The Biofilm Inhibitor Carolacton Enters Gram-Negative Cells: Studies Using a TolC-Deficient Strain of Escherichia coli

The emergence of pathogens resistant against most or all of the antibiotics currently used in human therapy is a global threat, and therefore the search for antimicrobials with novel targets and modes of action is of utmost importance. The myxobacterial secondary metabolite carolacton had previously been shown to inhibit biofilm formation and growth of streptococci. Here, we investigated if carolacton could act against Gram-negative bacteria, which are difficult targets because of their double-layered cytoplasmic envelope. We found that the model organism Escherichia coli is susceptible to carolacton, similar to the Gram-positive Streptococcus pneumoniae, if its multidrug efflux system AcrAB-TolC is either inactivated genetically, by disruption of the tolC gene, or physiologically by coadministering an efflux pump inhibitor. A carolacton epimer that has a different steric configuration at carbon atom 9 is completely inactive, suggesting that carolacton may interact with the same molecular target in both Gram-positive and Gram-negative bacteria.

negative cell envelope and that the lack of sensitivity of wild-type E. coli to carolacton is due to export from the cell by TolC-mediated efflux. However, mutations in TolC can have different effects on substrate export, and there have even been reports that a misassembled TolC protein may result in an open channel which allows influx of antibiotics into the cell, resulting in an increased sensitivity (25). The TolC-deficient strain used in our screenings has been propagated as a glycerol stock in laboratories since at least 1980 (B. Kunze, personal communication), and so far it has not been characterized genetically. Over a period of 37 years, massive genetic changes could have occurred (26). Moreover, although TolC-deficient strains are used by many laboratories, they were constructed with different methods and in different genetic backgrounds (25,27,28), making it hard to compare results. We here determined the genome sequence of E. coli TolC with high resolution by using a combination of PacBio and Illumina sequencing. With these methods, an insertion of a natural transposon at the tolC locus was identified, and genetic changes were recorded that had occurred in this strain in comparison to its closest relative, which was identified as E. coli K-12 MG1655 (NZ_CP014225.1). We determined MICs for E. coli K-12 MG1655 and E. coli TolC and deposited E. coli TolC with the DSMZ as a tool and reference for future studies. We then studied the influence of carolacton on E. coli TolC by using transcriptome sequencing (RNA-seq), the carolacton C-9 (R) epimer, and the EPI PA␤N. The data clearly showed that carolacton easily penetrates the Gram-negative cell envelope. Once inside the cell, it inhibits E. coli at similar concentrations as for streptococci, suggesting that the molecular target of carolacton is highly conserved and might be highly similar even in distantly related bacterial phyla, such as Firmicutes and Proteobacteria. The export of carolacton from the cell can be overcome by blocking the AcrAB-TolC efflux complex with the EPI PA␤N. This finding highlights the potential use of carolacton in combinatorial treatment with EPIs.

RESULTS
E. coli TolC is an ancient natural derivative of E. coli K-12 and is closely related to K-12 MG1655. PacBio single-molecule real-time (SMRT) sequencing and Illumina MiSeq short-read sequencing were combined to obtain a high-quality genome sequence of E. coli TolC. By Illumina MiSeq sequencing, 2,623,454 reads were obtained, totaling~656 Mb and resulting in~138-fold genome coverage. The PacBio SMRT sequencing data set consisted of 74,571 reads with an N50 read length of 17,770 bp and was used for de novo genome assembly. For the correction of indel errors, Illumina reads were mapped onto the newly assembled genome.
