Transcriptional and functional characterizations of multiple flagellin genes in spirochetes

Abstract The flagellar filament is a helical propeller for bacterial locomotion. In external flagellates, the filaments are mostly homopolymers of a single flagellin protein. By contrast, the flagellar filaments of spirochetes are mostly heteropolymers of multiple flagellin proteins. This report seeks to investigate the role of multiple flagellin proteins using the oral spirochete Treponema denticola as a model. First, biochemical and genetic studies uncover that the flagellar filaments of T. denticola mainly comprise four proteins, FlaA, FlaB1, FlaB2, and FlaB3, in a defined stoichiometry. Second, transcriptional analyses reveal that the genes encoding these four proteins are regulated by two different transcriptional factors, sigma28 and sigma70. Third, loss‐of‐function studies demonstrate that each individual flagellin protein contributes to spirochete motility, but none of them is absolutely required. Last, we provide genetic and structural evidence that FlaA forms a “seam”‐like structure around the core and that deletion of individual flagellin protein alters the flagellar homeostasis. Collectively, these results demonstrate that T. denticola has evolved a unique mechanism to finely regulate its flagellar filament gene expression and assembly which renders the organelle with the right number, shape, strength, and structure for its distinct motility.


| INTRODUC TI ON
The bacterial flagellum is a sophisticated nanomachine that consists of three mechanical units: the basal body (motor), the rodhook complex (shaft-joint), and the filament (propeller) (Armitage & Berry, 2020;Chevance & Hughes, 2008;Erhardt et al., 2010). The basal body is imbedded within the cell envelope and works as a reversible rotary motor powered by an inward-directed electrochemical gradient of protons or sodium ions (Atsumi et al., 1992;. The torque generated by the motor is mechanically transmitted to the flagellar filament, which rotates at hundreds of revolutions per second to propel bacterial locomotion (Berg, 2000). The bacterial flagellar filament has been intensively studied due to its critical role in motility and host-pathogen interactions, for example, activating host innate immunity through Toll-like receptors (Hayashi et al., 2001;Yoon et al., 2012). In addition, the flagellar filament has served as an enlightening system for understanding how a polymer composed of a single flagellin protein is assembled and functions as an Archimedean screw to generate mechanical propulsion (Maki-Yonekura et al., 2010;Samatey et al., 2001;Wang et al., 2017;Yonekura et al., 2003). Assembly of the flagellar filament starts after the flagellar hook is completed, whereby thousands of copies of flagellin proteins are exported through the flagellar-specific type III secretion system (fT3SS) (Chevance & Hughes, 2008;Erhardt et al., 2010). With the assistance of the cap protein FliD (Al-Otaibi et al., 2020), an assembling chaperone, the flagellin proteins polymerize to a long helical structure that consists of 11 protofilaments that can adopt both left-handed (L-type) and right-handed (R-type) conformations due to mechanical forces, such as when the motor switches the sense of rotation (i.e., from counterclockwise to clockwise), allowing bacteria to swim forward or backward or to tumble (Maki-Yonekura et al., 2010;Pandini et al., 2016;Wang et al., 2017;Yonekura et al., 2003).
Flagellin proteins are highly conserved among different bacterial species, as is their quaternary structure. By using cryo-electron microscopy (cryo-EM) and helical reconstruction, Galkin et al. reported that the structures of the flagellar filaments from six different bacteria are all composed of 11 protofilaments (Galkin et al., 2008;Wang et al., 2017). In addition to the conservation of amino acid sequences and structures, the regulatory mechanism of flagellin gene expression is also conserved. In most external flagellates, the genes encoding flagellin proteins belong to class III genes in the flagellar regulatory hierarchy, which are finely regulated by sigma 28 (FliA), a flagellar-specific alternative transcription activator, and its antagonist, FlgM (Chevance & Hughes, 2008;Lertsethtakarn et al., 2011;Subramanian & Kearns, 2019). Some bacterial flagellin genes, such as Campylobacter jejuni and H. pylori, are regulated by both sigma 28 and sigma 54 (Hendrixson & DiRita, 2003;Prouty et al., 2001;Suerbaum et al., 1993). In addition to transcriptional regulation, the biosynthesis of flagellin proteins in some bacteria can also be regulated at the translational level. For example, CsrA, a small RNA binding protein, negatively regulates the translation of flagellin proteins of B. subtilis, Borrelia burgdorferi, and several other bacteria (Dugar et al., 2016;Kao et al., 2014;Mukherjee et al., 2011;Romeo & Babitzke, 2018;Sze et al., 2011).
