Analysis of a flagellar filament cap mutant reveals that HtrA serine protease degrades unfolded flagellin protein in the periplasm of Borrelia burgdorferi

Abstract

Unlike external flagellated bacteria, spirochetes have periplasmic flagella (PF). Very little is known about how PF are assembled within the periplasm of spirochaetal cells. Herein, we report that FliD (BB0149), a flagellar cap protein (also named hook-associated protein 2), controls flagellin stability and flagellar filament assembly in the Lyme disease spirochete Borrelia burgdorferi. Deletion of fliD leads to non-motile mutant cells that are unable to assemble flagellar filaments and pentagon-shaped caps (10 nm in diameter, 12 nm in length). Interestingly, FlaB, a major flagellin protein of B. burgdorferi, is degraded in the fliD mutant but not in other flagella-deficient mutants (i.e., in the hook, rod, or MS-ring). Biochemical and genetic studies reveal that HtrA, a serine protease of B. burgdorferi, controls FlaB turnover. Specifically, HtrA degrades unfolded but not polymerized FlaB, and deletion of htrA increases the level of FlaB in the fliD mutant. Collectively, we propose that the flagellar cap protein FliD promotes flagellin polymerization and filament growth in the periplasm. Deletion of fliD abolishes this process, which leads to leakage of unfolded FlaB proteins into the periplasm where they are degraded by HtrA, a protease that prevents accumulation of toxic products in the periplasm.

Graphical Abstract

The flagellar cap protein FliD forms a pentagon-shaped structure (10 nm in diameter, 12 nm in length), promoting flagellin (FlaB) polymerization and flagellar filament formation in the periplasm of Borrelia burgdorferi. Deletion of fliD leads to hypersecretion of FlaB into the periplasm in which unfolded FlaBs are degraded by a serine protease HtrA.

Introduction

The flagellum is a sophisticated nano-machine that propels bacteria to swim toward a favorable environment and away from a toxic one; thus, flagellum-driven motility plays a critical role in bacterial pathophysiology [for review, see references (Chevance & Hughes, 2008, Mukherjee & Kearns, 2014)]. The assembly and structure of the flagellum have been well studied in two prototypical bacteria, Escherichia coli and Salmonella typhimurium. The flagellum is composed of at least 25 different proteins in copy numbers ranging from a few to more than a thousand. Those proteins form three mechanical parts: the basal body, the hook, and the filament. The basal body is imbedded within the cell envelope and works as a reversible rotary motor. The flagellar hook and filament extend outwards to the cell exterior and function as a universal joint and a propeller, respectively. The basal body is complex and consists of several functional units: the MS/C ring (rotor), the rod (driveshaft), the L/P rings (bushings), the stator (torque generator), and the flagellar export apparatus (also known as flagella-specific Type III secretion system, fT3SS) [for review, see references (Aldridge & Hughes, 2002, Paul et al., 2008)]. The motor is powered by an inward-directed electrochemical gradient of protons or sodium ions (Meister et al., 1987, Atsumi et al., 1992). The torque generated by the motor is mechanically transmitted to the filament via the rod-hook complex, leading to the rotation of flagellar filament, which confers locomotion to bacterial cells.

Flagellar assembly is a finely orchestrated sequential process [for review, see references (Aldridge & Hughes, 2002, Chevance & Hughes, 2008)]. In E. coli, the flagellar assembly begins with the MS ring assembly. Built onto the MS ring is a hollow rod that spans the periplasmic space. After formation of the MS ring/rod complex, the FlgI and FlgH proteins assemble around the rod, forming the P and L rings, respectively. The flagellar rod begins with the MS ring and stops at the hook. The hook (FlgE) proteins are secreted through the fT3SS and assembled on the rod, forming a tubular helical structure (average 55 nm in length) that connects the basal body to the filament (Samatey et al., 2004). Once the hook assembly is completed, FliD (also known as hook-associated protein 2, HAP2) forms a cap on the distal end of the hook. Then, the flagellin proteins are secreted via the fT3SS and polymerized below the cap one after another, forming long helical filaments that are composed of more than a thousand of copies of a single flagellin protein, FliC (Ikeda et al., 1987, Maki-Yonekura et al., 2003). During filament elongation, the cap stays attached at the distal end, promoting flagellin polymerization and preventing flagellin from leaking out (Yonekura et al., 2000). Deletion of fliD genes leads to mutant cells that are non-motile and lack flagellar filaments, highlighting its essential role in bacterial motility and flagellar filament assembly (Yokoseki et al., 1995, Ikeda et al., 1993, Inaba et al., 2013).

Spirochetes are a group of bacteria responsible for several human diseases, including Lyme disease (Borrelia burgdorferi), syphilis (Treponema pallidum), and leptospirosis (Leptospira interrogans) [for review, see references (Rosa et al., 2005, Radolf et al., 2016, Picardeau, 2017)]. Spirochetes can be readily recognized by their flat-waved or coiled cell morphology and distinct form of corkscrew-like motility (Charon et al., 2012, Charon & Goldstein, 2002). It is well recognized that motility plays a critical role in the pathogenicity of spirochetes, including colonization, dissemination, and immune evasion, as well as in bacterial transmission among different hosts (Motaleb et al., 2015, Sultan et al., 2015, Sultan et al., 2013, Li et al., 2010, Harman et al., 2012, Norman et al., 2008). Unlike external flagellated bacteria, spirochetes swim by means of rotating two bundles of periplasmic flagella (PF) that reside between the outer membrane and cell cylinder (Li et al., 2000b, Charon et al., 2012). In general, spirochetal PF are structurally similar to the flagella of other bacteria, as each consists of a basal body-motor complex, a hook, and a filament. The filament is the largest component of the PF (Zhao et al., 2013, Liu et al., 2009, Kudryashev et al., 2009). Compared to E. coli and S. typhimurium, the structure and protein composition of flagellar filaments are more complex in spirochetes (Li et al., 2000b, Charon et al., 2012). In most spirochetes such as Brachyspira and Treponema species, the filaments comprise at least one sheath protein FlaA and multiple flagellin proteins (i.e., FlaB1, FlaB2, and FlaB3) (Kurniyati et al., 2017, Li et al., 2008, Li et al., 2000a). FlaBs are homologs of FliC and form a filament core that is sheathed by FlaA.

B. burgdorferi is the causative agent of Lyme disease, which is the most prevalent tick-borne disease in the United States (Sanchez et al., 2016, Kuehn, 2013). B. burgdorferi is relatively long (10 to 20 µm) and thin (0.3 µm) and has a flat-wave shape (Hovind-Hougen, 1984, Goldstein et al., 1994, Kudryashev et al., 2009). Approximately 7 to 11 PF are subterminally attached at each of the cell poles, forming a tight-fitting ribbon that wraps around the cell cylinder (Charon et al., 2009, Liu et al., 2009, Sze et al., 2011). The flagellar filament of B. burgdorferi is composed of a major flagellin protein, FlaB, and a minor sheath protein, FlaA (Ge et al., 1998, Fraser et al., 1997). Deletion of flaB leads to mutant cells that are rod-shaped and lack flagellar filaments (Motaleb et al., 2000). B. burgdorferi motility-defective mutants fail to establish mammalian infection, highlighting an important role of motility in the pathogenicity of spirochetes (Motaleb et al., 2015, Charon et al., 2012, Wolgemuth, 2015). Due to its medical importance and genetic tractability, B. burgdorferi has emerged as a model system to study the genetic regulation, structure and assembly of PF, as well as chemotaxis in spirochetes (Charon et al., 2012, Zhao et al., 2014, Brisson et al., 2012). Though significant progress has been achieved, our understanding of PF structure and assembly remains limited. Two intriguing questions are how multiple flagellin proteins are assembled and how this assembly process is regulated in the periplasm of spirochetes. In this report, we investigated the role of FliD in the assembly of PF by using an approach consisting of bioinformatics, genetics, biochemistry, and cryo-electron tomography (cryo-ET). The results indicate that the cap protein FliD is essential for the flagellar filament polymerization of B. burgdorferi and abrogation of the cap leads to leakage of unfolded flagellin proteins into the periplasm where they are degraded by a serine protease HtrA.

