Gliding is energized by the proton gradient

It is well known that gliding motility is arrested upon dissipation of the PMF [31–35]. The PMF consists of 2 gradients across the cytoplasmic membrane: an electrical potential (Δψ) and a chemical potential (ΔpH). To better define the energy source that powers gliding motility, F. johnsoniae gliding cells in a well chamber with glass bottom were subjected to valinomycin/K⁺ or nigericin, to specifically dissipate Δψ or ΔpH, respectively, and single-cell gliding motility was quantified (Fig 1A and 1B). In agreement with previous observations [6], cells glide with an average speed of 1.7 μm.s⁻¹ (Fig 1B). As a control, injection of 10 μM of CCCP rapidly blocked all cell displacement in a reversible manner. By contrast, no significant inhibitory effect was observed upon addition of 40 μM valinomycin (+50 mM KCl). However, when cells were treated with 7 μM nigericin, motility was strongly impaired (Fig 1B). Instead of gliding, cells appeared to jiggle around the same location, possibly because nigericin did not totally abolish the ΔpH (Fig 1A). When nigericin was washed out, cells resumed gliding motility at normal speed. Therefore, we conclude that the proton gradient, but not the electrical potential, is the source of energy used by the gliding machinery.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Effects of PMF dissipating drugs on F. johnsoniae cells gliding.

(A) Rainbow traces of cell motility on glass recorded by phase contrast microscopy over time (2 minutes) in the absence of drug or in the presence of CCCP, nigericin, or valinomycin. Individual frames from time-lapse acquisition were colored from red (start) to yellow, green, cyan, and blue (end) and merged into a single rainbow image. Scale bar, 20 μm. (B) Combined jitter plots/boxplots of mean cell gliding velocity (in μm.s⁻¹) of n > 38 WT cells before (−), during a pulse of 10 μM CCCP (+CCCP), or 7 μM nigericin (+Nig), or 40 μM valinomycin/+50mM KCl (+Val), and after wash with fresh CYE medium (Wash). Statistical significance relative to the nontreated condition (−) is indicated above the plots (ns, nonsignificative; ***, p < 0.001; Wilcoxon t test). Data underlying this figure can be found in S1 Data. CCCP, carbonyl cyanide m-chlorophenyl hydrazone; PMF, proton motive force; WT, wild-type.

https://doi.org/10.1371/journal.pbio.3001443.g001

GldL and GldM constitute the molecular motor that couples PMF to GldM conformational changes

Bacterial molecular motors such as the MotAB flagellar stator, the ExbBD and TolQR related transport systems, and the AglRQS gliding motor generate mechanical energy by harvesting the chemical gradient through the cytoplasmic membrane [40,43–45]. These complexes usually comprise 2 subunits organized in a 5:2 stoichiometry that interact via their TMHs [46,47]. A conserved acidic residue, located in one TMH and facing the other TMHs, plays a key role in proton transit [46–50]. It is proposed that proton flow through the channel triggers protonation–deprotonation cycles of the side chain of this residue and induces rearrangements in the TMHs, ultimately leading to the production of mechanical torque in the form of conformational changes in the extramembrane regions [51–53].

GldL and GldM interact via their transmembrane segments.

In agreement with the recent cryo-EM structure of the GldLM complex, bacterial 2-hybrid analyses show that GldL and GldM interact (Fig 2A). As previously shown for the P. gingivalis PorLM complex [23], these interactions likely involve the transmembrane segments of both proteins. In agreement with the cellular localization of the soluble domains of GldL and GldM, which reside in the cytoplasm and periplasm, respectively, no GldL-M interaction was observed when only the soluble domains of these proteins (GldL_C and GldM_P) were tested (Fig 2A). Finally, similarly to the P. gingivalis PorKLMN complex [23,24], the GldLM module is implicated in interactions with the putative OM-associated GldKN/O ring (S1A Fig) via contacts between the GldM periplasmic domain and GldK, GldN, and GldO (S1B Fig). To test the contribution of TMHs to GldLM interactions, we conducted GALLEX and BLA approaches. GALLEX is based on the repression of a β-galactosidase reporter by 2 LexA DNA binding domains with different DNA binding specificities (LexA^WT and LexA⁴⁰⁸). If 2 TMHs interact, LexA^WT and LexA⁴⁰⁸ heterodimeric association causes repression of β-galactosidase synthesis [54,55]. We tested interactions between TMHs displaying in-to-out topologies (GldL-TMH1 and GldM-TMH; Fig 2B). GldL-TMH1 and GldM-TMH specifically formed homodimers but no interaction was detected between GldL-TMH1 and GldM TMH (Fig 2C). To test the interaction with GldL-TMH2, which exhibits an out-to-in topology, we used the BLA assay [56,57]. GldL-TMHs and GldM-TMH were fused to either the N- or carboxyl-terminal domain of β-lactamase (Bla). If an interaction between TMHs occurs, a functional β-lactamase is reconstituted, and its activity can be quantified using a chromogenic substrate-based assay. In this assay, GldL-TMH2 homodimerization was not observed. However, GldL-TMH2 specifically interacted with both GldL-TMH1 and GldM-TMH (Fig 2D). Taken together, our data show that GldL TMH1 and TMH2 interact with each other, and GldL TMH2 interacts with GldM single TMH in the motor complex. With the exception of GldL-TMH1/GldL-TMH1 contacts detected by GALLEX, these data are in agreement with the position of the TMHs in the recent cryo-EM structure of the GldLM complex [42].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Interactions between GldL and GldM TMHs.

