FadA from Fusobacterium nucleatum Utilizes both Secreted and Nonsecreted Forms for Functional Oligomerization for Attachment and Invasion of Host Cells*

Fusobacterium nucleatum is a Gram-negative anaerobe associated with various human infections, including periodontal diseases and preterm birth. A novel FadA adhesin was recently identified for host-cell binding. It consists of 129 amino acid residues, with an 18-amino acid signal peptide. Expression of FadA in Escherichia coli enhanced bacterial binding to host epithelial and endothelial cells. In both E. coli and F. nucleatum, FadA exists in two forms, the intact pre-FadA and the secreted mature FadA (mFadA), with pre-FadA anchored in the inner membrane and mFadA secreted outside the bacteria. Pre-FadA and mFadA formed high Mr complexes. When each form was purified to a single species, mFadA was soluble at neutral pH, whereas pre-FadA was insoluble. Pre-FadA became soluble when mixed with mFadA or under acidic pH. When fluorescence-labeled mFadA alone was added to the epithelial cells, no binding was detected. However, when mixed with nonlabeled pre-FadA, binding and invasion of mFadA into epithelial cells was observed. FadA is a unique bacterial adhesin/invasin in that it utilizes its own two forms for both structural and functional purposes. The pre-FadA-mFadA complex is probably anchored in the inner membrane and protrudes through the outer membrane. Internalization of the pre-FadA-mFadA ensures invasion of the bacteria into the host cells.

Fusobacterium nucleatum is a Gram-negative anaerobe associated with various human infections. It is ubiquitous to the oral cavity and is implicated in periodontal diseases (1). The organism coaggregates with microbial species in the oral cavity, playing a critical role in periodontal plaque formation (2). It is also isolated from infections and abscesses of other parts of the body and is one of the most prevalent species in human intrauterine infections (3)(4)(5). F. nucleatum may translocate from the oral cavity to different sites in the body hematogenously and cause localized abscess or infection (6). The organism binds to and invades different types of host cells. Attachment and invasion of epithelial and endothelial cells by F. nucleatum was observed in vitro, which elicited proinflammatory responses (7,8). Attachment and invasion of endothelial cells was also observed in vivo in infected mouse placentas, leading to bacterial colonization in the placenta and resulting in adverse pregnancy outcomes (7).
It has been postulated that F. nucleatum may possess lectinlike and nonlectin-like adhesins for binding to various partners (9 -12). Different F. nucleatum strains may bind to the same partner via different adhesins (9). The same adhesin(s) may also be involved in binding to different partners (12). Several putative adhesin molecules have been suggested for F. nucleatum for involvement in binding to other microbial species or human IgG (13)(14)(15)(16)(17). However, none has been characterized, and it was not known if they were involved in bacterial binding to the host cells.
A novel adhesin, FadA, from F. nucleatum 12230 was recently identified to be involved in attachment to host epithelial cells (18). The fadA gene was highly conserved among oral fusobacterial species, including F. nucleatum, Fusobacterium periodonticum, and Fusobacterium simiae, but was absent from the nonoral fusobacteria (18). FadA consists of 129 aa residues, with the first 18 encoding a typical signal peptide (MKKFLL-LAVLAVSASAFA) (18). It was not known, however, if the signal peptide was cleaved during secretion in F. nucleatum. Based on the amino acid sequence, FadA appeared to be predominantly ␣-helical. Using a novel gene disruption technique, sonoporation, we constructed the first double cross-over allelic exchange mutant of F. nucleatum, US1, carrying a deletion of fadA. Binding of US1 to the oral mucosal cell KB and Chinese hamster ovarian (CHO) 3 cells were each reduced by 70 -80% compared with the wild-type strain. Therefore, FadA was involved in binding to epithelial cells (18). In the current study, the FadA adhesin was expressed in Escherichia coli as a His tag fusion protein, which was purified and characterized. The recombinant FadA not only attached to but also invaded the host cells as a heterogenous complex. The formation of such a complex required both the intact and secreted forms of FadA, and the complex is probably anchored in the inner membrane and protrudes through the outer membrane.

EXPERIMENTAL PROCEDURES
Bacterial Strains, Plasmids, and Culture Conditions-The bacterial strains and plasmids used for this study are listed in Table 1. All F. nucleatum strains were propagated as described (19). All E. coli strains were maintained in LB broth (Difco) or on LB agar (Difco) and incubated at 37°C in air. Plasmid pYWH417-6 was constructed as follows. The fadA gene from F. nucleatum 12230 was amplified from pYWH401 using primers fadAtop-NdeI (5Ј-GGAATTCCATATGAAAAAATTTTTA-TTATTAGCAG-3Ј) and fadAbottom-XhoI-His tag (5Ј-CCG-CTCGAGGTTACCAGCTCTTAAAGCTTG-3Ј), using Vent DNA polymerase (New England BioLabs, Ipswich, MA). The PCR fragments contained NdeI and XhoI sites (underlined) at their ends, respectively. Following endonuclease digestion and purification using the QIAquick gel extraction kit (Qiagen, Valencia, CA), the fragment was cloned into pET21(b) ( Table 1) at the NdeI and XhoI sites, followed by transformation into competent E. coli DH5␣ (Invitrogen).