Results

Transcriptional analyses of fliD gene.

The fliD gene (gene locus, bb0149) resides upstream of flaB (gene locus, bb0147), and the two genes are adjacent and co-oriented (Fraser et al., 1997). To determine if they are co-transcribed, a previously described co-RT-PCR was conducted (Ge et al., 1997). Three pairs of primers spanning nagA-fliD, fliD-flaB, and bb0146-bb0145 were used for the RT-PCR analysis, as illustrated in Fig. 1A. As expected, no product was detected between nagA and fliD, as they are divergently transcribed. There was no product detected between fliD and flaB (Fig. 1B), indicating that the fliD gene is monocistronic and not co-transcribed with its downstream flaB gene. There is a 191 bp intergenic region upstream of fliD. A previous in vivo transcription analysis mapped a transcription start site (TSS) at the position of 150,792 bp on the chromosome (Adams et al., 2017), which is 9 nt from the start codon of fliD (Fig. 1C). To understand the genetic regulation of fliD, we searched upstream of the TSS for consensus sequences of different promoters (i.e. sigma⁷⁰ and sigma⁵⁴) and transcriptional regulators (i.e., RpoS and BosR) that have been identified in B. burgdorferi (Brisson et al., 2012, Radolf et al., 2012, Charon et al., 2012). A promoter-like sequence was identified: its −12 region having 100% identity to the consensus sequence (TATAAT) of the E. coli sigma⁷⁰ promoter and the −35 region (TTGTTT) being less similar (Fig. 1D). The identified promoter was designated as pfliD. To determine if pfliD is a functional promoter in vivo, a transcriptional reporter assay using lacZ was conducted. For this assay, a previously identified flaB promoter (pflaB) was included as a positive control (Sze et al., 2011). As shown in Fig. 1E, the average β-galactosidase activity for pfliD is 184.0 ± 2.3 Miller Units, which is slightly weaker than that of pflaB (299.7± 3.9 Miller units), indicating that pfliD was able to initiate reporter gene expression in E. coli. A reporter assay using green fluorescence protein (GFP) showed that pfliD is also functional in B. burgdorferi (Fig. 1F, G).

FIG. 1. Transcriptional analyses of fliD_Bb gene.

(A) Genes adjacent to fliD_Bb. Black arrows represent three pairs of primers for co-RT-PCR. Vertical arrows denote promoters of flaB (pflaB) and fliD_Bb (pfliD). (B) fliD_Bb is monocistronic. DNA gel electrophoresis analysis of RT-PCR (top panel) and PCR (bottom panel) products. The lane numbers correspond to the primers labeled in panel A. (C) Upstream DNA sequence of fliD_Bb. Underlined sequences are the predicted −12 and −35 regions of pfliD. Asterisk denotes the previously mapped transcription start site (TSS) (Adams et al., 2017). (D) Comparison of pfliD with the consensus sequences of σ⁷⁰ promoters identified in E. coli and B. burgdorferi. Asterisk represents TSS. (E) Transcriptional analysis of pfliD using lacZ as a reporter. For this assay, pfliD or pflaB was fused to the promoterless lacZ gene in the pRS414 plasmid. The empty pRS414 was used as a negative control, and the plasmid containing pflaB used as a positive control. β-galactosidase activity was measured and expressed as the average Miller units of triplicate samples from two independent experiments, as previously described (Bian et al., 2011). (F) & (G) GFP reporter assay. For this assay, pfliD-gfp (F) and gfp alone (G) were cloned into pJSB275, respectively. The resultant vectors were transformed into B31A strain. The expression of GFP was visualized with fluorescence microcopy (Axiostar Plus, Ziess). All of the primers used here are listed in Table S1.

Homology modeling analyses reveal an extra variable region in the D3 domain of FliD.

As a flagellar cap protein, the function and structure of FliD proteins have been well studied in E. coli and S. typhimurium (ecFliD and stFliD, respectively) (Yonekura et al., 2000, Song et al., 2017, Postel et al., 2016). Both ecFliD and stFliD are composed of 468 amino acids (aa) and harbor three functional domains (D1, D2, and D3). FliD proteins self-polymerize to a stool-like structure in which the D2 and D3 domains form the cap plate and the D1 domain resembles the leg. Similar to bacterial flagellin proteins, the cap proteins are also highly conserved. B. burgdorferi FliD (hereafter named FliD_Bb) consists of 665 aa with a predicted molecular weight (MW) of 74.73 kDa and is 198 aa longer than ecFliD and stFliD. Sequence alignment revealed that FliD_Bb harbors a highly conserved D1, moderately conserved D2 and D3 domains, and an extra variable region (VR, 185 aa) inserted in D3 (Fig. 2A, Figs. S1 & S2). Despite low sequence identity (16%) between ecFliD and FliD_Bb, a homology structural model that had ~72% coverage (480 out of 665 residues) was established by using the crystal structure of ecFliD (5H5V) as a template (Song et al., 2017). The results revealed that FliD_Bb has a similar topology to ecFliD, consisting of a highly conserved D1 domain that is primarily composed of α-helices and less conserved D2 and D3 domains that primarily comprise β-strands (Fig. 2B). The 185 aa VR forms a large disordered loop (dotted lines, Fig. 2B) at the interface between the D2 and D3 domains. Homology modeling using the ecFliD homo-hexamer as a template suggests that FliD_Bb forms a stool-like structure that consists of a leg and a cap plate whereby the disordered VR loop is distributed along the peripheral region of the cap plate (Fig. 2C).

FIG. 2. Domain organization and overall structure of FliD_Bb.

(A) The domain organization of FliD_Bb. Domains were determined by multiple sequence alignments of FliD proteins (Fig. S2), including E. coli and S. typhymurium FliD (ecFliD and stFliD). (B) Monomeric structure of FliD_Bb. An all-atom model of FliD_Bb was generated using the crystal structure of ecFliD (PDB ID 5H5V) (Song et al., 2017) as the template. Despite the low sequence homology (16% identity), the model has ~72% coverage (480 of 665 aa). Due to a lack of sequence similarity, the variable region (VR) of 185 residues (dotted lines) was unable to be modeled. (C) Hexamer structure of FliD_Bb. The homology model was generated using the hexamer of ecFliD as the template. The domains that resemble the cap plate and leg of FliD are labeled.

Deletion of fliD_Bb leads to mutant cells that are rod-shaped and non-motile.

The function of FliD remains unknown in spirochetes. To understand its role, the fliD_Bb gene (bb0149) was inactivated by targeted mutagenesis as illustrated in Fig. 3A. Kanamycin-resistant colonies that appeared on the agar plates after 15 days of incubation were inoculated into BSK-II media. PCR analysis showed that all of the examined colonies had kan inserted into the fliD_Bb gene as expected (data not shown). One of those colonies (designated as ∆fliD) was further confirmed by immunoblotting using a specific antibody against FliD_Bb. As shown in Fig. 3C, a band near 75 kDa was detected in the wild type but not in ∆fliD, indicating the cognate gene product was abolished. To cis-complement this mutant, a full-length fliD_Bb gene was first fused to P_flgB, a previously identified promoter, and then inserted into an intergenic region between bb0445 and bb0446 on the chromosome (Fig. 3B). The complementation successfully restored the expression of FliD_Bb in the mutant (Fig. 3C). ∆fliD and one of its complemented clones (designated as ∆fliD^com) were selected for further characterizations as described below.

