The structure of the efflux pump AcrB in complex with bile acid

Abstract

Gastrointestinal bacteria, like Escherichia coli, must remove bile acid to survive in the gut. Bile acid removal in E. coli is thought to be mediated primarily by the multidrug efflux pump, AcrB. Here, we present the structure of E. coli AcrB in complex with deoxycholate at 3.85 Å resolution. All evidence suggests that bile acid is transported out of the cell via the periplasmic vestibule of the AcrAB-TolC complex.

• Multidrug efflux protein
• AcrB
• bile acid
• E. coli
• structure

Abbreviations
DDM	=	dodecyl-B-D-maltopyranoside
DOC	=	deoxycholate, E. coli, Escherichia coli
GFP	=	Green Fluorescent Protein
MIC	=	minimum inhibitory concentration
Ni-NTA	=	nickel-nitrilotriacetic acid
RND	=	resistance nodulator cell division
TEV	=	tobacco etch virus protease
IMAC	=	Immobilized Metal Affinity Chromatography

Introduction

Gastrointestinal bacteria such as Escherichia coli must remove bile acid to survive in the gut [1–3]. In E. coli the resistance nodulator cell division (RND) transporter AcrB works in concert with the outer membrane protein TolC, and a membrane fusion partner AcrA, to export various antibiotics in a proton-driven process [4]. The AcrAB-TolC complex also removes bile acid [2], [3], [5], a detergent like amphipathic molecule that makes up ∼ 1% of the fluid within the intestine. Bile acid significantly induces expression of the acrAB operon by acting on specific transcriptional activators [6], [7], and triggers expression of virulence genes to alter pathogenic behaviour [8]. Deletion of the acrAB operon leaves cells highly sensitive to bile acid, with a 10-fold decrease in the minimum inhibitory concentration (MIC) to deoxycholate [2]. The bile acid deoxycholate was found to be a direct transport competitor of the substrate fluorophospholipid in an AcrAB reconstituted system [5]. The high-resolution structure of AcrB was first solved as a symmetrical homotrimer in 2002 [9], and later as a physiological asymmetrical trimer [10], [11], [12]. Each protomer is thought to represent an alternate single state of a three-step transport cycle designated loose (L), tight (T), and open (O) [13]; (i) in the L state the periplasmic vestibule is open for substrate entry (ii) in the T state the vestibule tunnel is extended to include a hydrophobic binding pocket and (iii) in the O state, the periplasmic vestibule closes, and the central pore opens for substrate to exit. Most likely the sequential states occur by a rotating mechanism driven by proton translocation across the membrane-embedded portion of AcrB [10–12].

Here, to confirm that AcrB is unequivocally required for bile acid efflux in E. coli we first constructed a single acrB deletion mutant and monitored the effect of bile acid to growing cells. To elucidate the export pathway for bile acid, we solved the structure of AcrB in complex with deoxycholate.

Methods and materials

Construction of AcrB deletion strain

The DacrB::Km^R mutation from the Keio collection was moved into strain BL21(DE3) by bacteriophage P1-mediated transduction [14]. After P1 transduction the BL21(DE3)DacrB::Km^R mutant was checked by PCR as described previously [14].

Bile acid growth assay

E. coli strains BL21(DE3) and BL21(DE3)▵ acrB::Km^R were cultured at 37°C in 50 ml of Luria-Broth (LB) containing kanamycin 50 mg/ml in a 250 ml shaker flask set at 220 r.p.m. Densities of cells were monitored at 600 nm at regular time intervals. At OD₆₀₀=0.7 – 0.8, a final concentration of 5 mM deoxycholate (Sigma, UK) was added and optical densities were continuously measured.