Spirochetes are a unique group of bacteria readily recognized by their flat-waved or coiled cell morphology and distinct form of corkscrew-like motility (Charon et al., 1981;Charon et al., 2012;Goldstein et al., 1994) and are responsible for several human diseases, including Lyme disease (B. burgdorferi), syphilis (Treponema pallidum), and leptospirosis (Leptospira interrogans) (Picardeau, 2017;Radolf et al., 2016;Rosa et al., 2005). Unlike external flagellates, spirochetes swim by means of rotating two bundles of periplasmic flagella (PFs) that reside between the outer membrane and cell cylinder (Charon et al., 2012;. The number and length of PFs vary from species to species. In general, spirochetal PFs are structurally similar to the flagella of other bacteria as each consists of a basal body-motor complex, a hook, and a filament. However, the spirochetal flagellar filaments are distinct in terms of their protein composition and structures (Charon et al., 2012;Zhao et al., 2013). For instance, the flagellar filaments of Brachyspira and Treponema species are composed of three FlaBs (e.g., FlaB1, FlaB2, and FlaB3), homologs of FliC, and a sheath protein, FlaA, that has no sequence similarity to the flagellin proteins (Kurniyati et al., 2017;Li et al., 2008;Norris et al., 1988).
The flagellar filament of Leptospira is even more complex as it is composed of two FlaA proteins, at least one FlaB protein, and two novel Leptospira-specific flagellar filament proteins, FcpA and FcpB (Gibson et al., 2020;Lambert et al., 2012;Wunder Jr. et al., 2016;Wunder Jr. et al., 2018).
Why spirochetes have evolved multiple flagellar filament proteins and what their roles are with respect to flagellar filament assembly, structure, and bacterial motility are longstanding questions (Charon et al., 2012). We first sought to address this question by using Brachyspira (formerly known as Treponema and Serpulina) hyodysenteriae, the first spirochete that could be genetically manipulated. Our results indicate that, while each individual flagellin protein contributes to motility, FlaA appears to be more important than individual FlaBs, for example, deletion of flaA, but not flaBs, impairs the filament diameter and helicity Li et al., 2008). A different scenario occurs in Leptospira biflexa, in which deletion of flaB eliminates filament assembly entirely (Picardeau et al., 2001). In L.
interrogans, deletion of flaA only affects the flagellar filament curvature, but the mutant becomes non-motile and non-infectious, highlighting the importance of FlaA in spirochete motility and infectivity (Lambert et al., 2012). Deletions of fcpA and fcpB abolish the ability of flagella to assume their characteristic supercoiled form and thus dramatically reduce motility and virulence (Wunder Jr. et al., 2016;Wunder Jr. et al., 2018). By integrating high-resolution cryo-electron tomography (cryo-ET) and X-ray crystallography, Gibson et al presented structural evidence that the Leptospira filaments are coated by a highly asymmetric, multi-component sheath that resides at the filament inner and outer curvatures and defines the supercoiling geometry, a key functional attribute of Leptospira PFs (Gibson et al., 2020).
Similar to other spirochetes, Treponema denticola, an oral spirochete and keystone pathogen of human periodontitis, possesses flagellar filaments composed of at least one sheath protein FlaA, three flagellin proteins (FlaB1, FlaB2, and FlaB3), and a filamentassociated protein, FlaG (Kurniyati et al., 2017;Kurniyati et al., 2019;Ruby et al., 1997;Seshadri et al., 2004). The three FlaB proteins are modified by a novel 450.2 Da glycan, which is essential for flagellar filament assembly and motility (Kurniyati et al., 2017). The genes encoding these filament proteins were inactivated in our previous report (Kurniyati et al., 2017); however, their roles in flagellar filament assembly, structure, and bacterial motility remain largely unknown.
This report aims to fill this knowledge gap by using a multidisciplinary approach of genetics, biochemistry, and cryo-ET. The results shown here further highlight the complexity and uniqueness of flagellar filaments in spirochetes and provide several new perspectives on what we have learned from the paradigm model organisms of external flagellates such as E. coli and B. subtilis.

