and out to accommodate the passage of the filament, increasing the outer pore diameter from ~65 Å in the closed state to ~70 Å in the open state (Fig. 5a). The presence of the needle supports a more upright lip orientation consistent with the recent in situ cryo-electron tomography structure of the S. Typhimurium T3SS which demonstrated the membrane invagination we had noted previously13,37,38 to be more pronounced in a mutant lacking the needle compared to the assembled needle complex13 (Supplementary Fig. 4). Interestingly, recent structures of several T2SS secretins have revealed their lip region consists of 4 rather than 3 β-strands per monomer and as such present a non-kinked continuation of the outer β-barrel wall36,39 (Supplementary Fig. 5a, b). Sequence analysis suggests the three-stranded lip architecture observed for InvG will be common in other T3SS secretins and may reflect the necessity for a closer fitting and more stable interface with the needle, which is more statically attached in the injectisome than, by comparison, the dynamically attached (pseudo)pilus of the T2SS or T4PS which are in a fluctuating state of extension and retraction.
The N3 domain abuts the periplasmic base of the inner β-barrel formed by GATE1 and GATE2 (Figs. 4, 5) and has been demonstrated through mutagenesis to be essential for secretin oligomerization35,40,41. N3 belongs to the family of small mixed α/β domains we previously termed ring building motifs4 (RBMs) which have been observed in numerous ring forming components from different secretion systems. Although often sharing a common general oligomerization interface4,5,32,33 the variation of non-covalent interactions and surface areas observed suggest inherent degrees of stability in RBM oligomerization that may reflect their role in assembly. The structure of InvG7 and subsequent T2SS secretins36,39 demonstrated that the N3 domain forms a ring oligomer with a consistently hydrophobic RBM interface that is also intimately associated with the underside of the inner β-barrel of the neighboring i + 1 monomer (Fig. 4c, Supplementary Fig. 5a, b), providing a stable foundation for secretin oligomerization. Here, in the open state, the N3 domain is rotated outwards slightly (Fig. 5a). The β-hairpin insertion specific to T3SS secretin orthologues (referred to here as β-INSERTION), observed to form a strut supporting the conformation of GATE1 and GATE2 in the closed state7, is pulled away and twisted, disrupting the specific interactions of conserved Arg198 at the β-INSERTION turn with Asp384 and Glu429 at the kinked regions of GATE1 and GATE2, respectively (Fig. 5b–e). The neighboring conserved Asp199 also shifts to interact with Arg411 at the base of GATE1 in the open state (Fig. 5b–e). Mutation of Asp199 completely abrogated secretion with Arg198 having a milder phenotype in our previous assays7 supporting the importance of these observed interactions in the two functional open/closed conformational states of InvG.
The quality of the needle complex local reconstruction in the OM secretin region has also permitted the observation of further structural features previously not resolved: First, we have been able to model additional sequences of the variable insertion (residues 217–267) between the first helix and second strand of the N3 domain. The now resolved 217–227 and 252–267 residues demonstrate an unexpected contribution to the structural stabilization of the secretin N domains (Fig. 5a, Supplementary Fig. 5c, d). In the prior basal body reconstruction, we suggested this unresolved loop might account for the map features forming a periplasmic constriction evident at lower contour levels7. However, with the needle filament occupying the secretin lumen in the assembled state here, the now clear atomic details of the loop show it threads out between the N3 and N1 domains and participates in both intra and inter domain interactions. The N-terminal region (residues 217–227) contributes to the N3 domain RBM interface, effectively chelating between the first strand of monomer i–1 and the second helix of monomer i (Supplementary Fig. 5c,d) while residues Met252-Gln260 form an α-helix which packs between the N1-N3 domain loops of monomer i and i−1 and interacts with the N1 domain of monomer i (Supplementary Fig. 5c,d). Collectively these additional interfaces serve to further anchor and stabilize the needle bound form of the injectisome at the point of greatest structural change. Second, we have been able to position the peripheral N0 and N1 domains, which bridge the periplasmic space to the IM rings with the N-terminal residues of the N0 domain interacting with the C-terminus of PrgH7,12,17 (Fig. 1a, Supplementary Fig. 5e; see above). Interestingly, while the structure validates our previous Rosetta density guided symmetry models of the InvG34–172 crystal structure5,7, the data here confirms the direction of the N1 domain RBM interface is indeed opposite to the N3 domain (Supplementary Fig. 5a). This is in contrast to the recent cryo-EM structures of several T2SS secretins36,39 where the N1 and N2 (a distinct RBM motif not present in T3SS secretins) domains are shown to have the same orientation with the relative inter N domain packing more staggered compared to InvG (Supplementary Fig. 5a, b). These rather unexpected distinctions again clearly underlie the plasticity of the RBM in modulating periplasmic span and inner membrane coupling in these functionally diverse secretion systems.
