The inactivation of tolC sensitizes Escherichia coli to perturbations in lipopolysaccharide transport

Summary

The Escherichia coli outer membrane channel TolC complexes with several inner membrane efflux pumps to export compounds across the cell envelope. All components of these complexes are essential for robust efflux activity, yet E. coli is more sensitive to antimicrobial compounds when tolC is inactivated compared to the inactivation of genes encoding the inner membrane drug efflux pumps. While investigating these susceptibility differences, we identified a distinct class of inhibitors targeting the core-lipopolysaccharide translocase, MsbA. We show that tolC null mutants are sensitized to structurally unrelated MsbA inhibitors and msbA knockdown, highlighting a synthetic-sick interaction. Phenotypic profiling revealed that tolC inactivation induced cell envelope softening and increased outer membrane permeability. Overall, this work identified a chemical probe of MsbA, revealed that tolC is associated with cell envelope mechanics and integrity, and highlighted that these findings should be considered when using tolC null mutants to study efflux deficiency.

Graphical abstract

Highlights

• •tolC null mutants are sensitized to structurally unrelated MsbA inhibitors
• •A synthetic-sick interaction exists between tolC and msbA
• •The benzophenones represent a distinct class of MsbA inhibitors
• •Inactivating tolC induces cell envelope softening and increased permeability

Biochemistry; Microbiology; Applied microbiology

Introduction

Diderm bacteria represent a unique challenge for compound permeation and transport since they possess two membranes with orthogonal properties.¹ In Escherichia coli, the inner membrane (IM) is a phospholipid bilayer permeable to amphiphilic and hydrophobic compounds yet restrictive to polar and highly charged.^2,3 In contrast, the outer membrane (OM) is atypical, with an unusual asymmetric lipid distribution that impedes entry; the outer leaflet comprises lipopolysaccharide (LPS) and the inner leaflet phospholipids.¹ Polyanionic LPS molecules are often bridged by lateral interactions with divalent cations, imparting integrity to the OM and reducing the permeation of large and hydrophobic compounds.¹ Small hydrophilic compounds can diffuse through the OM via the OmpA, OmpF, and OmpC porins, which form a relatively static network.⁴ OM proteins such as OmpA, Braun’s lipoprotein (Lpp), and peptidoglycan lipoprotein (Pal) stabilize the OM through interactions with the peptidoglycan.¹ In addition to the barrier functions of the OM, this structure provides physical strength to the cell envelope, compensating for the relatively thin layer of peptidoglycan in gram-negative bacteria.⁵ The OM also plays a role in determining the shape of E. coli.⁶

Importantly, the impermeable nature of the OM augments the exporting activities of drug efflux pumps, which collectively render E. coli intrinsically resistant to a broad spectrum of antimicrobial agents.^2,7 In E. coli, an extensive network of conserved drug efflux pumps contributes to antimicrobial detoxification, and these proteins can be broadly categorized into two main groups: (1) single-component efflux pumps residing in the IM and (2) multicomponent drug efflux systems that span the entire cell envelope.⁸ The latter comprises an efflux pump driving export from the IM, periplasmic adaptor proteins, and a nonspecific OM channel for transport across the OM. Eight different E. coli efflux pumps are known to complex with the OM channel TolC,⁹ and tolC null mutants are commonly used efflux-deficient strains that display increased sensitivity to a wide variety of antimicrobial agents.¹⁰ TolC is associated with several other cellular functions, including metabolite extrusion and acid tolerance.¹¹ Additionally, the protein serves as a component of type I secretion systems, enabling virulence factor secretion.¹¹

We recently reported the generation of a mutant strain, Efflux KnockOut-35: EKO-35, lacking 35 IM efflux pumps comprising the E. coli drug efflux network, including all pumps complexing with TolC.⁹ We anticipated EKO-35 would be equally, if not more, susceptible to antimicrobial agents than a tolC null mutant (harboring an in-frame marker-less deletion) since this strain lacks all efflux pumps complexing with TolC and all single-component efflux pumps. Unexpectedly, while both strains were highly susceptible to a broad range of antimicrobials, the ΔtolC mutant was more susceptible than EKO-35 in several instances.⁹

Here, we explored the mechanistic basis of these susceptibility differences, identifying an inhibitor of MsbA, the ATP-binding cassette (ABC) transporter responsible for “flipping” core-LPS to the periplasmic leaflet of the IM.¹ We subsequently uncovered a synthetic-sick interaction between msbA and tolC and revealed that the cell envelope of ΔtolC exhibits softening and increased permeability, which does not appear to be associated with known drug efflux pumps that form complexes with this OM channel. Collectively, this work reveals that tolC inactivation sensitizes E. coli to perturbations in LPS transport and impacts cell envelope mechanics and integrity. Our findings highlight that these factors should be considered when using tolC mutants to study efflux since increased sensitivity is not necessarily attributable to decreased efflux. Furthermore, we propose that the tolC-associated synthetic-sick interaction could be exploited to identify new cell envelope biogenesis machinery inhibitors.

Results

Exploring the mechanistic basis of antimicrobial susceptibility differences observed between efflux-deficient strains

As described, we reported susceptibility differences between efflux-deficient E. coli K-12 (BW25113) strains, ΔtolC and EKO-35, despite the latter lacking the IM efflux pumps that complex with TolC.⁹ Compared to EKO-35, we reasoned that a mutant lacking only the IM efflux pumps complexing with TolC would be a more suitable comparison to a tolC null mutant since single-component efflux pumps are still produced. Therefore, using a combination of λ-Red recombineering and CRISPR-Cas9-mediated counter selection, here we inactivated eight genes (acrB, acrD, acrE, mdtF, mdtB, emrY, emrB, macB) encoding efflux pumps that complex with TolC in E. coli K-12 (BW25113),¹² generating the tripartite efflux (TE) mutant. Three non-synonymous mutations, identified via long-read next-generation sequencing, were repaired in the TE mutant to ensure they were not responsible for antimicrobial susceptibility differences (Table S1). Finally, Illumina sequencing confirmed that the repaired TE mutant and ΔtolC were isogenic except for disruptions in the efflux-associated genes, as expected (Table S1).

