Introduction

The outer membrane of Gram-negative bacteria serves as a permeability barrier that prevents the entry of noxious compounds into the cell,1 thereby having the crucial role in multidrug resistance. Ensuring the barrier function of the outer membrane requires not only appropriate biogenesis of the outer membrane, that is, the biosynthesis, transport and assembly of outer membrane components such as lipids and proteins,2, 3 but also structural proteins that maintain its stability and integrity. Murein-lipoprotein (Lpp) and the Tol–Pal complex are widely conserved in Gram-negative bacteria and form a physical connection between the outer membrane and peptidoglycan, thereby maintaining the structural stability of the outer membrane.4, 5, 6, 7, 8, 9 Lpp is a small outer membrane lipoprotein that consists of 58 amino acids and is the most abundant outer membrane lipoprotein. About one-third of Lpp molecules are covalently linked to the peptidoglycan, and the remaining exist as free form.9 The Tol–Pal complex forms the structural bridge between the outer membrane and the cytoplasmic membrane.10, 11, 12 It comprises (at least) five proteins: TolA, TolQ and TolR form the cytoplasmic membrane component, TolB is a periplasmic protein and Pal is an outer membrane lipoprotein. Mutants defective in Lpp or any component of the Tol–Pal complex are known to show an increased susceptibility to drugs, leakiness of periplasmic proteins and partial detachment of the outer membrane from peptidoglycan, resulting in the formation of blebs or vesicles.7, 8, 13

Although the increased drug susceptibility of Δlpp and Δtol–pal mutants suggest that these proteins can be the potential target to permeabilize the outer membrane and therefore to combat multidrug-resistant Gram-negative bacteria, the significance of Lpp and Tol–Pal function on the outer membrane permeability and the drug resistance has not been fully characterized in a quantitative manner. Here, we determined the outer membrane permeability of these mutants by the assay using intact cells and liposomes reconstituted with the outer membrane proteins. We determined that there was 3- to 5-fold increase of the permeability in Δtol–pal mutants, but not in Δlpp mutant, compared with that in parental strain. The permeability increase in Δtol–pal mutants occurred without affecting the function of outer membrane diffusion channels, and was most pronounced in the cells at exponential growth phase. The impact of tol–pal deletion on the drug resistance was revealed to be almost comparable with that of deletion of acrAB, a major multidrug efflux transporter of E. coli that makes a predominant contribution to drug resistance. Our observations highlight the importance of Tol–Pal as a possible target to combat multidrug-resistant Gram-negative bacteria.

Materials and methods

Bacterial strains

LA51 is an E. coli K-12 derivative harboring a mutation in the promoter region of AmpC β-lactamase, which upregulates AmpC expression.14 The construction of the LA51 ΔacrAB mutant was described previously.15 LA51 mutants, in which the tolA, tolB, tolQ, tolR, pal or lpp genes are deleted, were constructed by transducing disrupted gene cassette containing Kmr from JW0729, JW5100, JW0727, JW0728, JW0731 or JW1667 (Keio collection),16 respectively. The transductants were selected with 35 μg ml−1 kanamycin. To construct the LA51 ΔacrAB double mutants with tol–pal or lpp deletion, we first transduced acrAB::Kmr gene from RAM1337,17 and the kanamycin-resistant gene was removed using pCP20 according to the method of Datsenko and Wanner.18 Next, the tol–pal or lpp genes were deleted as described above. The transduction was performed using the P1cml,clr100 phage and standard protocols.19

