Switch-Loop Flexibility Affects Transport of Large Drugs by the Promiscuous AcrB Multidrug Efflux Transporter

Abstract

Multidrug efflux transporters recognize a variety of structurally unrelated compounds for which the molecular basis is poorly understood. For the resistance nodulation and cell division (RND) inner membrane component AcrB of the AcrAB-TolC multidrug efflux system from Escherichia coli, drug binding occurs at the access and deep binding pockets. These two binding areas are separated by an 11-amino-acid-residue-containing switch loop whose conformational flexibility is speculated to be essential for drug binding and transport. A G616N substitution in the switch loop has a distinct and local effect on the orientation of the loop and on the ability to transport larger drugs. Here, we report a distinct phenotypical pattern of drug recognition and transport for the G616N variant, indicating that drug substrates with minimal projection areas of >70 Ų are less well transported than other substrates.

INTRODUCTION

The Escherichia coli AcrAB-TolC tripartite multidrug efflux system confers resistance against a wide variety of drug compounds, including antibiotics, dyes, organic solvents, and detergents (1–3). The pump utilizes the proton motive force in an antiport fashion to energize the extrusion of compounds from the periplasm of the Gram-negative cell to the extracellular environment. Drug transport and energy transduction are catalyzed by the homotrimeric inner membrane component AcrB, belonging to the large superfamily of resistance nodulation and cell division (RND) transporters. AcrB contains two large periplasmic loops between transmembrane helices 1 and 2 plus 7 and 8. The protomers within the asymmetric trimeric setup adopt three different states, designated loose (L), tight (T), and open (O) (Fig. 1) (4–6). Recent crystallographic analysis of AcrB in complex with antibiotic molecules indicated that large drug molecules bind to the periplasmic access pocket (AP) of the protomer in the L state (7, 8). These and earlier studies (4) provided further insight into drug binding into the periplasmic deep binding pocket (DBP) in the protomer adopting the T state. Since only a few structures of AcrB in complex with antibiotics (minocyclin, doxorubicin, rifampin, erythromycin, and a pyridopyrimidine derivative inhibitor [4, 7–9]) have yet been obtained, it remains elusive whether the AP and DBP can bind the same drugs. Nakashima et al. (8) showed an AcrB trimer structure with rifampin bound in the L monomer in addition to binding of minocycline in the T monomer, and Eicher et al. (7) showed a trimer structure with doxorubicin bound to both the DBP and AP, with the latter occupied by a doxorubicin dimer. The question therefore arises whether these pockets represent sequential binding sites within the rotational transport cycle and if both pockets bind the same set of drugs. Both studies (7, 8) also indicated that a flexible switch loop of 11 amino acid residues residing between the AP and DBP might play an essential role in drug transport during the conformational cycling of the protomers through the L, T, and O states. Comparison of all asymmetric wild-type AcrB structures shows that the switch loop in the L protomer can adopt multiple alternative conformations, but these are very distinct from the switch-loop conformations observed in T and O protomers (see Fig. S1 in the supplemental material).

FIG 1.

Substrate binding in the deep binding pocket is observed exclusively in the AcrB T protomer. (A) View of the porter domain with the indicated subdomains PN1, PN2, PC1, and PC2 of the AcrB trimer from the membrane plane (PDB entry 4DX7 [AcrB/doxorubicin cocrystal structure]). The three monomers L (blue), T (yellow), and O (red) represent conformational states along a functional rotation cycle. In the L conformation (blue), a doxorubicin dimer (green spheres) is depicted in the access pocket. In the T conformation (yellow), a doxorubicin molecule (green spheres) is located in the deep binding pocket. The switch loop (including the side chains) is indicated in magenta. (B) The PN2/PC1 subdomains of the T protomer (represented as a yellow cartoon) have a larger distance from each other than the distance of the subdomains in the L protomer (represented as a blue cartoon). The distance between G290 (located on the PN2 subdomain) and I626 (located on the PC1 subdomain), at 15.2 Å (orange line), is larger in the T protomer than the distance in the L protomer, at 11.7 Å (blue line). The location of the deep binding pocket is represented by a minocycline molecule, shown in orange spheres (PDB entry 4DX5 [AcrB/minocycline cocrystals structure]). (C) Side view of the AcrB porter domain of the L protomer in a light-blue cartoon representation. The AP with bound doxorubicin dimer (blue spheres) is separated from the superimposed doxorubicin molecule in the DBP in the T protomer (yellow spheres) by the switch loop (in red) (PDB entry 4DX7 [AcrB/doxorubicin cocrystal structure]). All pictures were made using the software program PyMol (Schrödinger LLC).

