Characterization of Treponema denticola Major Surface Protein (Msp) by Deletion Analysis and Advanced Molecular Modeling

ABSTRACT

Treponema denticola, a keystone pathogen in periodontitis, is a model organism for studying Treponema physiology and host-microbe interactions. Its major surface protein Msp forms an oligomeric outer membrane complex that binds fibronectin, has cytotoxic pore-forming activity, and disrupts several intracellular processes in host cells. T. denticola msp is an ortholog of the Treponema pallidum tprA to -K gene family that includes tprK, whose remarkable in vivo hypervariability is proposed to contribute to T. pallidum immune evasion. We recently identified the primary Msp surface-exposed epitope and proposed a model of the Msp protein as a β-barrel protein similar to Gram-negative bacterial porins. Here, we report fine-scale Msp mutagenesis demonstrating that both the N and C termini as well as the centrally located Msp surface epitope are required for native Msp oligomer expression. Removal of as few as three C-terminal amino acids abrogated Msp detection on the T. denticola cell surface, and deletion of four residues resulted in complete loss of detectable Msp. Substitution of a FLAG tag for either residues 6 to 13 of mature Msp or an 8-residue portion of the central Msp surface epitope resulted in expression of full-length Msp but absence of the oligomer, suggesting roles for both domains in oligomer formation. Consistent with previously reported Msp N-glycosylation, proteinase K treatment of intact cells released a 25 kDa polypeptide containing the Msp surface epitope into culture supernatants. Molecular modeling of Msp using novel metagenome-derived multiple sequence alignment (MSA) algorithms supports the hypothesis that Msp is a large-diameter, trimeric outer membrane porin-like protein whose potential transport substrate remains to be identified.

IMPORTANCE The Treponema denticola gene encoding its major surface protein (Msp) is an ortholog of the T. pallidum tprA to -K gene family that includes tprK, whose remarkable in vivo hypervariability is proposed to contribute to T. pallidum immune evasion. Using a combined strategy of fine-scale mutagenesis and advanced predictive molecular modeling, we characterized the Msp protein and present a high-confidence model of its structure as an oligomer embedded in the outer membrane. This work adds to knowledge of Msp-like proteins in oral treponemes and may contribute to understanding the evolutionary and potential functional relationships between T. denticola Msp and the orthologous T. pallidum Tpr proteins.

INTRODUCTION

Oral spirochetes, most notably Treponema denticola, are associated with 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 on host cells, particularly the intracellular responses, remain to be determined. Edwards et al. (4) reported that recombinant polypeptides consisting of Msp residues 14 to 202 and 203 to 259, respectively, 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. reported that a recombinant polypeptide encompassing Msp residues 272 to 406 inhibited neutrophil chemotactic responses (7). The location of a specific active Msp domain(s) relative to Msp structure and the mechanisms by which Msp affects neutrophils and other cells are as yet poorly understood. Determining the topology of the Msp complex in the outer membrane is a crucial step toward understanding its role 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 three of the seven phylogroups of human oral treponemes (11, 12), with genotypes corresponding to T. denticola 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 best studied and can be divided into three groups: two that are very closely related (represented by strains 35405 and 33520, which are greater than 90% identical, differing only in a 70-residue central domain corresponding to residues 202 to 271 in 35405) and one represented by strain OTK, which has only about 40% total homology with the other two (13, 19). A group of orthologous Treponema pallidum rare outer membrane proteins (TprA to -K) shares evolutionary history with the Msps, though sequence homology is fairly limited (20). Functional characterization suggests involvement, as with Msps, in outer membrane permeability (21) and, unlike with Msps, in rapid in vivo generation of intrastrain antigenic variants in the TprK locus during infection that likely contributes to disease persistence (22, 23).

Outer membrane topologies of both T. denticola Msp and the T. pallidum Tpr family remain unresolved, with little consensus either on overall structure or on organization and localization of polypeptide domains within the proteins. Several recent studies have reported predictive molecular models of T. pallidum Tprs and T. denticola Msps as comprised of structurally and functionally distinct domains: an N-terminal periplasmic domain and a C-terminal porin domain with β-barrel structure (21, 24, 25). These models are largely supported by studies of recombinant T. pallidum TprC/D and T. denticola 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 comprised of a large β-barrel structure with a central surface-exposed domain (4, 26) that is divergent between strains (13, 19). Our recent immunotopological analysis showed that the surface-exposed antigenic epitope of Msp is localized between residues 229 and 251, which comprise an extracellular loop in a predictive model of Msp as a membrane-spanning β-barrel protein (27). While native T. denticola Msp is a detergent-stable trimeric oligomer, porin-active Msp expressed in E. coli appears to be monomeric (3). Similarly, a recent study reported that T. pallidum TprK expressed in E. coli is a porin-active monomer (28). The purpose of the present study was to further characterize immunological and topological characteristics of T. denticola Msp that contribute to its expression as a trimeric outer membrane complex by mutational analysis of its N-terminal, C-terminal, and antigenic domains (Fig. 1) and by advanced molecular modeling of the native protein complex. Here, we report data consistent with identification of Msp as a classic trimeric β-barrel outer membrane protein complex.

FIG 1.

Map of T. denticola 35405 Msp protein summarizing individual deletion mutations made in msp for expression in T. denticola. The complete 543-residue Msp protein is illustrated above, with the signal peptide cleavage site (SP, residue 20) indicated. Amino acids deleted or replaced by a FLAG tag in the N-terminal and antigenic regions are indicated with numbering from the complete amino acid sequence. Deletions of 0 to 7 amino acids from the C terminus are indicated (Δ0 to Δ7). Predicted N-glycosylation sites flanking the antigenic region (residues 229 to 251) are indicated at Asn residues 194 and 297.