Susceptibility testing revealed that both efflux-deficient strains exhibited increased sensitivity to a broad range of antimicrobial agents (Figure S1; Table S2); however, while ΔtolC was highly sensitized to several of the uncharacterized synthetic antimicrobial compounds,⁹ the TE mutant exhibited high-level resistance, consistent with our previous study using the EKO-35 efflux-deficient strain (Figure 1A; Table S2). In most cases, increasing OM permeability by introducing a non-selective FhuA-derivatized pore^9,10,13 did not impact these susceptibility differences, suggesting they are not solely attributable to increased OM permeability in ΔtolC (Figure 1A).

Figure 1.

The inactivation of tolC impacts antimicrobial sensitivity differently than the disruption of genes encoding inner membrane efflux pumps

(A) Heatmap depicting the susceptibility of unporinated (−) and porinated (+) wild-type E. coli K-12, TE mutant, and ΔtolC strains to a series of synthetic compounds. The minimum inhibitory concentration (MIC) values were log₂ transformed and normalized to 100% for each compound tested, where orange on the heatmap represents the highest value and white represents the lowest value.

(B) Spontaneous mutations conferring resistance in ΔtolC to synthetic compound #20 (labeled with a red asterisk in panel (A)). The amino acid mutations localized to or near the MsbA elbow helix (labeled and colored light blue). Three different MsbA conformations are shown: the inward-facing open (PDB ID: 7SEL), occluded (PDB ID: 7BCW), and outward-facing open conformation (PDB ID: 3B60). The protomers of MsbA are colored pink and green, with the mutations highlighted in red. TM, transmembrane helix.

(C) Susceptibility testing of the quinoline MsbA inhibitors against the K-12, ΔtolC, and TE mutant strains. The chemical structures of each MsbA inhibitor are shown. Growth was normalized to 100%, where orange represents the greatest OD_600nm and white represents the lowest OD_600nm. Related to Table 1.

tolC inactivation sensitizes E. coli to MsbA inhibition

To gain insight into the mechanistic basis of this phenomenon, we next sought to generate spontaneous mutations conferring resistance to the synthetic compounds with increased ΔtolC activity (Figure 1A). In the presence of a benzophenone compound, synthetic compound #20 (syn. #20), suppressor mutations in ΔtolC emerged at a frequency of 1.1 × 10⁻⁷. Next-generation sequencing of different spontaneous mutants identified several nonsynonymous mutations that localized to the msbA gene, and these amino acid substitutions clustered within the same region of the protein (Figure 1B). All the suppressor mutants exhibited high-level resistance to the benzophenone compound, with >16-fold increases in the minimum inhibitory concentrations (MICs) (Table 1).

Table 1.

MsbA inhibitor susceptibility testing

Strain	Pore	MsbA inhibitor	Class	MIC (μg/mL)	Fold change
K-12	−	Synthetic compound #20	Benzophenone	>80	>256
ΔtolC	−	Synthetic compound #20	Benzophenone	0.3125	–
EKO-35	−	Synthetic compound #20	Benzophenone	>80	>256
TE	−	Synthetic compound #20	Benzophenone	>80	>256
TE + tolC disrupted (insertion of 3 tandem stop codons)	−	Synthetic compound #20	Benzophenone	0.3125	1
K-12	+	Synthetic compound #20	Benzophenone	>80	>256
ΔtolC	+	Synthetic compound #20	Benzophenone	0.3125	1
EKO-35	+	Synthetic compound #20	Benzophenone	>80	>256
TE	+	Synthetic compound #20	Benzophenone	>80	>256
ΔbamB	−	Synthetic compound #20	Benzophenone	>5	>16
ΔbamB ΔtolC	−	Synthetic compound #20	Benzophenone	0.3125	1
Spontaneous suppressor mutants
ΔtolC MsbAA30V	−	Synthetic compound #20	Benzophenone	>5	>16
ΔtolC MsbAF13V	−	Synthetic compound #20	Benzophenone	>5	>16
ΔtolC MsbAG94A	−	Synthetic compound #20	Benzophenone	>5	>16
ΔtolC MsbAD6Y	−	Synthetic compound #20	Benzophenone	>5	>16
ΔtolC MsbAD6A	−	Synthetic compound #20	Benzophenone	>5	>16
Target overexpression and knockdown
ΔtolC pGDP2	−	Synthetic compound #20	Benzophenone	0.3125	1
ΔtolC pGDP2:msbA	−	Synthetic compound #20	Benzophenone	>5	>16
K-12 Empty vector CRISPRi	−	Synthetic compound #20	Benzophenone	>5	>16
K-12 msbA CRISPRi	−	Synthetic compound #20	Benzophenone	5	16
ΔtolC Empty vector CRISPRi	−	Synthetic compound #20	Benzophenone	0.3125	1
ΔtolC msbA CRISPRi	−	Synthetic compound #20	Benzophenone	0.0049	−64
TE Empty vector CRISPRi	−	Synthetic compound #20	Benzophenone	>5	>16
TE msbA CRISPRi	−	Synthetic compound #20	Benzophenone	5	16
Other MsbA inhibitors
K-12	−	G717	Benzophenone	>80	>128
ΔtolC	−	G717	Benzophenone	0.625	–
TE	−	G717	Benzophenone	>80	>128
K-12	−	G662	Benzophenone	>40	>32
ΔtolC	−	G662	Benzophenone	1.25	–
TE	−	G662	Benzophenone	>40	>32
K-12	+	G662	Benzophenone	>40	>32
ΔbamB	−	G662	Benzophenone	>40	>32
K-12	−	G615	Benzophenone	>80	>256
ΔtolC	−	G615	Benzophenone	0.3125	–
TE	−	G615	Benzophenone	1.25	4
K-12	−	G332	Quinoline	2.5	32
ΔtolC	−	G332	Quinoline	0.0781	–
TE	−	G332	Quinoline	1.25	16
ΔtolC MsbAA30V	−	G332	Quinoline	0.0391	−2
ΔtolC MsbAF13V	−	G332	Quinoline	0.1563	2
ΔtolC MsbAG94A	−	G332	Quinoline	0.0781	1
ΔtolC MsbAQ316K	−	G332	Quinoline	0.0391	−2
ΔtolC MsbAD6A	−	G332	Quinoline	0.0195	−4
ΔtolC MsbAR190L	−	G332	Quinoline	0.0781	1
K-12	−	G913	Quinoline	10	128
ΔtolC	−	G913	Quinoline	0.0781	–
TE	−	G913	Quinoline	0.3125	4
ΔtolC MsbAA30V	−	G913	Quinoline	0.0391	−2
ΔtolC MsbAF13V	−	G913	Quinoline	0.1563	2
ΔtolC MsbAG94A	−	G913	Quinoline	0.0781	1
ΔtolC MsbAQ316K	−	G913	Quinoline	0.0391	−2
ΔtolC MsbAD6A	−	G913	Quinoline	0.0195	−4
ΔtolC MsbAR190L	−	G913	Quinoline	0.0781	1

Strains were assessed in technical triplicate. For each strain, the minimum inhibitory concentration (MIC) values were compared to the ΔtolC mutant to determine the fold change. Pore, fhuA ΔC/Δ4L, representing a permeabilized outer membrane; K-12, the wild-type strain; ΔtolC, tolC inactivated mutant; EKO-35, Efflux KnockOut 35 mutant; TE, Tripartite Efflux mutant.