Structural analysis of a variant of AcrB harboring a Gly616-to-Asn substitution (AcrB_G616N) within the switch loop indicated a conformation of this modified switch loop in the L conformer to be akin to the wild-type T-conformer switch-loop conformation (7). Moreover, the structure of MexB (10) revealed that its L-switch loop (also containing Asn at the homologous position 616) likewise adopts a T-like conformation, which is very similar to the T-like L-switch-loop conformation seen in the AcrB_G616N structure (see Fig. S1 in the supplemental material). This structural correlation is also reflected in the substrate specificity of the G616N AcrB variant and MexB. Both RND pumps confer lower resistance against erythromycin and other macrolides but show resistance profiles similar to those of wild-type AcrB for other drugs tested (11).

Structurally, the switch loop provides significant interaction with drugs bound to either the AP or the DBP (7, 8). It has been postulated that alteration of the switch loop (like the G616N substitution) restricts its conformational freedom, therefore limiting entrance to the drug binding sites. This implies the existence of a mechanism regulating the selective occupation of the AP in the L monomer and the DBP in the T monomer. Recognition and dwelling time of the drug inside the L-conformer AP (12) are possible parameters for initial distinction between substrates and nonsubstrates. The anticipated substrate transport from the AP to the DBP is most likely dependent on the flexibility of the switch loop and the volume of the DBP in the T or L monomer. Apparently, the volume of the DBP is sufficiently large in its opened T state to allow binding of substrate molecules, whereas the reduced volume in the collapsed DBP (i.e., closed state) in the L (and O) conformation prevents binding of substrate molecules to this site.

The DBP is defined by the PN2 and PC1 subdomains (PN2/PC1) (Fig. 1). These two subdomains seem to significantly rearrange their conformation relative to each other during functional rotation, where the PN2/PC1 distance in the T protomer is much larger than the distance observed in the L protomer (13, 14). The substantial rearrangements are reflected by a high root mean square deviation (RMSD) (≥2.0 Å) of the PN2/PC1 subdomains in the L, T, and O conformations. In contrast, the PN1 and PC2 subdomains (Fig. 1) appear to move substantially but remain a rigid body throughout the conformational cycle (RMSD < 0.5 Å). The opening of the DBP in the T protomer is therefore accomplished by widening of the PN2/PC1 distance and renders the DBP accessible for substrate molecules (Fig. 1).

In this work, we extended the spectra of drugs tested for the phenotypical differences of wild-type AcrB and the G616N variant via MIC assays, 12-h growth profile determination, plate dilution tests, and dye accumulation assays. Whereas physicochemical parameters of the tested compounds, like charge or hydrophobicity, could not be correlated to the observed phenotype, the minimal projection areas (MPAs) and to some extend the molecular masses of the tested drugs clearly showed a correlation to the diminished capability of the G616N variant to confer drug resistance or dye accumulation prevention.

MATERIALS AND METHODS

Strains and growth conditions.

Functional data were produced either using E. coli BW25113ΔacrB cells complemented with pET24acrB, encoding wild-type AcrB (15), complemented with pET24acrB_G616N, encoding a G616N variant (7), or complemented with pET24acrB_D407N, encoding a nonfunctional D407N variant (16), or E. coli 3-AG100 (parental gyrA marR strain, overproducing AcrAB-TolC) and its acrB::rpsL-neo and acrB_G616N (acrB encoding the G616N substitution) genomic variants (11). AcrB, AcrB with the D407N substitution (AcrB_D407N), and the AcrB_G616N variants were equally well expressed from plasmids (7). Cells were plated on LB agar plates with kanamycin (50 μg ml⁻¹) (LB-K50) or without kanamycin in the case of E. coli 3-AG100, grown overnight at 37°C, and stored at 4°C for at least 3 days before usage. Overnight cultures were grown from a single colony for 12 to 16 h (3 ml in a Falcon tube [15-ml volume] at 37°C and 180 rpm). For MIC determination and growth curve analysis, overnight-grown cultures were diluted to an optical density at 600 nm (OD₆₀₀) of approximately 0.018 and kept on ice before inoculation.