RESULTS AND DISCUSSION

The Msp C terminus is required for native expression and outer membrane oligomerization.

Translocation to and oligomerization of porins in the outer membrane of didermal bacteria are generally dependent on specific periplasmic chaperones such as SurA and the outer membrane β-barrel assembly machinery (BAM) complex, which includes BamABCDE and several associated proteins (reviewed in reference 29). This system is present in Treponema and is under study both in T. pallidum (30–34) and in oral treponemes (35–37). Since recognition of unfolded outer membrane porins in the periplasm by BamA is reported to be dependent on a motif of aromatic residues at the extreme C terminus (38, 39), we tested the effects of C-terminal truncations on expression of Msp in T. denticola. Both T. denticola Msps and T. pallidum Tprs show a pattern of conservation of aromatic residues in the C terminus (Fig. 2A). We previously reported construction of an isogenic msp mutant in which DNA encoding the last 48 Msp amino acid residues was replaced in frame with 47 extraneous residues from a partial open reading frame in the 5′ end of the ermF/B cassette used to make the mutant. This strain produced only a small amount of a monomeric protein reactive with antibodies to Msp (40). To examine this further in a more defined manner, we first constructed isogenic T. denticola mutants with deletions of 8, 16, 24, 32, and 40 C-terminal Msp residues. Msp protein could not be detected in any of these strains (data not shown). We then constructed a series of isogenic T. denticola mutants with deletions of between 1 and 7 C-terminal Msp residues (Fig. 1). These mutant strains, as well as a control strain with no deletion, each carry an erythromycin resistance cassette 3′ to the msp stop codon for selection purposes. In Western immunoblotting assays (Fig. 2B), strains carrying deletions of 1 or 2 C-terminal residues (Δ1 and Δ2) expressed native oligomeric Msp similarly to the wild-type strain and the control strain (Δ0). Only monomeric Msp was detectable in the Δ3 mutant, and mutants lacking more than 3 C-terminal residues produced no detectable Msp. Immunofluorescence assays of intact and permeabilized cells (Fig. 2C) confirmed these results, showing loss of cell surface Msp reactivity in mutants with deletions of more than two C-terminal residues, with some minimal reactivity in permeabilized cells lacking three and four C-terminal residues. These results strongly support a role for the conserved BamA pathway in outer membrane localization and oligomerization of Msp and complement a recent report that identified BamA as one of 49 proteins that coimmunoprecipitated with Msp in a T. denticola outer membrane preparation (25). Further studies will be required to characterize interactions between the BamA complex and Msp during its translocation to and assembly in the outer membrane.

FIG 2.

Analysis of C-terminal Msp deletions. (A) Aligned C-terminal amino acid sequences of Msps of three representative T. denticola strains and T. pallidum TprK. Aromatic amino acid residues are shaded. (B) Immunoblots of unheated samples of T. denticola strains carrying deletions of 0 to 7 amino acids from the C terminus (Δ0 to Δ7) probed with antibodies to native Msp (top) and FlaA (bottom). Oligomeric Msp is recognized at 150 kDa to 200 kDa, while monomeric Msp reacts at approximately 53 kDa. FlaA reacts at approximately 37 kDa. (C) Immunofluorescence microscopy of T. denticola strains showing loss of expression and surface localization due to Msp C-terminal deletion mutagenesis. T. denticola 35405 and derivatives grown to an OD₆₀₀ of 0.2 were fixed on glass slides with 1% paraformaldehyde, incubated with PBS (intact cells) or PBS plus 0.5% Triton X-100 (TX-100) (permeabilized cells), and probed with rabbit antibodies to native 35405 Msp or recombinant FlaA, followed by goat anti-rabbit Alexa Fluor 488, or with 4′,6-diamidino-2-phenylindole (DAPI). Images were obtained at 600× using an Olympus BX40 microscope fitted with an Olympus DP73 camera.

Construction of complete msp deletion mutant.

In previously reported T. denticola Msp mutants, the msp coding region was disrupted by an antibiotic cassette such that Msp protein was either not detectable or greatly reduced (40). Cloning, expression, and mutagenesis studies of Msp have been complicated due to the fact that expression of the full-length protein from its native promoter is toxic in E. coli (3, 8). Furthermore, the msp gene contains the sequence ⁴⁶⁷GGGCCC⁴⁷³ encoding two overlapping recognition sites for the TdeIII restriction enzyme system (41) that cannot be altered while conserving the native amino acid sequence. Prior reports indicated that TdeIII restriction activity is a major factor limiting genetic transformation in T. denticola (41, 42). Because these features interfere with construction of fine-scale mutations in the N-terminal and central regions of Msp (data not shown), we constructed a complete Δmsp strain for use as a mutagenesis host in a T. denticola 35405/ΔTDE0911 strain that lacks TdeIII restriction activity (41). In the resulting Δmsp/ΔTDE0911 strain (T. denticola CF1090; Table 1), the msp coding region and promoter are replaced by an aphA2 gene cassette encoding kanamycin resistance (Km^r) (43). Subsequent specific deletion constructs were made in this strain. As described in the Supplemental Materials and Methods, CF1090 was transformed with linear DNA fragments consisting of the msp promoter and msp coding region containing defined mutations, followed by the aacC1 gene encoding gentamicin resistance (Gm^r). In each case, the introduced msp gene carried a mutation of interest, and for each mutant strain recovered, the entire msp coding region and promoter sequence were validated by DNA sequencing.

