Serotype-Specific Sugars Impact Structure but Not Functions of the Trimeric Autotransporter Adhesin EmaA of Aggregatibacter actinomycetemcomitans

ABSTRACT

The human oral pathobiont Aggregatibacter actinomycetemcomitans expresses multiple virulence factors, including the trimeric, extracellular matrix protein adhesin A (EmaA). The posttranslational modification of EmaA is proposed to be dependent on the sugars and enzymes associated with O-polysaccharide (O-PS) synthesis of the lipopolysaccharide (LPS). This modification is important for the structure and function of this adhesin. To determine if the composition of the sugars alters structure and/or function, the prototypic 202-kDa protein was expressed in a non-serotype b, emaA mutant strain. The transformed strain displayed EmaA adhesins similar in appearance to the prototypic adhesin as observed by two-dimensional (2D) electron microscopy of whole-mount negatively stained bacterial preparations. Biochemical analysis indicated that the protein monomers were posttranslationally modified. 3D electron tomographic reconstruction and structure analyses of the functional domain revealed three well-defined subdomains (SI, SII, and SIII) with a linker region between SII and SIII. Structural changes were observed in all three subdomains and the linker region of the adhesins synthesized compared with the known structure. These changes, however, did not affect the ability of the strain to bind collagen or form biofilms. The data suggest that changes in the composition of the glycan moiety alter the 3D structure of the molecule without negatively affecting the function(s) associated with this adhesin.

IMPORTANCE The human oral pathogen A. actinomycetemcomitans is a causative agent of periodontal and several systemic diseases. EmaA is a trimeric autotransporter protein adhesin important for colonization by this pathobiont in vivo. This adhesin is modified with sugars associated with the O-polysaccharide (O-PS), and the modification is mediated using the enzymes involved in lipopolysaccharide (LPS) biosynthesis. The interaction with collagen is not mediated by the specific binding between the glycans and collagen but is attributed to changes in the final quaternary structure necessary to maintain an active adhesin. In this study, we have determined that the composition of the sugars utilized in the posttranslational modification of this adhesin is exchangeable without compromising functional activities.

INTRODUCTION

Bacterial adhesion is mediated by proteinaceous surface structures, including fimbrial and nonfimbrial adhesins. The extracellular matrix protein adhesin A (EmaA) expressed by the human oral pathogen Aggregatibacter actinomycetemcomitans is a nonfimbrial adhesin promoting colonization (1–3). EmaA is classified as a trimeric autotransporter protein (type V_c secretion system) and belongs to a family of proteins which is represented by YadA of Yersinia enterocolitica (1). Three monomers of EmaA form antenna-like projections, extending up to 150 nm from the bacterial outer membrane, as visualized by transmission electron microscopy (TEM) (4, 5).

The functional domain of the adhesin is associated with the ellipsoidal end of the projections, which corresponds to the first 760 amino acids of the protein monomer after cleavage of the 56-amino-acid signal peptide (5–8). The remainder of the sequence forms the stalk, including the membrane pore-forming domain (4, 5). The distal end of the structure is further subdivided into three subdomains (SI, containing amino acids 57 to 225; SII, containing amino acids 226 to 433; and SIII, containing amino acids 448 to 627) with a linker region between SII and SIII containing 14 amino acids (6, 8). Deletion of the whole functional domain or specific subdomains negatively affects the collagen binding activity of this adhesin (5, 8). The functional domain is also involved in cell-cell interactions in microcolony development during biofilm formation (3).

The emaA gene is ubiquitous in A. actinomycetemcomitans; however, heterogeneity exists in the protein sequence that correlates with the serotype of the strain (9). Serotypes b and c strains express the prototypic (202-kDa, b-EmaA) protein, whereas serotype a and d strains encode a 173-kDa (a-EmaA) homolog. The a-EmaA isomer shares 75% amino acid sequence similarity in the functional domain and 95% sequence similarity within the stalk and pore-forming sequence. The reduction in the size of the monomer is due to a 279-amino-acid deletion in the region between the head and stalk domains of the prototypic sequence (9). Both protein species form structures that bind to collagen and participate in biofilm formation (3, 9).