DNA Sequencing Analysis-Plasmids were purified from DH5␣ using Wizard Plus Midipreps DNA Purification System (Promega, Madison, WI) and their DNA sequences analyzed at the Molecular Biotechnology Core (Lerner Research Institute, Cleveland, OH) using the T7 terminator and promoter primers. Once the constructs were verified, the plasmids were transformed into BL21(DE3) for protein expression.
Expression of the Recombinant Proteins-E. coli BL21(DE3) carrying pET21(b), pYWH417-6, or the mutant plasmids were grown to an A 600 of 0.5. The cultures were induced with isopropyl ␤-D-1-thiogalactopyranoside (IPTG) (Sigma) at a final concentration of 0.1 mM for 2-3 h unless otherwise specified.
Tissue Cell Attachment Assay-CHO (ATCC CRL10154), human umbilical vein endothelial cells (ATCC CRL-1730), oral mucosal cell line KB (ATCC CCL-17, known to be contaminated with HeLa markers), and primary cultures of human gingival epithelial cells were all maintained as previously described (7,8,18). Immortalized human oral keratinocytes OKF6/Tert cells were obtained from Dr. J. Reinwald (Harvard University, Boston, MA) and were grown as previously described (20). The attachment assays were carried out as previously described (8,18). Briefly, cells were seeded into 24-well trays and allowed to grow to near confluence. The bacteria were then added to the monolayers at a multiplicity of infection of 50 -150:1. For inhibitory attachment assays, purified mFadA or reconstituted pre-FadA-mFadA complex was added to the monolayers prior to the addition of bacteria. Following a 1-h incubation at 37°C under 5% CO 2 , the monolayers were washed with phosphate-buffered saline (PBS; Sigma) and lysed with water. It was shown previously that water lysis under the test conditions did not affect bacterial viability (8,18). Serial dilutions were plated onto agar plate, followed by incubation under appropriate conditions to allow for growth of the total cell-associated bacteria (7,8). The bacterial colonies were enumerated, and the levels of attachment were expressed as the percentage of bacteria recovered following cell lysis relative to the total number of bacteria initially added. Each experiment was performed in triplicate.
Purification of Recombinant FadA Mixture under Denaturing Conditions-The bacterial pellet (about 1 g) from 250 ml of IPTG-induced culture was collected by centrifugation at 3000 ϫ g for 10 min, resuspended in 5 ml of buffer A (50 mM NaH 2 PO 4 , 0.3 M NaCl, 8 M urea, pH 8.0), and incubated at room temperature for 1 h. Clear lysate was collected by centrifugation (7000 ϫ g, 20 min) and mixed with 1 ml of TALON cobalt resin (BD Clontech, Mountain View, CA) for 1 h at room temperature. The mixture was transferred to a small column. After all of the solution was removed, the resin was washed with 6 ml of buffer A and eluted with 4 ml of buffer B (50 mM NaH 2 PO 4 , 0.3 M NaCl, 8 M urea, pH 5.0). The eluted sample (4 ml) was dialyzed extensively against 10 mM Tris-HCl, pH 7.4, at 4°C, in a dialysis tubing with M r cut-off of 6000 -8000. The protein concentration was determined using the BCA kit (Pierce).
Matrix-assisted Laser Desorption Ionization Time of Flight Mass Spectrometry (MALDI-TOF MS)-Protein bands from SDS-PAGE were excised and eluted from the gel slices. The gel-purified proteins were dialyzed against PBS and analyzed by MALDI-TOF MS. The mass spectrometer (Biflex III; Bruker Daltonics, Billerica, MA) was used at an acceleration voltage of 20 kV operated in the linear mode with delayed extraction and a pulsed nitrogen laser source ( ϭ 337 nm). Samples were mixed with an equal volume of a saturated matrix solution (15 mg/ml sinapinic acid in 60% acetonitrile and 0.3% trifluoroacetic acid). One microliter of the mixture was applied onto the laser target probe and was air-dried before being introduced into the mass spectrometer. A total of 200 -300 spectra were obtained for each sample. The spectra were averaged and analyzed using the Xacq software provided by the manufacturer. The instrument was calibrated using a mixture of standard proteins. A mass accuracy of 0.02-0.08% was routinely obtained.