| Measuring the stoichiometry of T. denticola flagellar filament proteins
Previous studies indicated that the flagellar filaments of T. denticola consist of a sheath protein, FlaA (TDE1712), and three core flagellin proteins, FlaB1 (TDE1477), FlaB2 (TDE1004), and FlaB3 (TDE1475) (Kurniyati et al., 2017;Ruby et al., 1997); however, their stoichiometry in the assembled flagellar filaments remained unknown. To fill this gap, we first purified the flagellar filaments from T. denticola ATCC 35405 strain (wild type, WT) to homogeneity (Figure 1a), which were then subjected to SDS-PAGE for protein separation and quantification. However, the molecular weights (MWs) of the three FlaBs were too similar (30.9-31.6 kDa, Table 1) to be completely separated by SDS-PAGE ( Figure 1b). To overcome this issue, the purified flagellar filaments were subjected to 2D gel electrophoresis followed by Coomassie blue staining ( Figure 1c) and immunoblotting probed against FlaA and FlaB antibodies (αFlaA and αFlaB) (Figure 1d). On 2D gels, the four filament proteins were completely separated, and their positions on the gels correlate with the protein isoelectronic points (pIs); for example, FlaB3 (pI 5.3) is more acidic than FlaA (pI 5.39), FlaB1 (pI 5.4), and FlaB2 (pI 6.54) (Figure 1c,d and Table 1). We then measured the densitometry of individual flagellar filament proteins, and the data is expressed as averaged ratios of individual FlaB proteins relative to FlaA (Figure 1e). The stoichiometry of FlaA, FlaB1, FlaB2, and FlaB3 is approximately 1.0:0.7:0.6:0.2 ( Figure 1e). We also examined the protein profile of purified PFs by using mass spectrometry. In addition to FlaA, FlaB1, FlaB2, and FlaB3, we also detected a trace amount of other flagellar proteins, such as FlgE (hook protein), FlgK (hook-associated protein), FliD (cap protein), FliF (MS-ring protein), and several flagellar rod proteins (e.g., FlgB, FlgC, FlgG, and FlgF) (Table S1). Compared to the flagellin proteins, the abundance of these flagellar proteins is extremely low in the purified PFs. For example, FlgE, the most abundant among these detected minor flagellar proteins, accounts for only 4.2% of FlaB1 (calculated based upon protein molar percentage). Taken together, these results indicate that the flagellar filament of T. denticola is a multiple proteinaceous structure that is mainly composed of FlaA, FlaB1, FlaB2, and FlaB3.

| The flagellar filament genes of T. denticola are regulated by different transcriptional factors
Our previous study showed that the flaB1 and flaB3 genes reside in a large motility gene operon regulated by a sigma 70 -like promoter, P flaB1 (Kurniyati et al., 2019). However, the regulation of flaA and flaB2 remains unknown. The flaA gene (TDE1712) is located in a large gene cluster of 11 open reading frames (orfs) that starts from TDE1716 and ends at TDE1706. Interestingly, flaA is an orphan flagellar gene in this cluster (Seshadri et al., 2004). Co-reverse trancription-PCR (Co-RT-PCR) analysis using five pairs of primers (P 1 to P 10 ) that bridge flaA and its four orfs upstream revealed that these five genes are co-transcribed ( Figure 2a,b). There is a 252 bp untranslated region (UTR) upstream F I G U R E 1 Characterizations of T. denticola flagellar filaments. (a) a representative transmission electron microscopic (TEM) image of flagellar filaments isolated from WT. 5 μl of the purified flagellar filaments applied to formvar-carbon copper grids and then stained with 1% uranyl acetate for 1 min (pH 4.2). The samples were subjected to a JEOL JEM-1400 plus TEM at an acceleration voltage of 120.0 kV. (b) SDS-PAGE analysis of WT flagellar filaments. (c) 2D gel electrophoresis and (d) immunoblotting analyses of WT flagellar filaments. For the immunoblotting, antibodies against T. denticola FlaA (αFlaA) and T. pallidum FlaB (αFlaB) were used. (e) the average stoichiometry of the flagellar filament proteins, FlaA, FlaB1, FlaB2, and FlaB3. The densitometry of four flagellar filament proteins was measured from three sets of 2D gels using Bio-Rad image lab software, and the data is expressed as ratios of FlaB1, FlaB2, and FlaB3 relative to FlaA.  Collectively, these results indicate that the flagellar filament genes of T. denticola are differently regulated, whereby flaB2 is controlled by sigma 28 , and the other genes are regulated by sigma 70 .