(A) Bacterial 2-hybrid assay. GldL–GldM interaction is dependent on their TMHs. BTH101 cells producing the indicated proteins (GldL and GldM) or domains (GldL_C, cytoplasmic domain of GldL, amino acid 59 to 189; GldM_P, periplasmic domain of GldM, amino acid 36 to 513) fused to the T18 and T25 domain of the Bordetella adenylate cyclase were spotted on X-Gal-IPTG reporter LB agar plates. The blue color of the colony reports interaction between the 2 partners. Controls include T18 and T25 fusions to TolB and Pal, 2 proteins that interact but unrelated to the T9SS. (B) Schematic representation of GldL and GldM domains and topologies in the IM. The N-terminus (N) and carboxyl terminus (C) of proteins are indicated in gray. The GldL TMH1 (L1, blue) and TMH2 (L2, brown) and GldM TMH (M, green) are indicated. (C) Homo- and heterodimerization of in-to-out TMHs of GldL and GldM probed with the GALLEX method. Jitter plots of β-galactosidase activity reporting the dimerization of TMHs fused to LexA^WT or LexA⁴⁰⁸. Measurements (transcriptional repression) are reported as 1/β-galactosidase activity. Data are combined from technical triplicates of 4 independent measurements (2 colonies from 2 independent transformations each). Interactions with TssL1 (in gray) served as negative control or positive control for self-interactions. TssL1 is the TMH of TssL, a protein of the E. coli T6SS that homodimerizes. Data underlying this figure can be found in S1 Data. (D) Homo and heterodimerization of the in-to-out and out-to-in THMs of GldL and GldM probed with the BLA method. Jitter plots of CENTA chromogenic substrate hydrolysis after 10 minutes of incubation. The activity is reported as the A_405nm value per A_600nm. Controls include interaction assays with the TMHs of TssM (TssM1 and TssM2, in gray), a subunit of the E. coli T6SS. Data underlying this figure can be found in S1 Data. T6SS, type VI secretion system; T9SS, type IX secretion system; TMHs, transmembrane helices; WT, wild-type.

https://doi.org/10.1371/journal.pbio.3001443.g002

GldM changes conformation depending on the proton gradient.

We next sought to understand how the GldLM complex responds to the proton gradient. Our previous structural characterization of GldM and its homolog PorM from P. gingivalis revealed a conformational flexibility in the periplasmic region of GldM [25]. The extracellular region of GldM assembles as homodimer, spanning most of the periplasm, composed of 4 domains, D1 to D4 [25,58]. Interestingly, the homolog PorM presents a kink between domains D2 and D3 [25], while a kink between the GldM D1 and D2 domains has been proposed in a composite model [42], suggesting that GldM/PorM may alternate between several conformational states. Indeed, recent proteolytic susceptibility assays showed that the PMF regulates conformational changes in GldM: upon PMF dissipation by CCCP, 2 cleavages at the interface of domains D2 and D3 were identified by mass spectrometry after limited trypsinolysis [59]. Here, we extend these observations by showing that the in vivo conformation of GldM is altered by drugs that perturb the proton gradient such as CCCP and nigericin, but remained unaffected upon treatment with the F₁F₀ ATPase inhibitors sodium azide and sodium arsenate, nor upon dissipation of the Δψ by valinomycin (Fig 3A), demonstrating that GldM undergoes a structural transition dependent on the IM proton gradient (Fig 3B).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Conformational changes in GldM periplasmic domain in response to the proton gradient.

(A) GldM protease accessibility assay. Spheroplasts of WT F. johnsoniae or the ΔgldM mutant were treated (+) or not (−) with the trypsin protease and 10 μM CCCP (PMF inhibitor), 7 μM nigericin (ΔpH inhibitor), 40 μM valinomycine/K+ (Δφ inhibitor), 1.5 mM sodium azide, or 20 mM sodium arsenate (F₁F₀ ATPase inhibitors). GldM was analyzed by SDS-PAGE and immunoblot with anti-GldM antibodies. The full-length GldM protein is indicated, as well as degradation products (* and **). The molecular mass markers (in kDa) are indicated on the left. (B) Schematic model of GldM conformational transition dependent on the proton gradient. WT, wild-type.

https://doi.org/10.1371/journal.pbio.3001443.g003

A conserved glutamate residue in GldL TMH2 is critical for harvesting the proton gradient.