Production of Polyclonal Antibodies against FadA-Two 12-week-old male New Zealand White rabbits were each injected intradermally with 500 g of mFadA in Freund's complete adjuvant (Sigma), followed by three weekly boosts 3 weeks later, by injecting 500 g of FadA in Freund's incomplete adjuvant each time. Antisera were collected 1 week after the final boost.
Production of Monoclonal Antibodies against FadA-The mouse anti-FadA monoclonal antibody (mAb) 5G11-3G8 was produced at the Hybridoma Core (Lerner Research Institute) following standard procedures. Briefly, the hybridomas secreting mAb were derived from the BALB/c mice immunized with recombinant mFadA. Antibodies of the desired specificity were identified by their specific binding to FadA in enzyme-linked immunosorbent assay, followed by Western blotting analysis using purified FadA proteins and F. nucleatum 12230. One of the hybridoma clones was designated as 5G11-3G8. The mAb from this clone was obtained from the serum-free culture, followed by purification using the protein G column, and stored at a final concentration of 4 mg/ml.
Gel Filtration Column Chromatography-Approximately 1.8 mg of protein in 1 ml of elution buffer (20 mM Tris-Cl, 0.1 M NaCl, pH 7.5) was applied to a Sephacryl S-300 column (1.6 ϫ 60 cm; Amersham Biosciences), connected and operated by a BioLogic LP system (Bio-Rad). The protein was eluted at a flow rate of 0.5 ml/min at 4°C, fractions were collected at 2 or 2.5 ml/tube, and the absorption at 280 nm was recorded. A standard curve was generated using blue dextran (2000 kDa), bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa), and cytochrome c (12.3 kDa), all purchased from Sigma.
Bacterial Fractionation-Bacteria were collected by centrifugation, washed twice with PBS, and resuspended in 20 mM Tris-HCl, 0.1 mM phenylmethylsulfonyl fluoride, pH 7.4. All samples were kept at 4°C unless otherwise stated. Bacteria were disrupted by sonication in an ice-water bath using a model 60 Sonic Dismembrator (Fisher) (15 s with a 45-s interval, 20 cycles) with an output power of 18 watts for a total of 5 min. The lysate was centrifuged at 4°C at 3000 ϫ g for 15 min and again at 7000 ϫ g for 10 min to remove unbroken cells. The supernatant was then subjected to ultracentrifugation at 100,000 ϫ g for 1 h at 15°C, producing the clear supernatant (i.e. cell extract). The pellet was washed with ice-cold 20 mM Tris, 0.1 mM phenylmethylsulfonyl fluoride, pH 7.4, resuspended in the same buffer containing 0.8% N-lauroylsarcosine (Sigma), and incubated at 30°C for 30 min. After ultracentrifugations at 100,000 ϫ g for 1 h at 15°C, the clear supernatant was saved as the inner membrane fraction. The procedure was repeated once, and the pellet was resuspended in 20 mM Tris, 0.1 mM phenylmethylsulfonyl fluoride, pH 7.4, and was saved as the outer membrane fraction. The protein concentrations were measured by BCA. To detect mFadA in the culture supernatant of E. coli, 20 ml of the culture supernatant was filtrated through a Nalgene 0.2-m syringe filter, followed by 80% ammonium sulfate precipitation in the presence of 1 mM phenylmethylsulfonyl fluoride at 4°C for 16 h. Sedimented protein was obtained by centrifugation at 10,000 ϫ g at 4°C for 5 min. The proteins were dissolved in 1 ml of 20 mM Tris-HCl buffer, pH 8.0, followed by washing with 20 mM Tris-HCl buffer containing 80% ammonium sulfate and dialyzed against 10 mM Tris-HCl buffer at 4°C for 16 h.
Purification of mFadA-The bacterial pellet from 3 liters of IPTG-induced culture was collected by centrifugation and resuspended in 300 ml of PBS. The suspension was incubated at 60°C for 30 min, followed by centrifugation at 3000 ϫ g for 10 min. The bacterial cell pellet was saved for purification of pre-FadA (see below). The clear supernatant was mixed with 6 ml of TALON cobalt resin in 50 mM NaH 2 PO 4 , 0.3 M NaCl, pH 8.0. After transferring to a column and washing, mFadA was eluted with 20 ml of washing buffer containing 0.15 M imidazole. The eluate was suspended in 8 M urea, pH 4.0, and applied to the High S cationic exchange column (Bio-Rad) pre-equilibrated with buffer C (50 mM acetic acid, 0.3 M NaCl, 8 M urea) at a flow rate of 0.5 ml/min. The column was then washed with 50 ml of buffer C to collect mFadA. The eluted sample was dialyzed against 50 mM phosphate buffer, pH 8.0, at 4°C for 16 h. The protein concentration was determined by BCA.