| The flagellar filament genes of T. denticola have different expression patterns
Given that the four flagellar filament genes of T. denticola are regulated by two different promoters, we speculated that they may have different expression dynamics during growth. To determine if this is the case, we isolated the RNA samples from T. denticola cells harvested at days 3 (middle log), 4 (late log), and 5 (stationary) ( Figure 3a Table 4; asterisk (*) represents the transcriptional start site. (b) co-RT-PCR analysis. This experiment was performed as previously documented (Kurniyati et al., 2019). Five pairs of primers that bridge flaA and its upstream genes were designed and used for co-RT-PCR. For each co-RT-PCR reaction, a parallel PCR reaction was performed and used as a positive control. Of note, the pair of P 9 /P 10 that bridges TDE1716 and TDE1717, two divergently transcribed genes, was used as a negative control to rule out the possibility of DNA contamination in RNA samples. The numbers below the primers are the predicted sizes of PCR products, and the numbers below (e.g., 1712-1713) illustrate the genes that are bridged by each pair of primers (e.g., P 1 /P 2 ) for the co-RT-PCR analysis. The resultant co-RT-PCR and PCR products were detected in 2% agarose gel electrophoresis. (c) a diagram showing the genes adjacent to flaB2 (TDE1004) and its upstream sequence containing a promoter-like sequence. Red colored ATG is the start codon of flaB2. (d) Mapping the transcriptional start sites of flaA and flaB2 genes using 5′-RLM-RACE. This experiment was performed using the FirstChoice RLM-RACE kit (Ambion) according to the manufacturer's protocol. Arrows show the sequencing direction; asterisks (*) are the TSS. (e) Sequence comparison between the E. coli sigma 70 (top panel), sigma 28 (lower panel) promoters, and the three flagellar filament gene promoters mapped in T. denticola, including the promoter sequences upstream of TDE1716 (P flaA ), TDE1004 (P flaB2 ), and a previously identified promoter for flaB1, P flaB1 . (f) Transcriptional analysis of P flaA , P flaB1 , and P flaB2 using lacZ as a reporter in E. coli. For this assay, the three promoters were fused to the promoterless lacZ gene in the pRS414 plasmid. The empty vector was used as a negative control. β-Galactosidase activity was measured and expressed as the average Miller units of triplicate samples from three independent experiments, as previously described. All the primers used here are listed in Table 4.
relative to that of dnaK (TDE0628), a housekeeping gene of T. denticola, using quantitative reverse transcription-PCR (qRT-PCR), as previously documented (Kurniyati et al., 2019). The flaA gene was constantly expressed and not affected by the growth phase, whereas the three flaB genes were actively expressed at the middle log phase, and then flaB2 and flaB3 started to decline at the late log phase, as did flaB1 at the stationary phase (Figure 3), suggesting that the expression of the three flagellin genes is affected by the growth phase of T. denticola.

| Assessing the role of individual flagellar filament genes in the motility of T. denticola
We previously in-frame replaced four individual flagellar filament genes with an erythromycin resistance marker (ermB) (Kurniyati et al., 2017); the resulting mutants are designated as: ΔflaA, ΔflaB1, ΔflaB2, and ΔflaB3. Western-blot analyses using whole-cell lysates showed that the cognate gene products were abolished in these mutants as expected, for example, FlaA was absent in ΔflaA, and FlaB3 was abolished in ΔflaB3 ( Figure 4a). These four mutants were less motile than WT under dark-field microscopy (Videos 1-5), suggesting that motility is impaired in these mutants. We then quantitatively assessed the impact of individual flagellar filament genes on T. denticola motility by using swimming plate assays and a computer-based bacterial tracking system, as previously described (Kurniyati et al., 2019). Swimming plate assays showed that the swimming rings formed by the four mutants are significantly (p < .01) smaller than those of WT but larger than those of Δtap1, a non-motile mutant previously constructed (Limberger et al., 1999) (Figure 4b). By using the bacterial tracking system,