A nucleotide-based genome BLAST distance phylogeny (GBDP) tree with branch support values inferred from both the nucleotide and amino acid data is depicted in Fig. S1 of our supplementary data posted on figshare (https://doi.org/10.6084/m9 .figshare.5395471). The average branch support of the nucleotide tree was 47.3%, and branch support for the amino acid tree was 37.6%. Target strain E. coli TolC was placed in a highly supported subtree containing 14 strains, most of them K-12 strains. Figure 1 shows the nucleotide sequence identity of E. coli TolC in comparison to the five most similar E. coli strains, as reported in BLASTϩ. Most notably, E. coli TolC contains the bacteriophage and the fertility plasmid F integrated into its chromosome. Phage was located between genes ybhB and ybhC at positions 3,079,545 to 3,128,200 of the E. coli TolC chromosome, and the F plasmid was integrated into an insertion sequence element (IS3C) within the cryptic prophage DLP12 (positions 3,368,702 to 3,467,447). This is in contrast to the most closely related E. coli strains, which encode neither the fertility plasmid nor phage , the only exception being NCM3722, which still carries phage (Fig. 1A). In comparison to MG1655, an rph-1 mutation is absent in TolC, and the rpoS gene is present as the 33Am variant. Like other derivatives of E. coli K-12, strain E. coli TolC is also valine sensitive (ilvG deficient) (30). Similar to E. coli MG1655, an early deletion of two nucleotides (c.977_978delAT) that results in an Ile327-Glu substitution and subsequent insertion of a premature TGA translation termination site at position c.982_984 were found. As a common marker of all E. coli K-12 derivatives, E. coli TolC additionally carries an IS5 insertion (IS5I) in the last gene of the O-antigen cluster encoding the rhamnosyltransferase WbbL (rfb-50 mutation) (31). Although these strains are closely related, large structural rearrangements within their chromosomes were found (Fig. 2).
The tolC locus (btd92_00696) was inspected in detail, and the absence of a functional copy of the tolC gene was confirmed. The E. coli TolC strain carries a transposon insertion after base 1309 (c.1309_1310insIS5*) of the tolC gene, and this causes a disruption of the CDS (Fig. 1B). Genes of the three additional OMF proteins in E. coli (cusC, mdtQ, and mdtP) were not affected (see Table S1 at https://doi.org/10.6084/m9 .figshare.5395471). The transposon within tolC was identified as transposable element IS5, which contains three protein-coding genes: the transposase gene insH1 (ins5A) and two genes (ins5B and ins5C) opposite insH1 with unknown function (32,33). Altogether, the E. coli TolC chromosome contained 12 insertions of IS5 elements, of which only the one integrated into the tolC locus (IS5*) disrupted a functional gene. Additionally, IS5 insertions were also located within the sequences of cryptic prophages, e.g., the IS5Y element was inserted into the cryptic prophage Rac, interrupting lomR=. The E. coli TolC  (64) comparison of the E. coli TolC genome (innermost black ring) to the closely related genomes of E. coli strains K-12 MG1655, ER1821R, NCM3722, K-12 W3110, and JW5437-1 (the four outermost rings), shown in blue to red, respectively, as identified by isDDH (29). Shading of the four outermost rings is according to their respective percent nucleotide identity to the query sequence (E. coli TolC), determined by BLASTϩ. The second and third innermost rings show the GC skew (purple/green) and the GC content (black). IS5 elements are numbered according to annotations for E. coli K-12 MG1655 (NC_000913.3). The location of the fertility plasmid on the chromosome of E. coli TolC is indicated by the letter F (on left side of diagram). (B) Close-up comparison of the tolC locus of E. coli TolC and its closest relative, E. coli K-12 MG1655, drawn by using Easyfig (66) The tolC locus (tolC_1 and tolC_2) in E. coli TolC is interrupted by insertion of an IS5 element (IS5*) that codes for the transposase insH1 (ins5A). ins5B (**) and ins5C (*) are indicated by arrows in reverse orientation, underneath insH1. A BLASTϩ comparison of the tolC locus for each of the two strains indicated 100% nucleotide identity. strain described here was deposited at the Leibniz Institute DSMZ German Collection of Microorganisms and Cell Cultures (Braunschweig, Germany) and assigned strain number DSM 104619.

Role of TolC for MICs of carolacton and different classes of antibiotics.
To evaluate the effect of TolC inactivation on antibiotic susceptibility of E. coli, the MICs of selected antibiotics against E. coli MG1655 and E. coli TolC were determined (Table 1). We included two RNA polymerase inhibitors, corallopyronin A and sorangicin, previously isolated from myxobacteria at our institution (34,35).
E. coli TolC was at least 64 times more sensitive to carolacton than E. coli MG1655. The MIC of carolacton against E. coli TolC was in the same range as that reported by Jansen et al. (19). For S. pneumoniae TIGR4, the MIC of carolacton was determined to be 0.06 &micro;g/ml (24), similar to the value reported for E. coli TolC. In comparison to E. coli MG1655, E. coli TolC showed a strong increase in sensitivity (Ն4-fold) to antibiotics from all functional groups. The determined MICs were in the same range as those reported previously for E. coli W3110 and its tolC null mutant (11), indicating that the presence of the F plasmid and phage do not affect antibiotic susceptibility. Rifampin and vancomycin are not substrates of the pump; thus, E. coli TolC is not expected to be hypersensitive to these compounds, which was confirmed. The data indicated that carolacton penetrates the two membranes of the Gram-negative cell envelope and that its intracellular inhibitory effect is comparable to that of Grampositive cells. Transcriptional response of E. coli TolC to carolacton. We analyzed the transcriptome of carolacton-treated cultures of E. coli TolC in comparison to untreated cultures during the first 30 min of growth.
