The MacA–MacB–TolC tripartite complex is a transmembrane machine that spans both plasma membrane and outer membrane and actively extrudes substrates, including macrolide antibiotics, virulence factors, peptides and cell envelope precursors. These transport activities are driven by the ATPase MacB, a member of the ATP-binding cassette (ABC) superfamily. Here, we present the crystal structure of MacB at 3.4-Å resolution. MacB forms a dimer in which each protomer contains a nucleotide-binding domain and four transmembrane helices that protrude in the periplasm into a binding domain for interaction with the membrane fusion protein MacA. MacB represents an ABC transporter in pathogenic microorganisms with unique structural features.
Multidrug resistance caused by export proteins is a serious problem in antibiotic treatment of numerous bacterial infections1. The envelope of Gram-negative pathogens, such as Acinetobacter baumannii and Pseudomonas aeruginosa, contains unique tripartite machineries that export noxious compounds from the cell. These machineries are composed of a plasma membrane transporter and outer membrane porin (OMP) that are connected by a periplasmic adaptor protein2. Proton motive force-dependent tripartite multidrug efflux transporters belonging to the resistance-nodulation-cell division (RND) transporter family have been well characterized3, 4. ABC family and major facilitator superfamily transporters can also be part of tripartite complexes, and share similar or identical components with RND transporters, but their structures are still unknown5, 6. Crystal structures of ABC transporters have been reported for various organisms for both substrate exporters and importers7, 8. In type I bacterial ABC exporters for antibiotics and cytotoxic agents, the minimum functional unit has two transmembrane domains (TMDs) and two cytosolic nucleotide-binding domains (NBDs). One TMD is often fused to one NBD on a single polypeptide that assembles into a homodimer or heterodimer to form a functional unit. The TMDs can contain six transmembrane helices (TMs) per monomer, but two helices out of six are combined with four helices of the second half-transporter, thus enlarging the dimer interface and providing a tight interaction between the monomers during transport. The TMs are extended at the cytosolic side, forming intracellular domains that interact with NBDs via coupling helices (CHs)9,10,11,12.
The tripartite MacA-MacB-TolC transporter in Gram-negative bacterial pathogens including Escherichia coli, Neisseria gonorrhoeae, P. aeruginosa, Vibrio cholerae, Klebsiella pneumoniae, Yiersinia pestis and A. baumannii 5, 13 is an important efflux pump that mediates the extrusion of macrolides14, peptide toxins15, virulence factors16, siderophores17, lipopolysaccharides18 and protoporphyrins19. MacA-MacB-TolC contains similar components as the relatively well-characterized tripartite RND transporters AcrA-AcrB-TolC in E. coli and MexA-MexB-OprM in P. aeruginosa. The membrane fusion protein (MFP) MacA is homologous to AcrA and MexA, whereas the OMP TolC is shared by both RND and ABC tripartite systems in E. coli, and is homologous to OprM20. However, the ABC protein MacB is unrelated to AcrB and MexB, which form large trimeric complexes in the plasma membrane, raising questions about the structure and domain organization of MacB in the tripartite efflux pump. Here, we present the crystal structure of MacB, and show that, by analogy to RND transporters, the MacB homodimer contains two large periplasmic domains (PLDs) that enable the interactions with the other components in this tripartite efflux pump.
