Structure of an antibacterial peptide ATP-bindingcassette transporter in a novel outward occluded stateHassanul G. Choudhurya,b,c,1, Zhen Tongd,1, Indran Mathavana,b,c, Yanyan Lie, So Iwataa,b,c,f, Séverine Zirahe,Sylvie Rebuffate,2, Hendrik W. van Veend,2, and Konstantinos Beisa,b,c,2

aDivision of Molecular Biosciences, Imperial College London, London SW7 2AZ, United Kingdom; bMembrane Protein Lab, Diamond Light Source, HarwellScience and Innovation Campus, Chilton OX11 0DE, United Kingdom; cRutherford Appleton Laboratory, Research Complex at Harwell, Didcot OX11 0FA,United Kingdom; dDepartment of Pharmacology, University of Cambridge, Cambridge CB2 1PD, United Kingdom; eLaboratory of Communication Moleculesand Adaptation of Microorganisms, Unité Mixte de Recherche 7245, Muséum National d’Histoire Naturelle, Centre National de la Recherche Scientifique,75005 Paris, France; and fDepartment of Cell Biology, Graduate School of Medicine, Kyoto University, Kyoto 606-8501, Japan

Edited by Douglas C. Rees, Howard Hughes Medical Institute, Caltech, Pasadena, CA, and approved May 9, 2014 (received for review November 14, 2013)

Enterobacteriaceae produce antimicrobial peptides for survival un-der nutrient starvation. Microcin J25 (MccJ25) is an antimicrobialpeptide with a unique lasso topology. It is secreted by the ATP-binding cassette (ABC) exporter McjD, which ensures self-immunityof the producing strain through efficient export of the toxic maturepeptide from the cell. Here we have determined the crystal struc-ture of McjD from Escherichia coli at 2.7-Å resolution, which is tothe authors’ knowledge the first structure of an antibacterial pep-tide ABC transporter. Our functional and biochemical analyses dem-onstrate McjD-dependent immunity to MccJ25 through efflux ofthe peptide. McjD can directly bind MccJ25 and displays a basalATPase activity that is stimulated by MccJ25 in both detergentsolution and proteoliposomes. McjD adopts a new conformation,termed nucleotide-bound outward occluded. The new conforma-tion defines a clear cavity; mutagenesis and ligand binding studiesof the cavity have identified Phe86, Asn134, and Asn302 as impor-tant for recognition of MccJ25. Comparisons with the inward-openMsbA and outward-open Sav1866 structures show that McjD hasstructural similarities with both states without the intertwining oftransmembrane (TM) helices. The occluded state is formed by rota-tion of TMs 1 and 2 toward the equivalent TMs of the oppositemonomer, unlike Sav1866 where they intertwine with TMs 3–6 ofthe opposite monomer. Cysteine cross-linking studies on the McjDdimer in inside-out membrane vesicles of E. coli confirmed the pres-ence of the occluded state. We therefore propose that the outward-occluded state represents a transition intermediate between theoutward-open and inward-open conformation of ABC exporters.

antimicrobial peptide ABC exporter | membrane transporter crystalstructure | substrate binding | transport mechanism |microcin immunity protein

Microcins are gene-encoded antibacterial peptides of lowmolecular weight (<10 kDa), produced by Enterobacteriacea

(1). They are secreted under conditions of nutrient exhaustionthrough dedicated ATP-binding cassette (ABC) exporters and ex-ert potent antibacterial activity against closely related species (2).Microcin J25 (MccJ25) is a plasmid-encoded, ribosomally synthe-sized, and posttranslationally modified 21-aa antimicrobial peptide(3). Its 3D structure shows a unique lasso topology (4–6), with theC-terminal tail threading through an N-terminal eight-residuemacrolactam ring, where it is locked by bulky amino acid sidechains, thus forming a compact interlocked structure called thelasso fold (Fig. S1). This extraordinarily stable structure is appor-tioned into two regions: a loop involved in uptake of the microcininto sensitive bacteria and a ring/tail region that interacts with thecytoplasmic target of the antimicrobial peptide (1, 7, 8). MccJ25enters the target cell using the siderophore receptor FhuA (9), andinside the cell it inhibits the bacterial RNA polymerase (7, 8, 10).Four genes are required for the biosynthesis and export of MccJ25(11). The lasso topology is acquired by modification of a linear 58-aa precursor peptide (McjA) by two dedicated enzymes (McjB andMcjC) (12). The ABC transporter McjD ensures efficient export of

the toxic mature peptide out of the cell and simultaneously servesas a self-immunity strategy for the producing strain (11). Homologsof McjD and MccJ25-like defense systems can be identified inseveral genomes of bacterial pathogens (13).ABC exporters form a large superfamily of transmembrane

proteins responsible for the translocation across the membraneof a large diversity of substrates, ranging from small ions toamino acids, sugars, lipids, or peptides, using the energy of ATPhydrolysis. Some ABC exporters contribute to multidrug re-sistance. Bacterial ABC exporters are dimers, with each mono-mer composed of a transmembrane domain (TMD) consistingof six TM helices, which forms the translocation pathway acrossthe membrane bilayer and ensures the substrate specificity, anda nucleotide-binding domain (NBD) where binding and hydro-lysis of ATP take place. Biochemical and modeling studies, andthe crystal structures of the Escherichia coli lipid A transporterMsbA (14), the Staphylococcus aureus exporter Sav1866 (15),and others suggest that ABC exporters extrude their substratesout of the cell via an alternating access mechanism. However, thecurrent structures do not explain how the transition betweeninward-open and outward-open conformations occurs mechanis-tically. Here we have determined the high-resolution structure

Significance

ATP-binding cassette (ABC) exporters transport substrates byan alternating access mechanism that is driven by ATP bindingand hydrolysis. The general mechanism is a motion from aninward to an outward state, with a different intertwining ofthe half-transporters in both states. In this study we de-termined the function and crystal structure of the ABC exporterMcjD that exports the antibacterial peptide microcin J25. Ourstructure represents a novel nucleotide-bound, outward-occluded state. It does not possess subunit intertwining andshows a well-defined binding cavity that is closed to all sides,consistent with it being an intermediate between the inward-and outward-facing state. Our structure provides valuableinsights in a transition state of an ABC exporter.

Author contributions: S.R., H.W.v.V., and K.B. designed research; H.G.C., Z.T., I.M., Y.L., S.Z.,and K.B. performed research; S.I., S.R., H.W.v.V., and K.B. contributed new reagents/analytictools; H.G.C., Z.T., I.M., Y.L., S.Z., S.R., H.W.v.V., and K.B. analyzed data; and H.G.C., Z.T., Y.L.,S.Z., S.R., H.W.v.V., and K.B. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.

Freely available online through the PNAS open access option.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org (PDB ID code 4PL0).1H.G.C. and Z.T. contributed equally to this work.2To whom correspondence may be addressed. E-mail: kbeis@imperial.ac.uk, hwv20@cam.ac.uk, or rebuffat@mnhn.fr.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1320506111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1320506111 PNAS | June 24, 2014 | vol. 111 | no. 25 | 9145–9150

BIOCH

EMISTR

Y

at 2.7-Å resolution of the E. coli immunity-conferring ABCexporter McjD that is responsible for the export of the lassopeptide MccJ25. It displays a new conformation, outward-occludedand without intertwining of the TMDs, which is intermediatebetween the outward-open and inward-open state. In addition,the structure defines a clear binding cavity that can accommo-date one MccJ25 molecule. Our functional data in detergentsolution and proteoliposomes demonstrate that McjD mediatesMccJ25 transport in an ATP-dependent fashion and that in theabsence of Mccj25, the protein can mediate the transport oftypical substrates of multidrug transporters.

ResultsFunctional Characterization. Growth curves of E. coli strains pro-ducing MccJ25 with or without McjD showed that the immunityof cells to MccJ25 required the expression of McjD, whereassensitivity to MccJ25 was observed in the absence of McjD ex-pression (Fig. 1A). McjD-mediated resistance was also observedfor MccJ25 added to growth medium (Fig. S1). Transport mea-surements in these cells with fluorescent, BODIPY-labeledMccJ25 demonstrated that the MccJ25 resistance is due to re-duced peptide accumulation and active extrusion by the McjD

transporter (Fig. 1B). The interaction between McjD and MccJ25was further examined using purified McjD in detergent solutionand proteoliposomes (Figs. S2 and S3). Binding studies using mi-croscale thermophoresis (MST) showed that McjD bound MccJ25with a KD of 104 ± 52 μM (Fig. S4). McjD exhibited a basal ATPaseactivity in the absence of added transport substrate (Fig. S1); thebasal activity of McjD was measured at different ATP concen-trations and displayed a Km of 169.3 ± 6.7 μM and a Vmax of 44.4 ±0.5 nmol min−1 mg protein− 1. A Walker B mutant, E506Q,was used as a control to subtract ATP background hydrolysis andshowed significantly reduced activity, 2% relative to the wild-typeMcjD. The ATPase activity of McjD was inhibited by the specificinhibitors adenosine 5′-(β,γ-imido)triphosphate (AMP-PNP) andortho-vanadate, with an IC50 of 39.5 ± 0.6 μM and 61.6 ± 1.1 μM,respectively (Fig. S5). The ATPase activity of McjD reconstituted inproteoliposomes was stimulated by MccJ25 in a concentration-dependent manner (Fig. 1C). The ATPase activity of detergentpurified McjD could also be stimulated by MccJ25 at slightly lowerlevels compared with the reconstituted protein (Fig. 1D). Surpris-ingly, the McjD ATPase activity was also stimulated by Hoechst33342 (Fig. 1D), which is a typical substrate for multidrug trans-porters such as Sav1866 and MsbA (16, 17). Our observations onMcjD-mediated Hoechst 33342 transport in proteoliposomes, andMcjD-mediated Hoechst 33342 and ethidium transport in intactcells, underscore the relevance of structural comparisons betweenMcjD and these ABC exporters (Fig. 1 E and F and Fig. S1).