Measurement of outer membrane permeability

The protocol was essentially the same as described previously.15, 20 Briefly, the cells were grown in 50 ml of LB medium supplemented with 5 mm MgCl2 at 37 °C with shaking, and collected at an OD at 600 nm (OD600) of 1~1.2 by centrifugation at 5000g for 10 min at room temperature (RT). The cells were washed once with 50 mm potassium-phosphate buffer (pH 7.0) containing 5 mm MgCl2 (K-P buffer) and suspended in the same buffer. The OD600 of the suspension was adjusted to 4.0 (corresponding to 1.5 mg (dry cell weight) per ml). To measure the influx of benzylpenicilliin (PEN), the cell suspension was mixed with various concentrations of PEN and incubated at RT for 10 min, after which the mixture was heated in a boiling water bath for 1 min to stop hydrolysis by AmpC. During this period, any PEN that permeated outer membrane was subsequently hydrolyzed by AmpC. The amount of hydrolyzed PEN was quantified by microiodometry.21 The net rate of influx (Vin) was calculated as the rate of hydrolysis (Vh) per milligram dry cell weight. It should be noted that when acrAB+ strains are used, the Vin value is affected by the rate of efflux (Ve). Accordingly, ΔacrAB strains (with additional deletion of tol–pal or lpp genes) were used to determine the permeability coefficient (P) of the outer membrane. P was calculated using Fick’s first law of diffusion: Vin=P·A·(Co−Cp), where A, Co, and Cp represents the cell surface area (132 cm2 mg−1 (dry cell weight)) for E. coli,22 extracellular and periplasmic concentrations of PEN, respectively. Cp was calculated using Michaelis–Menten equation: Cp=Vh × Km/(Vmax−Vh), where Vmax and Km are kinetic parameters for AmpC, which were determined using cell extract as described previously.15

Nitrocefin permeability was examined by mixing 50 μm nitrocefin with the cell suspension prepared as described above. Hydrolysis of nitrocefin results in the increase in absorbance at 486 nm (A486), and the complete hydrolysis of 100 μm nitrocefin gives the increase of A486 by 2.05. The A486 of the reaction mixture was monitored for 2 min, and hydrolysis rate, which is equal to Vin, was calculated.

Measurement of the leakage of periplasmic β-lactamase

The cell suspensions prepared as above were incubated at RT for 15 min and the supernatant was collected after pelleting the cells by centrifugation at 20 000 g for 5 min at RT. The supernatant was then mixed with 50 μm PEN and incubated at RT for 10 min. The amount of hydrolyzed PEN was quantified by microiodometry.

MIC determination

The protocol was the same as previously described.15 Briefly, approximately 104 cells were inoculated to liquid LB broth or linear gradient plates containing drugs to be tested, and incubated for 18 h at 37 °C.

Preparation of outer membrane proteins

The protocol was the same as described previously.20

Liposome swelling assay

The protocol was essentially the same as described by Nikaido et al.23 Briefly, 2 μg of outer membrane proteins were reconstituted into liposomes comprising 2.4 μmol of phosphatidylcholine and 0.15 μmol of dicetylphosphate. In this assay, the influx of substrates into the proteoliposome causes the proteoliposome to swell, resulting in the reduction in the absorbance (at OD400) of the reaction mixture. Thus, the rate swelling, that is, the rate at which the OD400 decreases upon mixing of the proteoliposome with substrates, is taken as the rate of influx.

Electron microscopy

The samples were prepared according to the protocol of Fung et al.24 with several modifications. The cells were collected from 50 ml of LB medium (at OD600~1) by centrifugation at 400 g for 10 min at RT, washed once with K-P buffer, and suspended in the same buffer. The cell suspension was then mixed with 2% (final concentration) of glutaraldehyde, and incubated at RT for 1.5 h. The cells were pelleted by centrifugation at 400 g for 10 min at RT, and then embedded in 2% agar. After the agar solidified, it was washed with K-P buffer and soaked in 1% osmium tetroxide in Veronal buffer for 1 h at RT. The agar blocks were then dehydrated by soaking in increasing concentrations of ethanol (60, 70, 80, 90 and 100% for 30 min each). The agar blocks were then soaked in propylene oxide for 30 min at RT. This step was repeated three times to completely replace the ethanol with propylene oxide. The blocks were then embedded in Quetol651 (Nisshin EM, Tokyo, Japan) and the samples were thin-sectioned (60 nm) using an ultramicrotome and a diamond knife. The samples were stained with Ti blue (Nisshin EM; diluted 10-fold) and 0.4% lead citrate. The samples were observed under a transmission electron microscope (H-7650 Hitachi, Tokyo, Japan) at an accelerating voltage of 100 kV.

