1. Structural Biology and Molecular Biophysics

An allosteric transport mechanism for the AcrAB-TolC multidrug efflux pump

  1. Zhao Wang
  2. Guizhen Fan
  3. Corey F Hryc
  4. James N Blaza
  5. Irina I Serysheva
  6. Michael F Schmid
  7. Wah Chiu  Is a corresponding author
  8. Ben F Luisi  Is a corresponding author
  9. Dijun Du  Is a corresponding author
  1. Baylor College of Medicine, United States
  2. The University of Texas Health Science Center at Houston Medical School, United States
  3. MRC Mitochondrial Biology Unit, Cambridge Biomedical Campus, United Kingdom
  4. University of Cambridge, United Kingdom
https://doi.org/10.7554/eLife.24905
Share this article
Cite this article
  1. Zhao Wang
  2. Guizhen Fan
  3. Corey F Hryc
  4. James N Blaza
  5. Irina I Serysheva
  6. Michael F Schmid
  7. Wah Chiu
  8. Ben F Luisi
  9. Dijun Du
(2017)
An allosteric transport mechanism for the AcrAB-TolC multidrug efflux pump
eLife 6:e24905.
https://doi.org/10.7554/eLife.24905
  2. Download BibTeX
  3. Download .RIS

Abstract

Bacterial efflux pumps confer multidrug resistance by transporting diverse antibiotics from the cell. In Gram-negative bacteria, some of these pumps form multi-protein assemblies that span the cell envelope. Here, we report the near-atomic resolution cryoEM structures of the Escherichia coli AcrAB-TolC multidrug efflux pump in resting and drug transport states, revealing a quaternary structural switch that allosterically couples and synchronizes initial ligand binding with channel opening. Within the transport-activated state, the channel remains open even though the pump cycles through three distinct conformations. Collectively, our data provide a dynamic mechanism for the assembly and operation of the AcrAB-TolC pump.

https://doi.org/10.7554/eLife.24905.001

Introduction

Antibiotic resistance of pathogenic bacteria is a growing clinical problem, exacerbated by insufficient development of new antibiotics. Drug efflux pumps play important roles in intrinsic or acquired drug resistance to a wide variety of currently available antimicrobial agents (Pu et al., 2016). In Gram-negative bacteria, some of these pumps form multi-protein assemblies that span the cell envelope (Du et al., 2015a). The components of these assemblies include an outer membrane protein, a plasma membrane-spanning protein and a periplasmic protein that connects the two trans-membrane components (Du et al., 2015b). AcrAB-TolC is a RND-based tripartite efflux pump, comprised of the outer membrane protein TolC, the periplasmic membrane fusion protein AcrA, and the inner membrane transporter AcrB, which cycles through three different conformational states during the drug transport process: access (L), binding (T) and extrusion (O) (Du et al., 2015a). A small peptide, AcrZ, has been identified that modifies the activity of AcrB (Hobbs et al., 2012). Previous electron microscopy studies of the AcrAB-TolC pump have revealed the overall shape of the pump and the relative arrangement of its components (Daury et al., 2016; Du et al., 2014; Jeong et al., 2016). Due to the limited resolution, these studies were unable to identify the detailed interaction interfaces between the components. Moreover, the mechanism of pump assembly and channel opening is presently unclear (Müller and Pos, 2015; Zgurskaya et al., 2015) and requires detailed structural information of the pump in the different conformational states that accompany drug translocation. In this study, we report the near-atomic resolution cryoEM structures of the AcrAB-TolC pump captured in both resting and drug transport states. The data reveal the detailed tertiary and quaternary structural changes associated with the transition from apo to ligand-bound states. The quality of the cryoEM maps enable unambiguous modeling of all components and identification of the conformational asymmetry in AcrB in the presence of antibiotic that underpins the molecular mechanism of drug translocation. Our study provides a structural framework for understanding the mechanism underlying drug efflux.

Results

CryoEM structure determination

Preparation of the intact AcrAB-TolC for structural studies has been a challenge. Previously, the fully-assembled pump was stabilized by tandem fusion and cross-linking of its components using the ‘GraFix’ method (Du et al., 2014). The fusion was co-expressed with AcrZ, a hydrophobic helical peptide that affects AcrB efflux activity for a subset of antibiotics (Hobbs et al., 2012) and interacts with AcrB in the transmembrane portion (Du et al., 2014). To pursue a higher resolution structure, we improved the purification procedure. The ratio of detergent to total membrane protein was optimized to extract AcrAB-TolC from the cellular membrane, and the detergent was exchanged with amphipol A8-35 (Baker et al., 2015; Liao et al., 2013; Popot, 2010; Tribet et al., 1996; Zoonens and Popot, 2014). Ligands were used in an effort to favor conformational uniformity and/or to capture structures corresponding to different ligand transport states. The purified pump was imaged in the presence of pyranopyridine inhibitor MBX3132 (Sjuts et al., 2016) (Figure 1—figure supplement 1A and B; Supplementary file 1) or substrate puromycin, as described in Materials and methods.

We also have developed a new procedure to stabilize the full pump using disulphide-linkages that were chosen from the proximity of residues S273 in the β-barrel domain of AcrA and S258 in the β-hairpin motif of the DN subdomain of AcrB seen in the cryo-EM structure of the full pump. We introduced single cysteine-substitutions in the individual AcrA and AcrB components (AcrA-S273C and AcrB-S258C) without flexible engineered linkers. Consistent with the model based on the fusion constructs, we observed disulfide bond formation between the two free components. Moreover, the disulfide bond-stabilized AcrAB can recruit TolC (and three additional AcrAs, which are not expected to be cysteine-crosslinked) to form a stable tripartite complex in vivo (Figure 1—figure supplement 2A). This complex was prepared without co-expression of AcrZ. The full pump assembly, stabilized through the disulfide-bond linked AcrAB, was imaged in the absence of ligand (Figure 1—figure supplement 2, B and C).

The cryo-EM structures of the apo pump and the pump/ligand complexes reveal different conformational states of the full pump. These structures have been solved at different subnanometer and near atomic resolutions (Supplementary file 1). The maps of the full pump at 3.6 Å resolution can be readily segmented into a TolC trimer, a periplasmic AcrA hexamer, and an AcrB trimer (Video 1). The corresponding domains and subdomains of the four protein components derived from the cryoEM maps are delineated in Figure 1—figure supplement 3. We will describe each structure further below.

Video 1
Download asset
Structure of AcrABZ-TolC multidrug efflux pump with inhibitor MBX3132 bound.

Individual protein subunits of the density map are highlighted and their corresponding models are then shown.

https://doi.org/10.7554/eLife.24905.002

The apo structure of the pump adopts a closed channel

The reconstruction of the apo pump at 6.5 Å resolution (Supplementary file 1; Figure 1—figure supplement 4) revealed TolC to be in the closed conformational state (Figure 1), like that seen in the crystal structure of the isolated protein (Koronakis et al., 2000). The periplasmic component of the pump, AcrA, forms a hexamer through the trimerization of two conformationally distinct protomers (Figure 1A). We refer to the two distinct subunits as ‘protomer-I’ and ‘protomer-II’. In both AcrA protomers, a short helix-turn-helix (HTH) motif interacts in a tip-to-tip manner with the HTH motif of the TolC coiled-coils at their periplasmic ends (Figure 1A, right panel). Because each subunit of TolC has an internal structural repeat, the trimer has 6 HTH motifs, which interact with the 6 HTH motifs of the AcrA hexamer in a quasi-equivalent manner: AcrA protomer-I and the adjacent protomer-II interdigitate with the intra- and inter-protomer grooves of TolC, respectively. The two distinct conformations of the HTH motif loops of AcrA protomers I and II structurally conform to the slightly different contact surfaces of TolC (Figure 1A, right panel). This interaction is consistent with functional (Lee et al., 2012; Song et al., 2014; Xu et al., 2010) and structural data (Daury et al., 2016; Kim et al., 2015), but differs in detail from that proposed for the open state of the pump reported earlier (Jeong et al., 2016), as we will describe further below. A gap is present at the interfaces between adjacent AcrA dimer pairs, so that the AcrA helical hairpin, lipoyl and β-barrel domains do not pack tightly to seal the channel from the periplasm (Figure 1—figure supplement 5). This gap must be closed in the transport process to prevent leakage of substrates into the periplasm. AcrB adopts a symmetric state (LLL) in the pump assembly, as seen in the crystal structure of apo AcrB (Murakami et al., 2002).

Figure 1 with 5 supplements see all
Download asset Open asset
The structure of disulfide-bond stabilized AcrAB-TolC pump with closed channel at 6.5 Å resolution.

(A) CryoEM map of the pump with closed-state TolC, visualized in side view along the membrane plane (left). The four components within one asymmetric unit of the C3 symmetric assembly are color-coded: TolC (red), AcrA (orange and green) and AcrB (blue). (middle) A sliced view of the pump shows the closed channel. (right) Ribbon representation of the pump with closed-state TolC with the same color code as in left panel. The black arrow indicates the closed site in TolC. (B) Comparison of reference- free 2D averages (left) and slice view of 3-D maps at the arrow (right) of the apo pump with closed TolC (top), pump in presence of puromycin (middle) and pump with MBX3132 (bottom).

https://doi.org/10.7554/eLife.24905.003

The transport state of the full pump with opened channel

The crystallographic studies of AcrB in complex with natural substrates reveal asymmetric structures, which suggest consecutive steps of a transport cycle (Murakami et al., 2006; Seeger et al., 2006). We explored how the full pump accommodates AcrB in this functional state in the presence of puromycin, which is a validated transport substrate (Hobbs et al., 2012).

