Molecular architecture of Streptococcus pneumoniae TIGR4 pili
2009; Springer Nature; Volume: 28; Issue: 24 Linguagem: Inglês
10.1038/emboj.2009.360
ISSN1460-2075
AutoresMarkus Hilleringmann, Philippe Ringler, Shirley A. Müller, Gabriella De Angelis, Rino Rappuoli, Ilaria Ferlenghi, Andreas Engel,
Tópico(s)Glycosylation and Glycoproteins Research
ResumoAlthough the pili of Gram-positive bacteria are putative virulence factors, little is known about their structure. Here we describe the molecular architecture of pilus-1 of Streptococcus pneumoniae, which is a major cause of morbidity and mortality worldwide. One major (RrgB) and two minor components (RrgA and RrgC) assemble into the pilus. Results from TEM and scanning transmission EM show that the native pili are approximately 6 nm wide, flexible filaments that can be over 1 μm long. They are formed by a single string of RrgB monomers and have a polarity defined by nose-like protrusions. These protrusions correlate to the shape of monomeric RrgB–His, which like RrgA–His and RrgC–His has an elongated, multi-domain structure. RrgA and RrgC are only present at the opposite ends of the pilus shaft, compatible with their putative roles as adhesin and anchor to the cell wall surface, respectively. Our structural analyses provide the first direct experimental evidence that the native S. pneumoniae pilus shaft is composed exclusively of covalently linked monomeric RrgB subunits oriented head-to-tail. Various types of filamentous surface appendages, pili, have been identified in Gram-negative and Gram-positive bacteria (Wu and Fives-Taylor, 2001). Pili fulfill manifold functions during bacterial life cycles, such as host cell invasion, biofilm formation, cell aggregation, DNA transfer and twitching motility (Proft and Baker, 2009). Their structure has to withstand both environmental stress and the activities of the host immune system. The role of pili as adhesive organelles is crucial to the survival of pathogenic bacteria, which have to attach to specific host cells for colonisation and to establish an infection. While many Gram-negative pili have been studied in detail over the last decades (Fronzes et al, 2008), the majority of Gram-positive pili have been discovered only recently and their study, initiated through pioneering work by Schneewind and co-workers on Corynebacterium diphtheriae pili (Ton-That and Schneewind, 2003; Ton-That et al, 2004), is in its infancy. In contrast to Gram-negative pili, which are typically formed by non-covalently linked subunits, Gram-positive pili are extended polymers assembled from covalently cross-linked pilin subunits and tethered to the cell wall peptidoglycan (reviewed by Ton-That and Schneewind, 2004; Telford et al, 2006; Mandlik et al, 2008b). As demonstrated for the major pilin subunit of C. diphtheriae, the conserved genetic requirements necessary for pilus formation include the pilin motif (WXXXVXVYPKN), the E-box domain (YXLXETXAPXGY) and the cell wall sorting signal (LPXTG), followed first by hydrophobic and then by charged amino acids (Ton-That and Schneewind, 2003, 2004; Ton-That et al, 2004). Mass spectrometric studies of pilus fragments of Bacillus anthracis have confirmed the existence of intermolecular amide bonds between the C-terminal threonine of cleaved sorting signals and the conserved lysine residue (YPKN) within the pilin motif (Budzik et al, 2008). However, the structure of the backbone pilin Spy0128 of Streptococcus pyogenes and mass spectrometric analysis of pilus fractions showed the isopeptide bond to link the threonine of the sorting signal EVPTG with a conserved lysine that is close to but not within the pilin-like motif (Kang et al, 2007). Sortases catalyse the reaction between the threonine of the LPXTG motif and the conserved lysine of the next backbone-forming protein (Marraffini et al, 2006; Manzano et al, 2008; Neiers et al, 2009), and also anchor pili in the cell wall, as demonstrated for several bacterial genera (Swaminathan et al, 2007; Budzik et al, 2008; Mandlik et al, 2008a; Nobbs et al, 2008; Neiers et al, 2009). The Gram-positive bacterium Streptococcus pneumoniae, also known as pneumococcus, is a major human pathogen (Lode, 2009). The clinical serotype-4 strain S. pneumoniae TIGR4 (TIGR4) forms long pili (Barocchi et al, 2006; Hilleringmann et al, 2008) and its virulence depends on them (Barocchi et al, 2006; Rosch et al, 2008). However, the second pneumococcal pilus found recently (Bagnoli et al, 2008) has not been detected. The long S. pneumoniae TIGR4 pili are encoded by the rlrA pathogenicity islet that includes a Rof-A-like transcriptional regulator (RlrA), three sortases (SrtC-1, SrtC-2 and