Biochemical and genetic studies support the hypothesis that the b-EmaA is glycosylated with O-polysaccharide (O-PS) sugars, mediated by the WaaL ligase of the lipopolysaccharide (LPS) pathway (10). This posttranslational modification is necessary for the collagen binding activity and protein stability but is not essential for biofilm formation (3, 10–12). The glycan is not directly involved in collagen binding but is required for the functional quaternary structure of the adhesin. However, the role of the specific glycan composition in the structural stability and function of the adhesin remains unknown.

To analyze the functional and structural role of the sugars modifying EmaA, we modified the glycan moiety by transforming a serotype a strain (O-PS composed of d-talose disaccharide repeats), which does not express the endogenous a-EmaA protein because it is a spontaneous mutant that does not express any EmaA protein, with a plasmid encoding the prototypic b-EmaA protein sequence under the control of the endogenous promoter. The transformed strain displayed antenna-like projections, similar to prototypic EmaA adhesins expressed in serotype b strains, as visualized by TEM, and the EmaA monomers displayed characteristics of a posttranslationally modified protein. Furthermore, electron tomography and subvolume averaging were utilized to determine the substructure of the functional region of this prototypic b-EmaA expressed in the serotype a strain. The overall structure was found to be similar to that of the prototypic b-EmaA, but clear differences in the density of the subdomains and the linker region were observed. Despite the difference in EmaA structures, the transformed strain was active in binding to collagen and forming biofilms. The data suggest that the compositional change in associated glycans does not influence the known functions of this adhesin.

RESULTS

TEM images of the A. actinomycetemcomitans serotype a strain transformed with prototypic b-EmaA.

The serotype a strain transformed by electroporation with the plasmid expressing the full-length b-emaA gene (pKM11) under the control of the endogenous promoter sequence was examined by two-dimensional electron microscopy (2DEM) of whole-mount negatively stained bacterial preparations. Antenna-like projections were observed associated with the rugose membrane of the transformed cells. These surface projections extended well away from the bacterial surface and contained bends similar to those observed in the serotype b wild-type strain expressing the chromosomal copy of the gene. The increased number of projections associated with the transformed strain ATCC 29523/pKM11 (Fig. 1A) is due to the overexpression of the gene associated with the plasmid. The overall structure is similar to the prototypic EmaA expressed in serotype b (Fig. 1B).

FIG 1.

Transmission electron micrographs of b-EmaA expressed in different O-PS backgrounds. (A) The EmaA structures (white arrows) were expressed in a serotype a emaA mutant strain (ATCC 29523) using the plasmid pKM11. (B) Prototypic b-EmaA (black arrows) expressed in the serotype b strain VT1169. Bar = 100 nm.

Biochemical characterization of EmaA.

The electrophoretic mobility of the EmaA monomers isolated from the membrane of the transformed serotype a strain was similar to the mobility of the monomers isolated from the serotype b wild-type strain (Fig. 2A). The monomers associated with both strains had a reduced mobility compared with the monomers derived from the membrane of a serotype b strain deficient in O-PS synthesis (rmlC) (Fig. 2A). The immunoreactive material at the top of the wells represents aggregates of EmaA that do not enter the running gel, which is also consistent with the biochemical feature of this adhesin (10, 13). No immunoreactive material was observed in the membrane fraction of the serotype a bacteria transformed with the empty plasmid. In addition, a comparison of the EmaA monomers isolated from the cytosol with those associated with the membrane demonstrated a difference in the apparent electrophoretic mobilities (Fig. 2B). This difference is indicative of a change in the mass of the protein. Taken together, the data suggest that the monomers isolated from the serotype a strain transformed with b-emaA exhibit characteristics resembling those of the wild-type monomers isolated from serotype b strains.

FIG 2.

Biochemical characterization of EmaA expressed in different O-PS backgrounds. (A) Immunoblot of membrane EmaA monomers in different O-PS backgrounds. Wild type, VT1169 (serotype b); rmlC, isogenic rmlC mutant of VT1169; transformed EmaA, serotype a emaA mutant strain ATCC 29523 transformed with a plasmid expressing b-emaA (pKM11); emaA, ATCC 29523 transformed with the empty plasmid (pKM2). (B) Immunoblot of cytosol- and membrane-localized EmaA of the transformed emaA strain (ATCC 29523/pKM11). Solid arrows indicate electrophoretic mobility of the membrane, and dashed arrows indicate electrophoretic mobility of monomers isolated from an O-PS mutant (rmlC) (A) or from the cytosol (B).