TABLE 1.

Treponema denticola strains used in this study

Strain	Genotype	Source/reference
35405	Wild-type parent strain	ATCC
CF678	35405; msp + 6×His tag; Emr	This study
CF975	35405; msp with no C-terminal deletion; Emr	This study
CF988	35405; msp with 1-residue C-terminal deletion; Emr	This study
CF989	35405; msp with 2-residue C-terminal deletion; Emr	This study
CF983	35405; msp with 3-residue C-terminal deletion; Emr	This study
CF985	35405; msp with 4-residue C-terminal deletion; Emr	This study
CF986	35405; msp with 5-residue C-terminal deletion; Emr	This study
CF990	35405; msp with 6-residue C-terminal deletion; Emr	This study
CF987	35405; msp with 7-residue C-terminal deletion; Emr	This study
CF637	35405; ΔTDE0911; Emr	41
CF1090	CF637; Δmsp promoter→coding region; Emr/Kmr	This study
CF1094	CF1090; msp (Δ26–33); Emr/Gmr	This study
CF1095	CF1090; msp (Δ26–33) FLAG tag; Emr/Gmr	This study
CF1096	CF1090; msp (Δ229–235) FLAG tag; Emr/Gmr	This study
CF1098	CF1090; msp (Δ229–235); Emr/Gmr	This study

Deletion within the N-terminal Msp domain blocks oligomer formation.

To initiate characterization of the role of the Msp N terminus, we constructed two strains carrying a deletion of DNA encoding Msp residues 26 to 33, such that the signal peptide and first five residues of the mature protein were retained in order to ensure proper processing of the Msp signal peptide in mutant strains (Table 1 and Fig. 1). The Δmsp/ΔTDE0911 strain T. denticola CF1090 was transformed with linear DNA fragments carrying a deletion of DNA encoding residues 26 to 33 (VTAKASVN) or with a fragment in which the same DNA was replaced with DNA encoding the FLAG tag epitope (DYKDDDDK; Fig. 3A). One strain (CF1094) carries a deletion of 8 residues (residues 26 to 33), while in the other (CF1095), the same deletion is replaced by the FLAG tag epitope. Protein expression in each strain was analyzed by immunoblotting using antibodies raised against native Msp (27), T. denticola FlaA (44), and the FLAG tag epitope (GenScript-USA, Piscataway, NJ). As shown in Fig. 3C, monomeric Msp was detected in the T. denticola parent and in both mutants using antibodies raised against native Msp. Sodium dodecyl sulfate (SDS)-stable oligomeric Msp was present in unheated samples of 35405 but was not detected in either mutant strain. Some Msp degradation products were detected in all strains but were more evident in the mutants. As expected, the FLAG tag epitope was detected in CF1095 as a band of the same apparent molecular weight as Msp. These results show that the Msp N terminus is required for Msp oligomer formation and that in the absence of oligomerization, Msp is more susceptible to degradation. A FlaA immunoblot was included as a loading control. It should be noted that we often observe that the unheated portion of a single T. denticola protein preparation shows reduced reactivity to FlaA antibody compared with the heated portion of the same sample. This is evident for all four strains shown in Fig. 3C and to a lesser degree in Fig. 3D.

FIG 3.

Western blots of N-terminal and antigenic domain mutants. (A) N-terminal residues of Msp in T. denticola 35405 (wild type) and N-terminal mutant strains CF1094 (Δ26–33) and CF1095 (Δ26–33/FLAG tag). The Msp signal peptide is shaded. The FLAG tag sequence in CF1095 is underlined. (B) Msp antigenic region in T. denticola 35405 (wild type) and Ag1 mutant strains CF1098 (Δ229–235) and CF1096 (Δ229–235/FLAG tag). The Msp antigenic domain in 35405 is indicated by a line over residues 239 to 251. The FLAG tag sequence in CF1096 is underlined. (C and D) Western immunoblots of T. denticola lysates probed with the indicated antibodies against Msp, FLAG tag, or FlaA. Samples were unheated (−) or heated (+) prior to electrophoresis. Panel C shows N-terminal mutants CF1094 (ΔN) and CF1095 (ΔN-FLAG), and panel D shows antigenic domain mutants CF1096 (ΔAg1) and CF1098 (ΔAg1-FLAG). Each blot includes wild-type strain T. denticola 35405 and isogenic Δmsp strain CF1090. The gray bar in panel D obscures an irrelevant sample on the blots. MW, molecular mass in kDa.

Mutagenesis of the Msp surface-exposed epitope.