FIG. 3. Construction of a fliD_Bb deletion mutant and its complemented strain.

(A) Construction of fliD::kan for targeted mutagenesis of fliD_Bb. To construct this vector, a kanamycin-resistance marker (kan) was inserted into fliD_Bb at a HindIII cut site. (B) Construction of cis-complementation vector of fliD/cisCom by inserting fliD_Bb into an intergenic region between bb0445 and bb0446 on the chromosome of B. burgdorferi, as previously reported (Sze et al., 2013, Li et al., 2007). (C) Characterization of WT, a fliD_Bb mutant (∆fliD), and its complemented strain (∆fliD^com) by immunoblotting with a specific antibody against FliD (αFliD). DnaK was used as an internal control, as previously described (Sze et al., 2013).

We first examined the impact of FliD_Bb on mutant cell motility and morphology. Under the dark-field microscope, while the wild-type B31A (WT) and complemented ∆fliD^com strains were fully motile (Videos 1 & 2), the ∆fliD mutant was non-motile and unable to migrate in the growth medium containing 1% methylcellulose (Video 3). Swimming plate assays further showed that ∆fliD failed to swarm on soft agarose plates (Fig. 4B), whereas WT and ∆fliD^com spread quite readily, which was evident by the formation of large swimming rings (Fig. 4A, C). These results indicate that deletion of fliD_Bb results in mutant cells that are non-motile. We also found that the ∆fliD mutant cells are rod-shaped (rather than wave-shaped like both WT and ∆fliD^com), and mutant cells grew as long chains with incomplete division (Fig. 4F). Collectively, these results demonstrate that the phenotype of ∆fliD is similar to previously reported B. burgdorferi flagella-deficient mutants such as the mutants of flaB and flgE (Motaleb et al., 2000, Sal et al., 2008), indicating that FliD_Bb is essential for the motility and probably for the formation of flagella in B. burgdorferi.

FIG. 4. Deletion of fliD_Bb leads to mutant cells that are non-motile and rod-shaped.

(A-D) Swimming plate assays. These assays were conducted by inoculating 5 μl of B. burgdorferi cells into 0.35% agar plates, as previously described (Li et al., 2002). ΔflaB, a previously constructed non-motile mutant (Motaleb et al., 2000), was used as a control to determine the initial inoculum size. At least four plates were included for each strain and the data are expressed as averaged diameters of the rings ± standard deviations (n=4): WT (12 ± 0.4 mm), ΔfliD (7.25 ± 0.6 mm), ΔfliD^com (11.75 ± 0.3 mm), and ΔflaB (7.5 ± 0.5 mm). (E-F) Dark-field microscopy analysis of WT and ∆fliD.

∆fliD fails to assemble the flagellar filaments.

FlaB is a major flagellin protein of B. burgdorferi (Motaleb et al., 2000, Motaleb et al., 2004). Deletion of flaB leads to mutant cells that are non-motile, rod-shaped, and lack flagellar filaments (Motaleb et al., 2000). The phenotype of ∆fliD is very similar to that of ∆flaB, a previously characterized flaB-deletion mutant (Motaleb et al., 2000), suggesting that deletion of fliD_Bb might abolish the formation of flagellar filaments. To ascertain if this is the case, cryo-electron tomography (cryo-ET) was used to directly visualize the PF in situ. To obtain more detailed structural information, we also included ∆flaB in the cryo-ET analysis. Previous studies have shown that B. burgdorferi has ~7–11 long helical PF that reside at the cell poles, extend towards the center of the cells, and form a tight-fitting ribbon that wraps around the cell cylinder (Sze et al., 2011, Liu et al., 2009). However, in both ∆fliD and the ∆flaB mutant cells, only short curved and disoriented flagellar hooks (9 ± 2 PF per cell) rather than long helical PF were detected by cryo-ET (Fig. 5, Fig.S4). These observations indicate that deletion of fliD_Bb results in a mutant that lacks flagellar filaments, even though the flagellar number remains unchanged. In addition, cryo-ET and sub-tomogram averaging detected a cap-like structure at the distal end of the flagellar hooks in ∆flaB (Fig. 5B, D) but not in ∆fliD (Fig. 5A, C). Due to the lack of these cap-like structures, the flagellar hooks of ∆fliD (average length, 51.85 ± 3.06 nm, n=10) are approximately 11.20 nm shorter than those in ∆flaB (average length, 63.05 ± 1.96 nm, n=10), suggesting that the cap formed by FliD_Bb is approximately 11.20 nm in length. Collectively, these results demonstrate that FliD_Bb attaches to the distal end of flagellar hooks, forming a cap-like structure that is required for the formation of flagellar filaments in B. burgdorferi.

FIG. 5. Visualization of ∆fliD and ΔflaB mutant cells by cryo-ET.

(A, B) Representative tomograms of ∆fliD and ΔflaB mutant cells. The insertion images show enlarged views of the flagellar hook structure and length. (C, D) Three-dimensional surface views of the cells shown in (A) and (B). The insertion images show the enlarged views of the hooks and hook-associated caps.

FliD_Bb forms a pentagon-shaped cap.

The structure of FliD was initially established by visualization of self-assembled ecFliD recombinant proteins using electron microscopy along with single-particle image analysis (Yonekura et al., 2000). Here, we were able to solve the structures of the hook tip in the ΔflaB and ∆fliD mutants using cryo-ET and sub-tomogram averaging of frozen hydrated samples in situ. The hook structure of the ΔflaB mutant was averaged from 1,492 sub-tomograms out of 211 tomograms, and the hook structure of ∆fliD mutant was averaged from 599 sub-tomograms from 83 tomograms. Similar to other bacteria, the flagellar hooks of B. burgdorferi are twisted tubular structures (Fig. 6). The hook region of both mutants showed a helical symmetry of approximately 11 subunits per 2 turns (Fig. 6A, E). The diameter of the hook is about 16 nm. Compared to the structure from the ∆fliD mutant (Fig. 6E), the one from the ΔflaB mutant has an extra density on top of the hook (Fig. 6A), which resembles the FliD cap. The dimension of the cap is about 10 nm in diameter and 12 nm in length. The cap structure is asymmetric but close to 5-fold symmetry, as the cross-section view shows 5 subunits (Fig. 6C). There is symmetry mismatching from the hook to the cap, and the entire structure is asymmetric. Interestingly, an extra density was observed near the hook tips of ∆fliD (Fig. 6E, G), which might be substrates (i.e., FlaB) or hook-associated proteins such as FlgK and FlgL leaking from the hook channel due to the lack of cap protein.

FIG. 6. Three-dimensional averaged structures of ΔflaB and ∆fliD show difference in the FliD cap region.

(A) A central section of the averaged structure from the ΔflaB mutant. The averaged structure was generated from 1492 sub-tomograms. (B) A cross section on the hook region from (A) shows six subunits with a diameter of 16 nm. (C) A cross section on the cap region from (A) shows five subunits with a diameter of 10 nm. (D) Two different views of solid-surface rendering map on the structure in (A) show the symmetry mismatching between hook and cap. The hook region has helical packing at about 11 subunits per two turns. The cap region has an asymmetric structure that is close to 5-fold symmetry. The top part of the cap that interacts with the hook is not symmetric. (E) A central section of the averaged structure from ∆fliD. The averaged structure was generated from 599 sub-tomograms. (F) A cross section on the hook region from (E) also shows six subunits with a diameter of 16 nm. (G) Two different views of the surface rendering of the hook structure from (A). It has the same helical symmetry as the hook region in (D).

Decrease of flagellar filament proteins in ∆fliD.