Theoretical calculation of the movement and concentration of PEN across the outer membrane

When β-lactam antibiotics are added to E. coli cells, they penetrate the outer membrane and then undergo either efflux transport via AcrAB-TolC or hydrolysis via AmpC β-lactamase. In a steady state, the relationship between Vin, Ve and Vh is described by the following equation: Vin=Ve+Vh (equation (1)). Previous study showed that this steady-state assumption was applicable to explain the movement of β-lactam antibiotics across the outer membrane.15 For PEN, all of the kinetic parameters required to describe Vin, Vh and Ve are available, and these can be described as follows: Vin=P·A·(Co−Cp), Fick’s first law of diffusion; Vh=Vhmax × Km/(Km+Cp), the Michaelis–Menten equation; and Ve=Vemax·Cph/(K0.5h+Cph), the Hill equation, in which Vemax, K0.5 and h are the maximum rate of efflux that AcrAB-TolC can produce, the substrate concentration that yields half-maximal efflux rate and the Hill coefficient, respectively. The parameters for PEN used in this theoretical analysis were as follows: P=0.07 × 10−5 cm s−1, Vemax=0.085 nmol mg−1 s−1, K0.5=0.30 μm, h=4.0, Vhmax=0.330 nmol mg−1 s−1 and Km=5.28 μm.15 Substituting these parameters into equation (1) gives the function of Cp against Co, which was used to create the theoretical Cp versus Co curve. Using this function, the influence of an increase in outer membrane permeability was assessed by substituting 4-fold higher value of P (0.28 × 10−5 cm s−1).

Results

Measurement of the outer membrane permeability in Δtol–pal and Δlpp mutants using the intact cells

Although both of Δlpp and Δtol–pal mutants are known to impair the outer membrane stability, scrutiny of their phenotype highlights the several lines of evidence that suggest the difference in their outer membrane barrier property. (1) Lpp deletion increases susceptibility to drugs mainly those with high hydrophobicity, whereas there is no apparent change in the MIC for relatively hydrophilic drugs, such as ampicillin.7 Also, the permeation rate of 6-aminopenicillanic acid across the outer membrane was shown to be unaltered by the deletion of Lpp.25 By contrast, deleting either of the tol–pal genes increases the susceptibility to hydrophobic as well as hydrophilic drugs that even include vancomycin, which is usually excluded by the outer membrane as it is both large and hydrophilic.26 These susceptibility patterns are also supported by a study by Nichols et al.27 which quantified the growth phenotype of nearly 4000 E. coli deletion mutants (Keio collection) in the presence of various drugs. (2) The addition of Mg2+, a divalent cation known to stabilize the outer membrane by reinforcing the lateral interactions between lipopolysaccharide (LPS) molecules,1 to the growth medium represses bleb or vesicle formation in lpp mutant,24 whereas tol–pal mutants are not apparently responsive to the external Mg2+ concentration.28, 29