Overall McjD Structure. We determined the structure of McjD incomplex with the ATP analog AMP-PNP at 2.7-Å resolution(Table S1) by molecular replacement using the Salmonella typhi-murium MsbA (14) and S. aureus Sav1866 (15) as search models.The quality of the electron density allowed us to build the entiresequence, and the structure was refined to an Rfactor of 24.7% andRfree of 26.6%. Clear electron density was also observed for AMP-PNP, Mg2+, and two detergent molecules (Fig. S6). The overallarchitecture of McjD is similar to Sav1866 and MsbA multidrugABC exporters (Fig. 2A); the structure is a homodimer, and eachmonomer is composed of an N-terminal TMD (6 TM helices) anda C-terminal NBD. The transporter is 124 Å long, 55 Å wide, and51 Å deep. The dimer interface at the TMDs is formed betweenTM2 and TM5/TM6 from one monomer, with the equivalent TM5/TM6 and TM2 from the opposite monomer, burying a surface areaof ∼7,100 Å2 per subunit. The TM helices are connected by ex-tracellular (ECL) and intracellular (ICL) loops; ICL1 (betweenTM2 and TM3) and ICL2 (between TM4 and TM5) form thecoupling helices that interact with the NBD to transmit ATPbinding and hydrolysis-induced conformational changes to theTMDs. For this purpose ICL1 contacts both NBDs, whereas ICL2only interacts with the NBD from the opposite monomer, as ob-served for other ABC exporters. The overall conformation of theMcjD NBD is very similar to that of Sav1866 and MsbA. BecauseMcjD was cocrystallized with the ATP analog AMP-PNP andMgCl2, the NBDs are dimerized in the ATP-bound state with theP-loop and ABC signature motifs involved in the binding of thenucleotide at each nucleotide-binding site.

McjD Ligand Binding Cavity. The 12 TM helices form a large cavity,∼5,900 Å3, that is occluded from both sides of the membrane(Fig. 2 B–D). The cavity is aligned by ionizable (six Arg) residues, aswell as polar and hydrophobic residues (Fig. 2C). The cavity ofMcjD is approximately 40 Å long, with its widest point at 21 Å andthe narrowest point at 10 Å, perpendicular to the membrane. Inthe absence of direct structural information about the McjD–

MccJ25 complex, we constructed a model to identify potentialresidues that coordinate peptide binding. The NMR structureof MccJ25 shows that the lasso peptide is 23 Å long with a looplength of 10 Å, and is 18 Å wide at the lariat ring (4–6). In thecurrent conformation of McjD, we could only position MccJ25

Fig. 1. Functional characterization of McjD. (A) Growth of cytotoxic MccJ25-expressing E. coli requires coexpression of ABC exporter McjD (green, pro-duction of MccJ25 plus McjD; red, MccJ25 minus McjD; black, without MccJ25production). (B) Active efflux of fluorescent BODIPY-labeled MccJ25 fromcells results in reduced MccJ25 accumulation for McjD compared with theinactive E506Q McjD with equal expression (Fig. S3). (C) Stimulation of pu-rified McjD-ATPase by MccJ25 in proteoliposomes. (D) Ligand-stimulatedATPase activity of McjD in detergent solution. (E) Hoechst 33342 transport inproteoliposomes containing an equal amount of McjD or the E506Q mutant.ATP-dependent Hoechst 33342 transport from the membrane into the acidiclumen causes quenching of dye fluorescence for McjD. (F) Hoechst 33342efflux in intact cells. Inhibition of McjD with ortho-vanadate (Vi) increasesHoechst 33342 accumulation for McjD but not E506Q-McjD. Fluorescencetraces are typical for data obtained in three independent experiments usingindependent batches of cells and proteoliposomes. Error bars are shown forall measurements (mean ± SEM; n = 3); when error bars are not visible, theyare contained within the symbols.

9146 | www.pnas.org/cgi/doi/10.1073/pnas.1320506111 Choudhury et al.

in one orientation without further manipulation. The volume ofMccJ25 from the NMR structure was calculated to be 1784 Å3;it thus could fit inside the McjD cavity with some minor sterichindrance (Fig. S7); the lariat ring of MccJ25 is a very rigidstructure, and it could only be positioned facing toward the widerand more charged cytoplasmic side of the cavity. The cavity at theperiplasmic face is more narrow and hydrophobic and thus canaccommodate and bind residues in the loop region of MccJ25(Phe10, Ile13, Ile17, and Phe19). We have recently shown thatMccJ25 loses its β-strand structure at the loop region uponbinding to the outer membrane receptor FhuA (9). The lessstructured and wider loop conformation allows MccJ25 to in-teract with the FhuA barrel wall and extracellular loops for in-ternalization. This new conformation of MccJ25 could also beaccommodated within the McjD cavity. On the basis of the po-sition of MccJ25 into the McjD cavity, we constructed pointmutations along the cavity to identify residues involved in thebinding of MccJ25. Unlike wild-type McjD, the presence ofMccJ25 did not enhance the ATPase activity of the mutants.Furthermore, the binding affinity of the McjD mutants forMccJ25 was reduced by almost 10-fold (Fig. S4). These singlepoint mutations are unlikely to completely abolish MccJ25binding, but the absence of ligand-induced ATPase activity and

reduced binding affinity support the proposition that the resi-dues lining the McjD cavity are involved in the binding andtransport of MccJ25.Inspection of the McjD binding cavity revealed elongated

electron density close to the bottom of the ligand-binding cavitythat is formed between the two McjD monomers (Fig. S6); thehigh resolution allowed us to model and refine this density asnonyl-glucopyranoside (derived from the crystallization buffer).The residues in this region can bind inhibitors, as shown by thelow-resolution inward-facing structure of the mammalian multi-drug resistance ABC transporter P-glycoprotein in the presenceof the cyclic peptide inhibitors QZ59-SSS and QZ59-RRR (18).The two nonyl-glucopyranoside detergent molecules in the McjDstructure are bound in a similar location as the P-glycoproteininhibitors. However, the ATPase activity of purified McjD indetergent solution and in proteoliposomes was not significantlystimulated or inhibited by nonyl-glucopyranoside (Fig. S8). Fur-thermore, McjD expression in intact cells was not associated withenhanced resistance to nonyl-glucopyranoside compared with cellsexpressing nonactive E506Q-McjD (Fig. S8); McjD expression doesconfer resistance on cells to MccJ25 under these conditions (Fig. 1Aand Fig. S1C). We therefore conclude that nonyl-glucopyranosideis not a substrate or inhibitor of McjD. Under the current

Outside

Inside

TMD

NBD

Elbow Helix

McjD Sav1866

1 2

3 4

5 6

1

2

3

6 5

4

1

2

5

4

6

3

A B

C D

Fig. 2. Structure of the outward occluded conformation of McjD. (A) The McjD structure is viewed in the plane of the membrane. The membrane is shown ingray. The transmembrane helices of one subunit are numbered. Bound AMP-PNP is shown as sticks. (B) The cavity of McjD is shown as gray surface. (C) Theelectrostatic surface calculation of McjD cavity displays positive and negative charges to bind MccJ25. The cavity is outlined with a broken yellow line. Thesurfaces are colored from blue (positively charged regions) to red (negatively charged regions). Hydrophobic regions are shown as white. (D) Periplasmic viewof the outward occluded McjD (Left) and “winged” outward-open Sav1866 (pdb 2ONJ) (Right) structures show clear differences in helix packing and ac-cessibility to the interior chamber. The NBDs have been omitted for clarity.

Choudhury et al. PNAS | June 24, 2014 | vol. 111 | no. 25 | 9147

BIOCH

EMISTR

Y

crystallization conditions, the detergent contributes to the stabilityof the McjD dimer in this occluded state. The crystal packing mayfurther stabilize the current conformation, even though we do notobserve significant crystal contacts along the TMs (Fig. S6).