Determination of MICs.

MIC assays were carried out in 96-well plates (BRANDplates pureGrade S, no. 781662). A Hamilton Microlab Starlet pipetting robot was used to prepare 100 μl of serial 2-fold substrate dilutions in LB or LB-K50. Ninety-six-well plates were inoculated with 50 μl of culture dilutions (see above; from overnight-grown cultures diluted to an OD₆₀₀ of 0.018, with a final OD₆₀₀ of approximately 0.006). MIC values were determined in different ways for the two strains in use. Cultures were grown in the case of E. coli BW25113ΔacrB for 24 h at 37°C and 300 rpm (Infors HAT Multitron). E. coli 3-AG100 strains were cultured at 37°C and at 1,000 rpm for 20 h (Infors HAT Multitron). In both cases, final net absorption values (net OD₆₀₀) were determined by measuring the OD₆₀₀ at the appropriate time point and subtracting the OD₆₀₀ value at t = 0. For the E. coli BW25113ΔacrB strains, net OD₆₀₀ values were averaged for each condition. MIC values correspond to the substrate concentration, resulting in a net OD₆₀₀ value of <0.18. Net OD₆₀₀ was determined with 4- to 10-fold redundancy for each substrate. For E. coli 3-AG100 strains, MIC values were assigned to the substrate concentration, resulting in a net OD₆₀₀ value of <0.18 for each single experiment. Final MIC values are given as modal values out of 8 replications. Internal controls (wild type [WT] and D407N variant) were used as quality standards. If the internal controls differed by more than one dilution step in comparison to the average (this was only the case for single plate measurements with azithromycin, ciprofloxacin, and bleomycin), the entirety of measurements on that plate were not taken into account.

Growth curve assays.

For growth curve assays, inoculation into LB-K50 started from an OD₆₀₀ of 0.006 (1:3 dilution of the overnight-diluted culture at an OD₆₀₀ of 0.18; see above) in the presence of substrates at 37°C, shaking (30 s linear, 5-mm amplitude), and measurement of the OD₆₀₀ every 20 min (TecanReader Infinite M200) for 12 h. OD₆₀₀ values at 12 h were used for further data analysis. Relative growth values for E. coli BW25113ΔacrB/pET24acrB_G616N (G616N_RG) were calculated by normalizing the final OD₆₀₀ values (G616N_fOD) against the signal difference between E. coli BW25113ΔacrB/pET24acrB (WT_fOD) and E. coli BW25113ΔacrB/pET24acrB-D407N (D407N_fOD): G616N_RG = (G616N_fOD − D407N_fOD)/(WT_fOD − D407N_fOD).

The following substrate concentrations were used: acriflavine (16 μg/ml), azithromycin (1 μg/ml), berberine (256 μg/ml), chloramphenicol (1 μg/ml), ciprofloxacin (0.0025 μg/ml), dicloxacillin (64 μg/ml), erythromycin (8 μg/ml), fusidic acid (16 μg/ml), Hoechst 33342 stain (0.5 μg/ml), linezolid (16 μg/ml), novobiocin (1.5 μg/ml), rifampin (4 μg/ml), SDS (128 μg/ml), and spiramycin (32 μg/ml). Each data set consisted of 5 to 9 single measurements repeated independently on 3 to 7 days. Three single measurements (2 out of 7 experiments with erythromycin and 1 out of 6 measurements with fusidic acid) displayed standard deviations from the average of more than 30% and were excluded from further calculations.

Plate dilution assay.

For this solid medium-based growth method, LB agar plates were supplemented with substrates to be tested at a concentration that affected growth of both the wild type and the G616N variant. E. coli 3-AG100 cultures grown overnight were adjusted to an OD₆₀₀ of 1. For each variant, droplets (5 μl) with an OD₆₀₀ of 1 and with 10-fold dilutions (up to 10⁻³) were placed on LB agar plates. Plates were analyzed using the GE ImageQuant LAS 4000 instrument after approximate 14 h of incubation at 37°C (see Fig. S3 in the supplemental material). Data shown in Table 1 for E. coli BW25113ΔacrB were taken from the work of Eicher et al. (7).

TABLE 1.