In total, 4,730 transcripts of E. coli TolC were investigated using Rockhopper (see Data Set S1 in the supplemental material). At 30 min after addition of carolacton, 71 transcripts showed a strong differential abundance (log 2 fold change [FC] of ՆϮ2), corresponding to 1.6% of all open reading frames of E. coli TolC (Data Set S2). At this time point, E. coli TolC grows at the same rate with or without carolacton (see below). The data therefore provide additional proof that carolacton immediately enters the Gram-negative cell. At a log 2 FC of ՆϮ0.8, approximately 29% of all genes were differentially abundant, comparable to the degree of differential transcript abundance in S. mutans (31.3%) and S. pneumoniae (22.8%) in the presence of carolacton when we used an identical cutoff (21,24). The most strongly differentially abundant transcripts encoded components for flagellar assembly, heat shock and cold shock proteins, and chaperones (Fig. 3). Transcription of the alternative sigma factor F ( 28 ) was upregu-lated~7.4-fold (log 2 FC, 2.88), and the putative helix-turn-helix (HTH)-type transcriptional regulator RhmR was downregulated. Moreover, precursors of the outer membrane pore proteins NmpC (btd92_03329) and PhoE (btd92_03746) were upregulated. Interestingly, all 7 StyR-44 family small noncoding RNAs encoded in the genome were strongly (log 2 FC, Ն6.5) upregulated after only 5 min of growth with carolacton. The data showed that interaction of E. coli TolC with carolacton triggers global transcriptional adaptations already after 5 min, suggesting a molecular target in a central metabolic pathway.
Stereospecificity of carolacton activity and inhibition of efflux. Subsequently, the differences in carolacton susceptibility between E. coli TolC and E. coli MG1655 were investigated in detail over all growth phases. E. coli MG1655 with and without carolacton and TolC without carolacton grew similarly and reached their maximal optical density at 600 nm (OD 600 ) of~6 after 7 h (Fig. 4A). In the presence of carolacton (added at t ϭ 0), growth was indistinguishable from the controls for 1 h. At this time point, growth of the carolacton-treated culture of the E. coli TolC strain was strongly inhibited, while all other strains entered the exponential growth phase. The carolacton-treated culture of the E. coli TolC strain grew linearly over the next 5 h to an OD 600 of approximately 0.8, which did not increase much farther and reached a maximal OD 600 of around 1 after 24 h. Complementation of E. coli TolC with a plasmid-borne copy of the OMF TolC was able to restore insensitivity to carolacton, confirming indeed the absence of TolC-mediated efflux of carolacton as the sole cause for sensitivity (Fig. 5).
epi-Carolacton is a carolacton epimer with an inversion of the stereocenter at C-9 from the native (S) to the (R) configuration. This carolacton derivative lacks biological activity in S. pneumoniae TIGR4 and S. mutans UA159 (22,24). Here, we tested the inhibitory properties of epi-carolacton against E. coli TolC. Figure 4B shows that epi-carolacton had no influence on growth of E. coli TolC. Since epi-carolacton was dissolved in dimethyl sulfoxide (DMSO), we investigated its effect on growth as an additional control, but we did not detect any. The loss of growth inhibition of epi-carolacton shown here suggests that the molecular target of carolacton might not only be conserved in the genus Streptococcus but also in the phyla Firmicutes and Proteobacteria.
Antibiotics that are substrates of TolC have to be administered in high doses to overcome the intrinsic resistance mediated by efflux (13). Alternatively, they could be applied in combination with efflux pump inhibitors. Therefore, we investigated the influence of PA␤N, a competitive inhibitor of AcrAB-TolC (16), on carolacton sensitivity in E. coli. Table 1 shows that the MIC of E. coli MG1655 toward carolacton was reduced from Ͼ8 &micro;g/ml to 4 &micro;g/ml when PA␤N was coadministered at 40 &micro; g/ml. Lower concentrations of PA␤N had no effect on the MIC of carolacton. The susceptibility of the TolC mutant was also increased by PA␤N. The MIC of E. coli against PA␤N has been shown before to be strongly reduced in an efflux-deficient strain (ΔacrAB); moreover, PA␤N can cause membrane destabilization as an unspecific side effect (16). Accordingly, we observed a growth reduction of~45% for the effluxdeficient E. coli TolC strain when grown with 40 &micro;g/ml PA␤N, but not for the wild-type (Fig. 5).
