Cryo-EM structures of a MexB–MexY chimeric efflux pump reveal that large open clefts are intrinsic to the MexY porter domain

RND-type multidrug-efflux pumps are major contributors to multidrug resistance in Gram-negative bacteria, with MexY from Pseudomonas aeruginosa playing a central role in aminoglycoside resistance. Unlike other RND transporters, MexY exhibits unusually large open clefts in the binding and extrusion states. To determine whether this feature is intrinsic to its drug-recognition porter domain, we created a chimeric protein, MexBYB, by replacing the funnel-like and transmembrane domains of MexY with those of the homologous transporter MexB, and determined its structures by cryoEM under apo and kanamycin-supplemented conditions. Under both conditions, MexBYB was reported to adopt symmetric-like and asymmetric conformations. Structural comparisons reveal that the unusually large open clefts are retained in MexBYB, indicating that this feature is intrinsic to the MexY porter domain.

1. Introduction

In recent years, the global increase in multidrug-resistant bacteria that exhibit resistance to multiple classes of antibiotics has become a serious public health concern (Tacconelli et al., 2018). These pathogens often acquire cross-resistance to major clinical antibiotics, including β-lactams, aminoglycosides and fluoroquinolones, leading to treatment failure, severe complications and increased mortality. Among such pathogens, Pseudomonas aeruginosa is a major cause of nosocomial infections and is known for its remarkable environmental adaptability and diverse resistance mechanisms, making it one of the most problematic species in clinical settings (Lister et al., 2009; Pang et al., 2019). One of the major contributors to its drug resistance is the overexpression of resistance–nodulation–cell division (RND)-type multidrug-efflux complexes composed of three proteins: an outer membrane factor (OMF), a membrane-fusion protein (MFP) and an RND transporter (Li et al., 2015).

RND transporters assemble into trimers in the inner membrane and are responsible for drug selection and efflux, using the proton motive force as the energy source (Nikaido & Pagès, 2012). They consist of three major structural regions: the funnel-like (FL) domain, which stabilizes the trimeric architecture, the porter domain, which recognizes and binds drugs, and the transmembrane (TM) domain, which anchors the RND to the inner membrane and mediates proton translocation through the inner membrane. The porter domain consists of four subdomains: PN1, PN2, PC1 and PC2. Among them, PC1 and PC2 form the entrance, whereas PN1 and PN2 form the exit (Murakami et al., 2002).

Studies of the RND transporter AcrB from Escherichia coli have revealed that the three protomers adopt distinct conformations corresponding to the access, binding and extrusion states (Murakami et al., 2006). These conformational states interchange sequentially among the protomers, establishing the functional rotation mechanism that leads to drug efflux. In the access-state protomer, PC1 and PC2 separate from each other to create an open cleft, while the PN1–PN2 domains remain closed. This conformation prepares the protomer for substrate recognition. In the binding-state protomer, both the PC1–PC2 and PN1–PN2 domains open to form open clefts, allowing the substrate to interact with the drug-binding pocket. In the extrusion-state protomer, the cleft between PC1 and PC2 is closed and the central helix shifts outwards, enabling the substrate to be expelled.

Among the RND transporters of P. aeruginosa, MexY is the only complex known to confer resistance to aminoglycoside antibiotics and is an important target for antimicrobial drug development (Morita et al., 2012). Recently, an apo-state MexY trimer structure has been reported at 3.63 Å resolution (PDB entry 9e9f; Gregor et al., 2025). Unlike that of AcrB, the porter domain of MexY has been described in resting, binding and extrusion states. Unlike the protomers in other functional states, the resting-state protomer has both the PC1–PC2 and PN1–PN2 domains closed, forming no open cleft. By contrast, in the binding and extrusion states of MexY homologs, the open clefts of MexY are unusually large (Supplementary Figs. S1, S2 and S3). To determine whether this is a characteristic feature of its drug-recognition porter domain, we created a chimeric protein, MexBYB, by replacing the FL and TM domains of MexY with those of the homologous RND transporter MexB.

In this study, we determined cryo-EM structures of MexBYB in the absence (apo state) and presence of kanamycin (Ka-supplemented state) and performed a comparative structural analysis. Here, we present two different conformational structures, an asymmetrical structure and a symmetry-like structure, in the apo state and the Ka-supplemented state. Here, by elucidating the conformational features of the porter domain in these structures, our work provides new insight into the architecture of the porter domain of MexY.

2. Methods

2.1. Construction of the MexBYB plasmid

The chimeric protein MexBYB was constructed by fusing the FL domain and TM domain of MexB (residues 1–39, 182–273, 351–556, 581–585, 723–802 and 860–1043) with the porter domain of MexY (residues 40–181, 274–350, 557–580, 586–722 and 803–859). The plasmid pUC118mexBYB with a 6×His-tag at the C-terminus was extracted from the pUC118 vector using HindIII and NdeI, and the resulting fragment was ligated into the pET-21b(+) vector at the HindIII and NdeI sites. Transformants were propagated in XL-10 Gold cells, and the resulting expression plasmid was purified using a Monarch Spin Plasmid Miniprep Kit (New England Biolabs).