The activity of bacterial MotAB-like molecular motors characterized so far depends on conserved acidic residues located in their TMHs [39,40,48,60]. We first tested the effect of the drug N-N’-Dicyclohexyl-carbodimide (DCCD), which covalently reacts with carboxylic groups located in a hydrophobic environment [61]. Addition of DCCD abrogated gliding motility (Fig 4A and 4B) and had the same effect on GldM conformation as CCCP (Fig 4C). This effect was irreversible since DCCD remains covalently bound (Fig 4B). These results therefore suggested that aspartate or glutamate residues are involved in coupling PMF to GldM conformational change and gliding motility. Sequence alignment showed that 3 acidic residues are conserved in the GldL and GldM N-terminal regions (S2A and S2B Fig): a glutamate at position 31 in the vicinity of GldM TMH (GldM-E31) and 2 glutamates in GldL, one located in TMH2 (E49; strictly conserved in all GldL homologs in the OrthoDB database), and one located between TMH2 and the cytoplasmic domain (E59) (S2C Fig). Substitution of GldM E31 (GldM^E31A) did not exhibit any defect in gliding motility compared to wild-type (WT) GldM, indicating that this residue does not play a significant role in T9SS-dependent secretion or gliding (Fig 4D). By contrast, cells producing GldL^E49A failed to adhere to the glass surface, while substitution of GldL E59 abolished gliding motility without affecting adherence (i.e., SprB secretion by the T9SS) (Fig 4E). Interestingly, the GldL^E49A and GldL^E59A variants presented distinct phenotypes regarding SprB secretion and dynamics as shown by live cell immunolabeling using polyclonal anti-SprB antibodies and fluorescence time-lapse microscopy on agarose pads. As previously reported [6], WT cells exhibited surface-exposed SprB fluorescent foci that describe an overall helicoidal pattern along the long cell axis, with dispersed velocity in the order of 1 μm.s⁻¹ (Fig 4F). No fluorescent focus was observed in ΔgldL mutant cells, which are unable to secrete SprB, indicating that SprB immunolabeling was specific. A similar observation was made with gldL^E49A mutant cells (Fig 4F), demonstrating that T9SS-dependent SprB secretion to the cell surface requires residue E49 in GldL TMH2. By contrast, the GldL^E59A substitution supported SprB secretion but abolished the dynamic cell surface movements of the adhesin (Fig 4F and 4G). It is noteworthy that all GldL and GldM variants were produced in F. johnsoniae at levels comparable to the WT proteins (S2D and S2E Fig), although GldL^E49A migrated with lower apparent size than the WT protein (S2E Fig), an aberrant migration already observed in a separate study [42] and likely caused by the difference in detergent binding in SDS-PAGE between the TMH2 variants [62]. Taken together, these results support a model in which GldL E49 is required for secretion of the SprB adhesin and constitutes a key determinant of T9SS, whereas GldL E59 is dispensable for secretion and plays a specific function in gliding because it is only required for SprB movement. We next tested the contribution of these acidic residues for the regulation of GldM conformation. Protease accessibility assays showed that GldL and its Glu49 residue are required to maintain GldM in the conformation required for T9SS activity (Fig 4H). By contrast, the GldL E59A substitution did not impact GldM proteolytic susceptibility (Fig 4H), suggesting that the GldM conformation change observed by limited proteolysis is specifically linked to effector secretion rather than gliding motility.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Roles of conserved GldL TMH2 glutamates in T9SS-dependent secretion and dynamics of SprB.

