Topology and Structure/Function Correlation of Ring- and Gate-Forming Domains in the Dynamic Secretin Complex of Thermus thermophilus

Secretins are versatile outer membrane pores used by many bacteria to secrete proteins, toxins or filamentous phages, extrude type IV pili (T4P) or take up DNA. Extrusion of T4P and natural transformation of DNA in the thermophilic bacterium Thermus thermophilus requires a unique secretin complex comprising six stacked rings, a membrane-embedded cone structure and two gates that open and close a central channel. To investigate the role of distinct domains in ring and gate formation, we examined a set of deletion derivatives by cryo-microscopy techniques. Here, we report that maintaining the N0 ring in the deletion derivatives led to stable PilQ complexes. Analyses of the variants unravelled

reduced number of pili of the N3 mutant. All mutants exhibit a wild-type natural transformation phenotype, leading to the conclusion that DNA uptake is independent of functional T4P.
Secretins in the outer membrane of Gramnegative bacteria are key components for the assembly of type IV pili (T4P) machineries or type II and type III secretion systems (T2SS, T3SS), and are required for the export of filamentous phages (1)(2)(3)(4). Secretin pores can determine bacterial pathogenicity by their role in T4P extrusion, which is crucial for bacterial adherence and motility on solid surfaces, termed twitching motility (5)(6)(7). Natural transformation systems, in bacteria, mediating the uptake of free DNA are also dependent on secretin proteins (8,9). The latter contribute significantly to the acquisition and distribution of antibiotic resistance, which is an emerging threat to human health (10).
A wide range of secretins have been discovered in different bacteria. Some of them have been subject to biochemical and structural characterization, revealing significantly different architectures. Distinct features of secretin complexes are proposed due to their unique functions and corresponding interaction partners and substrates (1,2,(11)(12)(13)(14)(15). Important structural information has been derived from electron cryomicroscopy of isolated GspD, the T2SS secretin from Vibrio cholerae. GspD forms a 12-fold symmetrical cylinder with a diameter of 155 Å and a length of 200 Å (16,17). The conserved β-sheet rich C-terminal domains (CTD) of GspD form a channel through the outer membrane, which is closed by an extracellular and a periplasmic gate (2,17). The periplasmic N-terminal domains (NTDs) N0, N1, N2 and N3 were mapped to the density of the wall of the periplasmic chamber observed by single particle electron microscopy (SP-EM). This revealed that the N3 domain occupies a constriction site that was suggested to be important for the initiation of conformational changes during protein secretion (16).
Recently, we reported the first in situ structure of an entire T4P machinery by electron cryotomography (cryo-ET) using the thermophilic bacterium Thermus thermophilus HB27 as a model organism (18). The central membrane-embedded part of the structure is formed by the secretin PilQ, which plays a dual role in T4P extrusion and natural transformation (9,(18)(19)(20).
PilQ was found to undergo substantial conformational changes between the closed and the open, pilus extruding state (18). SP-EM analyses of purified PilQ complexes revealed that the structure comprises six stacked rings (N0-N5) and a cone structure (20), consistent with the in situ data of the entire T4P machinery (18). Structural analyses of PilQ complexes formed by different PilQ variants led to the identification of an unusual ααβαββα fold as the N0 ring-forming domain. Production of the truncated PilQ variant in a pilQ negative background led to a mutant which was non-piliated and non-motile but still naturally transformable (19) leading to the conclusion that pilus structures are not required for natural transformation. However, identification of other ring-and gate-forming domains was not possible due to instability of the PilQ variants. Moreover, the assignment of specific functions to distinct ring structures in secretins has not been reported so far. Here, we found that maintaining the N0 ring domain in PilQ deletion derivatives leads to stable PilQ complexes, which we used to perform a thorough structure/function analysis, which unravelled one of the gate-forming, and four ringforming domains. We further demonstrated that the individual rings play separate roles in T4P extrusion and T4P dynamics, but are dispensable for natural transformation. These results demonstrate that individual ring structures within the secretin complex exhibit distinct functions during T4P extrusion.

RESULTS
The four conserved βαββα folds and the preceding βββαβ domain are not essential for PilQ complex formation -Secondary structure analyses revealed that there are four contiguous βαββα folds (N2 to N5) in the secretin PilQ from T. thermophilus suggested to form ring-building motifs (20). To analyze the role of these motifs in ring formation, we generated a set of internal PilQ deletion variants (ΔN2, ΔN3, ΔN4 and ΔN5), which were produced in a T. thermophilus pilQ deletion background (Fig. 1). Western blot analyses of membrane fractions revealed that all variants assembled into SDS-stable PilQ complexes ( Fig. 2) demonstrating that deletion of individual βαββα domains does not abolish complex formation.