Cysteine Cross-Linking. In the presence of AMP-PNP, both theMsbA and Sav1866 structures adopt a “winged” outward-openconformation. Because the McjD structure in the presence ofAMP-PNP is in an occluded state, we used a predictive cysteine-cross approach to further characterize this new conformation. Asthe wild-type protein contains four cysteines, we first constructeda Cys-less McjD protein (McjD-cl). This mutant is functionalbecause it shows a basal ATPase and ethidium efflux activitycomparable to wild-type protein and mediates cellular resistanceto MccJ25 (Fig. S1). Next we introduced a single cysteine at po-sition 53 (McjD-L53C), which allows intermolecular cross-linkingbetween the extracellular loops connecting TM1-TM2 and TM1’-TM2’ in the ATP-bound McjD dimer in a manner that is specificfor the occluded state and that excludes the “winged” outward-facing state. ISOVs containing McjD-L53C and detergent purifiedprotein were preincubated with nucleotides for 10 min before theaddition of the oxidizing agent CuCl2 (SI Materials and Methods).McjD-cl did not display any cross-linking. In addition, the apoMcjD-L53C preparation did not contain significant amounts ofpreformed cross-linked dimer in the absence of CuCl2. In thepresence of the oxidizing agent CuCl2 and AMP-PNP or ATP ±Mg2+, intermolecular cross-linking in the McjD dimer verified theexistence of the nucleotide-bound occluded state in both E. colimembranes and detergent solution (Fig. 3). Apo McjD can becross-linked as effectively as the AMP-PNP–bound state becausethe L53C residues are also in close proximity in an inward-facing,MsbA-based model of McjD. Because the new McjD conformationwas crystallized in the absence of MccJ25, we also tested the effectof MccJ25 addition on intermolecular cross-linking in the McjDL53C dimer in E. coli membrane vesicles. Preincubation of McjDwith MccJ25 did not alter the cross-linking behavior in the absenceof nucleotides (Fig. S1), suggesting that, upon binding of MccJ25to inward-facing McjD, the TMs do not undergo major confor-mational changes.

DiscussionComparison with Other ABC Exporters. McjD was cocrystallizedwith the ATP analog AMP-PNP and MgCl2. The two NBDs arein the ATP-bound state in which the P-loop and ABC signaturemotifs tightly interact with the nucleotide. ATP binding andhydrolysis at the NBD site trigger conformational changes atthe TMDs, and these changes are transmitted through the twocoupling helices ICL 1 and 2, that link the NBDs to the TMs.ICL1 contacts both NBDs, whereas ICL2 only interacts with theNBD from the opposite monomer, as with other ABC exporters.The overall conformation of the McjD NBDs is very similar tothat of Sav1866.The most striking feature of McjD is that it is occluded on

both sides of the membrane. Comparison of McjD with theoutward open structures of Sav1866 and MsbA reveals that boththe cytoplasmic and periplasmic sides of the McjD dimer areoccluded, unlike the Sav1866 and MsbA structures, which areopen at the periplasmic side (Figs. 2D and 4 A and B). A modelof MsbA and Sav1866 based on McjD results in an occludedcavity of similar size as found in McjD. McjD can be aligned withthe outward-open Sav1866 and MsbA at the TM region (330 Caatoms) with an rmsd of 2.3 Å and 2.2 Å, respectively. TheSav1866 and MsbA dimer structures contain two “wings” formedby the TMs 1 and 2 of one subunit and TMs 3–6 of the othersubunit. In McjD, TMs 1 and 2 have rotated symmetrically by 26°along Leu29 and Gln90 and shifted by 6 Å at ECL1 relative toSav1866 (Fig. 4C). As a result, TM1 is located in the vicinity ofTM6 of the same monomer and TM1 of the adjacent monomer.

TM2 is close to TM5 from the adjacent monomer; TM5 bendsby 25° at Leu264 toward TM2. This configuration of TMs causesthe occlusion at the periplasmic side and is consistent with the

McjD-L53C + CuCl2

- ATP AMP-PNP

- Mg - Mg

CLD

M

A

98 63

40

98 62 49

CLD

M

CLD

M

B

C

D

McjD-L53C - CuCl2

- ATP AMP-PNP

- Mg - Mg

98-

62-

49-

F G

- + - +

E

McjD-cl CuCl2

pET28b CuCl2

Fig. 3. Predictive cross-linking of McjD. The reaction conditions for eachlane are indicated above the gels. McjD-L53C was preincubated with ap-propriate nucleotides for 10 min before cross-linking reactions; cross-linkingof McjD-L53C was performed in the presence of 1 mM CuCl2 for 30 min atroom temperature. (A) Western-blot analysis of McjD-L53C cross-linked inISOVs. Nucleotide dependence of intermolecular cysteine cross-linking be-tween the McjD monomers at position 53 is consistent with the formation ofan outward-facing occluded state in the ATP-bound state. CLD denotes theformation of the cross-linking dimer and M the monomer in the absence ofCuCl2. Leu53Cys mutation is located in the loop that connects TM 1 and 2that form the occluded structure; the formation of the cross-linking dimerin the presence of nucleotides verifies that the nucleotide-bound occludedstate can also exist in the plasma membrane of E. coli. (B) Coomassie-stainedSDS/PAGE gel of cross-linked detergent purified McjD. (C) The degree ofcross-linking of detergent-purified McjD was estimated in the presence of7-diethylamino-3-(4’-maleimidylphenyl)-4-methylcoumarin (cpm) that becomesfluorescent upon binding to free cysteines. Formation of the cysteine cross-links resulted in reduced available free cysteines for labeling by the cpm;a small degree of nonspecific binding is observed for the dimeric species bycpm. (D) The relative intensity for each band was estimated by densitometry ofthe Western blot, (E) Coomassie-stained, and (F) in-gel fluorescence gel; thesignal is displayed as fold change in intensity relative to the non–cross-linkedreactions. In the absence of CuCl2, the fluorescent signal intensity is high forthe monomeric species, M, whereas formation of the CLD has resulted ina very reduced signal. Error bars are shown for all densitometry measurements(mean ± SEM; n = 3). A very small percentage of monomeric species is stillpresent when the cross-linking reactions are terminated. Gray bars representfraction of monomeric species relative to the total protein present. (G) Viewfrom the periplasm of the Leu53 side chain position. Coloring scheme as in Fig.2A. The L53 side chain is shown as magenta stick.

9148 | www.pnas.org/cgi/doi/10.1073/pnas.1320506111 Choudhury et al.

observations on cysteine cross-linking at position L53C in theapo and AMP-PNP bound conformations of McjD (Fig. 3 andFig. S1 F and G). Another feature of McjD is the presence ofthree salt bridges between Lys80 and Glu301 (from the oppositemonomer), Arg141 and Glu309, and Arg195 and Asp135; thesesalt bridges seem to stabilize packing of TM2 to TM6 (of theopposite monomer), TM3 to TM6, and TM3 to TM4 in ouroccluded state (Fig. 4D). For comparison, the outward-openSav1866 is stabilized by salt bridges between Glu78 from TM2and Arg295 from TM6 of the opposite monomer. We havepreviously shown that Glu314 in the ABC multidrug transporterLmrA from Lactococcus lactis, which is equipositional to Glu309

in McjD, is important for LmrA-mediated efflux of ethidium andother substrates (19). McjD shares 19% sequence identity at theTMD (residue 1–342) with LmrA. We investigated the role ofGlu309 for McjD activity by mutating this residue to Gln andAla. Our ethidium transport data in whole cells (Fig. 4E) arereminiscent of those obtained for LmrA (19) and show a lowresidual ethidium efflux activity for the mutants compared withwild-type protein. Therefore, the stabilization of the TMs in theMcjD dimer by the salt bridges supports the important role ofour occluded state in the propagation of the transport cycle. Wealso investigated whether McjD can adopt an outward-open stateby performing the Leu53Cys cross-linking experiment at pH 4.5(Fig. S1 H and I), where salt bridges in the McjD dimer can bedisrupted through protonation of relevant carboxylates. In the ab-sence of nucleotides, a substantial population of McjD formeda cross-linked dimer under these conditions, whereas dimer for-mation was not observed in the presence of ATP or AMP-PNP.These results suggest that McjD can sample an outward-openconformation similar to Sav1866.Our work on McjD describes a novel conformation for ABC

exporters, a nucleotide-bound outward-facing occluded state.A nucleotide-bound occluded conformation has been previouslyreported for the type I ABC importer MBP-MalFGK2 (20) andtype II ABC importer BtuCD-F (21). The cavity in MalFGK2 is notcompletely occluded at the TMD, but the MBP protein occludes itat the periplasmic side. In contrast, the BtuCD-F cavity is occludedat the TMD by the periplasmic and cytoplasmic gate. The existenceof a conformation in which the NBDs are dimerized withoutopening of the TMD at their external side was also recently observedfor MsbA in cysteine cross-linking and FRET studies (22). OurMcjD conformation is therefore relevant for other ABC exporters.

