Influx and efflux properties of the deletion mutants
bamB (ΔbamB), tolC (ΔtolC), and bamB and tolC (ΔbamBΔtolC) deletion mutants of E. coli were created. N-Phenyl-1-naphthylamine (NPN), a probe compound, was used to investigate the influx and efflux properties of the mutants. It is a small hydrophobic fluorescent dye (molecular weight, 219; 1-octanol/water partition coefficient [log Pow], 4.2 [https://pubchem.ncbi.nlm.nih.gov]) that exhibits high fluorescence in hydrophobic environments, such as lipid bilayers of cell membranes17.
The influx properties of the wild-type strain and the mutants were analyzed by adding NPN to the cultures and monitoring fluorescence over time (Fig. 1a). In this experimental setup, an increase in fluorescence was interpreted as an increase in the influx rate or a decrease in the efflux rate. Fluorescence increased in ΔbamB even at time zero. This result showed that the uptake in ΔbamB occurred too fast to monitor a gradual increase in fluorescence. In ΔbamBΔtolC, fluorescence also immediately increased although it gradually further increased. The fluorescence of ΔtolC at time zero was the same as that of the wild-type strain, and it increased gradually over time. These results suggested that influx was increased by bamB deletion but not by tolC deletion; tolC deletion might cause a decrease in the efflux rate.
NPN influx and efflux analysis. (a) Result of the influx analysis; fluorescence intensity of NPN is presented as a function of time. The data are presented as the mean ± standard error of mean from triplicate experiments. (b) Efflux analysis. The arrow indicates the time of glucose addition. NPN, N-phenyl-1-naphthylamine; WT, wild-type strain; ΔbamB, outer membrane lipoprotein deletion mutant; ΔtolC, outer membrane efflux protein deletion mutant; ΔbamBΔtolC, bamB and tolC double deletion mutant.
An efflux analysis was performed using NPN and carbonyl cyanide 3-chlorophenylhydrazone (CCCP) that inhibits the proton motive force (PMF) and hence the function of efflux pumps. The cells subjected to CCCP treatment in the absence of glucose, which do not have any PMF driving efflux pumps, were resuspended in buffer with NPN, and fluorescence transition was monitored (Fig. 1b). Because of a lack of PMF, fluorescence intensity was gradually increased solely by the influx in all strains. After the addition of glucose that restored PMF and efflux flow, the fluorescence intensity immediately deceased in ΔbamB and the wild-type strain but not in ΔtolC and ΔbamBΔtolC. This result indicated that tolC deletion affected efflux property. Therefore, bamB and tolC deletion resulted in an increase in influx and a decrease in efflux properties, respectively.
Scaffold size-dependent effect of antibiotics on uptake efficiency in the deletion mutants
The uptake efficiency of foreign compounds in the deletion mutants was further evaluated by determining the MIC of various antibiotics (Table 1). Regarding large-scaffold antibiotics, bamB deletion effectively improved susceptibility: the MICs of vancomycin and actinomycin D decreased by 16- and 8-fold, respectively. Conversely, small-scaffold hydrophobic antibiotics (chloramphenicol, triclosan, and 5-ketoclomazone) were more efficiently taken up by the tolC mutant than those by the other mutants (8-, 64-, and over 4-fold, respectively). An inverse trend of scaffold size dependency of MIC fold change (mutant/wild-type) was observed between ΔbamB and ΔtolC with statistical significance (Fig. 2a). Conversely, no clear correlation was observed in ΔbamBΔtolC. bamB or tolC deletion likely decreased the MIC by improving the inefficient uptake of antibiotics by the wild-type strain due to a low influx or high efflux rate. The uptake of moderate-sized scaffold hydrophobic antibiotics (rifampicin, erythromycin, novobiocin, and fusidic acid) was improved via both deletions (64-, 256-, 256- and 512-fold, respectively). These results suggested that the effect of each deletion on MIC depended on the scaffold size of antibiotics. The hydrophobicity dependency of MIC fold change was commonly observed in the mutants (Fig. 2b), indicating that compounds with high hydrophobicity were more susceptible to influx and efflux changes by bamB or tolC deletion than compounds with low hydrophobicity.