2.2. Expression and purification of MexBYB

The plasmid pET-21b(+)mexBYB was transformed into E. coli C43(DE3) Condon Plus RIPL cells. The bacterial cells were cultured in Terrific Broth (TB) medium supplemented with 100 mg ml⁻¹ ampicillin in a jar fermenter at 37°C and 270 rev min⁻¹. When the optical density at 600 nm (OD₆₀₀) reached 0.45, the temperature was lowered to 20°C and protein expression was induced with 1.2 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). The cells were then harvested within 18 h of induction. The collected bacterial cells were resuspended in lysis buffer (50 mM sodium phosphate pH 6.5, 100 mM NaCl, 2 mM MgCl₂, 10 µg ml⁻¹ DNAse I) and disrupted with a French press (SMT Co. Ltd) at 90 MPa four times. The membrane fraction was collected by ultracentrifugation and washed four times with high-salt buffer (50 mM sodium phosphate pH 6.5, 2 M NaCl). The washed membrane was resuspended in freezing buffer (50 mM sodium phosphate pH 6.5, 300 mM NaCl, 15% glycerol), flash-frozen using liquid nitrogen and stored at −80°C. The membrane was solubilized in 3%(w/v) n-dodecyl-β-D-maltoside (DDM), 10 mM imidazole pH 6.5. Insoluble material was removed by ultracentrifugation at 30 000 rev min⁻¹ for 30 min at 4°C (45 Ti rotor, Beckman). The supernatant was subjected to Ni–NTA column chromatography (Qiagen) in an Econo-Column and the resin was then washed with wash buffer (50 mM sodium phosphate pH 6.5, 300 mM NaCl, 100 mM imidazole, 0.0348% DDM, 0.0007% PE [E. coli total extract lipid)]. The purified MexBYB was eluted with elution buffer (50 mM sodium phosphate pH 6.5, 300 mM NaCl, 250 mM imidazole, 0.0348% DDM, 0.0007% PE), concentrated using Spin-X 20 ml concentrators (100 000 molecular-weight cutoff) and subjected to a Superdex 200 Increase 10/300 GL column (Cytiva) in size-exclusion chromatography (SEC) buffer (20 mM sodium phosphate pH 6.5, 150 mM NaCl, 0.0348% DDM, 0.0007% PE). Peak fractions were concentrated to 3 mg ml⁻¹.

2.3. MSP nanodisc preparation

To assemble MexBYB into nanodiscs, a mixture consisting of 15.9 µM MexBYB, 31.8 µM membrane-scaffold protein 1E3D1 (MSP1E3D1) and 795 µM PE was incubated at room temperature for 15 min. 8.8 mg ml⁻¹ prewashed Bio-Beads (Bio-Rad) were added to the mixture and incubated for 1 h at 4°C to remove the DDM detergent. The Bio-Beads were then supplemented to adjust the final concentration to 46.7 mg ml⁻¹. The resultant mixture was then incubated overnight at 4°C. The protein–nanodisc solution was filtered through 0.22 µm nitrocellulose filter tubes to remove the Bio-Beads. The filtered protein–nanodisc solution was further purified using a Superdex 200 Increase 10/300 GL column (Cytiva) in SEC buffer without DDM. Fractions corresponding to the size of the trimeric MexBYB–nanodisc complex were collected for cryo-EM sample preparation.

2.4. Mass photometry

Proteins obtained from the peak fractions in SEC were measured. Before measurement, the proteins were diluted to a final concentration of 5–10 nM in SEC buffer without detergent. Mass photometry was performed with a Refeyn OneMP mass photometer (Refeyn Ltd, Oxford, UK) and Refeyn AcquireMP (version 2.3), using the default parameters set by Refeyn AcquireMP. Aldolase and thyroglobulin were used as controls to determine the molecular weight. The results were subsequently analyzed using Refeyn DiscoverMP (version 2.3) and graphs were prepared to visualize the data.

2.5. Cryo-EM sample preparation

To obtain the two data sets for MexBYB in the absence of kanamycin (apo state), 2.5 µl MexBYB–nanodisc solution (1.08 mg ml⁻¹ MexBYB, 20 mM sodium phosphate pH 6.5, 150 mM NaCl, 0.0007% PE) and 3 µl desalted MexBYB–nanodisc solution (1.1 mg ml⁻¹ MexBYB, 20 mM sodium phosphate pH 6.5, 0.0007% PE) were applied onto a glow-discharged holey carbon film (Quantifoil R1.2/1.3R, 300 mesh, copper grid). The grids were blotted for 2.5 s at 7.4°C and 97% humidity and for 2.5 s at 4°C and 100% humidity, respectively,and were then plunge-frozen in liquid ethane using a Vitrobot Mark IV (Thermo Fisher Scientific, Oregon, USA).

To obtain the two data sets for MexBYB in the presence of kanamycin (Ka-supplemented state), the MexBYB–nanodisc sample was supplemented with kanamycin to a final concentration of 10 mM and the mixture was incubated on ice for 30 min before grid preparation. Subsequently, 3 µl MexBYB–nanodisc solution with kanamycin (1.63 mg ml⁻¹ MexBYB, 20 mM sodium phosphate pH 6.5, 150 mM NaCl, 0.0007%PE, 10 mM kanamycin) and 3 µl of the desalted MexBYB–nanodisc solution with kanamycin (2.2 mg ml⁻¹ MexBYB, 20 mM sodium phosphate pH 6.5, 0.0007% PE, 10 mM kanamycin) were applied to glow-discharged holey carbon grids (Quantifoil Cu R1.2/1.3, 300 mesh). The grids were blotted for 2.5 s at 4.7°C and 88.3% humidity and for 2.5 s at 4°C and 100% humidity, respectively, followed by plunge-freezing in liquid ethane using a Vitrobot (Thermo Fisher). The grids were then transferred into cartridges before data collection.

