The TolQRA knockout strain fails to undergo proper cell division and displays a characteristic multi-segmented, chain-like morphology under low osmotic conditions13,14. This phenotype can be suppressed by overexpression of soluble PG hydrolase13, highlighting the role of TolQRA in coordinating OM invagination and PG remodeling. However, the mechanism by which TolQRA mediates transient tethering between the cell wall and the OM remains unclear34. In this study, we captured cryo-electron microscopy (cryo-EM) structures of the TolQRA complex in multiple conformational states. Structural comparisons allowed us to investigate the rotary motion of TolQR driven by PMF. Combined with molecular dynamics (MD) simulations of the proton conduction pathway within TolQR and functional analysis of the periplasmic domains of TolRA, we proposed a model for proton-induced mechanical transduction by the Tol-Pal system. Additionally, key residues crucial for the function of TolQ, TolR, and TolA were identified through cell viability and morphological analysis. These findings provide critical insights that could inform the development of novel antimicrobial strategies.

Previous live fluorescent microscopy results suggest that TolA is dynamically recruited to the cell septum via TolQR. However, the assembly of the complex has not been directly visualized. To address this, we co-expressed TolQR and TolA from Escherichia coli (strain K12), purified, and determined the structure of the TolQRA complex reconstituted in phospholipid nanodiscs by cryo-EM (Supplementary Figs. S1b, S2 and Table S1). The transmembrane portion of the complex, including TolQ and the transmembrane helices of TolR and TolA, is well resolved at 3.52 Å resolution. Densities for the periplasmic domains of TolA and TolR were obtained at lower resolution due to inherent flexibility (Fig. 1b and Supplementary Fig. S2). These regions were subsequently modeled through AlphaFold3 prediction. TolQ forms a pentameric structure similar to that of MotA and ExbB, exhibiting a volcano-like configuration with a narrow periplasmic end and a wider cytoplasmic base (Fig. 1b). The pentameric TolQ encapsulates the dimeric TolR through its N-terminal transmembrane helices at the periplasmic end. Each TolQ monomer comprises three transmembrane helices (TM1-TM3). Adjacent monomers interact through their TM2 and TM3 helices to assemble a hollow central core composed of ten TM helices, while TM1 is positioned peripherally, interfacing with the TM2-TM3 junction (Fig. 1b). The periplasmic ends of the central pore exhibit a slight clockwise twist when viewed from the periplasm, while the cytoplasmic ends remain largely perpendicular to the membrane plane. The N-terminus of TM1 extends into a short periplasmic helix (helix 1) that lies parallel to the membrane, and the linker between TM1 and TM2 forms two cytoplasmic helices (helix 2 and helix 3) oriented obliquely at the outer base of the central core (Fig. 1b).

In the TolQRA nanodisc complex, we identified three major reconstructions showing up to three TolA molecules associated with TolQ subunits TolQ1, TolQ3, and TolQ4, respectively (Fig. 1b and Supplementary Fig. S2). TolA interacts with TolQ through their transmembrane helices (TolA_I) in a crossing configuration, stabilized by hydrogen bonds between TolQ_I29 and TolA_S18, TolQ_S28 and TolA_H22, and TolQ_N5 and TolA_S33 (Fig. 1c). These interactions are consistent with the conserved SHLS motif shared by TolA and its homolog TonB. MD simulations revealed that the hydrogen bonds between TolQ_I29 and TolA_S18 and TolQ_S28 and TolA_H22 are stable (Supplementary Fig. S1d). Site-directed mutagenesis disrupting the TolA_H22A-TolQ_S28A interaction impaired bacterial viability in the presence of the membrane-permeable agent sodium deoxycholate (DOC) (Fig. 1d), indicating that this interaction is vital for proper Tol complex function. Despite this, the variant proteins can still be co-purified as a complex (Supplementary Fig. S1c), suggesting that TolA and TolQR interact through additional interfaces, likely involving the periplasmic domain of TolR (Fig. 1b).