Correlation between MIC fold change and scaffold size or hydrophobicity. The MIC fold change (each mutant/the wild type strain) and (a) scaffold size or (b) hydrophobicity of antimicrobial compounds are plotted for each mutant. The dotted line represents a regression line. The correlation coefficient (r) and p value by Pearson’s correlation analysis are indicated within each panel; p value in bold means significant at p = 0.05.
bamB and tolC deletions elicited combined effects on the uptake of some antibiotics (erythromycin, novobiocin, fusidic acid, and triclosan; Table 1). Therefore, interactions between bamB and tolC deletions were investigated by using a fractional inhibitory concentration index (FICI), which is used to evaluate interactions (synergism, no interaction, or antagonism) between two inhibitory agents (Supplementary Table S1). The results showed that two deletions worked synergistically in moderate-sized scaffold antibiotics showing a combined effect. Moreover, triclosan, a small-scaffold antibiotic with a combined effect, also had an FICI of 0.502, which is close to the criterion for synergism (FICI ≤ 0.5).
Uptake efficiency of CPP-PNAs by the deletion mutants
Antisense CPP-PNAs are considered new anti-infective or antimicrobial drugs. PNA is a charge-neutral oligonucleotide analog that possesses good hybridization properties and resistance to degradation by nucleases and proteases19. The uptake efficiency of PNA alone by bacteria is usually poor; thus, CPPs such as KFFKFFKFFK are frequently conjugated to PNAs19. Therefore, we evaluated whether the uptake efficiency of KFF-acpP, a CPP-PNA, also increased in the deletion mutants. acpP is essential for growth, and KFF-acpP has bactericidal effects19. Subinhibitory KFF-acpP concentrations (0.3 μM) caused a slight growth delay in the single deletion mutants (ΔbamB and ΔtolC) and a relatively strong growth delay in the double deletion mutant (ΔbamBΔtolC). However, it did not affect the growth curve of the wild-type strain, indicating that KFF-acpP was taken up more efficiently in the mutants than in the wild-type strain (Fig. 3). The differential effect of KFF-acpP on growth between the mutants and the wild-type strain was also supported by the decrease in MIC in the mutants (Supplementary Table S2).
Growth of the wild-type and mutant strains in the presence of KFF-acpP. Cell growth in the presence or absence of CPP-PNAs is indicated. The CPP-PNAs used were KFF-NC (□) and KFF-acpP (○). In a control experiment, distilled water was added instead of CPP-PNAs (△). The final CPP-PNA concentration was 0.3 μM. Data are presented as the mean ± standard error of the mean from triplicate experiments. KFF-NC, negative control; KFF-acpP, CPP-PNA targeting acyl carrier protein.
Proof-of-concept usages: detection of antibiotic compounds at low concentrations
In screening novel antibiotics from natural samples, antibiotic concentrations in samples are generally low20. Therefore, the presence of antibiotic compounds in such samples is often difficult to be detected; consequently, sensitized strains are preferable as indicator strains. On this basis, we tested the sensitivity of the present mutants to the extracts of environmental soil samples as follows: potential antibiotic-containing samples were prepared by inoculating a soil sample with a liquid medium and extracting the culture with ethyl acetate; drops of the extracts were placed onto agar plates containing the wild-type or mutant cells. We found halos (zones of inhibition) only in the plates with ΔbamBΔtolC cells (Fig. 4a). Therefore, ΔbamBΔtolC was more sensitive to antibacterial compounds than the wild-type strain and could be used for detecting active compounds at low concentrations.