Mechanistic Implication.ABC exporters need to alternate betweeninward-facing and outward-facing conformations to transporttheir substrates from the cytoplasm to the periplasm in an ATP-dependent manner. The proposed mechanism for exporters is an

C TM1 TM2

TM1 TM2

Before rotation

TM1 TM2

After rotation

A BTM1 TM2 TM1 TM2

D K80

D135

R141

R195

E309

E

2.6Å

3.4Å

3.0Å

3.4Å

2.9Å

Fig. 4. Conformational changes in McjD. The McjD transmembrane helicesof one subunit are colored as in Fig. 2A. Superposition of McjD with (A)Sav1866 (outward open, gray) and (B) Vibrio cholerae MsbA (inward closedapo, gray). McjD shows structural similarities to the cytoplasmic TM region ofSav1866 and the periplasmic region of MsbA. The most significant confor-mational changes between McjD and Sav1866 are in TMs 1 and 2. (C) Super-position of TMs 1 and 2 of Sav1866 onto McjD (Left); the Sav1866 TMs 1 and 2have been rotated around the axis shown (Right). (D) The salt bridges that arestabilizing the occluded conformation are shown as sticks. (Upper) Residuesinvolved in stabilizing TM2 to TM6 (of the opposite monomer, shown in gray).(Lower) Residues that stabilize TM3 to TM6 and TM3 to TM4. The salt-bridgedistances are in Å and shown as black broken lines. Color scheme as in Fig. 2A.(E) Disruption of the Glu309-R141 salt bridge by E309Q and -A mutationsresults in reduced ethidium bromide efflux in whole cells compared with wild-type McjD. The transport-inactive E506Q mutant is used as a control.

(i) Inward-open (iii) Outward-occluded (ii) Inward-closed

TMD

NBD 2 x ATP

2 x (ADP + Pi)

(v) Outward-occluded

(iv) Outward-open

Fig. 5. Proposed role of the outward-occluded state in the mechanism ofABC exporters. Ligand binding to the inward-open (apo) conformation (state i)facilitates transition of the transporter into the inward-closed (apo) con-formation (state ii). ATP binding is associated with a transition into a hypo-thetical occluded conformation (state iii), and subsequent formation of anucleotide-bound outward-open conformation (state iv). Upon release ofthe ligand, the ABC exporter adopts an outward-occluded conformation(state v) (current McjD structure). ATP hydrolysis resets the transporter backto the inward-facing conformation (state i). The TMs 1 and 2 from eachmonomer are highlighted in dark blue and red. The substrate is shown asa yellow diamond.

Choudhury et al. PNAS | June 24, 2014 | vol. 111 | no. 25 | 9149

BIOCH

EMISTR

Y

alternating access model with subunit intertwining (14, 15); thetransporter cycles between an inward- and outward-facing con-formation in which the TM 4 and 5 hairpin is exchanged betweenthe TMDs of the half-transporters in the inward-facing dimer,whereas the TM 1 and 2 hairpin is exchanged in the outward-facing dimer. In the inward conformation the substrate can bindin the open cavity between the TMDs, which in turn stimulatesbinding of ATP (22). The affinity of McjD for the mature MccJ25peptide in detergent solution (apparent KD of 104 μM) mightindicate that under physiological conditions substrate binding toMcjD occurs from the cytoplasmic membrane in which MccJ25accumulates. Our point mutations are associated with an almost10-fold reduction in the experimentally determined binding af-finity. In the inward-facing apo conformation the two NBDs areseparated, and binding of the ATP brings the two NBDs together.Formation of the ATP-bound outward-open state disrupts thebinding cavity causing release of MccJ25 at the external side of themembrane. MccJ25 would then be exported extracellularly in aTolC-dependent fashion (23). ATP hydrolysis finally resets theMcjD transporter back to the inward-facing conformation.Previously published models do not explain in detail by what

mechanism the transition between the two conformations occurs.The nucleotide-bound outward occluded McjD has structural sim-ilarities to the closed cytoplasmic side of outward-open Sav1866and MsbA in which the NBDs are dimerized, and the closedperiplasmic side of inward-open apo MsbA (Fig. 4 A and B). Assuch, it represents an intermediate state between the outward-facingconformation and the inward-facing conformation (Fig. 5). After theligand has left the outward-open cavity, TMs 1 and 2 rotate awayfrom TMs 3–6 of the opposite monomer to form the outward-occluded state without TMD intertwining (current McjD structure).We propose that the initial conformational changes associatedwith this transition are ATP-independent.Transition from the outward occluded state to the inward-open

conformation will require ATP hydrolysis, which disrupts the NBDdimer interface and induces intertwining of TMs 1–3, 6 from onemonomer with TMs 4 and 5 from the opposite monomer. After thesystem reverses back to the inward conformation, a substrate-bound

form of the nucleotide-bound outward occluded state might occurduring the ATP-dependent transition from the inward-facing stateto the outward-facing state (Fig. 5); in BtuCD-F the equivalentoccluded ATP-bound state was trapped with engineered disulphidebonds (21).In summary, to the authors’ knowledge we have generated the

first high-resolution structure of an antibacterial peptide (microcin)ABC exporter. Our structure of McjD represents a novel outward-occluded intermediate between the outward-facing state andinward-facing state of ABC exporters.

Materials and MethodsDetailed materials and methods are available online (SI Materials andMethods). McjD was cloned as a fusion with a C-terminal GFP that containeda His6-tag and tobacco etch virus (TEV) site. The protein was expressed inE. coli and purified in 0.03% n-Dodecyl-β-D-maltopyranoside. The GFP-tag wasremoved by TEV protease before crystallization and biochemical analyses indetergent solution. For functional analyses in intact cells and proteolipo-somes, non–GFP-tagged McjD was expressed in a drug-hypersensitive E. colithat lacks the major endogenous drug efflux pump AcrAB-TolC. The ATPaseactivity of purified proteins in detergent solution and proteoliposomeswas measured in an enzymatic-coupled assay and by the malachite greenmethod, respectively. Transport in intact cells and proteoliposomes wasdetermined by fluorimetry. Ligand binding was assessed by MST. Cysteinecross-linking studies were performed in ISOVs and detergent-purified McjDprotein. McjD crystals were grown in the presence of 10 mM AMP-PNP, 2.5mM MgCl2, and nonyl-glucopyranoside. The structure of McjD was solved bymolecular replacement using the MsbA (14) and Sav1866 (15) structures assearch models. The final structure was refined to a resolution of 2.7 Å.

ACKNOWLEDGMENTS. We thank Diamond Light Source for beam timeallocation and access; Prof. Xiaodong Zhang for access to the MonolithNT.115 for MST measurements; David Charles for technical assistance;Dr. Alexander Cameron and Yusuke Sekiguchi for kindly providing pureAsbT-GFP protein for MST measurements; Christophe Goulard for purifyingthe MccJ25 peptide ligands used in this study; and Prof. Robert Ford andDr. Alexander Cameron for critical reading of the manuscript. Financialsupport was provided by the Biotechnology and Biological Sciences Re-search Council (Grants BB/H01778X/1 to K.B., BB/G023425/1 to S.I., andBB/I002383/1 to H.W.v.V.) and Wellcome Trust (MPL: WT/099165/Z/12/Z toS.I.). I.M. is supported by a Ministry of Science, Technology and Innovationpostgraduate scholarship.

1. Duquesne S, Destoumieux-Garzón D, Peduzzi J, Rebuffat S (2007) Microcins, gene-encoded antibacterial peptides from enterobacteria. Nat Prod Rep 24(4):708–734.

2. Baquero F, Bouanchaud D, Martinez-Perez MC, Fernandez C (1978) Microcin plasmids:A group of extrachromosomal elements coding for low-molecular-weight antibioticsin Escherichia coli. J Bacteriol 135(2):342–347.

3. Rebuffat S, Blond A, Destoumieux-Garzón D, Goulard C, Peduzzi J (2004) Microcin J25,from the macrocyclic to the lasso structure: Implications for biosynthetic, evolutionaryand biotechnological perspectives. Curr Protein Pept Sci 5(5):383–391.

4. Rosengren KJ, et al. (2003) Microcin J25 has a threaded sidechain-to-backbone ringstructure and not a head-to-tail cyclized backbone. J AmChem Soc 125(41):12464–12474.

5. Wilson KA, et al. (2003) Structure of microcin J25, a peptide inhibitor of bacterial RNApolymerase, is a lassoed tail. J Am Chem Soc 125(41):12475–12483.

6. Bayro MJ, et al. (2003) Structure of antibacterial peptide microcin J25: A 21-residuelariat protoknot. J Am Chem Soc 125(41):12382–12383.

7. Destoumieux-Garzón D, et al. (2005) The iron-siderophore transporter FhuA is thereceptor for the antimicrobial peptide microcin J25: Role of the microcin Val11-Pro16beta-hairpin region in the recognition mechanism. Biochem J 389(Pt 3):869–876.

8. Semenova E, Yuzenkova Y, Peduzzi J, Rebuffat S, Severinov K (2005) Structure-activityanalysis of microcinJ25: Distinct parts of the threaded lasso molecule are responsiblefor interaction with bacterial RNA polymerase. J Bacteriol 187(11):3859–3863.

