Structure of the Neisseria meningitidis outer membrane PilQ secretin complex at 12 A resolution.

The bacterial pathogen Neisseria meningitidis expresses long, thin, retractile fibers (called type IV pili) from its cell surface and uses these adhesive structures to mediate primary attachment to epithelial cells during host colonization and invasion. PilQ is an outer membrane protein complex that is essential for the translocation of these pili across the outer membrane. Here, we present the structure of the PilQ complex determined by cryoelectron microscopy to 12 A resolution. The dominant feature of the structure is a large central cavity, formed by four arm features that spiral upwards from a squared ring base and meet to form a prominent cap region. The cavity, running through the center of the complex, is continuous and is effectively sealed at both the top and bottom. Analysis of the complex using self-orientation and by examination of two-dimensional crystals indicates a strong C4 rotational symmetry, with a much weaker C12 rotational symmetry, consistent with PilQ possessing true C4 symmetry with C12 quasi-symmetry. We therefore suggest that the complex is a homododecamer, formed by association of 12 PilQ polypeptide chains into a tetramer of trimers. The structure of the PilQ complex, with its large and well defined central chamber, suggests that it may not function solely as a passive portal in the outer membrane, but could be actively involved in mediating pilus assembly or modification.

N. meningitidis is the causative agent of meningococcal meningitis and septicemia. In common with many other Gram-negative pathogens, including Neisseria gonorrhoeae and Pseudomonas aeruginosa, it constitutively produces long, thin fibers, termed type IV pili (Tfp) 1 (1). The expression of Tfp is vital to the pathogenicity of these organisms, and they play a pivotal role in cellular adhesion to host epithelial cells. Neisseria also utilizes Tfp in several diverse processes, including cellular agglutination (2), twitching motility (3), and DNA uptake during transformation (4). Tfp are composed predominantly of a helical polymer of the pilin subunit PilE; a model for the helix with five pilin subunits/turn has been proposed based on the crystal structure of PilE (5). Tfp are dynamic structures, capable of rapid extension as well as retraction and can be subjected to substantial motor forces in the region of 100 piconewtons (6). The Tfp biogenesis system can therefore be thought of as a highly sophisticated molecular machine, capable of working in two opposite directions. Formation of Tfp is a complex process that involves at least 12 different proteins, but PilQ is the only protein component that is an integral outer membrane protein. Expression of the PilQ complex is mandatory for Tfp formation, and inactivation of the pilQ gene in Neisseria sp. leads to a non-piliated phenotype (7,8). PilQ is a member of a family of outer membrane proteins termed secretins, which are involved in secretion of a variety of substrates from Gram-negative bacteria. There is a significant level of sequence homology between the C-terminal regions of the PilQ monomer and the Klebsiella oxytoca PulD secretin involved in the type II secretion system (9) and the pIV protein, which plays a vital role in assembly of filamentous bacteriophage (10). This observation has led to suggestions that the secretins may function in a similar way, by forming a passage through the outer membrane for the secreted substrate. Purification and subsequent studies by electron microscopy of PulD, pIV, and PilQ have shown that they form high molecular mass complexes (9 -12). Transmission electron microscopy (TEM) images in negative stain show ring-like structures in projection with a central stained feature, interpreted as a cavity, measuring up to 75 Å in diameter. This observation has lent weight to the hypothesis that the secretins act as a conduit for passage of the secreted substrate. There is also evidence from studies of reconstituted PulD that the channel formed by the secretin is gated, suggesting that the pore formation is regulated (9). The K. oxytoca PulD oligomer was shown to exhibit C12 rotational symmetry, and a three-dimensional model of the structure of the complex produced from face-on and side-on views of the complex has been presented. The structure is composed of a central channel with a minimum diameter of 70 Å, enclosed by a cylindrical protein complex with radial spokes. More recent work on the pIV protein complex used cryoelectron microscopy to produce a full threedimensional structure of the secretin at 22 Å resolution (10). Interestingly, this structure had C14 (rather than C12) symmetry and appears to have a blockage in the central pore.