3D structure of the EmaA functional domain.

The structure of the EmaA functional domain was determined by electron tomography and subvolume averaging techniques of whole-mount negatively stained bacterial preparations of the transformed serotype a strain. Eighty-four tomographic single-axis tilt series were collected with an angular range of at least ±64° in 2° increments; the defocus was selected so that the first zero of the contrast transfer function (CTF) would lie at a resolution better than 15 Å. A total of 151 EmaA functional domains were selected from the tomograms (Fig. 3A) and reconstructed from the corresponding subprojections using Radon transform algorithms (Fig. 3B) (14). After an initial interactive alignment, using the prototypic EmaA as a reference, 15% of the subvolumes were discarded due to the close proximity of either fiducials or proteinaceous contamination. The data were subjected to a first round of classification using principal-component analysis (PPCA-EM subroutine implemented in EMIRA) (15, 16) followed by clustering to aid subsequent alignment steps within this data set. A representative for each class was used as a new reference in an interactive multireference alignment step. Since the data exhibited conformational heterogeneity, a final round of classification was performed on the remaining 128 EmaA subvolumes, and eight classes were obtained with different numbers of subvolumes in each class (Fig. 4). The majority of the class subvolume averages, with the exception of classes I and VI, show a resolution better than 15 Å (see Fig. S1 in the supplemental material).

FIG 3.

Functional N terminus of b-EmaA expressed in the serotype a transformed strain versus the wild-type EmaA expressed in a serotype b strain. (A) The antenna-like EmaA projections are visualized extending from the bacterial surface. The functional N terminus comprises approximately the 30-nm distal end region of the EmaA appendage (white square). Bar = 5 nm. (B) Structures of the functional N terminus: the average wild-type EmaA (pink mesh) versus one individual transformed EmaA (blue surface). The functional terminus includes three subdomains, SI (amino acids 57 to 225), SII (226 to 433), and SIII (448 to 627), and a more flexible linker region between SII and SIII (6, 8).

FIG 4.

Surface representation of subvolume averages of the functional N terminus of the b-EmaA expressed in the transformed serotype a strain. (A) Average EmaA subvolume of each class, in a total of eight classes based on 128 EmaA subvolumes extracted from 84 reconstructed tomograms, is demonstrated in blue surface versus the wild-type EmaA (from the serotype b strain) in pink mesh. Bar = 3 nm. (B) Distribution of eight different classes in b-EmaA expressed in the transformed serotype a strain.

The overall structure of the EmaA subvolumes (from the serotype a strain transformed with the serotype b-emaA gene) calculated from the combined subprojection sets displayed a subdomain architecture similar to that of the prototypic b-EmaA average (Fig. 4A; also, see Movie S1 in the supplemental material). However, in general, subdomain SII in all classes was significantly smaller and smoother than in the prototypic EmaA; this effect was more visible in classes I and II. Moreover, 73% of the EmaA population had an enlarged subdomain SI. This appears most evident in classes V and VI, which have concomitantly lower density in subdomains SII and SIII. An additional feature present in a few of these classes (i.e. III, IV, VII, and VIII) is a bend situated between the linker region and subdomain SIII, which rarely appears in the prototypic EmaA (5). An analysis of direct class averages and variances calculated from the reconstituted volumes after the last round of PPCA-EM revealed that the largest variance is located at the core of the structure within subdomains SI and SII (data not shown). Only class II and to a minor extent classes V and VIII show variances on the external surface of the adhesin, as would be expected.

Volume segmentation was carried out to characterize the continuity of the density along the functional domain of the EmaA adhesins (17–19). Segmentation of the EmaA class average subvolumes in Chimera using three smoothing steps resulted in subdivisions of the SI-SIII region that contained two segments as for the prototypic EmaA either with or without 3-fold symmetrization (Fig. 5; also, see Movie S2 in the supplemental material). However, the results of the segmentation differed in the specific start and end of the segments. The first prototypic EmaA segment contained SI, SII and a third of the linker region (Fig. 5, WT), while the b-EmaA expressed in the serotype a strain contained SI, SII, and either half of the linker region for classes II, III, and VIII or the complete linker region for the rest of the classes. Taken together, these data indicate that there are differences between the prototypic b-EmaA and the EmaA structures when they are expressed in different O-PS backgrounds.