To evaluate the outer membrane permeability of the Δtol–pal and Δlpp mutant, we measured the influx rates of PEN across the outer membrane of intact cells by coupling it with subsequent hydrolysis by periplasmic AmpC β-lactamase, a reaction that can be experimentally quantified according to the method of Kojima and Nikaido.15 For this assay, the net rate of PEN influx (Vin) was determined as its rate of hydrolysis (Vh). PEN was chosen as a substrate because (i) it is sensitive enough to AmpC hydrolysis,30 so Vh can be easily quantified and is expected to reflect precise differences in outer membrane permeability, even if they are only marginal and (ii) PEN diffuses through porin channels, but because it is not too hydrophilic, it slowly penetrates the phospholipid bilayer;31 therefore, possible outer membrane defect(s) in the lipid bilayer region and in proteinaceous channels are expected to be reflected by Vin. A prerequisite for the assay was that the cells did not leak a significant amount of β-lactamase into the external environment during the experiment. We examined β-lactamase activity in the supernatant after suspending cells in buffer at RT. We confirmed that using a buffer supplemented with 5 mm MgCl2 prevented significant β-lactamase leakage, at least within the 15-min incubation period (Supplementary Table S1). The rate of PEN influx into LA51 (parental strain), Δtol–pal and Δlpp cells is shown in Figure 1. Deleting any component of the Tol–Pal complex led to marked increase of Vin. By contrast, we observed no detectable difference of Vin in Δlpp mutant. The increase in Vin in Δtol–pal mutants was substantial, and this led us to compare it with the Vin for a mutant lacking AcrAB, an efflux transporter that pumps out a broad range of drugs from the periplasm (or inner membrane) and makes a predominant contribution to the multidrug efflux activity of E. coli.32 We found that the Vin for Δtol–pal mutants was comparable with that for ΔacrAB mutant (Figures 1a and b). The sigmoidal nature of the data, which was apparent in all the strains except ΔacrAB mutant, was derived from the positive cooperativity of the efflux activity by AcrAB-TolC.15, 33 To obtain the value for the permeability coefficient (P), a robust index that can be used to quantify outer membrane permeability, we constructed double deletion mutants (Δtol–pal/ΔacrAB or Δlpp/ΔacrAB) to eliminate the influence of efflux activity on Vin. We observed a 3- to 5-fold increase in P for the Δtol–pal mutants, but not for the Δlpp mutants (Table 1). To clarify whether similar increase of permeability occurs with other class of β-lactam antibiotics, we tested the permeability of nitrocefin (Figure 1c). The Vin of nitrocefin in Δtol–pal and ΔacrAB was markedly increased compared with that in parental strain, whereas only modest increase was observed in Δlpp mutant. This result confirmed that the permeability increase in Δtol–pal mutants was not limited to PEN.

Figure 1
figure 1

Rate of PEN influx into LA51 cells and derivative mutant cells. (a) Co is the extracellular PEN concentration. For clarity, a fitted line is shown to reflect the approximate Vin for the ΔacrAB strain. Data obtained below 500 μm of Co is shown in an enlarged graph (b). (c) Rate of nitrocefin influx into LA51 cells and its derivative mutants cells. Concentration of nitrocefin was 50 μm. Asterisks show that the difference compared with LA51 was statistically significant according to the Student’s t-test (P<0.05). The data are expressed as the mean±s.d. of triplicate measurements.

Full size image
Table 1 Permeability coefficient of benzylpenicillin
Full size table

The increase in the outer membrane permeability of the Δtol–pal mutants is growth phase-dependent

Recent studies showed that Tol–Pal complex is a part of cell division machinery, being involved in outer membrane invagination and modulation of peptidoglycan biosynthesis during the cell constriction.28, 34 Establishing a method for the quantitative measurement of the outer membrane permeability of the Δtol–pal mutant allowed us to examine the link between Tol–Pal function and cell division. With this aim, we examined the outer membrane permeability of Δtol–pal mutants during the various growth phases using PEN as a substrate (Figure 2). We found that increases in outer membrane permeability were most pronounced in cells during the exponential growth phase, whereas it became almost equivalent to that of the parental strain during the stationary phase. There was no significant change in the permeability of Δlpp or LA51 cells. These results suggest that only actively dividing Δtol–pal mutant cells accompany outer membrane defects, supporting the proposed link between cell division and the outer membrane stabilizing effect by Tol–Pal function.

Figure 2
figure 2

Rate of PEN influx into cells at various growth stages. Vin was measured with 300 μm of PEN. The data are expressed as the mean±s.d. of triplicate measurements from two independent experiments.