2.6. Data collection

The data sets for MexBYB were collected at 81 000× magnification, corresponding to a pixel size of 0.675 Å per pixel, at 300 kV on a Titan Krios (Thermo Fisher Scientific, Oregon, USA) equipped with an FEI Falcon III direct detector. For the apo-state data sets, micrographs were acquired using SerialEM with a total dose of 50 e⁻ Å⁻², corresponding to 44 and 50 movie frames, respectively. For the Ka-supplemented state data sets, micrographs were acquired using SerialEM with a total dose of 50 e⁻ Å⁻², corresponding to 42 and 50 movie frames, respectively.

2.7. Data processing

For the data sets of MexBYB in the absence of kanamycin (apo state), the image stacks were aligned and binned by patch motion correction. The contrast transfer function (CTF) was estimated using `patch CTF estimate' in CryoSPARC (version 4.7.0; Punjani et al., 2017). A procedure for blob picking followed by two-dimensional (2D) classification was applied to generate templates for automated template picking. Initially, 1 525 339 particles and 2 641 349 particles were extracted using `extract from micrographs' after autopicking. Four iterative rounds of 2D classification, followed by ab initio reconstruction and heterogeneous refinement, were performed to remove false picks. Three 3D masks were created based on the refined map, focusing on the three porter domains of each protomer. Heterogeneous refinement and 3D classification were performed again to remove false picks and classes with unclear features. The two data sets were mixed into one data set. 3D variability analysis resulted in three distinct classes of MexBYB images. In the 3D variability display, there are two states of the MexBYB trimer, one with asymmetry and one with a symmetry-like conformation. Non-uniform refinement resulted in 4.72 and 5.82 Å resolution cryo-EM maps for the symmetry-like MexBYB-Apo I trimer and the asymmetric MexBYB-Apo II trimer based on the gold-standard Fourier shell correlation (FSC 0.143), respectively.

In the same way as the processing of the MexBYB apo state, for the data sets of MexBYB in the presence of kanamycin (Ka-supplemented state) the image stacks were aligned and binned by patch motion correction. The contrast transfer function (CTF) was estimated using `patch CTF estimate' in CryoSPARC (version 4.7.0; Punjani et al., 2017). A procedure for blob picking followed by 2D classification was applied to generate templates for automated template picking. Initially, 3 722 536 particles and 3 166 482 particles were extracted using `extract from micrographs' after autopicking. Four iterative rounds of 2D classification, followed by ab initio reconstruction and heterogeneous refinement, were performed to remove false picks. Three 3D masks were created based on the refined map, focusing on the porter domain of each protomer. Heterogeneous refinement and 3D classification were performed again to remove false picks and classes with unclear features. The two data sets were mixed into one data set. 3D variability analysis resulted in three distinct classes of MexBYB images. In the 3D variability display, there are two states of the MexBYB trimer, one with asymmetry and one with a symmetry-like conformation. Non-uniform refinement resulted in 3.60 and 3.55 Å resolution cryo-EM maps fo the symmetry-like MexBYB Ka-supplemented I trimer and the asymmetric MexBYB Ka-supplemented II trimer, based on the gold-standard Fourier shell correlation (FSC 0.143), respectively.

2.8. Model building and refinement

Model building was based on the cryo-EM maps. An MexBYB structure predicted using AlphaFold3 (Abramson et al., 2024) was used and fitted into the corresponding density maps using ChimeraX (version 1.8; Meng et al., 2023). Structural refinements were performed using manual adjustment and rebuilding in Coot (version 0.9.8.95 EL; Emsley et al., 2010). Final modelling was performed using the phenix.real_space_refine program from Phenix (version 1.21.2-5419; Afonine et al., 2018).

3. Results

3.1. The cryo-EM structure of apo MexBYB

The obtained cryo-EM structure of MexBYB closely resembles the known structures of RND-family efflux transporters such as MexB (Nakashima et al., 2013), AcrB (Eicher et al., 2012) and AdeB (Morgan et al., 2021), forming a homotrimer oriented perpendicular to the membrane surface. Each protomer consisted of 12 transmembrane helices (TM1–TM12) and six periplasmic subdomains (PN1, PN2, PC1 and PC2 in the porter domain; DN and DC in the FL domain). 3D variability analysis performed in CryoSPARC (version 4.7.0; Punjani et al., 2017) revealed two distinct conformational states of MexBYB. One was an asymmetric structure (MexBYB-Apo I), in which the PC1–PC2 cleft of protomer A was closed while its PN1–PN2 cleft was opened, whereas both clefts remained open in protomers B and C. The other was a C3-symmetry-like structure (MexBYB-Apo II) in which all three protomers adopted a binding-like state conformation with both the PC1–PC2 and PN1–PN2 clefts open (Fig. 1).