FIG 5.

Segmentation of subvolume averages of the functional N terminus of b-EmaA expressed in the transformed serotype a strain. Segmentation of the subvolume average of each class of the b-EmaA, expressed in the transformed serotype a strain. Segmentation was carried out using three smoothing steps and resulted in subdivisions of the functional region of EmaA between SI and SIII. WT, wild-type EmaA (serotype b); I to VIII, classes labeled as in Fig. 4; SI, SII, and SIII, three well-defined subdomains of the structure; L, linker region between SII and SIII; N, neck sequence; S, stalk domain (6, 8). Bar = 3 nm.

Functional activities of EmaA expressed in a noncognate serotype strain.

Changes in the structure of the functional domain may correlate with an alteration in the functions of the adhesin. The transformed serotype a strain expressing b-emaA was observed to form a more robust biofilm than the strain transformed with the empty plasmid (Fig. 6). The biofilm formed was 5- and 10-fold greater in mass at 48 and 72 h, respectively, suggesting that changes in structures did not impact the ability of the strain expressing EmaA to mediate biofilm formation. Collagen binding activity, however, is dependent on the posttranslational modification. In collagen binding assays, a 3-fold increase in collagen binding activity of the transformed strain was observed compared with the parent mutant strain (Fig. 6). These data demonstrate that b-EmaA modified with d-talose retains collagen binding activity.

FIG 6.

Functional analyses of b-EmaA expressed in the transformed serotype a strain. (A) Collagen binding activity. **, P < 0.01. (B) Biofilm formation. emaA, serotype a emaA mutant strain transformed with an empty plasmid; transformed emaA, serotype a strain transformed with a plasmid expressing b-emaA. *, P < 0.05.

DISCUSSION

The multifunctional collagen adhesin EmaA is a recognized virulence factor contributing to the early colonization of the oral pathogen A. actinomycetemcomitans (1–3). EmaA is proposed to be glycosylated using a novel mechanism via the O-PS biosynthetic pathway (10). The posttranslational modification is required for collagen binding activity (13), and there is mounting evidence that the sugars are required for the formation of the quaternary structure of the adhesin (11, 12). To investigate the role of specific sugars in the folding of the protein, we replaced the sugars of serotype b (branched l-Rha, d-Fuc, and d-GalNAc) with those expressed in serotype a strains (disaccharide units of d-talose), which do not express EmaA on the surface due to the presence of an amber mutation in the 3′ end of the gene (9).

The 2D electron microscopy images of the serotype a emaA mutant strain transformed with a replicating plasmid containing the endogenous promoter with the intact b-emaA gene sequence validated the competency of the strain to express and form EmaA surface structures. These structures resembled those of the prototypic adhesin forming antenna-like appendages and contained multiple bends along the length of the structures, which are important for the flexibility of the adhesin (4–6, 20).

The electrophoretic mobility of the b-EmaA monomers expressed in the serotype a strain is similar to that of the prototypic b-EmaA expressed in the cognate serotype b background. The molecular mass of the protein expressed in the serotype a strain is also greater than that of the prototypic monomer expressed in a glycosylation-deficient mutant strain (Fig. 2) (10, 13). The glycosylation state of the protein is reinforced by the difference in the electrophoretic mobility of the membrane-associated form and the monomer found in the cytoplasm (Fig. 2B), and it is consistent with the protein monomers isolated from these two fractions associated with the prototypic EmaA expressed in the serotype b background (13). Direct evidence for the glycosylation state of the protein is lacking due to difficulties in purifying membrane bound EmaA adhesins. This glycosylated trimeric adhesin tends to form aggregates trapped in the stacking SDS-PAGE gel, with only small portions of monomers running into the separating gel, which was demonstrated in the prototypic EmaA expressed by the wild-type serotype b strain (10, 13), as well as here with the b-EmaA expressed in serotype a strains (Fig. 2). Additional support stems from the structural data. Adhesins expressed on O-PS mutant strains seem to “hug” the cell surface, which might ensue from modifications of the electrostatic properties of both surfaces, and exhibit a strong curvature along the whole length of the functional domain (13). The adhesins in this study extend well away from the bacterial surface (Fig. 1A) and have relatively straight functional domains (Fig. 3 to 5). Thus, the biochemical and structural data presented in this study suggest that the prototypic protein expressed in this transformed serotype a strain is glycosylated.