Purification of Pre-FadA-The bacterial pellet obtained above was resuspended in 120 ml of buffer A and incubated with agitation at room temperature for 1 h. The clear lysate was collected by centrifugation at 7000 ϫ g for 20 min, followed by cobalt column purification as described above. The eluted sample was adjusted to pH 4 with acetic acid and passed through the High S cationic exchange column as described above. Following washes with buffer C, the column was eluted with buffer D (50 mM acetic acid, 1 M NaCl, 8 M urea). Following dialysis, the eluate was applied to preparative 13% SDS-polyacrylamide gels. The pre-FadA band was excised, and the protein was eluted. The eluted sample was dialyzed extensively against 10 mM Tris-HCl, pH 7.4, at 4°C. The pre-FadA precipitation was collected by centrifugation and dissolved in 50 mM acetic acid when needed. The protein concentration was determined by BCA.
Fluorescent Labeling of Proteins-Purified mFadA or cytochrome c were labeled with Alexa-fluor 488 (Molecular Probes, Inc., Eugene, OR) according to the manufacturer's instructions. Briefly, purified mFadA was mixed with Alexa-fluor 488 stock solution at a molar ratio of 5:1 and incubated at room temperature in the dark for 1 h. The labeled protein was purified by a D-Salt dextran desalting column (Pierce) and concentrated by Microcon YM-3 (Millipore).
Reconstitution of pre-FadA-mFadA Complex-Purified mFadA was mixed with purified and acid-dissolved pre-FadA at a molar ratio of 5:1 by slowly adding pre-FadA into mFadA, followed by incubation at 4°C for at least 2 h. The mixture was dialyzed against 10 mM Tris-Cl, 0.1 M NaCl, pH 7.4, with an M r cut-off of 6000 -8000. For the epithelial cell binding assays, fluorescence-labeled mFadA was used for the complex reconstitution (below).
Binding of Fluorescence-labeled Recombinant FadA to OKF6/Tert Cells-Immortalized human oral keratinocytes OKF6/Tert cells were inoculated into a Lab-Tek II chamber slide system (Nalge Nunc International Co., Naperville, IL) and grown until 50% confluent. Following preincubation in culture medium containing 5% bovine serum albumin for 1 h, the fluorescence-labeled proteins were added and incubated overnight at 37°C under 5% CO 2 . The cells were washed four times with Hanks' buffer. The actin and cell nuclei were sequentially stained with 20 M Alexa-fluor 568 phalloidin (Molecular Probes) at 37°C for 15 min and 10 g/ml 4Ј,6-diamidino-2-phenylindole (Molecular Probes) at room temperature for 5 min, respectively. For the competition assays, nonlabeled pre-FadA-mFadA complex (mFadA/ pre-FadA ratio of 5:1) in 20-fold excess or mFadA alone in 20-fold excess was added to the monolayers 8 h prior to the addition of the fluorescence-labeled FadA complex. The fluorescence was observed under an Olympus IX71 microscope using emission filter BA420 and excitation filter BP330-385 for 4Ј,6-diamidino-2-phenylindole (exposure time 2.5 ms), emission filter BA515IF and excitation filter BP460-490 for Alexa-fluor 488 (exposure time 250 ms), and emission filter BA590 and excitation filter BP510-550 for Alexa-fluor 568 (exposure time 500 ms). The images were captured using an Olympus DP70 camera operated with DP controller software version 1.2.1.108.
Scanning Confocal Laser Microscopy-The binding assay described above was performed in glass bottom culture dishes (MatTek Co., Ashland, MA). Following the assay, the cells were washed three times with Hanks' buffer and fixed with 3.7% formaldehyde at room temperature for 15 min. The cells were then permeabilized with 0.1% Triton X-100 at room temperature for 5 min. The cell nuclei and actin were stained with 20 M Draq5 (AXXORA, San Diego, CA) and Alexa-fluor 568 phalloidin, respectively, at 37°C for 15 min.
After washing with Hanks' buffer, the slides were examined with a scanning confocal laser microscope (model LSM510; Carl Zeiss, Jena, Germany). The argon and helium/neon-2 lasers were used to excite the fluorescent dyes, Alexafluor 488 and 568 and Draq5, respectively. The resulting corrections of images were processed for display by using Java-based program ImageJ for OSX (version 1.33u; W. Rasband, National Institutes of Health) with the LSM reader plug-in.