Figure 1 Cryo-EM density maps of MexBYB-Apo. (a) Side view of MexBYB-Apo I density in the asymmetric conformation at an overall resolution of 4.72 Å. (b) Side view of MexBYB-Apo II density in the symmetry-like conformation at 5.82 Å resolution. (c) Top view of the porter domain of MexBYB-Apo density in the asymmetric conformation. (d) Top view of the porter domain of MexBYB-Apo II density in the symmetry-like conformation. The approximate boundary of the inner membrane-embedded part of MexBYB-Apo is indicated by dashed lines. The dimensions of the densities corresponding to the TM, porter and FL domains are indicated in (a) and (b). The open clefts between PC1 and PC2 are indicated by dashed lines.

Although both the asymmetric and symmetry-like conformations shared an overall similar fold in the three protomers, the density of the PC1 subdomain in protomer C was weaker than that in the other protomers, suggesting localized instability. Consequently, C3 symmetry could not be applied to the symmetry-like structure. Therefore, both MexBYB-Apo I and MexBYB-Apo II were reconstructed in C1 symmetry and were determined at 4.72 and 5.82 Å resolution, respectively (Supplementary Fig. S4).

Furthermore, to compare changes in the porter domains between the two conformational states, we extracted atomic models of the PC1, PC2, PN1 and PN2 subdomains from the published MexY structure (PDB entry 9e9f; Gregor et al., 2025). These subdomain models were independently fitted into each density map using UCSF ChimeraX (version 1.8; Meng et al., 2023) and the fitted components were then combined into an integrated model. By measuring the distances between the center of mass of each subdomain pair, PC1–PC2 and PN1–PN2, in the integrated models using PyMOL (version 3.0.0; Schrödinger), we quantitatively compared the structural differences within the porter domain (Tables 1 and 2).

Table 1
The distance (Å) between PC1–PC2 and PN1–PN2 in the extrusion state of MexY, MexBYB and their homologs

The center of mass of each subdomain and the distance between them were calculated by PyMOL (version 3.0.0; Schrödinger, USA). The amino-acid residues in the subdomain of each protein were defined as PC1 (MexY and MexBYB, residues 566–665; MexB, residues 567–668; AcrB, residues 567–668; AdeB, residues 564–660), PC2 (MexY and MexBYB, residues 666–715 and 806–853; MexB, residues 669–717 and 810–857; AcrB, residues 669–718 and 810–858; AdeB, residues 66–709 and 801–849), PN1 (MexY and MexBYB, residues 39–134; MexB, residues 41–135; AcrB, residues 39–135; AdeB, residues 39–135) and PN2 (MexY and MexBYB, residues 135–179 and 272–326; MexB, residues 136–180 and 273–327; AcrB, residues 136–180 and 273–327; AdeB, residues 136–180 and 273–327).

	Distance (Å)
Protein (PDB code)	PC1–PC2	PN1–PN2	Orientation of Lys on TM10
MexBYB-Apo I protomer A	23.2	30.0
MexBYB-Ka-supplemented I protomer A	23.4	30.1	TM11 Thr
MexY (9e9f)	22.9	30.9	TM11 Thr
MexB (3w9i)	22.3	28.5	TM11 Thr
AcrB (4dx5)	22.9	28.1	TM11 Thr
AdeB (7kgh)	23.1	29.9	TM11 Thr

Table 2
The distance (Å) between PC1–PC2 and PN1–PN2 in the binding-like state of MexBYB

The center of mass of each subdomain and the distance between them were calculated by PyMOL (version 3.0.0; Schrödinger, USA). The amino-acid residues in the subdomain of each protein were defined as PC1 (residues 566–665), PC2 (residues 666–715 and 806–853), PN1 (residues 39–134) and PN2 (residues 135–179 and 272–326).

	Distance (Å)
Protein (PDB code)	PC1–PC2	PN1–PN2	Orientation of Lys on TM10
MexBYB-Apo I protomer B	28.9	31.1
MexBYB-Apo I protomer C	31.4	33.3
MexBYB-Apo II protomer A	26.6	32.2
MexBYB-Apo II protomer B	28.7	33.6
MexBYB-Apo II protomer C	30.4	33.0
MexBYB-Ka-supplemented I protomer B	29.8	30.5	TM4 Asp-Asp
MexBYB-Ka-supplemented I protomer C	30.8	33.6	TM4 Asp-Asp
MexBYB-Ka-supplemented II protomer A	28.8	31.5	TM4 Asp-Asp
MexBYB-Ka-supplemented II protomer B	29.6	33.5	TM4 Asp-Asp
MexBYB-Ka-supplemented II protomer C	31.3	32.7	TM4 Asp-Asp

In the symmetry-like conformation, MexBYB-Apo II, the overall appearance of the trimer was close to C3 symmetry. However, analysis of the measured center-of-mass distances revealed a slight asymmetry: the PC1–PC2 distance in protomer A (26.6 Å) was reduced by 2.1 and 3.8 Å relative to those in protomer B (28.7 Å) and protomer C (30.4 Å), respectively. This change indicates that MexBYB-Apo II does not represent a strictly C3-symmetric structure but rather reflects a symmetry-like state.