F I G U R E 3
Measuring the expression of four flagellar filament genes during T. denticola growth. (a) the growth curve of T. denticola. For this experiment, T. denticola wild-type ATCC 35405 (WT) was grown in TYGVS medium and enumerated every 24 h using a Petroff-Hausser counting chamber. (b) the expressional level of flaA, flaB1, flaB2, and flaB3 at different growth phases. T. denticola cultures were collected at the mid-log, late-log, and stationary phase as indicated in (a) and subjected to RNA isolation followed by qRT-PCR, which was performed using iQ SYBR green supermix and a MyiQ thermal cycler, as previously described (Kurniyati et al., 2017;Kurniyati et al., 2019). The molecular chaperone DnaK gene (dnaK, TDE0628) was used as an internal control to normalize the mRNA level of each individual genes. The results were expressed as mRNA level (ΔΔC T ) of the mid-phase relative to the late-log phase or stationary phase. Triplicates were included for each experiment. The data shown here is the average of three independent experiments. The primers for qRT-PCR are listed in Table 4.

F I G U R E 4
Characterizations of four flagellar filament gene deletion mutants. (a) Whole-cell lysate immunoblotting analysis of WT and the four flagellar filament gene deletion mutants. The blots were probed with antibodies against T. denticola DnaK (αDnaK), FlaA (αFlaA), and T. pallidum FlaB (αFlaB), respectively. DnaK was used as a loading control. (b) Swimming plate assay. This assay was carried out on 0.35% agarose plates containing the TYGVS medium diluted 1:1 with PBS. The plates were incubated anaerobically at 37°C for 3 days to allow the cells to swim out. Δtap1, a previously constructed non-motile mutant (Limberger et al., 1999), was used as a control to determine the initial inoculum sizes. The sizes of swimming rings from 10 different plates were measured and averaged. (c) Cell tracking analysis. T. denticola cells were tracked in the presence of 1% methylcellulose, as previously described (Kurniyati et al., 2017). The results are expressed as the mean of μm/s ± standard errors of mean (SEM). The data were analyzed by one-way ANOVA followed by Tukey's multiple comparison at p < .01.

| FlaA forms a "seam"-like sheath structure
It has been proposed that FlaB proteins form the core that is sheathed by FlaA (Charon et al., 2012;; however, there is lack of solid genetic and physical evidence to support this proposition. To fill this gap, we first measured the flagellar filament diameters in T. denticola cells using cryo-ET. Both WT and three flaB mutants have two types of PFs, referred to as "thick-PFs" and "thin-PFs". The thick-PFs are the predominant species and average 20 nm in diameter; the thin-PFs constitute a minority and average 13.5 nm in diameter (Table 2). By contrast, the ΔflaA mutant has only thin-PFs (13.9 ± 1.5 nm). We next isolated the PFs from WT and ΔflaA and examined their ultrastructure using cryo-EM.
As in the whole-cell cryo-ET analysis, the majority (>95%) of PFs in WT are thick (average 20 nm in diameter) (Figure 5a). By contrast, the PFs of ΔflaA are uniformly thin (average 13.5 nm in diameter) ( Figure 5b). Interestingly, the WT PFs have a rough surface which is covered by a "seam"-like structure (Figure 5a). In intact PFs, this "seam"-like structure starts from the filament tip, extends to the entire filament, and stops at the interface between the filament and hook (Figure 5c). By contrast, the PFs of ΔflaA are uniformly thin and smooth and have no such "seam"-like structures observed (Figure 5b,d). We also examined the PFs isolated from the three flaB mutants. They are morphologically indistinguishable from WT ( Figure S2). Based on these observations, we propose that FlaA, as a sheath protein, forms the "seam"-like structure around the core formed by three FlaB proteins.

| Characterizations of individual flaB mutants
In the above study, we measured the stoichiometry of four flagellar filament proteins in the WT PFs ( Figure 1). This assay is conducted as described in Figure 4b. ΔflaB2/B1 and ΔflaB2/B3 are two mutants in which the flaB2 gene was in-frame replaced with either flaB1 or flaB3. These two mutants were confirmed by PCR and western-blot analyses.  Similarly, T. denticola can also alter the level of individual flagellar filament proteins in response to environmental changes, for example, the levels of FlaA and FlaB2 increased considerably when T. denticola formed dual-species biofilms with the oral bacterial pathogen

TA B L E 3 The stoichiometry of four flagellar filament proteins in the PFs of WT and four mutants
Porphyromonas gingivalis (Zhu et al., 2013).