Similar to flagella in other bacteria, the PF of B. burgdorferi consist of three parts: the basal body, the hook, and the filament (Liu et al., 2009, Zhao et al., 2013). Cryo-ET analysis revealed that deletion of fliD_Bb only impaired the filament but not the other parts of the PF. To confirm this, we conducted quantitative immunoblotting analyses to measure the levels of the five flagellar proteins that make up the three parts of the PF – the filament (FlaA and FlaB), the hook (FlgE), the rod (FlgG), and the MS-ring (FliF). The immunoblotting results showed that while the levels of FlgE, FlgG, and FliF were not affected in ∆fliD, the levels of the two filament proteins were substantially decreased – no FlaA was detected and only a trace of FlaB was present (Fig. 7A). qRT-PCR analysis further revealed that the level of flaB transcripts in the ∆fliD mutant is similar to that in WT (Fig. 7B), suggesting that the reduction of FlaB in the mutant occurs at the post-transcriptional level such as protein turnover, as previously reported in B. burgdorferi flagella-deficient mutants (Sal et al., 2008). In support of this possibility, protein turnover assays showed that 12 hrs after addition of spectinomycin, more than 80% of FlaB was degraded in the mutant (Fig. 7C). In contrast, the level of FlaB remained unchanged in WT after addition of spectinomycin (Fig. 7D). Taken together, these results indicate that FliD_Bb is essential for the filament protein accumulation and flagellar filament formation in B. burgdorferi.

FIG. 7. Deletion of fliD_Bb impairs the level of flagellar filament proteins.

(A) Immunoblotting analyses to measure the levels of five flagellar proteins that represent three parts of the PF –the filament (FlaA and FlaB), the hook (FlgE), and the rod (FlgG) – as well as the MS-ring (FliF). For this experiment, similar amounts of whole cell lysates of WT, ∆fliD, and ∆fliD^com were analyzed by SDS-PAGE and then probed against different antibodies as indicated. α- Abbreviation of antibody. DnaK was used as a loading control. Immunoblots were developed using horseradish peroxidase secondary antibody with an ECL luminol assay, as previously described (Sze et al., 2013). (B) Detection of flaB transcripts by qRT-PCR. This experiment was conducted, as previously described (Sze et al., 2013). The transcript of the enolase gene (eno) was used as an internal control to normalize the qPCR data. The results were expressed as the mean threshold cycles (CT) of triplicate samples. (C, D) Measuring FlaB stability by protein turnover assays. To arrest protein translation, 100 µg/ml spectinomycin was added to the WT and ∆fliD mutant cultures. Samples were collected at the indicated time points for immunoblotting analyses. DnaK was used as a sample loading control. The levels of FlaB were measured by densitometry and expressed as the percent protein density at zero time compared to subsequent time points.

FlaB is degraded in ∆fliD but not in other flagella-deficient mutants.

The above results showed that FlaB is turned over in the ∆fliD mutant. To investigate the mechanism by which FlaB is degraded, we generated three flagella-deficient mutants by disrupting the MS-ring (∆fliF), the rod (∆flgG), and the hook (∆flgE) (Fig. 8A). We then measured the level of FlaB in these mutants by quantitative immuno-blotting analyses. The results showed that the level of FlaB in the ∆fliF, ∆flgG, and ∆flgE mutants was less than that in WT but significantly higher than that in the ∆fliD mutant (Fig. 8B). Follow-up protein turnover assays revealed that FlaB was turned over only in ∆fliD but not in the other three mutants (Fig. 8C). Collectively, these results indicate that as a flagellar cap protein, FliD_Bb is essential for maintaining the stability of FlaB. Furthermore, a lack of FliD_Bb leads to FlaB turnover, which in turn cripples flagellar filament formation.

FIG. 8. Detection of FlaB in different flagellar mutants by immunoblotting.

(A) Overall structure of PF. Red arrows denote five flagellar genes that were inactivated by targeted mutagenesis. (B) Detection of FlaB in WT and five flagellar mutants by immunoblotting. Similar amounts of whole cell lysates were subjected to SDS-PAGE, followed by immunoblotting using antibodies against FlaB (αFlaB) and DnaK (αDnaK). DnaK was used as an internal control. (C) Measuring the stability of FlaB in different flagellar mutants by protein turnover assays.

Deletion of HtrA_Bb increases the level of FlaB in ∆fliD.

Recent studies have shown that HtrA_Bb (BB0104) functions as a serine protease that plays an important role in the pathophysiology of B. burgdorferi (Coleman et al., 2013, Russell et al., 2013, Coleman et al., 2018, Russell & Johnson, 2013, Gherardini, 2013). HtrA_Bb can be detected in the cytoplasm and periplasm, as well as in the outer membrane vesicles (OMV) of B. burgdorferi. The PF are assembled in the periplasm of B. burgdorferi and the role of FliD_Bb is to promote flagellin polymerization and prevent the accumulation of unassembled flagellin in the periplasm. Thus, deletion of fliD_Bb could lead to FlaB monomers leaking into the periplasm where they would be degraded by HtrA_Bb. To test this hypothesis, we constructed a double mutant of fliD_Bb-htrA_Bb (designated as ∆fliD-htrA). Immunoblotting analysis showed that HtrA_Bb was detected in the WT and ∆fliD mutant but not in ∆fliD-htrA and ∆htrA, a single-deletion mutant of htrA_Bb (Fig. 9A). Next, we examined the level of FlaB in ∆fliD-htrA by immunoblotting analysis and found that the level of FlaB in the double mutant was significantly increased compared to ∆fliD (Fig. 9B), suggesting that HtrA_Bb is responsible for, at least in part, the degradation of FlaB in the ∆fliD mutant.

FIG. 9. Deletion of htrA_Bb or csrA_Bb partially restores the level of FlaB in the ∆fliD mutant.

(A) Characterization of WT, ∆fliD, ∆htrA (a single-deletion mutant of htrA_Bb, kanamycin resistance), and ∆fliD-htrA (a double mutant of fliD_Bb and htrA_Bb). (B) Detection of FlaB in ∆fliD, ∆fliD-htrA, and ∆fliD-csrA (a double mutant of fliD_Bb-csrA_Bb) by immunoblotting. DnaK was used as a sample loading control.

HtrA_Bb cleaves FlaB monomers but not polymerized FlaB.

Deletion of htrA_Bb protects FlaB from degradation, suggesting that HtrA_Bb is responsible for the FlaB turnover in ∆fliD. To verify this, we carried out in vitro proteolytic assays to determine if HtrA_Bb directly cleaves FlaB. For this assay, we prepared recombinant HtrA_Bb (rHtrA_Bb) and a mutated HtrA_Bb (rHtrA_Bb^S226A) in which Ser226, a residue essential for the activity of HtrA_Bb (Coleman et al., 2013), was replaced with Ala. The flagellar filaments of B. burgdorferi are composed of a minor sheath protein FlaA and a major filament core protein FlaB (Ge et al., 1998). To obtain polymerized FlaB (pFlaB), we isolated flagellar filaments from a mutant of flaA, as previously described (Miller et al., 2014). SDS-PAGE showed that the isolated filaments contained only FlaB (Fig. 10, lane 2). In addition, we boiled the isolated filaments to produce denatured FlaB (dFlaB). Blue native PAGE (BN-PAGE) showed that pFlaB is polymeric and dFlaB is monomeric (data not shown). As shown in Fig. 10, while pFlaB remained intact (lane 3), dFlaB was completely degraded by rHtrA_Bb (lane 5) after a 12-hour incubation but not by rHtrA_Bb^S226A (lane 8). These observations indicate that dFlaB is specifically degraded by HtrA_Bb, rather than by other factors such as self-autolysis. Collectively, these results demonstrate that HtrA_Bb cleaves FlaB monomers but not polymerized FlaB.

FIG. 10. Proteolytic assay of HtrA_Bb against FlaB proteins.