Finally, we investigated the role of PA␤N (Fig. 6A) under the same conditions as those used for studying the effect of TolC deletion. The effect of PA␤N on growth inhibition of E. coli MG1655 by carolacton was dependent on the concentration of PA␤N used (Fig. 6B). At concentrations of 20 and 40 &micro;g/ml PA␤N, a maximal inhibition of 59% and 78%, respectively, was found, in comparison to a culture treated with only carolacton. The observations concerning MICs and a PA␤N-mediated growth inhibition by carolacton were reproducible for the tolC-complemented E. coli TolC strain ( Table 2 and Fig. 5, respectively). For comparison, inhibition of growth of E. coli TolC treated with carolacton is shown, which reached a maximum of 90% in comparison to the untreated culture (Fig. 6B). Thus, in E. coli, addition of 40 &micro;g/ml PA␤N, together with carolacton, causes a growth reduction similar to that with treatment with carolacton in a TolC-deficient strain.
The observed growth inhibition characteristics of carolacton-and PA␤N-treated cultures of E. coli TolC and E. coli MG1655 were also reflected in drastic changes in the maximal doubling time (t D ) of cells during exponential growth ( Table 3). The t D of E. coli TolC after treatment with carolacton increased from 25 to Ͼ372 min ( Fig. 4 and 7). A comparable decrease of the doubling time was also observed after coadministration of PA␤N and carolacton to cultures of E. coli MG1655 (t D ,~257 min), supporting the

DISCUSSION
Here we studied the role of TolC, a component of the major multidrug efflux system of E. coli, in its susceptibility to carolacton. To this end, we determined the genome  sequence of the genetically uncharacterized, highly carolacton-susceptible E. coli TolC strain and revealed that it (i) shares the highest nucleotide sequence homology with E. coli MG1655 and (ii) is also phylogenetically reliably placed in a highly supported group that primarily harbors other K-12 strains. Originally, in the 1950s, the chromosome of the wild-type E. coli K-12 was cured from phage , generating E. coli K-12 W1485. E. coli K-12 W1485 was subsequently cured of its F ϩ factor to make MG1655 (36). Thus, as E. coli TolC still contains the phage and a chromosomal copy of the F plasmid, our TolC strain appears to be an ancient prototrophic derivative of the original wild-type E. coli K-12. The profile of MIC resistance of E. coli TolC provided further evidence for an impairment of the efflux function in the mutant strain, rather than a change in the permeability of the outer membrane (25). As the biological function of the TolC OMP is of great scientific interest, tolC deletion mutants of E. coli are often generated anew, elaborately and with varied techniques for every study and in different, often-undescribed genetic backgrounds (25,27,28). The E. coli TolC strain sequenced here has now been thoroughly characterized. It is closely related to the ancestral E. coli wild-type strain K-12 and publicly available and thus could be used as a standard tool in the future.
A strong growth inhibition of E. coli TolC occurred at 0.25 &micro;g/ml (0.54 &micro; M). At this concentration, growth of S. pneumoniae TIGR4 is inhibited in a similar way, indicating a bacteriostatic role of carolacton (24). The same concentration of carolacton caused cell death in biofilms of S. mutans (20). A carolacton epimer, C-9 (R) (epicarolacton), lacked biological activity in all organisms tested so far (22,24). Here, we showed that it was also inactive when testing growth of the highly carolacton-sensitive E. coli TolC strain. The complete loss of biological activity of this carolacton derivative, with a mere inversion of a single stereogenic center at C-9, indicates a specific interaction of carolacton with a cellular target. A target that is present not only in streptococci (24) but also in Gram-negative bacteria like Aggregatibacter (22) and E. coli, and thus might be conserved in the phyla Firmicutes and Proteobacteria.