Phenotypic drug efflux and accumulation measurements^a

^aReduced phenotypes/activities mediated by AcrB_G616N compared to those mediated by wild-type AcrB are indicated in bold text. The table is sorted according to the minimal projection area (MPA) of the used compounds. Drugs that resulted in an overall reduced phenotype of the AcrB_G616N variant with two or more methods (MIC, plate dilution, growth curve, or accumulation) (highlighted by gray shading) clustered well in the range of MPA > 70 Ų, except for dicloxacillin. ND, not done.

^bData taken from the work of Eicher et al. (7).

Fluorescent dye accumulation assays.

In addition to growth-based methods, fluorescent dye accumulation was monitored via fluorescence of berberin (50 μM), Hoechst 33342 stain (5 μM), doxorubicin (26.8 μM), and rhodamine 6G (50 μM). Overnight cultures were diluted 1:100 and grown until OD₆₀₀ values reached 0.7 to 0.8 (50 ml LB, 50 μg/ml kanamycin, 250-ml flask, 37°C, 180 rpm), cooled down on ice for 15 min, washed in 1.5 ml 50 mM KP_i (pH 7.5)–1 mM MgSO₄, centrifuged for 3 min at 4.000 × g at 4°C, and adjusted to an OD₆₀₀ value of 20 using ice-cold 50 mM KP_i (pH 7.5)–1 mM MgSO₄. Measurements were started by adding 20 μl of 10-fold-concentrated substrate solution to cells (180 μl) placed in a well of a 96-black-well plate (BRANDplates pureGrade S) and monitored over time at the experimentally determined maximal excitation (λ_ex) and emission (λ_em) wavelengths (TecanReader Infinite M200, excitation/emission bandwidth of 9/20 nm, 25 flashes). Accumulation of berberine (λ_ex, 365 nm/λ_em, 540 nm) and Hoechst 33342 stain (λ_ex, 365 nm/λ_em, 450 nm) were measured with cells at an OD₆₀₀ value of 2 in the presence of 0.2% d-glucose for 30 min at 28°C. Experiments with rhodamine 6G (λ_ex, 480 nm/λ_em, 558 nm) and doxorubicin (λ_ex, 480 nm/λ_em, 600 nm) were conducted with cells at an OD₆₀₀ value of 20 for 30 min or 50 min, respectively. For further data analysis, fluorescence intensity values (usually every 5 s) of the G616N variant (G616N_fl) were double normalized against fluorescence intensity values of the WT (WT_fl) and D407N variant (D407N_fl) between t = 1,600 s and t = 1,800 s for berberine, rhodamine 6G, and Hoechst33342 stain or between t = 2,000 s and t = 3,000 s for doxorubicin using the following equation, relative activity of E. coli BW25113ΔacrB/pET24acrB-G616N: G616N_RA = (G616N_fl − D407N_fl)/(WT_fl − D407N_fl).

Deviations for G616N_RA for each time point were usually less than 0.05 relative units and were not considered for calculation of the standard deviations for the final relative activities. Relative activities of E. coli BW25113ΔacrB/pET24acrB_G616N (G616N_RA) were averaged over all single measurements (n = 4 to 7 for each substrate).

Computational analysis.

All physicochemical parameters of the substrates were calculated using the web site at http://www.chemicalize.org/ (ChemAxon, Budapest, Hungary). Superimposition of the switch loops (residues 613 to 623) of all published asymmetric AcrB crystal structures and root mean square deviation (RMSD) values were calculated using the software program SUPERPOSE (17) from the CCP4 program suite (18) using PDB entry 4DX5 (AcrB in complex with minocycline, 1.9-Å resolution) as the template. Visualization of the (superimposed) structures was done using the PyMOL program (The PyMOL Molecular Graphics System, version 1.5.0.3; Schrödinger, LLC).