We targeted the AcrB protomers using a focused classification procedure (Bai et al., 2015) (See Materials and methods) and could identify a subset of the full pump assembly (30.2% of the entire dataset) containing exactly one protomer whose density matches the extrusion state (O) of AcrB, as shown in Figure 3—figure supplement 1. A 5.9 Å density map of the full pump assembly without any imposed symmetry was reconstructed from this particle subset (Figure 2A; Figure 2—figure supplement 2). The correlation analysis of the map region corresponding to the AcrBZ components shows that the three AcrB protomers in L, O and T states are in a fixed spatial relationship with respect to each other in this full pump assembly (Figure 2B). This spatial disposition fits very well to the crystal structures of numerous other drug-bound AcrB structures (Eicher et al., 2012; Murakami et al., 2006; Nakashima et al., 2011). Though the AcrB subunit densities match well with those in the crystal structure, no puromycin density is detected in the cryoEM map. A similar situation occurs for some crystal structures of the asymmetric state, where drug is present in the buffer but discernable electron density for the drug is absent from the ligand binding pockets (Seeger et al., 2006). The crystal structure of AcrBZ in the presence of puromycin at 3.2 Å (Figure 2—figure supplement 3; Supplementary file 3) shows that the AcrB subunits adopt three distinct conformational states (Figure 2—figure supplement 3B) as seen in our cryoEM structure in the presence of this drug. Poorly defined density for puromycin is found in the binding pocket of the T protomer (Figure 2—figure supplement 3C). In the case of the cryoEM map, the lack of density can be rationalized as either a resolution limit in the C1 map or the drug being bound only transiently and flexibly.

Figure 2 with 3 supplements see all
Download asset Open asset
5.9 Å resolution cryoEM asymmetric structure of the AcrABZ-TolC pump in the presence of puromycin.

(A) Structure of asymmetric AcrABZ-TolC pump visualized in side view; subunits are color-coded accordingly: TolC (red and pink), AcrA (orange and green), AcrB (blue) and AcrZ (purple). (B) Model variation of the C-alpha trace between the three subunits of AcrB within the full pump colored in gradient from blue to red with increasing variation.

https://doi.org/10.7554/eLife.24905.009

TolC adopts a fully opened state via tip-to-tip interactions with AcrA in the pump assembly and AcrAs pack tightly to form a sealed channel, which is similar to that seen in the complex of the pump with the potent inhibitor MBX3132, to which we now turn.

The structure of pump in complex with inhibitor MBX3132 reveals the interaction interfaces between pump components at atomic resolution

The inhibitor MBX3132 likely binds AcrB tightly, since it is active in the nanomolar concentration range (Sjuts et al., 2016), and the compound was used to lock the pump in a more homogeneous conformation. The 3D map for the pump/MBX3132 complex was initially generated without any imposed symmetry and exhibited an apparent 3-fold symmetric pattern (Figure 1—figure supplement 1C). Thus, refinements were performed with 3-fold symmetry, which produced the final 3D reconstruction at ~3.6 Å resolution. This dataset was further subjected to focused 3D classification (Bai et al., 2015) of a targeted region of AcrB to reveal its structural variations in response to the presence of the inhibitor. We found that 73% of the particles within the dataset have their three AcrB subunits in the same (TTT) conformation (Materials and method and Figure 3—figure supplement 1A–E). Subsets of particles show AcrB conformations in LLL (1.4%), LLT (6.2%) and LTT (18.6%) states, respectively, while none of the subunits was identified in an ‘O’ state. Therefore, this inhibitor-bound pump conformation primarily represents a T-saturated state of the AcrB. The C3 symmetry imposed reconstruction using the subset of particles with TTT state (24,597 particles) also achieved 3.6 Å resolution (Figure 3A, Figure 3—figure supplement 1F and G). We observed clear density for the inhibitor in a 'hydrophobic trap' site adjacent to the drug binding of the AcrB trimer in both the asymmetric and symmetry imposed maps (Figure 3B). Side chains can be readily assigned and modeled in the protruding densities along the peptide backbone (Figure 3—figure supplement 2). This cryo-EM structure of pump/MBX3132 enables us to visualize the interaction interfaces between the pump components at atomic resolution.

Figure 3 with 2 supplements see all
Download asset Open asset
3.6 Å resolution cryoEM structure of the AcrABZ-TolC pump with inhibitor MBX3132 bound AcrB in the TTT state.

(A) Three-fold symmetry imposed density map (left) and model (right) of the AcrABZ-TolC complex. (middle) A sliced view of pump. The five protein components within one asymmetry unit of the pump are color-coded. TolC (red), AcrA (orange and green), AcrB (blue) and AcrZ (purple). (B) Density of the MBX3132 seen in the hydrophobic trap of each AcrB subunit. The location of the trap is indicated by the star in the model in the right panel of (A).

https://doi.org/10.7554/eLife.24905.013

Comparing the structure of TolC in the crystal or in the apo pump with that in the inhibitor-bound pump assembly reveals that TolC adopts a fully opened state through an iris-like dilation of the periplasmic end. This dilation is presumably necessary to maintain TolC-AcrA connectivity and to permit drug molecules to pass through the pump (Koronakis et al., 2000). Little conformational change is shown in the TolC TM region and the adjacent α-helical portion. However, the structures start to deviate at the equatorial domain, where the superhelical trajectories of coiled-coil helices change with the iris-like dilation (Figure 4A and B; Figure 4—figure supplement 1A and B). In the ‘closed state’ conformation, seen in the apo pump, the constriction is maintained by an inter-protomer hydrogen-bonding network, involving R367 in one subunit and T152, D153 and Y362 in the adjacent subunit (Andersen et al., 2002; Augustus et al., 2004; Bavro et al., 2008)(Figure 4—figure supplement 1A). This network of interactions is broken in the TolC of pump/MBX3132 complex (Figure 4—figure supplement 1B) and the pump/puromycin complex. Accordingly, we refer to TolC in the pump-ligand complexes as being in an ‘open state’ conformation.

Figure 4 with 3 supplements see all
Download asset Open asset
Interactions between TolC and AcrA in the AcrABZ-TolC pump with inhibitor MBX3132 bound.

(A) Segments of the 3.6 Å resolution cryoEM density map of the pump with fitted model showing the tip-to-tip interaction between a TolC protomer (red) and two α-helical hairpins of AcrA (orange and green). (B) Detail of the tip-to-tip interface of TolC and AcrA. Residue pairs of complementary mutations (Kim et al., 2015) identified at this interface are shown (Xu et al., 2011).

https://doi.org/10.7554/eLife.24905.016

We observe tip-to-tip interactions of the AcrA and TolC HTH motifs (Figure 4), which is similar to that proposed for the open state of the pump reported earlier (Jeong et al., 2016). These interactions account for the importance of the conserved Val-Gly-Leu/Thr element of TolC (Jeong et al., 2016; Song et al., 2014). At the AcrA protomer-I/TolC interface, contacts involve the backbone of G365 in the conserved VGL motif of TolC with the backbone of K140 and side chain of S139 in AcrA (Jeong et al., 2016). The N145 and T366 of TolC directly contact AcrA L132 in the pump, consistent with the results of in vivo site-specific cross-linking experiments (Xu et al., 2011) (Figure 4B). Our cryoEM structure shows the presence of residue pairs at the interface between AcrA and TolC (Figure 4—figure supplement 2) that were predicted to have co-evolved in the homologous proteins in a different efflux pump MdtNOP (Harley and Saier, 2000; Ovchinnikov et al., 2014). 3D classification identified a small subset of particles, in which TolC rotates by 60 degrees with respect to AcrA, but the alternative interface still retains the key contacts described above. As observed in the apo state, there are two distinct conformations of the HTH motif loops of AcrA protomers I and II that structurally conform to the slightly different contact surfaces of TolC (Figure 4). The flexibility of the HTH is consistent with observations from crystal structures (Figure 4—figure supplement 3) (Mikolosko et al., 2006).

In the six protomers of AcrA, the lipoyl and β-barrel domains form a stack of two continuous rings. The lipoyl domains have no interaction with either AcrB or TolC, and the β-barrel domains contact only the docking domain of AcrB (Figure 5). The membrane-proximal (MP) domain of AcrA interacts with both the docking domain and the top of the pore domain of AcrB. Each AcrB protomer makes two non-equivalent interactions with AcrA (Ntreh et al., 2016); i.e. each pair of AcrA protomers makes two conformationally distinct interfaces with AcrB, as was found with the interaction between AcrA and TolC (Figures 4 and 5). AcrA protomer-I makes extensive interactions with AcrB. The β-barrel domain of AcrA protomer-I docks on the DN and DC subdomains of AcrB, while its MP domain interacts with the PC1 subdomain, the linker region between PC2 and DC, and an extended loop of the DN subdomain from an adjacent AcrB protomer (Figure 5A and C). The spatial orientation of a β-hairpin motif, in the DN subdomain of AcrB, differs between the cryoEM and crystal structures. This conformational difference facilitates the interactions between AcrB and the β-barrel domain of AcrA protomer-II and a short α-helix of the DC subdomain from the adjacent AcrB protomer (Figure 5A and D). This contact resembles the interactions observed between the metal transporter subunits CusB and CusA, which are homologs of AcrA and AcrB, respectively (Su et al., 2011). This β-hairpin motif in the DN subdomain of AcrB has been shown to be important for the proper assembly of a functioning pump (Weeks et al., 2014). The linking region between AcrB subdomains PN2 and DN contacts the end of a β-sheet in the MP domain of AcrA protomer-II (Figure 5A and D).