We recently identified the primary Msp surface-exposed epitope located between residues 229 and 251 (27), which is within a 70-residue Msp central domain that contains the only differences between the antigenically distinct Msps of strains 33405 and 33520 (4, 19, 27). To determine the role of the central antigenic region in Msp expression, oligomerization, and membrane topology, we deleted residues 229 to 235 (designated Ag1 [27]), which encompass the first eight residues of the antigenic domain. As illustrated in Fig. 3B, the Δmsp strain T. denticola CF1090 was transformed with linear DNA fragments carrying a deletion of DNA encoding residues 229 to 235 (yielding ΔAg1 strain CF1098) or with a fragment in which the DNA encoding these residues was replaced with DNA encoding a FLAG tag (yielding ΔAg1-FLAG strain CF1096). Protein expression in each strain was analyzed by immunoblotting using antibodies raised against native Msp (27), T. denticola FlaA (44), and the FLAG tag epitope (GenScript-USA, Piscataway, NJ). As shown in Fig. 3D, anti-FLAG tag antibodies reacted only with CF1096, while anti-Msp antibodies reacted with parent strain 35405 and CF1096, but not with CF1098. While the high-molecular-weight Msp complex was strongly reactive at >200 kDa in unheated 35405 samples, the monomeric form of Msp prevailed in unheated CF1096 samples, with only very weak reactivity at >200 kDa corresponding to the oligomer band in 35405. The complete lack of anti-Msp reactivity with CF1098 was expected, as we had previously shown that, in recombinant constructs expressed in E. coli, deletion of either residues 229 to 235 or residues 236 to 251 abrogated recognition of Msp by either anti-Msp antibodies or antibodies raised against T. denticola cells (27). However, the strong reactivity of CF1096 with anti-Msp antibodies suggests that replacement of the Ag1 peptide with the FLAG tag preserves conformation of the adjacent downstream sequence sufficient for antibody binding to the Msp epitope. Further planned fine-scale mutagenesis throughout the antigenic domain (residues 229 to 251) will aid in more precise definition of both the specific antigenic epitope and the role of this specific domain in expression of native oligomeric Msp.

The Msp surface epitope is resistant to proteolysis.

Proteinase K (PK) treatment of intact cells has been used to identify surface-exposed proteins in a wide range of bacteria, including spirochetes (45, 46). It is particularly useful in determining cellular localization of their abundant lipoproteins, some of which are localized to the outer leaflet of the outer membrane (47). Two recent studies reported that treatment of intact T. denticola with relatively high levels of PK resulted in partial cleavage of Msp, releasing a polypeptide of approximately 25 kDa (24, 25) that was recognized by antibodies raised against an Msp polypeptide encompassing residues 77 to 282 (24), which includes the Msp surface epitope (residues 229 to 251) recently identified by Godovikova et al. (27). Anand and coworkers concluded that, since this polypeptide was not susceptible to proteolysis, it must be localized to the periplasm and have no surface exposure (24, 25). To further explore this issue, we conducted expanded surface proteolysis experiments in which both T. denticola cell pellets and cell-free supernatants were analyzed for Msp and Msp fragments following treatment of intact cells with PK. As shown in Fig. 4, PK treatment of T. denticola released a 25 kDa Msp polypeptide fragment into the supernatant that was recognized by rabbit antibodies raised against Msp (3). The 25 kDa band was not detected in untreated control cells, pellets of PK-treated cells, or PK-treated sonic lysates (Fig. 4B). As expected, PK (29 kDa) was detected in culture supernatants and sonic lysates but not in cell pellets (Fig. 4A). It is of interest that a considerable amount of full-length Msp remained in all samples after PK treatment of intact cells, consistent with its localization primarily within the outer membrane but with an extracellular antigenic domain region having limited susceptibility to PK cleavage.

FIG 4.

Proteinase K treatment of T. denticola cells releases a 25 kDa Msp fragment containing the Msp surface epitope. Following treatment of T. denticola cells with proteinase K (PK) for 25, 50, or 100 min, cell pellets, supernatants, and sonicates of cell cultures were analyzed by SDS-PAGE and immunoblotting with anti-recombinant Msp antibodies. Untreated T. denticola served as control. Msp full-length protein and 25 kDa fragment are indicated in both panels. (A) Silver-stained gel. (B) Immunoblot in anti-Msp antibodies. Proteinase K (PK; 29 kDa) was detected in supernatants and sonicates of PK-treated cell cultures but not in pellets of PK-treated cells. Full-length Msp and a 25 kDa Msp fragment are indicated. The 25 kDa Msp fragment was detected only in culture supernatants. Numbers at left are molecular masses in kDa.

To determine the boundaries of the immunoreactive 25 kDa Msp polypeptide in the supernatant fraction of PK-treated T. denticola cells, the corresponding region was excised from an SDS-PAGE gel and subjected to liquid chromatography-mass spectrometry (LC-MS/MS) analysis. The highest number of spectra obtained were assigned to Msp (64% total coverage). Msp spectra corresponding to residues 151 to 364 (80% coverage) correspond to a 22.1 kDa polypeptide including the identified antigenic surface epitope (residues 239 to 251), consistent with the observed apparent molecular weight of this PK-resistant polypeptide (Fig. 4B) as well as with the prior PK studies in T. denticola (24, 25). PK proteolysis and LC-MS/MS data confirm that the surface-exposed Msp epitope is present on a polypeptide released from T. denticola by PK digestion and that the epitope itself is not susceptible to PK under the experimental conditions utilized. This is consistent with reports from several investigators that the high level of T. denticola intrinsic secreted protease activity can be utilized to remove other proteins from T. denticola crude outer membrane extracts, resulting in highly purified Msp oligomeric complex (6, 26, 48).