Purification of AcrB

AcrB forms a natural hexahistidine tag (i.e. AcrB forms a trimer and has 2 histidines at its very C-terminus), which binds to Ni-NTA resin in addition to targeted membrane protein-GFP-His⁸ fusions in our GFP-based pipeline [15], in which targets are typically purified from BL21(DE3) derived membranes with buffer containing 20 mM Tris/HCl pH 7.5, 150 mM NaCl, and 1% dodecyl-β-D-maltopyranoside (DDM, Anatrace, USA). To isolate AcrB we collected protein that bound to the second Ni-NTA step in our standard purification pipeline in buffer containing 250 mM imidazole, 20 mM Tris/HCl pH 7.5, 150 mM NaCl, and 0.03% dodecyl-β-D-maltopyranoside. To separate AcrB from other contaminants and tags, AcrB was loaded onto a Superdex 200 10/300 (GE healthcare, UK) gel filtration column in 20 mM Tris/HCl pH 7.5, 0.1 M NaCl, and 0.03% DDM at a flow-rate of 0.4 ml/min. The protein peak corresponding to AcrB was collected and concentrated to 8 mg/ml.

Structure determination of AcrB in complex to bile acid

Native AcrB crystals were grown from mixing equal volumes of a protein solution containing 8 mg/ml of AcrB in 20 mM Tris/HCl pH 7.5, 150 mM NaCl, 0.03% DDM and 0.01% DM and reservoir solution containing 10% PEG 4000, 0.1 M HEPES pH 7.5, 0.1 M ammonium sulphate and 22% v/v glycerol (MemGold, Molecular dimensions, UK) ^[16] . Crystals appeared within two days and grew to final size within a week. Crystals were soaked with 50 mM sodium deoxycholate (Sigma, UK) for one month prior to data collection. The crystals were flash frozen in liquid nitrogen prior to data collection. A full data set was collected to 3.85 Å resolution at beamline I04 at Diamond Light Source ID14-4 at ESRF (Table I) at 100K. The data were integrated with XDS [17], scaled with SCALA [18] and phased by molecular replacement using Phaser [19] from the CCP4 suite of crystallography programs. PDB file 1 1IWG (native AcrB) [9] was used for molecular replacement. The structure was refined using REFMAC [20]. Model building and electron density inspection was carried out in COOT [21]. Figures were prepared with PYMOL [22] and CCP4mg [23]. Table I summarizes the data collection and refinement statistics for the AcrB-deoxycholate complex. Coordinates and structure factors have been deposited in the RCSB Protein Data Bank with PDB ID code 2W1B.

Table I.  Data collection and refinement statistics.

Data collection	Deoxycholate	Native
Space group	H32
Beamline	Diamond I04	ESRF ID 14 4
Cell dimensions
a, b, c (Å)	145.49, 145.49, 515.18	146.29, 146.29, 513.7
α, β, γ (°)		90.0, 90.0, 120.0
Resolution (Å)	3.85 (3.85)	3.2 (3.2)
Rsym or Rmerge^#	23.4 (56.4)	7.6 (52.2)
I/σ	2.5 (1.3)	8.3 (1.4)
Completeness (%)	96.4 (96.2)	98.8 (100)
Redundancy	4.1 (4.2)	3.5 (3.5)
Refinement
Resolution (Å)	3.85	3.2
No. reflections	18558	32895
Rwork/Rfree *	27.7/34.9	30.98/36.5
No. atoms
Protein	7841
Ligand	28
Average B-factors (A^2)	75	100
R.m.s. deviations
Bond lengths (Å)	0.012	0.01
Bond angles (°)	1.39	1.3

Values in parentheses are for the high-resolution shells; ^#R_merge=Σ_hkl Σ_I ∣I_i(hkl)- < I(hkl)>∣/Σ_hkl Σ_I I_i(hkl); ^*R_factor and R_free =Σ_hkl [∣F_o(hkl)∣-∣F_c(hkl)∣]/ Σ_hkl ∣F_o(hkl)∣; R_free was calculated using 5% of the data.

Results and discussion

Single acrB E. coli deletion strain is hypersensitive to bile acid

A single acrB knockout E. coli strain was constructed to complement previous growth experiments measured in an acrAB operon deletion mutant strain [2].