(A) The addition of the drug DCCD inhibited cell motility. Rainbow traces of cell motility on glass recorded by phase contrast microscopy over time (2 minutes) in the presence of 10 μM DCCD. (B) Combined jitter plots/boxplots of mean cell gliding velocity (in μm.s⁻¹) of n > 50 WT cells before (−), during a pulse of 10 μM DCCD (+DCCD), and after wash with fresh CYE medium (Wash). Statistical significance relative to the nontreated condition (−) is indicated above the plots (ns, nonsignificative, ***, p < 0.001; Wilcoxon t test). Data underlying this figure can be found in S1 Data. (C) Effect of DCCD on GldM protease susceptibility. GldM protease accessibility assay on spheroplasts of WT F. johnsoniae in absence (−) or presence (+) of 10 μM DCCD. (D) GldM conserved glutamate 31 does not play an important role in gliding motility. Combined jitter plots/boxplots of mean cell gliding velocity (in μm.s⁻¹) of WT cells (n = 50), and strains expressing ectopically gldM^wt (n = 49) or gldM^E31A point mutant (n = 18) in a ΔgldM mutant background. The ΔgldM mutant has been placed in the graph for relevance but cell velocity has not been measured for that strain because it did not adhere to glass. Data underlying this figure can be found in S1 Data. (E) Combined jitter plots/boxplots of mean cell gliding velocity (in μm.s⁻¹) of WT cells (n = 50) and a gldL^E59A point mutant (n = 26). Strains were cultivated in CYE and single-cell gliding was observed on a free glass coverslip by phase contrast microscopy during 2 minutes. Gliding of the ΔgldL mutant and of the gldL^E49A point mutant was not measured because cells did not adhere to the glass substratum. Data underlying this figure can be found in S1 Data. (F) Localization and dynamics of SprB on the cell surface in the WT strain, a ΔgldL mutant, a gldL^E49A point mutant, and a gldL^E59A point mutant. A representative cell is shown. Strains were cultured in CYE and, after SprB immunolabeling, were sandwiched between an agarose pad and a glass coverslip to significantly limit cell movement and facilitate SprB signal acquisition and analysis. SprB was immunolabeled using a primary serum directed against SprB and Alexa-488 fluorescent secondary antibodies. Fluorescence was recorded with 100-ms intervals for several seconds. The brightfield image (top panel), the first frame (middle panel, in gray levels), and the rainbow trace of SprB motion over time (bottom panel, not available for the ΔgldL mutant and the gldL^E49A point mutant) are shown. Scale bar, 2 μm. (G) Combined jitter plots/boxplots of mean displacement velocity (in μm.s⁻¹) of SprB in WT cells (n = 69) and a gldL^E59A point mutant (n = 85). SprB fluorescent spots were detected and tracked over time (>2 seconds) using the Trackmate plugin. Data underlying this figure can be found in S1 Data. (H) GldM protease accessibility assay in WT F. johnsoniae, the ΔgldL mutant, and GldL^E49A and GldL^E59A point mutants. Spheroplasts were treated with (+) or not (−) with the trypsin protease. GldM was analyzed by SDS-PAGE and immunoblot with anti-GldM antibodies. The full-length GldM protein is indicated, as well as degradation products (* and **). The molecular mass markers (in kDa) are indicated on the left. DCCD, N-N’-Dicyclohexyl-carbodimide; T9SS, type IX secretion system; WT, wild-type.

https://doi.org/10.1371/journal.pbio.3001443.g004

Protonation of GldL glutamate residues.

To address the question whether GldL E49 and E59 residues undergo protonation and deprotonation cycles, we determined their pKa values. A ¹⁵N/¹³C Glu-labeled synthetic peptide corresponding to GldL TMH2 (L2, residues Val40 to Val61) was solubilized in deuterated dodecylphosphorylcholine (DPC) micelles and analyzed by nuclear magnetic resonance (NMR). pKa values of 5.54 ± 0.04 and 5.65 ± 0.13 for the carboxylic groups of residues E49 and E59, respectively, were measured by the pH-dependent chemical shifts in two-dimensional ¹³C-HSQC experiments (Fig 5A and 5B). In the presence of peptides corresponding to GldL TMH1 (L1, residues Lys6 to Thr29) and GldM TMH (M, residues Leu15 to Leu38), the pKa values slightly increased to 5.83 ± 0.02 for both glutamates (Fig 5B). The behavior of the ¹³C chemical shifts for the GldL TMH2 glutamate residues was then monitored in presence of the different peptides. At pH 5.2 (i.e., protonated glutamates), we observed chemical shift variations in the presence of L1, M, or both (Fig 5C). These data confirm that GldL TMH2 interacts with GldL-TMH1 and GldM-TMH and that the presence of these TMH peptides influences the environment of the glutamate residues. However, at pH 6.7 (i.e., deprotonated glutamates), no chemical shift was observed upon addition of the L1, M or both peptides (Fig 5C), suggesting that GldL TMH1 and GldM TMH are not in the environment of the glutamate residues. Taken together, these results suggest that the protonation state of the glutamic acids regulates contacts between TMH2 and the other TMHs in the GldLM complex and hence that GldLM helix organization is likely to be modified during motor function, as evidenced for the MotAB and TolQR motors [63,64].

Altogether, our results support a model in which GldL and GldM form an IM proton channel with conserved critical glutamates that are protonated and deprotonated in response to the proton gradient to power both T9SS-dependent secretion and gliding motility. Our results also demonstrate that the protonation state of GldL E49 controls changes within the GldLM TMHs packing that are likely transmitted to the GldM periplasmic domain.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. GldL glutamate residues protonation probed by NMR.

pKa determination of ¹³C-Glu of free and complexed L2 peptide. (A) Two-dimensional ¹³C-HSQC spectra of 1 mM L2 peptide (¹³C-Glu labeled) in 150 mM deuterated DPC in 50 mM phosphate buffers at different pH (pH 2.9 (yellow), 5.0 (orange), 5.8 (red), 6.7 (brown)). (B) The pH dependent chemical shift variations of C_γ carbons of E49 and E59 of the L2 peptide free or complexed with L1 and M in a 1:1:1 molar ratio, were measured, fitted, and apparent pKa values were calculated using the Henderson–Hasselbach equation. Data underlying this figure can be found in S1 Data. (C) Two-dimensional ¹³C-HSQC spectra of 1 mM L2 peptide (¹³C-Glu labeled) in 150 mM deuterated DPC in 50 mM phosphate buffer at pH 5.2 (left panel) and pH 6.7 (right panel), in the absence (brown) and presence at molar ratio 1:1 of GldL-TMH1 peptide (L1, orange), GldM-TMH peptide (M, blue), and both L1 and M peptides (green). DPC, dodecylphosphorylcholine; NMR, nuclear magnetic resonance.