The mechanism by which the IgG-accessible epitope is resistant to PK needs to be further explored. The present results suggest several areas for further investigation. The absence of the 25 kDa Msp fragment in PK-treated sonic lysates suggests that the oligomeric structure of Msp in its native form, including the predicted arrangement of the individual Msp epitopes in the oligomeric structure (discussed in below), may contribute to its documented resistance to protease activity. In this context, membrane protein glycosylation can decrease or block proteolytic susceptibility, presumably by steric hindrance or by contributing to stability in the membrane (49). Protein glycosylation is well documented to interfere with PK sensitivity of bacterial surface proteins (50–52). Msp is reported to be a glycoprotein (53), and deglycosylation of purified native Msp inhibited its ability to bind to Fusobacterium nucleatum, a T. denticola coaggregation partner in the oral cavity (54). Thus, it is of particular interest that the PK-resistant Msp polypeptide includes both the antigenic surface domain and the two predicted Msp N-glycosylation sites flanking it (N¹⁹⁴ and N²⁹⁷; Fig. 1). We did not deglycosylate the excised PK-resistant band before MS analysis, which may have affected the quality of MS reads. It may be necessary to apply alternative techniques such as negative-mode mass spectrometry (55) to acquire complete sequence coverage of the PK-resistant polypeptide. Further analysis of this region is of high interest, both for better understanding of the outer membrane architecture of T. denticola and for its potential clinical applications. For example, a recent report described characterization of an Msp polypeptide (Ala¹⁷⁴-Leu²⁸²) consisting of the region approximately bounded by the two glycosylation sites described here (N¹⁹⁴, N²⁹⁷; Fig. 1) as a potential component of a multivalent periodontal disease vaccine (56).

Enhanced structural modeling of Msp.

Past efforts to model the T. denticola Msp protein and related proteins, including the T. pallidum Tpr family, are prime examples of the limitations of predictive structure modeling of proteins that lack closely related homologs with solved structures. Modeling of these Msp-like proteins over the past decade has yielded two incompatible models, both of which have some level of support in experimental data (6, 21, 24–27). We noted recently that Msp structure predictions generated by I-TASSER (57) “evolved” considerably over a 3-year period (27), presumably due to ongoing expansion of the Protein Data Bank (PDB; https://www.rcsb.org), which is the source of the potential structural alignment templates in this algorithm. Several models of Msp structure previously generated in standard I-TASSER (25, 27) are of marginal reliability based on their rather low confidence scores (C-scores [57]) and template modeling scores (estimated template modeling [eTM] scores [58, 59]).

Traditional protein structure prediction algorithms rely heavily on the expanding array of empirically determined PDB tertiary structures. I-TASSER, one of the more robust algorithms currently available, utilizes multiple iterative threading alignments with known structures in the regularly updated PDB to generate potential structure models from the primary amino acid sequence of each query protein. However, the reliability of predictive modeling is limited when studying proteins that lack reasonably homologous PDB templates. This issue was recently addressed by Pearce and Zhang (60), who reviewed recent advances in high-resolution structure prediction independent of the PDB template availability, including novel metagenome-derived multiple sequence alignment (MSA) algorithms.

Modeling the Msp monomer.

Msp protein is a single-domain protein as predicted by FUpred (61). The D-I-TASSER (62) method, which interplays the conformations from AlphaFold2 (63) as initial conformations, was utilized for modeling the entire single-domain monomer structure of Msp protein, and the AlphaFold2-multimer pipeline was used for modeling the trimer complex of the Msp proteins, where the default threading templates were replaced by D-I-TASSER monomer models. In all cases, the mature Msp protein (residues 21 to 543) lacking its type I signal peptide was analyzed. Details of the structure prediction methodology are found in Materials and Methods. Briefly, multiple sequence alignments (MSAs) in the D-I-TASSER and AlphaFold2 monomer pipelines were generated by DeepMSA2 (62) with the MetaSource (64) selected metagenomics data source, which is an extended pipeline based on the DeepMSA methodology (62) by adding pipelines to search larger metagenome sequence databases. Three individual MSA generation pipelines, dMSA, qMSA, and mMSA, in DeepMSA2, were used in D-I-TASSER and AlphaFold2. This provides a much broader range of potentially similar structural elements in addition to the PDB library (https://www.rcsb.org), which consists of empirically determined protein structures. The initial conformations (templates) used in the D-I-TASSER simulation were generated by AlphaFold2 monomer modeling and LOMETS3 (65). In total, 50 models from 10 MSAs were generated by AlphaFold2 and ranked by the pLDDT score. These were used as the initial conformations, and collect contact/distance/hydrogen-bond restraints for the D-I-TASSER folding simulations.

As shown in Fig. 5, the predicted monomeric Msp is a large porin-like β-barrel in the outer membrane, similar to the model we recently reported using the standard I-TASSER algorithm (27), with only the antigenic domain (residues 229 to 251, Fig. 3B and Fig. 5) exposed on the cell surface. In contrast to our prior report (27), the present model generates an eTM score of 0.8 to 0.9, well above the generally accepted reliability threshold of 0.5 (62). In the present model, the Msp N and C termini are in close proximity, suggesting that they may interact in the properly folded native molecule. This is consistent with data from the N- and C-terminal mutant strains shown in Fig. 2 and 3. Deletion of only a few residues (eight residues at the N terminus; three residues at the C terminus) reduced stability of the protein and prevented oligomerization of the remaining “full-length” polypeptide, consistent with predicted interaction with the BamA complex during translocation to the outer membrane (29). As we previously demonstrated (27), the antigenic domain (residues 229 to 251, shown in green and magenta in Fig. 5) is clearly exposed on the cell surface in this model. Interestingly, while deletion of residues 229 to 235 in strain CF1098 (ΔAg1; green in Fig. 5) results in loss of reactivity with anti-Msp antibodies, replacement of residues 229 to 235 with a FLAG tag in CF1096 (ΔAg1-FLAG) rescues anti-Msp reactivity in Western immunoblot assays (Fig. 3D). However, the oligomeric form of Msp was not detected in CF1096, suggesting that the full native epitope region contributes to oligomer formation. Thus, while the FLAG tag in CF1096 may permit presentation of the remainder of the antigenic epitope (residues 236 to 251; magenta in Fig. 5) in proper conformation for antibody recognition, it is not sufficient for oligomerization. As noted above, further studies are required to more specifically define this epitope and its potential role in Msp oligomeric structure and function.