Cells were grown to mid-log phase and challenged with the bile acid deoxycholate. As shown in Figure 1, cells continue to grow in the presence of 5 mM bile acid in a wild-type E. coli strain (thin dashed line), but growth is severely impaired in the acrB deletion strain (thick-dashed line). Within two hours the cell density is declined to almost background levels, OD₆₀₀=0.08. As expected, there was some recovery of cell growth in the acrB deletion strain, reaching an OD₆₀₀ of 0.4 after 12 h. Putative RND transporter genes mdtABC, are noted to form some limited resistance against bile acid in an acrAB deletion strain [24], [25]. Nevertheless, recovery of cell growth was minimal as monitored for 24 h, and cells failed to grow in freshly inoculated media (results not shown).

Figure 1.  Comparison of growth of an E. coli wild-type strain with an acrB deletion strain in the presence of deoxycholate (DOC). thin line, BL21(DE3) in absence of bile acid; thin dashed line, BL21(DE3) with 5 mM addition of deoxycholate; thick line, BL21(DE3)ΔacrB::Km^R in absence of bile acid, thick dashed-line, BL21(DE3)ΔacrB::Km^R with 5 mM addition of deoxycholate.

To understand the transport mechanism for bile acid, we purified and crystallized native AcrB in the presence of deoxycholate.

Isolation of AcrB

AcrB forms a natural hexahistidne tag due to the fact the last two residues are histidines and it oligomerizes into a trimer. AcrB is difficult to remove from IMAC purifications of your target protein, and is considered a problem for structural biologists as it crystallizes at only 5% of total purified membrane protein [26]. One approach to avoid this problem is to cleave off the His-tag with a His-tagged site-specific protease and then to re-bind cleaved material. This is our standard approach for membrane protein purification in our laboratory as we work with a GFP-His₈ fusion tag that is cleaved and removed prior to crystallization [15]. Thus, we took advantage of this and recovered AcrB by eluting re-bound material from a second IMAC step, i.e. contaminants, GFP-His₈ and His₆-TEV protease (Figure 2). AcrB was further purified by size exclusion chromatography and concentrated to 8 mg/ml for crystallization.

Figure 2.  Fractions of a standard membrane protein-GFP-His₈ purification were loaded onto a 12% SDS-polyacrylamide gel and then stained with Coomassie to illustrate isolation of native AcrB. (1) MP-GFP-His₈ elution from first IMAC; (2) overnight His₆-TEV protease digested material; (3) flow-through of second IMAC step; and (4) elution of bound material from second IMAC (bands corresponding to AcrB, MP-GFP-His₈, His₆-TEVp, and MP are as indicted on gel).

Biding site of deoxycholate

AcrB is a trimer with an overall ‘jellyfish’ shape’ composed of a central transport proton-dependant pump cavity, vestibule, and pore funnel [9] (Figure 3). After molecular replacement using the native AcrB structure and some model building of the overall structure, 2F_o-F_c and F_o-F_c electron density maps were calculated. Although native AcrB without ligand diffracts to 3.2 Å resolution in this preparation, bile acid complex crystals diffract only to 3.85 Å. This is typical for AcrB complexes with almost all obtained at lower than 3.5 Å resolution [27], [28].

Figure 3.  Overall structure of AcrB trimer. The bound deoxycholate in the periplasmic vestibule is shown in yellow stick. Boxed area is a close-up view of the vestibule site with the deoxycholate and side chains that are involved in the binding shown in grey stick.

Inspection of the periplasmic binding cleft reveals positive density for bile acid, Figure 4b. Binding of deoxycholate to the periplasmic cavity there were no observed conformational changes, as the structure can be superimposed with the other published substrate-bound structures with an rmsd of 1.06 Å over 960 C_a atoms. The binding site for deoxycholate is in a similar position to the published complexes [27] rhodamine 6G [27], ethidium bromide [27], Phe-Arg-β-naphthylamide [27], nafcillin [27], and ciprofloxacin [27] despite their molecular diversity (Figure 5). The hydroxyl group of C-13 of the steroid ring of deoxycholate forms hydrogen bonds to Ser⁷¹⁵, 2.51 Å, and carbon atoms C-17, C-20, C-21 and C-22 pack by van der Waals interactions to Phe⁶⁶⁴ (Figure 4c). mutational analysis of these vestibule ligand binding residues showed that the transport activity of structurally different substrates was compromised in most cases [27]. The only mutation to affect (almost) all substrates was the Phe⁶⁶⁴ to Ala mutant. Phe⁶⁶⁴ is clearly an important hydrophobic residue for substrate binding, and likewise for deoxycholate, coordinates most interactions. The C-13 OH-coordination with Ser⁷¹⁵ is novel, as previous complex structures have no hydrogen bonding to their substrate. Although, the Ser⁷¹⁵Ala mutant was noted to have no effect on broad activity, Ser⁷¹⁵ could be a substrate-specific residue for deoxycholate. Indeed, this is the first AcrB structure with a naturally occurring substrate.