Figure 5
Download asset Open asset
Interactions between AcrA and AcrB in the AcrABZ-TolC pump with inhibitor MBX3132 bound.

(A) A pair of AcrA protomers interacts with one AcrB protomer. Subdomains of the AcrB are labeled with different colors. (B) Cartoon schematic of the domains/subdomains shown in (A). (C) Interaction between AcrA protomer-I and AcrB. (D) Interaction between AcrA protomer-II and AcrB. There are 120 and 180 degrees rotations in (C) and (D) relative to (A), respectively.

https://doi.org/10.7554/eLife.24905.020

Overlaying the model of the apo pump with the pumps in the presence of ligands reveals profound conformational changes at the AcrA-TolC, AcrA-AcrA and AcrA-AcrB interfaces (Figure 6; Video 2). The quasi six-fold symmetry of the interface between AcrA helical hairpin and open-state TolC reduces to three-fold symmetry in the apo-form. The rearrangements of AcrA repack the β-barrel and membrane proximal domains in a different orientation on the surface of the AcrB and seal the gaps to the periplasm. Conformational changes in the periplasmic headpiece of AcrB instigate repacking of all four of the AcrA domains, and this causes the reorganization of the coiled-coiled domain of AcrA, which is the key step to organize the HTH motifs and open the TolC channel. Thus, the pump is a highly allosteric system in which conformational changes associated with ligand binding by the apo state of AcrB are communicated over a long distance to TolC through the repacking of protein-protein interfaces (Fischer and Kandt, 2013; Zgurskaya and Nikaido, 1999).

Figure 6
Download asset Open asset
Comparison of AcrA-TolC interactions in puromycin-bound and apo state.

(A) Structure of AcrAB-TolC/puromycin pump visualized in side view. (B) Top view (top panel) and side views (middle and bottom panels) of AcrA-TolC interface in puromycin-bound state. (C). Structure of apo AcrAB-TolC pump visualized in side view. (D) Top view (top panel) and side views (middle and bottom panels) of AcrA-TolC interface in apo state. Subunits are color-coded accordingly: TolC (red), AcrA (orange and green) and AcrB (blue).

https://doi.org/10.7554/eLife.24905.021
Video 2
Download asset
Morph animation between the disulfide-engineered pump-derived model and the derived model of the pump in the presence of puromycin.

Intermediate states are interpolated between the two resolved states.

https://doi.org/10.7554/eLife.24905.022

Discussion

In this study, we have captured structures of the full pump assembly with a closed channel in the apo-state and with an opened channel in the presence of inhibitor or antibiotic. We propose that the apo-form of the pump with a closed TolC represents a resting state (Figure 1), while the pump assemblies with opened TolC in the presence of transportable ligands represent transport states, in which the AcrB trimer adopts asymmetric LLT, LTT or LTO conformations. The three pump components undergo significant conformational changes when the assembly switches from the resting-state to the transport-states. In the transport-states, ligand binding is associated with large conformational changes in AcrB for both PN1/PC2 and PN2/PC1 structural modules that form the drug-binding pocket (Eicher et al., 2014) (Figure 3B). However, the interfaces between TolC and AcrA, and between AcrA and AcrB do not change significantly in the different transport states (LTO, LTT, LLT). We propose that ligand binding in the apo state instigates quaternary structural changes in AcrB that are communicated to AcrA, which in turn repacks to trigger tertiary structural changes in TolC that open the channel from a sealed resting state. The repacking to AcrA is critical to seal the gaps in the pump, which would otherwise allow substrate to leak into the periplasm. Thus, the long-distance allosteric coupling between AcrB and TolC, mediated through AcrA, ensures that TolC channel opening is synchronized with closing the assembly to the periplasm. When substrate is absent, the pump assembly returns to its resting state. The above conformational changes are accompanied by a contraction of the pump along the long axis in its transport state by nearly 10 Å (Figure 7). We anticipate that the transition from the resting state to the transport states will entail a local compression of the periplasm to accommodate the axial contraction of the pump, and local curvature of the outer membrane and inner membrane near the portal of AcrB. These changes may have impact on the energetics and kinetics of the transport process.

Figure 7
Download asset Open asset
Schematic cartoon of the transport mechanism.

(A) The resting state of the apo pump with TolC in closed-state and the AcrB trimer in LLL conformation. (B) The apo pump switches to a transport-state in the presence of transport substrate (s), opening the TolC channel (right arrow). In the transport-state, AcrB cycles through three, structurally distinct states (L, T and O), two of which are shown in the left panel (T and O). Cycling is obligatory for unidirectional transport, driven by coupling with transmembrane proton conduction through the TM domain (red arrow). In the absence of substrate, the pump reverts to the resting state and closes the TolC channel (left arrow). The views are cross-sections through the cell envelope, with only two protomers shown for each of the pump components. The inset cartoons on the left in (A) and the right in (B) show views down the molecular axis of the AcrB trimer, indicating the states with the configuration inferred from the cryoEM reconstructions. The model predicts a contraction along the long axis of the pump with the switch from apo- to transport-states.

https://doi.org/10.7554/eLife.24905.023

RND-type tripartite multidrug efflux pumps in Gram-negative bacteria are one of the major contributors to multidrug resistance. Efflux pump inhibitors could be used as adjunct therapies to increase the potency of existing antibiotics and counter the emergence of multidrug resistant bacteria. Efforts have been focused on the development of inhibitors for the transporter components. Due to the capacity of the transporters to recognize a broad range of substrates, such compounds are particularly difficult to identify (Opperman and Nguyen, 2015). In the presence of inhibitor MBX3132, the AcrB trimer is predominantly trapped in the symmetric TTT conformation (Figure 2—figure supplement 1E), possibly because the inhibitor has high affinity to AcrB and cannot be transported. Thus, the cryoEM structure suggests that inhibitors like MBX3132 could effectively inhibit AcrAB-TolC efflux through saturated binding and blocking all sites and prevent AcrB from cycling through different states. It may be possible to block drug translocation by targeting the assembly or conformational switching of the tripartite pumps themselves. In this regard, the interaction interfaces between components, which are well resolved in our cryoEM structure, may be important drug-targeting sites to inhibit the assembly of this and other tripartite pumps.

Materials and methods

Construction of vectors for overexpression of disulfide-bond stabilized AcrAB–TolC complex

Request a detailed protocol

The cysteine-substitutions were introduced for residues AcrA_S273 and AcrB_S258 into the individual components by site-directed mutagenesis using plasmid pAcBH as a template and primers AcrAS273C_F: 5’- GAT CAG ACC ACT GGG TGT ATC ACC CTA CGC GCT ATC ttc-3’/AcrAS273C_R: 5’-GAA GAT AGC GCG TAG GGT GAT ACA CCC AGT GGT CTG ATC-3’, AcrBS258C_F: 5’-GTG AAT CAG GAT GGT TGT CGC GTG CTG CTG CGT GAC-3’ /AcrBS258C_R: 5’-GTC ACG CAG CAG CAC GCG ACA ACC ATC CTG ATT CAC-3’, resulting in the construct pAcBH-AcrAS273CAcrBS258C. The tolC gene was amplified using primers TolCinf_F: 5’-AAG GAG ATA TAC ATA TGA AGA AAT TGC TCC CCA TTC TTA TCG GCC-3’ and TolC1392inf_R: 5’-TTG AGA TCT GCC ATA TGT CAA TCA GCA ATA GCA TTC TGT TCC GGC GT-3’. The PCR product was then inserted into the NdeI site of pRSFDuet-1 using the In-Fusion cloning method (Clontech), generating the construct pRSFDuet-1-tolC1392.

Protein preparation

Request a detailed protocol

The components of AcrABZ-TolC pump were co-expressed and the cellular membrane was prepared as described previously (Du et al., 2014). The purification procedure was optimized to improve the homogeneity and stability of the protein complex. 3.5 g of cellular membrane was re-suspended in 50 ml lysis buffer (20 mM Tris pH 8.0, 400 mM NaCl) and was solubilized with 1.5% DDM. The AcrABZ–TolC complex purified by nickel affinity chromatography was loaded onto a Superose 6 column equilibrated with GF buffer (20 mM Tris pH 8.0, 400 mM NaCl, 0.03% DDM). Fractions containing purified AcrABZ–TolC complex were pooled and concentrated to 0.1 mg ml−1 using a Vivaspin column (MWCO: 100 kDa). Amphipol A8-35 (100 mg ml−1) was mixed with the protein solution with a mass ratio of amphipol A8-35 to protein of 4:1. The mixture was incubated at 4°C for 3 hr. Polystyrene beads (Bio-Beads SM-2) were then added to the protein/DDM/amphipol A8-35 mixture with a mass ratio of Bio-Beads SM2 to detergent of 10:1. The mixture was gently rotated at 4°C overnight to remove DDM. The detergent-exchanged AcrABZ–TolC complex was concentrated to 2 mg ml−1 using a Vivaspin column (MWCO: 100 kDa). Puromycin (100 mM) or pyranopyridine inhibitor MBX3132 (50 mM) was added to the protein sample at a final concentration of 1 mM or 0.4 mM, respectively, and the mixture was incubated on ice for 1 hr before embedding in vitreous ice.

