ABSTRACT
Treponema denticola, one of several recognized periodontal pathogens, is a model organism for studying Treponema physiology and host-microbe interactions. Its major surface protein Msp (or MOSP) comprises an oligomeric outer membrane-associated complex that binds fibronectin, has cytotoxic pore-forming activity, and disrupts several intracellular responses. There are two hypotheses regarding native Msp structure and membrane topology. One hypothesis predicts that the entire Msp protein forms a β-barrel structure similar to that of well-studied outer membrane porins of Gram-negative bacteria. The second hypothesis predicts a bipartite Msp with distinct and separate periplasmic N-terminal and porin-like β-barrel C-terminal domains. The bipartite model, based on bioinformatic analysis of the orthologous Treponema pallidum Tpr proteins, is supported largely by studies of recombinant TprC and Msp polypeptides. The present study reports immunological studies in both T. denticola and Escherichia coli backgrounds to identify a prominent Msp surface epitope (residues 229 to 251 in ATCC 35405) in a domain that differs between strains with otherwise highly conserved Msps. These results were then used to evaluate a series of in silico structural models of representative T. denticola Msps. The data presented here are consistent with a model of Msp as a large-diameter β-barrel porin. This work adds to the knowledge regarding the diverse Msp-like proteins in oral treponemes and may contribute to an understanding of the evolutionary and potential functional relationships between Msps of oral Treponema and the orthologous group of Tpr proteins of T. pallidum.
IMPORTANCE Treponema denticola is among a small subset of the oral microbiota contributing to severe periodontal disease. Due to its relative genetic tractability, T. denticola is a model organism for studying Treponema physiology and host-microbe interactions. T. denticola Msp is a highly expressed outer membrane-associated oligomeric protein that binds fibronectin, has cytotoxic pore-forming activity, and disrupts intracellular regulatory pathways. It shares homology with the orthologous group of T. pallidum Tpr proteins, one of which is implicated in T. pallidum in vivo antigenic variation. The outer membrane topologies of both Msp and the Tpr family proteins are unresolved, with conflicting reports on protein domain localization and function. In this study, we combined empirical immunological data derived both from diverse T. denticola strains and from recombinant Msp expression in E. coli with in silico predictive structural modeling of T. denticola Msp membrane topology, to move toward resolution of this important issue in Treponema biology.
INTRODUCTION
Oral spirochetes, most notably Treponema denticola, are associated with the most severe forms of periodontal diseases (1). Their numbers are highly elevated in the deepest recesses of active periodontal lesions, and they persist in cases that are refractory to standard treatment regimens (2). T. denticola Msp is a highly expressed, outer membrane-associated, oligomeric protein that binds fibronectin (3, 4), has cytotoxic pore-forming activity in epithelial cells (5, 6), disrupts intracellular cytoskeletal and calcium responses in fibroblasts (reviewed in reference 2), and inhibits neutrophil chemotaxis (7). While Msp pore-forming activity has been proposed to be responsible for its cytotoxicity, the molecular mechanisms responsible for other effects of Msp, particularly the intracellular responses, remain to be determined. Edwards et al. (4) reported that recombinant polypeptides consisting of Msp residues 14 to 202 or 203 to 259 bound immobilized fibronectin, keratin, laminin, collagen type I, fibrinogen, hyaluronic acid, and heparin, while the C-terminal recombinant polypeptide (residues 272 to 543) had no binding activity. In contrast, Jones et al. (7) reported that a recombinant polypeptide encompassing Msp residues 272 to 406 inhibited neutrophil chemotactic responses. The locations of specific active Msp domains, relative to Msp structure, and the mechanism of Msp effects on neutrophils are as yet unknown. Determining the topology of Msp in the outer membrane is crucial for understanding the role of Msp in Treponema biology, as well as its cellular effects in the host environment.
Msp or Msp-like proteins have been characterized in Treponema spp. (8–10), representing 3 of the 7 phylogroups (composed of over 60 distinct phylotypes) of human oral treponemes (11, 12). A recent study of 626 clinical samples identified 21 distinct msp genotypes within at least 5 Treponema species in phylogroups 1 and 2, with genotypes corresponding to T. denticola being the most frequently detected (13). Putative msp genes have also been identified in Treponema spp. associated with digital dermatitis and related diseases of domesticated and wild ungulates, most of which fall within phylogroup 1 (representative, Treponema vincentii), phylogroup 2 (representative, T. denticola), or Treponema phagedenis (14–18). T. denticola Msps are the most well studied and can be divided into 3 groups, 2 of which are very closely related (represented by strains ATCC 35405 and ATCC 33520, which are >90% identical and differ only in a 70-residue central domain) and 1 of which is represented by strain OTK, which has only about 40% total homology with the other 2 (13, 19).
