Kinetic Behavior of the Major Multidrug Efflux Pump Acrb of Escherichia Coli

Kinetic behavior of the major multidrug efflux pump AcrB of Escherichia coli

Keiji Nagano1 and Hiroshi Nikaido2

Department of Molecular and Cell Biology, University of California, Berkeley, CA 94720

Communicated by Alexander N. Glazer, University of California System, Oakland, CA, February 13, 2009 (received for review November 26, 2008) Multidrug efflux transporters, especially those that belong to the With intact cells, it is difficult to determine the substrate resistance-nodulation-division (RND) family, often show very concentration in the periplasm, a location where most of the broad substrate specificity and play a major role both in the substrates are thought to be captured (1, 8–10). When ␤-lactams intrinsic antibiotic resistance and, with increased levels of expres- are used as substrates, however, their periplasmic concentrations sion, in the elevated resistance of Gram-negative bacteria. How- can be calculated from their hydrolysis rate by intact cells and ever, it has not been possible to determine the kinetic behavior of from the kinetic constants of periplasmic ␤-lactamases (11). these important pumps so far. This is partly because these pumps Using this approach, we determined the kinetic behavior of form a tripartite complex traversing both the cytoplasmic and several cephalosporins in AcrB-catalyzed efflux and indeed outer membranes, with an outer membrane channel and a obtained indications of positive cooperativity in the ligand– periplasmic adaptor protein, and it is uncertain if the behavior of pump interaction. an isolated component protein reflects that of the protein in this multiprotein complex. Here we use intact cells of Escherichia coli Results containing the intact multiprotein complex AcrB-AcrA-TolC, and Principle of the Assay. When a ␤-lactam is added to the external measure the kinetic constants for various cephalosporins, by as- medium at a given concentration Co (␮M), it diffuses sponta- sessing the periplasmic concentration of the drug from their rate of neously across the outer membrane into the periplasm (Fig. 1). ␤ hydrolysis by periplasmic -lactamase and the rate of efflux as the The influx Vin is given by Fick’s first law of diffusion difference between the influx rate and the hydrolysis rate. Nitro-