To further investigate whether MsbA was the target, we assessed the effect of target gene overexpression and knockdown. Consistent with MsbA being the target of the benzophenone, the overexpression of msbA in ΔtolC (pGDP2:msbA) conferred high-level resistance (>16-fold increase in the MIC) (Table 1). Additionally, CRISPR interference (CRISPRi)-mediated msbA knockdown, confirmed using real-time quantitative reverse-transcription PCR (Figure S2), further sensitized ΔtolC, increasing sensitivity >64-fold (Table 1). Additionally, the wild-type and TE mutant strains were sensitized to syn. #20 when msbA levels were reduced (Table 1).

Finally, tolC was disrupted in the TE mutant using CRISPR-Cas9-mediated counterselection and the introduction of three tandem stop codons, which increased the susceptibility of the TE mutant to the same level as the ΔtolC mutant, revealing that the syn. #20 susceptibility differences are directly associated with TolC (Table 1).

MsbA is an essential translocase that “flips” nascent LPS to the outer leaflet of the IM; as such, the protein is an attractive antibacterial target.^14,15,16 There are two known classes of MsbA inhibitors: tetrahydrobenzothiophene 1 (TBT-1)¹⁷ and a quinoline-containing compound series.^18,19 To further probe the relationship between TolC and MsbA, we conducted susceptibility testing with members of the quinoline series (G332 and G913), revealing that ΔtolC was also hypersensitized to these structurally dissimilar MsbA inhibitors compared to the TE mutant (Figure 1C; Table 1).

Taken together, these data indicate that the benzophenone compound likely targets MsbA and highlights a striking association between tolC and MsbA, which appears to be distinct from the drug efflux pumps known to form complexes with this OM channel.

The benzophenones represent a distinct class of MsbA inhibitors

To further substantiate MsbA as the target of the benzophenone compound, ATPase inhibition was assessed using purified MsbA reconstituted into amphipols, as previously described.^18,19 The compound exhibited dose-dependent MsbA inhibition, with a half-maximum inhibitory concentration of 0.24 μM (Figure 2A), which was comparable to the minimum inhibitory concentration in ΔtolC (0.80 μM). We then profiled benzophenone analogs, revealing similar activity against E. coli MsbA. Yet, we did not identify any compounds with increased potency in the biochemical assay (Figure 2B). Attempts to replace the biaryl ketone reduced the activity of these compounds, highlighting the importance of this moiety (Figure S3). While the benzophenones exhibit poor antibacterial activity against wild-type E. coli strains, increased activity was observed against E. coli expressing the imp4213 allele of lptD^19,20 (Table S3), enabling phenotypic analysis. The benzophenones induced membrane defects with vesicle-like invaginations due to core-LPS accumulation (Figure 2C), consistent with the phenotypes observed with the quinoline class of MsbA inhibitors and an arabinose-inducible conditional knockout of msbA.¹⁹

Figure 2.

Biochemical and phenotypic characterization of the benzophenone MsbA inhibitors

(A) Synthetic #20 dose-response curve showing inhibition of MsbA-ATPase activity using purified E. coli MsbA reconstituted into amphipols. Data points represent the average ± standard deviation of three independent replicates.

(B) Benzophenone analogs exhibit comparable activity to synthetic compound #20.

(C) Representative thin-section transmission electron micrographs of strain CFT073 lptD (imp4213) treated with 20 μM of each inhibitor for 3 h. Black arrows indicate areas of vesicle-like formations of the IM, and orange arrows indicate membrane stacking due to LPS accumulation. Indentations of the cell envelope were also observed in some cells. The scale bar represents 0.2 μm for the untreated cells and 0.1 μm for the treated cells.

The benzophenones are structurally distinct from other described MsbA inhibitors, which bind to the transmembrane regions of this protein.^18,19,21,22 In contrast, the nonsynonymous resistance-conferring mutations identified within this study localized within or adjacent to the MsbA cytoplasmic elbow helices running parallel to the plane of the IM, a region of the protein that is proposed to facilitate ligand entry (Figure 1B).^23,24 In addition to recognizing physiological lipid substrates, MsbA is proposed to interact with drugs such as daunorubicin,^23,25,26,27 and the daunorubicin-binding site is proposed to localize near the elbow helices.²³ However, a chequerboard assay revealed no interactions between the benzophenones and daunorubicin (Figure S4A). In contrast, additive effects were observed between the benzophenones and a quinoline MsbA inhibitor (Figures S4B and S4C), supporting MsbA inhibition via the benzophenone and revealing that the binding mode and inhibitory actions of the two inhibitor types are not antagonistic or synergistic. Additionally, the mutations conferring resistance to the benzophenones (Figure 1B) did not provide cross-resistance to the quinolines (Table 1).

In summary, we have identified a distinct structural class of MsbA inhibitors that may exhibit different binding modes than previously described inhibitors.

Identification of a synthetic-sick interaction between tolC and msbA

Since the inactivation of tolC rendered E. coli hypersusceptible to two different classes of MsbA inhibitors (Figure 1), we hypothesized that a negative synthetic interaction existed between tolC and MsbA. Negative synthetic, or synthetic-sick, interactions refer to combinations of perturbations (e.g., either genetic or chemical) that induce a negative fitness defect different from the anticipated phenotype based on the effects of single perturbations.²⁸ As anticipated, CRISPRi-mediated knockdown of msbA revealed a synthetic-sick interaction between msbA and tolC, where the growth of the tolC-inactivated mutant was greatly reduced in response to msbA knockdown compared to the TE mutant and the wild-type strain (Figure 3A). Phenotypic analysis revealed that both the chemical inhibition of MsbA and the genetic repression of msbA induced severe morphological defects in ΔtolC, with filament formation, bulging, and tadpole-shaped defects emerging at the cell poles (Figures 3B–3F and S5). In summary, tolC inactivation sensitizes E. coli to both the chemical inhibition of MsbA and the genetic knockdown of msbA, identifying a synthetic-sick interaction.