2 μg of rHtrA_Bb (wild type recombinant HtrA_Bb) or rHtrA_Bb^S226A (a site-directed mutant of rHtrA_Bb in which Ser226 was replaced by Ala) was incubated with equal amounts of polymerized FlaB (pFlaB) or denatured FlaB (dFlaB) for 12 hrs. The resultant samples were subjected to SDS-PAGE, followed by Coomassie blue staining. pFlaB, flagellar filaments isolated from ΔflaA; dFlaB, heat-inactivated pFlaB. The flagellar filaments of B. burgdorferi are composed of a major flagellin protein, FlaB, and a minor sheath protein, FlaA. The filaments isolated from the ΔflaA mutant only contain FlaB, as shown in lanes 2, 4 and 7. Some minor bands were observed below rHtrA_Bb (lanes 1, 3, and 5) but not rHtrA_Bb^S226A (lanes 6 and 8), indicative of autolysis of rHtrA_Bb, as previously reported (Coleman et al., 2013).

Discussion

FliD is universal in all flagellated bacteria, yet its role in flagellation and motility has been investigated only in S. typhimurium, Clostridium difficile, Helicobacter pylori, Pseudomonas aeruginosa, and B. subtilis (Yokoseki et al., 1995, Arora et al., 1998, Kim et al., 1999, Mukherjee et al., 2013, Tasteyre et al., 2001). In these bacteria, deletion of fliD genes abolishes flagellin protein polymerization and filament growth, which leads to flagellin monomers leaking into the extracellular environments such as growth media. Due to a lack of filaments, fliD mutants of these bacteria are non-motile. Additionally, deletion of fliD genes also affects bacterial pathogenicity. For instance, a fliD mutant of H. pylori is unable to colonize mouse gastric mucosae (Kim et al., 1999), and a fliD mutant of P. aeruginosa is non-adhesive to mucin (Arora et al., 1998). In this report, we found that deletion of fliD_Bb leads to mutant cells that are non-motile (Fig. 4A) and deficient in flagellar filament formation (Fig. 5), which is similar to previously reported fliD mutants and highlights a critical role of FliD_Bb in flagellin polymerization and filament assembly of B. burgdorferi. However, in contrast to those externally flagellated bacteria, B. burgdorferi has flagella that are assembled and function within the periplasmic space rather than on cell surfaces. Thus, deletion of fliD_Bb results in FlaB monomers leaking into the periplasm, which raises several intriguing questions as below discussed.

How does B. burgdorferi protect nascent flagella from proteolytic enzymes in the periplasm?

Bacterial periplasm contains an array of proteolytic enzymes that control protein quality and prevent toxic molecule accumulation by removing misfolded and mislocalized proteins (Merdanovic et al., 2011, Backert et al., 2018). For instance, there are at least 19 proteases (9 serine proteases, 6 metalloproteases, and 4 cysteine proteases) in the periplasm of E. coli. Among these proteases, the biochemical and structural features of DegP, Tsp, and PtrA have been well characterized [for review, see (Backert et al., 2018)]. The number of proteases in the periplasm of B. burgdorferi remains unknown. We searched the genome of B. burgdorferi for known proteases. In addition to HtrA_Bb, we found 7 additional putative proteases – two Lon proteases (BB0253 and BB0613), two Clp proteases (BB0611 and BB0757), one zinc metalloprotease (BB0118), and two unclassified proteases (BB0359, a homolog of Tsp, also named peptidase; BB0536, a homolog of PtrA, also named insulinase) (Fraser et al., 1997). HtrA_Bb is a homolog of DegP and can be detected in the periplasm of B. burgdorferi (Coleman et al., 2013, Russell et al., 2013). Though the cellular localization of the other six proteases remains unknown, it is very likely that some of them reside in the periplasm. Cryo-ET analysis of the ∆flaB mutant showed that FliD_Bb forms an asymmetric pentagon-shaped cap at the hook tip in the periplasm (Figs. 5 & 6). The cap stays attached at the distal end during filament growth (Maki-Yonekura et al., 2003, Yonekura et al., 2000). Thus, it must remain stable until the filament assembly is completed. During this process, how does B. burgdorferi protect FliD_Bb from proteolytic enzymes while promoting flagellin polymerization and filament growth in the periplasm? FliD_Bb is much larger than its counterparts from other bacteria. In addition to the conserved D1, D2 and D3 domains, it also harbors a 185 aa VR insert that forms a disordered loop along the peripheral region of the cap plate (Fig. 2). We speculate that this loop may protect the FliD cap from enzymatic degradation. If so, this should be true for all spirochetes. To explore this possibility, we conducted multiple sequence alignment and phylogenetic analyses of 60 FliD proteins from different bacterial species. The results showed that FliD proteins from all of the spirochetes form a distinct clade due to having an extra VR insert at the same position as FliD_Bb (Fig. 11). Interestingly, in addition to spirochetes, the FliD of H. pylori also contains a 195 aa insert (aa 133 to 328) in D3 (Fig. S2). A conceivable explanation is that this insert may protect H. pylori FliD from the acidic environment and various proteolytic enzymes in the gastric mucosae. In the future, it would be very interesting to investigate the impact of this VR insert on the function and structure of FliD by combining genetic, biochemistry, and structural biology approaches.

FIG. 11. Phylogenetic analysis of FliD proteins.

Protein sequences were aligned using MUSCLE (Edgar, 2004). A maximum likelihood tree was inferred using FastTree (Price et al., 2010). The tree was subsequently re-rooted at the mid-point, and branches with lower than 90% bootstrap support were removed using the BIOTREE utility of the BpWrapper package (Hernandez et al., 2018). The tree was visualized and graphed using the APE package in R (Paradis et al., 2004). To reconstruct the evolutionary origin of the central domain, we inferred its gains or losses based on the principle of maximum parsimony. FliD homologs were identified using BLASTp with FliD_Bb (BB0149) as a query, and protein sequences were downloaded from GenBank. Note: Borrelia burgdorferi was recently named Borreliella burgdorferi.

How does B. burgdorferi degrade unpolymerized FlaB in the periplasm?

The bacterial periplasm is a narrow compartment between the cytoplasmic membrane and the outer membrane (Miller & Salama, 2018). The periplasm of B. burgdorferi cells is only 20–40 nm in width (Kudryashev et al., 2009, Liu et al., 2009). Within this narrow space are 7–11 long helical PF, which constantly rotate against the cell cylinder and generate torque to propel the cells forward (Wolgemuth, 2015). Assembling these large locomotive organelles in such a narrow space imposes tremendous physical stresses on the cells (Dombrowski et al., 2009, Yang et al., 2011). Thus, B. burgdorferi and other spirochetes must have evolved unique mechanisms to finely orchestrate the assembly process of PF in the periplasm. Misassembled PF not only impair cell motility but can also cause detrimental effects, including overgrowth of PF in the periplasm, which may protrude through the cell membranes and lyse the cells. As evident in the ∆fliD mutant, deletion of fliD_Bb leads to FlaB leakage into the periplasm. FlaB accounts for 10–14% of the total cellular protein content (Motaleb et al., 2004). Accumulation of such an amount of FlaB in the periplasm can be toxic to the cells. Along with speculation, we found protein aggregates accumulated in the periplasm of ΔfliD-htrA (Fig. S4) but not in that of single fliD and htrA mutants. In addition, the double mutant grew slower than the single mutants (Fig.S5). The HtrA (high-temperature-requirement) family of serine proteases plays an important role in regulating bacterial stress responses by turning over damaged or incorrectly folded proteins that are generated under stress conditions or during abortive cellular localization (Backert et al., 2018). In B. burgdorferi, HtrA_Bb plays an important role in cell physiology (membrane protein processing and motility) and virulence (dissemination and degradation of host factors) (Gherardini, 2013). In this report, we found that HtrA_Bb degrades unfolded FlaB, a major flagellin protein of B. burgdorferi. First, we found that FlaB is degraded in the fliD_Bb mutant but not in other flagella-deficient mutants (i.e., ∆fliF, ∆flgG, and ∆flgE) in which the export route of FlaB is blocked (Fig. 8). Second, we demonstrated that deletion of htrA_Bb in the background of a fliD_Bb mutant increases the level of FlaB (Fig. 9). Lastly, in vitro proteolytic assays showed that HtrA_Bb degrades unpolymerized FlaB proteins but not polymerized ones (Fig. 10). Based on these results, we propose that HtrA_Bb degrades unpolymerized FlaB when the flagellar filament polymerization is abrogated, thereby preventing accumulation of toxic products in the periplasm of B. burgdorferi. We are also aware that deletion of htrA_Bb only partially blocks FlaB degradation, suggesting that there might be other proteases involved in FlaB turnover in the periplasm.