https://doi.org/10.1371/journal.pbio.3001443.g005

GldLM motors and SprB adhesin do not have the same dynamics

SprB adhesins follow a closed right-handed helical track at the cell surface [5,6]. Two models have been proposed on how the T9SS controls SprB motion [3,4]. In the first model, fixed rotary motors may activate treads to which SprB adhesins are connected. In the second model, SprB adhesins are directly connected to moving GldLM motors. The first scenario requires a network of motor complexes along the SprB helicoidal path. The second scenario implies that SprB colocalizes with dynamic GldLM complexes and that SprB and GldLM move concomitantly along the helicoidal path. To explore these possibilities, we characterized the localization of the GldLM motor complex. Structured illumination microscopy (SIM) recordings of F. johnsoniae fixed and permeabilized cells immunolabeled with polyclonal primary antibodies against the GldL, GldM, GldK, or GldN proteins, and fluorescent secondary antibodies showed that each protein was distributed in many foci along the cell body (Fig 6A). For example, we numbered 33 ± 10 GldL foci per cell in average (n = 11, Fig 6A). These data suggest that multiple Gld motors decorate the cell envelope. A similar number of GldM foci were numbered (27 ± 5), in agreement with the assembly of functional GldLM complexes. For GldK and GldN, 13 ± 7 and 12 ± 4.5 foci were observed in average per cell, respectively, again in agreement with the formation of GldKN ring complexes. Based on the higher number of GldL/M foci as compared to GldK/N foci, one may hypothesize that GldLM complexes may exist in 2 subpopulations, one free and one associated with GldKN rings. To provide further information, we sought to perform live observations. However, none of the plasmid-borne or chromosomal fluorescent fusions to GldL or GldM we generated supported WT gliding, possibly due to the size of the fluorescent protein tags. We therefore turned to a more sophisticated method to generate functional and time-trackable proteins, using the alfa technology [65].

The alfatag is a 13–amino acid peptide that is specifically and almost irreversibly bound by the NBalfa nanobody with an affinity of approximately 0.26 pM [65]. The sequence encoding the alfatag was introduced in frame at the carboxyl terminus of the GldL-coding sequence at the native locus. GldL-alfa was functional and supported single-cell gliding (S3A and S3B Fig). In addition, no cleavage of the alfa tag was observed (S3C Fig). We then engineered a replicative plasmid expressing NBalfa-sfGFP under the control of an IPTG-inducible promoter in F. johnsoniae. Expression of NBalfa-sfGFP did not perturb cell gliding, either in a WT background or in a strain expressing GldL-alfa (S3B Fig). NBalfa-sfGFP was diffuse in WT cells that do not express GldL-alfa (S3D Fig). By contrast, NBalfa-sfGFP exhibited a punctate pattern in GldL-alfa cells (Fig 6B). Distinct foci were visible as well as more patchy signals, rendering quantification difficult. Remarkably, these foci were not all static relative to the cell, as opposed to the SprA translocon [26], nor did they behave like SprB adhesins that travel along the entire cell length. Time-lapse microscopy and kymograph analyses of signal dynamics indicated that some foci remain static while others moved quickly but at varying speed relative to the cell (Fig 6B, S1 Movie). In addition, fluorescence recovery after photobleaching (FRAP) experiments in nonmoving cells suggest that GldL-alfa movement was restricted to short distances within the cell because fluorescent signal could not be recovered over a large bleached cell region (Fig 6C and 6D). Based on these results and on the number of GldLM and GldKN foci per cell, we propose that these 2 GldL populations could correspond to GldLM complexes engaged into static complexes with the OM translocon to secrete substrates and to free GldL proteins or to GldLM complexes following a track to energize SprB motion. However, colocalization experiments in live cells with immunolabeled SprB showed that GldL-alfa and SprB do not follow the same trajectories (Fig 6E). These results support the tread model [66,67] in which GldLM proton channels convert the proton gradient into mechanical force to displace or activate treads involved in SprB movement (Fig 6F).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Dynamic localization of Gld complexes.