Figure 3.

Identification of a synthetic-sick interaction between tolC and msbA

(A) CRISPRi-mediated msbA knockdown in the wild-type K-12, ΔtolC, and TE mutant strains propagated in the presence of increasing concentrations of the inducer anhydrotetracycline. The empty plasmid (pFD152) represents the negative control. Data points represent the average OD_600nm ± standard deviation of three technical replicates and are representative of three independent experiments.

(B–F) Scanning electron micrographs of (B) untreated ΔtolC and (C) K-12 during the mid-exponential growth phase. (D, E) CRISPRi-mediated repression of msbA (D) in ΔtolC and (E) K-12. (F) Chemical inhibition of MsbA in the presence of 0.5x the MIC of synthetic compound #20 in ΔtolC. Related to Figure S5. Scale bars represent 10 μm.

TolC impacts the permeability and mechanical properties of the outer membrane

Based on the identified synthetic-sick interaction (Figure 3), we speculated that tolC inactivation could induce cell envelope defects that sensitize the strain to perturbations in LPS transport. As such, we investigated OM integrity by measuring the uptake of the fluorescent probe 1-N-phenylnaphthylamine (NPN).²⁹ This hydrophobic dye is ordinarily excluded by the OM, and disruption of the OM barrier function enables entry into the phospholipids, increasing fluorescence. Consistent with OM defects, NPN rapidly accumulated in the phospholipids of ΔtolC compared to the TE mutant (Figure 4A). We reasoned that the increase in OM permeability could explain the increased sensitivity of ΔtolC to the benzophenones compared to the TE mutant and the wild-type strain (Figure 1); however, we noted the susceptibility of the wild-type and TE mutant strains was not impacted by OM porination (Figure 1A; Table 1). To investigate further, we also profiled a bamB-inactivated mutant exhibiting a permeabilized OM³⁰ (Table 1) and wild-type cells treated with EDTA³¹ (Figure S4D). Overall, increased OM permeability, as assessed using two different permeabilized strains and EDTA treatment, did not affect the sensitivity of E. coli K-12 (BW25113) to the benzophenones.

Figure 4.

The inactivation of tolC reduces outer membrane integrity, impacting cell surface morphology

(A) NPN accumulation assays in the ΔtolC and TE mutants. Data points are background-subtracted averages of three technical replicates and are representative of two independent experiments.

(B) Cell length measurements of 325 cells from n = 3 biological replicates of scanning electron micrographs of the wild-type K-12, ΔtolC, and TE mutant strains. p values were calculated using a two-tailed Student’s t test (∗∗∗∗p < 0.0001, p = 9.70 × 10⁻⁶ and 4.50 × 10⁻⁶ for ΔtolC and the TE mutant, respectively).

(C–E) Scanning electron micrographs of (C) wild-type K-12, (D) the TE mutant, and (E) ΔtolC during the mid-exponential phase of growth. A subset of the ΔtolC cells displayed phenotypic heterogeneity, with regions of undulations and smoothness, which are indicated by the orange and yellow arrows, respectively. Images are representatives of three biological replicates. Scale bars represent 1 μm. Related to Figure S6.

Next, the morphology of the ΔtolC mutant was characterized using scanning electron microscopy (SEM), revealing that the surfaces of ΔtolC cells were relatively “smoother” than the undulated cell surface of the wild-type and TE mutant strains (Figures 4C–4E and S6). Both efflux-deficient mutants were also moderately, yet significantly, longer compared to the wild-type cells (Figure 4B). Overall, it was not possible to statistically assess the extent of smoothness across the ΔtolC population since there was a high level of phenotypic heterogeneity across the cell population and even within a single bacterial cell (Figure 4E). However, we did not identify smooth phenotypes among the wild-type or TE mutant cells. While these phenotypic observations could be a consequence of the cell preparation method used for SEM, all strains were handled equally, revealing cell surface structural differences between the strains.

To further analyze the morphological structural changes between these strains in aqueous conditions, and to assess the contribution of TolC to cell envelope mechanics, single-cell atomic force microscopy (AFM)^32,33 was employed. High-resolution images of the cell surface revealed that the wild-type cells exhibited high-frequency undulations (Figures 5A, 5B, and S7), consistent with the SEM findings (Figure 4C). In contrast, ΔtolC cells appeared to be either very smooth or exhibited low-frequency undulations (Figures 5A, 5B, and S7). The TE mutant exhibited combinations of both characteristics (Figures 5A, 5B, and S7). The mean roughness values of the wild-type strain exhibited a much wider dispersion; however, these differences were not found to be significant (Figure S8B). Finally, we used AFM-based indentation experiments to quantify cell envelope mechanical properties, or OM stiffness (also called tensile elasticity).^34,35 The Young’s modulus (E) and the spring constant (k) for the wild-type strain were consistent with previous studies^34,36,37 (Figures 5C–5F). Supporting the notion that ΔtolC exhibits a compromised OM, cell stiffness was shown to be significantly reduced in ΔtolC cells, with decreases in both E and k values (Figures 5C–5F and S9). In summary, our results show that tolC impacts the integrity and mechanical properties of the OM.

Figure 5.

Inactivation of tolC impacts surface morphology and envelope mechanics of single live cells

(A) Representative low-resolution AFM height images of the wild-type K-12, ΔtolC, and TE mutant strains during the mid-exponential growth phase. Images were obtained in PBS, in quantitative imaging (QI) mode (8 μm × 8 μm, color scale: 800 nm).

(B) High-resolution height images. Images were obtained in QI mode (200 nm × 200 nm, color scale: 25 nm). Representative distributions of (C) Young’s modulus and (D) spring constant (k) values obtained by AFM force–volume measurements in PBS across the surface of cells.