ϭ ͑ Ϫ ͒ MICROBIOLOGY cefin efflux showed a Km of about 5 ␮M with little sign of Vin P * A * Co Cp cooperativity. For other compounds (cephalothin, cefamandole, and cephaloridine) that showed lower affinity to the pump, how- where P, A, and Cp denote the permeability coefficient (cm/s) of 2 ever, kinetics showed strong positive cooperativity, which is con- the outer membrane, the surface area of bacterial cells (cm /mg ␮ sistent with the rotating catalysis model of this trimeric pump. For dry weight), and the periplasmic drug concentration ( Mor 3 2 the very hydrophilic cefazolin there was little sign of efflux. nmol/cm ). We used 132 cm /mg (12) for A. We measure the rate of hydrolysis by the periplasmic ␤-lactamase (Vh) spectropho- cephalosporin ͉ drug efflux pump ͉ multidrug resistance tometrically with suspensions of intact cells. We then solve the Michaelis-Menten equation for hydrolysis for Cp, using the values of Km (determined earlier) and Vmax (determined by using ultidrug efflux pumps of the resistance-nodulation- cell extracts). Mdivision (RND) family in Gram-negative bacteria can With nitrocefin, P was determined by coupling influx with the pump out a surprisingly wide range of antimicrobial compounds enzymatic hydrolysis in periplasm (11), with the active efflux (1, 2). For example, AcrB of Escherichia coli, which has been inactivated by deenergization of the cytoplasmic membrane with studied most extensively as a prototype of similar pumps, exports a proton conductor, 40 ␮M carbonyl cyanide m-chlorophenyl- a number of dyes, detergents, chloramphenicol, tetracyclines, hydrazone (CCCP). When this was done in the presence of, e.g., macrolides, novobiocin, fluoroquinolones, and organic solvents. ␮ ␤ 25 M of nitrocefin, the rate of hydrolysis increased severalfold Importantly, it also pumps out -lactams, some of which cannot upon the addition of CCCP, giving us qualitative confirmation easily cross the cytoplasmic membrane and thus stay in the that proton-motive-force-dependent pump(s) were pumping out periplasm (3). However, in the 13 years since their discovery (4, much of the nitrocefin from the periplasm in nonpoisoned cells. 5), no success was achieved in the determination of kinetic The influx rate, Vin, can then be calculated from the values of behavior of these pumps, although the affinity of some substrates P, A, and C , as described above. In our model, the rate of efflux, could be assessed on a relative scale by their activity as com- p V , is the difference between the (calculated) V and the petitive inhibitors in a proteoliposome reconstitution assay (6). e in (measured) V (Fig. 1). Recently an attempt to measure the binding of dyes to purified h One possible complication is that some of the periplasmic AcrB was made by using fluorescence polarization (7); however, cephalosporin may cross the cytoplasmic membrane [V (dashed it is unclear if the experiments measured binding to active sites c arrow) in Fig. 1]. Such diffusion is known to occur with relatively or to generally hydrophobic pockets in the protein. lipophilic cephalosporins such as cephalothin (3, 13). The first- The knowledge of kinetic constants is essential in our attempts order rate constant for this influx into cytosol can be calculated to understand the contribution of the pumps to drug resistance (14) from the half-equilibration time determined earlier (13). in a quantitative manner. Furthermore, recent crystallographic Ϫ For cephalothin, this is 2.9 ϫ 10 9 cm3/s/mg cells. When C ϭ 10 studies of AcrB (8–10) suggest that each protomer in this p trimeric transporter undergoes a succession of conformational changes that are dependent on the conformations of the neigh- Author contributions: H.N. designed research; K.N. performed research; K.N. and H.N. bors, that is, a functionally rotating mechanism. This mechanism analyzed data; and H.N. wrote the paper. predicts that positive cooperativity may exist in the export of The authors declare no conflict of interest. drugs by AcrB. In this study we measured the kinetic constants 1Present address: Department of Microbiology, Aichi-gakuin University School of Dentistry, of AcrB. We used intact cells, because AcrB, which exists as a Nagoya 464-8650, Japan. tripartite complex with the periplasmic protein AcrA and the 2To whom correspondence should be addressed. E-mail: [email protected]. outer membrane channel TolC (1, 2), might be altered in its This article contains supporting information online at www.pnas.org/cgi/content/full/ behavior when it is separated from these physiological partners. 0901695106/DCSupplemental.

www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901695106 PNAS Early Edition ͉ 1of5 Downloaded by guest on September 30, 2021 Co 0.04

V 0.03 0.05 in V (mM) Cp Cp e 0.04

β- Tol C- Cp V lactamase , Cp h AcrA- OM e 0.03 V , V h h 0.02 Ve Cp AcrB V 0.02 0.01 Ve