Overall structure of MacB
The ABC exporter MacB from A. baumannii was functionally expressed in E. coli (Fig. 1, Tables 1, 2, Supplementary Figs. 1 and 14). MacB was purified and its crystal structure was solved at 3.4-Å resolution (Fig. 2a–e, Table 3, Supplementary Fig. 2). A non-critical polypeptide segment between position 248 and 259 at the exterior of the NBD was found to be disordered. MacB shares 83% amino-acid sequence similarity (53% identity) with well-studied MacB from E.coli (Fig. 3a) and has a domain arrangement similar to yeast PDR521 and mammalian ABCG transporters22 with an N-terminal NBD followed by a C-terminal TMD (Fig. 3b). Two MacB monomers have a biological twofold rotation axis at the centre of the dimer. Each monomer in the dimer has essentially an identical structure (~1.1-Å RMSD in Cα position), and has elongated TMs (TM1 and TM2) protruding about 25 Å at the periplasmic side of the phospholipid bilayer (Fig. 2, Supplementary Fig. 3). The distal end of these TMs forms the PLD (Fig. 3c, Supplementary Fig. 3) that is important for the interaction with the MFP MacA23. The interface between the monomers in the MacB dimer is extensive (2188.9 Å2) with major contributions from the TMDs (1053.0 Å2, 48%), PLDs (676.1 Å2, 31%) and NBDs (417.0 Å2, 19%). Although MacB shares structural motifs with other members of the ABC superfamily, its architecture is unique among transporters of known structures (Fig. 2f).
MacB was crystallized in the presence of adenosine-5′-(β-thio)-diphosphate (ADPβS) (Fig. 4a, b, Supplementary Fig. 4). The two NBDs are dimerized in a head-to-tail arrangement, and both subunits contribute conserved Walker A, Walker B and ABC signature motifs and residues to form two composite nucleotide-binding sites at the dimer interface. Electron density was observed in the difference Fourier map (Fig. 4a, b) at the expected position for the nucleotide-binding site. Although the shape of the difference Fourier map is not optimally fitted by the ADPβS model, the phosphor and sulphur atoms of the bound ADPβS are clearly identified in the anomalous difference Fourier map (7σ, 0.013 electron Å−3) (Fig. 4a, b, Supplementary Fig. 5). The binding of the adenine moiety of the nucleotide involves a π-stacking interaction with a conserved aromatic residue (Phe16) in the A-loop of the NBD (Fig. 4b, Supplementary Fig. 4)24. When compared with crystallized nucleotide-bound and apo states of various ABC exporters (Fig. 4c, Supplementary Table 1), the arrangement of the two NBDs in the MacB dimer is most similar to the arrangement of the NBDs in the ADP-bound outward-occluded state of the lipid-linked oligosaccharide flippase PglK25.
Transmembrane domain structure
The TMD of MacB encompasses four TMs per monomer, which is the smallest number among currently characterized ABC exporters7. The TMs are not swapped between the half-transporters (Fig. 2f) and lack the helical extensions into intracellular domains as found in most type-I ABC exporters. MacB shares this property with ABC importers and the type-II exporter ABCG5/G826 and ABCG227. Even though the MacB monomer has only one cytoplasmic loop, MacB preserves two CHs as observed in other ABC exporters. CH1 is present in the intracellular loop between TM2 and TM3, whereas CH2 is located in the carboxy terminus of MacB (Fig. 3b). Unlike Sav1866 and MsbA, in which CH1 contacts the NBDs of respective monomers and CH2 interacts with the NBD of the opposite monomer exclusively9, 10, the two CHs in MacB only contact one NBD within the same monomer. However, even though the CHs in MacB are topologically different and the arrangement of connected TMs shares no similarity with any other bacterial exporter, the placement of the two CHs on the NBD surface is quite similar to the arrangement found in crystallized ABC exporters (Fig. 5, Supplementary Fig. 6, Supplementary Movie 1).The positions of the CHs and their interactions with the NBDs are similar, but the connecting TMs are completely different. Furthermore, the order of the CHs is opposite due to the topological differences (Fig. 5b–d, Supplementary Fig. 6). Removal of CH2 by truncation of the carboxy terminus of MacB leads to inactivation of drug export (Fig. 1a, Table 1, Supplementary Figs. 7 and 14), demonstrating the importance of the CHs for functionality. Furthermore, a loop connects the TMD at its N-terminus to the NBD. This loop contains an amphiphilic α-helix running parallel to the cytosolic surface of the plasma membrane in which it is partially embedded (Fig. 3b, Supplementary Fig. 3). This helix is equivalent to the ‘elbow helix’ in many other ABC exporters, and is also observed in a similar location as the ‘connecting helix’ in the heterodimeric ABCG-type exporter ABCG5/ABCG8 (Fig. 2f)26. Finally, the PLD of MacB has a similar overall structure as the crystallized isolated PLD derived from Actinobacillus MacB (r.m.s. deviation of Cα-positions of 2.2 Å over 214 residues)28 and YknZ, a homologue from Bacillus amyloliquefaciens (r.m.s. deviation of Cα-positions of 2.3 Å over 187 residues)29, except for one loop region containing α-helix, Pα1 (Supplementary Fig. 8). The variable structure of the loop containing Pα1 might point to molecular movements during catalytic activity at the NBDs, and propagation of these movements to MacA and TolC in the transport process. Given the differences in the number of TMs and their interactions in MacB compared to other ABC transporters, and the presence and structure of periplasmic extensions of TM1 and TM2 forming the PLDs, we conclude that the fold of MacB is unique among known structures of ABC exporters (Fig. 2f).
