Analysis of the cell surface layer ultrastructure of the oral pathogen Tannerella forsythia

The Gram-negative oral pathogen Tannerella forsythia is decorated with a 2D crystalline surface (S-) layer, with two different S-layer glycoprotein species being present. Prompted by the predicted virulence potential of the S-layer, this study focused on the analysis of the arrangement of the individual S-layer glycoproteins by a combination of microscopic, genetic, and biochemical analyses. The two S-layer genes are transcribed into mRNA and expressed into protein in equal amounts. The S-layer was investigated on intact bacterial cells by transmission electron microscopy, by immune fluorescence microscopy, and by atomic force microscopy. The analyses of wild-type cells revealed a distinct square S-layer lattice with an overall lattice constant of 10.1 ± 0.7 nm. In contrast, a blurred lattice with a lattice constant of 9.0 nm was found on S-layer single-mutant cells. This together with in vitro self-assembly studies using purified (glyco)protein species indicated their increased structural flexibility after self-assembly and/or impaired self-assembly capability. In conjunction with TEM analyses of thin-sectioned cells, this study demonstrates the unusual case that two S-layer glycoproteins are co-assembled into a single S-layer. Additionally, flagella and pilus-like structures were observed on T. forsythia cells, which might impact the pathogenicity of this bacterium.


Introduction
The periodontal pockets of humans harbor more than 500 bacterial species. Among them is a group of bacteria that constitute the "red complex", comprising Tannerella forsythia, Porphyromonas gingivalis, and Treponema denticola, with the latter being strongly implicated in the onset of periodontitis (Socransky et al. 1998). Periodontitis is a chronic inXammation of the periodontium with multifactorial etiology. T. forsythia meets the criteria for being considered a periodontal pathogen (Socransky 1979) because of (1) its association with and increased levels in periodontitis (Socransky et al. 1998), (2) the evidence of host responses to its antigens (Bird et al. 2001;Yoo et al. 2007), (3) its ability to cause disease in animal models (Sharma et al. 2005;Kesavalu et al. 2007), and (4) the existence of distinct virulence factors that can contribute to the disease process (Sharma 2010).
Tannerella forsythia is an anaerobic Gram-negative bacterium belonging to the Cytophaga-Bacteroidetes-Firmicutes cluster of bacteria. It was initially named Bacteroides forsythus (Tanner et al. 1986) and later reclassiWed as T. forsythia (Sakamoto et al. 2002). Its cell surface is covered with a regularly arrayed surface (S-) layer (for review see Messner et al. 2010), and early electron microscopic investigations have shown the presence of an orthogonal Communicated by Harald Huber.
S-layer lattice (Kerosuo 1988). SDS-PAGE of intact cells revealed that two high-molecular-mass glycoproteins of 230 and 270 kDa are present in T. forsythia, which are encoded by the tfsA and tfsB gene, respectively (Higuchi et al. 2000;Lee et al. 2006). The 1,179-amino acid TfsA and the 1,364-amino acid TfsB proteins have a calculated molecular mass of 132 and 154 kDa, respectively, with pI values of 7.8 and 9.9, respectively. Comparison with database entries indicated that the S-layer proteins of T. forsythia apparently have unique sequences exhibiting no homology to other known S-layer proteins of prokaryotic organisms. Only recently, these S-layer proteins were shown to be covalently modiWed with identical O-glycosidically linked oligosaccharides .
The exact function of the T. forsythia S-layer is not yet known, but there are indications that it might be an important virulence factor (Sakakibara et al. 2007;Sharma 2010;Sekot et al. 2011). These include the demonstration that the S-layer mediates adhesion and/or invasion to human gingival epithelial cells (Sakakibara et al. 2007) as well as its potential to delay the recognition of T. forsythia by the innate immune system of the host (Sekot et al. 2011). This underlines the importance of the bacterial cell surface in conferring to pathogenicity. For analyzing the role of the T. forsythia S-layer proteins in adhesion to and invasion of human gingival epithelial cells, deWned insertional inactivation mutants of either of the S-layer genes (named T. forsythia tfsA and T. forsythia tfsB) as well as an S-layer double mutant (T. forsythia tfsAB) have been constructed (Sakakibara et al. 2007). Results from this study and a more recent study conducted in our laboratory suggest that the S-layer of T. forsythia plays an important role in the initiation stage of oral infection, including periodontal disease (Sakakibara et al. 2007;Sekot et al. 2011).