The disulfide-bond stabilized AcrAB–TolC complex was overexpressed and purified as follows: Genes encoding AcrA and AcrB were deleted from the chromosome of Escherichia coli strain C43 (DE3). The resulting C43 (DE3) ∆acrAB strain was transformed with plasmids pAcBH-AcrAS273CAcrBS258C and pRSFDuet-1-tolC1392. The culture was grown in 2xYT medium with 100 mg ml−1 carbenicillin and 50 mg ml−1 kanamycin at 37°C until the culture reached an absorbance, at 600 nm, of 0.5 and was then induced by the addition of 0.1 mM isopropyl 1-thio-β-D-galactopyranoside (IPTG) at 25°C overnight. Cells were harvested by centrifugation, and pellets from 10 L culture were re-suspended in 150 ml of lysis buffer (400 mM sodium chloride, 20 mM Tris-HCl, pH: 8.0) with 1 tablet per 50 ml EDTA free protease inhibitor cocktail tablets, 5 U ml−1 DNase I and 5 mg ml−1 lysozyme, and the mixture was stirred at 4°C for 1 hr to digest the cell wall. The cells were lysed by eight passages through a high-pressure homogenizer (EmulsiFlex) at 15,000 psi. Cell debris was pelleted by centrifugation at 9000 g for 30 min. Cellular membrane was pelleted by ultracentrifugation at 125,755 g for 3 hr. 3.5 gram of membrane pellets were re-suspended in 50 ml of lysis buffer with protease inhibitors and were solubilized by adding 1.5% DDM and stirring at 4°C for 3 hr. Debris was pelleted by ultracentrifugation at 125,755 g for 30 min. Imidazole was added to the membrane solution to a final concentration of 15 mM. Histidine-tagged AcrAB-TolC complex was purified by nickel affinity chromatography using a HiTrap 1 ml chelating column (GE Healthcare Life Sciences) equilibrated with GF buffer (400 mM sodium chloride, 20 mM Tris-HCl, pH 8.0, 0.03% DDM) containing 20 mM imidazole. The column was washed with 50 mM imidazole added to GF buffer. Purified AcrAB-TolC complex was eluted with 500 mM imidazole in GF buffer, concentrated and loaded onto a Superose 6 column equilibrated with GF buffer. Fractions 11–15 containing purified AcrAB-TolC complex were pooled and concentrated to 1 ml using a Vivaspin concentrator (MWCO 100 kDa). Amphipol A8-35 (100 mg/ml) was added to a final concentration of 10 mg/ml and the final volume was adjusted to 2 ml using lysis buffer. The mixture was incubated on ice for 3 hr. 250 mg of Polystyrene beads (Bio-Beads SM2) was then added to the protein/DDM/Amphipols A8-35 mixture and rotated at 4°C overnight to remove DDM. The mixture was loaded onto a mini chromatography column to remove the Polystyrene beads (Bio-Beads SM2). The detergent-exchanged AcrAB–TolC complex was concentrated to 2 mg ml−1 using a Vivaspin column (MWCO: 100 kDa) before embedding in vitreous ice.

Crystallization of AcrBZ–DARPin

Request a detailed protocol

The AcrBZ–DARPin complex has been found to yield well diffracting crystals in the presence of antibiotics. The complex was purified as described previously (Du et al., 2014). The protein was diluted to 10 mg ml −1 using sample buffer (10 mM HEPES pH: 7.5, 50 mM sodium chloride, 0.03% DDM) and was incubated with 1 mM of puromycin for 3 hr at 4°C before crystallization trials. The crystals of AcrBZ–DARPin complex were grown at 20°C using the hanging-droplet vapor diffusion method by mixing 4 μl of protein with 2 μl of reservoir solution (80 mM Bis-Tris, pH 6.0, 50 mM sodium citrate, 120 mM KCl, 10% PEG 4000, 0.5% N,N-dimethyldodecylamine N-oxide). 200 μl of oil (mixture of 40% silicon oil and 60% paraffin oil) was applied over 1 ml reservoir solution to control the rate of vapor diffusion. Crystals appeared 24 hr after setting up the crystallization trials and reached maximal size in 1 week. The crystals were transferred briefly into reservoir solution supplemented with 25% v/v glycerol as cryo-protectant before flash freezing in liquid nitrogen.

Crystallographic data collection and structure refinement

Request a detailed protocol

The AcrBZ–DARPin complex in the presence of puromycin crystallized in space group P212121. Data sets were collected using beamline I24 at the Diamond Light Source. The diffraction data were processed using iMosflm (Battye et al., 2011) and scaled using SCALA (Evans, 2006). As the crystals are sensitive to radiation damage, diffraction data of contiguous 15-degree wedges were collected from multiple crystals and merged to obtain a full dataset. Structures were solved by molecular replacement using Phaser with AcrB–DARPin complex (PDB accession: 4DX5) as a search model and refined using PHENIX and REFMAC5 (Murshudov et al., 1999). Coot was used for modeling (Emsley et al., 2010). Maps calculated from molecular replacement using the AcrB-DARPin complex revealed clear electron density for AcrZ. Data collection and refinement parameters are presented in Supplementary file 3. Structures are shown in Figure 2—figure supplement 3.

Electron cryo-microscopy

Request a detailed protocol

For the AcrABZ-TolC/Puromycin sample, a 2.0 μl aliquot at 2 mg/ml was applied onto holey carbon film supported by a 200-mesh R2/1 Quantifoil grid (Quantifoil) that had been previously washed and glow discharged. The grid was blotted and rapidly frozen in liquid ethane using a Vitrobot IV (FEI) with constant temperature and humidity. The grid was stored in liquid nitrogen before imaging. Images of frozen-hydrated AcrABZ-TolC/puromycin particles were acquired on a FEI Tecnai G2 Polara electron microscope (FEI) operated at 300 kV using a K2 Summit direct electron detector camera (Gatan).

For the samples AcrABZ-TolC/MBX3132 and apo AcrAB-TolC, a 3 μl aliquot at a concentration of 2 mg ml−1 was applied onto glow-discharged holey carbon grid (Quantifoil Au R1.2/1.3, 300 mesh). The grid was blotted and flash frozen in liquid ethane using a Vitrobot IV (FEI) with constant temperature and humidity. The grid was stored in liquid nitrogen before imaging. Zero-energy-loss images of frozen-hydrated AcrABZ-TolC/MBX3132 or apo AcrAB-TolC particles were recorded automatically on an FEI Titan Krios electron microscope at 300 kV, using a slit width of 20 eV on a GIF Quantum energy filter and a Gatan K2-Summit direct electron detector.

The data collection parameters for all three specimens are summarized in Supplementary file 1.

Image processing and 3D reconstruction

Request a detailed protocol

For the AcrABZ-TolC/puromycin pump, dose-fractionated super-resolution raw image stacks were binned 2 × 2 by Fourier cropping resulting in a pixel size of 1.62 Å for further image processing. Each image stack was subjected to motion correction using dosefgpu_driftcorr (Li et al., 2013), and a sum of sub-frames 1–29 in each image stack was used for further image processing. The signal in the motion-corrected images extends beyond 4 Å. Defocus and astigmatism were determined for each micrograph by CTFFIND3 (Mindell and Grigorieff, 2003). Each image was binned 2 × 2 to enhance image contrast for particle picking. 99,385 particles were boxed out manually from 6456 micrographs using e2boxer.py. An initial map was generated with 3-fold symmetry imposition from 2-D reference-free averages using EMAN2 (Tang et al., 2007). The initial reference map was low pass filtered to 60 Å resolution and was used as a starting point for the RELION-1.4 refinement (Scheres, 2012). The first round of refinement resulted in a sub-nanometer resolution map. After this step, several rounds of iterative 3D classification and 3D auto-refinement were run to extract a self-consistent subset of the particle data. 67,436 particles were selected after 3D classification and further refined with averages from sub-frames 2–16 to achieve 3.9 Å resolution based on the gold standard criterion (Henderson et al., 2012). The final refinement was done using oversampling by a factor of two on the whole dataset. A soft mask in RELION post-processing was applied before computing the FSCs. The final resolution was estimated by 0.143 cutoff of FSCt. Local resolution variations were estimated with ResMap using the two independent maps (Kucukelbir et al., 2014).