SrtC-3) and three structural proteins RrgA (Swiss-Prot Q97SC3), RrgB (Swiss-Prot Q97SC2) and RrgC (Swiss-Prot Q97SC1), all of which contain an LPXTG motif (or variants thereof). As pneumococcal RrgB possesses the conserved motifs necessary for pilus formation (Ton-That and Schneewind, 2003, 2004; Ton-That et al, 2004), it has been suspected to form the backbone of the pneumococcal TIGR4 pilus, as indeed implied by immunoelectron microscopy (immuno-EM; Barocchi et al, 2006; LeMieux et al, 2006; Hilleringmann et al, 2008). This supposition was recently proved correct by the work of Fälker et al (2008), which revealed that only RrgB is required for pilus formation. Roles as ancillary proteins have been suggested for the two other components, RrgA and RrgC (Barocchi et al, 2006; LeMieux et al, 2006, 2008; Nelson et al, 2007; Hilleringmann et al, 2008). Immuno-EM has been used to unravel the location of the various pilus proteins since the first molecular description of Gram-positive pili (Ton-That and Schneewind, 2003). Antibodies targeted to RrgB, the major pilus building block of pneumococcal TIGR4 pili, were consistently found to bind along the filaments (Barocchi et al, 2006; LeMieux et al, 2006; Hilleringmann et al, 2008). Results concerning the location of the ancillary proteins RrgA and RrgC are less consistent: anti-RrgA is reported to bind in regularly spaced clusters along the pili (LeMieux et al, 2006) and close to the cell surface in the absence of both RrgB and RrgC (LeMieux et al, 2006; Nelson et al, 2007). Triple labelling has shown RrgA clusters to decorate pilus assemblies randomly and independent of RrgC distribution (Nelson et al, 2007; Hilleringmann et al, 2008), whereas a colocalisation of RrgA and RrgC in clusters has been observed by double-labelling experiments (Fälker et al, 2008). In addition, surface topographs acquired by atomic-force microscopy allowed identification of RrgC at the ends of detached pili (Fälker et al, 2008). Given these, in part contradictory, data the localisation of the three proteins in the pilus is still uncertain. In fact, both local association of RrgA and RrgC to the pilus shaft and their incorporation within it have been proposed (LeMieux et al, 2006; Nelson et al, 2007; Fälker et al, 2008; Hilleringmann et al, 2008). The pilus-specific TIGR4 sortases, SrtC-1, SrtC-2 and SrtC-3 (formerly SrtB, SrtC and SrtD) diverge in sequence from the housekeeping sortase, SrtA, the latter being dispensable for pilus assembly and localisation to the cell wall (LeMieux et al, 2008). Class-C sortases exhibit functional redundancy concerning pilus assembly and cell wall localisation (LeMieux et al, 2008; Manzano et al, 2008; Neiers et al, 2009). One study showed SrtC-1 and SrtC-3 to be required for incorporation of the ancillary subunits (LeMieux et al, 2008), while another report suggested SrtC-1 and SrtC-2 to differ only in their ability to incorporate RrgC (Neiers et al, 2009). Manzano et al (2008) demonstrated that SrtC-1 assembles RrgB fibres with high efficiency in vitro, whereas SrtC-3 has a much smaller fibre-assembling capacity and SrtC-2 has none at all. Negative-stain EM showed that these in vitro fibres mimic the pneumococcal pilus backbone with its beaded appearance. The structures of SrtC-1 and SrtC-3 both exhibit a flexible lid that shields the active site (Manzano et al, 2008). Based on this information a universal mechanism for pilus biogenesis was proposed, where class-C sortases with encapsulated active sites require activation by their specific substrate for pilus assembly (Manzano et al, 2008). Most recently, the structure of SrtC-2, showing the lid for this sortase as well, corroborated this hypothesis (Neiers et al, 2009). Although genetically based functional and epidemiological studies have substantially increased our understanding of Gram-positive pili (Telford et al, 2006), information on their native structure is lacking. Crystal structures of single pilus subunits of Streptococcus agalactiae (GBS) and S. pyogenes (GAS) have given novel insights into Gram-positive pilus structure: Krishnan et al (2007) describe the ancillary protein GBS52 as having a typical adhesin fold with two immunoglobulin-like domains. The first crystal structure of a Gram-positive pilus backbone protein showed the GAS shaft subunit, Spy0128, as an extended protein comprising two Ig-like domains and two intramolecular isopeptide bonds (Kang et al, 2007). These intramolecular bonds are likely to dictate pilus integrity, and a model based on Spy0128 describes the pilus fibre as a chain of individual subunits covalently linked head-to-tail by intermolecular peptide bonds (Kang et al, 2007; Kang and