Visualized under negative stain TEM, preparations of purified PilQ have a doughnut-like appearance with an apparent cavity ϳ60 Å in diameter (12). Further analysis by single particle averaging methods suggested that the complex has C12 symmetry, and a three-dimensional reconstruction from a sample in negative stain at 25 Å resolution shows a bowl-like structure (11). To obtain more detailed structural information on the PilQ complex, we have applied cryo-negative staining in conjunction with single particle averaging methods to extend the resolution of the structure considerably to 12 Å. These data are complemented by supporting results obtained from analysis of two-dimensional crystals of the reconstituted PilQ complex. The PilQ characterization at 12 Å resolution is the highest level of structural detail yet obtained for a protein from the secretin family; the results are not easily reconciled with a model for PilQ as a passive sheath within the outer membrane and suggest that the protein could play an active role in pilus assembly or modification.

MATERIALS AND METHODS
Bacterial Strains and Growth Conditions-Meningococcal strain M1080 was grown overnight on 5% blood agar in an atmosphere containing 10% CO 2 before harvesting. The M1080 pilQ gene sequence is available in the GenBank TM /EBI Data Bank under accession number AJ564200. The plasmid pMF121 (generously provided by Dr. Mathias Frosch) was used to generate a capsule-negative mutant also expressing truncated lipo-oligosaccharide of N. meningitidis strain M1080 by transformation with erythromycin selection (11).
Purification of Oligomeric PilQ-Meningococcal cells were disrupted in a French press in the presence of protease inhibitors. The oligomeric PilQ secretin complex was isolated from meningococcal cells using the Zwittergent 3-10 purification procedure described previously (11). The efficiency of membrane solubilization and subsequent yields of PilQ were increased using capsule polysaccharide-deficient bacteria. PilQcontaining fractions were selected by immunoblotting using rabbit polyclonal antibodies raised against the purified N. meningitidis PilQ complex from strain M1080 (7). The conditions used to verify PilQ sample purity by SDS-PAGE and immunoblotting with antibodies against PilE, PilP, PilC, and lipopolysaccharide have been described previously (7).
Cryo-negative Staining-Variations on previously reported methods (13) were used to prepare PilQ samples for cryoelectron microscopy. This method involves suspending the protein in a thin solution containing trehalose with 5-10% (w/v) ammonium molybdate and then recording data under low temperature conditions. The use of low temperature preserves the sample in a hydrated condition, and the addition of molybdate provides a significant gain in contrast enhancement of the sample over more traditional cryological methods. 20 l of PilQ (100 g/ml in 10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, and 0.1% (w/v) Zwittergent 3-10) were mixed with 20 l of freshly prepared 5-10% (w/v) ammonium molybdate containing 1% (w/v) trehalose. Carbon-coated copper grids or holey carbon grids (No. 400) were placed on the surface of the droplet for several seconds, briefly blotted, and immediately frozen in liquid N 2 to ϳ100 K in an Oxford System cryo-stage. Preparation of PilQ in this fashion routinely suspended PilQ particles in a layer of stain/cryoprotectant ϳ300-500 Å thick. Table I summarizes all additional information pertinent to cryo-TEM low dose data collection, micrograph scanning, and volume calculation used in this study.
Two-dimensional Crystallization of PilQ and Crystal Analysis-Twodimensional crystals of oligomeric PilQ were obtained by reconstitution into lipid bilayers. 50 l of PilQ (100 g/ml in 10 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, and 0.1% (w/v) Zwittergent 3-10) were mixed with 1,2-dimyristoyl-sn-glycero-3-phosphatidylcholine to provide a final protein/lipid ratio of 10:1. Hydrophobic polystyrene beads, or Bio-Beads, have been shown to provide a surface for nonspecific adsorption of detergent and are a convenient method for detergent extraction (14). Hydrated SM2 Bio-Beads (Bio-Rad) were added to the PilQ/1,2dimyristoyl-sn-glycero-3-phosphatidylcholine solution, and the mixture was incubated at 20°C for 24 -48 h. Large two-dimensional crystal sheets of PilQ were formed at the air-water interface of the solution; the crystals were examined using TEM by gently applying glow-discharged carbon-coated copper grids to the surface of the solution, followed by negative staining with 4% (w/v) uranyl acetate as described previously (11). TEM was then performed under low dose conditions on a Tecnai 10 transmission electron microscope operating at 100 kV. Images were recorded on Kodak SO-163 film and digitized on a UMAX Power Look 3000 densitometer at 4.81 Å/pixel at the specimen level. Lattice unbending and contrast transfer function corrections were applied as described (15) using the MRC-LMB software suite (16,17). Projection maps were generated using CCP4 software (18).