Full size image

Increase of drug susceptibility in Δtol–pal mutants, but not Δlpp mutant, was almost comparable with that of ΔacrAB mutant

Although it is known that deleting tol–pal or lpp leads to increased drug susceptibility, extent of the increase and the susceptibility pattern over various drugs has not been investigated in a comprehensive and quantitative manner using isogenic strains. The unexpectedly large increase in outer membrane permeability observed only in the Δtol–pal mutants led us to investigate and determine the MICs for various drugs against each mutant in the presence or absence of 5 mm MgCl2 (Tables 2 and 3). We confirmed that there was no significant difference in the growth rates among the mutants in the absence of drugs (Supplementary Figure S1). To obtain the precise values of MIC, we used a gradient-plating method when the conventional serial-dilution method failed to yield precise and reproducible MIC values. As expected, the Δtol–pal mutants showed markedly reduced MICs for almost all of the antibiotics tested, regardless of their size or hydrophobicity. Ciprofloxacin was the exception, presumably reflecting the fact that fluoroquinolones can penetrate the outer membrane rapidly1 even in LA51 and thus the MIC was not affected significantly by the mutation that alters the outer membrane permeability. The MICs for the Δtol–pal mutants were almost comparable with that of the ΔacrAB mutant. By contrast, the MICs for Δlpp were almost identical to those of the parental strain. These results confirmed that increased outer membrane permeability, as measured by PEN influx (Figure 1), correlated with drug susceptibility. These results also indicate that the phenotype of the Δlpp mutant, at least in terms of outer membrane barrier function, was almost completely suppressed by 5 mm Mg2+ (Table 2).

Table 2 MICs (μg ml−1) of various antibiotics determined in LB medium supplemented with 5 mm MgCl2
Full size table
Table 3 MICs (μg ml−1) of various antibiotics determined in LB medium without MgCl2
Full size table

Defining the outer membrane defect

Measurement of the permeability of outer membrane diffusion channels in reconstituted liposomes

The pattern of drug susceptibility shown by the Δtol–pal mutants was notable because increased susceptibility occurred regardless of the size and/or hydrophobicity of the drug. In general, defects in lipid bilayer do not significantly affect susceptibility to small, hydrophilic drugs, whereas defects in diffusion channels do not cause significant changes in susceptibility to large, hydrophobic drugs.35, 36 To clarify tol–pal specific outer membrane defect, we carried out (i) permeability measurement of liposomes reconstituted with outer membrane proteins and (ii) electron-microscopic observation of the cell surface structure. We reconstituted the outer membrane proteins from each mutant into liposomes and examined their permeability in a liposome-swelling assay23, 37 using various sugars as substrates (Figure 3). We observed no significant difference between the LA51, Δtol–pal and Δlpp mutants. This suggests that the high level of outer membrane permeability observed in Δtol–pal mutants was not due to proteinaceous factors, or at least not those that can be reconstituted into liposome. We examined the outer membrane proteins by SDS-PAGE (SDS polyacrylamide gel electrophoresis), but found no significant differences between the strains (Figure 3). As reported previously,38 the protein level of OmpF was slightly reduced in Δtol–pal mutants, but this did not affect the permeability of proteoliposomes. In E. coli, the OmpF and OmpC porin channels are the main pathways through which hydrophilic solutes diffuse.1 The channel properties of these porins can be examined by measuring the influx rates of various substrates, that is, the rate of swelling of proteoliposome. We found that all proteoliposomes showed similar swelling rates and substrate size dependency. Thus, we concluded that deletion of the tol–pal or lpp genes had no effects on the function of the outer membrane diffusion channels.

Figure 3
figure 3

(a) Outer membrane proteins of LA51 and its derivative mutants. Outer membrane proteins were analyzed on 10% SDS-PAGE gel containing 6 m urea. The gel was stained with Coomasie Brilliant Blue. M, molecular mass standards. (b) Liposome swelling assay. Outer membrane proteins (2 μg) from LA51 and its derivative mutants were reconstituted into liposomes, and swelling rates were measured upon mixing with the substrates of different MW (Mr). Substrates: arabinose (Mr: 150), glucose (Mr: 180), galactose (Mr: 180), N-acetylglucosamine (Mr: 221) sucrose (Mr: 342), lactose (Mr: 342). The data are expressed as the mean of triplicate measurements.