Baker, 2009). Given the fundamental structural difference between Gram-positive pili and their Gram-negative counterparts, and the significance these pili have during the bacterial life cycle, the elucidation of their native structure is of importance, not only to increase our understanding of the biology of Gram-positive bacteria, but also of related human disease. Here we study the molecular architecture of a Gram-positive pilus. We visualise native pneumococcal TIGR4 pili using a combination of electron microscopic techniques and show the pilus as a chain of RrgB proteins covalently linked head-to-tail, with the ancillary proteins RrgA and RrgC at its distal and proximal ends, respectively. Our results provide the first direct electron microscopic evidence for a simple Gram-positive pilus architecture, and resolve some of the open questions concerning the location and function of RrgA and RrgC. As documented earlier, the surface of the S. pneumoniae TIGR4 bacteria examined was covered by a non-homogenous distribution of pili (Barocchi et al, 2006). Different phenotypes of varying complexity have been described for these long structures (Barocchi et al, 2006; LeMieux et al, 2006; Hilleringmann et al, 2008). When TIGR4 bacteria were imaged by negative-stain EM after minimum perturbation (see section Materials and methods), the pili were seen as fine flexible filaments ∼6 nm in diameter (Figure 1A–C, arrowheads) that could be at least 1.5 μm long. In some cases changes in the degree of negative staining caused them to appear wider close to the bacterial surface (Figure 1B, arrow). A distinct tendency of the pili to associate into bundles of various diameters (Figure 1C, #) or, at larger distances from the bacterium, to intertwine to form tangles was also detected (Figure 1C, *). Both types of association were random and probably dependent on the bacterium, the negative-staining agent and grid handling. The resulting aggregates correlate well with the phenotypes of varying complexity detected previously primarily by immuno-EM (Barocchi et al, 2006; Hilleringmann et al, 2008), a technique that cannot match the structural resolution obtained by negative-stain transmission EM (TEM). Our results suggest that pneumococcal TIGR4 pili are single filaments and that the aggregates and super-helical assemblies previously observed are sample preparation artefacts. The degree to which the bacterial capsule and attached filaments could be visualised in the present experiments depended on the negative stain used. Under optimum staining conditions the fine ∼6-nm-wide pili could be followed all the way to the cell boundary (Figure 1C). Figure 1. TIGR4 bacteria were treated with the murein-hydrolysing enzyme mutanolysin to release their peptidoglycan-anchored pili into the supernatant. Using a modified procedure of Hilleringmann et al (2008), the liberated pili were then isolated in 10 mM Tris–HCl (pH 8), 1 mM EDTA and 1 mM DTT (see section Materials and methods). Under the transmission electron microscope TIGR4 pili appeared as long, flexible, ∼6-nm-wide filaments (Figure 2A), with the general morphology observed in situ (Figure 1). The structure revealed in enlarged views (Figure 2A, inset) bears some similarity to that of RrgB filaments assembled in vitro (Manzano et al, 2008). Analysis of the same high-molecular-weight (HMW) fractions by SDS–PAGE and Western blotting documented the presence of all three pilus proteins, RrgA, RrgB and RrgC (Figure 2B). In agreement with previous reports (LeMieux et al, 2006, 2008; Fälker et al, 2008; Hilleringmann et al, 2008) RrgB was found to be the major constituent and RrgA and RrgC only accounted for a minor fraction of the total protein present. The HMW fractions observed for RrgB confirm the covalent association of Gram-positive pilus subunits previously reported (Telford et al, 2006). Figure 2. Immunolabelling was used to localise all three proteins within the pilus structure. In contrast to previous reports, we directly visualised primary antibodies to overcome the resolution limit imposed by the size of the primary and secondary gold-bearing antibody complex (Barocchi et al, 2006; LeMieux et al, 2006; Fälker et al, 2008; Hilleringmann et al, 2008). The polyclonal antibodies generally used for Western blots were purified further according to the protocol of Mueller et al (2005), were highly specific and did not show cross-reactivity (Supplementary Figure S1). Their concentrations were adjusted to show binding but minimise the number of free antibodies in the solutions. Anti-RrgB–His antibodies decorated the pilus shaft at irregular intervals and, having two binding sites, often linked pili together