For AcrABZ-TolC/MBX3132, the software MotionCorr (Li et al., 2013) was used for whole-frame motion correction, Ctffind4 (Rohou and Grigorieff, 2015) for estimation of the contrast transfer function parameters, and RELION-1.4 package for all other image processing steps. For apo AcrAB-TolC, the software MotionCor2 (Zheng et al., 2016) was used for whole-frame motion correction and dose weighting, Gctf (Zhang, 2016) for estimation of the contrast transfer function parameters, and RELION-2.0/beta package for all other image processing steps. A particle subset was manually selected to calculate reference-free 2D class averages, which was then used as templates for automated particle picking of the entire data set. The templates were lowpass filtered to 20 Å to limit model bias. Then initial runs of 2D and 3D classifications were used to remove the heterogeneous particles, as well as the false positive particles from the auto-picking. We selected good particles for further analysis based on the quality and high resolution in the 2D and 3D classification.

For the AcrABZ-TolC/MBX3132 pump, 65,256 particles were picked automatically from a total of 1150 micrographs. After initial 2D and 3D classifications, a homogenous subset of 33,587 particles was selected for a first 3D auto-refinement, generating a reconstruction with a resolution of 6.67 Å. After per-particle motion correction and radiation-damage weighting, the polished particles were submitted to a second round of 3D auto-refinement by applying a soft mask around the TolC, AcrA and the periplasmic headpiece of AcrB (Scheres, 2016). These polished particles gave a reconstruction with a resolution of 3.6 Å based on the Gold-standard FSC 0.143. In the case of apo AcrAB-TolC pump, 95,410 particles were picked automatically from a total of 2292 micrographs. After initial 2D classification and two rounds of 3D classifications, 13,544 homogeneous particles were selected for 3D auto-refinement, which generated a map with a resolution of 6.5 Å based on the Gold-standard FSC 0.143. 3D classification also identified a small subset of particles, in which TolC is rotated by 60 degrees with respect to AcrA.

All 3D classifications and refinements were started from a 50–60 Å low-pass filtered initial model, the first of which was made from our previous 16 Å resolution map. The density map was sharpened by applying a negative B-factor estimated by automated procedures (Rosenthal and Henderson, 2003). Local resolution variations were estimated using ResMap and visualized with Chimera (Pettersen et al., 2004).

Symmetry release and focused classification

Request a detailed protocol

The density map generated from RELION was used to assess the structural asymmetry of the pump by performing focused classification. A soft mask of the targeted region of this map was generated using relion_mask_create with a soft edge extension of 5 pixels (Figure 2—figure supplement 1A; Figure 3—figure supplement 1A). We used this masked volume to set to zero the density of the targeted region from the 3D map of the whole pump, and generated a modified map (Figure 2—figure supplement 1B; Figure 3—figure supplement 1B). This altered density map was rotated by 120 and 240 degrees about the molecular three-fold axis. The three resulting 3D density maps represent the entire pump with the deletion of each of the three-targeted regions (i.e. each of the three subunits of AcrB) (Figure 2—figure supplement 1C; Figure 3—figure supplement 1C). Next, each of these three altered density maps was projected in 2D in the corresponding orientations of each of the raw particle images and subtracted from the original image (Figure 2—figure supplement 1C; Figure 3—figure supplement 1C). Altogether, the particle images (33,337 for the pump with MBX3132 and 67,436 for the pump in presence of puromycin) used in the final reconstruction generated three times that many subtracted particle images. Each of the three subtracted projections from an individual particle image (Figure 2—figure supplement 1C; Figure 3—figure supplement 1C) was assigned appropriate Euler angles so that they are all in the same orientation in the 3D map to permit 3D classification (Figure 2—figure supplement 1D; Figure 3—figure supplement 1D). The soft mask generated in the beginning of this process (Figure 2—figure supplement 1A; Figure 3—figure supplement 1A) was used for the 3D classifications to remove density outside the mask. We ran rounds of 3D classifications using the particle alignment parameters of the symmetry-imposed map of the pump. Each round was iterated 25 times and classified into several subsets (Figure 2—figure supplement 1D; Figure 3—figure supplement 1D). Focused classification of pump with MBX3132 result in two populations different in the binding site of MBX3132 according to crystal structure (Figure 3—figure supplement 1D). Subsets of original and unmodified particle images with different putative ligand binding sites were selected for 3D refinement with no imposed symmetry. We carried out the similar approach on the data of the pump in presence of puromycin. After 3 rounds of 3D classification, we identified two significant conformations. Both class 1 and 2 have a conformation in the pore domain that is closed towards the periplasm. In contrast, the pore domain in class four shows an opened conformation facing the periplasm. We selected the original and unmodified pump particle images that assign only one member from class 1 and 2 for another round of 3D refinement without symmetry constraint (Figure 2—figure supplement 1E).

Model docking and optimization

Request a detailed protocol

The crystal structures of trimeric AcrBZ (PDB code: 4C48) and trimeric TolC (PDB code: 1EK9) were docked into the cryoEM map by using Chimera. Both the TolC and AcrBZ models were adjusted manually to optimize the local fit to density using Coot. Individual domains of AcrA from the crystal structure (PDB code: 2F1M) including β-barrel domain, lipoyl domain and α-helical hairpin domain were fitted to the density map using Chimera (Pettersen et al., 2004). A homology model of the MP domain of AcrA was built based on the structure of MexA (PDB code: 2V4D) and fit to the density. Four conformationally different structures of the α-helical hairpin domain (Chain A-D in the crystal structure) (Figure 4—figure supplement 3) were evaluated for their fit into the density map of our two protomers. The α-helical hairpin structures from chains A and D fit best into the density map for AcrA protomer-I and for protomer-II in the pump assembly, respectively. The model for the complete pump was optimized using Phenix real-space refinement (Wang et al., 2014) with three-fold symmetry imposed (Supplementary file 2). The model of the whole pump complex was validated by computing a FSC with the density map. It is 4.1 Å at 0.5 FSC (Figure 3—figure supplement 2A). MolProbity (Chen et al., 2010) statistics were computed to ensure proper stereochemistry (Supplementary file 2).

The 5.9 Å resolution map calculated from the particle subset in class 1 and 2 without symmetry restraint was fit with the symmetric model based on the 3.9 Å density map. We found that TolC and AcrA did not show conformational change. We noticed during the modeling process of TolC alone that the TolC and AcrA interface is polymorphic, with a slight majority of TolC rotated 60 degrees with respect to the dominant configuration seen in the inhibitor bound pump structure, as shown in Figure 3.

However, the region for the AcrB in the new map revealed significant mismatch to the symmetric model. A better fit was obtained using the asymmetric AcrBZ crystal structure. Model optimization was then done with Phenix real-space refine (Wang et al., 2014) using stronger secondary-structure restraints parameters to maintain proper stereochemistry where weak density exists. Moreover, atomic displacement parameters (B-factors) were computed to determine the level of resolvability throughout the map, which correlated well with Resmap results. The Coot adjustments and Phenix model optimization were iterated one additional round to ensure that the model was a good representation of the map.

For the modeling of the apo state AcrAB-TolC pump, resolved to 6.5 Å resolution, a similar method to modeling of the asymmetric maps was utilized. Crystal structures were rigid-body fit into the density map and model optimization was then carried out with Phenix real-space refine. Again, due to the weaker resolution stronger stereochemical and secondary structure restraints were used to ensure that α-helices and β-sheets did not deviate far from their expected geometry. Manual adjustments were kept to a minimum to reduce human bias in the modeling procedure, with Coot only being used to fix obvious errors such as C-beta deviations. A final check of MolProbity and cross-correlation was done to ensure model quality.

Accession numbers

Request a detailed protocol

Accession codes: cryoEM maps and models of AcrABZ-TolC have been deposited in EMDB under accession code EMD-8636 (apo form), EMD-8640 (puromycin), EMD-3636 (MBX3132 inhibitor), and in the Protein Data Bank under accession code 5V5S(apo form), 5V78 (puromycin), and 5NG5 (MBX3132). The crystal structure of AcrBZ has been deposited to Protein Data Bank under accession code 5NC5.

Data availability

The following data sets were generated

References

    1. Emsley P
    2. Lohkamp B
    3. Scott WG
    4. Cowtan K
    (2010) Features and development of coot
    Acta Crystallographica Section D Biological Crystallography 66:486–501.
    https://doi.org/10.1107/S0907444910007493
    1. Evans P
    (2006) Scaling and assessment of data quality
    Acta Crystallographica Section D Biological Crystallography 62:72–82.
    https://doi.org/10.1107/S0907444905036693
    1. Harley KT
    2. Saier MH
    (2000)
    A novel ubiquitous family of putative efflux transporters
    Journal of Molecular Microbiology and Biotechnology 2:195–198.

Decision letter

  1. Olga Boudker
    Reviewing Editor; Weill Cornell Medical College, United States

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

Thank you for submitting your article "An allosteric transport mechanism for the AcrAB-TolC Multidrug Efflux Pump" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom, Olga Boudker, is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Richard Aldrich as the Senior Editor. The following individual involved in review of your submission has agreed to reveal his identity: Martin Picard (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission. The reviewers were all very enthusiastic about your work and had suggestions that were editorial in nature:

Reviewer 1:

In the manuscript, the authors present a series of Cryo-EM and crystal structures that illuminate the allosteric coupling in RND family of transporters. They describe the structures of the apo, and the substrate- or inhibitor-bound states of AcrAB-TolC complex. In the apo state, the protein channel that extends through the periplasmic space is closed at the AcrA-TolC juncture. Binding of the substrate or inhibitor to AcrB is coupled to conformational changes through AcrA to the opening of the constriction at AcrA-TolC interface. The presented structures are informative and the work seems solid. I have only comments that are editorial in nature:

1) The physiological role of AcrZ and why it was used in the studies of the substrate-bound pump need to be described in more detail.