Three-dimensional Structure Calculation-The three-dimensional volume of PilQ was calculated using common lines projection matching methods employed in EMAN (19). EMAN requires a preliminary model to calculate a three-dimensional structure and then implements a model-based iterative refinement process to generate the final structure. To generate the preliminary model, a series of reference-free class averages corresponding to different views of the top and sides of a particle must first be generated from the data set. An initial data set of 7588 PilQ particles was interactively selected using the graphics interfaces of either SPIDER (WEB) (20) or EMAN (BOXER) (19). The contrast transfer function for each particle in the data set was determined, and a correction was applied (21). Data were then converted into a format suitable for EMAN with contrast normalization applied. The selected particles were band pass-filtered to 8 Å and centered, and a set of reference-free class averages corresponding to a range of particle orientations was generated. Following established procedures in the EMAN software suite, a preliminary three-dimensional model was determined from class averages that represented distinct views of the PilQ particle. Volumes were then separately calculated with either C1, C4, or C12 symmetry applied. The relative orientations of the characteristic views were determined using a Fourier common lines routine, and the averages were combined to generate the preliminary threedimensional model, the starting point for the main refinement loop. The class averages generated in the refinement loop were produced by a projection matching routine, whereby projections with uniformly distributed orientations of the preliminary three-dimensional model were used as references for classification of the raw data set, with the class averages from this step used to construct a new three-dimensional model. This is an iterative procedure with convergence assessed by examining the Fourier shell correlation (FSC) of the three-dimensional models generated from each iteration. The final three-dimensional volume with C4 symmetry applied was fully converged after six rounds of iterative refinement. To assess that the correct structure had converged, several different starting models were used. For a C4 symmetry reconstruction, the EMAN starting model is automatically determined from the top and side views, whereas for C1, these views must be selected by the user; in both cases, equivalent structures were generated regardless of the starting model. As an additional check, an unsymmetrized three-dimensional volume, previously calculated from a separate data set (11), was used as a starting model and produced a similar convergence.
Symmetry Analysis of PilQ-Rotational symmetry in volumes was initially assessed in EMAN using STARTCSYM and independently validated in SPIDER (20) using a variation on rotational power methods (22,23). For the real space three-dimensional correlation rotation analyses, the volumes obtained with C1 or C4 symmetry applied were rotated around the central symmetry axis by a specified angle using the program MAPROT (24), and the density map correlation coefficient was calculated using OVERLAPMAP (25), both from the CCP4 suite (18).
Resolution Determination and Calculation of Variance Maps-Resolution was determined by FSC analysis by comparing the correspondence in reciprocal space of two subaverages generated from half of the final data set (i.e. 3794 particles) (20). Using the same subvolumes, a variance map for the final three-dimensional structure was calculated to reveal regions of the structure with a high degree of flexibility.