The serotype a modified proteins also displayed the activities associated with functional EmaA adhesins. The transformed strain demonstrated greater collagen binding and biofilm formation compared with the parent strain, which is a spontaneous emaA mutant. Collagen binding is dependent on the glycosylation state of the protein based on the decrease in binding activity of strains mutant for OPS metabolic, transport, or ligase enzymes (10, 13). Since binding is dependent on the posttranslational modification, the increase in activity lends additional support to the nature of the modification and the role of this modification in the formation of an active adhesin. The adhesin is also important in biofilm biogenesis (3), and the functionality of the adhesin is further supported by the observation of increased biofilm biogenesis of the transformed strain (Fig. 6B). Biofilm formation is independent of the glycosylation state and the specific 3D structure of the adhesin (3, 6, 8); thus, expression of the b-EmaA modified with different sugars does not affect this activity of the adhesin.

The collagen binding and biofilm formation abilities of the transformed strain in this study (Fig. 6) imply that it is glycosylation itself that is relevant for the proper function of EmaA, rather than the specific sugars bound to the protein. In Escherichia coli, O-PS is synthesized as undecaprenyl-diphosphate (Und-PP)-linked O-antigen, transferred from the cytoplasm to the periplasmic space, and covalently ligated to the lipid A-core oligosaccharide using the WaaL ligase (21). Since EmaA utilizes the WaaL ligase for glycosylation, the linkages between glycans and the protein involve at least O-linked glycosylation. The potential sites for O-linked glycosylation, Ser and Thr residues, are widely distributed across the whole protein. Glycosylation is required for EmaA binding to collagen and stability of the protein (11–13). A comparison between glycosylation-deficient and prototypic b-EmaA structures (6, 8, 11, 12, 17) reveals structural changes within the functional domain, including subdomains I, II, and III and the linker region, as demonstrated in the serotype a modified protein (Fig. 4 and 5). Therefore, taken together, these data indicate that the functional domain of EmaA contains glycosylation sites (Fig. 7).

FIG 7.

Predicted O-linked glycosylation sites of the EmaA functional domain. Sequence of the N-terminal 627 amino acids of EmaA: subdomain I (amino acids 57 to 225), subdomain II (226 to 433), and subdomain III (448 to 627). Pentameric repeats are in bold and underlined. Two neck sequences are in blue. The possible O-linked glycosylation sites (in red) are S (Ser) and T (Thr).

The structural integrity of the EmaA adhesin has been shown to be critical for optimal function (6, 8, 11, 12). Deletion mutants of the protein sequence located in the functional domain (subdomains SI to SIII) decrease or abolish the collagen binding ability of the adhesin. Even the mutation of a single amino acid (G162S) located at the core of the structure resulted in a negative functional outcome and concomitantly in a structural change (6, 8). The bend in the adhesin between subdomains SII and SIII observed in the transformed strain in this study was reminiscent of the bend displayed by the adhesin bound to collagen (8, 20). Mutants defective in the O-PS biosynthetic pathway (rmlC and waaL mutant strains) exhibited overall a reduced density of the functional domain. Specifically, subdomain SI had a diameter comparable to that of subdomain SII (~4.4 nm) instead of the prototypical SI diameter of 5 nm. In this study, we observed an enlarged subdomain SI concomitantly with weaker densities in subdomains SII and SIII. The larger volume occupied by subdomain SI might be the result of either an increased density or a more loosely packed structure resulting from the different sugars. In addition, the functional domain of O-PS mutants presented a strong curvature and multiple bends not commonly observed in the prototypical EmaA (11, 12). None of those bends were observed in the EmaA structures in the present study (Fig. 4). The data from this study indicate that the correlation between structure and function of this adhesin is less straightforward than originally suggested and that a certain structural permissiveness is allowed for proper function.