0 04080120Co (nmol/mg/s); e , V

IM h 0.01

V V c Vh

Fig. 1. Principle of the assay. A ␤-lactam compound is added to the external medium bathing the intact cells of E. coli at a concentration, Co. At steady state, the drug crosses the outer membrane at the net rate Vin, which, 0 Ϫ 0 50 100 150 200 according to Fick’s first law of diffusion, is equal to P * A *(Co Cp) (see text). Co This influx is balanced by the sum of 3 processes: (i) hydrolysis by the periplas- (µM) ␤ mic -lactamase at the rate Vh,(ii) periplasmic capture followed by efflux into Fig. 2. Changes of various processes at different external concentrations of the external medium through AcrAB-TolC complex at the rate Ve, and (iii) slow nitrocefin. Results from one typical experiment with strain HN1157 are shown. diffusion across the cytoplasmic membrane into cytosol at the rate Vc. The Inset shows similar results from an experiment with HN1159 (a ⌬acrAB derivative of HN1157), where efflux is undetectable at least within the range of C shown. Ϫ5 p ␮M, Vc will then be 2.9 ϫ 10 nmol/s/mg. This is about 1,000 times smaller than the rates we are concerned with here, Ve, Vin, and Vh (see below), and we neglected Vc in our analysis. rate Vh and the efflux rate Ve, it was not a simple function of Co. In practice, however, this approach demands certain condi- Examination of Vh and Ve (obtained by subtracting Vh from Vin) ␤ tions. For example, the Km of the -lactamase must be reason- showed that the values of Ve had a shape suggesting some ably high, so that Vh becomes a sensitive indicator of Cp, in the saturation, but those of Vh did not in this concentration range of range of the concentrations we are interested in. Ideally, Ve, Vh, Cp, as expected from the large value (340 ␮M) of Km. Efflux and Vin should all be of the same order of magnitude so that high represented by Ve was dominant in the low concentration range precision can be attained in our calculation. The Vh values must for Co, but the enzymatic hydrolysis represented by Vh became be reasonably high to allow for precise measurement of hydrolysis. more important at high values of Co. As an important control, a similar analysis was carried out by Optimization of Assay Conditions for Nitrocefin. In our initial stud- using the isogenic strain HN1159, where the main efflux genes ␤ ies, we used nitrocefin (15) as the -lactam substrate, because the acrAB were absent. As expected, at least up to 40 ␮M Cp, the changes in absorption occur at 486 nm, where the light scattering efflux rate Ve was insignificant (Fig. 2, Inset). Furthermore, the by intact cells is smaller than in the UV range. Upon hydrolysis, addition of CCCP to deenergize the efflux pump had almost no OD486 changes more strongly (15) than the changes occurring at effect on the rate of nitrocefin hydrolysis Vh. ␤ 260 nm with other cephalosporins (16). As the -lactamase, we In 4 independent assays using HN1157, the Ve vs. Cp curves used the endogenous AmpC ␤-lactamase of E. coli. It has been followed the Michaelis-Menten kinetics (an example is shown in reported that this enzyme shows a complex behavior for nitro- Fig. 3), and curve fitting resulted in the V values of 0.0235 Ϯ ␮ max cefin, with 2 Km values at 2.4 and 360 M, the latter fraction 0.003 nmol/mg/s, which corresponds to the kcat of Ϸ10/s based on contributing about 95% of the total kcat value (17). However, in the cellular content of AcrB. Km was 4.95 Ϯ 1.14 ␮M. Data our hands, crude sonicates of E. coli showed no sign of the showed no sign of positive cooperativity. existence of the high-affinity component, and the kinetics could ␮ be explained completely by the presence of a single Km of 340 M Effect of Efflux and Hydrolysis on Nitrocefin Resistance. To examine (supporting information Fig. S1). whether the fluxes determined affect the resistance of intact cells When nitrocefin was added to intact cells of E. coli [of the AG100 series (18)], the hydrolysis was extremely slow because of the slow influx of this bulky compound through the porin 0.025 channel. Because reliable data cannot be obtained under such r = 0.988 conditions, we used a strain of E. coli producing a wider mutant 0.020 porin channel (19, 20). This strain, RAM121, allows an influx of larger maltodextrins and large antibiotics that are excluded by 0.015

the normal E. coli porins (19) and shows a larger single channel (nmol/mg/s) e

conductivity (21). Indeed, it allowed a nearly 10-fold faster influx V of nitrocefin, with the permeability coefficient of about 0.2 ϫ 0.010 10Ϫ5 cm/s. This strain was further improved by increasing the expression of AcrAB severalfold (22) by the deletion of acrR 0.005 repressor, resulting in the strain HN1157. Efflux Rate 0.000 Nitrocefin Fluxes in HN1157. When intact cells of HN1157 were 0 5 10 15 20 25 30 35 40 incubated with different concentrations of nitrocefin, results C (µM) summarized in Fig. 2 were obtained. The periplasmic concen- p

tration Cp increased following the increase in the external Fig. 3. Curve fitting of Ve vs. Cp data for 1 nitrocefin experiment with the concentration Co, but because Cp was affected by the hydrolysis Michaelis-Menten equation. r denotes the correlation coefficient.