9. Mathavan I, et al. (2014) Structural basis for hijacking siderophore receptors by an-timicrobial lasso peptides. Nat Chem Biol 10(5):340–342.

10. Mathavan I, Beis K (2012) The role of bacterial membrane proteins in the in-ternalization of microcin MccJ25 and MccB17. Biochem Soc Trans 40(6):1539–1543.

11. Solbiati JO, et al. (1999) Sequence analysis of the four plasmid genes required toproduce the circular peptide antibiotic microcin J25. J Bacteriol 181(8):2659–2662.

12. Duquesne S, et al. (2007) Two enzymes catalyze the maturation of a lasso peptide inEscherichia coli. Chem Biol 14(7):793–803.

13. Maksimov MO, Pelczer I, Link AJ (2012) Precursor-centric genome-mining approachfor lasso peptide discovery. Proc Natl Acad Sci USA 109(38):15223–15228.

14. Ward A, Reyes CL, Yu J, Roth CB, Chang G (2007) Flexibility in the ABC transporterMsbA: Alternating access with a twist. Proc Natl Acad Sci USA 104(48):19005–19010.

15. Dawson RJ, Locher KP (2006) Structure of a bacterial multidrug ABC transporter.Nature 443(7108):180–185.

16. Reuter G, et al. (2003) The ATP binding cassette multidrug transporter LmrA and lipidtransporter MsbA have overlapping substrate specificities. J Biol Chem 278(37):35193–35198.

17. Velamakanni S, Yao Y, Gutmann DA, van Veen HW (2008) Multidrug transport bythe ABC transporter Sav1866 from Staphylococcus aureus. Biochemistry 47(35):9300–9308.

18. Aller SG, et al. (2009) Structure of P-glycoprotein reveals a molecular basis for poly-specific drug binding. Science 323(5922):1718–1722.

19. Shilling R, et al. (2005) A critical role of a carboxylate in proton conduction by theATP-binding cassette multidrug transporter LmrA. FASEB J 19(12):1698–1700.

20. Oldham ML, Khare D, Quiocho FA, Davidson AL, Chen J (2007) Crystal structure of acatalytic intermediate of the maltose transporter. Nature 450(7169):515–521.

21. Korkhov VM, Mireku SA, Locher KP (2012) Structure of AMP-PNP-bound vitamin B12transporter BtuCD-F. Nature 490(7420):367–372.

22. Doshi R, van Veen HW (2013) Substrate binding stabilizes a pre-translocation in-termediate in the ATP-binding cassette transport protein MsbA. J Biol Chem 288(30):21638–21647.

23. Delgado MA, Solbiati JO, Chiuchiolo MJ, Farías RN, Salomón RA (1999) Escherichia coliouter membrane protein TolC is involved in production of the peptide antibioticmicrocin J25. J Bacteriol 181(6):1968–1970.

9150 | www.pnas.org/cgi/doi/10.1073/pnas.1320506111 Choudhury et al.

Supporting InformationChoudhury et al. 10.1073/pnas.1320506111SI Materials and MethodsGene Disruption. Gene replacement of mcjD on the pTUC202plasmid by a kanamycin resistance gene was performed using λRed recombinase-mediated PCR targeting mutagenesis (1). TheKanR cassette was amplified from pKD4 plasmid, and primersused were as follows: forward 5′-CATTCAACTTCAATCCAT-TGATTATAAAGGTTAATTTGTGTAGGCTGGAGCTGCT-TC-3′ and reverse 5′-ACTCCCTTTCTGAATACAGAGCAG-GAGTCAGATATTCATATGAATATCCTCCTTA-3′ for mcjDdisruption (homologous regions to the targeted genes are un-derlined). The resulting plasmid pTU202-ΔmcjD:: kan were verifiedby PCR.

Growth Curve Determinations. For the data in Fig. 1A, Escherichiacoli MC4100 strain harboring pTUC202 or pTUC202-ΔmcjD::kan for plasmid-dependent production of MccJ25 (2) was grownovernight in LB medium supplemented with kanamycin (34μg/mL) at 37 °C. E. coliMC4100 harboring an inactive pTUC202delta6-6 plasmid, which does not lead to the production ofMccJ25, was used as a control. The cells were washed three timeswith M63 medium [13.6 g/L KH2PO4, 2 g/L (NH4)2SO4, 1 g/Lcasamino acids] and added to 200 μL M63 medium supple-mented with 0.2% glucose, 2 mM MgSO4, and 1 mg/L vitaminB1 in a 96-microtiter plate to OD600 around 0.1. Cells weregrown at 37 °C with shaking, and the optical density at 600 nmwas recorded in a microplate reader POLARstar Omega (BMGLabtec) every hour over a period of 36 h.For growth experiments in the presence of exogenously added

MccJ25 (Fig. S1C) or nonyl-glucopyranoside (NG) (Fig. S8C),wild-type McjD, cys-less McjD, or E506Q-McjD were expressedin drug-hypersensitive E. coli ΔacrA ΔacrB (DE3) as describedbelow (“Substrate transport in intact cells”) and diluted in thewells of a 96-well plate under a thin layer of silicon oil to an OD660of 0.05 in growth medium supplemented with 1 mM IPTG andMccJ25 or nonylglucoside at concentrations as indicated in figurelegends. The optical density at 600 nm was then followed over timeat 25 °C in a Versamax microplate reader (Molecular Devices).

Protein Expression and Purification. The mcjD was PCR amplifiedfrom the pTUC202 vector that carries the mcjABCD genes forMccJ25 expression (2) and cloned into pWaldoGFPd (3). Forthe generation of purified McjD protein for crystallography andbiochemical analyses in detergent solution, the protein was ex-pressed overnight at 25 °C in E. coli C43 (DE3) cells by theaddition of 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG)at OD600 0.4. All purification procedures were performed at 4 °Cunless otherwise stated. Inside-out membrane vesicles (ISOVs)were isolated from 10-L cultures and solubilized in 1% n-Dodecyl-β-D-maltopyranoside (DDM) for 1 h in 1× PBS buffer.The solubilized suspension was spun at 120,000 × g for 1 h toremove unsolubilized material. The supernatant was brought to20 mM imidazole and incubated with 1 mL Ni-NTA Superflowresin (Qiagen) per 1 mg of GFP. The slurry was incubated withstirring for 4 h. The slurry was loaded on an Econo-Column(Biorad) and washed with 10 CV of 1× PBS containing 30 mMimidazole and 0.1% DDM. The McjD-GFP-His8 fusion waseluted in the same buffer containing 250 mM imidazole. Theeluted protein was dialyzed overnight in the presence of stoi-chiometric amounts of His6-tagged TEV protease in 20 mM Tris(pH 8.0), 150 mM NaCl, and 0.03% DDM. The dialyzed samplewas passed over a 5-mL His-Trap column (GE Healthcare)to remove uncleaved material, TEV-His6, GFP-His8, and the

flow-through containing McjD was collected. The protein wasconcentrated using a 100-KDa cutoff concentrator and injectedin a Superdex S200 10/300 size exclusion column (GE Health-care) equilibrated with the dialysis buffer. The fraction corre-sponding to McjD was concentrated to 10 mg mL−1 for crys-tallization (Fig. S2).

Peptide Production and Purification. The peptide MccJ25 (2) wasexpressed and purified as previously described. Briefly, MccJ25was first purified from the supernatants of the producing strainby solid phase extraction using Sepak Plus C8 cartridges (Waters).The fractions containing the peptide were further purifiedby HPLC on a Capcell reverse-phase C18 semipreparativecolumn (300 × 7 mm, 5 μm, 120 Å; Shiseido). To label MccJ25with the BODIPY fluorophore, a single-substituted variant ofMccJ25[I13K] was generated by site-directed mutagenesis. La-beling reaction was done by mixing 300 μL of MccJ25[I13K] (10mg/mL) in 0.1 M NaHCO3 (pH 8.3) with 100 μL BODIPY FL-XSE (Life Technologies) (10 mg/mL) in N,N-Dimethylformamidewith agitation at 37 °C for 1 h. The resulting fluorescent MccJ25was purified by HPLC on a Capcell reverse-phase C18 column(250 × 4.6 mm, 5 μm, 120 Å; Shiseido).

Site-Direct Mutagenesis. Mutants were generated using the Quik-Change Site-Directed Mutagenesis Kit (Agilent Technologies).DNA sequencing of the resulting plasmids confirmed the presenceof themutants. All mutants were purified at similar purity as judgedby SDS/PAGE (Figs. S2 and S3).