The nature of the sugars posttranslationally modifying EmaA affects the continuity of the density along the functional domain of the adhesins. Classification and segmentation of subvolumes calculated from 3D reconstructions of the prototypic EmaA show that SI and SII segment together independently of subdomain SIII (Fig. 5, WT). Additionally, the top SI/SII segment of the prototypic EmaA contains a third of the linker region, while the bottom SIII segment contains the remaining two thirds of the linker region. In most of the classes from the modified EmaA in this study (all except class VIII), we observed that the whole or full linker region is part of the top SI/SII segment, with the bottom segment containing mainly SIII. This indicates that the different sugars modifying the EmaA protein play a role in the interconnectivity and intraconnectivity of the subdomains, which are relevant for the functional activity of the adhesin.

In summary, we have determined that changes in the composition of the glycan moiety associated with the protein do not impact the function of this multifunctional adhesin. However, the data suggest that the overall 3D structure of the adhesin is dependent on the composition of the glycan moiety, without concomitantly negatively affecting the function(s) associated with EmaA.

MATERIALS AND METHODS

Strains and plasmids.

All bacterial strains used in this study are listed in Table 1. A. actinomycetemcomitans strains were grown statically in 3% Trypticase soy broth, 0.6% yeast extract (TSBYE) with or without 1.5% agar (Becton, Dickinson and Company), and the required antibiotics in a humidified 37°C incubator with 5% carbon dioxide. All mutant strains in this study retained growth characteristics similar to those of the wild-type strain. Escherichia coli was grown in broth containing 1% Bacto tryptone, 0.5% yeast extract, and 1% sodium chloride (lysogenic broth [LB]) with appropriate antibiotics at 37°C under aerobic conditions with agitation. All sequencing was performed at the University of Vermont Cancer Center DNA Analysis Facility. The shuttle plasmid pKM2 was used for transformation and expression of EmaA (22).

TABLE 1.

Strains and plasmids

Strains	Genotype or descriptiona	Reference or source
A. actinomycetemcomitans
VT1169	Nonfimbriated, laboratory strain, serotype b, expressing endogenous b-EmaA	1
ATCC 29523	Nonfimbriated, laboratory strain, serotype a; spontaneous a-EmaA mutant (a-emaA)	9
ATCC 29523/pKM2	ATCC 29523 transformed with empty plasmid pKM2; Cmr	This study
ATCC 29523/pKM11	ATCC 29523 transformed with plasmid pKM11; Cmr	This study
rmlC (rmlC::pLOF/Sp)	Transposon pLOF/Sp inserted into the rmlC (TDP-4-keto-6-deoxy-d-glucose 3,5-epimerase) gene of VT1169	10
KM708	Nonfimbriated variant of strain IDH1062 expressing endogenous a-EmaA	9
E. coli DH10B	F− mcrA Δ(mrr-hsdRMS-mcrBC) ϕ80lacZΔM15 ΔlacX74 recA1 endA1 araD139 Δ(are, leu)7697 galU galK λ− rpsL nupG tonA	Invitrogen, Carlsbad, CA
Plasmids
pLOF/Sp	Tn10-based transposon vector; Apr Spr	1
pKM2	pPK1 containing chloramphenicol acetyltransferase; Cmr	22
pKM11	pKM2 containing ~500-bp upstream sequence of emaA (the putative promoter region) plus full-length b-emaA sequence from VT1169; Cmr	10

^aCm, chloramphenicol; Sp, spectinomycin; Ap, ampicillin.

Fractionation of membrane and cytosol proteins.

Membrane fragments of A. actinomycetemcomitans were prepared as described previously (1, 9, 10, 13). Briefly, 200 mL late-logarithmic-phase cells were harvested, washed with phosphate-buffered saline (PBS; 10 mM sodium phosphate, 150 mM sodium chloride [pH 7.4]), and resuspended in 2.0 mL of 10 mM HEPES (pH 7.4) with 1 mM phenylmethylsulfonyl fluoride (PMSF; USB Corporation, Cleveland, OH) and Pierce protease inhibitors with EDTA (Thermo Fisher Scientific, Rockford, IL). Bacteria were lysed in a French pressure minicell by three cycles of 9,000 lb/in² (62,100 kPa) at 4°C. Whole-cell lysates were centrifuged at 10,000 × g for 30 min to remove cell debris, followed by ultracentrifugation at 100,000 × g for 1 h to separate the membrane and cytosol fractions. The membrane pellet was washed with HEPES and centrifuged, and the resulting pellet was suspended in HEPES. The cytosolic fraction was recentrifuged to removed contaminated membrane proteins. Protein concentrations were measured at 280 nm using an Evolution 201 UV-visible spectrophotometer (Thermo Fisher Scientific, Waltham, MA).