2) It is not clear whether there are any structural changes that would propagate through AcrA between substrate- and inhibitor-bound pumps.

3) More detailed figures fro Apo pump showing differences between two conformations of AcrA and their packing with TolC. Also, showing the openings to the periplasm.

4) Show puromycin density in Figure 3—figure supplement 3.

5) Figure 3—figure supplement 3: spelling "deepinding".

6) Show more detailed comparison of AcrA- TolC interactions in ligand-bound and apo state.

7) A bit more detailed description of what are the overall structural changes between closed apo to open substrate-bound state?

8) Switch the order of Figures 2 and 3.

9) Figure 4 could be clearer if the structures in A and B were shown on the same scale and in the same orientation.

10) In Figure 4—figure supplement 1, which structure is shown in B?

11) Subsection “The structure of pump in complex with inhibitor MBX3132 reveals the interaction interfaces between pump components at atomic resolution” paragraph two, did you mean Figure 4—figure supplement 1B?

12) Please show in Figure 4 the location of the VGL motif in TolC and the interacting residues of AcrA.

13) In subsection “The structure of pump in complex with inhibitor MBX3132 reveals the interaction interfaces between pump components at atomic resolution” paragraph three what are the residue pairs?

Reviewer 2

The Cambridge group earlier was successful in elucidating, for the first time, the structure of the AcrAB-TolC tripartite efflux complex. Here they report, together with other scientists, on a more refined structure of this complex both in its resting state and in the presumably transport-engaged state, and show how the closed end of the TolC channel may become opened in association of the altered conformation of the pump AcrB. Improvement in technology that led to these wonderful results included many features, such as substitution of Amphipol for conventional detergents and use of focused classification in selecting images. The manuscript is fine. I have only mostly minor comments.

Subsection “CryoEM structure determination” paragraph one. Citing the two papers where amphipol was successfully used would not be enough, especially the journal is targeted toward diverse audience (and also because these two papers do not even cite the original Jean-Luc Popot 1996 PNAS paper on amphipol!).

Paragraph two of the same section. I suppose that disulfide-linked AcrAB recruits not only TolC, but also additional AcrA monomers, as seen in Figure 1—figure supplement 2A.

In the last line of paragraph two: I believe that this should be Figure 1—figure supplement 2, B and C.

Even with the non-intuitive naming convention of supplementary figures, it does not help that Figure 3 supplement comes before Figure 2 supplement. Please fix this confusing numbering system for the supplementary figures.

Subsection “The structure of pump in complex with inhibitor MBX3132 reveals the interaction interfaces between pump components at atomic resolution”. Inhibitor MBX3132 gave such wonderful results precisely because it binds very tightly to the binding pocket, as seen in the very large negative values of its calculated binding energy in Sjuts et al., 2016 The high affinity of this compound must be mentioned somewhere.

In paragraph three of the same section: conserved Val-Gly-Leu/Thr element in AcrA? It may be obvious, but it would be better to specify AcrA because TolC is also discussed in the same paragraph.

Also in this paragraph, the tip-to-tip interaction of AcrA and TolC HTH motifs is made very clear with Figure 4. However, non-specialists will miss the absence of discussion (and especially figure!) on what happens to these residues in the complex containing a closed TolC conformer. Also, the closed conformer of TolC is supposed to be stabilized by multiple interactions between coiled coils. Are the tip-to-tip interactions enough to overcome these pre-existing multiple interactions?

Paragraph four of the same section and Figure 5. This is very nice and informative, but I miss the comparison with the situation in the apo complex. The next paragraph talks about "overlays", but again there is no figure showing such overlays (especially for the AcrA-TolC interface) even in the supplement.

Discussion section. If TolC-AcrA interface does not change significantly by the binding of the substrate to AcrB, how could the TolC gate become open in the normal operation of the pump? This reminds us that the "open" complex was seen only when MBX3132 was bound to all three of the trimers, and could be an artifact of the experimental setup.

Reviewer 3

The manuscript by Wang et al.et al. describes the near atomic cryoEM structure of the AcrAB-TolC efflux pump from E. coliE. coli. Structures of the full pump assembly have been obtained in the apo state and in the presence of a substrate or of an inhibitor. For the first time the authors describe the pump under its closed and open conformations. To do so, they have optimized the biochemical procedure to stabilize the genuine complex, by inserting cysteine residues where the contacts between partners of the pumps are anticipated to take place and by subjecting the pump to disulphide crosslinking. The assembly is stabilized by amphipols. The structures have been obtained by crystallography (AcrBZ in the presence of substrate) and cryoEM (tri-quadripartite structures) using the latest technical developments and methodologies.

This work is a real tour de force and will undoubtedly have a great impact in the field. Hence we fully recommend it to be published in eLIFEeLife without further additional experiments to be performed. We only point below minor comments in order for the manuscript to be fully acceptable.

AcrZ seems to be present in the structures obtained in the presence of puromycin and in that in the presence of the inhibitor but absent in the apo states. The authors do not comment upon that point. Could it be that indeed AcrZ is recruited differently whether the substrate is present or not? What could be the possible role of AcrZ on the transport and/or activity?

Note that we advise the authors not to color AcrZ in purple because it is difficult (in particular for a daltonian reviewer) to see it in the very middle of an AcrB molecule colored in blue!

Subsection “The transport state of the full pump with opened channel” first paragraph: Could the authors give the reference to the paper mentioning that puromycin is a validated transport substrate of AcrB?

We believe that here and there, additional papers could be mentioned, see below:

Subsection “CryoEM Structure determination”: There are more and more papers showing that amphipols are indeed very useful in the EM field. The authors refer to the work by Baker, Fan and Serysheva, 2015 and by Liao, Cao, Julius and Cheng, 2013. We suggest to also cite the following review by J-L Popot (Zoonens and Popot, 2014), so that the lab where this molecule was developed is also acknowledged.

Subsection “The structure of pump in complex with inhibitor MBX3132 reveals the interaction interfaces between pump components at atomic resolution”, paragraph two: Maintenance of TolC thanks to the interprotomer network involving R367, T152, D153 and Y362 was already shown thanks to electrophysiology (Andersen et al., 2002), microbiology (Augustus et al., 2004) and crystallography (Bavro et al., 2008).

Paragraph four of the same section: Non equivalent interactions between each AcrB promoter and each pair of AcrA promoter is reminiscent to the work by the laboratory of Helen Zgurskaya on the related pump TriABC-OpmH (Ntreh, Weeks, Nickels and Zgurskaya, 2016)

In paragraph five of the section: Wang et al. comment on the fact that the pump is a highly allosteric system in which conformational changes associated with ligand binding by AcrB are communicated to TolC. Conversely, it was shown that the MFP is mandatory for the activity of the pump (Zgurskaya and Nikaido, 1999), a fact that was nicely discussed by Fischer and Kandt (in section 4.3.3 of Fischer and Kandt, 2013, BBA – Biomembranes 1828, 632-641).

https://doi.org/10.7554/eLife.24905.041

Author response

Reviewer 1:

[…] 1) The physiological role of AcrZ and why it was used in the studies of the substrate-bound pump need to be described in more detail.

We have expanded the description in the text about the known function of AcrZ. Regarding why we used AcrZ in the studies of the substrate-bound pump, AcrZ is a modulator of AcrB for the efflux of substrate puromycin, and the his-tagged version of AcrZ helped to purify the assembled pump from the expression host.

2) It is not clear whether there are any structural changes that would propagate through AcrA between substrate- and inhibitor-bound pumps.

Within the resolution limit of the substrate- and inhibitor-bound structures, we cannot see significant structural changes propagated through AcrA. In the substrate- and inhibitor-bound pumps, AcrB adopts LTO and TTT states respectively. The interaction interfaces between AcrA and AcrB in these two pump assemblies are slightly different, and the conformation of AcrA protomers is slightly different to adapt to the conformation of AcrB. However, the interfaces between AcrA and TolC are similar. In the case of the switch from apo- to ligand-bound states, the AcrA helical hairpins repack, as a result of changes in the interactions between the AcrA and AcrB. We suggest that these are predominantly quaternary structural changes.

3) More detailed figures fro Apo pump showing differences between two conformations of AcrA and their packing with TolC. Also, showing the openings to the periplasm.

We have included a new figure that shows these features (Figure 1—figure supplement 5).

4) Show puromycin density in Figure 3—figure supplement 3.

We have now included the puromycin density in the new Figure 2—figure supplement 3 (note that the order of the figures have been changed in the revised text).

5) Figure 3—figure supplement 3: spelling "deepinding".

This is now corrected.

6) Show more detailed comparison of AcrA- TolC interactions in ligand-bound and apo state.

The comparison is shown in the Video 2, and in addition, we have included a new Figure 6.

7) A bit more detailed description of what are the overall structural changes between closed apo to open substrate-bound state?

We have included a new Figure 6 and reference to Video 2, which help to visualise the overall structural changes between the apo and substrate-bound states.

8) Switch the order of Figures 2 and 3.