RESULTS
The fadA gene was amplified from pYWH401 by PCR, cloned into the expression vector pET21(b), and transformed into E. coli DH5␣ ( Table 1). The recombinant plasmid carried a fusion gene expressing FadA with eight additional residues, Leu-Glu-His 6 , at the carboxyl end. Through DNA sequencing analysis, a wild-type FadA fusion construct was identified and designated as pYWH417-6 ( Table 1). A mutant construct was also identified and designated as pYWH417-2, with a Glu-to-Lys substitution at position 26 in the mature form of FadA (mFadA E26K). This mutation was presumably introduced during PCR amplification (Table 1). All three plasmids, pYWH417-2, pYWH417-6, and pET21(b), were transformed into E. coli BL21(DE3). Expression of the recombinant fusion proteins was tested with different amounts of IPTG and induction times, followed by Western blotting analysis using INDIA TM His-Probe-HRP specific for His tag. An optimal amount of FadA-His tag fusion protein was detected from BL21(DE3)/ pYWH417-6 following a 2-h induction with 0.1 mM IPTG (Fig. 1A). Under all conditions tested, no His tag fusion was detected from BL21(DE3)/pET21(b) (Fig. 1A). The apparent M r of the recombinant fusion protein was between 16,000  This study pYWH417-2 Same as above except the E26K mutation in FadA This study and 17,000, higher than the calculated M r of 15,700 for the intact or 13,700 for the mature (secreted) protein (Fig. 1A). The stability of the expressed recombinant proteins was also examined. As shown in Fig. 1B, wild-type FadA-His tag fusion was consistently induced even after 5 h of incubation with 0.1 mM IPTG. The expressed protein appeared to be rather stable, with no detectable degradation following 1 h of incubation on ice. In contrast, expression of the FadA E26K-His tag fusion was much weaker and unstable (Fig. 1B). Moderate expression was detected following 0.5 h of induction with either 0.1 or 1 mM IPTG. However, the protein level quickly decreased with time. It was only weakly detected after 1 h of induction and was not detected at all after 2 h, indicating degradation of the expressed protein. Thus, BL21(DE3)/pYWH417-2 was no longer used in the subse-quent studies. For each strain tested, an equal amount of colony-forming units was loaded onto each lane, as determined by A 600 and verified by SDS-PAGE (data not shown). BL21(DE3) carrying pET21(b) or pYWH417-6 was tested for attachment to host epithelial and endothelial cells following IPTG induction. BL21(DE3)/pET21(b) demonstrated varying background levels of attachment to human gingival epithelial cells, KB, CHO, and human umbilical vein endothelial cells (Fig.  2). For each cell type, the level of attachment was consistently enhanced 3-4-fold when BL21(DE3)/pYWH417-6 was tested (Fig. 2). Thus, the expressed FadA-His tag fusion protein was functional and facilitated E. coli attachment to different host cells.
The recombinant FadA-His tag fusion protein was purified from E. coli BL21(DE3)/pYWH417-6 following IPTG induction using the cobalt column under denaturing conditions. When examined by 15% SDS-PAGE, a major upper band and a minor lower band were detected in the cobalt column eluate, with the upper band migrating near 16 kDa and the lower band migrating faster (Fig. 3A). Each band was gel-purified and subjected to NH 2 -terminal peptide sequencing (Lerner Research Institute) and mass spectrometry analysis. The first 5 aa residues of the upper band were ATDAA, matching those immediately following the putative 18-aa signal peptide. Thus, the upper band is presumably the secreted/mature form of FadA (mFadA). Surprisingly, the first 5 aa residues of the lower band were MKKFL, matching those of the intact form of FadA (pre-FadA). The average molecular mass was 13,641 Da for the upper band and 15,509 Da for the lower band, as determined from the detection of the singly and doubly protonated molecules by MALDI-TOF MS (Fig. 3, B and C). These results were consistent with the calculated molecular mass of the presumed mFadA (13,650 Da) and pre-FadA (15,513 Da), respectively. The noise level of the mass spectra was low, indicating that the two bands did not cross-contaminate. Since the outcome was the opposite of what was expected, with the smaller component migrating more slowly than the larger component on SDS-PAGE, the experiments were repeated to ensure that no mislabeling of samples occurred. Therefore, despite their anomalous migration on SDS-PAGE, two forms of FadA expressed as intact (pre-FadA) and secreted (mFadA) species were correctly assigned following NH 2 -terminal protein sequencing and mass spectrometry. Rabbit polyclonal anti-mFadA antibodies were raised and used to detect the presence of FadA in various F. nucleatum strains. FadA was present in all seven wild-type  strains, F. nucleatum 12230, F. nucleatum ATCC10953, F. nucleatum ATCC25586, F. nucleatum ATCC23726, F. nucleatum ATCC49256, F. nucleatum ATCC51190, and F. nucleatum PK 1594, but was absent in F. nucleatum 12230-US1, the fadA deletion mutant constructed previously in our laboratory (Fig.   4). These results agreed with the previous report on the conservation of fadA among oral fusobacteria and confirmed that the fadA genes were actively transcribed in different strains (18). In all seven strains, an upper and a lower band were detected, although their respective intensities varied from strain to strain. In F. nucleatum ATCC23726, the lower band was more prominent than the upper band, whereas in F. nucleatum ATCC51190, the former was hardly visible (Fig. 4). This could be due to differential expression of FadA in different strains or due to the varying specificity of the polyclonal antibodies for each FadA variant. Nonetheless, the results suggested co-existence of mFadA and pre-FadA in F. nucleatum.