Biochemical analysis of EmaA.

Equivalent amounts of protein were resuspended in buffer to a final concentration of 5% beta-mercaptoethanol, 0.06 M Tris (pH 6.8), 10% glycerol, 0.02% bromophenol blue, and 2% SDS, boiled for 5 min, and applied to wells of 4 to 15% polyacrylamide Tris-glycine minigels (Bio-Rad, Hercules, CA) for electrophoresis (9, 10, 13). The separated proteins were transferred to 0.2 μm Immobilon-PSQ polyvinylidene difluoride (PVDF) membranes (MilliporeSigma, Burlington, MA) and probed with an anti-EmaA stalk monoclonal antibody (9). The immune complex was detected using horseradish peroxidase (HRP)-conjugated goat anti-mouse IgG (Jackson Laboratory, Bar Harbor, ME) and visualized using the SuperSignal West Pico plus chemiluminescent substrate (Thermo Fisher Scientific, Waltham, MA).

TEM.

Bacteria were recovered from storage at −80°C and grown on TSBYE plates for 2 to 3 days. A single colony of each strain was inoculated into 10 mL of TSBYE broth and grown for 16 h; the cultures were diluted 1:10 and grown for an additional 150 min. Bacteria were collected by centrifugation of 1 mL bacterial suspension at 1,000 × g for 1 min at 4°C and resuspended in 100 μL PBS (pH 7.4). Electron microscopy grids were prepared as previously described (4–6, 8). Briefly, a 5-μL aliquot of bacterial suspension was placed on either 300 or 200 mesh carbon-coated grids and deeply stained with NanoW (Nanoprobes, Yaphank, NY). For 3D electron tomography, the grids were pretreated with poly-l-lysine (1,000 to 5,000 Da; Sigma, St. Louis, MO) and colloidal gold (SPI, West Chester, PA) to be used as fiducial markers.

TEM data acquisition.

Data were collected using a Tecnai12 electron microscope (FEI, Hillsboro, OR) equipped with a LaB6 cathode (Kimball Physics, Wilton, NH), operated in point mode (4, 23), a 2,048- by 2,048-pixel charge-coupled device (CCD) camera with a pixel size of 14 μm (TVIPS, Gauting, Germany), and a dual-axis tilt tomography holder (Fischione, Export, PA). All images were recorded on the CCD camera at an acceleration voltage of 100 kV and a nominal magnification of ×42,000, which corresponds to a 0.308-nm pixel size on the specimen scale. Tomographic tilt series were acquired at least within a ±64° angular range in 2° angular intervals. Data were collected under low-dose exposure conditions (10 e⁻/Ų for 2D imaging and 3 e⁻/Ų per image for the tomographic tilt series data) as previously described (8, 20).

3D reconstruction of the EmaA functional domain.

Tomographic tilt series were processed using IMOD (24). Projections were initially aligned by cross correlation and further refined using fiducial markers located outside the bacterial surface. EmaA functional domains were orientationally selected from the tomograms by marking two points along their long axis (11, 12, 17, 25), with the first point located at the tip of the adhesin. For each selected molecule, a tilt series of subprojections was extracted, subprojection angles were recalculated, and subvolumes were reconstructed with the molecules’ long axis oriented approximately parallel to the y axis using algorithms implemented in EMIRA (15). The 3D reconstruction algorithms based on Radon transforms, as implemented in EMIRA (15), contain an occupancy index that keeps track of the number of 2D transforms averaged into it and is used to determine the location of the missing data (26). Subvolumes were visualized in Chimera (27) and further aligned to a reference subvolume of the wild-type EmaA (6) using the “fit in map” command. Angles and shifts were applied to the subprojections, and aligned subvolumes were reconstructed. The aligned subvolumes were grouped using probabilistic principal-component analysis with expectation maximization (PPCA-EM), an algorithm to analyze 3D volumes with missing data, followed by clustering using Diday’s method of moving centers as implemented in EMIRA (15, 17, 26, 28, 29). Outliers (with a sigma value of 5) were identified and excluded from this stage of the analysis. A representative of each group was selected, and all representatives were aligned to each other. The members of each group were subsequently realigned to the aligned representative subvolume using the “fit in map” command in Chimera (27). Angles and shifts were applied to the subprojections, and newly aligned subvolumes were reconstructed. The newly aligned subvolumes were grouped anew using PPCA-EM followed by nonlinear mapping and Diday’s method of moving centers for clustering (15, 26, 28, 30). Final 3D structures were calculated by de novo reconstruction from 2D Radon transforms of the combined subprojections of all subvolumes in each group. In addition, to assess variability within groups, average subvolumes and variances were calculated from the reconstituted subvolumes after processing with PPCA-EM, which results in reconstituted subvolumes with the missing data filled in (15, 26). The resolution for the average subvolume of each class was calculated using Fourier shell correlation methods (31). The subprojections of each subvolume belonging to a class were divided into two groups (odd and even) and combined to calculate two subvolumes per class. The Fourier shell correlation was calculated using Spider (32), and the 0.143-nm resolution criterion was used to determine resolution (33). Averaged subvolumes and variances were low pass filtered and visualized in Chimera (27). Segmentations were performed in Chimera using the Segger watershed algorithm, applying initially a 2-step smoothing and finally a 3-step smoothing (34).