Data Collection and Single Particle
Classification-A previous electron microscopy study on the PilQ oligomer was performed at room temperature on air-dried protein samples that had been negatively stained with uranyl acetate (11). The three-dimensional structure generated from these data, using the random conical tilt method, was limited to 25 Å resolution. Under these conditions, the PilQ oligomer adopted a preferential orientation with respect to the carbon support film and presented a toroidal appearance in projection. In this study, we used the relatively new technique of cryo-negative staining to extend the resolution and detail of the molecular envelope (13). This method provided several advantages over data collected previously with negatively stained PilQ recorded at ambient temperatures. Cryo-negatively stained PilQ particles adopted previously unobserved multiple orientations within the staining layer (Fig. 1A). The distinctive high contrast ring-shaped structures were readily identified, but additional projection profiles could also be observed. The views corresponded to the PilQ complex orientated in different directions from those observed previously (Figs. 1A and 2). This combination of multiple views of the PilQ complex with good signal/noise ratio allowed us to implement the Fourier common lines projection matching method (19) to determine the PilQ complex threedimensional structure. Fig. 1B illustrates the Fourier transform calculated from the cryo-negatively stained PilQ micrograph in Fig. 1A, showing concentric Thon rings reaching beyond 8 Å resolution. In principle, it is reasonable to anticipate a three-dimensional structure determination to ϳ13 Å resolution or better, if a sufficiently large and accurately aligned data set can be obtained. Fig. 2 shows a selection of contrast-enhanced PilQ particles with contrast transfer function correction applied. Examples are shown of reference-free class averages that were obtained when the raw data were aligned and are representative of the range of multiple orientations sampled. The "top" view of the PilQ complex, which displayed the highest Cn rotational sym-metry, has a distinctively square appearance in projection, and consequently, it was readily distinguished from partial top and side views. A small amount of positive density is apparent in the center of the projection, probably originating from the cap feature (see Fig. 6). The PilQ complex has a maximum diameter of 145 Å when viewed from this orientation, slightly smaller than the equivalent dimension (ϳ160 Å) derived from PilQ samples stained with uranyl acetate. This is most probably due to the fact that conventional negative stain often produces some sample flattening during drying (11,12). The majority of the data consisted of particles in intermediate orientations between the top and side views; an example is shown in Fig. 2 (middle panels). Particles with this type of orientation often had a slightly elliptical or flattened ring appearance. Examples of side views are shown in Fig. 2 (lower panels). For the side orientation, the central cavity of the complex still forms a high contrast feature, but the projection is more conical in appearance, with a rounded base. The side view has three distinct bands running across the projection, a feature that is shared by the PulD and pIV secretins (9,10).
Rotational Symmetry of Oligomeric PilQ-A three-dimensional volume for the PilQ complex was initially calculated using common lines projection matching methods employed in EMAN (19), assuming no rotational symmetry (i.e. C1). The three-dimensional volume was then examined for rotational symmetry by calculation of the map correlation coefficient (25) as a function of the angle of rotation about the main axis through the middle of the PilQ complex (Fig. 3). The results provide clear evidence for 4-fold rotational symmetry, with peaks of approximately equal height at 90°, 180°, and 270°. This conclusion was confirmed by calculation of two-dimensional class average projection maps of the unsymmetrized complex along the long axis of rotational symmetry and determination of the angular cross-correlation for each rotational symmetry (C1-C14) in Fourier space (data not shown) (26). The pronounced 4-fold rotational symmetry was also readily apparent from the square-shaped projections of individual particles when viewed from the top (Fig. 2). There was an excellent correlation between the example class averages where C4 symmetry had been applied and those without any rotational symmetry imposed. This correlation was not apparent when other higher rotational symmetries were applied (data not shown). This observation confirms that the application of C4 symmetry in the processing of the data does not introduce features that were not present in the unsymmetrized data set. The data set FIG. 2. PilQ particle selection and classification. Shown is a montage of representative PilQ particles (60-pixel mask) present in three of the different orientation classes. The resulting two-dimensional class averages for each of the particle classes following alignment are shown with the corresponding two-dimensional back-projections from the preliminary three-dimensional (3-D) reconstruction with C4 symmetry imposed. Averages were calculated using the software package EMAN (19). Scale bar ϭ 100 Å.
was also analyzed for possible D symmetry, but none was found.