(E and F) Statistical analysis performed for each strain shows that tolC-inactivation significantly reduces cell envelope stiffness compared to the K-12 (p = 9.13 × 10⁻⁴ for Young’s modulus and 3.92 × 10⁻² for spring constant) and TE mutant strains (p = 1.59 × 10⁻⁵ for Young’s modulus and 1.82 × 10⁻³ for spring constant). Boxplots show the individual (E) Young’s modulus and (F) spring constant values obtained from n = 12 cells for K-12, n = 13 cells for ΔtolC, and n = 14 cells for the TE mutant. p values were calculated using Tukey’s multiple comparison test (∗p ≤ 0.05, ∗∗p ≤ 0.01, ∗∗∗p ≤ 0.001, n.s. = not significant). Shown are mean values (stars), 25% and 75% quartiles (boxes), medians (center line of boxes), and the standard deviation (whiskers). Related to Figures S7–S9.

Discussion

The permeation and transport of compounds across the E. coli cell envelope is a complex phenomenon due to two membranes with orthogonal properties.³⁸ To facilitate the extrusion of toxic substances across the cell envelope, several IM efflux pumps complex with the OM channel TolC. In these instances, TolC essentially functions as a nonspecific passive conduit for expulsion across the OM. Supporting our previous findings,⁹ we demonstrate that tolC inactivation affects antimicrobial susceptibility differently from the inactivation of genes encoding the IM efflux pumps that form complexes with TolC. Specifically, a tolC null strain was more sensitive to a panel of poorly characterized synthetic compounds than mutants lacking TolC-dependent IM pumps (Figure 1). While investigating the molecular basis of these susceptibility differences, we identified benzophenone-containing inhibitors that target the LPS transporter, MsbA.

MsbA is an essential ABC transporter responsible for translocating LPS to the outer leaflet of the IM; MsbA antagonism induces LPS accumulation at the cytoplasmic interface of the IM, causing perturbations in cellular ultrastructure.^14,15,39 MsbA is an integral membrane protein, forming a homodimer alternating between inward-facing and outward-facing conformations during transport.⁴⁰ Two different MsbA antagonist scaffolds have been identified: the quinoline series^18,19 and TBT-1.¹⁷ Both classes bind the MsbA transmembrane region yet occupy adjacent and separate binding pockets.²² Interestingly, TBT-1 stimulates ATPase activity and induces a collapsed inward-facing MsbA conformation, targeting the LPS substrate binding site.²² The quinolines abolish ATPase hydrolysis,¹⁸ interacting with the inward-facing MsbA conformation, preventing the closure of the protein by displacement of the nucleotide-binding domains.²² In contrast, the benzophenones identified in this study appear to target a binding pocket localizing to the elbow helix of MsbA (Figure 1). One of the nonsynonymous mutations conferring high-level resistance to syn. #20 (Table 1) localizes to transmembrane helix 4 of MsbA (R₁₉₀L). In the occluded and outward-facing conformations, this amino acid comes into proximity with the proposed binding site (Figure 1), supporting a conformation-dependent binding site. However, we acknowledge that these mutations may not reside in the binding pocket and could instead affect the conformation of MsbA, reducing the drug’s affinity. Nonetheless, despite the binding site remaining putative, the benzophenones demonstrated dose-dependent inhibition of MsbA, inducing LPS accumulation, and are structurally distinct from existing classes of MsbA inhibitors.

Notably, despite MsbA being considered essential for growth,^14,39 the benzophenones do not exhibit growth-inhibitory activity against wild-type E. coli strains (Table 1). We consequently revealed that intrinsic resistance mediated by the OM and active efflux does not appear to be responsible for the weak activity against wild-type E. coli strains. Indeed, the efflux-deficient TE and EKO-35 strains were insensitive to these inhibitors (Figure 1). Additionally, permeabilizing the OM, either through the introduction of a large nonspecific FhuA-derivatized pore,¹⁰ through bamB inactivation, or using cells treated with EDTA, did not sensitize E. coli to the benzophenone-containing compound synthetic #20 (Table 1; Figure S4D). In contrast, antibacterial activity was observed in E. coli strains expressing the imp4213 allele of lptD, which increases the abundance of phospholipids in the OM, increasing OM permeability.^20,41 However, the increased antibacterial activity may be the result of collateral damage due to disruptions in both MsbA-mediated LPS translocation and LptD/E-mediated LPS assembly at the cell surface.^42,43

Compared to the quinolines, which are potent MsbA antagonists exhibiting low nM activity,^18,19 the benzophenones are relatively weak inhibitors (low μM activity) (Figure 2). We postulate that the level of MsbA antagonism provided by these inhibitors is not sufficient to inhibit growth in wild-type strains. Indeed, we also note that while CRISPRi-mediated knockdown of msbA—which reduced the msbA transcript levels by ∼50% compared to the basal levels (Figure S2)—induced a lethal phenotype in ΔtolC, the growth of the wild-type strain remained largely unaffected (Figure 3A). Thus, we hypothesize that antagonists of MsbA must be potent (in the low nM range), or msbA transcript levels must be reduced >50%, to elicit growth inhibition. We subsequently hypothesized that the ΔtolC mutant exhibits a compromised OM, which sensitizes the strain to MsbA antagonism. Consequently, a synthetic-sick interaction was identified between tolC and msbA (Figures 1C and 3). The latter indicates that the susceptibility differences (Figure 1) observed between ΔtolC and the TE mutant are not merely a reflection of altered drug permeation.

Lipoproteins and LPS within the OM of E. coli confer stiffness and strength to the cell envelope.⁴⁴ As such, when the composition of the OM is altered, as exemplified by E. coli strains expressing the imp4213 allele of lptD, OM stiffness is reduced.⁴⁴ Similarly, mutants lacking the abundant OM protein OmpA⁴⁴ and compromising the function of Braun’s lipoprotein, Lpp, also decrease cell envelope stiffness, the latter of which is attributed to the protein’s ability to link peptidoglycan to the OM, controlling the depth of the periplasm.^34,44 Consistent with the notion that tolC null mutants exhibit compromised cell envelopes, single-cell AFM revealed significant alteration in cell envelope mechanics, with reduced stiffness (Figure 5). Additionally, high-resolution SEM in conjunction with AFM height images revealed that the surface of ΔtolC cells is heterogeneous, with relatively “smooth” regions that were not observed in the wild-type or TE mutant strains (Figure 4). Taken together, our findings highlight that disruptions in tolC impact the integrity of the cell envelope, which we believe sensitizes E. coli to perturbations in LPS transport.