RESULTS AND DISCUSSION

We examined the drug susceptibility of E. coli BW25113ΔacrB complemented with plasmid-borne wild type or AcrB with the G616N substitution and the E. coli 3-AG100 parental strain (gyrA marR), its acrB_G616N chromosomal variant, and the acrB deletion strain E. coli 3-AG100 acrB::rpsL-neo (11). The combined analysis is summarized in Table 1 and comprises MIC assays (see Table S1 in the supplemental material), 12-h growth curve analysis on LB medium (E. coli BW25113ΔacrB complemented strains only) (Fig. 2; see also Fig. S2), and LB agar plate dilution tests with different antibiotics (see Fig. S3) (7). In addition, we measured efflux pump activity in whole cells by detection of the rate and final accumulation of dye substrate using fluorescence spectroscopy (E. coli BW25113ΔacrB complemented strains only) (Fig. 3). In addition to the known susceptibility of the G616N variant toward macrolides (11), MIC values for dicloxacillin, rhodamine 6G, and fusidic acid were decreased 2-fold in the G616N variant for both the E. coli BW25113ΔacrB and E. coli 3-AG100 test systems (Table 1). For acriflavin, SDS, ciprofloxacin, ethidium, Hoechst 33342 stain, oxacillin, and novobiocin, no MIC value differences were detected. For linezolid, chloramphenicol, tetraphenylphosphonium ion (TPP⁺), rifampin, and bleomycin, a 2-fold reduction in the MIC value was detected for either G616N variant (4 to 8 independent measurements) in E. coli BW25113ΔacrB or E. coli 3-AG100 (Table 1; see also Table S1). As previously reported (19), we also observed that AcrB-mediated resistance against carbenicillin is not well pronounced (not shown). The MIC value determination was done 4 to 10 times (indicated in Table S1) for each substrate, providing redundant measurements to distinguish between 2-fold differences found for growth on the relevant substrates.

FIG 2.

Correlation between drug minimal projection areas (MPAs) and the relative growth difference between AcrB wild type-producing E. coli and G616N variant-producing E. coli BW25113ΔacrB. (A) Relative growth (arbitrary units, au) of E. coli BW25113ΔacrB harboring the G616N mutant (black bars) compared to E. coli BW25113ΔacrB harboring acrB (see Materials and Methods) in the presence of different antibiotic substrates. Correlation of substrate minimal projection area (MPA) (given in Ų) is indicated by gray bars. The distinction between affected and nonaffected phenotypes was arbitrarily set to 0.8 relative growth units and corresponds to a critical MPA value of ∼70 Ų.

FIG 3.

Effect of the G616N substitution on the efflux of fluorescent dyes. (Upper panel) Accumulation of berberine and Hoechst 33342 stain in E. coli BW25113ΔacrB producing wild-type AcrB (gray lines), proton translocation-deficient AcrB_D407N (gray dotted lines), or the AcrB_G616N variant (black dashed lines) was monitored by assessing fluorescence (see Material and Methods). Accumulation of berberine or Hoechst 33342 stain in the cells results in fluorescence enhancement, whereas accumulation of doxorubicin or rhodamine 6G results in fluorescence quenching. Active efflux by a functional AcrB pump results in a reduced uptake rate and final accumulation of fluorescent dye compared to results for the inactive pump mutant (D407N). (Lower panel) Correlation between minimal projection area of the fluorescent dyes and their relative transport activities between the AcrB wild type and the G616N variant. Error bars indicate the standard deviations. The correlation between activity and substrate minimal projection area (MPA) (in Ų), is indicated by the black and gray bars. For details, see Material and Methods.

For the E. coli 3-AG100 strains, we observed clear growth inhibition on LB agar plates in a plate dilution experiment for bleomycin, erythromycin, rifampin, TPP⁺, oxacillin, and ethidium. The results are in line with observations for previous plate dilution assays, where growth on erythromycin and doxorubicin was most affected (7). From the MIC data and plate dilution tests presented here, it appears that cells producing the AcrB_G616N variant are more susceptible to large-molecular-mass substrate molecules. Due to the subtle phenotypic differences caused by the G616N substitution, exceptions are present in both assays and in both E. coli systems, despite redundant measurements. Exceptions can be found in the case of linezolid, chloramphenicol, rifampin, and bleomycin MIC values in the E. coli BW25113ΔacrB system and for ethidium in the E. coli 3-AG100 strain system (Table 1). Susceptibility to dicloxacillin, a compound containing two halogen atom substitutions, was reduced 2-fold in both systems and also showed reduced growth in the 12-h growth analysis (Table 1). Nevertheless, from Table 1 and Table S1 in the supplemental material, it appears that a correlation between the minimal projection areas (MPAs) of the compounds used and the reduction in the MIC value of the E. coli BW25113ΔacrB and E. coli 3-AG100 strains expressing the G616N variant exists. A clear exception is posed by the double-halogen-atom-substituted compound dicloxacillin, independent of MPA.