A mechanism for substrate mediated secretin gate opening
We previously envisaged a substrate induced gate opening mechanism for the OM secretin during needle assembly based on comparison of the basal body structure to previous low resolution (10–20 Å) studies of the needle complex17,42. The near-atomic resolution structure of the needle complex here provides details of the discrete structural changes occurring at the molecular level and permits further elaboration. We hypothesize a gate opening mechanism broadly divided into two stages: an initial allosteric transition to unlock the gate and a secondary steric phase to push the gate open. Rod/needle polymerization within the secretin lumen would bring the filament into contact with the N3 domain, including the regions of the N3 domain loop that account for the periplasmic constriction visible in the isolated cryo-EM basal body7 and in situ cryo-ET reconstruction13, triggering a reorientation/ordering to adopt the conformation we observe here and contributing lateral stability to withstand the continued force of the assembling needle (Fig. 6a, b). The accompanying change in N3 domain orientation we observe alters the interface with the inner β-barrel, which we propose triggers an allosteric transition that effectively unlocks the gate (Fig. 6b). Central to this appears to be the disruption of interactions of the β-INSERTION with GATE1 and GATE2: Arg198 at the β-INSERTION turn is at the center of a network of polar and hydrophobic interactions that serve to stabilize the closed gate conformation of GATE1 and GATE2 (Fig. 5b–e). The reorientation of β-INSERTION and disruption of this interface could enable the downstream conformational changes associated with gate opening. Supporting an important functional role, mutation of Arg198 and neighboring Asp199 impacted secretion in our previous assays7. GATE2, although not contributing directly to the physical barrier per se (the role of GATE1), appears to play rather a critical buttressing role for GATE1, undergoing a significant displacement between closed and open states (Fig. 5b–e). In the closed state, residues at the N-terminus of strand 7 in GATE2 pack between kinked GATE1 β-hairpins of monomers i and i–1 holding them in a closed position (Fig. 5e), whereas in the open state this same region is rotated 180° and packed against the outer β-barrel and the extended GATE1 of monomer i (Fig. 5d). A specific interaction between GATE2 Glu449 and Arg387 of GATE1 of monomer i–1 is maintained between the two states suggesting the movement of these two hairpins occurs in concert (Fig. 5d, e), with the rotation of GATE2 potentially initiating the lifting of the GATE1 and/or providing structural support to guide the extended β-strands during opening. Mutants found to increase the permeability of the filamentous phage secretin pIV map to GATE2 in addition to GATE1 (Supplementary Fig. 6) highlighting the influence its conformation has on the extent of gate opening. Continued needle polymerization within the secretin lumen could provide additional energy to sterically push the gate open with the resulting reorientation of the entire inner β-barrel we observe (Fig. 6c, d). The resulting complementary fit, notably at the T3SS specific adaptations of the N3 domain β-INSERTION and three-stranded collapsed β-barrel outer lip, and complementary electrostatics of the N3 and β-INSERTION (Supplementary Fig. 7) would contribute to the stable anchoring of the needle through the OM, essential for the assembled needle complex that must endure considerable extracellular forces in the infected host.