While most S-layer lattices are composed of a single (glyco)protein species (Messner et al. 2010), data from literature indicate that in T. forsythia, two glycosylated S-layer proteins are present (Higuchi et al. 2000;Lee et al. 2006). In the rare cases of bacterial species which possess more than one S-layer protein, expression of these S-layer proteins is strongly impacted by the growth stage of the bacterium as well as by the physiological and/or environmental conditions. One of the best investigated among those organisms is the Gram-positive bacterium Bacillus anthracis, for which regulatory studies have shown that the abundant S-layer proteins EA1 and Sap appear sequentially at the bacterial surface (Couture-Tosi et al. 2002). Within Gram-negative bacteria of the genus Aquaspirillum, several species are known to possess two superimposed S-layer proteins. The S-layer ultrastructure is best investigated for Aquaspirillum serpens strain MW5 where two, presumably identical, hexagonal S-layer lattices are superimposed (Stewart and Murray 1982). While the underlying layer is attached to the lipopolysaccharide of the outer membrane, the second layer appears to be attached directly to the Wrst layer. Whereas in A. serpens, both superimposed S-layers can be clearly distinguished in thin-sectioned bacterial cells, the situation has been less clear for T. forsythia according to the electron microscopy data published by Sakakibara et al. (2007). This poses the interesting questions of how the two diVerent S-layer glycoprotein species of T. forsythia are arranged, with the principal options for (1) superimposition of two individually assembled S-layers or (2) co-assembly of the two S-layer glycoproteins into a single S-layer. For building up a deWned S-layer ultrastructure, it has to be taken into account that the T. forsythia S-layer proteins are glycosylated , with glycans naturally occurring in an outward orientation, which would allow them to carry out yet to be identiWed biological roles.
To assess the potential of the T. forsythia cell surface in the bacterium-host cross talk the understanding of the S-layer ultrastructure as outermost cell envelope component is essential. Therefore, in this study, diVerent microscopic approaches were applied to characterize the native cell surface of T. forsythia wild-type and S-layer mutant cells as well as the self-assembly capability of isolated native S-layer glycoproteins (TfsA-GP, TfsB-GP) and their recombinant, glycan-free, counterparts (rTfsA and rTfsB). This included (1) electron microscopic investigations (TEM) of T. forsythia wild-type and S-layer mutant cells as well as of isolated, puriWed (glyco)proteins, using either negatively stained, freeze-fractured, freeze-dried, or ultra-thin sectioned samples; (2) immune-Xuorescence microscopy of intact bacterial cells using polyclonal antibodies raised against the individual recombinant T. forsythia S-layer proteins; (3) AFM studies of the S-layer under nondestructive conditions as present on intact bacterial cells as well as visible after recrystallization of recombinant TfsA and TfsB proteins on silica substrates. These microscopic data are interpreted in conjunction with a transcription and translation analysis of the S-layer genes. This study provides the Wrst comprehensive characterization of the cell surface ultrastructure of the oral pathogen T. forsythia.

Materials and methods
Bacterial strains, cultivation, and isolation of native TfsB-glycoprotein Tannerella forsythia ATCC 43037 was purchased from the American Type Culture Collection (ATCC, Manassas, VA, USA). Cultivation of the strains has been described previously (Sekot et al. 2011). Upon insertional inactivation of one or both S-layer genes, the mutant strains tfsA, tfsB, and tfsAB were obtained (Sakakibara et al. 2007). S-layer (glyco)proteins were detected after separation of samples by SDS-PAGE on 8% polyacrylamide (PA) gels upon staining with Coomassie Brilliant Blue R (CBB; GERBU, Wieblingen, Germany) and periodic acid-SchiV (PAS) for protein and carbohydrates, respectively . A relative quantitation of S-layer proteins was done directly from a CBB-stained SDS-PA gel using an Odyssey scanner (LI-COR, Bad Homburg, Germany) and integration of the intensity signal obtained from the S-layer bands at 700 nm using application software 3.0.
Cloning and expression of TfsA and TfsB, and puriWcation of rTfsA and rTfsB The tfsA and tfsB genes were ampliWed by PCR without their predicted signal sequences using chromosomal T. forsythia DNA as template. For gene ampliWcation, the primer pairs tfsA 23-1179 for/tfsA 23-1179 rev and tfsB 28-1364 for/ tfsB 28-1364 rev, respectively, were used (Invitrogen, Lofer, Austria; Table 1). The restriction sites NcoI/NotI were introduced into the for/rev primer for the tfsA and the sites BsaI-NcoI/NotI into the for/rev primer for the tfsB gene to allow directional cloning of the PCR product into the multicloning site of the expression vector pET28a(+) (Novagen, Darmstadt, Germany) via NcoI/NotI. Since the sequence of the tfsB gene contains an NcoI restriction site, a forward primer was designed in such a way that a BsaI site preceded the NcoI site. Thus, upon digestion of the PCR product with BsaI, an NcoI compatible overhang was achieved (Novagen, pET System Manual 11ed; http://lifeserv. bgu.ac.il/wb/zarivach/media/protocols/Novagen%20pET% 20system%20manual.pdf). The resulting expression plasmids, pET28a(+)-rtfsA and pET28a(+)-rtfsB, were analyzed by restriction enzyme digestion and Wnally conWrmed by DNA sequencing (AGOWA, Berlin, Germany). The rTfsA and rTfsB proteins are encoded by the sequence from 67 to 3,537 bp (contig TF2661-2662, Oralgen database at: http://www.oralgen.lanl.gov/_index.html) and from 82 to 4,092 bp (contig TF2663), respectively.