A group of orthologous Treponema pallidum rare outer membrane proteins (TprA to TprL) shares evolutionary history with the Msps, although sequence homology with the Msps is rather limited (20). Functional characterization of the Tprs suggests involvement in immunogenicity (21) and outer membrane permeability (22) (as with the Msps) and in relatively rapid generation of intrastrain antigenic variants at the TprK locus in vivo (23, 24) (unlike Msps). In addition to this genetic mechanism for generating a wide range of sequence variability in a single gene, the other 11 tpr gene sequences show a wide range of diversity while still having conserved domain architectures (25). The role of Tpr sequence variation and diversity in T. pallidum pathogenesis is an area of great interest. To date, there is no evidence that any Treponema species other than T. pallidum carries more than 1 gene encoding an Msp-like protein.
The outer membrane topologies of both T. denticola Msp and the T. pallidum Tpr family remain unresolved, with little consensus regarding either the overall structure or the organization and localization of polypeptide domains within the protein. Several recent studies have reported predictive molecular models for T. pallidum Tprs and T. denticola Msps, composed of structurally and functionally distinct domains, i.e., an N-terminal periplasmic domain and a C-terminal porin domain with a β-barrel structure (22, 26, 27). These models are largely supported by expression studies of recombinant T. pallidum TprC/D and T. denticola ATCC 35405 Msp expressed in Escherichia coli, as well as by cell compartment fractionation studies in T. denticola. In contrast, other studies of Msp reported data suggesting that the entire Msp molecule is composed of a large β-barrel structure (28) with a central surface-exposed domain (4) that is divergent between strains (13, 19). Similarly, sequence analysis and modeling of multiple tpr loci from the various subspecies of T. pallidum revealed colocalization of discrete variable regions with predicted surface-exposed loops consistent with a typical β-barrel porin structure (25). In this study, we combined empirical immunological data derived both from T. denticola strains and from recombinant Msp expression in E. coli with in silico predictive structural modeling of T. denticola Msp membrane topology, in order to advance toward resolution of this important issue in Treponema biology.
RESULTS
The Msp central domain contains a surface-exposed epitope.In all T. denticola strains studied to date, a single msp gene encodes a secreted protein, with a calculated molecular mass of 56 to 60 kDa, that forms a trimeric, outer membrane-associated complex. As illustrated in Fig. 1, the Msp amino acid sequences of strains ATCC 35405 and ATCC 33520 differ only at residues 202 to 271 (ATCC 35405 numbering). This domain encompasses nearly the same polypeptide as the recombinant “rV-Msp” central domain of ATCC 35405 (residues 203 to 259), which was reported to have binding activity for several extracellular matrix-associated substrates (4). We performed immunofluorescence microscopy with intact and detergent-permeabilized cells of these 2 T. denticola strains with polyclonal antibodies raised against native ATCC 35405 Msp (Fig. 2), using antibodies specific for the periplasmic FlaA protein (29) as a control for outer membrane integrity. As shown in Fig. 3, antibodies to native ATCC 35405 Msp reacted with both intact and permeabilized T. denticola ATCC 35405 cells. The same antibodies also reacted with permeabilized T. denticola ATCC 33520 cells but did not react with intact ATCC 33520 cells. We interpret this as strong evidence that the 70-residue region of ATCC 35405 Msp that differs from that of ATCC 33520 Msp contains a prominent surface-localized epitope.
Alignment of the Msp central domains of strains ATCC 35405 and ATCC 33520. Amino acid residue numbers are shown at the sequence ends. Purple-highlighted residues represent areas of high predicted antigenic index values (Ag1 and Ag2). Green-highlighted residues represent approximate boundaries of the central domain (D1 and D2). Blue-highlighted residues represent the region of greatest difference between the Msps other than Ag1 and Ag2 (Xr). Other than the region shown here, the amino acid sequences of the two proteins are identical, with 543 and 547 residues, respectively.