| Unique FlaA sheath proteins in spirochetes
It is of note that flagellar sheaths are also found in other Gramnegative bacteria, such as H. pylori, C. jejuni, and V. cholera (Chu et al., 2020;Lertsethtakarn et al., 2011); however, these sheaths are fundamentally different from the flagellar sheath of spirochetes.
Electron microscopy analyses uncovered that the flagellar sheaths in those non-spirochetal bacteria are an extension of the cell outer membrane (OM) that encases the flagella (Chu et al., 2020). The sheath found in these bacteria is loosely attached to the flagella and can be easily removed by treatment with detergents or even by physical methods such as washing (Sjoblad et al., 1983;Yang et al., 1977).
Little is known about the composition of these flagellar sheaths and their roles. Biochemical analyses of the flagellar sheaths from a few bacterial species suggest that they contain lipopolysaccharide, phospholipids, and OM proteins (Fuerst & Perry, 1988); however, these studies are inconclusive, and some are even controversial. Various functions for the flagellar sheath have been proposed, including preventing disassociation of flagellin subunits in the presence of gastric acid, avoiding activation of the host innate immune response by flagellin, adherence to host cells, and protecting the bacterium from bacteriophages (Carlsohn et al., 2006;Chu et al., 2020;Jones et al., 1997;Zhu et al., 2018). However, the experimental evidence that supports these speculations is limited.  (Lambert et al., 2012). Our previous studies indicated that deletion of flaA impairs flagellar the filament diameters, helicity, and motility of B. hyodysenteriae . Herein, a similar phenotype is observed in the flaA mutant of T. denticola (Figures 4-6). By contrast, deletion of flaA in B. burgdorferi has no obvious impact on flagellar morphology, assembly, or motility. Finally, emerging evidence suggests that FlaA proteins form a different structure. It has been proposed that FlaA proteins form a sheath around the filament core (Charon et al., 1992;Charon et al., 2012;. This model was initially built upon early immunoelectron microscopy studies using FlaA antibodies, whereby FlaA proteins are pinpointed at the surface of flagellar filaments, and later substantiated by our previous studies of using B. hyodysenteriae as a genetic model (Cockayne et al., 1987;Trueba et al., 1992). In these studies, FlaA was found to form a hollow tubular-like structure that confers extra thickness and strength upon the flagellar filament. The data shown in this report further corroborates this model. In addition, more structural details are revealed, for example, the sheath of FlaA is a "seam"-like structure that encases the filaments ( Figure 5). Interestingly, a recent structural study revealed that the FlaA proteins of Leptospira bind to FcpA and FcpB and form an asymmetric, lopsided architecture rather than the tubular-like sheath observed in B. hyodysenteriae and T. denticola (Gibson et al., 2020). B. burgdorferi has a minor FlaA protein (Ge et al., 1998;Ge & Charon, 1997). Our recent immuno-fluorescence microscopy study revealed that FlaA does not assembled around the filament core. Instead, it is located at the interface between the flagellar hook and filament (Li et al., unpublished data). Its role and structure in B. burgdorferi remain unknown. From these four exemplified spirochetes, we conclude that the role and structure of FlaA proteins vary from species to species despite their amino acid sequences being conserved among spirochetes.