FIG 5.

Predicted structure of mature Msp monomer. Panel A is a side view of the monomeric protein, presumably as inserted in the outer membrane as a β-barrel. Panel B is a top view of the monomeric protein showing the large-diameter pore. Eight N-terminal residues deleted in CF1094 are shown in blue. Seven residues at the C terminus encompassing progressive deletions in CF984 to CF990 (Table 1) are shown in red. The Ag1 domain (residues 229 to 235) deleted in CF1098 or replaced by FLAG tag in CF1097 is shown in green. The remainder of the antigenic domain (residues 236 to 251) is shown in magenta.

Modeling the Msp trimer complex.

Three copies of the monomer sequences were input to the AlphaFold2-multimer pipeline to build the trimer complex. The default threading templates used in the AlphaFold2-multimer pipeline were replaced by five D-I-TASSER top-ranked models, and the monomer MSA used in the AlphaFold2-multimer pipeline was replaced by DeepMSA2 MSA. The final trimer complex model from AlphaFold2-multimer with D-I-TASSER model and DeepMSA2 MSA has an eTM-score of 0.64, whereas running the default AlphaFold2-multimer alone yields an eTM-score of 0.42. The structural model of trimeric Msp as localized in the outer membrane is shown in Fig. 6. It is interesting that the portion of Msp that exhibits a lesser amount of β-sheet structure is the region at which the three monomeric β-barrels interact to form a membrane-embedded oligomer. Also of note is that the Msp antigenic domains of each monomer (residues 229 to 251, shown in red) appear to be closely associated in the oligomeric structure. This is consistent with prior observations that antibodies raised against native oligomeric Msp are more highly reactive with the oligomeric form of Msp than are antibodies raised against monomeric recombinant Msp expressed in E. coli (reference 19 and data not shown). Beyond the crucial role of the BamA complex, mechanisms involved in outer membrane trimer formation in Gram-negative bacteria in general (66–68), and specifically in spirochetes (33, 69), are not yet fully understood. Additional studies are planned to investigate the mechanism of Msp oligomer formation and stabilization in the outer membrane, including identification of intermonomer interactions and the potential contribution of T. denticola outer membrane glycolipid (lipooligosaccharide) to this process (70).

FIG 6.

Structural model of the Msp oligomeric complex. Panel A is a side view as inserted in the outer membrane. Panel B is a top view. Individual Msp monomers are shown in blue, green, and yellow. The antigenic region of each monomer (residues 229 to 251) is shown in red.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

T. denticola strains were grown under anaerobic conditions (37°C; 5% CO₂, 10% H₂ in N₂) in TYGVS broth or semisolid medium containing 0.8% Noble agar as previously described (71), and supplemented with erythromycin (Em; 40 μg mL⁻¹), kanamycin (Km; 25 μg mL⁻¹), or gentamicin (Gm; 20 μg mL⁻¹) as appropriate. All T. denticola growth media were incubated under anaerobic conditions for at least 18 h prior to use (72). Purity of spirochete cultures was monitored by darkfield microscopy. E. coli strains were grown in LB medium supplemented with appropriate antibiotics (Km, 50 μg mL⁻¹; Em, 200 μg mL⁻¹; Gm, 5 μg mL⁻¹; carbenicillin, 50 μg mL⁻¹). Routine cloning was done in E. coli strains JM109 and JM110 (73).

DNA constructs for mutagenesis studies.

Recombinant DNA molecules used for T. denticola mutagenesis were constructed either in E. coli plasmids or by a combination of overlap extension PCR (74, 75) and FastCloning (76). Plasmids constructed for T. denticola mutagenesis studies are listed in Table 2. Due to the high activity of the msp promoter and toxicity of Msp expression in E. coli, DNA constructs for mutagenesis in the N-terminal and antigenic domains of Msp were entirely made using PCR methodology. Details of construction of individual plasmids and PCR products are contained in the Supplemental Materials and Methods. All plasmids were constructed using either restriction fragments of previously verified plasmids or PCR products generated with high-fidelity DNA polymerases (Phusion [New England Biolabs, Beverly, MA] or Taq-HF [Invitrogen, Carlsbad, CA]). Recombinant constructs were confirmed by DNA sequencing (Eurofins Genomics, Louisville, KY) and analyzed using DNAStar sequence analysis software (DNAStar Inc., Madison, WI).

TABLE 2.

Plasmids used in this study

Plasmida	Description	Selectionb	Reference or source
pKX102	5.5-kb T. denticola DNA in pBluescript II KS	Cbr (Ec)	8
pVA2198	Bacteroides shuttle vector carrying ermF-ermB	Emr (Ec)	83
pCF31	msp coding region in pET17b	Cbr (Ec)	3
pCF639	3′-end msp-6×His/ermB/TDE0406 in pSTBlue-1	Kmr (Ec)	This study
pCF1014	pCF639, with ermB replaced by aacC1	Kmr (Ec)	This study
pCF844	pCF639, XbaI-SalI in MCSc deleted	Kmr (Ec)	This study
pCF846	Derivative of pCF844; mspΔ0	Kmr (Ec)	This study
pCF976	Derivative of pCF844; mspΔ1	Kmr (Ec)	This study
pCF977	Derivative of pCF844; mspΔ2	Kmr (Ec)	This study
pCF978	Derivative of pCF844; mspΔ3	Kmr (Ec)	This study
pCF979	Derivative of pCF844; mspΔ4	Kmr (Ec)	This study
pCF980	Derivative of pCF844; mspΔ5	Kmr (Ec)	This study
pCF981	Derivative of pCF844; mspΔ5	Kmr (Ec)	This study
pCF982	Derivative of pCF844; mspΔ7	Kmr (Ec)	This study
pCF931	1.2 kb 5′ to msp promoter in pGEM-T Easy	Cbr (Ec)	This study
pCF932	1.1 kb 3′ to the msp coding region in pCF932	Cbr (Ec)	This study
pCF933	For allelic replacement deletion of msp	Cbr (Ec); Kmr (Td)	This study

^aConstruction of these plasmids, including intermediate steps, is described in Supplemental Materials and Methods.