Silver-stained polyacrylamide gel of purified native ATCC 35405 Msp. Lane 1, unheated sample; lane 2, boiled sample. Arrows indicate oligomeric Msp (lane 1, 150 to 200 kDa) and monomeric Msp (lane 2, approximately 53 kDa). This preparation was used to raise rabbit antibodies against native Msp.
Immunofluorescence microscopy of T. denticola, revealing the Msp surface domain. T. denticola ATCC 35405 and ATCC 33520, grown to an OD600 of 0.2, were fixed on glass slides with 1% paraformaldehyde, incubated with PBS or PBS plus 0.5% Triton X-100, and probed with rabbit antibodies to native ATCC 35405 Msp or recombinant FlaA, followed by goat anti-rabbit IgG conjugated to Alexa Fluor 568 (Msp) or Alexa Fluor 488 (FlaA), or with DAPI. Images were obtained at a magnification of 600× using an Olympus BX40 microscope fitted with an Olympus DP73 camera.
The Msp surface epitope is localized to residues 229 to 251.To more precisely localize the immunogenic Msp surface epitope, we compared the sequences of the ATCC 35405 and ATCC 33520 Msp central regions (Fig. 1). Using the predictive method of Jameson and Wolf (30), we identified two 7-residue areas in ATCC 35405 Msp with high antigenicity scores (Ag1 and Ag2) that are separated by a 17-residue sequence (Xr) that is highly divergent from that of ATCC 33520 Msp. As illustrated in the schematic map in Fig. 4, we constructed and expressed in E. coli full-length recombinant Msp derivatives lacking Ag1 (deletion of residues 229 to 235), Xr (deletion of residues 237 to 251), Ag2 (deletion of residues 253 to 259), or D2 (deletion of residues 263 to 269, constituting the C-terminal end of the divergent domain). We also constructed and expressed in E. coli a series of recombinant Msp derivatives with deletions in the N-terminal region (deletions of residues 26 to 47, 26 to 69, or 26 to 92), all of which retained the native Msp signal peptide (residues 1 to 20) and the first 5 residues of the mature protein. The N-terminal deletion constructs were included in this experiment as further controls for expression and antibody detection of partial Msp polypeptides. As shown in Fig. 4, antibodies raised against a recombinant N-terminal fragment of Msp (residues 14 to 202) recognized all of the recombinant Msp deletion constructs expressed in E. coli, as well as native Msp from T. denticola ATCC 35405. Antibodies raised against either native ATCC 35405 Msp or whole T. denticola ATCC 35405 cells failed to react with recombinant Msps with deletions of residues 229 to 235 (Ag1) or 236 to 251 (Xr). The combined immunofluorescence and immunoblotting data suggested that a prominent surface-exposed epitope in ATCC 35405 Msp is within the 23-residue region of the predicted Ag1 and Xr domains, which includes the sequence that is most divergent between the ATCC 35405 and ATCC 33520 Msps. These results indicate that this is the predominant surface-localized Msp epitope of strain ATCC 35405. It should be noted that two previous studies proposed that this sequence was within a large N-terminal domain completely contained within the periplasmic space (26, 27).
Map and immunoblots of deletion mutations made in msp for expression in E. coli. (Upper) The 543-residue Msp protein, with the signal peptide (SP) cleavage site indicated (residue 20), is shown above the deletions made in the coding region. The amino acid residues deleted in the N-terminal region in each of the three constructs are indicated. Deletions in the central region within the approximate boundaries of the central domain (D1 through D2) include predicted antigenic domains (Ag1 and Ag2), the remaining region of greatest divergence between Msps of strains ATCC 35405 and ATCC 33520 (Xr), and D2. These mutant Msps were expressed in E. coli. (Lower) Immunoblots of E. coli strains expressing full-length Msps with the indicated Msp residues deleted are shown. T. denticola ATCC 35405 serves as a positive control. Blots were probed with rabbit polyclonal antibodies raised against native ATCC 35405 Msp, the Msp N-terminal domain, or whole T. denticola (Td) cells. All samples were boiled prior to electrophoresis. Lanes contain lysates of E. coli strains. Antibodies raised against native T. denticola ATCC 35405 Msp or whole T. denticola ATCC 35405 cells do not recognize recombinant Msps lacking residues 229 to 251.