Analysis of Two-dimensional Crystals of the Reconstituted PilQ Complex-Previous work on PilQ (11) and other related secretins (9, 10) has provided evidence for 12-and 14-fold rotational symmetries in these complexes. To establish unambiguously that C4 symmetry is appropriate, two-dimensional crystals of PilQ were obtained by reconstitution of the protein into lipid membranes (Fig. 4A). Mixtures of the detergentsolubilized PilQ complex and 1,2-dimyristoyl-sn-glycero-3phosphatidylcholine at a high protein/lipid ratio were incubated with Bio-Beads, resulting in the formation of very large two-dimensional crystalline membrane sheets (Ͼ5-m diameter). Analysis of Fourier transforms from these two-dimensional crystal membrane sheets revealed that they consisted of two overlapping crystalline layers. The better stained lower layer was deconvoluted from the upper layer, and two-dimensional crystal analysis was performed by selecting small, well ordered areas. Examination of the two-dimensional crystals revealed that the tightly packed lattice was composed of square particles, with one PilQ oligomer present in the unit cell (a ϭ 146 Å, b ϭ 146 Å, ␥ ϭ 90°). A computed Fourier transform of a well ordered area of the PilQ crystal subjected to three rounds of lattice unbending is shown in Fig. 4B. The transform shows reflections extending to the sixth order, corresponding to an approximate spatial resolution of ϳ24 Å. Analysis of the calculated structure factors using the MRC-LMB software ALL-SPACE (17) identified only p2 and p4 plane groups as acceptable, with p4 the most probable. This assessment was based on using the observed phase residual versus the target phase residual for a specified plane group based on statistics accounting for Friedel weight. Since there is no evidence for a dimeric arrangement of the PilQ complex, the p2 plane group was discounted. Calculated projection maps without (p1) and with (p4) crystallographic symmetry applied are shown in Fig. 4 (C  and D). The p4 symmetrized projection is entirely consistent in appearance and dimensions with the unsymmetrized single particle averaging top view of the PilQ complex (Fig. 2, upper panels) and shows a slightly rounded square particle with a large density in the center of the complex. Furthermore, these data support the assignment of C4 symmetry to the single particle averaging data set.
Three-dimensional Structure Calculation and Resolution Assessment-The large particle data set and imposition of C4 symmetry produced a complete and converged three-dimensional structure; matrix analysis of sampling over Euler angles , , and showed a good coverage of three-dimensional space (data not shown), as would be expected for an approximately random set of particle orientations. Fig. 5 shows the resolution estimation by the FSC method for the final volume. For this procedure, two separate three-dimensional structures were created from either even-or odd-numbered particles, so each structure was calculated from half of the full data set. The extent to which these volumes are identical was then compared in Fourier space by correlation analysis (26). The surface-rendered side views of these even and odd subvolumes used for the final resolution calculation are also shown in Fig. 5. Only minor differences in the surface-rendered volumes were apparent, even when only half of the data were used to calculate them, demonstrating the fidelity of the structure and the sampling of data. Previous estimates of resolution limits have recommended an FSC threshold of 0.5 (26,27), although this may be a conservative assessment. An argument for an FSC threshold of 0.33 has recently been made (28) as a way of compensating for using only 50% of the available data in the subvolumes. FSC thresholds of 0.5 and 0.33 gave resolution estimates of 12 and 10 Å, respectively, for the data used here. These estimates are consistent with the expected instrument limits; visible Thon rings to beyond 8 Å were observed in Fourier transforms of images. Similar resolution limits have been reported previously for cryo-negative staining using single particle averaging methods (29 -31) and two-dimensional crystals (13).
Three-dimensional Structure of the PilQ Complex-Three views of the three-dimensional reconstruction of the PilQ complex are shown in Fig. 6A. Viewed from the side, the PilQ complex can be perceived as four distinct structural portions, which we have named the "plug," "ring," "arms," and "cap." Assuming a protein packing density of 0.73 g/ml (32), the volume at 1 above the mean density would accommodate a FIG. 3. Real space three-dimensional correlation rotation analysis of the unsymmetrized three-dimensional volume. Correlation coefficients were calculated from a three-dimensional volume produced with no rotational symmetry averaging applied (C1); this volume was calculated using procedures described previously (19).