^bSelectable marker in E. coli (Ec) or T. denticola (Td).

^cMCS, multiple-cloning site.

Allelic replacement mutagenesis of T. denticola.

Defined isogenic mutants within the msp gene were constructed as described previously (40, 77), by electroporation of T. denticola with unmethylated linear DNA fragments (either linearized plasmid fragment or purified PCR product) consisting of a selectable antibiotic resistance cassette between DNA fragments flanking the mutagenesis target. Mutants were selected for growth in TYGVS agar containing appropriate antibiotics. Mutations were verified by PCR analysis and DNA sequencing of the target region in genomic DNA of the mutant strains.

Protein gel electrophoresis and immunoblotting.

Protein electrophoresis and immunoblotting were carried out as described previously (3). T. denticola cultures were harvested by centrifugation at 10,000 × g (10 min, 4°C) in the presence of 2 mM phenylmethylsulfonyl fluoride (PMSF), washed once in phosphate-buffered saline (PBS), suspended in 20 mM Tris (pH 8) in a volume equal to (0.5 mL × culture optical density at 600 nm [OD₆₀₀]), and then lysed by sonication, followed by repeated passage through a 26-gauge (G) needle to fragment chromosomal DNA. Equal volumes of samples were prepared in standard SDS-PAGE sample buffer containing dithiothreitol and 2 mM PMSF. Samples (either heated at 100°C for 5 min or held on ice) were separated by SDS-PAGE and transferred to nitrocellulose membranes, which were then probed with rabbit polyclonal antibodies against the protein of interest or monoclonal antibodies against FLAG tag (GenScript-USA, Piscataway, NJ), followed by horseradish peroxidase (HRP)-conjugated goat anti-rabbit or anti-mouse IgG (Thermo Scientific, Rockford, IL). Protein bands of interest were visualized using SuperSignal West Pico chemiluminescent substrate (Thermo Scientific) and imaged using a G:Box imaging system (Syngene, Frederick, MD).

Proteinase K treatment of Treponema cells.

Surface proteolytic treatment of T. denticola cells was done using standard protocols (50, 78), with minor variations. Five-milliliter cultures of T. denticola cells were harvested by centrifugation (1,500 × g, 10 min), resuspended in 1 mL proteolysis buffer (50 mM Tris-HCl, pH 8, 7.5 mM CaCl₂), and then split into four 250-μL aliquots. Aliquots of intact cells were either held on ice or disrupted by sonication on ice (three-30 s cycles at 30% duty; Branson sonifier; Branson Ultrasonics Corp., Danbury, CT). Intact cells and sonicated cells were then treated with proteinase K (PK; Sigma-Aldrich, St. Louis, MO) at 100 μg mL⁻¹ in 50 μL proteolysis buffer at 37°C for the indicated times. PK was then inactivated by adding 20 μL of 0.1 M phenylmethylsulfonyl fluoride (PMSF; Sigma-Aldrich). Tubes containing PK-treated intact cells were centrifuged at 1,500 × g for 10 min. Supernatants were removed, and the cell pellets were resuspended in 300 μL of proteolysis buffer. Aliquots from each PK-treated sample (sonicate, cell pellet, and supernatant) plus an untreated cell control were separated by SDS-PAGE and analyzed by silver staining or Western blotting using rabbit polyclonal antibodies raised against Msp.

Mass spectrometry.

Following PK treatment and SDS-PAGE analysis of supernatant fractions from intact cells, a gel slice containing a protein band of slightly less than 25 kDa immunoreactive with anti-N-terminal Msp antibody was excised and subjected to liquid chromatography-mass spectrometry (LC-MS/MS) analysis at the University of Michigan Protein Structure Core Facility. The submitted gel band was processed by in-gel digestion with trypsin using a ProGest robot (DigiLab). Half of each digest was analyzed by nano-LC-MS/MS with a Waters NanoAcquity high-performance liquid chromatography (HPLC) system interfaced with a ThermoFisher Q Exactive. Peptides were loaded on a trapping column and eluted over a 75-μm analytical column at 350 nL/min; both columns were packed with Luna C₁₈ resin (Phenomenex). The mass spectrometer was operated in data-dependent mode, with the Orbitrap operating at 70,000 full width at half maximum (FWHM) and 17,500 FWHM for MS and MS/MS, respectively. The 15 most abundant ions were selected for MS/MS. Data were searched against the UniProt T. denticola database using a local copy of the Mascot search engine (Matrix Science, Boston, MA).

Msp monomer modeling.