Modeling of Msp tertiary structure in a biological context.Having determined the location of the major surface-exposed Msp epitope, we next utilized the I-TASSER server (31) to identify three-dimensional structural models of Msp consistent with the biological and immunological data. Based on the primary amino acid sequence, the I-TASSER algorithm utilizes multiple, iterative, threading alignments to known structures in the regularly updated Protein Data Bank (PDB) (https://www.rcsb.org) to generate potential structural models from the primary amino acid sequence of each protein. Two sets of 5 predicted ATCC 35405 Msp structural models, generated 16 months apart, are shown in Fig. 5A and B. Residues 229 to 251, containing the identified Msp surface epitope, are highlighted in green in each model. The rather low confidence scores (C scores) and the wide range of Msp models reflect the absence of closely related proteins with known structures. Model 3 in Fig. 5B shows Msp as a β-barrel protein, with the identified surface epitope localized to a presumably exposed loop. The predicted Msp structure recently published by Puthenveetil et al. (27) as the optimal model for Msp is model 5 in Fig. 5A. In this model, the surface-exposed Msp epitope is within the portion of the protein predicted by those authors to constitute a large N-terminal periplasmic domain of Msp (25, 26). It should be noted that this model is absent in the set of predicted structures based on the most recent PDB template database (Fig. 5B).
Models for T. denticola ATCC 35405 Msp generated by I-TASSER. The I-TASSER algorithm generates 5 predicted structural models based on several parameters, of which model 1 is the overall best model. Overall model ranking does not necessarily follow the C scores for proteins lacking closely related known structures. All 5 models are displayed to show the N termini (blue) and C termini (red). The ATCC 35405 Msp surface-exposed epitope (residues 229 to 251) is highlighted in green in each model. (A) Models 1 to 5 generated using the PDB data available in June 2017. (B) Models 1 to 5 generated using the PDB data available in October 2018.
We then generated I-TASSER models of the Msps of T. denticola strains ATCC 33520 and OTK (Fig. 6), to determine whether the considerable amino acid sequence differences between these Msps were reflected in the predicted structures. In contrast to the Msps of ATCC 35405 and ATCC 33520, which share >90% sequence identity, ATCC 33520 Msp and OTK Msp share <40% overall homology (Fig. 6A). Interestingly, the most highly ranked structural models of ATCC 33520 and OTK Msps were very similar (C scores of −2.16 and −1.67, respectively), with each showing the Msp as a single β-barrel structure consistent with identification as an outer membrane porin (Fig. 6B). The complete set of I-TASSER models for ATCC 35405, ATCC 33520, and OTK Msps is shown in Fig. S1 in the supplemental material.
T. denticola Msp alignment and structural models for strains ATCC 33520 and OTK. (A) Clustal W alignment of ATCC 33520 and OTK Msps. Identical residues are highlighted in green; nonidentical residues and gaps are in red. Residues outlined in gray (residues 232 to 256 in ATCC 33520 Msp and residues 230 to 254 in OTK Msp) correspond to the Ag1-Xr epitope of ATCC 35405 Msp (Fig. 1). (B) Side and top views of I-TASSER models of ATCC 33520 and OTK Msps (C scores of −2.16 and −1.67, respectively) displayed to show the N termini (blue) and C termini (red). Residues corresponding to the ATCC 35405 Ag1-Xr surface-exposed epitope are highlighted in green.
Several studies have reported that Msp has porin-like channel-forming activity (5, 6, 26–28). The pore diameter of native ATCC 35405 Msp was estimated to be 3.4 nm, based on single-channel conductance studies in black lipid bilayer model membranes (28), and recombinant Msp has been shown to have similar conductance characteristics (6, 26). Based on the structural models shown here, the pore diameters of the ATCC 33520 and OTK Msps are approximately 3.6 nm, calculated as the average distance between three pairs of residues on opposite sides of the predicted barrel structure of each molecule.
To investigate potential factors contributing to the considerable differences in the predicted structure between ATCC 35405 Msp and the Msps of strains ATCC 33520 and OTK, we focused on the sequence of the 70-residue domain in ATCC 35405 (residues 202 to 271) that distinguishes the Msps of ATCC 35405 and ATCC 33520. We noted that the immunogenic domain in ATCC 35405 Msp contains a single diproline motif (residues 249 and 250) that is absent in ATCC 33520 Msp. Because the significant inherent flexibility of the proline ring (particularly when doubled) could potentially influence protein structure predictions (32), we performed an in silico analysis of an ATCC 35405 Msp sequence in which the unique diproline motif was replaced with the cognate glycine-alanine from ATCC 33520 Msp (Fig. 1). Two of the 5 models of ATCC 35405 with the Gly-Ala substitution for Pro-Pro predicted a single β-barrel structure consistent with identification of Msp as an outer membrane porin (data not shown). The I-TASSER model with the highest score (C score of −1.64) is shown in Fig. S2, with the surface epitope (residues 229 to 251) labeled as in Fig. 5.