Full size image

Electron-microscopic observation of the cell surface structure

As there was no detectable defect in the outer membrane diffusion channels, another possible explanation for permeability increase should be physical instability of the membrane, which may result in transient fragility of the barrier. Although many studies have identified structural defects in the outer membrane of Δtol–pal or Δlpp mutants,8, 13, 24, 39 none have addressed the specific defects or the preferential location of these defects for the Δtol–pal mutants at the electron-microscopic level. We noted that the buffer used to measure the outer membrane permeability in intact cells (Figure 1) suppressed the leakage of β-lactamase for at least within 15 min at RT; nonetheless, there was a notable difference between the outer membrane permeability of the Δtol–pal and Δlpp mutant under this condition. We expected that electron-microscopic analysis under this condition would reveal Δtol–pal specific defects in the outer membrane. To avoid the artificial damage during the sample preparation, we referred to the protocol of Fung et al.,24 by which large blebs in the outer membrane in Δlpp mutant grown in low-Mg2+ medium was observed. Accordingly, we performed all procedures at RT, and the cells were never subjected to centrifugal forces greater than 400 g. With this sample, we observed the outer membrane defects specific for Δtol–pal mutants (Figure 4). The outer membrane of the mutants lacking any component of Tol–Pal showed apparent rupture and/or large blebs at the cell division septum. Around 10~15% of the observed cells (n~30) showed this kind of defect. Blebs were also observed around the cell poles, although less frequently (around 6%), presumably reflecting an inherited defect that might have originated at the site of constriction. There was no significant defect in the overall cell morphology of these mutants. On the other hand, no apparent ruptures, blebs or detachment from peptidoglycan were observed in the outer membrane of LA51 or the Δlpp mutant. These results are consistent with the finding reported by Gerding et al.,28 in which GFP protein was expressed in the periplasm of Δtol–pal mutants, and an increase in periplasmic volume specifically at the site of constriction was observed. Taken together, the specific defect at the site of constriction, which presumably leads to structural instability of the outer membrane in Δtol–pal mutants, appears most likely to explain the permeability increase in these strains.

Figure 4
figure 4

Cell surface structure of LA51 and its derivative mutants. Samples were stained with Ti blue and 0.4% lead citrate. (a) Cell constriction site. Structural defects in the outer membrane are indicated by arrows. Bars, 500 nm. Overall morphology of the cells was shown in the inset. Bars in insets, 1 μm. (b) Cell poles. Three representative micrographs showing the cell poles of Δtol–pal mutants are shown. Images show a ΔtolA mutant (left and center), and a ΔtolB mutant (right).

Full size image

Discussion

The outer membrane of Gram-negative bacteria serves as a barrier that prevents the entry of toxic compounds into the cell. Thus, permeabilizing the outer membrane, if it is possible, is expected to be a promising strategy to combat multidrug-resistant Gram-negative bacteria. We determined that Δtol–pal mutants, but not Δlpp mutant, showed 3- to 5-fold increase in the outer membrane permeability compared with that of parental strain, and found that the drug susceptibility of Δtol–pal mutants was almost comparable with that of ΔacrAB mutant. We analyzed below the outer membrane barrier property and the drug susceptibility of Δtol–pal mutants in comparison with that of Δlpp and ΔacrAB mutants.

Comparison of tol–pal deletion and lpp deletion

Cascales et al. showed that the outer membrane instability in Δlpp mutant was rescued by overexpression of Pal protein, whereas overexpression of the Lpp protein did not suppress the Δpal phenotype.26 Gerding et al.28 revealed that Tol–Pal complex is a part of the cell division machinery and is required for proper invagination of the outer membrane during cell constriction, providing the evidence for Tol–Pal specific function. Our results demonstrated that the outer membrane barrier of Δtol–pal mutants was indeed impaired only in the Δtol–pal cells at exponential growth phase (Figure 2), supporting the proposed link between cell division and outer membrane stabilizing function by Tol–Pal.