Detection of antibiotic activity in the culture samples of natural soil and gene expression titration assay. (a) Aliquots of the extract from soil culture with R2A (R2A sample) and M9 (M9 sample) media and control (DMSO) were spotted on the lawn of indicator strains on an agar plate as indicated by a diagram. After incubation, antibiotic activity was detected as the inhibition zone formed on the plate. The arrow indicates a halo. (b) 5-Ketoclomazone was spotted on the lawn of indicator strains on an agar plate at concentrations indicated by a diagram. (c) Control experiment was performed with triclosan and fusidic acid as test antibiotic compounds. ΔbamB, outer membrane lipoprotein deletion mutant; ΔtolC, outer membrane efflux protein deletion mutant; ΔbamBΔtolC, bamB and tolC double deletion mutant; dxs AS, expression of dxs antisense RNA.
Proof-of-concept usages: gene expression titration assay (GETA)
A GETA can also be used to increase the sensitivity of cells to antibiotics at low concentrations21. In GETA, an antisense agent is used to partially silence the expression of a specific mRNA, whose translation product is a target for a certain antibiotic; consequently, host cells become sensitized because antibiotics can elicit their effects at low concentrations.
5-Ketoclomazone was used as a test antibiotic here because it showed a low uptake by the wild-type strain (Table 1). dxs encoding 1-deoxy-d-xylulose 5-phosphate synthase, the putative cellular target molecule of 5-ketoclomazone22, was silenced by the endogenous expression of a short antisense RNA against dxs by using an expression vector23. The cells became more susceptible to 5-ketoclomazone when the antisense RNA against dxs was expressed (Fig. 4b), and this effect was greater in ΔtolC and ΔbamBΔtolC than in the wild-type strain and ΔbamB even at low 5-ketoclomazone concentrations. In the negative control experiment involving antibiotics unrelated to Dxs (triclosan and fusidic acid), dxs silencing did not change the sizes of the halo (Fig. 4c). This result indicated that dxs silencing specifically increased the susceptibility to 5-ketoclomazone. Therefore, target gene silencing conferred the higher susceptibility of tolC deletion mutants to 5-ketoclomazone than that of the wild-type strain, and the present bamB/tolC mutants could be useful high-sensitive host strains for GETA.
Proof-of-concept usages: conversion of inactive VD3 to active hydroxylated VD3 in the mutants
VD3 is a hydrophobic vitamin that contains a steroid skeleton24. Several forms of VD3 exist; cholecalciferol (molecular weight, 385; log Pow, 7.5 [https://pubchem.ncbi.nlm.nih.gov]) is an inactive form of VD3, and its hydroxylated forms 25-hydroxycholecarciferol (calcifediol) and 1,25-dihydroxycholecalciferol (calcitriol) are active forms of VD3 (showing a hormonal activity) and valuable as pharmaceuticals24. An enzymatic conversion of inactive to active VD3 in microbes is desirable because an enzymatic hydroxylation approach is more cost-effective than chemical synthesis25. However, a low VD3 uptake rate in E. coli retards enzymatic conversion8. Therefore, cells expressing VD3 hydroxylase (vdh), ferredoxin (aciB), and ferredoxin reductase (aciC) genes were prepared to improve the conversion of VD3. When the wild-type strain was used as an expression host, no VD3 conversion occurred, whereas in ΔtolC and ΔbamBΔtolC, an effective conversion (1.4 and 4.6 μM, respectively) was detected (Fig. 5). Therefore, the conversion rate of inactive VD3 was improved by tolC deletion and further enhanced by double deletion with bamB.
VD3 conversion efficiency of the mutants. Inactive VD3 was converted to active VD3 by using cells transformed with pHN1387 (empty vector control; vdh-aciBC plasmid−) or pHN4136 (vdh-aciBC expression plasmid; vdh-aciBC plasmid+). The host cells used are indicated at the bottom of the graph. The quantification result of inactive VD3 that remained after the conversion reaction from triplicate experiments is presented as the mean ± standard error of the mean. Data were statistically analyzed using ANOVA, and significant differences were evaluated with Tukey’s multiple comparison test. One-way ANOVA showed a significant difference at F = 2.7 (p = 0.046). The bars marked with “#” correspond with the groups that did not differ significantly from each other at adjusted p < 0.05 in the post hoc test. The unmarked bars correspond with the group that significantly differed from the other groups. (b) The quantification result of active VD3 produced after the conversion reaction is shown as in (a). VD3, vitamin D3; ΔbamB, outer membrane lipoprotein deletion mutant; ΔtolC, outer membrane efflux protein deletion mutant; ΔbamBΔtolC, bamB and tolC double deletion mutant. One-way ANOVA showed a significant difference at F = 82 (p = 2.5 × 10–11). “##” is marked as in (a).