To provide further information validating the high-scoring structural models generated by I-TASSER that resembled outer membrane porins, we used the TMBpro server, which is specifically designed to predict the tertiary structure and membrane orientation of transmembrane β-barrel (TMB) proteins embedded in the outer membranes of didermal bacteria (33). It should be noted that the I-TASSER algorithm does not directly predict the orientation or membrane topology of the structural models generated. In TMBpro, the structures predicted for the Msps of ATCC 35405, ATCC 33520, and OTK very closely resembled the high-scoring I-TASSER models generated from Msps of ATCC 35405 (Pro249Pro250 or Gly249Ala250), ATCC 33520, and OTK. In each case, the TMBpro algorithm identified the surface-exposed ATCC 35405 Msp epitope as part of an outward-facing loop of the β-barrel porin molecule (Fig. S3).
DISCUSSION
In this study, we utilized several approaches to determining Msp topology in the T. denticola outer membrane. First, we identified a surface-exposed epitope of T. denticola ATCC 35405 Msp within a 23-residue domain in the central region of the protein that lies between the predicted Msp N-terminal (pfam02707, residues 109 to 208) and C-terminal (pfam02722, residues 360 to 543) domains. Then, using in silico protein structural modeling, we showed that (i) the proposed structural model in which the N-terminal half of Msp is localized to the periplasmic space is inconsistent with the immunological data and (ii) a proposed model of ATCC 35405 Msp as a β-barrel porin is both similarly predicted for very diverse T. denticola Msp molecules and wholly consistent with the immunological data.
The N- and C-terminal regions of Msp and Msp-like proteins are annotated as distinct domains with separate Pfam entries in the Conserved Domain Database (CDD; https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi), which has been regularly updated to incorporate new sequences and structures and improved analytical techniques (34, 35). This updating is reflected in several changes in the defined boundaries of the putative Msp N-terminal and C-terminal domains over the past few years. Two earlier reports on predicted Msp structure and topology (26, 27) described the MOSP-N domain as encompassing residues 113 to 348, including residues 204 to 270 (the central domain), based on CDD information current at the time. It should be noted that the latest updated version of the CDD shows that, in ATCC 35405 Msp, the MOSP-N domain (pfam02707) and the MOSP-C domain (pfam02722) encompass residues 109 to 208 and 361 to 543, respectively, such that the neither the N-domain nor the C-domain includes the portion of Msp containing the surface epitope identified here. This evolving understanding of the molecular architecture of Msp makes it even more important that bioinformatics-based structure and topology predictions are informed and supported by empirical biological evidence.
Identification of the predominant surface-exposed Msp epitope advances the knowledge of Treponema outer membrane biology in several ways. This epitope defines a major source of antigenic heterogeneity observed among T. denticola strains. This is consistent with previous studies that showed that (i) human sera from different subjects recognized antigenically distinct T. denticola Msp types (36) and (ii) the ATCC 35405 Msp central domain containing this epitope also contains a fibronectin-binding domain (4). In this context, it is relevant that the major immunogenic epitopes of T. pallidum TprK are localized to the variable regions of this molecule (37). It remains to be determined whether T. denticola Msp has more than a single surface-exposed epitope. Studies in progress will address this by in-frame deletion of all or part of the identified epitope. More importantly for the present study, localization of this epitope constrains the range of potential models that can accurately represent the native Msp structure, thus permitting rational choices among the wide range of potential structures predicted by in silico modeling.
Translocation to and oligomerization of bacterial β-barrel-containing proteins in the outer membrane are generally dependent on specific chaperones such as SurA and the outer membrane β-barrel assembly machinery (BAM) complex, which includes BamABCDE and several associated proteins (reviewed in reference 38). T. denticola Msp was recently reported to interact with both BamA and SurA (27), suggesting that its native expression may be BamA dependent. Because recognition of outer membrane protein precursors by the BAM complex typically involves a conserved C-terminal pattern of aromatic residues (38, 39), both the two-domain and porin models of the Msp structure appear to be consistent with BAM complex-dependent translocation and oligomerization of Msp in the outer membrane. Studies are in progress in our laboratory to address the role of the Msp C terminus in localization to the T. denticola outer membrane.