How does B. burgdorferi regulate its flagellar assembly?

In a given flagellum, the filament is the largest part, consisting of more than a thousand copies of flagellin proteins. Thus, the expression of flagellin genes must be finely tuned to ensure sufficient but not too much proteins are produced (Hughes, 2017, Chevance & Hughes, 2008, Mukherjee & Kearns, 2014). In most bacteria such as E. coli and S. typhimurium, the expression of flagellin genes (e.g., fliC) is controlled by a transcription regulatory cascade that consists of FliA (a motility-specific transcription factor, σ²⁸) and FlgM (an anti-σ²⁸ factor) (Aldridge et al., 2006, Hughes, 2017). FlgM and FliA remain bound as a complex in the cytoplasm during basal body and hook synthesis. Upon completion of the hook, the hook-basal body complex (also known as fT3SS) provides an export route for FlgM. As FlgM exits the cell, FliA is free to initiate the transcription of fliC and to support filament assembly. In addition to the transcriptional regulation by FliA-FlgM (Calvo & Kearns, 2015, Mukherjee & Kearns, 2014), B. subtilis also homeostatically regulates the level of flagellin (Hag) by a partner-switching mechanism between the FliW protein and either Hag or CsrA (carbon storage regulator A), an RNA binding protein that represses hag translation (Mukherjee et al., 2011, Yakhnin et al., 2007). FliW is an antagonist of CsrA and binds to both Hag and CsrA. Upon completion of the hook, a checkpoint of flagellar assembly, Hag is secreted via fT3SS, releasing FliW from a FliW-Hag complex. Free FliW in turn binds to CsrA, relieving CsrA-mediated repression on hag translation, thereby initiating Hag synthesis and filament assembly.

The regulation of flagellar synthesis in B. burgdorferi differs markedly from other bacteria. First, the gene encoding FliA and FlgM are not present in its genome, nor are σ²⁸ promoter consensus sequences associated with the motility genes (Fraser et al., 1997, Charon et al., 2012). Second, analysis of a hook-deficient mutant reveals that completion of the hook assembly does not serve as a checkpoint for transcriptional regulation of flagellar synthesis in B. burgdorferi (Sal et al., 2008). Thirdly, we previously demonstrated that BB0184 (CsrA_Bb), a homolog of B. subtilis CsrA, specifically represses FlaB synthesis by inhibiting the translation initiation of flaB mRNA (Sze et al., 2011). We recently also found that BB0183, a homolog of B. subtilis FliW, functions as an antagonist of CsrA_Bb and deletion of this gene inhibits the translation of flaB (Li et al., unpublished data), suggesting that B. burgdorferi regulates the level of FlaB by a partner-switching mechanism as evident in B. subtilis. Lastly, the data in this report reveal that deletion of fliD_Bb severely impaired the level of FlaB but had no impact on the transcription of flaB, indicating that in the absence of FliD_Bb, the reduction of FlaB occurs at the translational level. We also found that the level of FlaB in the fliD mutant is significantly less than that of other flagella-deficient mutants (i.e., ∆fliF, ∆flgG, and ∆flgE) and that deletion of htrA_Bb partially increases the level of FlaB (Fig. 9B). In vitro proteolytic analyses show that HtrA_Bb degrades unpolymerized FlaB proteins but not polymerized ones, indicating a role of HtrA in FlaB turnover in the periplasm of B. burgdorferi.

Based on above discussions, we propose a model (Fig. 12) to explain the phenotype of the fliD_Bb mutant and the regulatory mechanism of flagellar assembly in B. burgdorferi. In this model, B. burgdorferi controls the cytoplasmic level of FlaB by a partner-switching mechanism of CsrA_Bb-FliW-FlaB and governs the assembly of PF in the periplasm by removing unfolded or misfolded FlaB with HtrA_Bb. Upon completion of the hook assembly, FlaB monomers are transported from the cytoplasm into the periplasm via fT3SS, lowering the cytoplasmic level of FlaB. When FlaB falls below a threshold level, FliW is released and binds to CsrA_Bb, thereby relieving its translational repression to produce more FlaB proteins for assembling nascent filaments. When the filament assembly is completed or fT3SS is disrupted, as seen in the ∆fliF, ∆flgG, and ∆flgE mutants, the secretion of FlaB is blocked, which leads to FlaB accumulation in the cytoplasm. When FlaB rises above a threshold level, it binds to FliW, releasing CsrA_Bb, which in turn binds to the Shine-Dalgarno (SD) sequence of flaB mRNA and blocks translation. Thus, the FlaB levels in the ∆fliF, ∆flgG, and ∆flgE mutants resemble its basal level in the cytoplasm. Deletion of fliD_Bb leads to FlaB monomers leaking into the periplasm, where they are degraded by HtrA_Bb. As a result, the homeostatic equilibrium of CsrA_Bb-FliW-FlaB in the cytoplasm is perturbed, which further leads to repression of flaB translation. Collectively, HtrA_Bb-mediated protein turnover and CsrA_Bb-mediated translational repression reduces the basal level of FlaB in the fliD_Bb mutant.

FIG. 12. Model illustrating how flagellar assembly is regulated in B. burgdorferi.

In this model, the FlaB-FliW-CsrA_Bb interaction regulates flagellin homeostasis in the cytoplasm by a partner-switching mechanism, and HtrA_Bb functions as a house cleaner by removal of unpolymerized FlaB in the periplasm. Upon completion of the hook assembly, FlaB monomers are transported from the cytoplasm into the periplasm via fT3SS, reducing the cytoplasmic level of FlaB. When FlaB falls below a threshold level, FliW is released and binds to CsrA_Bb, thereby relieving the translational repression such that more FlaB proteins are synthesized and exported to assemble nascent filaments. When the filament assembly is completed or fT3SS is disrupted, as occurs in the ∆flgE mutant, the secretion of FlaB is blocked. This leads to FlaB accumulation in the cytoplasm. When FlaB rises above a threshold level, it binds to FliW, releasing CsrA_Bb, which in turn binds to the SD sequence of flaB mRNA, blocking translation. Deletion of fliD_Bb leads to FlaB monomers leaking into the periplasm, where they are degraded by HtrA_Bb. As a result, the homeostatic equilibrium of CsrA_Bb-FliW-FlaB in the cytoplasm is perturbed, which further leads to restriction of flaB translation.

Experimental Procedures

Bacterial strains and growth conditions.