When MacB is compared with outward-facing structures of MsbA and Sav1866, which participate in alternating access mechanisms for these transporters, the overall structure provides a similar V-shape opening with access to the periplasm(Fig. 6, Supplementary Movie 2). However, the locations of these openings are different. The cavity in Sav1866 is present in the outer leaflet of the phospholipid bilayer, whereas the cavity in MacB is present in the periplasmic space. The TMD of MacB does not exhibit a distinct cavity in the membrane-embedded sections of the TMs (Supplementary Fig. 9), suggesting that a substrate-binding pocket might not exist in this part of MacB. This notion is further corroborated in a search of structural similarities between the TMD of MacB and other membrane proteins using the fold match program DALI30, 31. Surprisingly, the first 30 significant hits all correspond to ABC transporters. In particular, the arrangement of TM1, 2 and 3 (H282-L321, N521-567T and A578-F616) at the dimer interface of the MacB dimer shows a similarity with the organization of TM1, 2 and 3 (V33-T211) in the mouse multidrug resistance P-glycoprotein ABCB1a (PDB accession code: 4M1M)32 (Fig. 7, Supplementary Table 2). In the DALI pairwise comparison, the resulting Z-score of 8.0 (RMSD = 4.1) suggests that these regions share a similar protein structure motif (Fig. 7)31. The dimer interface in MacB is not matched with the dimer interface of ABCB1a. Instead, TM1, 2 and 3 in ABCB1a are located at the side of the transporter in a region that faces the phospholipid bilayer, and that is important for conformational changes during transport (Fig. 7c). Similarly, MacB’s TM1, 2 and 3 are not part of a substrate-binding cavity in the membrane (Supplementary Fig. 9), and might therefore play an analogous role in energy transduction rather than in substrate interactions. In agreement with this view one of MacB’s endogenous substrates, the heat stable enterotoxin II, is matured from precursors in the periplasm and therefore most likely extruded by MacA-MacB-TolC from the periplasm rather than membrane or cytosol15. It is also noteworthy that the MacB PLD shares structural similarity with the PLD LolE in the lipoprotein transport system LolCDE (Z = 15.7, RMSD = 3.8) (Supplementary Figs. 10–12)33, which transfers lipoproteins from the outer leaflet of the inner membrane to the periplasmic carrier protein LolA33. Furthermore, MacB PLD is similar to the PLD of AcrB (Z = 7.8, RMSD = 9.0)28 for which substrate transport from the periplasm has been demonstrated2 (Supplementary Fig. 13). Thus, MacB might capture substrates from the periplasm. Further structural information about inward-facing states might provide clues regarding the possibility that MacB can also transport substrates from the cytoplasm.