We relied on two structural modeling algorithms to model potential Msp structures. We primarily utilized I-TASSER, a robust general homology-based prediction algorithm, to generate potential Msp structures consistent with the biological data. The I-TASSER algorithm produces 5 predicted structural models (ranked 1 to 5). This ranking is based on a C score combined with a template modeling (TM) score and a root mean square deviation (RMSD), which are two measures for quality prediction (40). For proteins having homology with known structures, the C score is highly correlated with the TM score and RMSD. For proteins (such as Msp) that lack solved homologous structures, overall model “quality” ranking does not necessarily follow the C score. We also used TMBpro, a porin-specific modeling algorithm (33), to compare and to help validate biologically germane Msp models generated by I-TASSER.
Bioinformatic analysis of Msp structure has been somewhat limited, due to two factors. First, general prediction algorithms such as I-TASSER rely heavily on databases of solved crystal structures to identify similarities between the protein of interest and known protein structures, in order to generate models with a high level of confidence. Proteins such as Msp and Msp-like proteins (including the Tprs of T. pallidum) are Treponema specific, and none has been crystallized to date. Second, structure prediction methodology for porins and porin-like molecules has lagged behind that for other protein classes because conventional methods relying on hydropathy patterns are not very useful for analyzing proteins that consist largely of amphiphilic β-strands that form a membrane-embedded, water-filled channel (41). Until fairly recently, relatively few porin structures had been described. While considerable progress has been made, there is still no clear consensus regarding porin structural modeling using general prediction algorithms (42, 43). However, because TMB structures conform to specific construction rules that drastically restrict the conformational space, it is possible for alignment-free methods such as TMBpro to successfully model this class of proteins (33).
One of the I-TASSER structural models generated for ATCC 35405 Msp was recently published as representing the optimal Msp structural model, based on consistency with sequence-based domain predictions (27). As shown in Fig. 5, that model (model 5) has the lowest C score of the 5 models generated. More importantly, in that model the surface-exposed Msp epitope resides within what was proposed as the Msp N-terminal periplasmic domain. Data presented herein clearly show that the bipartite model cannot accurately represent the actual structure and membrane topology of Msp.
The highest-ranked I-TASSER models for Msps of strains ATCC 33520 and OTK are surprisingly similar, given the low level of amino acid homology between these proteins. Both models strongly indicate an overall β-barrel structure typical of outer membrane porins, with the region corresponding to the surface-exposed epitope in ATCC 35405 Msp being localized to an extracellular loop. The finding that the predicted pore diameters of ATCC 33520 and OTK Msp closely match the empirically determined pore size of ATCC 35405 Msp (6, 26, 28) lends further support to the conclusion that these models closely represent the native Msp structure.
We were initially puzzled by the major differences between the structural models of ATCC 35405 Msp (on one hand) and ATCC 33520 and OTK Msps (on the other hand). We find it highly intriguing that replacement of the Pro-Pro motif in the Xr domain of ATCC 35405 with the cognate Gly-Ala motif from ATCC 33520 resulted in I-TASSER models for the modified ATCC 35405 Msp that closely resembled those for ATCC 33520 and OTK. Two of the 5 I-TASSER models of ATCC 35405 Msp with the Gly-Ala substitution for Pro-Pro show a single β-barrel structure consistent with the identification of Msp as an outer membrane porin (data not shown). I-TASSER modeling of proteins that lack structurally determined homologs remains somewhat problematic. The most likely reason for the dramatic change in structure prediction resulting from the Pro-Pro to Gly-Ala substitution is considerable loss of potential polypeptide chain flexibility at that point, which likely constrains the range of possible predicted structures. We speculate that this reflects both the iterative template-based approach and the structural flexibility of the diproline motif (32). As to any potential function of a diproline motif in Msp, it should be noted that there are no other proline doublets in any of the 3 Msps examined here. Proline doublets are present in a total of 4 of >60 sequenced Msps (13), and all are found at various locations within the predicted Msp central domain. Targeted in vivo mutagenesis of these residues is being pursued for better understanding of this issue.