A high-passage Borrelia burgdorferi sensu stricto strain, B31A (wild type) (Samuels et al., 1994, Samuels et al., 2018), and its isogenic mutants were grown in Barbour-Stoenner-Kelly II (BSK-II) liquid medium or on semi-solid agar plates at 34°C or 23°C in the presence of 3.4% carbon dioxide, as previously described (Li et al., 2002, Sze et al., 2013). The strains were grown in appropriate antibiotics for selective pressure as needed: kanamycin (300 μg ml⁻¹), gentamicin (40 μg ml⁻¹), or streptomycin (50 μg ml⁻¹). Escherichia coli TOP10 strain (Invitrogen, Carlsbad, CA) was used for DNA cloning and plasmid amplification. M15 and BL21 strains (Qiagen, Valencia, CA) were used for recombinant protein preparations. E. coli strains were cultured in lysogeny broth (LB) supplemented with appropriate antibiotics as needed.

RNA preparations and reverse transcription PCR (RT-PCR).

RNA isolation was performed, as previously described (Sze et al., 2013). Briefly, B. burgdorferi strains were cultivated at 34°C, and 100 ml of early stationary phase cultures (~10⁸ cells/ml) was harvested for RNA preparation. Total RNA was extracted using TRI reagent (Sigma-Aldrich, St. Louis, MO), following the manufacturer’s instructions. The RNA samples were then treated with Turbo DNase (Ambion, Austin, TX) at 37°C for 2 hours to eliminate genomic DNA contamination. The resultant RNA samples were re-extracted using acid phenol-chloroform, precipitated in isopropanol, and washed with 70% ethanol. The RNA pellets were re-suspended in RNase-free water. cDNA was generated from the purified RNA (300 ng) using AMV Reverse Transcriptase (Promega, Madison, WI). For RT-PCR, 1 µl of cDNA was PCR amplified with different pairs of primers (P19-P24) using Taq DNA polymerase (Qiagen). For quantitative PCR (qPCR), the transcripts of target genes (e.g., flaB and flgE) were amplified using iQ SYBR Green Supermix and a MyiQ thermal cycler (Bio-Rad, Hercules, CA). The enolase gene (eno, bb0337), a housekeeping gene, was included as an internal control to normalize target gene transcripts, as previously described. The results were expressed as threshold cycle (C_T) value between the wild type and mutant strains. The primers for RT-PCR and qRT-PCR are listed in Table S1.

β-galactosidase and green fluorescence protein (GFP) reporter assays.

The upstream intergenic region of fliD_Bb (gene locus, bb0149) was PCR amplified with primers of P43-P44 (Table S1), generating an amplicon with engineered EcoRI and BamHI cut sites at the 5’ and 3’ ends, respectively. The resultant amplicon was cloned into pGEM-T Easy vector (Promega) and then released by EcoRI and BamHI digestion. The released DNA fragment was cloned into pRS414 (Bian et al., 2011), a lacZ reporter plasmid (a gift from R. Breaker, Yale University). A previously identified flaB promoter (Ge et al., 1997) was also amplified and cloned into pRS414, which was used as a positive control. The resultant plasmids were transformed into the E. coli DH5α strain. Galactosidase activity was measured, as previously described (Bian et al., 2011). The results were expressed as the average Miller units of triplicate samples from two independent experiments. For the fluorescent protein reporter assay, the fliD promoter was fused to gfp gene and then cloned into the shuttle vector pJSB275 (Groshong et al., 2012). A promoterless gfp was also cloned into pJSB275, which is used as a negative control. The resultant construct was transformed into B31 A and the expression of GFP was visualized with fluorescent microcopy (Axiostar Plus, Ziess), as previously documented (Yang & Li, 2009).

DNA cloning, recombinant protein preparation, and antibody production.

Three recombinant proteins, including FliD_Bb, FlaB (gene locus, bb0147) and HtrA_Bb (gene locus, bb0104), were prepared by using different E. coli expression systems. The recombinant FliD_Bb (rFliD_Bb) was prepared using the pQE30 vector and E. coli M15 strain (Qiagen), and the other two (rFlaB and rHtrA_Bb) were prepared using the pET100 vector and E. coli BL21 Star (DE3) strain (Invitrogen). For rFliD_Bb, the entire fliD_Bb gene was PCR amplified with engineered BamHI and PstI cut sites at its 5′ and 3′ ends, respectively. The resultant amplicon was cloned into the pGEM-T vector (Promega) and then subcloned into pQE30 with an N-terminal His₆ tag. The resultant expression vector was transformed into M15 cells. For rFlaB and rHtrA_Bb, the genes encoding these two proteins were PCR amplified using Platinum pfx DNA polymerase (Invitrogen) and directionally cloned into pET100 with an N-terminal His₆ tag. The resultant expression vectors were transformed into E. coli BL21 Star (DE3). For rHtrA_Bb, a site-directed mutant of rHtrA_Bb^S226A was prepared. The mutation was introduced by using QuikChange II Site-Directed Mutagenesis Kit (Stratagene, San Diego, CA) and confirmed by DNA sequencing (Roswell Park Cancer Institute, Buffalo, NY). The primers for the recombinant protein preparations and site-directed mutagenesis are listed in Table S1.

The expression of recombinant proteins in E. coli cells was induced with 1 mM isopropyl-β-D-thiogalactoside (IPTG). The recombinant proteins were purified using nickel agarose columns (Qiagen) under native conditions per the manufacturer’s instructions. The purified proteins were dialyzed in a buffer containing 10 mM Tris-HCl at 4°C overnight. To raise antibodies against FliD_Bb and HtrA_Bb, approximately 5 mg purified recombinant proteins were used to immunize rats (2.5 mg for each animal) at General Bioscience Corporation (Brisbane, CA), following a standard immunization procedure.

Gene inactivation and complementation of fliD.

The fliD_Bb gene and its flanking DNA were first amplified by PCR with primers P1/P2 from chromosomal DNA of strain B31A, and the product obtained was cloned into plasmid pGEM-T Easy (Promega) to yield pGFD. A previously constructed 1.3 kb flgB-kanamycin-resistance cassette (kan) (Motaleb et al., 2000) was PCR amplified and inserted into pGFD at a HindIII cut site of fliD_Bb, yielding plasmid fliD::kan. To disrupt fliD_Bb, the fliD::kan construct was linearized and transformed into competent B31A cells by electroporation (Samuels et al., 1994). Transformants were selected on semi-solid agar plates containing kanamycin (300 μg/ml). The fliD_Bb mutant was cis-complemented, as previously described (Sze et al., 2013, Li et al., 2007). Briefly, the flgB promoter of B. burgdorferi (P_flgB) and the fliD_Bb gene were PCR amplified with primers P5/P6 and P7/P8, respectively. The resultant two fragments were fused by PCR, generating P_flgB-fliD with engineered XbaI and PstI cut sites at its 5′ and 3′ ends, respectively. P_flgB-fliD_Bb was first cloned into pGEM-T vector, released by XbaI and PstI, and subcloned into pCisCom, a previously constructed suicide plasmid that harbors an intergenic region of bb0445-bb0446 (Sze et al., 2013), yielding the construct of fliD/pCisCom. To complement the fliD_Bb mutant, fliD/pCisCom was linearized and transformed into fliD_Bb mutant cells. Complemented clones were selected on semi-solid agar plates containing both kanamycin (300 μg/ml) and gentamicin (40 μg/ml). The primers for construction of the fliD_Bb mutant and its complemented strain are listed in Table S1.

Generation of htrA_Bb and csrA_Bb deletion mutants.