E. coli BL21-(DE3) (Novagen) was used as host strain for the expression of the two recombinant S-layer proteins. E. coli BL21, containing either plasmid pET28a(+)-rtfsA or pET28a(+)-rtfsB, was cultured with continuous shaking at 200 rpm and 37°C for 12 h in Luria-Bertani (LB) medium (Invitrogen) containing 50 mg/L kanamycin. LB medium was inoculated with 1% (v/v) of overnight culture and grown until the optical density at 600 nm (OD 600 ) of the culture reached 0.6. Expression of the recombinant proteins was induced by addition of isopropyl--D-thiogalactopyranoside (IPTG; Fermentas, St. Leon-Rot, Germany) to a Wnal concentration of 1 mM. Bacteria were harvested 4 h past induction, and proteins were analyzed by SDS-PAGE on 8% PA-gels as described above.
Fluorescence was recorded at 470-nm extinction and 510nm emission. After ampliWcation, a melting curve analysis with a temperature gradient of 0.1°C/s from 65 to 99°C was performed to conWrm that only the speciWc products were ampliWed. As a standard for quantiWcation, the tfsAB operon (ampliWed from genomic DNA, »7.6 kbp, primers Af6, Br6; Table 1) was used. Primers applied in tfsA qPCR cycling were TfsA_RT2_for and TfsA_RT2_rev; in tfsB qPCR cycling, primers TfsB_RT1_for and TfsB_RT2_rev were used (Table 1). QuantiWcation of the transcription levels of tfsA and tfsB was done applying undiluted and 10-fold diluted cDNA samples. RNA (undiluted) was used to analyze for contaminating DNA after DNAseI treatment to rule out false-positive results.
Relative protein expression levels of TfsA and TfsB in T. forsythia wild-type cells and in S-layer gene single mutants (T. forsythia tfsA, tfsB) were analyzed after CBB staining as described above. Gels were loaded with standardized amounts of crude cell extracts prepared from the respective T. forsythia cells.
Raising, puriWcation, and labeling of polyclonal antibodies against T. forsythia S-layer proteins PuriWed rTfsA and rTfsB proteins, respectively, were dialyzed against 18 MOhm-water (MilliQ water; Millipore, Vienna, Austria), lyophilized and resuspended in PBS (phosphate-buVered saline) containing 0.1% SDS at a concentration of 0.85 mg/mL and used to raise polyclonal antibodies in rabbits. Antisera were tested by Western blot analysis for cross-reactivity prior to use. Pure antibody fractions were received upon HiTrap ™ Protein G 2 mL HP (GE Healthcare, Uppsala, Sweden) chromatography after washing with 20 mM sodium phosphate buVer (pH = 7.0). Elution was done with 0.1 M glycine solution (adjusted to pH = 2.7 with HCl), collecting 1-mL fractions into tubes containing 40 L of 1 M Tris solution. Antibodies were concentrated by centrifugation with Amicon ® Ultra 3 K centrifugal Wlters (Millipore) and stored in 20 mM sodium phosphate buVer (pH = 7.0). Fab fractions were produced by papain digestion using the Pierce ® Fab Micro Preparation Kit (Thermo ScientiWc, Waltham, MA, USA) and labeled with Pierce ® FITC (Xuorescein isothiocyanate) Antibody Labeling Kit (Thermo ScientiWc) according to the manufacturer's instructions.
Ultra-thin sectioning, negative-staining, freeze-etching, and freeze-drying experiments and transmission electron microscopy Ultrathin-sectioning of T. forsythia cells was carried out as described previously (Messner et al. 1984). BrieXy, the room-temperature processing included Wxation of samples with 2.5% (w/v) paraformaldehyde/2.5% (w/v) glutaraldehyde/0.5% (w/v) tannin in cacodylate buVer and osmium tetroxide Wxation without ruthenium red. Dehydration was performed in an increasing alcohol series before samples were embedded in Epon resin. Negative staining of crystallized S-layer proteins as well as of intact Tannerella cells was performed with 1% uranyl acetate solution for 30 s (Messner et al. 1986).
Freeze-etching of intact bacteria was carried out in diVerent freeze etching units, including a BAF 400-T unit (BAL-TEC, Balzers, Liechtenstein), a BAF 060 unit (Leica, Wetzlar, Germany), and a Cressington CFE-50 unit (Cressington, Watford, UK). Fracturing and etching of frozen samples was done at ¡95°C for 90 s prior to Pt/C shadowing. Replicas were puriWed for 30 min in 70% sulfuric acid, then neutralized in water and subsequently subjected to 14% sodium hypochloride solution (for 3-5 min), followed by three washing steps in distilled water and immobilization on 400-mesh TEM copper grids (Agar ScientiWc, Stansted, UK).
For freeze-drying, bacteria were resuspended in 50 mM Tris/HCl buVer (pH = 7.2), disrupted with a Branson model 250 soniWer (Branson Ultrasonics, Danbury, CT, USA) for 40 s with 50% output (no cooling), and centrifuged (1,300£g, 10 min). The pellet was resuspended in 50 mM Tris/HCl buVer (pH = 7.2, containing 1% Triton X-100), incubated for 20 min at 25°C, and centrifuged. After two further washing steps with Tris/HCl buVer (pH = 7.2), the pellet was resuspended in water. For adsorption of disrupted bacterial cells, a freshly prepared Formvar and carbon-coated 300-mesh TEM copper grid (Agar ScientiWc) was Xoated facedown for 1 min on one drop of bacterial suspension. After removal of excess of liquid from the grid with Wlter paper, it was rapidly frozen in nitrogen-cooled liquid CHClF 2 and immediately loaded onto a nitrogencooled specimen holder for freeze-drying in the BAF 060 unit (Leica). Freeze drying was performed at ¡80°C for 2 h as described previously (Messner et al. 1986).