(A) Localization of Gld proteins in fixed cells. Immunostaining of GldL, GldM, GldK, and GldN and observation by SIM. A typical specificity control is shown on the right panel with a ΔgldM mutant stained with antibodies directed against GldM. Scale bar, 2 μm. The mean number and standard deviation of foci per cell is indicated on top (n > 10). (B) Live cell dynamics of a functional GldL-alfa fusion bound to NBalfa-sfGFP. NBalfa-sfGFP expression was induced with 1 mM IPTG for 1 hour prior to observation. GldL-alfa dynamics was then followed by time-lapse fluorescence microscopy with 300-ms intervals. For a representative cell, a stack of individual frames is shown. Time is indicated in ms. At the bottom of the stack, a kymograph of the fluorescence signal in the same representative cell. The x-axis is the position of GldL-alfa/NBalfa-sfGFP signal with respect to the substratum (glass), and the y-axis is time. The red arrow shows an example of a static signal, and the yellow arrow points to an example of a moving GldL-alfa focus. Scale bar, 2 μm. (C) GldL-alfa does not travel long distance within the cell. Diffusion of GldL-alfa/NBalfa-sfGFP fluorescence signal over time assayed by FRAP. Cells expressing GldL-alfa and NBalfa-sfGFP were pulse-bleached in the region indicated by the blue rectangle. A representative cell is shown. Scale bar, 2 μm. (D) Fluorescence intensities were measured in the bleached region (blue rectangle) and a nonbleached region (red rectangle) for 60 seconds in 4 cells. Individual measurements are shown with spots (filled spots for bleached regions and empty spots for nonbleached regions). The green arrow indicates the moment of bleaching. The fluorescence intensity of each region of interest was normalized to the first prebleached intensity. The blue (bleached region) and red (nonbleached) lines indicate the average of all measurements. Data underlying this figure can be found in S1 Data. (E) GldL-alfa and SprB do not colocalize. NBalfa-sfGFP expression was induced with 1 mM IPTG for 1 hour prior to observation. Then, dynamic localization of GldL-alfa bound to NBalfa-sfGFP (in green) and immunostained SprB (in red) in a representative cell. Fluorescence was followed by hilo microscopy with 500-ms intervals. A stack of individual frames is shown. Time is indicated in ms. At the bottom of the stack, a kymograph of the fluorescence signal in the same representative cell. The x-axis is the position of the cell with respect to the substratum (glass), and the y-axis is time. Scale bar, 1 μm. (F) Model of GldLM molecular motors function in type IX secretion (top) and surface adhesin dynamics (bottom). GldLM motors (blue) are fueled by the proton gradient (H+), leading to conformational shifts of the periplasmic domain of GldM. When associated with the T9SS (top), GldLM motors generate mechanical torque to rotate a GldKNO ring to drive secretion of SprB (red) through the SprA translocon. GldLM motors may also be associated with the gliding machinery (bottom), in which they serve to displace SprB on the cell surface via the displacement of a track or baseplate machinery (in green). FRAP, fluorescence recovery after photobleaching; SIM, structured illumination microscopy; T9SS, type IX secretion system.

https://doi.org/10.1371/journal.pbio.3001443.g006

Concluding remarks

In this study, we provided evidence that type IX secretion and surface adhesin motion are energized by a molecular motor fueled by the proton gradient, like the flagellar motor and other bacterial molecular motors. Our data support the idea that interactions between the TMHs of GldL and GldM shift in response to the proton flux, eventually leading to conformation changes in the GldM periplasmic domain. Conserved glutamate residues in GldL are important in this process but are not equivalent in terms of function. While amino acid E49 in GldL is essential for secretion through the T9SS as observed by Hennell and colleagues [42], glutamate at position 59 is only required for gliding, indicating that in F. johnsoniae, T9SS secretion itself does not require concomitant SprB motion along the cell surface. Thus, it is tempting to speculate that secretion and SprB motion are not supported by the same mechanical rearrangements in GldM or that SprB motion may require more mechanical torque than the secretion process. Furthermore, since SprB motion and secretion are uncoupled in the GldL E59A point mutant, this mutation is an interesting tool to study secretion independently of gliding in F. johnsoniae. Nevertheless, GldL-E59 is also conserved in nongliding bacteria like P. gingivalis, suggesting that it may also serve for T9 secretion in other bacteria (S2A Fig).

Our data also support the idea that GldM conformational shift upon PMF sensing could be converted into mechanical torque through the periplasmic part of the T9SS. Indeed, we showed that GldM periplasmic domain is connected to the T9SS GldKNO subcomplex, similar to the PorKLMN complex in P. gingivalis. Two recent studies help understand how this could work. First, the structure of the GldLM motor showed that 10 GldL TMHs (5 GldL molecules) wrap 2 GldM TMHs in an asymmetric manner [42]. By similarity to other MotAB-like motors, it was proposed that GldM TMHs would rotate within a GldL ring in response to the PMF to generate mechanical movement of GldM periplasmic domain. These findings are consistent with our data and provide an explanation for why GldL-E49 is required for motor function. However, they do not explain the role of E59, which is located outside the membrane in the GldLM structure (S2C Fig) [42]. One may hypothesize that E59 enters the proton channel when GldM rotates. Second, in situ PorKN rings were observed by cryo-electron tomography [59,68]. These rings may serve to maintain T9SS subcomplexes in close proximity to allow sequential translocation, maturation, and attachment of the substrates [68]. Therefore, an attractive hypothesis is that GldM conformational changes in response to the proton gradient could generate mechanical torque for the rotation of GldKNO rings, similar to cogwheels, which directly or indirectly facilitate secretion of T9SS substrates.