Collagen binding assay.

A polyclonal antibody was developed against a serotype a strain of A. actinomycetemcomitans as described before (3). The collagen binding activity of different strains was evaluated using 96-well enzyme-linked immunosorbent assay (ELISA) plates, as described before (35). Briefly, type V collagen from human placenta (Sigma type IX, C3657; Sigma-Aldrich, St. Louis, MO) was presolubilized in aqueous acid and diluted in carbonate coating buffer, pH 9.6. A 100-μL aliquot at 10 μg/mL was loaded per well and stabilized overnight at 4°C. The wells were rinsed with PBS (pH 7.4) and blocked in PBS complemented with 0.5% bovine serum albumin. Approximately 1.0E7 bacteria per well were incubated for 1 h, washed with PBS, and incubated with a suspension of 4.6 μg/mL anti-serotype a polyclonal antibody. After 1 h of incubation, plates were washed with PBS containing 0.05% Tween 20 and incubated for 1 h with the secondary antibody, HRP-conjugated goat anti-rabbit immunoglobulin (Jackson Laboratory, Bar Harbor, ME). Immunoglobulin complexes were detected using citrate-phosphate buffer pH 5.0 containing 0.04% O-phenylenediamine and 0.012% hydrogen peroxide. The reaction was stopped by addition of 50 μL of 4 M H₂SO₄, and absorbance was measured at 490 nm. Data were analyzed using paired t test with GraphPad Prism 7.0a (GraphPad, San Diego, CA), and P values of <0.05 were considered significant.

Biofilm formation assay.

Biofilm formation was evaluated using 96-well cell culture plates with Nunclon Delta cell culture treatment (Thermo Fisher Scientific) as described before (3). Briefly, bacteria were recovered from storage at −80°C and grown on TSBYE plates for 3 days. A single colony of each strain was inoculated into 10 mL of TSBYE broth and grown overnight; subsequently, the cultures were diluted 1:10 and grown for one doubling time (~150 min). Each bacterial strain was loaded with ~1.0E6 bacteria in 200 μL/per well and grown for either 24 h, 48 h, or 72 h. The top 100 μL culture medium was gently replaced with fresh medium every 24 h. Wells were washed with PBS (pH 7.4) three times, stained with 0.35% crystal violet in 19% ethanol and 1% methanol, washed to remove extracellular dye, and destained using 95% ethanol. Semiquantification was performed using a spectrophotometer (absorbance at 562 nm). Proper dilution of the crystal violet was required to avoid instrument saturation. Data were analyzed using a paired t test with GraphPad Prism 9.3.1 (GraphPad, San Diego, CA), and P values of <0.05 were considered significant.

Data availability.

The 3DEM maps generated in this study have been deposited in the Electron Microscopy Data Bank (EMDB; https://www.ebi.ac.uk/emdb/) under accession codes EMD-28839 (class CI), EMD-28838 (class CII), EMD-28834 (class CIII), EMD-28840 (class CIV), EMD-28841 (class CV), EMD-28842 (class CVI), EMD-28843 (class CVII), and EMD-28844 (class CVIII).