We have changed the order in the new draft.

9) Figure 4 could be clearer if the structures in A and B were shown on the same scale and in the same orientation.

We have revised the figure to place the structures on the same scale.

10) In Figure 4—figure supplement 1, which structure is shown in B?

The figure shows the TolC structure in the inhibitor-bound pump (which is very similar to the drug-bound structure). We have modified the figure legend to explain that the structure is for the inhibitor bound state.

11) Subsection “The structure of pump in complex with inhibitor MBX3132 reveals the interaction interfaces between pump components at atomic resolution” paragraph two, did you mean Figure 4—figure supplement 1B?

Yes, we do. Thank you for spotting this. It has now been corrected in the modified text.

12) Please show in Figure 4 the location of the VGL motif in TolC and the interacting residues of AcrA.

These have now been annotated.

13) In subsection “The structure of pump in complex with inhibitor MBX3132 reveals the interaction interfaces between pump components at atomic resolution” paragraph three what are the residue pairs?

The pairs are co-evolving residues, and these are listed in Figure 4—figure supplement 2.

Reviewer 2

[…] Subsection “CryoEM structure determination” paragraph one. Citing the two papers where amphipol was successfully used would not be enough, especially the journal is targeted toward diverse audience (and also because these two papers do not even cite the original Jean-Luc Popot 1996 PNAS paper on amphipol!).

We have included a reference to the Popot, 2010 review and to other papers from Popot and colleagues.

Paragraph two of the same section. I suppose that disulfide-linked AcrAB recruits not only TolC, but also additional AcrA monomers, as seen in Figure 1—figure supplement 2A.

Based on the structure, only three AcrAs form a disulphide bond with the AcrB trimer, and the other 3 AcrAs with single-cysteine mutation in the assembly will be free, and not disulphide linked with AcrB or each other. (The AcrA and AcrB with single-cysteine mutations are overexpressed in an acrAB null strain, so only single-cysteine mutants will be present in the assembly). We have now made it more explicit in the revised text that there are free subunits in the cysteine engineered pump.

In the last line of paragraph two: I believe that this should be Figure 1—figure supplement 2, B and C.

We have changed this.

Even with the non-intuitive naming convention of supplementary figures, it does not help that Figure 3 supplement comes before Figure 2 supplement. Please fix this confusing numbering system for the supplementary figures.

In the revised version, we swapped the order of the Figure 2 and Figure 3, and the numbers for the corresponding supplement figures have also been changed.

Subsection “The structure of pump in complex with inhibitor MBX3132 reveals the interaction interfaces between pump components at atomic resolution”. Inhibitor MBX3132 gave such wonderful results precisely because it binds very tightly to the binding pocket, as seen in the very large negative values of its calculated binding energy in Sjuts et al., 2016. The high affinity of this compound must be mentioned somewhere.

We have changed the text to read:

“The inhibitor MBX3132 likely binds AcrB tightly, since it is active in the nanomolar concentration range (Sjuts et al., 2016), and the compound was used to lock the pump in a more homogeneous conformation.”

In paragraph three of the same section: conserved Val-Gly-Leu/Thr element in AcrA? It may be obvious, but it would be better to specify AcrA because TolC is also discussed in the same paragraph.

We have changed the text here to read:

“These interactions account for the importance of the conserved Val-Gly-Leu/Thr element of TolC “

Also in this paragraph, the tip-to-tip interaction of AcrA and TolC HTH motifs is made very clear with Figure 4. However, non-specialists will miss the absence of discussion (and especially figure!) on what happens to these residues in the complex containing a closed TolC conformer. Also, the closed conformer of TolC is supposed to be stabilized by multiple interactions between coiled coils. Are the tip-to-tip interactions enough to overcome these pre-existing multiple interactions?

We have added a new Figure 6 that compares the tip-to-tip interactions in the apo- and ligand-bound pumps. Based on the observation that the TolC channel is open in the ligand-bound states, we can conclude that the tip-to-tip interactions are sufficient to overcome the stabilising interactions that keep free TolC in a closed state.

Paragraph four of the same section and Figure 5. This is very nice and informative, but I miss the comparison with the situation in the apo complex. The next paragraph talks about "overlays", but again there is no figure showing such overlays (especially for the AcrA-TolC interface) even in the supplement.

We have added a new Figure 5—figure supplement 1 and accompanying text that compares the apo- and ligand-bound pumps. These are also shown in the accompanying Video 2.

Discussion section. If TolC-AcrA interface does not change significantly by the binding of the substrate to AcrB, how could the TolC gate become open in the normal operation of the pump? This reminds us that the "open" complex was seen only when MBX3132 was bound to all three of the trimers, and could be an artifact of the experimental setup.

It is not true that the open complex was seen only in the presence of the inhibitor; it is also seen in the presence of substrate puromycin. The interfaces are similar in the substrate-bound and inhibitor-bound pumps with an opened TolC channel. In the transport-state, the interface does not change significantly as AcrB cycles through the three conformations, LTO. However, the TolC-AcrA interface does change significantly between the apo and substrate-bound pumps, with a closed and opened TolC channel, respectively.

In the switch from the apo- to the ligand-bound states, the conformational changes in AcrA are effectively quaternary structural adjustments. These entail repacking of the AcrA helical hairpin domains between adjacent protomers in the tube-like portion of the pump, and repacking of the membrane proximal and β-barrel domains with the surface of AcrB. We think that ligand binding by AcrB triggers changes in its periplasmic headpiece, and this in turn changes the interaction surface with the AcrA membrane proximal and β-barrel domains that results in a rotation movement of those domains.

Reviewer 3

[…] AcrZ seems to be present in the structures obtained in the presence of puromycin and in that in the presence of the inhibitor but absent in the apo states. The authors do not comment upon that point. Could it be that indeed AcrZ is recruited differently whether the substrate is present or not? What could be the possible role of AcrZ on the transport and/or activity?

Yes, it is correct that AcrZ is present in the puromycin and inhibitor bound structures but not in the apo-structure. AcrZ is a modulator of AcrB for the efflux of substrate puromycin. We used the his- tagged AcrZ to pull down the full assembly to prepare the ligand-bound structures. We did not co- express AcrZ with the disulphide-stabilised construct. We have changed the text to make this clearer. AcrZ may not be recruited differently depending on the presence of ligand, based on comparisons of our crystal structure of AcrZ/AcrB complex in the absence of ligand (Du et al., 2014) and the AcrZ/AcrB/puromycin complex reported here. The functional role of AcrZ in the complex is still unclear.

Note that we advise the authors not to color AcrZ in purple because it is difficult (in particular for a daltonian reviewer) to see it in the very middle of an AcrB molecule colored in blue!

Subsection “The transport state of the full pump with opened channel” first paragraph: Could the authors give the reference to the paper mentioning that puromycin is a « validated transport substrate » of AcrB ?

The reference is Hobbs et al., (2012).

We believe that here and there, additional papers could be mentioned, see below:

Subsection “CryoEM Structure determination”: There are more and more papers showing that amphipols are indeed very useful in the EM field. The authors refer to the work by Baker, Fan and Serysheva, 2015 and by Liao, Cao, Julius and Cheng, 2013. We suggest to also cite the following review by J-L Popot (Zoonens and Popot, 2014), so that the lab where this molecule was developed is also acknowledged.

Subsection “The structure of pump in complex with inhibitor MBX3132 reveals the interaction interfaces between pump components at atomic resolution”, paragraph two: Maintenance of TolC thanks to the interprotomer network involving R367, T152, D153 and Y362 was already shown thanks to electrophysiology (Andersen et al., 2002), microbiology (Augustus et al.,2004) and crystallography (Bavro et al., 2008).

Paragraph four of the same section: Non equivalent interactions between each AcrB promoter and each pair of AcrA promoter is reminiscent to the work by the laboratory of Helen Zgurskaya on the related pump TriABC-OpmH (Ntreh, Weeks, Nickels and Zgurskaya, 2016)

In paragraph five of the section: Wang et al. comment on the fact that the pump is a highly allosteric system in which conformational changes associated with ligand binding by AcrB are communicated to TolC. Conversely, it was shown that the MFP is mandatory for the activity of the pump (Zgurskaya and Nikaido, 1999), a fact that was nicely discussed by Fischer and Kandt (in section 4.3.3 of Fischer and Kandt, 2013, BBA – Biomembranes 1828, 632-641).

We have included these references in the modified text.

https://doi.org/10.7554/eLife.24905.042

Article and author information

Author details

  1. Zhao Wang

    1. National Center for Macromolecular Imaging, Baylor College of Medicine, Houston, United States
    2. Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, United States
    Contribution
    ZW, Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft, Writing—review and editing
    Contributed equally with
    Guizhen Fan and Corey F Hryc
    Competing interests
    The authors declare that no competing interests exist.

  2. Guizhen Fan

    Department of Biochemistry and Molecular Biology, Structural Biology Imaging Center, The University of Texas Health Science Center at Houston Medical School, Houston, United States
    Contribution
    GF, Data curation, Formal analysis, Funding acquisition, Writing—editing draft
    Contributed equally with
    Zhao Wang and Corey F Hryc
    Competing interests
    The authors declare that no competing interests exist.