Sensitizing wild-type E. coli with anti-bamB and tolC CPP-PNAs
The aforementioned results depended on gene recombination techniques; thus, the generated mutant strains cannot be readily used in pharmaceutical and food industries because of restrictions and regulations for the use of genetically modified organisms. Instead of using gene recombinant mutants, we tested whether CPP-PNAs targeting bamB and tolC could function as efficient sensitizers for increasing the uptake of several antibiotics. Before CPP-PNAs as sensitizers were evaluated, their effect on the growth of the wild-type strain in the absence of antibiotics was assessed. Low CPP-PNA concentrations marginally affected growth, but it caused a slight growth delay at high concentrations, indicating a dose-dependent negative effect of CPP-PNAs on E. coli growth (Supplementary Fig. S1). The wild-type strain was then treated with antibiotics in the presence of the following CPP-PNAs: KFF-NC, KFF-bamB, and KFF-tolC (Table 2). Unexpectedly, the treatment with KFF-NC decreased the MIC of certain antibiotics, namely, actinomycin D, novobiocin, and fusidic acid (compared with the results in Table 1); this result indicated that the KFFKFFKFFK peptide fragment affected the uptake efficiency of some antibiotics. Nevertheless, KFF-bamB and KFF-tolC more effectively decreased the MICs than KFF-NC did. For example, KFF-bamB decreased the MICs of actinomycin D, novobiocin, and fusidic acid by 2-, 16-, and 4-fold, respectively; KFF-tolC decreased the MIC of 5-ketoclomazone by 2-fold. Furthermore, the combination of KFF-bamB and KFF-tolC decreased the MIC of the tested antibiotics largely than the single use of each CPP-PNA except for the MIC of vancomycin, which was not decreased by any CPP-PNAs. The interaction between KFF-bamB and KFF-tolC was evaluated by using FICI (Supplementary Table S3), and two CPP-PNAs were found to work synergistically (FICI ≤ 0.5) for actinomycin D and fusidic acid. These results indicated that KFF-bamB and KFF-tolC are effective sensitizers for E. coli against some antibiotic compounds.
Sensitizing Pseudomonas aeruginosa with anti-bamB and oprM CPP-PNAs
To explore the applicability of CPP-PNA approach, we tested the effect of CPP-PNAs on antibiotic susceptibility in a notorious opportunistic pathogen, P. aeruginosa. We additionally created CPP-PNAs that targeted bamB and oprM, a tolC homolog in the P. aeruginosa genome and evaluated the potential of these CPP-PNAs as sensitizers in the same manner as E. coli. CPP-PNAs negatively affected the growth of P. aeruginosa PAO1 in the absence of antibiotics; although the same trend was observed as in E. coli, the effect on the growth of P. aeruginosa PAO1 was stronger than that on the growth of E. coli (Supplementary Fig. S2). Similar to E. coli, P. aeruginosa PAO1 treated with the CPP-PNAs targeting bamB and oprM was sensitized by certain antibiotics (Table 2). For all antibiotics tested, the highest MIC decrease (4-, > 8-, and > 16-fold decrease for vancomycin, erythromycin, and carbenicillin, respectively) was observed when both CPP-PNAs were simultaneously added although a synergistic interaction between RXR-bamB and RXR-oprM was not detected (Supplementary Table S3). Therefore, CPP-PNAs targeting bamB and tolC homologs function as efficient sensitizers for improving the uptake of various antibiotics by Gram-negative bacteria.