In summary, we have identified the predominant surface-exposed epitope of T. denticola ATCC 35405 Msp as a 23-residue sequence within the central domain of Msp. Identification of this epitope constrains the range of possible three-dimensional structures that are also consistent with Msp porin activity to a β-barrel structure generally similar to that of Gram-negative outer membrane porins. Studies in progress include site-directed mutagenesis throughout the msp gene to characterize potential intramolecular interactions between the N-terminal and C-terminal portions of Msp that may contribute to the stability of the oligomeric β-barrel structure.
MATERIALS AND METHODS
Bacterial strains and growth conditions.T. denticola strains ATCC 35405 and ATCC 33520 (oral phylogroup 2) were grown under anaerobic conditions in new oral spirochete (NOS) broth, as described previously (44). All growth media were incubated under anaerobic conditions for at least 18 h prior to use (45). The purity of spirochete cultures was monitored by dark-field microscopy. E. coli strains were grown in LB broth or agar medium supplemented with kanamycin (50 µg ml−1), carbenicillin (50 µg ml−1), or chloramphenicol (34 µg ml−1), as appropriate. Routine cloning was performed in E. coli JM109 (46). Recombinant proteins were expressed in E. coli C41 (47) carrying pLysS (48).
Purification of T. denticola Msp.Native Msp was purified by preparative electrophoresis from 4-liter batch cultures of T. denticola ATCC 35405 as described previously (6), with minor modifications. Briefly, cells were collected by centrifugation (3,500 rpm for 10 min at 4˚C), washed twice in phosphate-buffered saline (PBS) containing 5 mM MgCl2 and 1 mM phenylmethylsulfonyl fluoride (PMSF), and then extracted overnight at 4˚C, with gentle stirring, in 100 ml 1% Triton X-114 (ANAPOE-X-114; Anatrace, Maumee, OH) containing 5 mM MgCl2 and 1 mM PMSF. The aqueous supernatant was collected by centrifugation (12,000 rpm for 10 min at 4˚C). Triton X-114 was added to a final concentration of 2%, the mixture was incubated at 37˚C for 15 min, and then phase partitioning was performed by centrifugation (4,000 rpm for 10 min at 37˚C). The aqueous phase was subjected to two more rounds of Triton X-114 extraction and then concentrated in a stirred-cell ultrafiltration unit fitted with an Amicon XM-50 filter (Millipore, Inc., Beverly, MA). The final aqueous phase (approximately 6 ml) was subjected to preparative SDS-PAGE (5% acrylamide) using a model 491 Prep Cell (Bio-Rad Laboratories, Richmond, CA). Samples were electrophoresed at 60 mA at 4°C. The cathode buffer contained 25 mM Tris (pH 8.3), 192 mM glycine, and 0.1% SDS, and the anode and elution buffers consisted of 25 mM Tris (pH 8.3) with 192 mM glycine. The eluate was collected in 5.0-ml fractions, at a flow rate of 1 ml min−1. Fractions were analyzed by SDS-PAGE with silver staining. Fractions containing Msp were concentrated approximately 10-fold with an Amicon XM-50 filter (Millipore) and then were subjected to buffer exchange by three washes with 10 volumes of PBS containing 0.1% 3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate (CHAPS). Following determination of the protein concentration using the bicinchoninic acid (BCA) protein assay (Thermo Scientific, Rockford, IL), the purified Msp protein was used to immunize rabbits to generate polyclonal antibodies, as described previously (29).
Immunofluorescence microscopy of T. denticola.T. denticola cells grown to an optical density at 600 nm (OD600) of 0.2 were fixed on glass slides with 1% paraformaldehyde, washed with PBS or PBS plus 0.5% Triton X-100, blocked, and probed with rabbit antibodies to native ATCC 35405 Msp or recombinant FlaA, followed by goat anti-rabbit IgG conjugated to Alexa Fluor 568 (Msp) or Alexa Fluor 488 (FlaA). Coverslips were sealed with ProLong Gold antifade mounting reagent with 4',6-diamidino-2-phenylindole (DAPI) (Life Technologies, Carlsbad, CA). Images were obtained at a magnification of 600×, using an Olympus BX40 microscope fitted with an Olympus DP73 camera (49).