FIG. 4. Analysis of reconstituted PilQ two-dimensional crystals.
A, a selected area from a large, well ordered region of two-layer overlapping two-dimensional crystals of PilQ reconstituted into 1,2dimyristoyl-sn-glycero-3-phosphatidylcholine. Data were contrast transfer function-corrected and contrast-enhanced for presentation. Scale bar ϭ 1000 Å. B, computed Fourier transform of a selected section of a two-dimensional crystalline area following lattice unbending. The crystallographic axes (h, k) in reciprocal space are indicated, and the size of the boxes provide an indication of the reflection signal/noise ratio (17). C and D, two-dimensional projection maps calculated from the Fourier transform shown in B in space groups p1 and p4, respectively. Scale bars ϭ 100 Å. molecular mass of ϳ800,000 Da. This figure is similar to that obtained from calculations based on the lower resolution volume of the PilQ complex reported previously (11). The plug region at the bottom of the complex is formed by a blunt point (30 Å wide), which gradually tapers upwards into a ring region with a diameter of 110 Å and a height of 40 Å. Emerging from the top of the ring, four arms coil over a height of 65 Å and meet to form a dome-shaped cap, which is 65 Å in diameter and 33 Å in height. Viewing the complex from the top shows that the arm and cap portions adopt a square-based pyramid shape extending from the ring base. Estimation of the volume proportions of each of these structural features gave values of 36% for the ring and plug sections combined, 48% for the arms, and 15% for the cap feature. The extent of the central cavity and its internal features within the complex are shown in Fig. 6B. The cavity has an overall height of 90 Å. At its widest point, it is 87-Å wide, and from there, it tapers upwards to join the cap region. The volume of the cavity is 377,000 Å 3 , determined at a density threshold of 1 above the mean density level. To provide a sense of scale, by comparison with other protein complexes with large internal voids, this value is larger than the volumes of the archaebacterial thermosome (260,000 Å 3 ) (33) or the cis-cavity of the GroE complex (175,000 Å 3 ) (34).
Five discrete slices through the PilQ complex, perpendicular to the C4 symmetry axis, are shown in Fig. 6B (right panels). The 4-fold symmetry relationship is particularly apparent in the arm and cap regions, but is less clear in the ring and plug sections of the map. A variance map was calculated to illustrate the regions of greatest structural deviation between the two subvolumes used in the resolution assessment (Fig. 6C). The highest variance, shown at 8 above the mean variance, is at the center of the structure in the cavity region surrounding the small central density. The high variance in this portion of the volume may be due to strong deviations from C4 symmetry or multiple conformations; this feature may explain why the mass appears to be unconnected to the rest of the PilQ structure. Other regions of the map that display considerable variation correspond to the section of the structure where the four arm portions join to the cap. The density is weaker at this point, and the high variance could be due to a higher flexibility of the protein in this region. DISCUSSION The structure of the PilQ complex described above has been determined to the highest resolution yet reported for an outer membrane secretin protein. The consequently higher level of structural detail has revealed new features of the PilQ complex that have important implications for the function of this class of outer membrane proteins. The structure of the PilQ complex differs in several important respects from PulD from K. oxytoca (9) and the phage pIV multimer (10), although all three proteins exhibit significant levels of sequence homology within their C-terminal domains. A three-dimensional model for the structure of PulD was determined from face-on and side-on views of unstained hydrated frozen images and showed an approximately cylindrical structure with C12 symmetry (9). Although a channel was visible through the entire length of the protein complex, conductance measurements indicated that the pore was gated. PulD is one of a complex of 15 proteins that form the type II secretion machinery, or secreton, and mediates the secretion of the enzyme pullulanase from K. oxytoca. The close similarity between the type II secretion pathway and Tfp biogenesis has been demonstrated by the observation that the pullulanase secreton components are capable of incorporating the type IV pilin protein PpdD into pili (35). The observation that PilQ and PulD differ significantly in their overall structure suggests a high degree of specialization in the architecture of the N-terminal part of the outer membrane protein components of the two respective secretory systems. The three-dimensional structure of the PilQ complex also contrasts with that of the phage pIV multimer, which mediates the secretion of filamentous bacteriophage across the outer membrane. pIV exhibits C14 (rather than C12) symmetry and has a central channel 60 -88 Å in diameter, which is blocked in the center by a slab of density associated with the M ring of the complex (10). Thus, in contrast to the PilQ complex, the pIV multimer presents two open cavities to the inner (periplasmic) and outer faces of the outer membrane.