To compare the MexBYB-Apo I and MexBYB-Apo II structures, the two structures were aligned by superposing PC2 of protomer A in MexBYB-Apo I onto that of protomer A in MexBYB-Apo II, and likewise for protomers B and C. This alignment revealed a clear conformational shift in PC1 when the maps were aligned based on PC2 of protomer A (Supplementary Fig. S5). To enable a quantitative assessment of this conformational change, we compared the PC1–PC2 distances for each protomer. In MexBYB-Apo I, the PC1–PC2 distances varied among the protomers (23.2, 28.9 and 31.4 Å in protomers A, B and C, respectively), indicating distinct open and closed states. In contrast, the PC1–PC2 distances in MexBYB-Apo II fell within a narrower range (26.6, 28.7 and 30.4 Å in protomers A, B and C), demonstrating a reduced protomer-to-protomer change compared with the asymmetric conformation. Together, these distance distributions support the assignment of MexBYB-Apo I as an asymmetric conformation containing distinct open and closed protomers, whereas MexBYB-Apo II represents a more symmetry-like state with reduced conformational heterogeneity.

For the PN1–PN2 distances, the degree of change was smaller both in the asymmetric MexBYB-Apo I (30.0, 31.1 and 33.3 Å in protomers A, B and C) and MexBYB-Apo II (32.2, 33.6 and 33.0 Å in protomers A, B and C). In the asymmetric MexBYB-Apo I, the PC1–PC2 clefts differ markedly, whereas the PN1–PN2 distances display only small differences between protomers. In the symmetry-like MexBYB-Apo II, these distances were nearly identical across all protomers. Thus, unlike PC1–PC2, the PN1–PN2 distances did not exhibit pronounced differences between protomers.

3.2. The cryo-EM structure of MexBYB-Ka-supplemented and comparison with MexBYB-Apo

To elucidate the aminoglycoside-recognition mechanism of MexY, MexBYB reconstituted in lipid nanodiscs was incubated with 10 mM kanamycin for 30 min before grid preparation. 3D variability analysis in CryoSPARC (version 4.7.0; Punjani et al., 2017) revealed two distinct conformational states of MexBYB-Ka-supplemented, as observed for the MexBYB-Apo structures: an asymmetric structure (MexBYB-Ka-supplemented I), characterized by a closed PC1–PC2 and an open PN1–PN2 cleft in protomer A, with both clefts open in protomers B and C, and a C3 symmetry-like structure (MexBYB-Ka-supplemented II), in which all protomers were in a binding-like state with both clefts open (Fig. 2). Also, localized instability was observed in the PC1 subdomain of protomer C. Consequently, both MexBYB-Ka-supplemented I and MexBYB-Ka-supplemented II were reconstructed with C1 symmetry and determined at resolutions of 3.60 and 3.55 Å, respectively (Supplementary Fig. S6). These resolutions represent improvements of 1.12 and 2.27 Å compared with their corresponding apo structures. However, no density corresponding to kanamycin was detected in the maps.

Figure 2 Cryo-EM density maps of MexBYB-Ka-supplemented. (a) Side view of MexBYB-Ka-supplemented I density in the asymmetric conformation at an overall resolution of 3.60 Å. (b) Side view of MexBYB-Ka-supplemented II density in the symmetry-like conformation at 3.55 Å resolution. (c) Top view of the porter domain of MexBYB-Ka-supplemented density in the asymmetric conformation. (d) Top view of the porter domain of MexBYB-Ka-supplemented II density in the symmetry-like conformation. The approximate boundary of the inner membrane-embedded part of the MexBYB-Ka-supplemented is indicated by dashed lines. The dimensions of the densities corresponding to the TM, porter and FL domains are indicated in (a) and (b). The open clefts between PC1 and PC2 are indicated by dashed lines.

In the asymmetric conformation of the kanamycin-supplemented sample (MexBYB-Ka-supplemented I), the PC1–PC2 distances differed substantially among the protomers (23.4, 29.8 and 30.8 Å in protomers A, B and C, respectively), indicating clearly distinct open and closed states, and the PN1–PN2 distances also showed low stepwise change (30.1, 30.5 and 33.6 Å in protomers A, B and C), similar to MexBYB-Apo I.

In contrast, in the symmetry-like conformation of the kanamycin-supplemented sample (MexBYB-Ka-supplemented II), the PC1–PC2 distances (28.8, 29.6 and 31.3 Å in protomers A, B and C, respectively) and PN1–PN2 distances (31.5, 33.5 and 32.7 Å in protomers A, B and C, respectively) all exhibit open clefts.

Comparison of the measured PC1–PC2 and PN1–PN2 distances (Tables 1 and 2) indicated that the asymmetric conformations of MexBYB exhibited no major structural changes upon addition of kanamycin. All clefts were markedly open, resembling the large open clefts observed in MexY, across the porter domains of all protomers in MexBYB-Ka-supplemented II and MexBYB-Apo II. Based on these observations, we focused on the higher resolution MexBYB-Ka-supplemented structures to delineate their detailed structural features.

3.3. The distances between PC1–PC2 and PN1–PN2 in RNDs

In previous studies, four conformational states (access, binding, extrusion and resting) were defined based on the combination of open or closed clefts between PC1–PC2 and PN1–PN2. To more clearly distinguish these states, we measured the center-of-mass distances of the PC1–PC2 and PN1–PN2 clefts in the homologous RND transporters MexB (PDB entry 3w9i; Nakashima et al., 2013), AcrB (PDB entry 4dx5; Eicher et al., 2012), AdeB (PDB entry 7kgh; Morgan et al., 2021) and BpeF (PDB entry 7wlv; Kato et al., 2023) (Tables 1, 3, 4 and 5).