Localization of mFadA and pre-FadA in E. coli BL21(DE3) and F. nucleatum, respectively, was assessed by bacterial fractionation. E. coli BL21(DE3) carrying either pYWH417-6 or pET21(b), and F. nucleatum strains 12230 and US1 were each sonicated, followed by centrifugation. The clear supernatants were saved as cell extracts. The pellets were treated with 0.8% N-lauroylsarcosine, followed by centrifugation. The supernatants were saved as the inner membrane fractions, and the pellets were saved as the outer membrane fractions. The recombinant mFadA and pre-FadA mixture purified with the cobalt column was included as positive control. mFadA was detected in the cell extracts of both E. coli and F. nucleatum but nonassociated with either the inner or outer membrane (Fig. 5, A and  B). Therefore, it was probably secreted or released from the bacteria upon sonication. Indeed, mFadA, but not pre-FadA, was detected in the culture supernatant of E. coli BL21(DE3)/ pYWH417-6, and the amount of mFadA secreted into media increased with the duration of IPTG induction (Fig. 5C). Pre-FadA was associated with the inner membrane and nonsecreted (Fig. 5,  A and B). No FadA was detected in the negative controls of E. coli BL21(DE3)/pET21(b) or F. nucleatum 12230-US1. Interestingly, although pre-FadA existed in lower quantities than mFadA, as indicated by Coomassie Blue stain, it reacted more strongly with the anti-FadA antibodies on Western blots (compare lanes 1 of Fig. 5, A and B).
In order to delineate the association between the two forms of FadA, the mFadA and pre-FadA mixture purified using the cobalt column (Fig. 3A) was subjected to Sephacryl S-300 size exclusion column chromatography, with a fractionation range of 10 -1500 kDa. The standard size positions were established using blue dextran (2000 kDa), bovine serum albumin (66 kDa), carbonic anhydrase (29 kDa), and cytochrome c (12.3 kDa), and the elution fractions were collected at 2 ml/tube. The blue dextran came out in the void volume, in fraction 23.   1, 3, and 5) or cell pellet (lanes 2, 4, and 6) prior to IPTG induction (lanes 1 and 2) or at 1 h (lanes 3  and 4) or 2 h (lanes 5 and 6) after IPTG induction, using mAb 5G11-3G8 as the primary antibodies. An aliquot of 12 l of supernatant concentrated by 20- fold (lanes 1, 3, and 5) or a total of 5 g of proteins (lanes 2, 4, and 6) was loaded onto each lane. The arrows indicate mFadA and pre-FadA, respectively.
When a total of 1.8 mg of the FadA mixture was applied to the column, FadA was eluted starting in fractions 19 -52, as determined in a chromatogram and by SDS-PAGE (Fig. 6, A and B). These results indicated that FadA formed oligomers with continuously varying sizes, from larger than 2000 kDa to smaller oligomers. Although mFadA was detected in all of fractions 19 -52, pre-FadA was only present in the early fractions, whose quantity in relation to mFadA decreased with the size of the complex. Therefore, pre-FadA was probably involved in formation of high M r FadA oligomers. A 30-kDa component often co-purified with FadA and was eluted in fraction 39 (Fig. 6B), nonassociated with the high molecular weight oligomers.
To assess their respective functions, pre-FadA and mFadA were separately purified (Fig. 7, A and B). To extract mFadA, IPTG-induced E. coli BL21(DE3)/pYWH417-6 was incubated with PBS at 60°C for 30 min. mFadA in the supernatant of the hot PBS extract was purified using the cobalt column. In order to eliminate the trace amount of pre-FadA in the preparation, high S cation exchange column chromatography was employed, utilizing the charge difference between pre-FadA and mFadA, since the signal peptide contained two positively charged lysine residues. mFadA came out in the washes, whereas pre-FadA was retained on the column (see below). To purify pre-FadA, the mFadA-and-pre-FadA mixture was first purified from BL21(DE3)/pYWH417-6 cell pellet using cobalt column. The eluate was then applied to the High S column to separate these two proteins. The column was washed extensively with buffer C until all mFadA was eluted. Pre-FadA was then eluted with buffer D. The eluted pre-FadA was applied to 13% preparative SDS-PAGE and was gel-purified. Silver staining of SDS-PAGE showed that pre-FadA and mFadA were each purified to a single species (Fig. 7, A and B). Purified mFadA was soluble in 50 mM phosphate buffer, pH 8.0. In contrast, purified pre-FadA was only soluble under acidic conditions (pH 2.6 -3.0). Under neutral conditions, it was insoluble and only became soluble when mixed with mFadA at a molar ratio of mFadA/pre-FadA Ͼ 3:1 (data not shown).