forming ladder- and net-like assemblies depending on the degree of lateral cross-linking (Figure 2C and Supplementary Figure S2). Views of single antibodies are shown in the inset of Figure 2D for comparison. In contrast to anti-RrgB–His antibodies, antibodies against RrgA–His only bound at the end of pili, generally clustering and, as they are divalent, frequently linking two pili together in typical v-shaped assemblies not otherwise observed (Figure 2D and Supplementary Figure S3). Anti-RrgC–His also bound at the end of the pilus shaft and sometimes clustered there (Figure 2E and Supplementary Figure S4). As RrgC is two times smaller than RrgA (see below and Figure 3), the capacity of anti-RrgC–His to link two fibres appears to be much diminished. As polyclonal antibodies were used, the formation of antibody clusters does not necessarily mean that there is more than one copy of the labelled protein present. Neither anti-RrgA–His nor anti-RrgC–His decorated the pilus shaft, and the ladder-like assemblies typical of anti-RrgB–His were not formed. Occasionally an antibody was extremely close to the side of a pilus, but this was a chance occurrence rather than specific binding as it was rare and reflected the free antibody concentration and distribution on the grid. Accordingly, the results show that RrgA and RrgC are present at the ends of the pilus shaft formed by RrgB, and strongly imply that these two ancillary proteins are neither incorporated in nor associated with it. In confirmation, mutant bacteria lacking either RrgA or, RrgC or both (Supplementary Figure S5) still form long pili with the shaft morphology of the wild type (see below; Supplementary Figure S6 and reference Fälker et al, 2008). Figure 3. To acquire further structural information, the three pilus components were expressed in Escherichia coli with an engineered C-terminal His6 tag but otherwise mimicking the predicted processed forms, which are lacking the N-terminal signal sequence and the C-terminal region starting from the respective LPXTG sorting motif (Supplementary Figure S7). The affinity purified RrgA–His (93.37 kDa), RrgB–His (66.29 kDa) and RrgC–His (40.26 kDa) proteins were examined by EM. First, the oligomeric state of RrgA–His was defined by scanning TEM (STEM) mass measurements. The measured mass of 108 (±42)kDa (n=319) clearly showed the large majority of the protein to be monomeric (Supplementary Figure S8A). Also, all three proteins were imaged by negative-stain TEM. The average projections calculated by single-particle analysis of TEM electron micrographs and STEM single-shot images are shown in Figure 3. RrgA–His is a flexible, ∼18-nm-long, elongated macromolecule with four domains of unequal size. These domains give the structure a distinct taper, one end being ∼5 nm wide and the other ∼3 nm wide (Figure 3A). RrgB–His is an ∼12-nm-long, elongated particle (Figure 3B). Up to three domains can be detected and, depending on the orientation, a lateral protrusion is sometimes discernible. The domains are almost 5 nm wide without and ∼6.5 nm wide with the protrusion. Being of about 2/3 the length of RrgA–His and roughly the same width, the imaged RrgB–His was monomeric. The overall shape of the RrgB–His particles is compatible with the partial 2.2-Å X-ray structure of Spy0128 (aa 18–308), the major pilin subunit of the Gram-positive human pathogen S. pyogenes (Kang et al, 2007). With 291 residues this construct is much smaller than RrgB, which comprises 608 residues in its predicted processed form. This explains the different dimensions of the two structures, the Spy0128 construct being 2–3 nm wide and 9.8 nm long, that is, half as wide and 20% shorter than the RrgB–His class averages. Images of RrgC–His, the smallest of the three pilus proteins, revealed up to 4-nm-wide and 10-nm-long elongated particles with 2–3 domains (Figure 3C). From these dimensions the protein was also monomeric, as confirmed by size-exclusion chromatography (data not shown). The mass-per-length of freeze-dried, unstained, isolated TIGR4 wt pili (Figure 4A) was measured by STEM to determine the stoichiometry of their RrgB subunits. The 395 segments evaluated gave a histogram with a single peak at 6.4 (±1.4)kDa/nm (standard error, 0.07 kDa/nm; Figure 4B). Given the ±5% overall precision of the STEM measurement (Müller and Engel, 2006), this indicates the presence of one 65.44-kDa RrgB monomer every 10.2 (±0.5)nm on average, which is slightly shorter than the length determined for a recombinant RrgB-His monomer, ∼12 nm. As expected the mass-per-length of TIGR4 ΔrrgA pili was comparable (Supplementary Figure S8B). Figure 4. The high signal-to-noise ratio of the STEM