Identification of the ring-forming domains -To unequivocally identify the ring-forming domains in PilQ, His-tagged deletion derivatives were produced in T. thermophilus and purified by affinity and anion exchange chromatography. SDS-PAGE analyses led to the detection of high molecular weight PilQ complexes (Fig. 3). Different size PilQ complexes and varying amounts of monomers were detected for most of the purified deletion derivatives (Fig. 3). The highest amount of monomer accumulation was detected in preparations of PilQ deletion derivatives devoid of the N1 domain and the N1Δβ 1 β 2 subdomain. This leads to the conclusion that in particular the N1β 1 β 2 plays a major role in PilQ complex stability.
SP-EM studies of PilQ showed six rings and a cone domain, which were also seen in situ by cryo-ET (18)(19)(20). To analyze the effect of the domain deletions on the structure of PilQ, we analyzed the purified deletion derivatives by SP-EM (Fig. 4A). In all cases, with the exception of the ΔN1 mutant, the amount of protein complex was sufficient for SP-EM analysis (Fig. 4A). The overall shape of the complex was comparable to the in situ cryo-ET structure of the PilQ complex. Deletion of the first subdomain of N1 (N1Δβ 1 β 2 ) did not result in an obvious change in the structure compared to the wild-type (Fig. 4D). In contrast, deletion of the second part of N1 (N1Δβ 3 αβ 4 ) changed the structure dramatically (Fig. 4E). The periplasmic end of the complex, corresponding to the N0, N1 and N2 rings, was significantly narrower. The N0 ring, which is the widest ring in the wild-type complex ( Fig. 4B and C), was no longer visible. In the wild-type structure, a second dense domain was observed close to the N0 ring that extends into the center of the complex, forming gate 2 ( Fig. 4B and C). This gate is still clearly present in PilQ complexes devoid of the partial deletions in the N1 domain. The N1 ring is followed by two further rings (N2 and N3), which are similar in size and are separated by a gap of ~12 Å, designated as the N2/N3-gap ( Fig. 4B and C). The ΔN2 and ΔN3 variants were structurally similar, with one ring and the N2/N3-gap missing ( Fig. 4F and G). Deletion of the βαββα domain downstream of the N3 domain led to a loss of the N4 ring, confirming its position in the complex ( Fig. 4C and H). The Cterminal part of all variants, as well as the wildtype, formed a trapezoid ("cone structure"), which shows a strong central density (Fig. 4B). Below the cone is a "cup" structure, with a shape characteristic of the secretin domain. Based on its position, the lower part of the "cone" is likely to be gate 1, as identified by cryo-ET and subtomogram averaging ( Fig. 4B and C) (18). In all ring-variants, the secretin domain and gate 1 were clearly visible ( Fig. 4B-I), suggesting that a specific motif in the secretin domain is responsible for forming this gate. The narrow ring beneath gate 1 in the cup is missing in the ΔN5 mutant complex (Fig. 4I), indicating that it is formed by the βαββα domain in N5 (Fig. 4I).
Identification of the gate 2-forming domain -Due to the poor stability of the ΔN1 mutant, it was not suitable for SP-EM analysis. To investigate the effect of the N1 mutation, the intact T4P machinery composed of PilQ and accessory proteins was analyzed in situ by cryo-ET and subtomogram averaging. The assembled machinery in the closed state (without pili) was identified in whole cells (Fig. 5F) and detected exclusively at the cell poles, in agreement with previous data (18). However, the number of complexes was severely reduced compared to wild-type cells (Fig. 5A, E, F, and J), probably as a result of reduced protein stability. Closed state complexes were selected, aligned and averaged to determine the 3D structure of the ΔN1 mutant in its native membrane-bound state. Comparison to the wild-type complex ( Fig. 5C and D) shows that the N1 domain is responsible for formation of gate 2 ( Fig. 5H and I). The outer membrane-embedded secretin domain was poorly resolved in the average, most likely due to the inherent flexibility of the complex and the reduced number of particles available. Removal of gate 2 did not affect the ability of PilQ to assemble with accessory proteins (designated P1, P2 and C1) into the entire T4P machinery (Fig. 5C, D, H, and I). In rare cases, cells lacking gate 2 in PilQ were also found to assemble pili (Fig. 5G). However, the shape of the PilQ complex was abnormal compared to the wild-type (compare Fig. 5B and C) and it was associated with extracellular material, possibly from cell leakage (Fig. 5B). Conventional Fourier shell correlation, applying the 0.5 FSC criterion are presented in Fig. 6.