In the AcrAB-TolC and MexAB-OprM assemblies, the proton motive force-dependent functional rotation of the trimetric AcrB and MexB components at the inner membrane is tightly coupled to conformational changes in the PLDs, which in turn drive periplasmic transport of substrate via AcrA/MexA and TolC/OprM34. In MacB, the tight coupling between NBD and TMD via CH1 and CH2 will not only be used for movement of TMs but will also be critical for long-range communication with the PLDs that are required for transport of substrate via MacA and TolC. The significant differences in size of the trimeric AcrB and MexB vs.dimeric MacB in otherwise homologous tripartite assemblies raise interesting questions about the stoichiometry of the components in the MacA-MacB-TolC complex35. Given the diversity in size and chemical properties of MacB substrates, it will also be interesting to study the mechanisms of substrate binding and transport by the transporter complex. Our crystal structure provides a framework for these and other studies on MacB and related ABC exporters in Gram-negative pathogenic bacteria.
The macB gene of A. baumannii was amplified from genomic DNA (ATCC no. BAA-1710D-5) by the polymerase chain reaction using KOD-plus-Neo DNA polymerase (TOYOBO, Japan) with the forward (5′-ggaattccatatgacaaaacaagctttgcttgaagtc-3′) and the reverse (5′-cgggatccttattctcgtgatagtgctgcaacag-3′) primers (Supplementary Table 3), and inserted via NdeI and BamHI restriction sites into a modified pET-22 (Novagen) expression vector. Using this method, an N-terminal hexa-histidine tag was added for purification by immobilized metal affinity chromatography. The DNA was sequenced to ensure that only intended changes were introduced. The resulting plasmid was transformed into E. coli BL21(DE3) (Novagen) for protein expression. The transformants were grown in ten 5-L flasks at 37 °C in the Davis minimal medium36 supplemented with 0.2% glucose and 0.1% casamino acid. Expression was induced for 3 h by the addition of 0.1 mM isopropyl-β-d-thiogalactopyranoside at an OD610 of 0.6. All subsequent procedures were performed at 4 °C unless indicated otherwise. Cells were collected by centrifugation, resuspended in 50 mM Tris (pH 7.0), 0.5 mM Na-EDTA, 1 mM MgCl2, and disrupted three times using a Microfluidizer M-110EH (Microfluidics Corp., NM, USA) at 15,000 psi. Cell debris was removed by low-speed centrifugation at 27,000×g for 10 min. To collect membrane fractions, the supernatant was subjected to ultracentrifugation at 145,000×g for 1 h, and washed with 5 mM Tris (pH 7.0), 0.5 mM EDTA. The plasma membrane was solubilized in 50 mM Tris (pH 7.0) buffer, 10% (v/v) glycerol containing protease inhibitors (Roche) and 2% (w/v) n-undecyl-β-d-maltoside (UDM, GlyconBiochemicals GmbH, Germany) on ice for 1 h. After a further step of ultracentrifugation at 145,000×g for 1 h, the detergent-solubilized fraction was collected and incubated with chelating sepharose resin immobilized with Ni2+ ion at 4 °C for 1 h. The resin was washed with the buffer containing 20 mM Tris (pH 7.5), 100 mM NaCl, 25 mM imidazole, 10% (v/v) glycerol and 0.05% (w/v) UDM. The protein was eluted from the affinity resin with the wash buffer plus 300 mM imidazole. The fractions containing MacB were collected, concentrated in Amicon Stirred Cell (Merck Millipore) with 100 kDa molecular weight cutoff Omega Ultrafiltration Membrane Disc Filter (Pall Corporation, USA), and filtered with Ultrafree-MC GV Centrifugal Filter (Merck Millipore). Further purification was performed by size-exclusion chromatography (Superdex
-200 Increase 10/300 GL; GE Healthcare) in the buffer containing 20 mM Tris (pH 7.5), 100 mM NaCl, 10% (v/v) glycerol and 0.05% (w/v) UDM at the flow rate of 0.3 ml min−1 using AKTA explorer 10 S (GE Healthcare). The peak fractions were collected and concentrated in the same way as described above to about 24 mg of MacB per ml for crystallization. Concentration of purified protein was determined by bicinchoninic acid (BCA) protein assay (Thermofisher) with bovine serum albumin as a standard. The purity of MacB was analysed by SDS-PAGE followed by Coomassie Brilliant Blue staining.