Several factors could underlie this phenomenon. TolC is intricately associated with Enterobacteriaceae physiology, being implicated in the export of potentially toxic metabolites that could accumulate, inducing membrane damage.^8,11,45,46 Indeed, enterobactin accumulation under iron-limited conditions in tolC null mutants leads to a metabolic shutdown, cell envelope stress, and severe growth defects.^8,11,45,46 Cell envelope stress and loss of integrity could underlie the increased sensitivity of ΔtolC; however, our phenotypes are evident in iron-replete, optimal growth conditions. Additionally, metabolite extrusion is associated with the efflux pumps absent in the TE mutant, and the mechanics of the cell envelope in the TE mutant are comparable to that of the wild-type strain (Figure 5). An alternative hypothesis is that tolC inactivation disrupts OM asymmetry and composition; TolC mutants exhibit decreases in the abundance of the OmpF and OmpC porins, whereas the OmpA abundance is only slightly reduced.⁴⁷ OmpF and OmpC interact with the OM lipoprotein MlaA, a component of the Mla system, preventing phospholipid accumulation in the outer leaflet of the OM and maintaining asymmetry.⁴⁸ As such, reductions in OmpC and OmpF could interfere with the Mla system. Overall, the role of TolC in the physiology of E. coli under iron-replete and optimal growth conditions remains to be established and will be the focus of future studies.

In summary, our findings reveal that tolC inactivation impacts cell envelope integrity, reduces OM stiffness, and sensitizes E. coli to perturbations in LPS transport. Additionally, we report the identification of a distinct class of inhibitors targeting MsbA, which could be optimized to improve activity against wild-type E. coli and could also be used as probes to study the function of MsbA since they target a region of the protein distinct from other known inhibitors. Interestingly, the identified synthetic-sick interaction appears distinct from TolC-associated drug-effluxing activities. Such an observation, in combination with the cell envelope defects induced by loss of tolC, suggests that these factors should be considered when using tolC null strains to study efflux deficiency. Finally, our findings reveal that the susceptibility differences observed between tolC null mutants and the efflux-deficient EKO-35 and TE mutant strains could be exploited to identify new inhibitors of cell envelope biogenesis. Indeed, in addition to syn #20, we have identified several synthetic compounds that exhibit increased activity against the tolC null mutant (Figure 1). Preliminary studies revealed msbA overexpression did not impact the susceptibility of ΔtolC to these compounds (Figure S10), indicating that MsbA may not be the target. We speculate that these inhibitors could target other components of the cell envelope biosynthesis machinery and that further tolC-associated synthetic interactions exist, which will be the focus of future studies.

Limitations of the study

The mechanistic work described in this study was mostly undertaken in E. coli K-12; future studies should ascertain whether our findings are also observed in other strains of E. coli.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
E. coli K-12 str. BW25113	Baba et al.12	N/A
E. coli CFT073	ATCC	ATCC 700928
E. coli CFT073 lptD(imp4213)	Ho et al.18	N/A
E. coli EKO-35	Teelucksingh & Thompson et al.9	N/A
E. coli pore	Teelucksingh & Thompson et al.9	N/A
E. coli TE mutant	This paper	N/A
E. coli ΔtolC	Baba et al.12	N/A
E. coli ΔbamB	Baba et al.12	N/A
E. coli TUC01	Thomason et al.49	N/A
Chemicals, peptides, and recombinant proteins
Ampicillin	Bioshop	Cat# AMP201
Kanamycin	Bioshop	Cat# KAN201
Spectinomycin	Bioshop	Cat# SPE201
Chloramphenicol	Sigma-Aldrich	Cat# C0378
Anhydrotetracycline	Sigma-Aldrich	Cat# J66688.MA
Synthetic compound #20	Fisher Thermo Scientific	Cat# RH01772SC
G332	Genentech	N/A
G913	Genentech	N/A
G662	Genentech	N/A
G717	Genentech	N/A
G511	Genentech	N/A
G615	Genentech	N/A
Diaminopimelic acid (DAP)	Sigma-Aldrich	Cat# D1377
Gentamicin	BioBasic	Cat# GB0217
Vancomycin	Bioshop	Cat# VAN990
1-N-phenylnaphthylamine (NPN)	Sigma-Aldrich	Cat# 104043
Carbonyl cyanide m-chlorophenyl hydrazone (CCCP)	Sigma-Aldrich	Cat# C2759
Bovine serum albumin	HyClone™	Cat# SH3057402
Critical commercial assays
2× GB-AMP™ PaCeR™ polymerase master mix	GeneBio Systems	Cat# PCR-002-01
Taq 2× polymerase master mix	FroggaBio	Cat# FBTAQM
GeneJET PCR purification kit	Thermo Fisher Scientific	Cat# K0701
GeneJET Plasmid Miniprep kit	Thermo Fisher Scientific	Cat# K0502
PureLink™ Genomic DNA Mini Kit	Invitrogen	Cat# K182001
RNeasy Mini Kit	Qiagen	Cat# 74104
TURBO DNA-Free Kit	Invitrogen	Cat# AM1907
Superscript™ IV First Strand Synthesis System Kit	Invitrogen	Cat# 18091050
Powertrack SYBR Green Master Mix kit	Thermo Fisher Scientific	Cat# A46012
Transcreener ADP2 FP Assay	BellBrook Labs	Cat# 3010
Deposited data
Raw whole genome sequencing data	This paper	BioProject ID PRJNA990483
E. coli K-12 str. BW25113 genome	Grenier et al.50	Accession no. CP009273.1
Inward-facing MsbA	Verma et al.21	PDB 7SEL
Occluded MsbA	Kehlenbeck et al.51	PDB 7BCW
Outward-facing MsbA	Ward et al.52	PDB 3B60
Oligonucleotides
Primers used in this work	This paper	See Table S5
Recombinant DNA
pKD46	Datsenko and Wanner.53	N/A
pCP20	Datsenko and Wanner.53	N/A
pCas	Jiang et al.54	Addgene Cat#62225
pTargetF	Jiang et al.54	Addgene Cat#62226
pFD152	Depardieu and Bikard.55,56	N/A
pGDP2	Cox et al.57	N/A
Software and algorithms
Prism 9.4.1	GraphPad	https://www.graphpad.com/
ImageJ (Fiji)	Schindelin et al.58	https://imagej.net/software/fiji/
Benchling	Benchling	https://www.benchling.com/
Excel 16.36	Microsoft	https://www.office.com/
Geneious Prime v2020.2.4	Dotmatics	https://www.geneious.com/
Flye de novo assembler	Kolmogorov et al.59	https://github.com/fenderglass/Flye
QuantiGene Plex Data Analysis Version 2.6.2	Thermo Fisher	https://apps.thermofisher.com/apps/quantigene
OriginPro 2021	Origin Lab	https://www.originlab.com/
Other
SeqCenter	SeqCenter	N/A
University of Guelph Advanced Analysis Center, Molecular and Cellular Imaging Facility – FEI Quanta FEG 250	Elecmi	N/A
QuantStudio 3 Real-Time PCR System	Applied Biosystems	Cat# A28567
BioTek Synergy H1 microplate reader	Agilent	N/A
JPK NanoWizard® 4 NanoScience	Bruker	N/A