Purified mFadA was labeled with Alexa-fluor 488 and tested for binding to immortalized human oral epithelial OKF6/Tert cells. As a control, cytochrome c, whose M r (12,300) is close to that of FadA, was also labeled with Alexa-fluor 488 and tested in parallel. No binding by either cytochrome c or mFadA was detected by epifluorescent microscopy following overnight incubation (Fig. 8). When nonlabeled pre-FadA was mixed with Alexa-fluor 488-labeled mFadA at a molar ratio of 5:1 (mFadA/ pre-FadA), formation of high M r complexes was detected by gel filtration column chromatography (data not shown), and binding of mFadA was observed (Fig. 8, A and B). The fluorescent binding was inhibited by the nonlabeled pre-FadA-mFadA complex in 20-fold excess, but not by mFadA alone, strongly suggesting specific binding by the complex (Fig. 8, A and B). Binding by pre-FadA alone was not testable due to its insolubility under neutral pH. However, association of the nonlabeled pre-FadA with OKF6/Tert cells was detected by Western blotting analysis (Fig. 8C). Confocal microscopy revealed that mFadA was internalized in the OKF6/Tert cells (Fig. 9). It is possible that the nonlabeled pre-FadA was also internalized.
To assess the importance of FadA in mediating F. nucleatum binding to the host cells, competitive attachment assays were performed. Although attachment of F. nucleatum 12230 to OKF6/Tert cells was not inhibited by mFadA alone, it decreased Ͼ80% in the presence of 0.5 or 1 g/l of reconstituted pre-FadA-mFadA (1:5 molar ratio) complex (Fig. 10). Thus, the pre-FadA-mFadA complex played a significant role in the attachment of F. nucleatum 12230 to the host cells. The threshold for inhibition was between 0.25 and 0.5 g/l of the reconstituted complex. Since the precise size of the complex  was unclear, the molar concentration of the inhibitory threshold could not be determined.

DISCUSSION
To the best of our knowledge, this is the first functional expression of a fusobacterial adhesin in E. coli. The expression vector pET21(b) was chosen so that the His tag fusion was conjugated to the carboxyl terminus of FadA and would not be lost due to secretion. FadA was stably expressed in F. nucleatum. Its expression in E. coli also appeared to be stable, making the subsequent characterization feasible. The recombinant FadA enhanced the ability of E. coli to bind to human gingival epithelial cells, KB, CHO, and human umbilical vein endothelial cells, each by 3-4-fold. Thus, a similar mechanism may be involved in FadA binding to these different cells. The FadA receptor may exist on all of these different cell types. The 3-4fold increase was consistent with the decrease in binding of the fadA deletion mutant US1, when compared with the wild-type F. nucleatum.
Two forms of recombinant FadA were identified: the intact pre-FadA and the secreted mFadA. Surprisingly, pre-FadA migrated faster than mFadA on SDS-PAGE. The calculated M r of both components were consistent with those determined by mass spectrometry. Thus, the aberrant migration was not due to post-translational modification. One possible explanation is that the hydrophobic signal peptide bound more SDS, enabling pre-FadA to migrate faster than predicted in the gel. Alternatively, the shape of pre-FadA and mFadA may differ, resulting in aberrant migration. The mobility of mFadA and pre-FadA in relation to the protein size markers also varied when different SDS-PAGE or running buffers were used (compare lanes 1 in Fig. 5). Furthermore, it was noticed that using INDIA TM HisProbe-HRP, only one FadA species was detected. However, both species were detected  using FadA antibodies, with pre-FadA reacting more strongly than mFadA.
Identification of pre-FadA and mFadA by amino-terminal protein sequencing and mass spectrometry also confirmed that the first 18 aa of the intact protein indeed encoded a signal peptide. The Sec, type IV, and type V secretion pathways have recently been found to be conserved in the genomes of F. nucleatum ATCC 25586 and ATCC 49256 (21). Therefore, it is likely that FadA in fusobacteria was secreted in a similar fashion as in E. coli. The co-existence of pre-FadA and mFadA in E. coli was not an artifact of overexpression, because two similar forms were also identified in different strains of F. nucleatum, although their relative quantities and ratios varied. Furthermore, the location of these two components in E. coli and F. nucleatum were consistent, with pre-FadA associated with the inner membrane and mFadA easily dissociated from the bacteria by either sonication or hot PBS extraction. The location of these two forms makes sense, because pre-FadA is not secreted, whereas mFadA is. Secretion of mFadA into the culture medium by E. coli BL21(DE3)/pYWH417-6 increased with the duration of IPTG induction. The native pre-FadA and mFadA each migrated faster than their recombinant counterparts. This is probably due to the lack of the His tag in the native proteins. Based on the size, it is unlikely that post-translational modification occurred to the native FadA in fusobacteria.