The data demonstrate that carolacton can enter the Gram-negative cell but is a substrate of the tripartite multidrug efflux pump AcrAB-TolC, the main component of intrinsic antibiotic resistance in Enterobacteriaceae. Its clinical application would therefore require high concentrations, or could be combined with efflux pump inhibitors. Treatment of the E. coli MG1655 with 40 &micro;g/ml of PA␤N, specific for inhibition of the AcrAB-TolC and AcrEF-TolC efflux complexes (16), rendered the strain susceptible to  carolacton in a similar way as the deletion of TolC. The effect of AcrEF for the export of carolacton can be neglected here, as its expression is very low and this exporter has a primary role in cell division (37); hence, deletion of acrEF does not change the antibiotic resistance phenotype of E. coli (11). Interestingly, lower concentrations of PA␤N did not influence the sensitivity to carolacton at all, which is puzzling, because carolacton was provided at 0.25 &micro;g/ml and inhibition by PA␤N has been reported to be competitive (16). The RNA-seq data for E. coli TolC indicated a strong regulatory response upon treatment with carolacton within the first 30 min, where growth is still unaffected, confirming the entry of carolacton into the cell and its likely immediate interaction with an intracellular target. The observed changes involved small regulatory RNAs, a sigma factor, chaperones, heat and cold shock proteins, flagellar components, and membrane transport proteins. The sigma factor F ( 28 in E. coli) is needed for flagellar assembly and motility (38), in accordance with the upregulation of the flagellar components fliL (btd92_01821) or fliJ (btd92_01823). Interestingly, all ncRNAs of the StyR-44 family were strongly upregulated already at the 5-min time point. Styr-44 ncRNAs are found in ribosomal operons located upstream of the 23S rRNA; their expression is dependent on the growth rate, but their specific function is unknown (39). As ncRNAs are known to act as global regulators of gene expression (40), their differential transcript abundance shows a fast and strong global regulatory response to carolacton. Carolacton treatment also caused upregulation of the outer membrane pore proteins NmpC (log 2 FC, 2.72) and PhoE (log 2 FC, 2.85), both of which play a role under heat shock and phosphorus starvation conditions, respectively (41,42). The transcriptome data showed that the molecular target of carolacton may be located within a central metabolic pathway in the cell, and inhibition of this target induces multiple metabolic and transcriptional adaptations.
In conclusion, we found that carolacton efficiently penetrates the Gram-negative cell envelope, and low micromolar concentrations are sufficient for growth inhibition of E. coli, unless it is exported by the tripartite AcrAB-TolC efflux system. Carolacton might potentially be used against Gram-negative bacteria in combination with EPIs.

MATERIALS AND METHODS
Bacterial strains and growth conditions. E. coli strains used for growth experiments (Table 4) were routinely grown under aerobic conditions in Luria-Bertani (LB) broth overnight (o/n) at 37°C (200 rpm). The cultures were then used to inoculate fresh LB medium to an OD 600 of 0.05, which was determined photospectrometrically (Ultrospec 3100 Pro; Amersham Biosciences, Inc.). Cultures with an OD 600 of Ͼ0.5 were diluted in LB broth to below 0.5 in order to maintain the linearity between the measured absorbance and cell density and to achieve the most exact results. The initial culture was then split into equal volumes and supplemented with carolacton, 9(R) epi-carolacton, or PA␤N, or maintained as untreated controls. For cryo-conservation, E. coli was grown in LB o/n, mixed with an equal volume of 50% (vol/vol) glycerol in cryovials, and frozen at Ϫ80°C.
Storage of carolacton, epi-carolacton, and PA␤N. Carolacton and its derivative 9(R)-carolacton were dissolved in methanol or DMSO to a final concentration of 5.3 mM (250 &micro;g/ml) or 2 mM (94.3 &micro;g/ml), respectively, and stored in small aliquots in amber glass vials at Ϫ20°C in the dark. PA␤N (25 mg/ml in H 2 O) was stored at Ϫ20°C and used at final concentrations between 5 and 40 &micro;g/ml, as indicated.