2of5 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901695106 Nagano and Nikaido Downloaded by guest on September 30, 2021 Table 1. Nitrocefin MIC values in the derivatives of HN1157 0.8 O r = 0.966 V H e Estimated nitrocefin MIC S 0.7 S HN Strain (␮g/mL)* N O 0.6 O HN1157 18.8 O ⌬ HO O HN1157 acrAB 1.9 0.5 HN1157 ⌬ampC 16.7 HN1157 ⌬acrAB ⌬ampC 1.1 0.4 (nmol/mg/s)

*The values shown are the averages of 3 experiments. h 0.3 V , e

V 0.2 Vh to nitrocefin as expected, we determined minimal inhibitory 0.1 concentration (MIC) of nitrocefin in HN1157 and its isogenic derivatives with the gradient plate method (see Experimental 0 Procedures). MIC in HN1157 decreased strongly when efflux was 0 20 40 60 80 100 120 140 abolished by introduction of deletion of acrAB (Table 1), Cp (µM) whereas there was only a small effect on MIC when the hydrolysis was eliminated by the deletion of the ampC gene. These results Fig. 5. Vh and Ve for cephalothin with strain HN1160. Cells were harvested at OD of 2.0 and incubated with 25 to 400 ␮M cephalothin. are consistent with the behavior of various fluxes in the low 600 range of Cp, where inhibition of growth occurs. compound was close to 500, if the cellular AcrB content is Efflux of Other Cephalosporins. Fluxes of several other cephalo- assumed to be similar to that in HN1157. The ratio Vmax/K0.5 sporins were examined by using a similar approach. Unlike showed only a severalfold increase from that of nitrocefin. nitrocefin, these are all compounds that are useful for the Cephaloridine (Fig. 6) was of special interest to us, as it was treatment of Gram-negative infections, and as such they pene- shown not to traverse the cytoplasmic membrane and remain in trate through the outer membrane relatively easily (23). Thus we the periplasm (3, 13). In 4 experiments with strain LA51, the Hill used E. coli strains producing normal porins. In fact, in most Ϯ Ϯ

coefficient was 1.75 0.14. The Vmax was high (1.82 0.85 MICROBIOLOGY cases we tried to decrease Vin by reducing the expression of the nmol/mg/s), and the K was also the highest among the com- Experimental Procedures 0.5 large channel porin OmpF ( ). For most pounds examined, 288 Ϯ 54 ␮M. compounds we used HN1160, in which the AmpC ␤-lactamase Cefazolin, a cephalosporin with 2 hydrophilic substituents (a was replaced with the TEM enzyme that had higher K values m tetrazole and a thiadiazole), showed in contrast very little (23) and gave a more sensitive indication of the C values. p evidence for active efflux (Fig. 7). Because the hydrolysis had to be observed at 260 nm, where the scattering and absorption by the cells were high, and because the Discussion absorbance changes were smaller than with nitrocefin at 486 nm, These are the first kinetic constants determined for a RND-type the overall quality of data was significantly lower. Nevertheless, extensive active efflux was observed both with multidrug efflux pump in intact cells. However, we have earlier cefamandole (Fig. 4) and cephalothin (Fig. 5). It is noteworthy examined the function of reconstituted AcrB in the extrusion of that the V vs. C curves were both clearly sigmoidal. In 6 NBD-labeled phospholipids, and found that conjugated bile e p acids, such as taurocholate and glycocholate, acted as presum- (cefamandole) and 4 (cephalothin) experiments, curve-fitting ␮ gave Hill coefficients of 3.2 Ϯ 1.2 and 1.9 Ϯ 0.26, respectively. ably competitive inhibitors in the range of 10–20 M (6). The observed Km for nitrocefin, 5 ␮M, suggests a rather strong Another interesting feature was that the K0.5 values, which correspond to the substrate concentration giving the half- affinity, an observation that is not surprising in view of the maximal rates and were 19.6 Ϯ 1.8 for cefamandole and 91.2 Ϯ presence of 2 aromatic nuclei within the nitrocefin molecule and of our knowledge that AcrB prefers lipophilic solutes (1, 2) as 25.8 ␮M for cephalothin, respectively, were higher than the Km substrates. value for nitrocefin. The Vmax for efflux (0.37 Ϯ 0.2 and 1.0 Ϯ 0.3 nmol/mg/s for cefamandole and cephalothin, respectively) The recently determined structure of the asymmetric trimer of was much higher than for nitrocefin, and the kcat for the latter AcrB (8–10) suggests that each protomer with its unique con- formation represents each of the 3 successive stages of transport,