The pentameric TolQ structure exhibits a hydrophobic region near the periplasmic end, delineating the transmembrane region, while the remaining surface is predominantly hydrophilic (Supplementary Fig. S3a). The internal cavities correspond to these hydrophobic and hydrophilic regions. TolR engages with TolQ within the hydrophobic cavity through the TolR transmembrane helices (TolR_TMs) (Supplementary Fig. S3a). The TolR_TMs dimerize through a leucine-rich motif (L21LDVLLVLLL30) (Fig. 2a). Substitution of these leucine residues with alanine resulted in cell death in the presence of DOC (Fig. 2c), highlighting their functional importance. Notably, the conserved charged residue D23 within this motif is shared among homologous prokaryotic motor proteins and is proposed to play a role in proton binding and transfer. The asymmetric architecture of TolQ leads to an uneven spatial arrangement of its pentameric subunits relative to the central TolR axis. While D23 from TolR chain A (TolR(A)_D23) points toward the interface between TolQ2 and TolQ3, TolR(B)_D23 forms hydrogen bonds with T145 and T178 of TolQ5 (Fig. 2b and Supplementary Fig. S3b). T145 and T178 from the TolQ pentamer form a polar ring within the hydrophobic transmembrane cavity. This striking structural feature, conserved in both TolQ and ExbB, likely facilitates proton conduction into the otherwise nonpolar environment. Site-directed mutagenesis of TolR_D23A, TolQ_T145A, and TolQ_T178A resulted in bacterial death in the presence of DOC and exhibited filamentous phenotypes, indicating that these residues are essential for function (Fig. 2c and Supplementary Fig. S3c, d). In MotAB, protonation of MotB_D22 in chain A and chain B alternately stabilizes its interaction with T189 in each corresponding MotA subunit, with each proton uptake event driving a 36° stepwise rotation of the complex. Similarly, our TolQRA structure reveals an analogous asymmetry: TolR_D23 in chain A and chain B adopt distinct conformations relative to TolQ, positioning them for protonation-dependent rearrangements reminiscent of the MotAB mechanism.

To investigate the proton transport path within the TolQRA complex, we conducted MD simulations around the putative proton acceptor TolR_D23 (Fig. 2d and Supplementary Fig. S3e). The threonine-interacting TolR(B)_D23 was fixed in the protonated state, and TolR(A)_D23 was simulated in either protonated or unprotonated forms. During microsecond simulations, a single-file water chain was observed from the periplasmic side between the TM2 and TM3 helices of TolQ3, approaching the unprotonated TolR(A)_D23 (Fig. 2d and Supplementary Fig. S3e). This linear water chain likely mediates proton translocation via a Grotthuss-like mechanism. In contrast, under the simulated protonation state of TolR(A)_D23, water occupancy around TM2 and TM3 of TolQ3 was markedly reduced. These results suggest that the electronegativity of TolR(A)_D23 is important for the formation of a proton-conductive pathway. Along the water channel, we also identified a polar residue TolQ_E173 at the periplasmic entrance of the channel (Fig. 2d). Its negatively charged side chain projects into the periplasm, potentially serving as a proton attractor for its translocation. Substitution of E173 with alanine (TolQ_E173A) caused bacterial lethality and disrupted cell division, highlighting the functional significance of this residue (Fig. 2c and Supplementary Fig. S3d).

To explore the dynamic motion of the TolQRA complex, we aimed to capture its distinct states using styrene-maleic acid (SMA) polymers. SMA enables the extraction of membrane proteins directly from their native lipid bilayer, preserving the complex in its biological state. Two distinct structural classes with high resolutions were determined at 2.92 and 3.28 Å, respectively, with one class lacking density for TolA (hereafter TolQR) and the other displaying density corresponding to one copy of TolA (hereafter TolQRA) (Fig. 3a and Supplementary Figs. S4, S5a, and Table S1). The TolA molecule bound in TolQRA is positioned consistently with its location on the TolQ1 subunit in the nanodisc TolQRA structure. The superimposition of the TolQR and TolQRA structures onto the TolQRA structure shows root-mean-square deviation (RMSD) values of 1.31 and 2.52 Å, respectively, based on the alignment of 217 Cα atoms (Fig. 3b and Supplementary Fig. S5b).