When pre-FadA and mFadA were co-purified, they formed oligomers with continuously varying sizes. The ratio of pre-FadA to mFadA decreased with the size of the oligomer. Pre-FadA was absent in the smaller oligomers, suggesting that it was only required for the formation of high M r oligomers. Pre-FadA and mFadA were each purified to a single species, following extensive procedures. Binding to epithelial cells was not detected with mFadA alone. However, when the same amount of fluorescence-labeled mFadA was mixed with pre-FadA at a ratio of 5:1, binding of mFadA was observed. The binding was unevenly distributed in the monolayer, consistent with what was previously observed with F. nucleatum binding to the epithelial cells (8). This could be due to heterogeneous expression of the FadA receptor among the cells. The fluorescent binding was inhibited by the nonlabeled pre-FadA-mFadA complex but not by mFadA alone. These results suggest that the co-existence of pre-FadA and mFadA served not only a structural but also a functional purpose.
Although pre-FadA is nonexposed on the bacterial surface, it plays an important role in FadA function. The peptide was insoluble under physiological pH and only became soluble at acidic pH values or when mixed with mFadA. This is consistent with the observation that pre-FadA was associated with the inner membranes. The insolubility of pre-FadA was presumably due to the high hydrophobicity of the signal peptide. In the presence of mFadA, oligomerization of pre-FadA with mFadA may prevent pre-FadA from autoaggregation. It is possible that the signal peptide is inserted in the inner membrane, serving as an anchor for the complex. Based on the results, we propose the following model. Due to its ␣-helical nature and low M r of the monomers, FadA may form a filamentous structure similar to some other adhesins, such as pili. The mFadA monomers are added at the base of the filament as they are secreted. As the complex builds, it extends through the outer membrane. Extension of the filament ceases when the nonsecreted pre-FadA is added at the base, with its signal peptide inserted in the inner membrane. The different ratios of pre-FadA and mFadA observed in different F. nucleatum strains may indicate different lengths of the FadA filaments. Extension and retraction of filamentous bacterial appendages was first observed in type IV pili of Pseudomonas aureginosa (22). The difference is that a large number of genes have been identified to be involved in the type IV pili assembly and function (23). Our model does not rule out the possibility of additional components to be involved in FadA function. One such candidate is the 30-kDa component, which was often co-purified with FadA. The identity of the 30-kDa component, as well as the validity of the model, is currently under investigation. The hydrophobicity of pre-FadA may cause aggregation of the individual filaments, forming high M r bundles, which may be the active form of FadA. Results of the inhibitory/competitive attachment assays indicate that FadA plays a significant role in mediating F. nucleatum attachment to host cells. It is possible to develop antagonist(s) to reduce binding of F. nucleatum to host cells based on the FadA structure. It should be pointed out, however, that residual binding by F. nucleatum 12230 was observed in the presence of pre-FadA-mFadA complex, suggesting that additional adhesin(s) may also exist on F. nucleatum. The current study also revealed that although FadA was first identified as an adhesin (18), it appeared to be involved in invasion as well. Upon overnight incubation of the epithelial cells with nonlabeled pre-FadA and Alexa-fluor 488-labeled mFadA, fluorescence was detected inside the cells. This internalization was unlikely to be due to release of Alexa-fluor 488 from the complex, because it was inhibited by nonlabeled pre-FadA-mFadA complex. Detection of fluorescence throughout the thickness of a cell suggests that a significant amount of protein was internalized. Due to its insolubility, pre-FadA was not labeled. The association of pre-FadA with OKF6/Tert cells was readily detected by Western blotting analysis. Thus, it is reasonable to speculate that pre-FadA was internalized into the host cells as part of the complex. Internalization of the pre-FadA-mFadA complex, rather than mFadA alone, is an efficient "quality control" mechanism. Since the complex is anchored in the bacterial membranes, its internalization inevitably leads to the invasion of the bacteria. In contrast, mFadA can be dissociated from the bacteria; thus, internalization of mFadA may not facilitate the bacterial invasion. The mechanism of FadA-mediated invasion is being further tested. Our recent success in crystallization of mFadA will undoubtedly facilitate understanding of the FadA structure and function (24).
In summary, our results indicate E. coli as a suitable host for functional expression of F. nucleatum proteins. FadA is a uniquely "self-sufficient" adhesin and invasin in that it utilizes its own two forms for anchoring, functional oligomerization, and attachment and invasion of the host cells.