8 O r = 0.988 0.1 H 6 HN S 0.1 Ve OH 1.4 V N S N h 4 N 0.1 O NN 1.2 2 0.1 O OH V 1 e 0 0.1 r = 0.996 8 V 0.8 0.0 h (nmol/mg/s) (nmol/mg/s) h .06 0.6 h V 0 , V

e 4 .0 , 0.4

0 e V 2 V 0.0 0.2 .00 0 0 10 20 0 0 100 200 300 400

Cp (µM) Cp (µM)

Fig. 4. Vh and Ve for cefamandole with strain HN1160. Cells were harvested Fig. 6. Vh and Ve for cephaloridine with strain LA51. Cells were harvested at at OD600 of 3.2 and incubated with 45 to 380 ␮M cefamandole. OD600 of 1.1 and incubated with 50 to 700 ␮M cephaloridine.

Nagano and Nikaido PNAS Early Edition ͉ 3of5 Downloaded by guest on September 30, 2021 H H pumped out very strongly at Cp values exceeding 100 ␮M (Fig. 0.14 N S N 6). This was not noticed by the MIC assays, because at the very N V N O N S S h N low Cp values that kill the cell [about 0.5 ␮M (21)], efflux is 0.12 O N N negligible especially because of the positive cooperativity O OH 0.10 shown by this substrate. A similar situation was also found for cephalothin for which no MIC change was seen in the ⌬acrAB 0.08 mutant of E. coli (34), although in Salmonella where the hydrolysis process is totally absent we could see small alter- (nmol/mg/s)