The ring-forming domains are important for Pilus Extrusion -To analyze the role of the ringforming domains in pilus biogenesis, the degree of piliation in different pilQ deletion mutants was analyzed in whole T. thermophilus cells by EM. Seventy-eight percent of wild-type cells were piliated with 6 ± 3 pili per cell, whereas only 1 % of the ΔpilQ::bleo mutant cells were piliated with one pilus at the cell surface, mostly on disrupted cells (Fig. 7A), in agreement with previous data (19). All PilQ variants could extrude pili, however the percentage of piliated cells decreased significantly to 20-34 % (Fig. 7A). Moreover, the number of pili per cell decreased dramatically (Fig.  7A). In the wild-type cells only 10 % showed less than 3 pili, whereas 41 % exhibited 4-7 pili and 27 % more than 7. In contrast, the piliated mutant cells extruded only 1-3 pili (Fig. 7A).
Twitching motility of the pilQ mutants -To test the role of the N1-N5 domains and the N1 subdomains (N1Δβ 1 β 2 and N1Δβ 3 αβ 4 ) in T4Pmediated twitching motility, we assessed the ability of cells to twitch. Both wild-type cells and the ΔpilQ::bleo mutant complemented with wildtype pilQ were able to move on solid surfaces with twitching zones of ~3.2 and ~3.5 cm, respectively (positive controls) (Fig. 7B). The noncomplemented ΔpilQ::bleo mutant (negative control) did not show twitching at all (Fig. 7B), in agreement with our previous findings (19). Depending on the specific pilQ mutation, the cells exhibited various changes in their motility. The ΔN3 mutant cells were as mobile as the wild-type [Student´s t-test (95%)] (Fig. 7B). The clear reduction in piliation but wild-type levels of twitching suggests that the remaining pilus structures must exhibit increased dynamics, thereby compensating for their reduced numbers. An alternative explanation would be that only a fraction of pili present on wild-type cells contributes to twitching motility.
The N1Δβ 3 αβ 4 , ΔN2 and ΔN4 mutants were impaired in twitching motility and showed only 34 %, 34 % and 21 % of wild-type motility, respectively (Fig. 7B). An even stronger effect was exerted by deletion of the N5 domain. These mutants exhibited only 5 % of wild-type motility (Fig. 7B). Moreover, the N1Δβ 1 β 2 and ΔN1 mutant were completely defective in twitching (Fig. 7B). These data suggest that mutations in close proximity to either the N-or C-terminus of PilQ have a strong effect on T4P-mediated twitching motility, and that the N5 domain and the Nterminal β 1 β 2 fold in the N1 subdomain are absolutely essential.
The ring-forming domains are not required for natural transformation -The pilQ mutants had transformation frequencies of 1 x 10 -3 ± 3.4 x 10 -4 to 2.1 x 10 -3 ± 8.2 x 10 -4 transformants per living count, which are comparable to those of the pilQ mutant complemented with the pilQ wild-type gene (1.9 x 10 -3 ± 9.4 x 10 -4 ). The negative control, a T. thermophilus ΔpilQ::bleo mutant, was not naturally transformable at all. The complete defect of the N1Δβ 1 β 2 and the ΔN1 mutants in twitching motility together with their wild-type transformation phenotype lead to the conclusion that pilus-mediated twitching motility is not required for natural transformation (19,25). This corresponds to our recent finding, that pilus structures on the cell surface are not essential for natural transformation (19,25)

DISCUSSION
PilQ in its monomeric state is only ~83 kDa and twelve to fourteen copies assemble to a remarkably long structure spanning the outer membrane and half of the periplasm (18,20). The complex has distinct structural features such as rings and gates, but the domains in the monomer responsible for these functions were so far elusive. Here, we provide the first insights into the topology of this macromolecular complex. Variants of PilQ described in this manuscript allowed us to identify the ring-and gate-forming motifs, based on SP-EM and in situ by cryo-ET. In a previous study, we identified an unusual N0 ringforming domain (ααβαββα fold) in PilQ (19). However, all previous attempts to identify further ring-forming domains (N1 to N5) by nested deletions beyond the N0 domain have failed, as mutants did not assemble PilQ complexes (19).