Adenosine-5′-(β-thio)-diphosphate of 10 mM (ADPβS, Jena bioscience GmbH, Germany) and 2 mM MgCl2 were added to the purified protein and incubate for overnight on ice before crystallization. MacB crystals were grown by the sitting drop vapour diffusion technique at 25 °C. The protein solution was mixed (1:1) with reservoir solution containing 1.2–1.3 M sodium citrate, 100 mM Na-HEPES (pH 7.2). Crystals were grown within 1 week to optimal size (0.3 × 0.3 × 0.2 mm3). The concentration of glycerol was gradually increased to 30% (v/v) by soaking in several steps for optimal cryo-protection. Crystals were picked up using nylon loops (Hampton Research, CA, USA) for flash-cooling in cold nitrogen gas from a cryostat (Rigaku, Japan). For Se-methionine labelling, E. coli cells were grown in M9 minimal medium (l-methionine was replaced with seleno-l-methionine), and the protein was purified and crystallized as described for native protein with the exception of the addition of 2 mM adenosine-5′-triphosphate (ATP, Nacalaitesque, Japan) instead of 10 mM ADPβS before crystallization.
Data collection and structure determination
Data sets were collected at 100 K using a Rayonix MX300-HE charge-coupled device detector on the BL44XU beamline at SPring-8. Diffraction images were processed with the HKL2000 package37. Further processing was carried out with programs from the CCP4 suite38 and Phenix39, 40. Data collection and structure refinement statistics are summarized in Table 3. Native data was collected at a wavelength of 1.7500 Å. The initial phases were obtained by SAD (the single anomalous dispersion method) using SeMet derivative crystals. The heavy atom sub-structure of 19 selenium atoms was determined by SHELXD, and initial phases were obtained using SHELXE and SHARP41, 42, and were used to calculate anomalous Fourier maps using programs from the CCP4 suite38. Density modifications with solvent flattening and non-crystallographic averaging to phase improvement was performed using DM43. Model building was carried out using programs O44 and COOT45. Chain tracing was aided by the known positions of methionines from the selenomethionine data. Model refinement was conducted using Refmac546 and Rosetta-Phenix47 and Phenix39, 40. Ramachandran analysis revealed 95.6% in the favoured region and 0.0% residues in the outliers with a Molprobity48 score of 1.49. The surface area between the MacB monomers was calculated using QT-PISA (CCP4 suite)49. Figures were prepared using Molscript50 rendered with Raster3D51, Chimera52 and PyMOL.
Determination of MIC values
The expression vector pBAD24 containing macA-macB-tolC genes of A. baumannii, deletion mutants, or E172Q mutant of macB (Supplementary Table 3), were transformed into E. coli strain W3104 ΔacrABΔmacAB 53. These expression hosts were additionally transformed with the pRARE2 plasmid, which supplies transfer RNAs for rarely used codons in E. coli. The pRARE2 plasmid was isolated from Rosetta 2 (Novagen). The minimum inhibitory concentrations (MICs) of macrolide antibiotics were determined as the concentrations that prevented bacterial growth after 30 h incubation at 30 °C on YT-agar (0.8% tryptone, 0.5% yeast extract, 0.5% NaCl, 1.5% agar, 100 μg ml−1 ampicillin and 34 μg ml−1 chloramphenicol) plates with sequential dilutions, according to standard protocol recommended by the Clinical and Laboratory Standards Institute54. To induce protein expression, 2% l-arabinose was added in the agar plates. Expression of MacA, MacB or TolC in the strains used for MIC measurements was confirmed by western blotting with anti-MacA polyclonal (1:8000 dilution), anti-MacB monoclonal (1:500 dilution) or anti-TolC polyclonal (1:8000 dilution) antibodies as the primary antibody, and horseradish peroxidase (HRP)-labelled goat anti-rabbit (1:20,000 dilution, cat# 111-035-003, Jackson ImmunoResearch, USA) or anti-mouse (1:10,000 dilution, cat# 115-035-062, Jackson ImmunoResearch) IgG antibodies as the secondary antibody.