All samples were investigated in a Tecnai G 2 20 Twin transmission electron microscope (TEM; FEI, Eindhoven, The Netherlands), operating at 80 keV. Pictures were taken with an FEI Eagle 4 k CCD camera (4,096 £ 4,096 pixels). The magniWcation was calibrated by using negatively stained catalase crystals (Wrigley 1986). Image processing was done with software developed in house based on Fourier domain techniques according to Amos et al. (1982) and Crowther et al. (1996). S-layer lattice parameters were obtained from the power spectra.
ImmunoXuorescence staining and Xuorescence microscopy of T. forsythia wild-type cells T. forsythia wild-type and mutant cells were washed twice with PBS buVer and unspeciWc-binding sites were blocked (1 h, 25°C, constant stirring) in PBS containing 5% BSA (blocking solution). Bacteria were exposed to primary polyclonal rabbit anti-TfsA antiserum diluted 1:20 in blocking solution (1 h, 25°C, constant stirring). After removal of unspeciWc binders by washing the bacteria three times with Tris-buVered saline containing 1% Tween-20 (TBST), the secondary anti-rabbit tetramethyl rhodamine isothiocyanate (TRITC)-conjugate antibody (Sigma-Aldrich) was diluted to a Wnal concentration of 0.15-0.3 mg/mL with blocking solution and added to the bacteria, followed by exposure for 1 h at 25°C under constant stirring. After three washing steps with TBST and one step with PBS, bacteria were resuspended in PBS and analyzed with an Eclipse TE2000-S Xuorescence microscope (Nikon, Vienna, Austria). Pictures were taken with a DS-Qi1 Mc-camera (Nikon), and the same image section was recorded in bright-Weld phase contrast, as well as in the TRITC channel, the respective overlay was generated with NIS Elements BR software (Nikon).
Atomic force microscopy (AFM) of T. forsythia wild-type cells T. forsythia wild-type cells were washed twice in PBS and immobilized by mechanical trapping on 0.8-m polycarbonate membranes (Millipore) for non-invasive in vivo imaging (Dupres et al. 2009). After Wltration of a washed bacterial suspension, the Wlters were gently rinsed with PBS buVer and attached to the sample holder using a double-side adhesive tape, and the mounted sample was measured in the AFM liquid cell. All AFM measurements were performed in PBS buVer and in deionized water, using a commercial Agilent 5500 AFM (Agilent Technologies, Chandler, AZ, USA). Magnetically coated AFM cantilevers with a nominal spring constant of 0.1 N/m were used for magnetic AC mode (MAC mode) imaging (Tang et al. 2008). The resonance frequency of the cantilever was selected between 9 and 11 kHz in liquid, and the measurement frequency was set to 20% below the resonance frequency. The Gwyddion image viewer open source program was used to analyze topographic images of the surface, as well as polynomial Xattening analysis.
For AFM experiments, the rTfsA and rTfsB proteins alone, an equimolar mixture of those as well as native TfsB-GP were dialyzed against MQ water; the protein monomers were diluted in crystallization buVer (10 mM CaCl 2 , 0.5 mM Tris/HCl, pH = 8.0) to a Wnal concentration of 1-10 mg/mL and applied to silicon wafers. For the rTfsB protein and TfsB-GP, the silicon wafer was additionally coated for 20 min with 1 mg/mL polyethylenimine (PEI) in distilled water. Crystallization was carried out for 12 h at 25°C, and AFM measurements were performed in contact mode with a Nanoscope IIIa multimode (Veeco Instruments, Santa Barbara, CA, USA) with a J-scanner (nominal scan size 130 mm). Scanning of samples was performed at approximately 4 Hz in 100 mM NaCl to avoid electrostatic repulsion between tip and sample, and the applied force during scanning was minimized to prevent the tip from modifying the sample. For optimal resolution, standard oxide-sharpened silicon nitrite cantilevers with a nominal spring constant of 0.06 N/m (NP-S, NanoProbes, Digital Instruments, Santa Barbara, CA, USA) were used for imaging. The Nanoscope 5.0 imaging software (Veeco) was used to analyze and process (Wrst order Xattening) topographic images of the surface.

Detection of T. forsythia S-layer (glyco)proteins in cell extracts
The presence of the two S-layer glycoproteins TfsA-GP and TfsB-GP was demonstrated by SDS-PAGE of SDSsolubilized whole-cell extracts upon CBB protein staining ( Fig. 1a) and PAS staining for carbohydrates , with these S-layer glycoproteins corresponding together to approximately 10% of whole cellular protein according to intensity signal integration. Western blotting using the polyclonal rabbit antibodies raised against the individual S-layer proteins corroborated the presence of the two glycoproteins TfsA-GP and TfsB-GP of T. forsythia exhibiting apparent molecular masses of 230 and 270 kDa, respectively, with the antibodies showing no cross-reactivity (Fig. 1b). In the deletion mutants T. forsythia tfsA, tfsB, and tfsAB, the respective bands were missing (Fig. 1b).