Resource availability

Lead contact

Further inquiries and requests should be directed to the lead contact, Dr. Georgina Cox (gcox@uoguelph.ca).

Materials availability

Reagents and bacteria used for the biological studies were obtained from the commercial or internal sources listed in the key resources table. The TE mutant strain generated in this study can be obtained from the lead contact.

Data and code availability

• •The raw genomic sequencing data of ΔtolC and the TE mutant (before and after the repair of non-synonymous mutations) were deposited in GenBank (BioProject ID PRJNA990483). All further data reported in this paper can be shared by the lead contact upon request.
• •The MsbA crystal structures used in this study were obtained from the Protein DataBank (PDB 7SEL; PDB 7BCW; PDB 3B60).
• •The E. coli BW25113 genome sequence was obtained from GenBank (Accession no. CP009273.1).
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Experimental model and study participant details

Strains, plasmids, and growth conditions

Bacterial strains and plasmids used in this study are provided in the key resources table and Table S4. E. coli K-12 str. BW25113^12,50 and the uropathogenic E. coli CFT073 were used as wild-type strains in this study. The E. coli TOP10 and DH5α strains were used as routine cloning hosts. The ΔtolC mutant and strains for resistance cassette amplification were obtained from the Keio Collection.¹² Plasmids for CRISPR–Cas9-mediated counter selection, pCas, and pTargetF, were purchased from Addgene.⁵⁴ Plasmids for the λ-Red recombinase system, pKD46 and pCP20,⁵³ and the constitutive expression of genes, pGDP-2,⁵⁷ were used as previously described. The kanamycin-resistance cassette from ΔtolC was removed using pCP20, the genome of this strain was then sequenced, and the strain was used for all aspects of this study. For the knockdown of msbA, sgRNA targeting msbA in E. coli was ligated into the conjugative CRISPRi plasmid, pFD152, using a previously described single-step golden gate assembly protocol.^55,56

Strains were routinely propagated in Lysogeny broth (LB) (Bioshop) at 37°C or 30°C with aeration (220 r.p.m.). When grown in microtiter plates, strains were aerated at 900 r.p.m. For the selection of resistance markers, ampicillin (100 μg/mL) (Bioshop), kanamycin (50 μg/mL) (Sigma-Aldrich), spectinomycin (150 μg/mL) (Bioshop), and chloramphenicol (15 μg/mL) (Bioshop) were used at the listed concentrations. Susceptibility testing was conducted in cation-adjusted Mueller Hinton II Broth (MHB II) (BD Difco).

Method details

Generation of the TE mutant

The TE mutant was generated using λ-Red recombination⁵³ and CRISPR–Cas9-mediated counter selection.⁵⁴ Eight efflux genes were inactivated in the order shown in Table S1. Nucleic acid amplification was achieved using the high-fidelity 2× GB-AMP PaCeR polymerase master mix (GeneBio Systems Inc) or the Taq 2× polymerase master mix (FroggaBio) according to the manufacturers’ guidelines. Amplicons were purified using the GeneJET PCR purification kit (Thermo Fisher Scientific) and plasmids were extracted and purified from saturated cultures using the GeneJET Plasmid Miniprep kit (Thermo Fisher Scientific) according to the manufacturer’s suggested guidelines.

For λ-Red recombineering, the pKD46 plasmid was transformed into recipient strains using electroporation.⁵³ Broth cultures were grown to the mid-exponential phase of growth (OD_600nm–0.5) in the presence of ampicillin and 4 mM arabinose. The genes of interest were replaced with kanamycin-resistance cassettes from the appropriate Keio strain, which were amplified using PaCeR and primers annealing 50 bp upstream and downstream of the desired region (Table S5). To disrupt the genes of interest, recombinase-induced electrocompetent cells were transformed with 500 ng of purified amplicons. Strains harboring successful gene deletions were transformed with pCP20 to excise the resistance cassette.

For CRISPR-Cas9-mediated counterselection, the methodology described by Jiang et al.⁵⁴ was modified as previously described.⁹ CRISPR guide software (Benchling) was used for the selection of sgRNA (N20) sequences, which were introduced into pTargetF by PCR (Table S5). Single-stranded DNA repair oligonucleotides were designed to introduce three tandem stop codons, a silent mutation to remove the PAM, and a silent mutation to introduce an AseI restriction enzyme site for screening of successful mutants. Electrocompetent cells were transformed with 50 ng of pCas. A 50 mL broth culture was grown at 30°C to an OD_600nm of 0.5 in the presence of kanamycin and 4 mM arabinose for recombinase induction. Recombinase-induced electrocompetent cells were then transformed with 100 ng of pTargetF modified with a sgRNA and 2000 ng of ssDNA repair oligonucleotide. The cells were recovered in 1 mL of LB at 30°C, then grown overnight on LB agar supplemented with kanamycin and spectinomycin. Amplicons were then digested with the AseI restriction enzyme (FastDigest enzyme series, Thermo Fisher Scientific) to screen for successful mutants. Gene disruptions were confirmed using Sanger sequencing (The Center for Applied Genomics, SickKids or Advanced Analysis Center, University of Guelph).