e 0.06 ations of MIC (3). , V

h These considerations also suggest a modified strategy for 0.04 V finding antibacterial agents that remain active in the presence of 0.02 efflux pumps. Thus far, it was assumed that an efficacious agent Ve should totally avoid efflux. However, these findings tell us that 0 the Vmax of the efflux does not matter, as long as it occurs in the 050 100 150 200 range of Cp that far exceeds the concentration that inhibits or C p (µM) kills bacteria, and give us renewed hopes that such novel agents may be found in a not-too-distant future. Fig. 7. Vh and Ve for cefazolin with strain HN1160. Cells were harvested at ␮ OD600 of 3.0 and incubated with 25 to 400 M cefazolin. Experimental Procedures Bacterial Strains. For nitrocefin efflux, we used a strain producing a mutant i.e., open access, ligand binding, and ligand extrusion. Extrusion OmpC porin producing a large channel, RAM121 (19), a kind gift from Rajeev Misra, Arizona State University. To increase efflux we transduced ⌬acrR::kan thus is unlikely to occur unless the neighboring protomer binds mutation from AG100B (16) (K-12 argE3 thi-1 rpsL xyl mtl ⌬(gal-uvrB) supE44 the second ligand, and we are led to anticipate a positive acrR::kan), to obtain HN1157. HN1159 was made by transducing ⌬acrAB::spc cooperative kinetics in transport. Positive cooperativity was marker from AG100YB (35) into HN1157. For analyzing efflux of cefamandole, indeed reported for ATP hydrolysis in the dimeric MalK in the cephalothin, and cefazolin, we used HN1160, that is AG100B, whose ampC maltose transporter complex (24) and in the hexameric ATPase gene was replaced by the amp gene coding for TEM1 ␤-lactamase [supporting TrwB involved in DNA transfer in conjugation (25), although information (SI)]. Cephaloridine efflux was determined by using a strain with an increased production of AmpC ␤-lactamase, LA51 (23). kinetic analysis has been difficult with the classical F1 ATPase (26, 27). Although it was disappointing to find no evidence for The Optimized Nitrocefin Efflux Assay. HN1157 was grown in modified LB broth positive cooperativity for nitrocefin, there was ample evidence (1% tryptone, 0.5% yeast extract, 0.87% NaCl, and 5 mM MgSO4)at30°C for that for cephalothin, cefamandole, and cephaloridine. The overnight with shaking, and the culture was diluted 100-fold in the fresh reason for this difference is not known. However, in one (out of medium, and the cultivation was continued at 30 °C with shaking till the OD600 many) interpretation, we may assume that this difference is reached 0.65. Cells were harvested by centrifugation at room temperature, related to the affinity of the drugs to the transporter. In contrast and after washing twice in 50 mM potassium phosphate buffer, pH 7.0, to nitrocefin, the other cephalosporins had higher values of K , containing 5 mM MgCl2, were resuspended in the same buffer at the OD600 of 0.5 0.8 (corresponding to 0.24 mg dry weight/mL). Nitrocefin was added at a which reached almost 300 ␮M in cephaloridine. Interestingly, desired final concentration, and the OD486 was measured at 10, 20, and 30 min even with the F1 ATPase, alteration of the adenosine-binding afterward after incubation at 25 °C, with a UVIKON 860 spectrophotometer. pocket (28) by site-directed mutagenesis not only reduced the Cp was calculated from Vh and the kinetic constants of the ␤-lactamase ϭ ␮ affinity of the substrate to the enzyme, but also created a clearly (Km 340 M; Vmax was determined by using the sonicated extract of cells). Vin detectable positive cooperativity (29). We also note that posi- and Ve were determined as described in Results. The kinetic constants were tively cooperative kinetics was reported in an early reconstitu- derived by curve fitting of the Ve vs. Cp data with the program CurveExpert, version 1.37 (http://curveexpert.webhop.com) by using the Michaelis-Menten tion study of an RND pump, CzcA transporter, which catalyzes max c c equation modified for cooperativity, Ve ϭ Ve *(Cp) /(K ϩ (Cp) ), where c is the export of toxic divalent metals (30), although the curves were the Hill coefficient. drawn on only 4 points and the K0.5 value reported was 6.6 mM, a concentration several orders of magnitude higher than that AcrB Content in HN1157 Cells. This content was determined by Western blot expected for the exclusion of toxic metals. analysis (35), by using a standard curve constructed by using known amounts The turnover number of the AcrB was rather small for of purified AcrB protein. nitrocefin, but it was about 2 orders of magnitude higher for cephalosporins like cephalothin and cephaloridine. One of us Determination of Fluxes of Cephalosporins Other than Nitrocefin. These com- pounds permeated through the porin channel rapidly (see ref. 23). Thus it was argued earlier that AcrB must be a slow pump, because AcrB often necessary to decrease the expression of the wider OmpF porin to needs to pump out only a small number of agents that trickled decrease Vin so that it will be similar to Ve. We thus used cells in early stationary in through the outer membrane barrier (31). But more recently phase (36). With cefazolin, OmpF was further repressed by adding 5 mM AcrB was found to confer resistance to solvents (1), which are sodium salicylate to the medium (37). Cells were grown at 30 °C (except for likely to traverse the outer membrane rapidly. Thus for these cephaloridine) because this appeared to improve the expression of the TEM ␤-lactamase in strain HN1160. compounds, a high Vmax would be needed, and we have seen that for some cephalosporins turnover numbers close to 1,000 are Hydrolysis by intact cells was followed by the decrease of optical density at 260 nm, using the end-on position of a UVIKON 860 spectrophotometer to possible. minimize the light scattering. Cells with 4 mm (for the cefamandole, ceph- Up to now, the usual criterion for active drug efflux in bacteria alothin, and cefazolin) or with 1 mm light path (for cephaloridine) were used was the alteration of MIC values upon the overexpression or to accommodate high concentrations of substrates and cells. Cp values were deletion of the pump genes (for example, see refs. 1, 5, 13, 18, calculated by using the Km values for the TEM and AmpC enzymes previously 32, 33). One of the important conclusions from this study was the determined (23). Raw data that produced the plots of Fig. 4 are shown in SI as realization that this criterion often misses the existence of a Table 1. The permeability coefficient P was obtained from CCCP-poisoned cells (see powerful efflux activity. For example, it has been believed that above) for cefazolin and cefamandole. For cephalothin and cephaloridine, P cephaloridine is not a substrate for the AcrAB efflux pump was obtained from the hydrolysis rates at the low substrate concentrations because the deletion of the acrAB genes had little effect on its where efflux made a negligible contribution (Figs. 4–6). Simulation with the MIC (3, 34). However, this study showed that cephaloridine is cefamandole data showed that increasing or decreasing the assumed value of