Here, we show that in the presence of the N0 domain, internal deletion derivatives of PilQ can still assemble into complexes (Fig. 2). This indicates that none of the rings are individually essential for complex assembly, which subsequently enables further detailed molecular analyses.
All PilQ variants were unstable, indicated by multiple bands on SDS-PAGE. Complexes generated from ΔN1 and N1Δβ 1 β 2 deletion derivatives were least stable, indicating that the β 1 β 2 subdomain of N1 plays a major role in PilQ complex stabilization. Cryo-ET of the in situ T4P machinery in the ΔN1 mutant revealed that the complexes were devoid of gate 2. This suggests that the unusual βββαβ fold in the N1 domain plays a dual role in ring and gate formation. Gate 2 might act as a seal to prevent leakage of periplasmic components into the secretin complex. This is supported by the observation of an in situ "leaky" phenotype (Fig. 5G). The β 1 β 2 β 3 αβ 4 fold is likely highly dynamic, supporting our earlier data demonstrating major conformational changes in the region of gate 2 (18). Conformational changes in gate 2, triggered by interactions with growing T4P, are suggested to allow extrusion of T4P.
SP-EM analysis of a set of PilQ deletion derivatives revealed that four distinct βαββα domains form the rings N2 to N5. These βαββα folds have also been detected in other ring-forming domains of secretin family proteins (40,41) and are similar to hnRNP K (nuclear ribonucleoprotein complex, K protein) homology (KH) domains. KH domains consist of a three-stranded β-sheet packed against two or three α-helices and are involved in DNA binding (40,41). A role of PilQ in DNA binding is also supported by the finding that a pilQ mutant was defect in DNA binding and uptake (23). Therefore, it is tempting to speculate that the βαββα domains in PilQ might be involved in DNA binding. However, the typical GXXG motif required for DNA interactions, which is present in the KH domain of the secretin HofQ and in eukaryotic proteins such as the neuronal splicing factor Nova-1 (41,42), is missing in PilQ.
The diameter of the PilQ rings differ significantly (~10 nm at N5 to ~16 nm at N0) in the closed state, as observed by both cryo-ET and SP-EM (18)(19)(20). These structural differences may be due to different mutual orientations of the domains, as seen for T2SS and T3SS (2). The invagination in ring N5 may form a stabilizer for the pilus within the central channel of PilQ.
Several T2SS and T3SS secretins have been studied by EM to date, revealing a number of common features. These include a cylindrical arrangement of 12-15 subunits per ring, and a large periplasmic vestibule comprising 2-4 rings, closed at the periplasmic end by one gate (2). The structure of the T. thermophilus secretin complex differs significantly from all other known secretin complexes by its composition of six stacked rings. Moreover, in T. thermophilus the PilQ channel was found to be sealed by a unique second N-terminal gate located in the N1 ring (18). These features may be a result of adaptation to the increased thickness of the cell envelope, which might be due to adaptation to the extreme environment T. thermophilus thrives in.
The ΔN1 domain deletion or the N1 subdomain deletion N1Δβ 1 β 2 abolished twitching motility, whereas deletion of the N1β 3 αβ 4 subdomain of the N1 fold had only a minor effect but changed the structure of the PilQ complex dramatically. The structural changes observed in this complex formed by the N1β 3 αβ 4 variant might be the result of a fusion of the N0 ring to the remaining N1 domain. This suggests that the interaction between the N1 β 1 β 2 subdomain and the T4P is important in determining pilus dynamics. We hypothesize that the first two β-sheets of the N1 ring might act as a lever controlling the opening and closing of the gate, analogous to VirB of the T4SS of Agrobacterium tumefaciens (43).
In conclusion, we find that secretins share many common features, such as cup forming domains, but also significant architectural differences, such as the number and shape of the rings and the number and localization of gates. The structure/function analyses of the PilQ mutants described here have helped to unravel the structural role of distinct domains (Fig. 8). These data provide a first insight into the function of distinct substructures of a secretin complex playing a dual role in T4P dynamics and DNA uptake. Our findings are an important milestone in the elucidation of the mechanism of secretin-mediated T4P extrusion and pilus dynamics.
Membrane preparation -Membranes were isolated as described recently (33).