Transcription of tfsA and tfsB genes and S-layer protein quantiWcation qPCR was applied to determine the transcription levels of the S-layer genes tfsA and tfsB in T. forsythia. Absolute quantiWcation was performed by calculating standard curves for both genes (Fig. 2a, b). The absolute copy numbers of tfsA and tfsB from a cDNA preparation were determined from the corresponding standard curves using the C T (cycle threshold) values. As the results suggest, both tfsA and tfsB genes, which form a putative operon structure on the T. forsythia genome (Lee et al. 2006), are uniformly transcribed into mRNA (Fig. 2c). For both genes, undiluted and 10-fold diluted samples gave comparable absolute copy numbers. Minor contamination of the cDNA template with genomic DNA was detected, but the overall inXuence on the experiment was negligible (data not shown). Melting curve analysis of both ampliWed products conWrmed the speciWcity of the respective PCRs (data not shown).
Semi-quantiWcation of the protein expression levels of TfsA-GP and TfsB-GP from T. forsythia wild-type cells and S-layer mutants after separation on an SDS-PA gel (compare with Fig. 1a) corroborated the qPCR results (Fig. 2d). Using T. forsythia wild-type cells, both S-layer proteins exhibited a comparable intensity signal. Interestingly, the intensity signals obtained for TfsA-GP and TfsB-GP glycoproteins from S-layer single mutants were comparable to those from wild-type cells, where both S-layer glycoproteins are present. Thus, deletion of one of the S-layer genes does not inXuence the overall expression level of the respective other protein, ruling out the option that one S-layer protein might replace the other in the mutant for formation of the mature S-layer lattice.
According to these data, the S-layer genes tfsA and tfsB are both transcribed into mRNA and expressed into protein in equal ratios. However, based on this analysis, it cannot be discerned whether a single S-layer would form upon co-assembly of the two S-layer glycoproteins species (TfsA-GP, TfsB-GP) or whether the two species would assemble into two independent, superimposed monolayers.
Visualization of the native S-layer lattice on thin-sectioned, negatively stained, freeze-etched, and freeze-dried preparations of T. forsythia cells TEM analysis of the S-layer on T. forsythia wild-type cells as well as on the S-layer deletion mutant cells was Wrst performed by ultra-thin sectioning (Fig. 3a-d).
Typically, T. forsythia wild-type cells are approximately 3.5 § 1 m in length and 0.6 § 0.1 m in diameter (not shown). They possess a typical Gram-negative cell envelope proWle consisting of a cytoplasmic membrane, a periplasm, and an outer membrane. Additionally, this outer membrane is completely covered by an S-layer with a thickness of approximately 22 nm, as determined from thin sections (Fig. 3a). Our data from cross-sectioned T. forsythia wild-type cells show a continuous, cross-sectioned S-layer, deWned as a 2D protein crystal, in a strikingly clear way (Fig. 3a). The cells in which one of the S-layer proteins was deleted, T. forsythia tfsA and T. forsythia tfsB, respectively, the S-layer as such is not visible (Fig. 3a, b). Instead, an additional "gray mass" without any periodicity (compare with the zigzag in Fig. 3a), which would reXect the presence of a 2D crystalline arrangement of the S-layer protein, is present on top of the outer membrane with a similar width as the S-layer on the wild-type cells. The T. forsythia tfsAB cells (Fig. 3d) are the devoid of any "stained mass" outside of the outer membrane (compare with Fig. 1a, lane 4).
A Wrst impression of the regularity of the T. forsythia S-layer lattice in a 2D projection was obtained after negative-staining of T. forsythia wild-type cells. A representative image exhibiting the regular array of glycoproteins was obtained from those areas of the cell where, due to the adsorptive attachment to the grid, an almost complete Xattening of the cell body took place leaving the "top" and the "bottom" S-layer in close vicinity (Fig. 4). The power spectrum of the resulting Moiré pattern (Glauert 1966;Valamanesh et al. 2011) conWrmed that in these preparations, two single S-layers are lying on top of each other (Fig. 4, inset). This analysis already indicates that a superimposition of two single S-layers on intact cells is unlikely, since this would have resulted in four superimposed monolayers in this negatively stained preparation, given that the uranyl acetate staining solution can penetrate between the double layers. Fourier domain image analysis was used to determine both the lattice symmetry and the lattice constant of the S-layer lattice on the wild-type organism. For the negatively stained samples, an S-layer lattice with square symmetry exhibiting an average (n = 3) lattice constant of 10.3 § 0.7 nm, with = 90° § 4.3° was determined. However, neither of the negatively stained preparations of the T. forsythia S-layer single-mutant cells showed a regular S-layer lattice in the TEM analysis, which lead us to speculate about a scenario of co-assembly of TfsA-GP and TfsB-GP to form the native S-layer lattice as present on the T. forsythia wild-type organism.