Our earlier work on the structure of the PilQ complex proposed that it is formed from 12 PilQ polypeptide chains (11); the argument was based on the overall volume of the complex, which indicated an enclosed molecular mass of between 800,000 and 900,000 Da, and also on the observation of 12-fold symmetry in the rotational power spectrum. Analysis of the unsymmetrized three-dimensional volume of PilQ in this study showed evidence of 4-fold symmetry (Fig. 3) but no evidence of 12-fold symmetry by regular peaks at 30°intervals. Analysis of the two-dimensional crystals of the reconstituted PilQ complex clearly supported the finding of 4-fold symmetry. Our previous work on the projection structure of the PilQ complex in uranyl acetate negative stain showed that the peaks for 4-fold symmetry were higher than those for 12-fold symmetry (12). To investigate the possibility that the 12-fold symmetry peaks were obscured by noise, calculation of the map correlation coefficient was repeated on the three-dimensional volume that had been generated with C4 symmetry applied. The result showed weak peaks at 30°and 60° (Fig. 7). The relatively high peak for 12-fold symmetry obtained from uranyl acetate negative stain data may be due to an uneven staining of the PilQ complex and compression around the periphery. The processed uranyl acetate stain data set consisted exclusively of circular ring-shaped particles, much more consistent with the slightly tilted views of PilQ presented here (Fig. 2, middle panels). This would suggest that this previous three-dimensional structure, although essentially correct, was actually calculated from particles partially tilted away from the true C4 top view. A meningococcal PilQ monomer has a predicted molecular mass of 82,000 Da, so a tetramer would have a mass of 328,000 Da, an octamer a mass of 656,000 Da, and a dodecamer a mass of 984,000 Da. A calculation based on the volume of the complex determined here indicated that it had a molecular mass of ϳ800,000 Da, a figure clearly incompatible with a tetramer. On the basis of the 12-fold symmetry we have observed (12), we propose that the PilQ complex is composed of 12 subunits within a complex of overall 4-fold symmetry (i.e. C4 symmetry with C12 quasi-symmetry). This implies that there is confor- mational flexibility in the PilQ monomer, with three different conformations. The unique region of repeated octamers in the N-terminal domain of the PilQ monomer (7) could facilitate this flexibility. Such quasi-symmetry is well characterized for viral capsids, but has been described relatively infrequently outside this group of structures. We note, however, that quasi-symmetry is a feature of some other membrane protein structures, including the EmrE homodimer (36), the BetP trimer (37), the polysaccharide K antigen transporter Wza (38), and also the membrane-associated "adaptor" protein PspA (39). There is also evidence that the quaternary structure of the PapC usher involves quasi-symmetry (40). Quasi-symmetry within the PilQ oligomer could contribute to an explanation of why at least two proteins, PilP (41) and Omp85 (42), have been implicated in its assembly or stability.
The three-dimensional model for the PilQ complex presented here was examined for consistency with the model for the complex at 25 Å resolution obtained using uranyl acetate negative stain (11). Recalculation of the original data assuming C4 (rather than C12) symmetry produced a three-dimensional model with overall dimensions, including the internal cavity, that were essentially the same as those for the higher resolution model presented here (data not shown). The main feature of the structure at low resolution was the sealing of the cavity at the bottom, which was clearly visible in the recalculated map. A small (Ͻ8-Å diameter) spherical density was apparent, suspended above the top of the symmetrized volume. It is now clear that this feature probably corresponds to the cap region, but it was present only at low (Ͻ1) density thresholds, and its significance was unclear at the time. The arm regions were not resolved in the lower resolution map because of the limited range of projections sampled.