4of5 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.0901695106 Nagano and Nikaido Downloaded by guest on September 30, 2021 P by even 30% caused no change in the calculated value of K0.5 and less than suspensions diluted to an OD660 of 0.1 were used as inoculum, as described 25% changes in the Hill coefficient. Thus we believe that our conclusions are earlier (35). Plates were observed after overnight incubation at 37 °C. not affected by the lack of precision in the estimation of P. ACKNOWLEDGMENTS. We thank Gary Hoang, Anthony Daulo, and Cheng Determination of MIC. Linear gradient plates containing nitrocefin in the Chen for their contributions. This study was supported in part by the United bottom layer were made with LB agar in square Petri dishes, and bacterial States Public Health Service Grant AI-09644.

1. Li X-Z, Nikaido H (2004) Efflux-mediated drug resistance in bacteria. Drugs 64:159–204. 20. Basle´ A, Rummel G, Storici P, Rosenbusch JP, Schirmer T (2006) Crystal structure of 2. Nikaido H (2009) Multidrug resistance in bacteria. Annu Rev Biochem (in press). osmoporin OmpC from E. coli at 2.0 Å. J Mol Biol 362:933–942. 3. Nikaido H, Basina M, Nguyen V, Rosenberg EY (1998) Multidrug efflux pump AcrAB of 21. Lakey JH, Lea EJA, Pattus F (1991) ompC mutants which allow growth on maltodextrins Salmonella typhimurium excretes only those ␤-lactam antibiotics containing lipophilic show increased channel size and greater voltage sensitivity. FEBS Lett 278:31–34. side-chains. J Bacteriol 180:4686–4692. 22. Ma D, Alberti M, Lynch C, Nikaido H, Hearst JE (1996) The local repressor AcrR plays a 4. Ma D, et al. (1995) Genes acrA and acrB encode a stress-induced efflux system of modulating role in the regulation of acrAB genes of Escherichia coli by global stress Escherichia coli. Mol Microbiol 16:45–55. signals. Mol Microbiol 19:101–112. 5. Li X-Z, Nikaido H, Poole K (1995) Role of mexA-mexB-oprM in antibiotic efflux in 23. Nikaido H, Normark S (1987) Sensitivity of Escherichia coli to various ␤-lactams is Pseudomonas aeruginosa. Antimicrob Agents Chemother 39:1948–1953. determined by the interplay of outer membrane permeability and degradation by 6. Zgurskaya HI, Nikaido H (1999) Bypassing the periplasm: Reconstitution of the AcrAB periplasmic ␤-lactamases: A quantitative predictive treatment. Mol Microbiol 1:29–36. multidrug efflux pump of Escherichia coli. Proc Natl Acad Sci USA 96:7190–7195. 24. Davidson AL, Laghaeian SS, Mannering DE (1996) The maltose transport system of Esch- 7. Su CC, Yu EW (2007) Ligand-transporter interaction in the AcrB multidrug efflux pump erichia coli displays positive cooperativity in ATP hydrolysis. J Biol Chem 271:4858–4863. determined by fluorescence polarization assay. FEBS Lett 581:4972–4976. 25. Tato I, Zunzunegui S, de la Cruz F, Cabezon E (2005) TrwB, the coupling protein 8. Murakami S, Nakashima R, Yamashita E, Matsumoto T, Yamaguchi A (2006) Crystal involved in DNA transport during bacterial conjugation, is a DNA-dependent ATPase. structures of a multidrug transporter reveal a functionally rotating mechanism. Nature Proc Natl Acad Sci USA 102:8156–8162. 443:173–179. 26. Boyer PD (1993) The binding change mechanism for ATP synthase–some probabilities 9. Seeger MA, et al. (2006) Structural asymmetry of AcrB trimer suggests a peristaltic and possibilities. Biochim Biophys Acta 1140:215–250. pump mechanism. Science 313:1295–1298. 27. Weber J, Senior AE (1997) Catalytic mechanism of F1-ATPase. Biochim Biophys Acta 10. Sennhauser G, Amstutz P, Briand C, Storchenegger O, Grutter MG (2007) Drug export 1319:19–58. pathway of multidrug exporter AcrB revealed by DARPin inhibitors. PLoS Biol 5:e7. 28. Shirakihara Y, et al. (1997) The crystal structure of the nucleotide-free ␣3␤3 subcom- 11. Zimmermann W, Rosselet A (1977) The function of the outer membrane of Escherichia plex of F1-ATPase from the thermophilic Bacillus PS3 is a symmetric trimer. Structure coli as a permeability barrier to ␤-lactam antibiotics. Antimicrob Agents Chemother 5:825–836. 12:368–372. 29. Muneyuki E, Odaka M, Yoshida M (1997) A single mutation at the catalytic site of 12. Smit J, Kamio Y, Nikaido H (1975) Outer membrane of Salmonella typhimurium: TF1-␣3␤3␥ complex switches the kinetics of ATP hydrolysis from negative to positive Chemical analysis and freeze-fracture studies with lipopolysaccharide mutants. J Bac- cooperativity. FEBS Lett 413:55–59.