Figure 4.  Binding of deoxycholate in the periplasmic vestibule of AcrB. (a) Electron density for native AcrB (purple), (b) deoxycholate bound prior to adding the ligand (light blue), (c) deoxycholate density after including it in the refinement (dark blue). The 2F_o-F_c maps are contoured at 1.0σ and deoxycholate (pink, contoured at 1.0σ) are shown for clarity, calculated after some model building (prior to deoxycholate being built). Some of the vestibule amino acid side chains are shown. Deoxycholate (yellow sticks) is included for reference. Nitrogen atoms are shown in blue and oxygen in red.

Figure 5.  Alignment of deoxycholate (yellow) with ethidium (light orange), Phe-Arg-β-naphthylamide (light blue), and ciprofloxacin (pink). The residues involved in the binding of deoxycholate are shown as grey sticks. Arg⁷¹⁷ is only shown as a reference since there is no direct contact with the deoxycholate.

In contrast to foreign compounds rhodamine 6G, ethidium, dequalinium, and ciprofloxacin, which bind to the periplasmic vestibule and the top of the lipid-interface at the central cavity [28], deoxycholate binds to the vestibule only. We suggest that the later interaction site is the biologically relevant interaction site, because it is in vectorial alignment to the binding site in the extended vestibule conformation (T).

The final electron density maps show positive density between the carboxyl group of C-23 and Leu⁸²⁸. A water molecule or sodium ion could be modelled and coordinated in the density. We prefer the latter as it would provide a counter-ion to deoxycholate; however, the resolution is insufficient to be certain. Interestingly, there is no density for YajC as observed in a recent native AcrB structure [29], perhaps due to the differences in culture conditions and/or strain background used.

Transport state of deoxycholate-bound AcrB

The AcrB-deoxycholate complex was crystallized in the rhombohedral space-group, H32, all molecules are symmetrical, with the three-fold crystallographic axis running down the centre of the trimer. Under these conditions, only the L state of AcrB is observed. This state is either a crystal artefact, in which the loose conformation is preferentially crystallized, or a true physiological state in which all three vestibules are open [13]. Detergents are also substrates for AcrB [13], but in our native AcrB structure there is no apparent density that could indicate the binding of detergent molecules in these regions (Figure 4a).

Concluding remarks

This is the first AcrB structure with a substrate that E. coli has had to evolve to continuously remove for its survival. It is apparent that deoxycholate is mainly exported by AcrB, and gains entry from the periplasmic space. Entry from the periplasm makes sense, slow passive diffusion into the cytoplasm would lead to high levels of bile acid, ∼ 1 mM is reported to be enough to compromise membrane integrity [30]. This entry site is an attractive area for rational drug design. Indeed, the pathogenic bacterium Campylobacter jejuni fails to colonize the chicken gastrointestinal tract when it lacks a component of the homologous AcrAB efflux system [1].

Acknowledgements

We would like to thank Dr Alexander Cameron for useful discussions. D.D was a recipient of an EMBO long-term fellowship. K.B is a RCUK Fellow. Funded by the Biotechnology and Biological Science Research Council (BBSRC) Membrane Protein Structure Initiative (to S.I.), the Swedish Research Council and the Center of Biomembrane Research through the Swedish Foundation of Strategic Research (to J.W.-d. G.). Data were collected at beam-line I04 at Diamond Light Source and ID14-4 at ESRF (France). We would also like to acknowledge the Membrane Protein Lab. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.