| Overlapping role of three FlaB proteins in T. denticola
Except for B. burgdorferi, most spirochetes have multiple flagellin proteins that are highly conserved with respect to their amino acid sequences (Charon et al., 2012;Li et al., 2008;Norris et al., 1988;Wunder et al., 2016). For example, the three FlaB proteins of T. denticola share at least 70% sequence identity and 80% sequence similarity. A similar scenario stands for the FlaB proteins in other spirochetes. It remains unclear why multiple FlaB proteins exist or whether these flagellins adopt distinct structural or functional roles.
The results shown in this report provide some answers to these questions. First, the loss-of-function study reveals that even though each individual FlaB protein contributes to the motility of T. denticola, none is absolutely required, for example, the three flaB mutants are still able to assemble flagellar filaments (Figures 7, S2, and S3), and their motility is only partially impaired (Figure 4). In addition, our recent studies disclosed that double deletion mutants of flaB1flaB2, flaB1flaB3, and flaB2flaB3 still retain partial motility ( Figure S5).
Second, 2D gel analyses reveal that deletion of one flaB leads to augmentation of other flagellins in the assembled flagellar filaments ( Figure 7). Last, genetic studies indicate that substitution of flaB2 with either flaB1 or flaB3 has no obvious impact on the motility of T. denticola (Figures 7g and S4). In contrast to those spirochetes that have multiple flagellin isoforms, B. burgdorferi has only one FlaB protein, which is essential for flagellar filament assembly and motility; for example, deletion of flaB leads to a mutant that has no flagellar filament and is thus non-motile . Taken together, these findings indicate that the multiple FlaB isoforms are, at least in part, functionally interchangeable which makes T. denticola, and perhaps other spirochetes as well, more resilient to mutations.
As mentioned in our introduction, bacterial flagellar filaments are structurally conserved and typically composed of 11 protofilaments.
It is also possible that these multiple flagellin isoforms render spirochete flagellar filaments a different structure. We are currently attempting to delineate the structural details of spirochetal flagellar filaments using the mutants constructed in this report.
In summary, this report further highlights the complexity and uniqueness of spirochete flagellar filaments and provides several new perspectives to understanding of the regulatory mechanism of multiple flagellar filament proteinaceous units as well as their roles in spirochete flagellar filament assembly, structure, and locomotion.
These perspectives, alongside the materials (e.g., mutants and the PFs isolated from these mutants) and methodologies (e.g., whole-cell cryo-ET) generated in this report, pave the way for future study, for example, elucidating the structural details and assembly mechanism of flagellar filaments by using single-particle cryo-EM and cryo-ET.

| Bacterial strains, culture conditions, and oligonucleotide primers
T. denticola ATCC 35405 (wild-type) and mutant strains were grown in tryptone-yeast extract-gelatin-volatile fatty acids-serum (TYGVS) medium at 37°C in an anaerobic chamber in the presence of 85% nitrogen, 5% carbon dioxide, and 5% hydrogen, as previously documented (Kurniyati et al., 2017;Seshadri et al., 2004). T. denticola isogenic mutants were grown with appropriate antibiotic for selective pressure as needed: erythromycin (50 μg/ml) and gentamicin (20 μg/ ml). Escherichia coli DH5α strain (New England Biolabs, Ipswich, MA) was used for DNA cloning. The E. coli strains were cultivated in lysogeny broth (LB) supplemented with appropriate concentrations of antibiotics for selective pressure as needed: ampicillin (100 μg/ ml). The oligonucleotide primers for PCR amplifications used in this study are listed in Table 4. These primers were synthesized by IDT (Integrated DNA Technologies, Coralville, IA).

| Gel electrophoresis
Two-dimensional (2D) gel electrophoresis was carried out as previously described (Kurniyati et al., 2017). Equal amounts of purified PFs were resuspended in a rehydration buffer (5 M urea, 2 M thiourea, 2% CHAPS, 2% SB 3-10, 0.2% Bio-Lyte 3/10 Ampholyte, 40 mM Tris, and 0.0002% Bromophenol Blue) and then subjected to separation using 7 cm long pH 5 → 8 linear IPG strips. The first dimension of isoelectric focusing (IEF) was performed using PROTEAN IEF (Bio-Rad Laboratories, Hercules, CA), followed by equilibration according to the manufacturer's protocol. The second-dimension separation was carried out using sodium-dodecyl-sulfate polyacrylamide-gel electrophoresis (SDS-PAGE) as described previously (Kurniyati et al., 2017). The resultant gels were subjected to Coomassie blue staining or immunoblotting analyses. The antibodies against T. pallidum FlaBs and T. denticola FlaA and DnaK are described in our previous publications (Kurniyati et al., 2017;Kurniyati & Li, 2016). The stoichiometry of each flagellar filament protein was analyzed using Image Lab software from Bio-Rad (Bio-Rad).

| Mass spectrometry analysis of isolated PFs
Isolated PFs were treated with trypsin in an S-Trap micro spin column (in solution digestion) (Zougman et al., 2014) and the resultant samples were subjected to nano-LC-ESI-MS/MS analysis for protein identification which was carried out using an Orbitrap Fusion™ Tribrid™ (Thermo-Fisher Scientific, San Jose, CA) mass spectrometer equipped with a nano-spray Flex Ion Source, and coupled with a Dionex UltiMate 3000 RSLCnano system (Thermo, Sunnyvale, CA) as previously documented (Yang et al., 2018). The raw files with MS and MS/MS were subjected to database searches using Proteome Discoverer (PD) 2.5 software (Thermo Fisher Scientific, Bremen, Germany) with the Sequest HT algorithm. The PD 2.5 processing workflow containing an additional node of Minora Feature Detector for precursor ion-based quantification was used for protein identification and relative quantitation of identified peptides and their modified forms. The database search was conducted against a Treponema denticola database downloaded from NCBI.