  3. Corey F Hryc

    1. National Center for Macromolecular Imaging, Baylor College of Medicine, Houston, United States
    2. Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, United States
    3. Graduate Program in Structural and Computational Biology and Molecular Biophysics, Baylor College of Medicine, Houston, United States
    Contribution
    CFH, Data curation, Validation, Visualization, Methodology, Writing—review and editing
    Contributed equally with
    Zhao Wang and Guizhen Fan
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7277-5249

  4. James N Blaza

    MRC Mitochondrial Biology Unit, Cambridge Biomedical Campus, Cambridge, United Kingdom
    Contribution
    JNB, Methodology, Writing—review and editing
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-5420-2116

  5. Irina I Serysheva

    Department of Biochemistry and Molecular Biology, Structural Biology Imaging Center, The University of Texas Health Science Center at Houston Medical School, Houston, United States
    Contribution
    IIS, Data curation, Formal analysis, Funding acquisition, Writing—editing draft
    Competing interests
    The authors declare that no competing interests exist.

  6. Michael F Schmid

    1. National Center for Macromolecular Imaging, Baylor College of Medicine, Houston, United States
    2. Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, United States
    Contribution
    MFS, Formal analysis, Validation, Methodology
    Competing interests
    The authors declare that no competing interests exist.

  7. Wah Chiu

    1. National Center for Macromolecular Imaging, Baylor College of Medicine, Houston, United States
    2. Verna and Marrs McLean Department of Biochemistry and Molecular Biology, Baylor College of Medicine, Houston, United States
    3. Graduate Program in Structural and Computational Biology and Molecular Biophysics, Baylor College of Medicine, Houston, United States
    Contribution
    WC, Supervision, Funding acquisition, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    wah@bcm.edu
    Competing interests
    The authors declare that no competing interests exist.

  8. Ben F Luisi

    Department of Biochemistry, University of Cambridge, Cambridge, United Kingdom
    Contribution
    BFL, Conceptualization, Data curation, Supervision, Funding acquisition, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    bfl20@cam.ac.uk
    Competing interests
    The authors declare that no competing interests exist.
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-1144-9877

  9. Dijun Du

    Department of Biochemistry, University of Cambridge, Cambridge, United Kingdom
    Contribution
    DD, Conceptualization, Data curation, Formal analysis, Supervision, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    dd339@cam.ac.uk
    Competing interests
    The authors declare that no competing interests exist.

Funding

U.S. National Library of Medicine (Training Program in Biomedical Informatics, T15LM007093 (PI Lydia Kavraki))

  • Corey F Hryc

University of Texas at Austin (ACI-1134872 (PI Daniel Stanzione))

  • Corey F Hryc

Medical Research Council (Mitochondrial Biology Unit, U105663141 (PI Judy Hirst))

  • James N Blaza

American Heart Association (16GRNT29720001)

  • Irina I Serysheva

National Institutes of Health (R01GM072804)

  • Irina I Serysheva

National Institutes of Health (S10OD016279)

  • Irina I Serysheva

National Institutes of Health (P41GM103832)

  • Wah Chiu

National Institutes of Health (R01GM079429)

  • Wah Chiu

Wellcome

  • Ben F Luisi

Human Frontier Science Program

  • Ben F Luisi

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Matthew L Baker and Grigore D Pintilie for their assistance with modeling and segmentation and to Richard Henderson for helpful discussions. We thank Martin Pos and Timothy Opperman for the reagent MBX3132 and for helpful advice. We thank Satoshi Matsumura for kindly providing the expression plasmid pAcBH. This work was supported by the American Heart Association (16RNT29720001), and Grants from National Institutes of Health (P41GM103832, R01GM079429, R01GM072804, S10OD016279), the Wellcome trust, the Human Frontiers Science Program and MRC Mitochondrial Biology Unit (Grant number: U105663141). CFH is supported by a pre-doctoral training fellowship from the Keck Center of the Gulf Coast Consortia, on the NLM Training Program in Biomedical Informatics (Grant No. T15LM007093). We acknowledge the computing resources provided by the Center for Computational and Integrative Biomedical Research of Baylor College of Medicine and the Texas Advanced Computing Center at the University of Texas at Austin funded by the National Science Foundation (NSF) through grant ACI-1134872.

Reviewing Editor

  1. Olga Boudker, Weill Cornell Medical College, United States

Publication history

  1. Received: January 5, 2017
  2. Accepted: March 14, 2017
  3. Accepted Manuscript published: March 29, 2017 (version 1)
  4. Version of Record published: April 25, 2017 (version 2)

Copyright

© 2017, Wang et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 10,515
    Page views
  • 1,665
    Downloads
  • 149
    Citations

Article citation count generated by polling the highest count across the following sources: Scopus, Crossref, PubMed Central.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Open citations (links to open the citations from this article in various online reference manager services)

Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)

  1. Zhao Wang
  2. Guizhen Fan
  3. Corey F Hryc
  4. James N Blaza
  5. Irina I Serysheva
  6. Michael F Schmid
  7. Wah Chiu
  8. Ben F Luisi
  9. Dijun Du
(2017)
An allosteric transport mechanism for the AcrAB-TolC multidrug efflux pump
eLife 6:e24905.
https://doi.org/10.7554/eLife.24905

Categories and tags

Research organism

Further reading

    1. Structural Biology and Molecular Biophysics

    Ion permeation pathway within the internal pore of P2X receptor channels

    Stephanie W Tam, Kate Huffer ... Kenton J Swartz
    Research Article

    P2X receptor channels are trimeric ATP-activated ion channels expressed in neuronal and non-neuronal cells that are attractive therapeutic targets for human disorders. Seven subtypes of P2X receptor channels have been identified in mammals that can form both homomeric and heteromeric channels. P2X1–4 and P2X7 receptor channels are cation-selective, whereas P2X5 has been reported to have both cation and anion permeability. P2X receptor channel structures reveal that each subunit is comprised of two transmembrane helices, with both N-and C-termini on the intracellular side of the membrane and a large extracellular domain that contains the ATP binding sites at subunit interfaces. Recent structures of ATP-bound P2X receptors with the activation gate open reveal the unanticipated presence of a cytoplasmic cap over the central ion permeation pathway, leaving lateral fenestrations that may be largely buried within the membrane as potential pathways for ions to permeate the intracellular end of the pore. In the present study, we identify a critical residue within the intracellular lateral fenestrations that is readily accessible to thiol-reactive compounds from both sides of the membrane and where substitutions influence the relative permeability of the channel to cations and anions. Taken together, our results demonstrate that ions can enter or exit the internal pore through lateral fenestrations that play a critical role in determining the ion selectivity of P2X receptor channels.

    1. Structural Biology and Molecular Biophysics

    Peptides that Mimic RS repeats modulate phase separation of SRSF1, revealing a reliance on combined stacking and electrostatic interactions

    Talia Fargason, Naiduwadura Ivon Upekala De Silva ... Jun Zhang
    Research Article Updated

    Phase separation plays crucial roles in both sustaining cellular function and perpetuating disease states. Despite extensive studies, our understanding of this process is hindered by low solubility of phase-separating proteins. One example of this is found in SR and SR-related proteins. These proteins are characterized by domains rich in arginine and serine (RS domains), which are essential to alternative splicing and in vivo phase separation. However, they are also responsible for a low solubility that has made these proteins difficult to study for decades. Here, we solubilize the founding member of the SR family, SRSF1, by introducing a peptide mimicking RS repeats as a co-solute. We find that this RS-mimic peptide forms interactions similar to those of the protein’s RS domain. Both interact with a combination of surface-exposed aromatic residues and acidic residues on SRSF1’s RNA Recognition Motifs (RRMs) through electrostatic and cation-pi interactions. Analysis of RRM domains from human SR proteins indicates that these sites are conserved across the protein family. In addition to opening an avenue to previously unavailable proteins, our work provides insight into how SR proteins phase separate and participate in nuclear speckles.

    1. Biochemistry and Chemical Biology
    2. Structural Biology and Molecular Biophysics

    Concurrent remodeling of nucleolar 60S subunit precursors by the Rea1 ATPase and Spb4 RNA helicase

    Valentin Mitterer, Matthias Thoms ... Roland Beckmann
    Research Article

    Biogenesis intermediates of nucleolar ribosomal 60S precursor particles undergo a number of structural maturation steps before they transit to the nucleoplasm and are finally exported into the cytoplasm. The AAA+-ATPase Rea1 participates in the nucleolar exit by releasing the Ytm1-Erb1 heterodimer from the evolving pre-60S particle. Here, we show that the DEAD-box RNA helicase Spb4 with its interacting partner Rrp17 is further integrated into this maturation event. Spb4 binds to a specific class of late nucleolar pre-60S intermediates, whose cryo-EM structure revealed how its helicase activity facilitates melting and restructuring of 25S rRNA helices H62 and H63/H63a prior to Ytm1-Erb1 release. In vitro maturation of such Spb4-enriched pre-60S particles, incubated with purified Rea1 and its associated pentameric Rix1-complex in the presence of ATP, combined with cryo-EM analysis depicted the details of the Rea1-dependent large-scale pre-ribosomal remodelling. Our structural insights unveil how the Rea1 ATPase and Spb4 helicase remodel late nucleolar pre-60S particles by rRNA restructuring and dismantling of a network of several ribosomal assembly factors.