Analysis of the ultrastructure of the cell surface of T. forsythia was also carried out by freeze-etching (Fig. 5a, b) and freeze-drying experiments (Fig. 5c). In both preparations, the wild-type cells exhibited a similar surface pattern for the cylindrical part of the cell envelope (Fig. 5a) as determined by negative-staining (Fig. 4), reXecting the architecture of a square S-layer lattice (Fig. 5a). Fourier domain image analysis conWrmed the presence of a square lattice on the cell surface of T. forsythia wild-type cells with an average (n = 5) lattice constant of 10.3 § 0.3 nm and = 90° § 5.0° according to the freeze-etching experiments. A rare view onto the region of the cell pole of a single cell showed grain boundaries between diVerent S-layer crystallites (Fig. 5b) as they have already often been observed on other S-layer-covered cells (Sleytr 1978;Pum et al. 1991;Kingl et al. 2011). In freeze-dried preparations, only T. forsythia wild-type cells showed a clearly recognizable S-layer lattice with average (n = 5) lattice parameters of 10.8 § 0.5 nm and » 90° § 3.0°, which is in agreement with the data from freeze-etching and negative-staining experiments (see above).
While visualization of the S-layer lattice was straightforward on the cell surface of the wild-type organism upon freeze-etching, this could be achieved only on a few cells of the S-layer single mutants T. forsythia tfsA and tfsB; with the lattice being less clearly visible; this conWrmed the results from the negative staining experiments. These mutant cells showed a faint regular surface structure with lattice parameters of 8.6 nm x 8.7 nm and » 90° for were obtained by amplifying the tfsA or tfsB genes using diVerent dilutions of the tfsAB operon (puriWed and quantiWed PCR product used as template) in ascending order. Black circles indicate calculated C T values for analyzed samples. c Absolute quantiWcation of tfsA/tfsB gene transcription in T. forsythia wt cells. Undiluted and 10-fold diluted cDNA was used as template and consistent absolute transcription levels (copy numbers) were determined for both genes. d Relative protein expression levels of TfsA/TfsB in T. forsythia wild-type cells and respective mutant strains. 0.7 mg of total crude cell extract applied to each lane in SDS-PAGE T. forsythia tfsA cells (Fig. 5e) and 9.7 nm £ 8.7 nm and » 90° for T. forsythia tfsB cells (Fig. 5f). Thus, obviously none of the mutant cell S-layer proteins alone is capable of forming a lattice identical to that on native cells, although both still possess some residual self-assembly capability.
Although T. forsythia has been described as a non-motile bacterium (Tanner et al. 1986;Sharma 2010), we found occasionally Xagella with diameters of approximately 20 nm and pilus-like structures, obviously randomly distributed over the S-layer surface during the sublimation of ice in the freeze-etching experiments (Fig. 6a-c). In addition, there is evidence of the presence of hook regions as required for anchoring the Xagella to the outer membrane (Fig. 6a, c). The number of the presumably peritrichous Xagella on T. forsythia cells seems to be low, whereas the pilus-like structures occur more frequently.

AFM imaging of bacterial cells
Scanning probe microscopy was used in the MAC mode to image the cell surface of T. forsythia wild-type cells by investigating the S-layer ultrastructure. Cells were trapped within a Wlter (Dufrêne 2002), because this method appears promising as a generally applicable immobilization method for bacteria of various shapes and sizes. Since it does not require exposure to chemicals, or drying, it does not aVect cell viability or trigger undesirable biological responses. AFM height and amplitude images of T. forsythia wild-type cells are shown in Figure 7. Because of the large curvature of the bacterial cells, the topographic image (Fig. 7a) has fairly poor resolution, whereas the amplitude image (Fig. 7b) is much more sensitive to changes in the surface topography. The inset in Figure 8a was Xattened with a twelfth-order polynomial. The latter image shows that In either S-layer single mutant the S-layer is not visible as a cross-sectioned 2D-crystal, but an unstructured "stained mass" without periodicity is present on top of the outer membrane, which is less densely packed than the S-layer on wild-type cells. d T. forsythia tfsAB double mutant without an S-layer. The white bar indicates that the overall thickness of the S-layer is identical in T. forsythia wt and T. forsythia S-layer single-mutant cells. SL S-layer, OM outer membrane, PP periplasmic space, CM cytoplasmic membrane, C cytoplasm native T. forsythia wild-type cells are featured with a periodic square lattice structure of a center-to-center spacing in the range of 9.1 § 0.8 nm and an angle of » 90.0° § 4.8°. These values are in agreement with the TEM analyses described above.
ImmunoXuorescence staining of S-layer proteins on T. forsythia cells To prove our hypothesis of the presence of both S-layer glycoproteins on the cell surface of T. forsythia wild-type cells, immunoXuorescence microscopy in conjunction with the polyclonal antibodies raised against the individual rTfsA and rTfsB proteins was used. Prior to this experiment, the puriWed antibodies were shown to recognize and bind the respective target protein and to possess no crossreactivity in Western blot analysis (Fig. 1b).