was also exploited to examine negatively stained TIGR4 pili. These images confirmed that the pilus is formed by a single string of subunits and revealed a well-defined protrusion extending at irregular intervals from the filament shaft like a 'nose' (Figure 4C and D). Most importantly, they showed that this 'nose' points in a defined direction giving the pilus a distinct polarity. Where the boundaries of single subunits could be discerned (Figure 4C and Supplementary Figure S9, lines), their shapes correlated well to those of the RrgB–His class averages and single projections implying that the pilus 'nose' is the protrusion observed in some orientations of the macromolecule. In agreement with the measured mass-per-length, the subunit spacing (10.2 (±0.6)nm; marked by lines in Figure 4C and Supplementary Figure S9) was somewhat less than the length of an RrgB–His monomer, implying that adjacent RrgB monomers may overlap slightly, as detailed in Figure 4E and F. Close examination of the pili formed by TIGR4 ΔrrgC, TIGR4 ΔrrgAC bacteria revealed the same sub-structure, that is, a single string of monomeric RrgB subunits, with 'noses' protruding at irregular intervals (Supplementary Figure S6, insets). As the 'nose' was not always visible at regular intervals along the pilus shaft under the preparation conditions used, the RrgB monomers do not appear to assemble according to a defined helical rule. In any case, the covalent bonds formed between the monomeric RrgB subunits and further possible molecular interactions not only allow the filaments to bend freely, but also permit a degree of rotation around the long filament axis under the forces encountered on adsorption to the EM grid. Knowledge of its existence allowed the fine 'nose' feature to be detected on TEM images and showed it to consistently point in one direction for longer pilus stretches, indicating that the individual RrgB pilus subunits are linked head-to-tail. Examination of the immunolabelling images with this information also clearly showed that anti-RrgA–His and anti-RrgC–His labelled opposite ends of the pilus shaft (Figure 5 and Supplementary Figure S10). The role of RrgA as a pilus-associated adhesin that is expected to locate at the distal end (Nelson et al, 2007; Hilleringmann et al, 2008), would suggest that RrgC is located at the proximal end of the pilus shaft. In accordance, the pili formed by a TIGR4 ΔrrgC genetic background (TIGR4 ΔrrgC and TIGR4 ΔrrgAC) detached more easily from the bacteria, resulting in the appearance of more HMW pili material in the culture supernatant (Supplementary Figure S11). Figure 5. Since Ton-That et al's description of the pili of C. diphtheriae (Ton-That and Schneewind, 2003), many Gram-positive bacteria have been shown to possess such filamentous appendages, including group-A Streptococci (Mora et al, 2005), group-B Streptococci (Lauer et al, 2005), S. pneumoniae (Barocchi et al, 2006), Enterococcus faecalis (Nallapareddy et al, 2006), Bacillus cereus (Budzik et al, 2007) and actinomyces (Wu and Fives-Taylor, 2001; Ton-That et al, 2004). In addition, probable pilus loci have been identified by genome sequencing of Streptococcus spp. (Osaki et al, 2002; Xu et al, 2007). The adhesive function of pili (Barocchi et al, 2006; Dramsi et al, 2006) is critical for the attachment of pathogens to specific host cells during colonisation, and explains why pilus expression increases the pathogenicity of various Gram-positive bacteria in animal models (Hava and Camilli, 2002; Abbot et al, 2007; Maisey et al, 2007; Rosch et al, 2008). This central function has promoted several EM analyses of native pili attached to bacterial cells and after purification (LeMieux et al, 2006; Nelson et al, 2007; Fälker et al, 2008; Hilleringmann et al, 2008), as well as X-ray studies of Gram-positive pilus components (Kang et al, 2007; Krishnan et al, 2007) and of the pilus assembly machinery (Manzano et al, 2008; Neiers et al, 2009). We have used TEM, antibody labelling and STEM to elucidate the structure of the native TIGR4 pilus. The shaft is an ∼6-nm-wide, single chain of slightly overlapping, head-to-tail covalently linked, monomeric RrgB subunits and can be at least 1.5 μm long. According to the mass-per-length measurements by STEM, a 1.5-μm-long pilus comprises approximately 150 RrgB monomers. TIGR4 pili assembled in a ΔrrgA, ΔrrgC or ΔrrgAC genetic background were also examined after purification and exhibit the same length and morphology as the wt pilus. The two ancillary proteins RrgA and RrgC are found at opposite ends of the shaft. RrgA is distal and consequently, RrgC proximal to the bacterium, as illustrated in Figure 6. Figure 6. The