| Measuring T. denticola growth rates
To measure the growth rates, 5 μl of the late-log phase T. denticola

| Constructions of T. denticola flagellin gene mutants
The flaA and three flaB genes were previously deleted through DNA allelic exchange (Kurniyati et al., 2017 genes were PCR amplified with primers P 39 /P 40 and P 41 /P 42 , respectively, and then cloned into the pGEM-T easy vector (Promega).
To replace flaB2 with flaB1 or flaB3, these two vectors were transformed into T. denticola wild-type competent cells via electroporation as previously described (Kurniyati & Li, 2021). The obtained two mutants, ΔflaB2/B1 and ΔflaB2/B3, were confirmed using PCR and immunoblotting.

| Bacterial swimming plate assay and motion tracking analysis
A swimming plate assay of T. denticola was performed as previously described (Kurniyati et al., 2013). In brief, 3 μl of cultures (10 9 cells/ml) were inoculated onto 0.35% agarose containing the TYGVS medium diluted 1:1 with PBS. The plates were incubated anaerobically at 37°C for 3-5 days to allow the cells to swim out.
The diameters of the swimming rings were measured in millimeters. As a negative control, a previously constructed T. denticola non-motile mutant, Δtap1, was included to determine the initial inoculum size (Limberger et al., 1999). The average diameters of each individual strains were calculated from three independent plates; and the results are represented as the mean of diameters ± standard error of the mean (SEM). The velocity of bacterial cells was measured using a computer-based bacterial tracking system, as previously described (Kurniyati et al., 2019). In brief, 100 μl of mid-logarithmic-phase T. denticola cultures was first diluted (

| Cryo-ET sample preparation, data collection, and image processing
The frozen-hydrated specimens of T. denticola were prepared as previously described (Kurniyati et al., 2017;Kurniyati et al., 2019). In brief, T. denticola cultures were mixed with 10 nm colloidal gold solutions and then deposited onto a freshly glow-discharged, holey carbon grid for about one minute. The grids were blotted with a small piece of filter paper for ~4 s and then rapidly plunged in liquid ethane using a gravity-driven plunger apparatus. The frozen-hydrated specimens of T. denticola were transferred to a 300 kV electron microscope (Krios, Thermo Fisher Scientific) or a 200 kV electron microscope (Glacios, Thermo Fisher Scientific) that is equipped with a field emission gun and a direct detection detector (K2, Gatan). To generate three-dimensional (3D) reconstructions of whole bacterial cells, tomographic package SerialEM was used to acquire multiple tilt series along the cell. The pixel size at the specimen level is 5.1 Å. All tilt series were collected in the low-dose mode with ~8 μm defocus.

| Statistics analyses
For quantitative experiments (e.g., swimming plate assay, tracking analysis, and measurement of PFs length, diameter, and helicity), multiple samples were included and at least three independent experiments were conducted. The results are expressed as mean ± standard errors of mean (SEM). Statistical significance was analyzed by one-way ANOVA followed by Tukey's multiple comparison at p < .01.

AUTH O R CO NTR I B UTI O N S
KK and YC performed the experiments and analyzed results. JL and CL designed the study and wrote the manuscript. All authors read and approved the manuscript.

ACK N OWLED G M ENTS
We thank Jennifer Aronson for critical reading of the manuscript.
We thank Meng Shao for cryo-ET data segmentation, Shenping Wu and Chunyan Wang for assisting with cryo-ET data collection. This research was supported by funding from the National Institute of Dental and Craniofacial Research (DE023080 to C.

Li) and National Institutes of Allergy and Infectious Diseases
(AI078958 to C. Li, AI087946 to J. Liu, and AI 148844 to B. Crane and C. Li), National Institutes of Health (NIH). Cryo-ET data were collected at the Yale Cryo-EM resources. Mass spectrometry analysis was performed at the Proteomics and Metabolomics Facility of Cornell University.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.

E TH I C S S TATEM ENT
All animal experimentation was carried out in strict accordance with the recommendations in the Guide for the Care and Use of