The presence of the TfsA-GP could be imaged by a Xuorescence signal derived upon probing T. forsythia wild-type and T. forsythia tfsB cells with a TRITC-conjugated anti-rTfsA antibody (Fig. 8a-d), while no signal was obtained from the T. forsythia tfsA and T. forsythia tfsAB control cells (Fig. 8e, f). Surprisingly, the presence of the TfsB-GP could not be visualized in an identical experimental setup using the corresponding TRITC-conjugated anti-rTfsB antibody (not shown). To exclude the possibility that due to steric hindrance the antibody could not access the TfsB protein, which we assumed to be part of the co-assembled S-layer lattice, the experiment was repeated using FITCconjugated Fab fragments of the TfsB antibody. However, again, no signal was obtained with either T. forsythia wildtype or T. forsythia tfsA cells, and also the controls were negative.
These data allow diVerent interpretations: (1) since the denaturated rTfsB protein was used for raising antibodies, it might well be that the antibody does not recognize the native epitopes of the folded S-layer protein as present on the cell surface; and (2) the recognition sites of the folded S-layer protein are not accessible for the antibody, because they might be orientated toward the underlying outer membrane or buried within the S-layer.

Re-assembly experiments of native and recombinant T. forsythia S-layer (glyco)proteins
Although the microscopic investigations of native T. forsythia cells already indicated that most likely coassembly of the TfsA-GP and the TfsB-GP does occur under native conditions, re-assembly experiments were also carried out in vitro, using puriWed S-layer glycoproteins as well as the recombinant S-layer proteins rTfsA and rTfsB, since both the S-layer proteins as well as the glycosylated forms, per deWnition, should possess inherent self-assembly capability.
For most of the chosen conditions (varying buVers, molarities, and pH values), self-assembly of individual native and recombinant T. forsythia (glyco)proteins could not be demonstrated in vitro according to analyses of negatively stained preparations. Only upon dialysis either against Tris/CaCl 2 buVer (pH = 8.0) or tap water, the (glyco)proteins were prone to impaired self-assembly with exclusively multilayered stacks of (glyco)proteins being formed, with Moiré patterns displaying striations of diVerent width and diVerent angles to each other, which cannot be clearly traced back a speciWc lattice symmetry (not shown). A comparable result was obtained when an equimolar mixture of native or recombinant T. forsythia (glyco)proteins, respectively, was analyzed for self-assembly behavior, attempting to mimic the most likely native situation (data not shown). Although it is generally known that covalently bound glycans, as present on the native T. forsythia proteins, can inXuence proper protein folding and, consequently, self-assembly, our results suggest that self-assembly of T. forsythia S-layer (glyco)proteins is not aVected by the glycosylation status.
The impaired self-assembly capability of the S-layer (glyco)proteins in vitro was also observed in the AFM studies. For these approaches, a monomer solution obtained by Fig. 4 Negative-staining of a T. forsythia whole cell preparation. During adsorption of the cell to the EM grid parts of the cell were Xattened so that the cytoplasm was squeezed away and, therefore, the "upper" and "lower" S-layer came into close distance. The inset shows the resulting power spectrum dialysis of a solution of the recombinant S-layer proteins rTfsA and rTfsB as well as of TfsB-GP against distilled water was used (Chung et al. 2010). In the AFM measurements, most of the recombinant S-layer protein was found to adsorb in the form of unstructured patches on the chosen support. Again, as a rare event, for rTfsA, an S-layer lattice with 13.9 nm £ 12.0 nm and = 80° could be resolved on a silica support (Fig. 7c, d), while for rTfsB, a regular protein lattice with the parameters of 10.0 nm £ 10.0 nm and = 73° could be detected on a PEI-coated silicon surface  (Fig. 7e, f). The same result was obtained for the corresponding glycoprotein TfsB-GP (not shown). It is important to note that the PEI coating of the support did not reveal any regular features (not shown). When co-assembly of rTfsA and rTfsB was attempted, exclusively unstructured patches were seen on the AFM support (not shown).
Generally, the AFM Wndings support the data from the negative staining and low-temperature preparation method experiments, indicating a strongly impaired self-assembly capability of T. forsythia S-layer proteins in vitro under the chosen conditions and/or a certain degree of structural Xexibility of the individual native and recombinant T. forsythia S-layer (glyco)proteins alone after self-assembly.

Discussion
Periodontitis is initiated due to colonization of the oral cavity by a group of Gram-negative anaerobes in the form of a subgingival bioWlm. The disease progresses as a result of the direct eVects of bacterial virulence factors on host tissues, as well as self-damaging host responses to the colonizing bacteria (Socransky et al. 1998). Among the few putative virulence factors that have been identiWed in the oral pathogen T. forsythia are two components which, due to their localization on the bacterial cell surface, can be expected to be involved in initial stages of the bacteriumhost cross talk; these are a leucine-rich repeat cell surfaceassociated and cell surface-secreted protein BspA (Sharma et al. 1998) and the S-layer (Sabet et al. 2003;Sakakibara et al. 2007;Sekot et al. 2011). This study focused on the ultrastructural characterization of the S-layer of T. forsythia as outermost cell surface layer by a set of microscopic, biochemical, and genetic methods, aiming at the determination of S-layer lattice parameters. T. forsythia presents a situation so far unknown for Gram-negative bacteria, which is the presence of two S-layer proteins that are additionally posttranslationally modiWed with so far unique O-glycosidically linked oligosaccharides .