proposed model agrees with functional implications derived from the structures of the S. pneumoniae sortases SrtC-1 and SrtC-3 (Manzano et al, 2008), and SrtC-2 (Neiers et al, 2009). All of them possess an encapsulated active site that is postulated to be activated by the specific LPXTG-like motifs found in RrgA, RrgB and RrgC. These motifs are known to have profound consequences for catalysis (LeMieux et al, 2008), and could be indicative of a controlled sequential pilus assembly process. Although the sequence of this process remains to be unveiled, it would ensure RrgA to be at the distal tip of the pilus shaft assembled from RrgB, which is the only one of the three pilins to have both the pilin and the LPXTG domain (Supplementary Figure S7B). Incorporation of the RrgC protein at the proximal end would then terminate the pilus assembly process and induce cell wall linkage similar to that observed for SpaB in C. diphtheriae (Mandlik et al, 2008a) or GBS150 in S. agalactiae (Nobbs et al, 2008). In spite of this agreement our model is in contrast to other reports, which state that ancillary proteins RrgA and RrgC are either incorporated in or associated with the pilus shaft (LeMieux et al, 2006; Nelson et al, 2007; Fälker et al, 2008; Hilleringmann et al, 2008). Immuno-EM showed RrgC and RrgA to be in clusters along the length of the pilus shaft (LeMieux et al, 2006; Hilleringmann et al, 2008), sometimes together as indicated by double-labelling studies (Fälker et al, 2008). The discrepancy to the observations presented here is explained by the lower resolution of the previous immuno-EM studies, which could not resolve single pili with certainty, but rather visualised immunogold-labelled pilus bundles. LeMieux et al (2008) came to a similar conclusion speculating that the observed clustering of RrgA simply manifests the bundling of different-length pili with RrgA at their tips. Immunogold labelling cannot prove this hypothesis because single pili within a bundle cannot be resolved. Negative-stain EM can achieve higher resolution. Using this technique we first demonstrate that pili emerge from the cell surface as single, ∼6-nm-wide filaments, which then form bundles or tangle at random (Figure 1). Second, by isolation of native pili using a modified protocol we obtain a higher-resolution definition of the pilus shaft by STEM dark-field imaging of single pili, and show their previously observed beaded structure (Manzano et al, 2008) to arise from single RrgB monomers linked head-to-tail with a periodicity of about 10 nm (Figure 4C–F and Supplementary Figure S9). This periodicity is compatible with STEM mass-per-length measurements; 6.4±1.4 kDa/nm (Figure 4B) also translates to about 1 RrgB monomer/10 nm. Third, the well-defined protrusions on the filaments, the 'noses', result in clear polarity and indicate a head-to-tail subunit assembly (Figure 4C–F). These 'noses' relate to lateral protrusions on images of negatively stained, recombinantly expressed RrgB–His monomers (Figure 3B). In agreement, antibody labelling demonstrated the pilus shaft to be exclusively comprised of RrgB proteins (Figure 2C–E and Supplementary Figure S2–4). Fourth, once these basic simple features of the TIGR4 pilus had been defined by high-resolution STEM, we could also recognise them on close inspection of TEM images recorded at lower magnification. Based on the polarity imposed by the 'noses' and primary antibody labelling it can be said that RrgA and RrgC are located at opposite ends of the pilus shaft (Figure 5 and Supplementary Figure S10). In contrast to immuno-EM, where a secondary, gold-labelled antibody is necessary, direct visualisation of the primary antibody provides the resolution required to locate a specific protein associated with a single pilus shaft. With the new higher resolution data obtained, the previously reported clusters of surface-located RrgA can be interpreted as the distal ends of several interacting individual pili that may emerge from different sites of the cell wall or have different length. The location of RrgA is assigned as distal as this protein has recently been described as an adhesin (Nelson et al, 2007; Hilleringmann et al, 2008). This implies that RrgC is at the proximal end of the pilus, which must be anchored in the bacterial cell wall. In C. diphtheriae the protein SpaB is proposed to act as the terminal subunit and cell wall anchor in pilus assembly, and in its absence the pili formed are largely found in the medium (Mandlik et al, 2008a). In S. agalactiae the pilus is covalently linked to the cell wall via the ancillary pilus subunit GBS150, and its absence provokes the release of pili into the culture supernatant (Nobbs et al, 2008). Similarly, less HMW pneumococcal TIGR4 pili are found in the supernatant of wt S. pneumoniae liquid cultures than in the supernatants of ΔrrgC and ΔrrgAC mutants, the apparent enhanced pilus loss of the latter over the ΔrrgC mutant probably resulting from the missing adhesive function of RrgA (Supplementary Figure S11). Accordingly, RrgC is likely to be the terminal pilus subunit and to warrant cell wall anchoring. The proposed pilus model (Figure 6) agrees with the fact that RrgB assembles in vitro in the presence of the pilus-polymerising transpeptidase SrtC-1 (Manzano et al, 2008) and in vivo in the absence of both RrgA and RrgC (Supplementary Figure S6C and Fälker et al, 2008; LeMieux et al, 2008). The observation that the accessory subunits RrgA and RrgC are found in similar quantities within each 'rung' of the ladder of bands arising from pili of different lengths on Western blots corroborates this model as well (LeMieux et al, 2008). Our results confirm the observation that RrgA and RrgC can form covalent heterodimers, but not higher order polymers in the ΔrrgB background. As indicated in our model, this implies that the IPQTG motif in the RrgB protein can be covalently linked to a critical lysine in RrgC, and that in the absence of RrgB the same site in RrgC is linked to the YPRTG motif of RrgA. Whereas RrgC must have another site to be anchored to the cell wall, RrgA appears to expose a single motif for sortase action (Figure 6 and Supplementary Figure S7). Our model of the pilus shaft (Figure 6) extends the previous model of the S. pyogenes pilus created on the basis of the 2.2-Å Spy0128 structure and mass spectrometry (Kang et al, 2007). No overlap of subunits is proposed in the Spy0128 filament model. In contrast, we predict an overlap of about 1 nm for RrgB, in accordance with the sequence-based pneumococcal pilus assembly model (Telford et al, 2006). The overlap is based on pilus images recorded at × 106 magnification, the mass-per-length values provided by STEM and the dimensions of RrgB–His monomers. A major difference in the mass of Spy0128 (aa 18–308; Kang et al, 2007) and the integrated RrgB monomer (608 aa) may explain this overlap, but a high-resolution structure of full-length RrgB is required to prove our hypothesis. Using quantitative EM techniques we have visualised the molecular details of a Gram-positive pilus for the first time. Together with the sequence analysis presented by Manzano et al (2008), our observations suggest a pilus architecture that is likely to be valid for other Gram-positive pili. It implies a simple pilus assembly mechanism, and indicates novel sites for therapeutic intervention. S. pneumoniae type-4 strain TIGR4 was used (Tettelin et al, 2001). The TIGR4 ΔrrgA mutant used initially was kindly donated by B Henriques-Normark (Karolinska Institutet, Stockholm). Later TIGR4 ΔrrgA, TIGR4 ΔrrgB, TIGR4 ΔrrgC and TIGR4 ΔrrgAC mutants were created by PCR-based overlap extension (Supplementary data and Supplementary Tables S1–S2). The pneumococcal strains were stored at −80°C in 12% glycerol and routinely grown at 37°C under 5% CO2 on Tryptic Soy Agar (Becton Dickinson) supplemented with 5% defibrinated sheep blood or in Todd–Hewitt Yeast Extract (THYE) broth. When appropriate, erythromycin and kanamycin (Sigma-Aldrich) were used as selection markers. Recombinant expression and purification of His6-tagged pilus proteins was performed as described previously (Hilleringmann et al, 2008). When necessary a size-exclusion chromatography step was performed after affinity purification. Purified proteins were finally dialysed against 10 mM Tris–HCl (pH 8), 1 mM EDTA and 0.5 mM DTT. The native pili of TIGR4 wt and TIGR4 ΔrrgA, TIGR4 ΔrrgC and TIGR4 ΔrrgAC were purified essentially according a protocol described by Hilleringmann et al (2008) treating bacteria with mutanolysin, a murein-hydrolysing enzyme (Sigma M9901), to liberate covalently peptidoglycan-anchored pili into the supernatant, but using a Tris buffer (10 mM Tris–HCl (pH 8), 1 mM EDTA, 1 mM DTT)-based procedure. In addition the following modifications were applied: harvested bacteria were washed in Tris buffer and resuspended in Tris containing protoplast buffer (Tris buffer, 20% sucrose). Final sample dialysis against Tris buffer was performed using a molecular weight cut-off of 300 kDa (Spectra/Por Biotech cellulose ester). SDS–PAGE analysis was performed using NuPAGE 3–8% Tris Acetate Gels (Invitrogen) according to the instructions of the manufacturer. HiMark pre-stained, HMW protein standard (Invitrogen) served as the protein standard. Western blot analysis was performed using standard protocols. Unless