Multiple sequence alignments (MSA) in D-I-TASSER and AlphaFold2 monomer pipelines were generated by DeepMSA2 (62) with the MetaSource (64) selected metagenomics data source, which is an extended pipeline based on the DeepMSA methodology (62) by adding pipelines to search larger metagenome sequence databases. Three individual MSA generation pipelines, dMSA, qMSA, and mMSA, in DeepMSA2, were used in D-I-TASSER and AlphaFold2. The dMSA pipeline is identical with the previous DeepMSA method. First, the query sequence was searched through Uniclust30 (version 2017_04) by HHblits2 to create dMSA-1. Next, the sequences identified by Jackhmmer and HMMsearch were used to construct a custom HHblits database, against which HHblits2 is run starting from the MSA generated in the previous stage to generate dMSA-2 and dMSA-3, respectively. A qMSA pipeline (quadruple MSA) was used to produce four more MSAs. First, HHblits2 was used to search against the Uniclust30 database (version 2020_01) to create qMSA-1. Next, the sequences detected by Jackhmmer, HHblits3, and HMMsearch through the UniRef90, BFD, and Mgnify databases were used to construct custom HHblits-style databases, against which HHblits2 were used to search starting from the MSA generated by the previous stage to create qMSA-2, qMSA-3, and qMSA-4, respectively. In mMSA (or multilevel MSA), the qMSA-3 alignment was used as a probe by HMMsearch to search through the IMG/M database and the resulting sequence hits were converted into a sequence database. This mMSA database was then used as the target database, which was searched by three seed MSAs (dMSA-2 and qMSA-2 and qMSA-3), to derive three new MSAs (mMSA-1, mMSA-2, and mMSA-3). The 10 MSAs (dMSA-1, dMSA-2, dMSA-3, qMSA-1, qMSA-2, qMSA-3, qMSA-4, mMSA-1, mMSA-2, and mMSA-3) were fed into AlphaFold2 to get 10 sets of full-length monomer models.

Spatial restraints of MSAs were predicted by DeepPotential. The 10 MSAs from DeepMSA2 were fed into DeepPotential (79) to get 10 predicted contact maps. The final MSA used in DeepPotential full set restraint generation is the MSA that has the highest cumulative probability for the top 10L predicted contacts, where L is the sequence length. Three types of spatial restraints can be predicted by DeepPotential, including (i) contact maps, (ii) distance maps, and (iii) hydrogen bond networks (62). In the DeepPotential pipeline, a set of coevolutionary features were extracted from the final MSA generated by DeepMSA2. These coevolutionary features, which are inherently two dimensional, include the raw coupling parameters from the pseudo-likelihood-maximized (PLM) 22‐state Potts model and the raw mutual information (MI) matrix. The PLM and MI matrices were extracted from the query‐specific coevolutionary information in the given MSA. The Potts model field parameters, hidden Markov model (HMM) features, and the self‐mutual information are the major one‐dimensional inputs, along with the one‐hot representation of the MSA and other descriptors, such as the number of sequences in the MSA. These two‐dimensional and one‐dimensional features were fed into deep convolutional residual neural networks separately, where each of them was passed through a set of one‐dimensional and two‐dimensional residual blocks, respectively, and subsequently tiled together. The tiled feature representations are considered the inputs of another fully residual neural network containing 40 two-dimensional (2-D) residual blocks which output several interresidue interaction terms, including the contact map, distance map, and hydrogen bond networks.

Initial conformations for monomeric Msp protein were generated by AlphaFold2 and LOMETS3. The initial conformations (templates) used in the D-I-TASSER simulation were generated by AlphaFold2 monomer modeling and LOMETS3 (65). Ten MSAs from DeepMSA2 were fed into AlphaFold2 monomer modeling pipeline to generate 10 sets of models. In total, 50 models are generated by AlphaFold2 and ranked by the pLDDT score. The pLDDT scores of the 50 models generated by AlphaFold2 with DeepMSA2 range from 0.71 to 0.91, while the pLDDT score of the AlphaFold2 model with the default MSA pipeline is only 0.81. LOMETS3 was used here for generating another 50 templates. LOMETS3 contains six profile-based threading methods and five contact-/distance-based methods. The MSA generated by DeepMSA2 was used to produce sequence profiles (or profile HMMs) for the six profile-based threading methods and to predict contact maps by DeepPotential for the five contact-based threading methods in LOMETS3. The top-ranked 50 templates associated with 50 AlphaFold2 models were used as the initial conformations and collect contact/distance/hydrogen-bond restraints for the D-I-TASSER folding simulations.

Atomic full-length models were generated by D-I-TASSER simulation. Fragments were extracted from initial conformations and assembled into models using a modified replica exchange Monte Carlo (REMC) simulation procedure. A force field, which combines the spatial restraints obtained from the AlphaFold2 and LOMETS3 initial conformations and DeepPotential with the inherent knowledge-based energy terms, was used to guide the D-I-TASSER structural assembly simulations. Five REMC simulations were performed, where the structural decoys from eight low-temperature replicas were submitted to SPICKER (80) for structure clustering and model selection. The SPICKER clusters were refined at the atomic level using fragment-guided molecular dynamics (FG-MD) simulations (81), and finally, the side chain rotamer structures were repacked by FASPR (82). Finally, the final D-I-TASSER model of monomer Msp protein has an estimated TM (eTM) score (62) of 0.82 and pLDDT score of 0.92. Both the eTM score and pLDDT score indicate that the monomer Msp model from D-I-TASSER is a highly accurate atomic-level model.

Msp trimer complex modeling.

Three copies of the monomer sequences were input to the AlphaFold2-multimer pipeline to build the trimer complex. The default threading templates used in the AlphaFold2-multimer pipeline were replaced by five D-I-TASSER top-ranked models, and the monomer MSA used in the AlphaFold2-multimer pipeline was replaced by DeepMSA2 MSA.