In the course of this study, it was demonstrated that the S-layer genes tfsA and tfsB are both transcribed into mRNA and expressed into protein in equal amounts, leaving the option for superimposition of two S-layers or co-assembly of TfsA-GP and TfsB-GP into a single S-layer lattice that covers the T. forsythia cells. While TEM analysis of wildtype cells revealed the presence of a square S-layer lattice with average lattice parameters of 10.3 § 0.3 nm and = 90° § 5.0° (freeze-etching) and of 10.3 § 0.7 nm and = 90° § 4.3° (negative staining), a more blurred lattice structure with a smaller lattice constant of approximately 9.0 nm (freeze-etching) was found on some surface areas of those Tannerella cells that were deWcient in expressing either of the two S-layer glycoproteins (T. forsythia tfsA, T. forsythia tfsB). We interpret this slight decrease in the lattice constant as a consequence of the individual S-layer glycoproteins assuming a diVerent conformation depending on the composition of the S-layer, that is, "mono-species" S-layer versus "co-assembled" S-layer lattice.
Despite the fact that we do not have a Wnal immunological proof for the presence of the TfsB-GP on the cell surface of T. forsythia wild-type cells, based on the results from the TEM investigations, we favor the assumption of TfsA-GP and TfsB-GP forming the mature S-layer lattice by equimolar co-assembly. This is vaguely supported by previous studies by others, where the presence and accessibility of the TfsB-GP on T. forsythia wild-type cells was shown by TEM using a gold-conjugated anti-TfsB-GP antibody (Higuchi et al. 2000;Moriguchi et al. 2008;Sakakibara et al. 2007). However, since that antibody also reacted with the native TfsA-GP, it might be that it preferentially recognized the S-layer glycans on the proteins, which we could show to be identical .
While the impaired self-assembly capability of the recombinant S-layer proteins might be explained by the fact that the chaotropic agent (5 M guanidine hydrochloride) used for their isolation causes irreversible conformational changes that might prevent, subsequently, proper selfassembly, no comparable stress is exerted on the S-layer glycoproteins of the intact T. forsythia S-layer single mutants. This together with the results from in vitro selfassembly studies indicate an impaired self-assembly capability of the individual S-layer (glyco) proteins, which makes them distinctly diVerent from most other known S-layer (glyco)proteins (Toca-Herrera et al. 2005;Messner et al. 2010), which usually retain their self-assembly capability in vitro.
Another Wnding of this study concerns, for the Wrst time, the visualization of Xagella and pilus-like structures on T. forsythia cells. Frequently, bacteria and archaea expose on their outer surfaces a variety of thread-like proteinaceous organelles with which they interact with their environments (Ghosh and Albers 2011;Sauer et al. 2000). These structures are repetitive assemblies of protein subunits and serve diVerent roles in cell motility, adhesion and Fig. 7 AFM a topography and b amplitude image of T. forsythia wild-type cells. The inset in a shows a zoomed part of the bacterial surface. AFM analysis of in vitro self-assembly experiments of the recombinant Slayer proteins rTfsA (c, d) on silicon and of rTfsB (e, f) on PEIcoated silicon. c, e Height images and d, f deXection images and image reconstruction of the 2D crystalline S-layer lattice (insets) of rTfsA (d) and of rTfsB (f) host invasion, protein and DNA secretion and uptake, conductance, or cellular encapsulation. With regard to oral bacteria, adherence to a surface is a key element for colonization of the human oral cavity (Whittaker et al. 1996). Thus, it is conceivable that these structures contribute to the pathogenicity potential of T. forsythia. There are many pathogens known for which, especially glycosylated, Xagella mediate virulence-associated phenomena (Thibault et al. 2001;Schirm et al. 2003). Considering the so far unique glycosylation of the S-layer proteins TfsA and TfsB from T. forsythia , one might speculate about glycosylation of these surface appendages. Such a situation is, for instance, known from archaea, where S-layer proteins and Xagella are decorated with the same glycan structures (Voisin et al. 2005;Jarrell et al. 2010Jarrell et al. , 2011. Consequently, these surface appendages might be of high importance for future studies with T. forsythia. Summarizing, the most likely situation on T. forsythia wild-type cells is the presence of the two intercalating S-layer glycoproteins TfsA-GP and TfsB-GP, forming a square S-layer lattice (Fig. 9a). The slight reduction in the lattice constant of the S-layer lattices present on the individual S-layer gene mutants might be explained by a denser packing of the protomers when the space-Wlling counterpart is missing (Fig. 9b, c). It is conceivable that this so far unique ultrastructure of an S-layer is of relevance for the complex reactions occurring during the bacterium-host cross talk, such as the adhesion in the oral cavity under physiological as well as pathological conditions.   , c). Scheme drawn not to scale hosting G. Sekot to perform experiments on the Cressington CFE-50 unit and E. Wäger, Baxter Innovations GmbH, Vienna, Austria, for production of T. forsythia S-layer antibodies. This work was supported by the Austrian Science Fund projects P20605-B20 and P21954-B20 (to C. S.) and P22791-B11 (to P. M.).
Open Access This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.