Finally, our results are consistent with the “rack and pinion” model proposed by Shrivastava and Berg to explain how the GldLM complex participates in SprB displacement [66]. Our microscopy data suggest the existence of static GldLM motors, which are presumably associated with static T9SS translocons, and GldLM complexes that are dynamic but that move differently than do SprB molecules. These motors could be linked to unidentified motion treads carrying SprB adhesins.

Bacterial strains, media, and chemicals

All strains are listed in S1 Table. Escherichia coli strains DH5α and BTH101 were used for cloning procedures and bacterial 2-hybrid assay, respectively. E. coli cells were grown in Lysogeny Broth, at 37°C or 28°C. For BACTH experiments, gene expression was induced by the addition of iso-propyl-β-D-thio-galactopyranoside (IPTG, Sigma-Aldrich, St. Louis, MI, USA, 0.5 mM), and plates were supplemented with 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-Gal, Eurobio, Les Ullis, France, 40 μg.mL⁻¹). F. johnsoniae CJ1827, a streptomycin-resistant rpsL2 derivative of ATCC 17061 (UW101), was used as model micro-organism. F. johnsoniae cells were grown at 28°C in Casitone Yeast Extract (CYE) medium [69] or Motility Medium (MM) [70] as indicated. For selection and maintenance of the antibiotic resistance, antibiotics were added to the media at the following concentrations: erythromycin, 100 μg.mL⁻¹; streptomycin, 100 μg.mL⁻¹; tetracycline, 20 μg.mL⁻¹, ampicillin, 100 μg.mL⁻¹, kanamycin, 50 μg.mL⁻¹, and chloramphenicol, 40 μg.mL⁻¹. Specific enzyme and chemicals source were as follows: trypsin (Sigma-Aldrich), CCCP (Sigma-Aldrich, 10 μM), nigericin (Nig, Sigma-Aldrich, 7 μM), valinomycine (Val, Sigma-Aldrich, 40 μM), sodium azide (Az, Sigma-Aldrich, 1.5 mM), arsenate (Ars, Sigma-Aldrich, 20 mM), and DCCD (Sigma-Aldrich, 100 μM).

Genetic constructs

All plasmids and oligonucleotide primers used in this study are listed in S1 Table. Enzymes for PCR and cloning were used as suggested by manufacturers.

Chromosomal mutants were generated as described [71]. The suicide plasmid designed to generate an in-frame deletion of gldM was built as follows. A 2.5-kb fragment containing the region upstream of gldM and gldM start codon was PCR amplified using oligonucleotide primers F1-ΔgldM and R1-ΔgldM. This fragment was digested with BamHI and SalI and inserted into plasmid pRR51 cut with the same restriction enzymes to generate an intermediate plasmid. A 2.5-kb fragment containing gldM stop codon and the region downstream of gldM was PCR amplified using oligonucleotide primers F2-ΔgldM and R2-ΔgldM. Similarly, it was digested with SalI and SphI and inserted into the previously generated plasmid cut with SalI and SphI to generate plasmid pRR51-ΔgldM.

The suicide plasmids designed to build gldL^E49A and gldL^E59A strains were constructed as follows. A plasmid with plasmid pRR51 backbone and carrying a 4-kb region centered around gldL E49A and E59A codon substitutions was synthesized (Geneart, Thermo Fisher Scientific, Waltham, MA, USA). The suicide plasmid pRR51-gldL^E49A was then built by restoring the codon for E59 in the previously synthesized plasmid by site directed mutagenesis with oligonucleotide primers Fw-GldL-A59E and Rv-GldL-A59E. Similarly, the suicide plasmid pRR51-gldL^E59A was built by restoring the codon for E49 in the previously synthesized plasmid by site directed mutagenesis with oligonucleotide primers Fw-GldL-A49E and Rv-GldL-A49E.

The suicide plasmid used to generate the GldL-alfa fusion was made as follows. A 1.5-kb fragment containing gldL and part of the alfa tag was PCR amplified using oligonucleotide primers oTM582 and oTM590. A 1.5-kb fragment containing part of the alfa tag, a stop codon, and the region immediately downstream of gldL stop codon was PCR amplified using oligonucleotide primers oTM591 and oTM592. These fragments were assembled using Gibson isothermal reaction and reamplified using oligonucleotide primers oTM582 and oTM592. This fragment was digested with BamHI and SphI and inserted into pRR51 cut with the same restriction enzymes. Plasmids for GALLEX and BLA were engineered by hybridizing complementary oligonucleotides corresponding to the GldL or GldM TMHs and inserting them into NheI-BamHI-digested target GALLEX or BLA vectors. BACTH plasmids were engineered by restriction and ligation as previously described [23]. The replicative plasmid designed for complementation and production of GldLM^WT were constructed as follows. A fragment containing gldL and gldM open reading frames was PCR amplified using oligonucleotide primers 5-BamHI-LM and 3-XbaI-LM. This fragment was digested with BamHI and XbaI and inserted into plasmid pCP11 [72] cut with the same restriction enzymes. The replicative plasmid designed to produce GldM^E31A was then generated by quick change site directed mutagenesis using oligonucleotide primers 5-GldM-E31A and 3-GLdM-E31A.