The superimposition of these structures, utilizing the TolR dimer as a reference, revealed that the TolQR structure slightly rotated compared to TolQRA. The morphing transition between TolQRA and TolQR demonstrates a prominent shutter-like rotation in the cytoplasmic termini of TolQ in relation to the central TolR dimer (Supplementary Videos S1, S2). The prominent conformational changes at the cytoplasmic ends of the motor unit indicate high dynamics in these regions. When viewed from the cytoplasmic side of the pentamer, TolQRA features an oval-shaped cytoplasmic opening, TolQR presents a round opening, and TolQRA displays a split between TolQ1 and TolQ5 as well as between TolQ3 and TolQ4 (Fig. 3a). Conformational variation primarily arises from the cytoplasmic half of the TM2 helix (Supplementary Fig. S6a), which contains three arginine residues (R110, R113, and R118). These residues form salt bridges with glutamic acid residues (E211 or E212) and hydrogen bonds with H218 in TM3 of adjacent TolQ subunits (Fig. 3c and Supplementary Fig. S6a, b). Substitution of these arginines with glutamic acid (R110E/R113E/R118E) disrupted the interactions and impaired bacterial viability (Fig. 3d and Supplementary Fig. S6c). Interestingly, introducing compensatory charge-reversal mutations in TM3 (E211R/E212R) restored viability when combined with the TM2 arginine substitutions (R110E/R113E/R118E) (Fig. 3d), indicating that electrostatic interactions between TM2 and TM3 are essential for TolQ function, likely by stabilizing the channel architecture and preventing proton leakage. Along the cytoplasmic end of TM3, additional negatively charged residues, E203 and D207, extend their side chains toward the channel lumen without forming direct interactions (Fig. 3c). Mutational analysis revealed that substitution of these residues (E203R, D207R, and E211R) significantly impaired bacterial viability, highlighting their functional importance (Fig. 3d and Supplementary Fig. S6c). In the homologous MotA protein, the acidic cytoplasmic region has been proposed either to serve as a proton reservoir or to mediate interactions with the N-terminus of MotB. Our result supports the former hypothesis, as deletion of the TolR N-terminus (TolRΔ2-10) had no detectable effect on cell viability (Fig. 3d and Supplementary Fig. S6c).

To investigate how D23 protonation in TolR influences conformational dynamics during TolQRA rotation, we performed MD simulations on different protonation states of the SMA-captured intermediate TolQR structure (Supplementary Fig. S7b). Protonation of D23 in both TolR(A) and TolR(B) enhanced structural stability, as reflected by consistent TolR_D23-TolQ_T178 distances in the simulations (Supplementary Fig. S7b). These results suggest that D23 protonation promotes conformational stabilization following a rotation step initiated by proton transduction at the previously deprotonated D23 site. Notably, stabilization of the TolR_D23-TolQ_T178 interaction correlated with a more consistent positioning of the resolved N-terminal region of TolR(A) (residues 14-17) relative to TolQ1-5 (Supplementary Fig. S7a, b), implying a structural link between D23 protonation and TolQ rotational motion. Further structural analysis revealed that TolR residues S14, I16, and N17 form contacts with TolQ residues G131 and Y139 (Supplementary Fig. S7a). Substitution of Y139 with alanine did not affect cell viability (Fig. 2c), whereas replacing it with a charged residue was lethal (Supplementary Fig. S5c), indicating that the hydrophobic character of Y139 is functionally essential, possibly to prevent proton leakage during transduction. Together, these findings, supported by MD simulations and structural morphing analysis, suggest a rotary mechanism analogous to that of the MotAB complex, but with distinct features of proton translocation.

In the cryo-EM map, we also observed a satellite periplasmic density located ~60 Å above TolQ5 and TolQ1, which likely corresponds to the periplasmic domains of TolR (TolR_PDs) (Fig. 4a). Previously reported crystal structures of the isolated TolR_PDs from E. coli (PDB: 5BY4) and H. influenzae (PDB: 2JWK) showed two distinct conformations: one with a flat interface and the other with a bent interface (Supplementary Fig. S8a). It has been proposed that the TolR_PD dimer undergoes "strand-swapping" remodeling upon activation, in which one monomer rotates 180°, converting the dimer from C2 symmetry to an asymmetrical architecture (Supplementary Fig. S8b). This remodeled interface has been experimentally shown to mediate PG binding and resembles the dimeric structures of other PG-binding proteins such as Pal, MotB, and ExbD. Although the resolution of the periplasmic density is relatively low, its flat shape and the sub-PG height imply that these TolR_PDs are in their pre-remodeled conformation (Fig. 4a and Supplementary Fig. S8c). MD simulations of the dimeric TolR_PDs reveal a restricted range of motion, with rotational flexibility limited to less than 360° (Fig. 4b). This motion is confined near TolQ1 and TolQ5, indicating that the TolR_PDs are anchored at that location, likely due to the interactions with the periplasmic loop connecting domains I and II of TolA1. A segment of this TolA1_loop is resolved in the cryo-EM map, protruding into the periplasmic space from the interface between helices 1 of TolQ1 and TolQ5, aligning with the TolR_PDs (Fig. 4a). Consequently, we predicted TolR_PDs-TolA1_loop structure using AlphaFold3, which revealed structural features resembling the crystal structure of the pre-remodeled TolR_PD dimer (PDB: 5BY4), supporting our interpretation of the observed periplasmic density (Fig. 4a).