ATPase Activity of Purified McjD in Detergent Solution. ATPase ac-tivities were measured in 0.03% DDM solution using a coupledassay (4) (Enzcheck, Molecular Probes). All ATPase activity datawere measured at 25 °C. The affinity and rate of ATP hydrolysis,Km and Vmax, were measured at different concentrations of ATP.All of the other ATPase measurements were performed at 1 mMATP. To subtract the background from impurities or hydrolysisof ATP by thermal degradation a Walker B mutant (E506Q) wasconstructed. The McjD concentration was 1 μM for all experi-ments. For ligand-induced ATPase activity assays, MccJ25 wasdissolved in 100% MeOH and added to the protein at a finalconcentration of 0.5 mM. To verify that MeOH had no inhibitoryeffect on the protein or the coupled assay, control ATPase assayswere performed in the presence of the same percentage ofMeOH as for the ligands. Hoechst 33342 was dissolved in waterand was added to a final concentration of 100 μM.The effect of NG on the ATPase activity of McjD was assessed

(Fig. S8 A and B). The effects of specific ATPase inhibitors,orthovanadate and adenosine 5′-(β,γ-imido)triphosphate (AMP-PNP), on the McjD activity (Fig. S5) were also measured. Ac-tivities were plotted as percentage of wild-type basal ATPaseactivity in the presence of 1 mM ATP (35 nmol min−1 mg pro-tein−1). All experiments were performed in triplicate. The datawere fitted to the Michaelis–Menten equation by nonlinear re-gression using GraphPad Prism software.

Substrate Transport in Intact Cells. For functional analyses in cellsand proteoliposomes, GFP-less wild-type McjD and E506Q-McjD were expressed from plasmids pET28-McjD and pET28-McjD-E506Q in drug-hypersensitive E. coli ΔacrA ΔacrB (DE3)cells lacking the tripartite multidrug transporter AcrAB-TolCand grown in Luria Broth containing 25 mM glucose and 50 μg/mLkanamycin. After the addition of 1 mM IPTG to the cell culture atOD600 of 0.4 and overnight incubation at 25 °C, the McjD proteins

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 1 of 14

were expressed at an equal level in the plasma membrane (Fig. S3).Cell pellets from 50 mL culture were washed once with 100 mMTris·HCl (pH 8.0). To increase the permeability of the outermembrane for drugs, cells were treated with 0.8 mM K-EDTA for10 min at 25 °C and finally resuspended to OD600 of 5 in 50 mMTris·Cl (pH 8.0) and 5 mM MgSO4 buffer. For measurements ofdye transport, cells were diluted to OD600 of 0.5 in a 2-mL glasscuvette and allowed to generate metabolic energy for 3 min afterthe addition of 25 mM glucose. Ethidium (2 μM) or Hoechst 33342(0.025 μM) was added, after which the dye fluorescence wasmeasured in a Perkin-Elmer LS 55B Fluorescence Spectrometer atexcitation and emission wavelengths of 500 nm and 580 nm withslit widths of 5 nm and 10 nm, respectively, for ethidium, and 355nm and 457 nm with slit widths of 5 nm and 2.5 nm, respectively,for Hoechst 33342. To inhibit the activity of McjD, 0.2 mM Na-orthovanadate was added to the external buffer before the additionof glucose and was allowed to accumulate in the cells for 8 minbefore the addition of Hoechst 33342.For measurements of BODIPY-MccJ25 transport by spec-

trofluorometry, cells were diluted to OD600 of 0.5 in a 2-mLEppendorf tube and allowed to generate metabolic energy for3 min through the addition of 25 mM glucose. Equal amounts ofcells were pipetted out at t = 0 min, 10 min, 20 min, and 30 min,centrifuged for 5 min at 12,000 × g, and then resuspended with200 μL buffer of 50 mM Tris·Cl (pH 8.0) and 5 mM MgSO4.MccJ25 was added to 5 μM after the first sample was taken att = 0 min. Samples were diluted 10-fold in Tris buffer in a 96-wellplate, and the fluorescence was monitored using SpectramaxGemini XS plate reader at excitation and emission wavelengthsof 488 nm and 513 nm, respectively.

Functional Reconstitution of McjD in Proteoliposomes. Twelve mil-ligrams of phospholipids (prepared from acetone/ether-washedE. coli total lipid extract and L-α-phosphatidylcholine in a ratioof 3:1) in chloroform were dried by evaporation under N2 gas.Lipids were then hydrated in buffer I (10 mM Tris·HCl, 10 mMKPi, 150 mM NaCl, and 150 mM NH4Cl, pH 6.8) to 4 mg/mLUnilamellar liposomes were prepared by extruding the lipidmixture 11 times through a polycarbonate membrane (pore di-ameter = 400 nm), followed by dilution to 1 mg/mL. DDM(0.05%) was then added to the liposome suspension. Purifiedwild type or mutant E506-McjD in elution buffer, or an equalvolume of elution buffer as a control, were added in a 1:50protein/lipid ratio (wt/wt). The mixture was incubated at 20 °Cunder mild shaking for 30 min. DDM was subsequently removedby incubation with 80 mg/mL and 8 mg/mL Bio-Beads for 2 h at20 °C and 4 °C, respectively, followed by an overnight incubationwith 8 mg/mL Bio-Beads at 4 °C. Proteoliposomes were collectedby centrifugation at 12,000 × g for 30 min, resuspended to 0.1 mgprotein/mL in buffer I, and immediately used in experiments. Theproteoliposomes contained equal amounts of McjD or E506Q-McjD(Fig. S3A).

Hoechst 33342 Transport in Proteoliposomes. Proteoliposomes con-taining wild-type or mutant E506Q-McjD (0.5 μg in buffer I) werediluted 1:200 in buffer II (10 mM Tris·HCl, 10 mM KPi, 300 mMNaCl, and 5 mM MgSO4, pH 7.5) and incubated in the presenceof 2 mMNa-ATP (pH 7.0) for 1 min. Subsequently, Hoechst 33342was added to a final concentration of 0.025 μM, and changes in thefluorescence were monitored in a 2-mL glass cuvette using a Per-kin-Elmer LS-55B Fluorescence Spectrometer at excitation andemission wavelengths of 355 nm and 457 nm, and slit widths of10 nm and 5 nm, respectively.

McjD ATPase in Proteoliposomes. The ATPase activity of McjD wasmeasured in reconstituted proteoliposomes in the absence orpresence of MccJ25 or nonyl-glucopyranoside, using the Mala-chite Green assay (5). Proteoliposomes (1 μg in buffer I without

150 mM NH4Cl, in which 10 mM KPi was replaced by 10 mMK-Hepes) were diluted 10-fold in 50 mM K-Hepes supplementedwith 5 mM MgSO4 and 2.5 mM Na-ATP (pH 7.0). MccJ25 wasdissolved in 100% MeOH and was added to the incubationmixtures together with solvent to give a constant MeOH con-centration of 25% (vol/vol). After incubation at 30 °C for 3 min,30-μL aliquots were mixed with 150 μL activated MalachiteGreen reagent in a 96-well plate for 5 min and subsequentlyincubated with 40 μL 34% citric acid for 25 min. The absorbancewas then determined at 600 nm and compared with values ob-tained for Pi standards.

Microscale Thermophoresis. Microscale thermophoresis measure-ments (Fig. S4) were performed using Monolith NT.115 (Nano-temper). All measurements were performed in standard coatedcapillaries at 25 °C. The laser power was set at 80% output andwas on for 30 s. Binding reactions for the McjD-GFP constructs,wild type and mutants, were set at a protein concentration of 60nM. For each protein measurement, a titration series was pre-pared with increasing concentration of MccJ25. MccJ25 could notbe titrated at higher concentration than 700 μM because it startprecipitating inside the capillaries; it could only be maintainedsoluble at high solvent concentrations, which significantly reducedthe ATPase activity of McjD. All curves were plotted usingGraphPad Prism. Under the experimental conditions KD is theconcentration at which half of the ligand is bound. Thermopho-resis signals were normalized to the fraction bound (X) by X = (Y(c) − Min)/(Max − Min). The bile acid transporter AsbT-GFP (6)was used to determine nonspecific binding of MccJ25; the KD wascalculated to be >4 mM. All experiments were performed intriplicate (mean ± SEM, n = 3).

Cysteine Cross-Linking. ISOVs containing McjD-cl, McjD-L53C, orISOVs carrying empty pET28b vector were used for cysteine cross-linking assays and were analyzed by Western blotting as previouslydescribed (7, 8). For cysteine cross-linking in ISOVs, the membraneswere adjusted to 10 μg/μL of total protein concentration, and fordetergent purified protein at 0.1 mg/mL. Where appropriate, McjDwas preincubated with 3 mM nucleotides (with or without MgCl2)for 10 min before treatment with CuCl2. For cross-linking experi-ments in the presence of peptide, the ISOVs were also incubatedwith 100 μM MccJ25. Cross-linking was performed in the presenceof 1 mM CuCl2 for 30 min at room temperature. Reactions werequenched with the addition of 2 mM N-ethylmaleimide. The degreeof cross-linking in the detergent-purified McjD was estimated inthe presence of 7-diethylamino-3-(4′-maleimidylphenyl)-4-methyl-coumarin) (cpm) that becomes fluorescent upon binding to freecysteines (no prior treatment with N-ethylmaleimide). The reactionswere analyzed on nonreducing SDS/PAGE, and the gels were im-aged using an ImageQuant LAS4000 (GE Healthcare). Densitom-etry analysis of the gel bands intensities was performed using theImageQuant TL software (GE Healthcare).