The replicative plasmid designed to express NbAlfa-sfGFP from an IPTG inducible promoter in F. johnsoniae was designed as follows. A first replicative plasmid was built with an IPTG-inducible promoter, a multicloning site and lacI constitutive expression for repression in the absence of IPTG in F. johnsoniae. A fragment containing the promoter of Fjoh_0697 [73] with lacO3 and lacO1 operator sites flanking the −33 and −7 promoter sequences, pCP23 multicloning site and Fjoh_0139 promoter after the PstI restriction site was synthesized (Geneart, Thermo Fisher Scientific). A fragment containing lacI open reading frame was PCR amplified using oligonucleotide primers oTM495 and oTM496. These fragments were assembled by Gibson isothermal reaction and reamplified using oligonucleotide primers oTM497 and oTM496. This fragment was digested with KpnI and SphI and inserted into pCP23 [69] cut with the same restriction enzymes to generate plasmid pCP-lac. Then, the gene encoding NbAlfa was synthesized (Geneart, Thermo Fisher Scientific) and reamplified using oligonucleotide primers oTM612 and oTM596. sfGFP, codon-optimized for translation in F. johnsoniae, was also synthesized (Geneart, Thermo Fisher Scientific) and then reamplified by PCR using oligonucleotide primers oTM595 and oTM593. These fragments were assembled using the Gibson isothermal reaction and reamplified using oligonucleotide primers oTM612 and oTM593. It was then digested with BamHI and NheI and inserted into plasmid pCP-lac cut with the same restriction enzymes.

Protein interaction assays

The adenylate cyclase-based bacterial 2-hybrid technique was used as previously published [23]. Briefly, the proteins to be tested were fused to the isolated T18 and T25 catalytic domains of the Bordetella adenylate cyclase. After introduction of the 2 plasmids producing the fusion proteins into the BTH101 reporter strain, plates were incubated at 28°C for 24 hours. Three independent colonies for each transformation were inoculated into 600 μL of LB medium supplemented with ampicillin, kanamycin, and IPTG (0.5 mM). After overnight growth at 28°C, 10 μL of each culture was spotted onto LB plates supplemented with ampicillin, kanamycin, IPTG, and X-Gal and incubated at 28°C. Controls include interaction assays with TolB and Pal, 2 protein partners unrelated to the T9SS. The experiments were done in triplicate and a representative result is shown.

GALLEX and BLA were performed as described [55].

Protease susceptibility assay

F. johnsoniae cells were grown in 5 mL of CYE medium to an A₆₀₀ = 0.8, harvested by centrifugation and resuspended in 100 μL of 20 mM Tris-HCl pH 8.0, 20% sucrose, 1 mM EDTA, and 100 μg.mL⁻¹ of lysozyme. After 30-minute incubation at room temperature (20°C), 100 μL of ice-cold sterile water was added and the mixture was carefully mixed by 3 inversions. Moreover, 50 μL of each spheroplast suspension was treated with trypsin (100 μg.mL⁻¹). After 5 minutes on ice, 17 μL of boiling 4× Laemmli loading buffer was added and immediately vortexed and boiled for 5 minutes prior to SDS-PAGE and immunoblot.

Western blot analyses

F. johnsoniae cells were grown to mid-log phase in CYE at 28°C. Whole cells were prepared for SDS-PAGE and western blotting assays were performed as previously described. Equal amounts of total proteins were loaded for each sample based on culture optical densities. Anti-GldL, anti-GldM [8], and anti-FLAG (Sigma-Aldrich, clone M2) antisera were used at 1/5,000, 1/5,000, and 1/10,000 dilutions, respectively.

Nuclear magnetic resonance

NMR experiments were carried out on a Bruker Avance III 600 MHz spectrometer, at 300 K. Three synthetic peptides L1 (GldL-TMH1: KKVMNFAYGMGAAVVIVGALFKITKK), L2 (GldL-TMH2: KKVMLSIGLLTEALIFALSAFEPVKK), and M (GldM-TMH: KKLMYLVFIAMLAMNVSKEVISAFGLKK), with ¹⁵N/¹³C-Glu-labeled, have been studied free and in complexes at molar ratio 1:1 or 1:1:1. NMR samples containing 1 mM peptide concentration in 150 mM deuterated DPC were used in different phosphate buffers (50 mM). The behavior of the ¹³C chemical shifts for glutamate residues in the different peptides as a function of pH (2.9 to 8.9) was monitored using a two-dimensional ¹³C-HSQC experiment. Chemical shift values as a function of pH were analyzed according to a single titration curve of the form where δ is the observed chemical shift at a given pH, δ_HA and δ_A are the chemical shifts for the various protonated forms of the peptide, and n is the number of protons transferred.

Fluorescence microscopy and image analysis