Each monomer of the TolR_PDs contains six β-strands (β1-β6) and three α-helices (α1-α3) (Fig. 4c, d). Studies of homologous proteins suggest that strands β1 and β6 are involved in TolR dimerization, and the remodeling of the TolR_PDs is achieved through rearrangement of the dimerization interface from β6/β1'/β1/β6' to β5/β5' (Fig. 4d). In our structure, the TolR_PD dimer exhibits C2 symmetry (Fig. 4d and Supplementary Fig. S8a). The N-terminal β1 and α1 from both TolR_PD monomers are shown as unfolded loops (Fig. 4e). The remaining strands β2-β6 form β-sheets, and the two α-helices (α2, α3) are positioned on the same side (Fig. 4d, e). While β6 of TolR chain A is disordered, β6 of TolR chain B interacts with the periplasmic TolA_loop through a β-strand (residues 52-55, TolA_β hereafter) at the dimeric interface (Fig. 4e, f). Functional assays showed that deletion of β6 (TolR) or β1-α1 (TolR) of TolR resulted in cell death in the presence of DOC (Fig. 4c, g). Furthermore, the TolR disrupted the complex assembly between TolA and TolQR, highlighting the importance of TolR_β6 in maintaining the interaction between TolA and TolR. Deleting either β1 (TolR) or α1 (TolR) of TolR independently was also lethal (Fig. 4g and Supplementary Fig. S9a, b). Interestingly, replacing the deleted β1 or α1 with an equally long alanine linker restored the lethality effects caused by TolR but not by TolR (Fig. 4g and Supplementary Fig. S9a, b). This suggests that TolR_α1 likely functions as a linker, but TolR_β1 plays a role that requires residue specificity.

At the dimeric interface, TolR_β6 interacts with TolA_β through backbone hydrogen bonds between TolR_L137 and TolA_V55 and between TolR_V135 and TolA_V53 (Fig. 4f). Cysteine mutations at these residues crosslinked the complex to a higher molecular weight (Supplementary Fig. S9b), validating the predicted interface between TolR_β6 and TolA_β. Preventing hydrogen bond formation at TolR_L137 by substituting it with the kinky proline (TolR_L137P) resulted in cell death in the presence of DOC (Fig. 4h and Supplementary Fig. S9a, b), suggesting that interactions between TolR_β6 and TolA_β1 are critical for mediating the transduction role of the Tol system. Moreover, crosslinking of this site also caused lethality (Fig. 4h and Supplementary Fig. S9a, b), suggesting that the interaction needs to be transient for proper functionality.

In the predicted structure, the periplasmic loop of TolA adopts a stretched configuration, while the loop of TolR is compressed (Fig. 4e). This suggests that the TolR_PDs could potentially extend further, but in the current conformation, their extension is restricted by the length of the TolA loop. In the crystal structure of activated TolR_PDs (PDB: 2JWK), β6 of TolQ_PDs is absent from the dimeric interface (Supplementary Fig. S8a). We speculate that rotation of the motor unit drives the elevation of the TolR_PDs toward the PG layer, and this upward movement facilitates interface remodeling, as the tethered TolR_β6 strand is pulled away from the dimer interface by TolA_β. Supporting this model, covalent crosslinking β6 to β5 at the TolR interface led to bacterial cell death in the presence of DOC, indicating that dissociation of TolR_β6 from the interface is essential for function (Fig. 4i and Supplementary Fig. S9a, b). This remodeling likely results in the formation of a new β5/β5 interface and a grooved, eight-stranded β-sheet that mediates PG binding. Notably, TolR lacking both β1/α1 and β6 (TolR) retained the affinity with TolA regardless of the presence of the periplasmic loop TolA_β (Supplementary Fig. S9c). This implies that the remodeled β5/β5 TolR_PD dimer also associates with TolA but through a region other than TolA_β. This observation is consistent with the idea that the remodeled TolR_PD dimer, upon associating with the PG layer, may provide mechanical support to the rigid helical domain II of TolA, enabling it to traverse the PG layer and engage TolB. Supporting this notion, a previous study demonstrated that replacing TolA_II with a flexible loop impaired its function.