otherwise stated, antibodies against recombinant RrgA–His and RrgB–His were used at 1:10 000 dilution, and against RrgC–His at 1:2000 dilution. Secondary goat anti-mouse HRP antibodies were diluted 30 000 ×. Polyclonal mouse antibodies against recombinant RrgA–His, RrgB–His and RrgC–His were produced in our laboratory. For immunolabelling of isolated native pili, antibodies were purified to 100% specificity against their respective proteins using the protocol described by Mueller et al (2005). The respective affinity-purified antibodies were individually incubated with wt or ΔrrgA TIGR4 pili (control experiments) overnight or during 50 min at 4°C in a series of runs covering a range of concentrations. Samples were inspected by negative-stain TEM. The conditions were optimised to yield good labelling and have the minimum number of free antibodies on the EM grids. For TEM of whole bacteria, 100–200 μl of PBS was added to the blood agar growth plate and agitated gently to delicately remove bacteria from the agar. The plate was tilted and an aliquot of the resulting bacterial suspension was removed from close to the liquid surface. Small aliquots of this stock suspension were then directly loaded onto carbon-coated Parlodion microscopy grids. The bacteria were allowed to settle (5 min) and then stabilised by addition of 2% paraformaldehyde (40 s). Grids were washed on droplets of water, negatively stained and examined. As dictated by grid quality, the stock was sometimes centrifuged gently for several minutes (3000 r.p.m. for 5–10 min), the pellet was then gently resuspended in PBS and grids were prepared; if necessary these steps were repeated. Samples of the isolated pili were diluted in buffer as required and adsorbed for 1 min to glow discharged 400 mesh carbon-coated Parlodion or STEM grids (see below). These were washed and negatively stained with 2% (w/v) uranyl acetate (UAc) or 2% (w/v) phosphotungstic acid (PTA) and imaged with a CM 100 transmission electron microscope (Philips, Eindhoven, the Netherlands) operating at 80 kV. Electron micrographs where recorded with a 2000 by 2000 pixel, charge-coupled device camera (Veleta; Olympus soft imaging solutions GmbH, Münster, Germany) at a nominal magnification of × 130 000, yielding a final pixel size corresponding to 0.36 nm on the specimen scale. Particles were manually selected for single-particle analysis and averaged using the EMAN software (Ludtke et al, 1999). Samples were prepared on glow-discharged, thin carbon films coating a perforated carbon layer on gold-coated copper grids, washed and either freeze-dried for mass measurement or negatively stained as above. Mass measurements were performed on pilus and recombinant RrgA–His samples as described (Broz et al, 2007), except that the grids were washed on eight drops of quartz double-distilled water. Images were recorded from the TIGR4 pili at doses ranging from 400 to 950 electrons/nm2 and from pili of the TIGR4 ΔrrgA mutant at a dose of 700±56 electrons/nm2, and evaluated using the MASDET program package (Krzyžáneka et al, 2009). A linear regression describing the dose dependence of the mass-per-length values determined for the former sample defined beam-induced mass-loss. Both the TIGR4 data set and the ΔrrgA pilus data were corrected accordingly and scaled to the mass-per-length determined in the same run for tobacco mosaic virus (TMV). Images were recorded from the RrgA–His sample at a dose of 1170±100 electrons/nm2 and evaluated using MASDET (Krzyžáneka et al, 2009). Beam-induced mass-loss was corrected according to the behaviour of proteins in the mass range 120–190 kDa (Müller and Engel, 2001 and unpublished results) and the data were scaled according to the mass-per-length measured for TMV. The corrected data sets were binned into histograms and described by Gauss curves. Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org). We thank Vladislav Krzyžáneka and Rudolf Reichelt, University of Münster, for continuous help with improving the mass determination software used in this project; Françoise Erne-Brand for STEM mass measurements; Ruben Diaz-Avalos, Florida State University, for providing Tobacco Mosaic Virus samples; and Monica Moschioni and Angela H Nobbs for valuable discussions and for supplying material. Further we thank Silvana Savino and Vega Masignani for promoting the project. This work was supported by the Swiss National Science Foundation (grant 3100A0-108299 to AE), the Maurice E Müller Foundation of Switzerland and Novartis Vaccines and Diagnostics s.r.l. MH, GD, RR and IF are employees of Novartis Vaccines and Diagnostics s.r.l. Supplementary Information Review Process File
Referência(s)