Crystallization. Before crystallization, the protein (10 mg/mL) wasincubated with 10 mM AMP-PNP and 2.5 mM MgCl2 at roomtemperature for 30 min. Crystals were grown at 20 °C using thevapor diffusion method by mixing protein and reservoir solutionat 1:1. Crystals were grown from a precipitant solution contain-ing 100 mM sodium citrate (pH 5.5), 45 mM NaCl, and 26%PEG 400. The crystals grew overnight and achieved maximumsize after 4 d. The crystals were directly frozen into liquid ni-trogen, and diffraction screening and data collection were per-formed at the Diamond Light Source synchrotron. These crystalsdiffracted X-rays to a maximum resolution of 3.2 Å. With ex-tensive additive and detergent screening, we found that 0.2%NG improved the crystal size and diffraction limits beyond 2.9 Å.

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 2 of 14

Data Collection.Diffraction data to 2.7 Å were collected on I02 atDiamond Light Source at a wavelength of 0.979 Å using a Pilatus6M detector and processed using the HKL suite of programs (9).Further processing was performed using the CCP4 suite (10).The resolution of the data and anisotropy analysis were evalu-ated by half-dataset correlation coefficient in Aimless (cutoff lessthan 0.5) (11). The space group was determined to be P212121with two copies of McjD in the asymmetric unit. The data col-lection statistics are summarized in Table S1.

Structure Solution and Refinement. The structure was solved bymolecular replacement in Phaser (12) using the outward openstructures of MsbA [Protein Data Bank (PDB) ID: 3B60] (13)and Sav1866 (PDB ID: 2ONJ) (14) as search models. Attemptsto molecular replace with the inward-facing MsbA (PDB ID:3B5X) did not produce a solution. Inspection of the electrondensity after molecular replacement allowed us to trace most ofthe protein except transmembrane (TM) helices 1 and 2 that didnot appear to fit in the electron density. The high resolution ofthe data allowed us to identify extra electron density not ac-counted for in the model, and TMs 1 and 2 were rebuilt (Fig.S6A). All model building was performed in Coot (15). Refine-ment was carried out in Buster (16) with NCS (17).Two AMP-PNP molecules and two MgCl2 ions were identified

in the nucleotide-binding domain (NBD) region (Fig. S6). Extraelectron density was observed in the cavity of the transporter,and the high resolution allowed us to identify it as NG from thecrystallization buffer (Fig. S6). Two NG molecules were builtand refined.

The final structure contains all residues from 9 to 579; theN-terminal region has very weak electron density to reliably buildthe first eight residues. The final model has an Rfactor of 24.76%and an Rfree of 26.60%. The refinement statistics are summa-rized in Table S1. The McjD structure has 94.4% of the residuesin the favored Ramachandran region and has four outliers asjudged by MolProbity version 3.19 (18). The outliers are Ser331and Pro335 from both chains, and they are in the linker regionbetween TM6 and the NBD. The electron density is too weak tomodel them correctly.

Positioning of MccJ25 Inside the McjD Cavity. The NMR structure ofMccJ25 (PDB ID: 1Q71) (19) was used for positioning it insidethe McjD cavity. The orientation selected for MccJ25 was basedon the minimum amount of steric clashes without manipulatingthe side chains of either MccJ25 or McjD (Fig. S7). In the cur-rent orientation, some of the MccJ25 chemistry is satisfied withhydrogen bonds with McjD. A couple of hydrophobic residuesare positioned toward the more hydrophobic pockets of thecavity. In the current orientation, Val6 and Tyr20 of MccJ25clash with Arg141 and Arg94, respectively. Rotation of themolecule by 90° would result in more clashes without manipu-lating either structure. Flipping the structure perpendicular tothe membrane does not seem to be a favorable orientation evenif the side chains of the cavity were manipulated.Based on the position of theMccJ25 inside theMcjD cavity, the

mutants were designed to probe if any of these residues are in-volved in the binding of MccJ25.All figures were generated using PyMol (20).

1. Datsenko KA, Wanner BL (2000) One-step inactivation of chromosomal genes inEscherichia coli K-12 using PCR products. Proc Natl Acad Sci USA 97(12):6640–6645.

2. Blond A, et al. (1999) The cyclic structure of microcin J25, a 21-residue peptideantibiotic from Escherichia coli. Eur J Biochem 259(3):747–755.

3. Drew D, Lerch M, Kunji E, Slotboom DJ, de Gier JW (2006) Optimization of membraneprotein overexpression and purification using GFP fusions. Nat Methods 3(4):303–313.

4. Lewinson O, Lee AT, Locher KP, Rees DC (2010) A distinct mechanism for the ABCtransporter BtuCD-BtuF revealed by the dynamics of complex formation. Nat StructMol Biol 17(3):332–338.

5. Venter H, Velamakanni S, Balakrishnan L, van Veen HW (2008) On the energy-dependence of Hoechst 33342 transport by the ABC transporter LmrA. BiochemPharmacol 75(4):866–874.

6. Hu NJ, Iwata S, Cameron AD, Drew D (2011) Crystal structure of a bacterialhomologue of the bile acid sodium symporter ASBT. Nature 478(7369):408–411.

7. Doshi R, et al. (2013) Molecular disruption of the power stroke in the ATP-bindingcassette transport protein MsbA. J Biol Chem 288(10):6801–6813.

8. Doshi R, van Veen HW (2013) Substrate binding stabilizes a pre-translocation intermediatein the ATP-binding cassette transport protein MsbA. J Biol Chem 288(30):21638–21647.

9. Otwinowski Z, Minor W (1997) Processing of X-ray diffraction data collected inoscillation mode. Methods Enzymol 276(20):307–326.

10. Collaborative Computational Project, Number 4 (1994) The CCP4 suite: Programs forprotein crystallography. Acta Crystallogr D Biol Crystallogr 50(Pt 5):760–763.

11. Evans P (2006) Scaling and assessment of data quality. Acta Crystallogr D BiolCrystallogr 62(Pt 1):72–82.

12. McCoy AJ, et al. (2007) Phaser crystallographic software. J Appl Cryst 40(Pt 4):658–674.

13. Ward A, Reyes CL, Yu J, Roth CB, Chang G (2007) Flexibility in the ABC transporterMsbA: Alternating access with a twist. Proc Natl Acad Sci USA 104(48):19005–19010.

14. Dawson RJ, Locher KP (2006) Structure of a bacterial multidrug ABC transporter.Nature 443(7108):180–185.

15. Emsley P, Cowtan K (2004) Coot: model-building tools for molecular graphics. ActaCrystallogr D Biol Crystallogr 60(Pt 12 Pt 1):2126–2132.

16. Blanc E, et al. (2004) Refinement of severely incomplete structures with maximumlikelihood in BUSTER-TNT. Acta Crystallogr D Biol Crystallogr 60(Pt 12 Pt 1):2210–2221.

17. Smart OS, et al. (2012) Exploiting structure similarity in refinement: Automated NCSand target-structure restraints in BUSTER. Acta Crystallogr D Biol Crystallogr 68(Pt 4):368–380.

18. Chen VB, et al. (2010) MolProbity: All-atom structure validation for macromolecularcrystallography. Acta Crystallogr D Biol Crystallogr 66(Pt 1):12–21.

19. Rosengren KJ, et al. (2003) Microcin J25 has a threaded sidechain-to-backbone ringstructure and not a head-to-tail cyclized backbone. J Am Chem Soc 125(41):12464–12474.

20. DeLano WL (2002) The PyMOL Molecular Graphics System, Version 1.5.0.4, Schrödinger,LLC (DeLano Scientific LLC, San Carlos, CA).

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 3 of 14

Fig. S1. (Continued)

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 4 of 14

(F)

CLD

M

(G)

McjD-L53C-MccJ25+ CuCl2

- ATP AMP-PNP

- Mg - Mg

McjD-L53C-MccJ25- CuCl2

- ATP AMP-PNP

- Mg - Mg

98-62-

49-

(H) (I)

98-

62-

49-

CLD

M

Apo ATP+

Mg

CLD

P +MMM

ggAM

PPNP

+ M

g

Fig. S1. Transport and cysteine cross-linking studies with McjD. (A) NMR structure of the lasso peptide MccJ25. (B and C) Ethidium transport in intact cellsexpressing wild-type (Wt) McjD or cys-less McjD (McjD-CL) at an external ethidium concentration of 2 μM (B), and growth of these cells in the presence of 30 μMMccJ25 (C) show that wild-type McjD and the cys-less mutant are transport-active compared with the inactive McjD E506Q mutant. The traces representobservations in three independent experiments (n = 3). (D) This conclusion is also supported by measurements of the basal ATPase activity of purified proteins.(E) Measurements of the basal ATPase activity of McjD in detergent solution as a function of ATP concentration. (F) Predictive cross-linking of the L53C mutantin the presence of MccJ25 in E. coli membranes. Western blot analysis of McjD-L53C cross-linked in ISOVs. CLD, formation of the cross-linking dimer; M,monomer in the absence of CuCl2 (Fig. 3). Preincubation of McjD with MccJ25 does not affect the formation of the CLD as observed in Fig. 3 in the absence ofMccJ25. (G) Relative intensity for each band was estimated by densitometry. A very small percentage of monomeric species is still present when the cross-linking reactions are terminated. Gray bars represent fraction of monomeric species relative to the total protein present. (H) Western-blot analysis of McjD-L53C cross-linked in ISOVs at pH 4.5. Reduced cross-linked dimer is observed in the presence of nucleotides when the cross-linking is performed at an acidic pH,suggesting that L53C is in an outward-open conformation. (I) The relative intensity for each band was estimated by densitometry. Red and gray bars representfraction of monomer and dimer species, respectively. A small fraction of nonspecific cross-linking is observed. Error bars are shown for all densitometrymeasurements (mean ± SEM; n = 3).