Since the MIC assay measures the OD₆₀₀ values after 20 to 24 h of growth on LB medium supplemented with the drug compounds, we anticipated for some of the antibiotics, detergents, or dyes pleiotropic effects that might obscure the true transport efficiency of WT AcrB or the G616N variants. Therefore, we analyzed the 12-h growth rate inhibition by monitoring the growth of cells in the presence of drugs in a time-resolved manner over 12 h. As summarized in Fig. 2 (see also Fig. S2 in the supplemental material), all macrolides but also rifampin, fusidic acid, and dicloxacillin caused inhibition of growth, whereas all other substrates, including TPP⁺, linezolid, and chloramphenicol, did not affect the growth of cells containing the G616N variant. For rhodamine 6G, solubility problems limited the reproducible determination of the MIC value (but see below). Control growth experiments in liquid LB (containing 50 μg ml⁻¹ kanamycin) without the addition of substrates confirmed that mutations in acrB did not significantly influence growth rates and final cell densities in the absence of added antibiotics (see Fig. S2).

For the fluorescent dyes berberine, doxorubicin, Hoechst 33342, and rhodamine 6G, we conducted accumulation assays in a whole-cell context on a time scale of approximately 30 min to trace the efflux activity of the AcrB WT or the D407N (as a control) or G616N variant (Fig. 3). Accumulation of doxorubicin or rhodamine 6G inside cells results in quenching of the fluorescence signal, whereas the fluorescence intensities of Hoechst 33342 or berberine increase upon accumulation inside cells. For all tested substrates, a clear difference in dye accumulation between cells containing wild-type AcrB and those containing the inactive AcrB D407N variant was measured, indicating that the efflux activity mediated by AcrB strongly counteracts the accumulation of all tested dyes (Fig. 3). In accordance with the plate dilution assays described by Eicher et al. (7), a decrease in doxorubicin efflux activity was observed for the G616N variant compared to that for wild-type AcrB. A similar decrease in G616N efflux activity was observed for rhodamine 6G. The accumulation of Hoechst 33342 stain was, within the limit of the standard deviations, also slightly affected in the G616N variant clone. In the case of berberine, however, no difference in dye accumulation between WT and G616N variant clones was observed. By comparing the fluorescence accumulation experiments for the tested dyes, we observed a tendency of reduced efflux activities for the G616N variant clones in relation to their calculated minimal projection areas (Fig. 3).

Table 1 summarizes for each drug the phenotypical difference between the WT and G616N variants for at least two of the four methods used (MIC determination, plate dilution, growth curve, and dye accumulation). As indicated, a clear tendency of transport deficiency of the G616N variant toward drugs with high MPA values is observed. Moreover, a clear exception is observed for dicloxacillin, a 2-fold chlorine-atom-substituted compound. For other halogenated compounds, linezolid and chloramphenicol, an effect could be observed for the MIC values (E. coli BW25113ΔacrB system only) but not in the 12 h-growth curve analysis assay. For rifampin, a high-MPA compound, the MIC determination did not reveal a difference for the E. coli BW25113ΔacrB system but reduced susceptibility was observed for E. coli 3-AG100_G616N; moreover, 12-h growth analysis (Fig. 2) and plate dilution tests see (Fig. S3 in the supplemental material) were unequivocal. For the high-molecular-mass and high-MPA compound bleomycin, the largest compound thus far known to be exported by the AcrAB-TolC system (see Table S2), no indication of growth reduction was observed for the E. coli BW25113ΔacrB_G616N variant; however, the MIC assay and plate dilution tests with the E. coli 3-AG100_G616N variant indicated a clear susceptibility compared to findings for the wild type.

Overall, the G616N amino acid substitution causes a reduction in conferred resistances for substrates with MPAs of <70 Ų (Fig. 2 and Table 1), with the notable exception of dicloxacillin. The difference in resistance between E. coli cells containing wild-type AcrB and those containing the G616N variant for this doubly halogenated substrate may indicate that additional physicochemical parameters besides the geometrical extent are likely to be important for substrate recognition and efflux by AcrB.

The G616N-induced rigidity of the switch loop in the L protomer to adopt a T-like conformation (see Fig. S1 in the supplemental material) might hinder passage of large molecules (i.e., compounds with MPAs of more than ∼70 Å) and therefore might indicate that the geometrical flexibility of this loop determines substrate specificity and substrate efflux.