Construction of expression plasmids with deletions in the Msp N-terminal and central domains.Plasmid pCF31, which carries the full-length msp gene under transcriptional control of the T7 RNA polymerase in pET17b (3), was used as the template for construction of several defined deletions by ligation-independent PCR cloning (50, 51). Briefly, using high-fidelity DNA polymerase (Phusion; New England Biolabs, Beverly, MA), partially complementary primers with sequences 5′ and 3′ of the desired deletions (Table 1) were used to amplify the entire pCF31 template, minus the deleted sequence. The PCR templates were removed by overnight digestion with DpnI. E. coli JM109 was then transformed with the resulting DNA, and transformants were screened by PCR and DNA sequencing to confirm the expected constructs. The resulting plasmid constructs are listed in Table 2. Recombinant constructs were confirmed by PCR and by DNA sequencing at the University of Michigan DNA Sequencing Core Facility and were analyzed using DNA sequence analysis components of the Lasergene Molecular Biology Suite (DNASTAR Inc., Madison, WI).
Oligonucleotide primers used in this study
Plasmids used for expression studies in E. coli
Expression of recombinant proteins in E. coli.Plasmid DNAs carrying various msp constructs were introduced into the E. coli expression strain C41/pLysS. Briefly, an overnight culture from a single colony was diluted 1:10 in fresh medium and incubated at 37˚C, with shaking, until reaching an OD600 of 0.5 to 0.6. Protein expression was induced by the addition of isopropyl-β-d-thiogalactopyranoside (IPTG) to 0.2 mM, and the culture was further incubated for 3 h. Cells were collected by centrifugation for 10 min at 4,000 rpm and lysed by suspension in 1× Laemmli sample buffer and repeated passage through a 26-gauge needle. Samples were subjected to 10% SDS-PAGE and analyzed by Western blotting with rabbit antibodies raised against the ATCC 35405 Msp N-terminal domain (residues 14 to 202) (4), native ATCC 35405 Msp (this study), recombinant Msp (3), T. denticola ATCC 35405 cells (8), or T. denticola FlaA (29).
Protein gel electrophoresis and immunoblotting.SDS-PAGE and Western immunoblotting were performed as described previously (3). Total cell lysates of T. denticola strains and E. coli strains expressing Msp constructs (either heated at 100° C for 5 min or held on ice) were separated by SDS-PAGE and then stained or transferred to nitrocellulose membranes, which were probed with rabbit polyclonal antibodies followed by horseradish peroxidase (HRP)-conjugated goat anti-rabbit IgG (Thermo Scientific). Protein bands of interest were visualized using the SuperSignal West Pico chemiluminescent substrate (Thermo Scientific).
Predictive bioinformatic analysis of Msp structure.Three-dimensional Msp structural models were generated from primary amino acid sequences using two public servers, i.e., the I-TASSER server (https://zhanglab.ccmb.med.umich.edu) and the TMBpro server (http://tmbpro.ics.uci.edu). I-TASSER generates 5 potential structural models for each protein, of which model 1 is the overall best model based on several parameters, including the RMSD of atomic positions of superimposed proteins, the TM score, and the C score (40). The TM score assesses the quality of protein structure templates and predicted full-length models (52), while the C score is calculated based on the significance of threading template alignments and the convergence parameters of the structure assembly simulations (31). C scores fall in the range of −5 to 2, with a C score of higher value signifying a model with higher confidence. In general, a C score greater than −1.5 indicates a model of correct global topology (31). While C scores are highly correlated with TM scores and RMSD values for proteins with closely related known structures, overall model ranking does not necessarily follow C scores for unknown structures (such as Msp).
TMBpro is a suite of specialized predictors for predicting the secondary structure, β-contacts, and tertiary structure of TMB proteins (33). Because homology-based modeling methods often fail with TMB proteins, TMBpro uses alignment-free methods within a set of specific construction rules that restrict the conformational space. Protein structural models generated with I-TASSER and TMBpro were labeled, scaled, and annotated using the Protean 3D component of the Lasergene Molecular Biology Suite (DNASTAR, Inc.).
ACKNOWLEDGMENTS
We thank Chengxin Zhang (Department of Computational Medicine and Bioinformatics, School of Medicine, University of Michigan) for helpful discussions. We also thank Howard Jenkinson and Angela Nobbs (University of Bristol, Bristol, UK) for providing rabbit polyclonal antibodies.
This study was supported by grants DE-025225 and DE-018221 (to J.C.F.) from the National Institute of Dental and Craniofacial Research.
FOOTNOTES
- Received 30 August 2018.
- Accepted 22 October 2018.
- Accepted manuscript posted online 29 October 2018.
Supplemental material for this article may be found at https://doi.org/10.1128/JB.00528-18.
- Copyright © 2018 American Society for Microbiology.