A breach in the permeability barrier during cell constriction in Δtol–pal mutants indicates that outer membrane integrity cannot be maintained just by passive tethering to peptidoglycan via Lpp, which distributes homogenously around the cell envelope.40 This may explain our observation that the permeability increase was evident only in Δtol–pal mutants, but not in Δlpp mutant, in the presence of 5 mm MgCl2. Loss of the tethering function of Lpp presumably led to only partial detachment of the outer membrane from the peptidoglycan; this might cause only limited instability that can be compensated for by the LPS-reinforcing properties of Mg2+. On the other hand, disorderly outer membrane invagination in Δtol–pal mutants presumably causes more drastic outer membrane instability, as also evident by electron-microscopic observation (Figure 4), which could not be compensated for by Mg2+.

Comparison of tol–pal deletion and acrAB deletion

We showed that the drug susceptibility of Δtol–pal mutants was almost comparable with that in the ΔacrAB mutant (Table 2). This was not surprising considering that the major factors underlying the antibiotic resistance of Gram-negative bacteria are the outer membrane permeability barrier and efflux pump activity, and that these factors act synergistically.15, 32, 41 This synergistic relationship was indeed evident in Figure 1, as Vin of PEN into Δtol–pal mutants at low concentration (Co<200 μm), a condition where the efflux activity by AcrAB remains low because the periplasmic PEN concentration is low,15 was higher than that into ΔacrAB. This difference reflects the fact that Δtol–pal mutants have defects in outer membrane barrier, on the other hand, the outer membrane of ΔacrAB is intact.

The efficacy of small, hydrophilic drugs that rapidly permeate the outer membrane (for example, ampicillin), is less affected by efflux activity than that of bulky, hydrophobic drugs with low outer membrane permeability.42 The faster a drug permeates across the outer membrane, the more difficult it is for the efflux pump to counteract it. We revealed that the P-value for PEN in the Δtol–pal mutants was 3- to 5-fold higher than that in the parental strain (Table 1). This allowed us to assess the impact of the permeability increase in a theoretical manner. PEN targets penicillin-binding proteins located in the periplasm.43 Thus, the level of resistance to PEN is dependent on the relationship between the PEN concentration in the periplasm (Cp) and that outside the cell (Co); if Cp remains low against a high Co level, it means that the cells are highly resistant to PEN. A theoretical Cp versus Co curve is shown in Figure 5. This curve was created using kinetic parameters representing influx, efflux via AcrAB-TolC efflux pump, and hydrolysis via AmpC β-lactamase.20 We assumed steady-state movement of PEN across the outer membrane, designated Vin=Ve+Vh, where Vin, Ve and Vh are the rate of influx, efflux and hydrolysis, respectively (see 'Materials and Methods' section).15, 33 A previous study showed that this steady-state assumption was applicable to explain the experimentally determined relationship between Cp and Co relationship.15 As shown in the figure, assuming a 4-fold increase in outer membrane permeability gave a Cp versus Co curve similar to that for ΔacrAB mutant, indicating that a 4-fold increase was indeed sufficient to cause increased susceptibility to PEN at a level comparable with that conferred by the acrAB deletion. Such a theoretical calculation is not possible for other classes of antibiotics because the kinetic parameters for influx and efflux are not available. However, considering that most antibiotics need to penetrate the periplasm to reach their targets, breaching the outer membrane barrier results in increased influx and an increase in Cp, which ultimately results in higher drug concentration in the cytoplasm and increased susceptibility. Inhibiting multidrug efflux pumps is a promising way to combat multidrug-resistant Gram-negative bacteria.32 Although direct inhibition of these pumps may be the most efficient strategy, our results show that inhibiting the function of the Tol–Pal complex can counteract efflux activity, thereby making this an alternative way to combat multidrug resistance.

Figure 5
figure 5

Theoretical relationship between the Cp and Co values for PEN in LA51 and its derivatives. Theoretical Cp versus Co curves for LA51 and the ΔacrAB mutant are shown. The effect of increasing outer membrane permeability was assessed by assuming a 4-fold increase in the permeability coefficient.

Full size image