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 5 of 14

Fig. S2. Size exclusion profile of the (A) wild-type McjD and (B) E506Q mutant. Both proteins show similar monodisperse profiles. (C) Both wild-type andE506Q samples have similar purity as judged by Coomassie-stained SDS/PAGE gel. (D and E) Cavity mutants were prepared to similar purity.

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 6 of 14

Fig. S3. McjD proteins used for functional analyses in intact cells and proteoliposomes. (A) Analysis on Coomassie-stained SDS/PAGE gel. Lane 1: molecularmass markers (kDa). Lanes 2 and 3: total membrane protein in inside-out membrane vesicles from E. coli ΔacrA ΔacrB (DE3) expressing non-GFP-tagged wild-type McjD (30 μg/lane) (lane 2) or E506Q-McjD (30 μg/lane) (lane 3). Lanes 4 and 5: affinity purified wild-type McjD (1.5 μg/lane) (lane 4) and E506Q-McjD (1.5μg/lane) (lane 5). This purified protein was used for functional reconstitution in proteoliposomes. Lanes 6 and 7: proteoliposomes containing purified wild-typeMcjD (1.0 μg/lane) (lane 6) or E506Q-McjD (0.6 μg/lane) (lane 7). Lane 8: empty liposomes (equivalent to the amount of proteoliposomes containing McjDproteins). (B) Western blot analysis of the expression levels of McjD mutants in the plasma membrane of E. coli ΔacrA ΔacrB (DE3) used in ethidium transportassays. Total membrane protein in inside-out membrane vesicles (20 μg/lane) containing wild-type McjD (lane 1), E309Q-McjD (lane 2), E506Q-McjD (lane 3), orE309A-McjD (lane 4). The blot was probed with anti-his-tag antibody.

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 7 of 14

Fig. S4. (A) Analysis by microscale thermophoresis of MccJ25 binding to McjD with increasing amounts of MccJ25. A plateau has not been reached becauseMccJ25 becomes insoluble at higher concentrations. (B) Ligand-induced ATPase activity of McjD cavity mutants. The results are shown as percentage activityrelative to the wild-type protein. All mutants display similar basal activity as the wild-type protein. None of the mutants display induced ATPase activity in thepresence of MccJ25, confirming their possible role in MccJ25 binding. Error bars are shown for all measurements (mean ± SEM; n = 3). (C and D) Analysis bymicroscale thermophoresis of the interaction of purified McjD mutants with MccJ25. The MccJ25 binding was plotted using the total binding equation inGraphPad Prism. The mutants have not reached complete saturation at the same MccJ25 concentration range as for the wild-type McjD. (C) The KD for theMcjD F86A mutant was estimated to be >2 mM and (D) for the N302A >1 mM (mean ± SEM; n = 3). The point mutations have significantly lowered the affinityof McjD for MccJ25 compared with the wild-type protein. These data are also in agreement with the ligand-induced ATPase activity measurements with themutants that did not display any increased activity.

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 8 of 14

Fig. S5. Inhibition of McjD ATPase activity in the presence of increasing amounts of (A) AMP-PNP and (B) ortho-vanadate. The ATP concentration in the assayswas 1 mM. The IC50 was determined to be 39.5 ± 0.6 μM and 61.6 ± 1.1 μM for AMP-PNP and vanadate, respectively. Error bars are shown for all measurements(mean ± SEM; n = 3); when error bars are not visible, they are contained within the symbols.

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 9 of 14

Fig. S6. (Continued)

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 10 of 14

Fig. S6. Electron density maps. (A) Omit map around TMs 1 and 2 (red mesh) calculated at 3.5 Å after molecular replacement with Sav1866. The map iscontoured at 1 σ. The two McjD TMs are only shown for clarity and were omitted from map calculation. (B and C) Electron density for the NG molecules. Themaps were generated with a B-factor scaling of −50 A2. NG is shown as red sticks. (B) 2Fo-Fc electron density map (blue mesh) contoured at 0.5 σ inside therefined McjD cavity, but in the absence of NG in the refinement. The NG molecule is only shown as reference. The electron density of the surrounding residuesis also shown. (C) 2Fo-Fc electron density map contoured at 0.9 σ after NG was included in the refinement. (D and E) Electron density along residues 25–38 of TM1 of the final refined McjD structure at 2.7 Å. (D) 2Fo-Fc map before map sharpening contoured at 1.0 σ. (E) 2Fo-Fc map after map sharpening contoured at 1.0 σ.The map was generated with a B-factor scaling of −50 A2. (F and G) Electron density along the NBD region of the final refined McjD structure at 2.7 Å. TheAMP-PNP is shown in stick format (dark red). (F) 2Fo-Fc map before map sharpening contoured at 1.0 sigma. (G) 2Fo-Fc map after map sharpening contoured at1.0 σ. The map was generated with a B-factor scaling of −50 A2. (H) Crystal lattice of McjD in space group P212121. The asymmetric unit contains one McjDdimer, shown in red ribbon.

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 11 of 14

(A)

(B)

Fig. S7. Position of MccJ25 inside the McjD cavity. (A) The MccJ25 volume (orange) can fit inside the McjD cavity (dark gray) without further manipulations.The McjD TM region is shown as cartoon (light gray). (B) The potential side chains from the TMs that might be involved in the binding of MccJ25 (shown as graysurface) are shown as sticks. The mutants (shown in bold red) were selected and designed according to this model. The ATPase activity for all of the mutantswas assessed in the presence of MccJ25.

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 12 of 14

(A) (B)

(C)

Fig. S8. Effect of NG on McjD activity. (A) Comparison of the basal ATPase activity of dodecyl-maltopyranoside-purified McjD in presence or absence of NG.The uncleaved GFP construct used for the microscale thermophoresis experiments shows the same activity as the cleaved protein. NG was used as a secondarydetergent during the crystallization to improve crystal diffraction (also found inside the cavity), but it does not interfere with the basal activity. Activities wereplotted as percentage of wild-type basal ATPase activity in the presence of 1 mM ATP (35 nmol min−1 mg protein−1). (B) Increasing concentrations of NG did notsignificantly stimulate or inhibit the ATPase activity of McjD in proteoliposomes. The NG concentration was increased up to 12 mM, which is 67% of theconcentration that completely dissolved the proteoliposomes. (C) Expression of wild-type McjD (solid lines) did not confer resistance on cells to NG comparedwith cells with equal expression of the inactive McjD-E506Q mutant (interrupted lines). Growth experiments were performed in a 96-well plate. Under similarconditions, McjD mediated cellular resistance to MccJ25 peptide (Fig. S1). Error bars in A and B refer to mean ± SEM and are based on data obtained with threeindependent protein preparations or batches of proteoliposomes. Growth curves in C are typical for observations in three separate experiments.

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 13 of 14

Table S1. Data collection and refinement statistics

Data collection statisticsBeamline Diamond Light Source I02Spacegroup P212121Resolution (Å) 48.96–2.70 (2.78–2.70)Cell dimensions a = 80.7, b = 107.8, c = 232.9No. of reflections 208,640 (17,507)No. of unique reflections 56,627 (4,554)Completeness (%) 99.8 (99.6)Redundancy 3.7 (3.8)Rmerge (%) 5.1 (133)Mean [I/σ(I)] 12.1 (1.5)Mn(I) half-set correlation CC(1/2) (%) 0.99 (0.77)

Refinement statisticsRfactor (%) 24.76 (35.39)Rfree (%) 26.60 (37.53)Protein atoms 9,109AMP-PNP 62Mg2+ 2Nonyl-glucopyranoside 42B-factors

Protein 100AMP-PNP 95Nonyl-glucopyranoside 110Ion 69

rmsd from ideal valuesBonds (Å) 0.01Angle (°) 1.05Ramachadran plot outliers (%) 0.35

Values in parentheses refer to data in the highest-resolution shell.

Choudhury et al. www.pnas.org/cgi/content/short/1320506111 14 of 14