teriol 124:942–958. 30. Goldberg M, Pribyl T, Juhnke S, Nies DH (1999) Energetics and topology of CzcA, a MICROBIOLOGY 13. Li X-Z, Ma D, Livermore DM, Nikaido H (1994) Role of efflux pump(s) in intrinsic cation/proton antiporter of the resistance-nodulation-cell division protein family. resistance of Pseudomonas aeruginosa: Active efflux as a contributing factor to J Biol Chem 274:26065–26070. ␤-lactam resistance. Antimicrob Agents Chemother 38:1742–1752. 31. Nikaido H (1996) Multidrug efflux pumps of gram-negative bacteria. J Bacteriol 14. Stein WD, Danielli JF (1956) Structure and function in red cell permeability. Disc 178:5853–5859. Faraday Soc 21:238–251. 32. Sulavik MC, et al. (2001) Antibiotic susceptibility profiles of Escherichia coli strains 15. O’Callaghan CH, Morris A, Kirby SM, Shingler AH (1972) Novel method for detection of lacking multidrug efflux pump genes. Antimicrob Agents Chemother 45:1126–1136. ␤-lactamases by using a chromogenic cephalosporin substrate. Antimicrob Agents 33. Nishino K, Yamaguchi A (2001) Analysis of a complete library of putative drug trans- Chemother 1:283–288. porter genes in Escherichia coli. J Bacteriol 183:5803–5812. 16. Nikaido H, Rosenberg EY, Foulds J (1983) Porin channels in Escherichia coli: Studies with 34. Mazzariol A, Cornaglia G, Nikaido H (2000) Contributions of the AmpC ␤-lactamase and ␤-lactams in intact cells. J Bacteriol 153:232–240. the AcrAB multidrug efflux system in intrinsic resistance of Escherichia coli K-12 to 17. Dubus A, Normark S, Kania M, Page MG (1994) Role of asparagine 152 in catalysis of ␤-lactams. Antimicrob Agents Chemother 44:1387–1390. ␤-lactam hydrolysis by Escherichia coli AmpC ␤-lactamase studied by site-directed 35. Takatsuka Y, Nikaido H (2007) Site-directed disulfide cross-linking shows that cleft mutagenesis. Biochemistry 33:8577–8586. flexibility in the periplasmic domain is needed for the multidrug efflux pump AcrB of 18. Okusu H, Ma D, Nikaido H (1996) AcrAB efflux pump plays a major role in the antibiotic Escherichia coli. J Bacteriol 189:8677–8684. resistance phenotype of Escherichia coli multiple-antibiotic-resistant (Mar) mutants. J 36. Liu X, Ng C, Ferenci T (2000) Global adaptations resulting from high population Bacteriol 178:306–308. densities in Escherichia coli cultures. J Bacteriol 182:4158–4164. 19. Misra R, Benson SA (1988) Isolation and characterization of OmpC porin mutants with 37. Rosner JL, Chai KJ, Foulds J (1991) Regulation of OmpF porin expression by salicylate in altered pore properties. J Bacteriol 170:528–533. Escherichia coli. J Bacteriol 173:5631–5638.

Nagano and Nikaido PNAS Early Edition ͉ 5of5 Downloaded by guest on September 30, 2021