Table 3
The distance (Å) between PC1–PC2 and PN1–PN2 in the binding state of MexY and its homologs

The center of mass of each subdomain and the distance between them were calculated by PyMOL (version 3.0.0; Schrödinger, USA). The amino-acid residues in the subdomain of each protein were defined as PC1 (MexY, residues 566–665; MexB, residues 567–668; AcrB, residues 567–668; AdeB, residues 564–660; BpeF, residues 575–676), PC2 (MexY, residues 666–715 and 806–853; MexB, residues 669–717 and 810–857; AcrB, residues 669–718 and 810–858; AdeB, residues 66–709 and 801–849; BpeF, residues 677–725 and 817–866), PN1 (MexY, residues 39–134; MexB, residues 41–135; AcrB, residues 39–135; AdeB, residues 39–135; BpeF, residues 41–137) and PN2 (MexY, residues 135–179 and 272–326; MexB, residues 136–180 and 273–327; AcrB, residues 136–180 and 273–327; AdeB, residues 136–180 and 273–327; BpeF, residues 138–182 and 276–330).

	Distance (Å)
Protein (PDB code)	PC1–PC2	PN1–PN2	Orientation of Lys on TM10
MexY (9e9f)	28.8	30.7	TM4 Asp-Asp
MexB (3w9i)	26.5	26.5	TM4 Asp-Asp
AcrB (4dx5)	28.7	27.9	TM4 Asp-Asp
AdeB (7kgh)	26.9	28.8	TM4 Asp-Asp
BpeF (7wlv)	27.0	26.8	TM4 Asp-Asp

Table 4
The distance (Å) between PC1–PC2 and PN1–PN2 in the access state of MexY homologs

The center of mass of each subdomain and the distance between them were calculated by PyMOL (version 3.0.0; Schrödinger, USA). The amino-acid residues in the subdomain of each protein were defined as PC1 (MexB, residues 567–668; AcrB, residues 567–668), PC2 (MexB, residues 669–717 and 810–857; AcrB, residues 669–718 and 810–858), PN1 (MexB, residues 41–135; AcrB, residues 39–135) and PN2 (MexB, residues 136–180 and 273–327; AcrB, residues 136–180 and 273–327).

	Distance (Å)
Protein (PDB code)	PC1–PC2	PN1–PN2	Orientation of Lys on TM10
MexB (3w9i)	24.2	25.2	TM4 Asp-Asp
AcrB (4dx5)	27.1	26.1	TM4 Asp-Asp

Table 5
The distance (Å) between PC1–PC2 and PN1–PN2 in the resting state of MexY and its homolog AdeB

The center of mass of each subdomain and the distance between them were calculated by PyMOL (version 3.0.0; Schrödinger, USA). The amino-acid residues in the subdomain of each protein were defined as PC1 (MexY, residues 566–665; AdeB, residues 564–660), PC2 (MexY, residues 666–715 and 806–853; AdeB, residues 66–709 and 801–849), PN1 (MexY, residues 39–134; AdeB, residues 39–135) and PN2 (MexY, residues 135–179 and 272–326; AdeB, residues 136–180 and 273–327).

	Distance (Å)
Protein (PDB code)	PC1–PC2	PN1–PN2	Orientation of Lys on TM10
MexY (9e9f)	22.5	26.7	TM4 Asp-Asp
AdeB (7kgh)	23.1	26.2	TM11 Thr

In MexB, the open-cleft distances were 24.2 Å (access PC1–PC2), 26.5 Å (binding PC1–PC2), 26.5 Å (binding PN1–PN2) and 28.5 Å (extrusion PN1–PN2). In AcrB, the corresponding open distances were 27.1 Å (access PC1–PC2), 28.7 Å (binding PC1–PC2), 27.9 Å (binding PN1–PN2) and 28.1 Å (extrusion PN1–PN2). In AdeB, the open distances were 26.9 Å (binding PC1–PC2), 28.8 Å (binding PN1–PN2) and 29.9 Å (extrusion PN1–PN2). In BpeF, the distances were 27.0 Å (binding PC1–PC2) and 26.8 Å (binding PN1–PN2).

For closed clefts, the PC1–PC2 and PN1–PN2 distances in MexB were 25.2 Å (access PN1–PN2) and 22.3 Å (extrusion PC1–PC2). In AcrB, the closed-state distances were 26.1 Å (access PN1–PN2) and 22.9 Å (extrusion PC1–PC2). In AdeB, the closed distances were 23.1 Å (extrusion PC1–PC2), 23.1 Å (resting PC1–PC2) and 26.2 Å (resting PN1–PN2).

From these results, we define the PC1–PC2 cleft as open when the center-of-mass distance is 24.2–28.8 Å and closed when it is 22.3–23.1 Å. Similarly, in PN1–PN2 the cleft is open at 26.5–30.9 Å and closed at 25.2–26.2 Å in these RND transporters.