Truncation of second coupling helix
To express MacB without CH2, a stop codon was introduced by replacing the Leu655 codon in the macB gene through the use of an inverse PCR-based site-directed mutagenesis kit using KOD DNA polymerase (TOYOBO) with the two primers (5′-caaaataagaccctgttgcagcactatcacgag-3′, 5′-ctgcattcttggcaggtaagaagcc-3′) (Supplementary Table 3) and pBAD24 vector containing macA-macB-tolC genes as a template. The resulting plasmid was sequenced to confirm successful truncation, and codes for a truncated MacB protein lacking the C-terminal 10 residues that correspond to CH2. Expression of this truncated MacB was confirmed by western blotting with anti-MacB polyclonal antibody (1:1000 dilution) and HRP-labelled goat anti-rabbit IgG antibody (1:20,000 dilution, cat# 111-035-003, Jackson ImmunoResearch). The mutant was used for the MIC measurement as well as the wild-type protein.
Measurement of cellular growth rates
Overnight cultures from glycerol-stocks of drug-hypersensitive E. coli ΔacrAB ΔmacAB, harbouring pBAD plasmid encoding A. baumannii MacAB-TolC, MacA-TolC (inactive control) or none of these genes (inactive non-expressing control), were grown in Luria-Bertani broth containing 100 µg ml−1carbenicillin. Cultures were grown for about 16 h at 37 °C, and then used to inoculate fresh medium (1:1000 dilution). Once E. coli cells had reached OD600 of 0.3–0.4, protein production was induced by addition of 2% arabinose for 3 h. The cells were then diluted to OD600 of 0.06 in fresh medium containing 2% arabinose in the wells of a 96-well plate to which antibiotics were added as indicated in Fig. 1b. Growth was followed over time at OD600 at 37 °C in a Versamax microplate reader (Molecular Devices, USA).
Coordinates and structure factors have been deposited in the Protein Data Bank with accession codes: 5GKO and 5WS4. Data supporting the findings of this study are available within the article and its Supplementary Information files. Data have also been deposited in the University of Cambridge data repository (https://doi.org/10.17863/CAM.12992) or are available from the corresponding author upon reasonable request.
Publisher's note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This work was supported by a Grant-in-aid in the funding program for Next Generation World-Leading Researchers (NEXT Program) from the Japan Society for the Promotion of Science (JSPS) (to S.M.), Targeted Proteins Research Program from the Ministry of Education, Culture, Sports, Science and Technology of Japan (MEXT) (to S.M. and E.Y.), ERATO “Lipid Active Structure Project” from Japan Science and Technology Agency (JST) (to S.M.) and the Advanced Research for Medical Products Mining Programme of the National Institute of Biomedical Innovation (NIBIO) Japan (to S.M.). We are also grateful for funding by the Strategic International Cooperative Program (JST, Japan) (to S.M. and H.W.v.V.), Royal Society (UK) (to H.W.v.V.), BBSRC (UK) (to H.W.v.V.) and International Human Frontier Science Program Organization (HFSPO) for collaborative research between S.M. and H.W.v.V. A.N. is the recipient of a Herchel-Smith Scholarship. We thank M. Kato and W. Tamaya for technical assistance. The synchrotron radiation experiments were performed at the BL44XU of SPring-8 with the approval of the Japan Synchrotron Radiation Research Institute (JASRI) (proposal no. 2012A6746, 2012B6746, 2013A6855, 2013B6855, 2013B6700, 2014A6954, 2014A6700, 2015A6700 and 2016A6700). This research is partially supported by the Platform Project for Supporting Drug Discovery and Life Science Research (Platform for Drug Discovery, Informatics, and Structural Life Science) from MEXT and Japan Agency for Medical Research and Development (AMED). The pBAD24 vector was provided by the National Institute of Genetics through the National Bio-Resource Project of the MEXT, Japan. This work was done with the approval of the Joint Research Committee of Institute of Protein Research, Osaka University.