Complementation of E. coli TolC. Chemo-competent cells of E. coli were prepared according to the TSS method described by Chung et al. (43). pIB166 was PCR amplified with Phusion polymerase (NEB) using primers (pIB166_fwd and pIB166_rev), thereby eliminating P 23 (Table 5). Genomic DNA of E. coli K-12 MG1655 served as a template for PCR amplification of the tolC locus (b3035), using primers (tolC_fwd and tolC_rev), additionally introducing flanks homologous to the linearized vector sequence. PCR products were purified with a PCR purification kit (Qiagen, Germany). The PCR-amplified tolC gene was cloned into pIB166 by using the CloneEZ kit (Genescript), and the reaction mix was transformed into E. coli DH5␣. Obtained plasmids were verified by sequencing and subsequently transformed into E. coli TolC. E. coli transformed with pIB166 or its derivatives were grown on LB agar plates or in liquid LB broth containing 20 &micro;g/ml chloramphenicol.
Determination of MIC values. MIC values of selected antibiotics and of carolacton against E. coli and E. coli K-12 MG1655 were determined by 2-fold serial microdilution in LB broth with incubation at 37°C for 20 h, as described previously (44). Antibiotics were tested in the following dilution ranges: ampicillin (32 to 0.25 &micro;g/ml), carolacton (8 to 0.03 &micro;g/ml), cephalotin (32 to 0.25 &micro;g/ml), cefotaxime (1 to 0.078 &micro;g/ml), cerulenin (32 to 0.25 &micro;g/ml), ciprofloxacin (0.25 to 0.0019 &micro;g/ml), chloramphenicol (64 to 0.5 &micro;g/ml), corallopyronin A (32 to 0.25 &micro;g/ml), erythromycin (64 to 0.5 &micro;g/ml), gentamicin (32 to 0.25 &micro;g/ml), kanamycin (8 to 0.03 &micro;g/ml), novobiocin (16 to 0.125 &micro;g/ml), penicillin G (32 to 0.25 &micro;g/ml), phosphomycin (32 to 0.25 &micro;g/ml), rifampin (32 to 0.25 &micro;g/ml), sorangicin (32 to 0.25 &micro;g/ml), sulfamethoxazole (256 to 2 &micro;g/ml), triclosan (1 to 0.078 &micro;g/ml), trimethoprim (2 to 0.015 &micro;g/ml), and vancomycin (256 to 2 &micro;g/ml), if not indicated otherwise. Corallopyronin A and sorangicin were kindly provided by Rolf Jansen (HZI, Braunschweig). Antibiotics were purchased from Sigma-Aldrich (Steinheim, Germany) or Carl Roth GmbH (Karlsruhe, Germany). MICs were the lowest concentrations that did not yield visible bacterial growth. The cell count of the initial inoculum was 5 ϫ 10 5 CFU/ml, which was confirmed by plating of serial cell dilutions and counting of CFU. MICs were confirmed in at least two independent experiments.  (54,55). False-discovery rate (FDR)-adjusted P values were calculated according to methods described previously (56). FDR values of Ͻ0.01 were considered significant. Heat maps were generated for genes that showed a log 2 FC of ՆϮ2 for at least one time point (FDR, Յ0.01), log 2 FC values of transcript abundance obtained with edgeR were used as input for the heatmap.2 function of the R package gplots (v.2.15.0) (57). Whole-genome-based phylogenomic analyses. To elucidate the phylogenetic positioning of strain TolC, and given its high sequence similarity to strain E. coli K-12 MG1655, a member of phylogroup A (58), a corresponding reference data set was defined. The latter included all 32 members of phylogroup A, according to methods described previously (58), and was further complemented by four recently genome-sequenced strains that had been found to be highly similar to TolC (accession numbers NZ_CP011495, NZ_CP014225, NZ_CP014348, and NZ_CP016018). Two whole-genome-based phylogenomic analyses were conducted using the genome BLAST distance phylogeny approach (59) in its latest version (29). The first analysis was based on the nucleotide data restricted to genes, whereas the second one used protein data only. Coding regions were determined via Prodigal under default settings (60). All pairwise intergenomic distances were calculated with GBDP under established settings (58), i.e., using the trimming algorithm, distance formula d 5 , and an E value cutoff of 10 Ϫ8 . A total of 100 pseudobootstrap replicates were calculated per distance and later used for the inference of branch support values (61). Phylogenetic trees were inferred from the original and pseudobootstrapped distance matrices by using FastME 2.1.4 (62) under the SPR branch-swapping option and rooted using the midpoint method (63).
Accession number(s). The genome sequences of E. coli TolC were deposited in NCBI's GenBank (67) under accession number CP018801.1. Raw and processed RNA-seq data were deposited in NCBI's Gene Expression Omnibus (GEO) database (68) and are accessible through GEO Series accession number GSE93125.