The new features of the structure of the PilQ complex described here have implications for the proposed mechanisms of Tfp biogenesis. The ring-like appearance of many secretins, as observed by TEM, has led to the inference that they act as a portal or gateway across the outer membrane for the secreted substrate. A structural model for the Tfp fiber from N. gonorrhoeae, which is closely related to that from N. meningitidis, has been deduced from the x-ray crystal structure of the pilus fiber subunit (5). The diameter of the pilus fiber is 65Å, with a helical pitch of 41 Å. The cavity within the PilQ complex is therefore large enough to accommodate the Tfp, although extension of the fiber outside the cavity would require a large structural change in the cap or plug region. It is interesting to note that the central cavity contains a small, approximately spherical density in the center (Fig. 6B). This feature was still present at high thresholds (5.5), but contributed only to ϳ1% of the total volume of the complex and appears to be uncon-FIG. 6. Three-dimensional structure of PilQ at 12 Å resolution. A, surface-rendered side, top, and bottom views of the PilQ complex with C4 symmetry applied. Volumes were low pass-filtered to 8 Å resolution, and high frequency signals were sharpened using a Fermi high pass filter with filter radii and temperature factors of 0.05 each. The volume is displayed at 2 above the mean density in purple. Scale bar ϭ 100 Å. B, two-dimensional projection slices through the C4symmetrized PilQ volume (6.8 Å thick) shown vertically (left panel) and horizontally (right panels) at the locations indicated on the first side view slab. Scale bar ϭ 100 Å. C, variance maps in green of the PilQ complex displayed at 5 and 8 showing the regions of high variation within the three-dimensional structure. For clarity, variance maps were first low pass-filtered to 18 Å resolution, and variance is displayed at 1 above the mean variance. Scale bar ϭ 100 Å.
FIG. 7. Real space three-dimensional correlation rotation analysis of the three-dimensional volume with C4 symmetry applied. Correlation coefficients were calculated from the C4-symmetrized three-dimensional volume. Other details were as described in the legend to Fig. 3. nected to the envelope. Phenotypic observations of the properties of pilus biogenesis mutants in diverse bacterial species have provided evidence that pilus formation occurs within the periplasm, rather than in the outer membrane (43,44). It is difficult, however, to reconcile the structure of the PilQ complex presented here with a role as a passive portal or pore within the outer membrane. Although the dimensions of the chamber within the protein are sufficient to accommodate at least one full helical turn of the Tfp, it is less clear how both the plug and ring sections could open out sufficiently to allow the passage of a pilus fiber measuring 65 Å in diameter. By contrast, the cap and arm portions of the structure appear to be more flexible (Fig. 6C) and could plausibly unfurl to allow passage of the pilus. The results could therefore indicate a novel role for the PilQ complex as a platform for pilus assembly, although such a role has yet to be demonstrated functionally. On the other hand, one cannot rule out the possibility that the cavity, under physiological conditions, might open into a pore to accommodate an already assembled pilus, in a similar fashion to the mechanism suggested for the pIV secretin (10). In this case, the PilQ complex could contribute to pilus modification as well as extrusion. Another point of uncertainty is the orientation of the PilQ complex relative to the outer membrane. It would appear more likely that the ring/plug section of the structure would be located within the membrane, but there is no direct evidence at present to establish whether the arm/cap section is orientated outwards or inwards, toward the periplasm, as indicated for the pIV secretin (10). Finally, it should be noted that the PilQ complex is likely to form part of a larger macromolecular assembly of several different proteins in vivo, all of which are lost during the purification procedure. There is some evidence to suggest, for example, that the assembly complex for type IV bundle-forming pili from enteropathogenic Escherichia coli associates in this fashion in vivo (45). At present, the manner in which other components of the Tfp biogenesis machinery associate with PilQ is unknown.