Two vectors were constructed to in-frame delete htrA_Bb, one with the kan cassette and the other one with aadA1, a streptomycin resistance gene. A kanamycin resistance mutant of csrA_Bb (gene locus, bb0184) was generated in our previous reports (Sze et al., 2013). To construct a double mutant of fliD_Bb-csrA_Bb, a new streptomycin resistance csrA_Bb mutant was generated. The constructs for deletion of htrA_Bb or csrA_Bb were constructed by using a PCR-based 3-way stitching method, as previously described (Motaleb et al., 2011). Briefly, the PCR primers (containing complementary overlaps to kan or aadA1 were designed immediately flanking the csrA_Bb gene or the htrA_Bb gene to generate approximate 1 kb products upstream and downstream of the coding sequences. The primers for the antibiotic genes (containing complementary overlaps to either csrA_Bb or htrA_Bb) were designed. An initial PCR for each of the three individual parts (i.e., 5’- and 3’-flanking DNA of csrA_Bb and kan) was performed, followed by a “stitching” PCR to connect all three pieces together. By using this method, csrA_Bb::aadA1, htrA_Bb::kan, and htrABb::aadA1 were constructed. The resultant constructs were transformed into competent wild type or fliD_Bb mutant cells by electroporation to delete the csrA_Bb and htrA_Bb genes. Mutant clones were selected on semi-solid agar plates containing kanamycin (350 μg ml⁻¹) alone or both kanamycin and streptomycin (50 μg ml⁻¹). The primers for these constructs are listed in Table S1.

Electrophoresis, immunoblotting, and protein turnover assays.

Sodium-dodecyl-sulfate polyacryl-amide gel electrophoresis (SDS-PAGE) and immunoblotting analyses were carried out, as previously described (Sze et al., 2013). Briefly, B. burgdorferi cells were cultured at 34°C and harvested at approximately 10⁸ cells/ ml. Equal amounts of whole-cell lysates (10 to 20 μg) were separated on an SDS-PAGE gel and transferred onto a polyvinylidene difluoride (PVDF) membrane (Bio-Rad). Immunoblots were probed with specific antibodies against various proteins, including DnaK, HtrA_Bb, FlaA, FlaB, FlgE, FlgG, FliF, and FliD_Bb. DnaK was used as an internal loading control. Monoclonal antibodies to FlaB, FlaA, and DnaK were provided by A. Barbour (University of California, Irvine, CA), B. Johnson (Centers for Disease Control and Prevention, Atlanta, GA), and J. Benach (State University of New York, Stony Brook, NY), respectively. HtrA_Bb, FliD_Bb, FlgG, and FliF antibodies were generated in this study, and a polyclonal antibody against FlgE was described in our previous publications (Sal et al., 2008). Immunoblots were developed using horseradish peroxidase-labeled secondary antibodies or protein A (GE Healthcare, Little Chalfont Buckinghamshire, United Kingdom) with Pierce ECL Western blotting substrate kit (Thermo Scientific, Rockford, IL).

Protein turnover assays were carried out, as previously described (Sal et al., 2008). Briefly, B. burgdorferi strains were grown in BSK-II medium at 34°C to late logarithmic phase (~10⁸ cells ml⁻¹), and then spectinomycin (100 µg ml⁻¹) was added to arrest protein synthesis. Then, 5 ml of cells were harvested at 0, 2, 4, 8, and 12 hours and subjected to SDS-PAGE, followed by immunoblots against different antibodies, as indicated. Immunoblots were developed using horseradish peroxidase-labeled secondary antibodies (GE Healthcare) with Pierce ECL Western Blotting Substrate Kit (Thermo Scientific). Densitometry of immunoreactive proteins in the blots was measured using the Molecular Imager ChemiDoc™ XRS Imaging system (Bio-Rad) to determine the relative amounts of proteins, as previously described (Sze et al., 2013).

Swimming plate assays.

Swimming plate analysis was performed, as previously described (Li et al., 2002). Briefly, 5 μl of culture (10⁸ cells ml⁻¹) was spotted onto 0.35% agarose plates containing BSK-II medium that was diluted 1:10 with Dulbecco’s phosphate-buffered saline (PBS) without divalent cations. Plates were incubated for 3 to 4 days at 34°C with 3.4% CO₂. The diameters of the swimming rings were measured and recorded in millimeters. The wild-type B31A strain was used as a positive control, and a previously constructed non-motile flaB mutant (Motaleb et al., 2000)was used as a negative control to define the initial inoculum size.

Enzymatic assays.

Purified rHtrA_Bb and its mutated version, rHtrA_Bb^S226A, were stored frozen in PBS containing 5% glycerol. The proteolysis assays of rHtrA_Bb against purified PF and recombinant FlaB (rFlaB) were conducted, as previously described (Coleman et al., 2013). Briefly, rHtrA_Bb (5 μg) was co-incubated with the same amount of rFlaB or PF in a final volume of 50 μl at 37°C overnight, and then 5x SDS-PAGE sample buffer was added (final 1 x). The resultant samples were boiled and subjected to 10% SDS-PAGE, followed by Coomassie blue staining. PF were isolated from a flaA deletion mutant, as previously described (Miller et al., 2014).

Cryo-ET data collection and 3D reconstruction.

The tilt series for the tomograms were collected at −170°C using a Polara G2 electron microscope equipped with a field emission gun and a direct detection device (Gatan K2 Summit), as previously described (Zhao et al., 2013, Liu et al., 2009). The microscope was operated at 300kV with a magnification of ×15,400, resulting in 2.5 Å /pixel. We used SerialEM to collect low-dose, single-axis tilt series with dose-fractionated mode at about −8 µm defocus. A cumulative dose of ~50 e⁻/Ų was distributed over 35 projections covering an angular range of −51° to +51° with 3° fixed increments. Each projection contains ~8–10 frames of images from the dose fractionation and has been drift-corrected using Motioncorr. The tilted series alignment was facilitated using Tomoauto with an automatic fiducial seed model, and reconstructed into both weighted back projection (WBP) tomograms and simultaneous iterative reconstruction technique (SIRT) tomograms using Tomo3D (Wolf et al., 2014, Agulleiro & Fernandez, 2011). The SIRT tomograms were binned by 4 so that there is enough contrast to pick the sub-tomograms manually. The WBP tomograms were used to process the sub-tomogram alignment and averaging.

Sub-tomogram analysis, 3D visualization, and modeling.

We used the tomographic package I3 for sub-tomogram analysis, as previously described (Zhao et al., 2013). The initial manual picking selected two points along each hook; thus, the initial position and orientation of each sub-tomogram were determined. There are 1,492 sub-tomograms that were extracted from 211 tomograms for the ΔflaB mutant, and 599 sub-tomograms from 83 tomograms for the fliD_Bb mutant. We use the template-free approach to avoid reference bias as well as the “alignment by classification” method with missing wedge compensation that is embedded in I3. The averaged structures from both mutants were asymmetric and visualized without any imposed symmetry (Fig. 6). For 3D visualization and modeling, the surface view of the cell tips (Fig. 5) and visualization of the electron density of sub-tomogram-averaged structures (Fig. 6) were done in IMOD (Kremer et al., 1996). The surface rendering of the averaged structures were visualized in UCSF Chimera (Pettersen et al., 2004). IMOD was used for the 3D length measurement of the hook and cap in raw tomogram using a two-step process: 1) An open contour containing multiple points was first drawn along the center of the hook and cap in bin4 SIRT tomograms and 2) The 3D length of the contour was calculated by the “Edit-point-distance” function in IMOD.

Bioinformatics analyses.

Homology structural modeling was conducted using the Fold Function Assignments (FFAS) (http://ffas.sanfordburnham.org/ffas-cgi/cgi/ffas.pl) (Jaroszewski et al., 2005) and SWISS-MODEL (https://swissmodel.expasy.org) web servers. The crystal structure of E. coli FliD was used as the template (Song et al., 2017). Protein sequences were aligned using MUSCLE (Edgar, 2004). A maximum likelihood tree was inferred using FastTree (Price et al., 2010). The tree was subsequently re-rooted at the mid-point, and branches with lower than 90% bootstrap support were removed using the BIOTREE utility of the BpWrapper package (Hernandez et al., 2018). The tree was visualized and graphed using the APE package in R (Paradis et al., 2004). To reconstruct the evolutionary origin of the central domain, we inferred its gains or losses based on the principle of maximum parsimony.