Transformation assays -Transformation frequencies were calculated as transformants per living count as described previously (34). Five µg chromosomal DNA of a spontaneous streptomycin resistant T. thermophilus mutant was used for transformation assays.
Twitching motility studies -T. thermophilus strains were grown for 3 days on minimal medium containing 1 % bovine serum albumin (BSA). The plates were stained with Coomassie Blue. T. thermophilus cells were detached from the agar and the twitching zones were defined as the resulting colorless areas Assays were performed as described previously (24).
Investigation of piliation phenotypes by electron microscopy The piliation phenotype of T. thermophilus cells was analyzed by electron microscopy as described recently (25). Cells were transferred to copper grids and shadowed at a unidirectional angel of 25° with platinum-carbon. The amount of pili was counted from 300 randomly selected cells.
Electron microscopy and image processing -PilQ complexes were diluted to a final concentration of 0.02 mg/ml and negatively stained with 2 % (w/v) ammonium molybdate. 3 µl of protein sample were applied onto freshly glow discharged carbon coated copper grids. The sample was reduced to a thin film by blotting and the staining solution was then immediately applied to the grid and blotted off from the same side. The negative stain process was repeated three times. Electron micrographs were collected using a FEI Tecnai G2 Spirit, operated at 120 kV at a nominal magnification of 28,000 x. Images were collected on a Gatan 4k x 4k CCD camera with a pixel size of 4.2 Å with a defocus range of -0.2 to -1.0 µm. Particles were selected using the boxer module in EMAN (35). Image processing and reference free two-dimensional classification were performed with Xmipp 3.1 using the classification algorithm CL2D (36) and with RELION (37) Electron Cryo-tomography and Subtomogram Averaging -Cryo-ET and subtomogram averaging were performed as previously described (18), with the following exceptions. Tomograms were collected at tilt steps of 2.5° and 7 µm defocus, using a JEM-3200FSC microscope (JEOL, Tokyo, Japan), equipped with a field emission gun operating at 300 keV and an in-column energy filter operated with a slit width of 40 eV. Tomograms were collected with a K2 Summit direct electron detection camera (Gatan, Pleasanton, USA) at a nominal magnification of 10,000x, corresponding to a pixel size of 3.9 Å. The total dose per tomogram was ~140e -/Å 2 . For subtomogram averaging, coordinates corresponding to the outer membrane and inner membrane domains of the T4P machinery were marked manually in IMOD from three-times binned tomograms and averaged using PEET (38). The final average was obtained from 432 particles (C12 symmetry applied), using a mask drawn around PilQ. The resolution estimate is ~50 Å using conventional 'even/odd' Fourier shell correlation (FSC), applying the 0.5 FSC criterion (Fig. 6). A mask was drawn around the protein to exclude the membrane from this estimate. Surface views were drawn with UCSF Chimera (39). For comparisons with the wild-type complex, data was used which has been previously published (18) and is now displayed in a different way.     PilQ complexes were negatively stained with 2 % ammonium molybdate and analyzed by EM. Images were taken at high defocus (-1.5-2.5 µm) to give an overview over the samples. In all cases the purifications yielded protein complexes that did not aggregate. However, the number and the shape of the complexes assembled from the deletion derivatives varied considerably. Scale bar, 100 nm.
In panels B and D-I reference-free class averages of a dataset of 9896 (PilQ wt ), 6648 (N1Δβ 1 β 2 ), 7109 (N1Δβ 3 αβ 4 ), 1568 (ΔN2), 5000 (ΔN3), 2028 (ΔN4) and 1974 (ΔN5) particle images revealed PilQ complex formation. Representative class averages are shown for each PilQ complex. The number of particles images per class for PilQ wt was 726, for N1Δβ 1 β 2 949, for N1Δβ 3 αβ 4 726, for ΔN2 341, for ΔN3 900, for ΔN4 and ΔN5 439, respectively. In C a model of the PilQ wt complex is shown. Scale bar, 10 nm.   Evaluation of the piliation and twitching motility data. Panel A shows the piliation of the pilQ mutants. In total 300-350 cells per strain were analyzed, and this value was set to 100 %. The percentage of non-piliated cells is shown in white, cells exhibiting 1-3 pili are colored light gray, cells exhibiting 4-7 pili are shown in dark gray, and cells exhibiting more than 8 pili are shown in black. Panel B shows the twitching motility of the pilQ mutants. The diameter of the twitching zones of wild-type T. thermophilus HB27 was set to 100 %.