We also measured the PC1–PC2 and PN1–PN2 distances for MexY (PDB entry 9e9f; Gregor et al., 2025; Tables 1, 3 and 5). In the resting state, the PC1–PC2 and PN1–PN2 distances were 22.5 and 26.7 Å, respectively. In the binding state both clefts were widely open, with distances of 28.8 Å in PC1–PC2 and 30.7 Å in PN1–PN2. In contrast, the extrusion state showed a closed PC1–PC2 cleft at 22.9 Å and an open PN1–PN2 cleft at 30.9 Å. Notably, MexY displayed the largest open-cleft sizes among the RND transporters measured, particularly in its binding and extrusion states. These measurements confirm our initial observation that MexY adopts unusually large open clefts in these conformations compared with previously reported RND transporters.

3.4. Comparison of MexY with the MexBYB-Ka-supplemented I and MexBYB-Ka-supplemented II structures

Compared with MexY and MexB, the FL domain in MexBYB is replaced with that from MexB, showing a clear positional shift in this domain. To visualize these differences, we created an equilateral triangle in each FL domain of MexB (EMDB entry EMD-10371; Glavier et al., 2020), MexY (EMD-47796; Gregor et al., 2025), MexBYB-Ka-supplemented I and MexBYB-Ka-supplemented II with dotted lines in red (Fig. 3).

Figure 3 Comparison of the FL domain between MexB, MexY, MexBYB-Ka-supplemented I and MexBYB-Ka-supplemented II. The side view and the top view of (a) MexB (EMDB entry EMD-10371), (b) MexY (EMDB entry EMD-47796), (c) MexBYB-Ka-supplemented I and (d) MexBYB-Ka-supplemented II are shown. An equilateral triangle, with the side length defined by connected reference points on protomers A and C (corresponding to protomers in the extrusion and binding states in MexY and to protomers in the close and tight states in MexB), was generated in red dotted lines in each FL domain.

In previously reported MexB structures, when two vertices of an equilateral triangle were aligned to the reference points of the binding-state and extrusion-state protomers, the third vertex fell very close to the reference point of the access-state protomer. This indicates that the position of the FL domains in the inflexible MexB is close to an ideal triangular configuration.

In contrast, when the same alignment was applied to MexY, the third vertex deviated substantially from the reference point of the resting-state protomer. This suggests that the FL domains of MexY remain in a relatively flexible configuration.

In MexBYB-Ka-supplemented I and MexBYB-Ka-supplemented II, the positions of the FL-domain reference points resemble those of MexB, forming a triangle-like arrangement. However, the distances between the reference points are overall larger than those observed in MexB. Moreover, when MexY was aligned with each MexBYB trimer map (Supplementary Fig. S7), the inward bending of the FL domain seen in MexY disappeared in MexBYB.

Taken together, these observations suggest that replacing the flexible FL domains of MexY with the inflexible FL domains of MexB causes the FL-domain orientations in the chimeric MexBYB to become more MexB-like, whereas the FL domains become more widely separated.

Next, we compared PC1–PC2 and PN1–PN2 of MexBYB-Ka-supplemented with those of MexY by overlapping the volume maps (Figs. 4 and 5). In MexBYB-Ka-supplemented I, protomer A exhibited a closed PC1–PC2 and an open PN1–PN2, which closely resembled the MexY extrusion state. In protomer B and protomer C of MexBYB-Ka-supplemented I, as well as all three protomers of MexBYB-Ka-supplemented II, PC1–PC2 and PN1–PN2 were open, but the sizes of each cleft are larger than the binding-state protomer of MexY. Notably, none of the protomers in any MexBYB-Ka-supplemented structure adopted a conformation corresponding to the resting state defined for MexY.

Figure 4 Overlay diagram of MexBYB-Ka-supplemented I and MexY protomers. (a, b, c) Protomers A, B and C of MexBYB-Ka-supplemented I (pink) fitted on protomers A, B and C of MexY (EMDB entry EMD-10371; gray) by aligning PC2. (d, e, f) Protomers A, B and C of MexBYB-Ka-supplemented I (pink) fitted on protomers A, B and C of MexY (EMDB entry EMD-10371; gray) by aligning PN1. The side view of the protomer and the top view of the porter domain are shown and significant changes are indicated by a dashed circle in red.

Figure 5 Overlay diagram of MexBYB-Ka-supplemented II and MexY protomers. (a, b, c) Protomers A, B and C of MexBYB-Ka-supplemented II (green) fitted on protomers A, B and C of MexY (EMDB entry EMD-10371; gray) by aligning PC2. (d, e, f) Protomers A, B and C of MexBYB-Ka-supplemented II (green) fitted on protomers A, B and C of MexY (EMDB entry EMD-10371; gray) by aligning PN1. The side view of the protomer and the top view of the porter domain are shown and significant changes are indicated by a dashed circle in red.

To evaluate the conformational states of each protomer in MexBYB, we compared the center-of-mass distances between PC1–PC2 and PN1–PN2 in MexY and MexBYB (Tables 1, 2, 3 and 5). As a result, protomer A of MexBYB-Ka-supplemented I (PC1–PC2, 23.4 Å; PN1–PN2, 30.1 Å) aligned well with the MexY extrusion state (PC1–PC2, 22.9 Å; PN1–PN2, 30.9 Å). Protomer B (PC1–PC2, 29.8 Å; PN1–PN2, 30.5 Å) closely aligned with the MexY binding state (PC1–PC2, 28.8 Å; PN1–PN2, 30.7 Å). Protomer C (PC1–PC2, 30.8 Å; PN1–PN2, 33.6 Å), however, exhibited a more open conformation that was not observed among the previously reported MexY states. In the symmetry-like MexBYB-Ka-supplemented II structure, the protomers showed PC1–PC2 distances of 28.8, 29.6 and 31.3 Å and PN1–PN2 distances of 31.5, 33.5 and 32.7 Å. These distances were consistently 1–2 Å larger than those of the representative MexY conformations.

Subsequently, we checked the location of specific residues involved in the proton-relay network in the TM domain by building the corresponding atomic models into the density maps (Fig. 6). Previous studies have shown that in the binding and resting states of MexY (PDB entry 9e9f; Gregor et al., 2025), Asp406 and Asp407 on TM4 form a salt bridge with Lys934 on TM10, with Lys934 positioned between Asp406 and Asp407. In contrast, in the extrusion state, Lys934 disengages from this salt bridge and instead shifts towards Thr971 on TM11 and Asn935 on TM10, interacting with these residues.

Figure 6 Locations of specific residues involved in the proton-relay network in the MexBYB protomers. The locations of specific residues involved in the proton-relay network of the three MexBYB protomers are shown in (a)–(d) MexBYB-Ka-supplemented I (PDB entry 22xk) and (e)–(h) MexBYB-Ka-supplemented II (PDB entry 22xm). Ribbon diagram of MexBYB-Ka-supplemented I and MexBYB-Ka-supplemented II showing the locations of the proton-relay network (black box). The residues are shown in cyan and the density map of each conformation is shown as a black mesh. The density maps of MexBYB-Ka-supplemented I and MexBYB-Ka-supplemented II are contoured at 5.0σ (sigma units) in PyMOL, corresponding to levels of 0.0878 and 0.0891 in ChimeraX, respectively.

In the MexBYB models, no density was observed between Asp406 and Asp407 in protomer A of MexBYB-Ka-supplemented I, and the side chain of Lys936 (corresponding to Lys934 in MexY) was oriented toward Thr973 and Asn937 (corresponding to Thr971 and Asn935 in MexY) (Fig. 6). This orientation closely resembles that of the MexY extrusion state, supporting the conclusion that the protomer A of MexBYB-Ka-supplemented I adopts an extrusion state.

By contrast, in protomer B and protomer C of MexBYB-Ka-supplemented I, as well as in all three protomers of MexBYB-Ka-supplemented II, Lys936 was positioned between Asp406 and Asp407 (Fig. 6), consistent with the salt-bridge geometry observed in the MexY binding and resting states. Together with the conformational features of the porter domain, these observations suggest that these protomers adopt a configuration similar to the binding state of MexY.

4. Conclusion and discussion

In this study, we determined the cryo-EM structures of the chimeric protein MexBYB, in which the trimer of P. aeruginosa MexB was combined with the porter domain of MexY. Under both apo and kanamycin-supplemented conditions, we captured two major conformations: an asymmetric conformation and a symmetry-like conformation. The kanamycin-supplemented structures consistently exhibited higher resolutions than the apo structures. Both the apo and kanamycin-supplemented samples were prepared under identical buffer conditions, with the only difference being the addition of kanamycin to the kanamycin-supplemented sample. Because kanamycin was supplemented in the form of kanamycin sulfate, sulfate ions were also introduced into the sample solution. Accordingly, the observed improvement in resolution suggests that the trimer could be stabilized by the presence of kanamycin and/or sulfate ions.

Although the structures of MexBYB resemble those of previously reported RND transporters such as MexY and MexB, the conformational changes of the porter domain differed substantially. In MexBYB-Ka-supplemented I, protomer A adopted an extrusion state similar to that of MexY. Meanwhile, protomers B and C of MexBYB-Ka-supplemented I, as well as all three protomers of MexBYB-Ka-supplemented II, resembled the binding state of MexY, with markedly large open clefts between PC1–PC2 and PN1–PN2.

Consistent with previous studies on MexY (PDB entry 9e9f; Gregor et al., 2025), MexB (PDB entry 3w9i; Nakashima et al., 2013), AcrB (PDB entry 4dx5; Eicher et al., 2012), AdeB (PDB entry 7kgh; Morgan et al., 2021) and BpeF (PDB entry 7wlv; Kato et al., 2023), we also observed specific residues involved in the proton-relay network in the MexBYB-Ka-supplemented structures. In protomer A of MexBYB-Ka-supplemented I, which resembles the extrusion state, the lysine on TM10 is oriented towards the threonine on TM11 (TM11 Thr). In contrast, in the other protomers that adopt a binding-like state, the TM10 lysine is positioned between the two asparagines on TM4 (TM4 Asp-Asp). These results indicate that even after replacing the FL and TM domains, the locations of specific residues involved in the proton-relay network are preserved. In addition, the characteristic features of the porter domain are also retained.

Overall, our studies show that the unusually large open clefts observed in both MexBYB and MexY represent a unique conformational feature of the MexY porter domain. This distinct conformation may be beneficial to the recognition, binding and extrusion of aminoglycoside antibiotics. Furthermore, understanding this conformation may accelerate structural and inhibitor-development efforts targeting clinically important multidrug-efflux pumps.

5. Related literature

The following references are cited in the supporting information for this article: Bharatham et al. (2021), Lawrence et al. (2025), Lyu et al. (2020), Ouyang et al. (2025), Su et al. (2017) and Zhang et al. (2021, 2023).