Four secondary mutations were identified in the TE mutant’s genome (Table S1), including non-synonymous mutations in hdfR, yjfC, and pitA. CRISPR-Cas9-mediated counterselection was used to repair the mutations in hdfR and yjfC, which introduced two intentional silent mutations to remove the adjacent PAM site and for restriction enzyme-guided screening purposes as described above. To screen for successful revertants, the hdfR and yjfC genes were PCR amplified (Table S5) and digested with AseI and SacI restriction enzymes, respectively. λ-Red recombineering was performed as described above with modifications to repair the mutated pitA gene using the cat-sacB selectable/counter-selectable cassette, as previously described.⁴⁹ The cat-sacB cassette was amplified from the E. coli TUC01 genome generously provided by Dr. Steven Kelly (Whitfield lab, University of Guelph), using PaCeR and primers that were designed to encompass regions immediately upstream and downstream of pitA (primers, Cat_pitA_Fwd/SacB_pitA_Rev; Table S5). The wild-type pitA gene was amplified using PaCeR (primers, pitA_76_Up/pitA_50_Low; Table S5). Recombinase-induced electrocompetent cells were transformed with 500 ng of the amplicon containing the cat-sacB cassette and grown overnight on LB containing chloramphenicol. Recombinase-induced electrocompetent cells of successful recombinants containing cat-sacB in the pitA region were then transformed with 2000 ng of the wild-type pitA amplicon and grown overnight on LB containing 6% (w/v) sucrose. Replacement of the cassette with the wild-type pitA gene was confirmed by the loss of resistance against chloramphenicol, growth on sucrose, and Sanger sequencing (The Center for Applied Genomics, SickKids or the Advanced Analysis Center, University of Guelph).

For the genomic integration of the ‘Pore’, the fhuA ΔC/Δ4L gene and an adjacent gentamicin resistance cassette were amplified from the genome of an E. coli K-12 str. BW25113 harboring the ‘Pore’ gene using PaCeR (primers, fhuA_40_Up/glmS_3680_Low; Table S5). Pore integration and removal of the resistance cassette were performed as described above using λ-Red recombineering. The pore was introduced into the intergenic region between glmS and pstS and verified using PCR (primers, pstS_520_Up/glmS_3680_Low, Table S5). The activity of the pore was confirmed through susceptibility testing with vancomycin, as described previously.⁹

Whole genome sequencing of ΔtolC and the TE mutant

Genomic DNA from the ΔtolC and TE mutants was extracted from overnight cultures using the PureLink Genomic DNA Mini Kit (Invitrogen) according to the manufacturer’s suggested guidelines. Nanopore long-read DNA library preparation was performed using a PCR-free V14 chemistry ligation sequencing kit by SeqCenter (Pennsylvania, USA), and sequencing was performed on a MinION platform. Raw sequencing reads were assembled with the Flye de novo assembler,⁵⁹ then annotated with the E. coli K-12 str. BW25113 reference genome (accession no. CP009273.1) from NCBI using Geneious Prime v2020.2.4.⁶⁰ Whole genome sequencing data of the TE mutant containing the non-synonymous mutations were deposited in the GenBank database (BioProject ID PRJNA990483). For short-read sequencing, Illumina DNA library preparation was performed using an Illumina Nextera kit by SeqCenter (Pennsylvania, USA), followed by Illumina sequencing on a NextSeq 2000 platform. Low-quality reads were trimmed using an in-suite BBDuk plug-in and raw reads were aligned to the E. coli K-12 str. BW25113 reference genome with bowtie2. Genomic differences between the ΔtolC and TE mutant genomes relative to wild-type E. coli K-12 strain were identified using the following thresholds: minimum variant frequency of 0.75, maximum variant P-value of 10⁻⁶, and minimum variant P-value of 10⁻⁵. Breseq⁶¹ was used to confirm the mutations identified using Geneious. Whole genome sequencing data of ΔtolC and the repaired TE mutant were also deposited in the GenBank database (BioProject ID PRJNA990483).

Generation of synthetic #20 resistant mutants

Resistance-conferring mutations in ΔtolC were generated using the single-step selection method.⁶² Briefly, a single colony was inoculated into MHB II and grown overnight at 37°C with aeration (220 rpm). The overnight culture was diluted 1:10 in fresh MHB II and 180 μL of the diluted culture was applied to the surface of MHB II agar containing 4X the MIC of the compound of interest. The plates were incubated at 37°C for 48 h. Isolated colonies were patch-plated onto MHB II agar containing 4X MIC of the compound of interest, and the susceptibility level of the mutants was assessed using the CLSI broth microdilution method⁶³ in MHB II as described below.

Antimicrobial susceptibility testing

Minimum inhibitory concentrations (MICs) were determined using the CLSI broth microdilution method.⁶³ Briefly, strains were applied to LB agar and incubated overnight at 37°C. To prepare the inoculum, colonies were suspended in sterile 0.85% (w/v) NaCl to an OD_600nm of ∼0.1, then diluted 1:100 in MHB II. In a round-bottom 96-well microtiter plate (VWR), compounds were serially titrated 2-fold in 50 μL of MHB II, followed by the addition of 50 μL of the diluted cell suspension, resulting in a final well volume of 100 μL. Plates were incubated at 37°C with aeration (900 rpm) for 18 h, and the OD_600nm was measured using a BioTek Synergy H1 microplate reader. Chequerboard assays were performed using the same methodology with modifications as described. Compound A was serially titrated 2-fold on the x axis while compound B was serially titrated 2-fold on the y axis. Cell inocula and final volumes remained at 50 μL and 100 μL, respectively.

RT-qPCR

The RT-qPCR procedure followed the MIQE Guidelines⁶⁴ for quality control. 3 mL of LB supplemented with spectinomycin and kanamycin was inoculated with a single colony and incubated at 37°C with aeration (220 rpm) for 18 h. To induce CRISPRi-mediated msbA knockdown using the pFD152 plasmid,^55,56 3.125 μg/mL and 0.05 μg/mL of aTc, equivalent to 0.5x the MIC, were added to the wild-type K-12 and ΔtolC cultures, respectively, at the time of inoculation. Saturated overnight cultures were diluted with LB to an OD_600nm of 0.4. 1 mL of diluted cells were harvested by centrifugation at room temperature and total RNA was extracted using the RNeasy Mini Kit (Qiagen) according to the manufacturer’s guidelines. RNA quality and quantity were determined using a Nanodrop 2000 Spectrometer (Thermo Fisher). The integrity of RNA samples was assessed by gel electrophoresis before cDNA synthesis. 400 ng of RNA was treated with the TURBO DNA-Free Kit (Invitrogen) according to the manufacturer’s guidelines to degrade contaminating genomic DNA. The samples were reverse transcribed into cDNA using the Superscript IV First Strand Synthesis System Kit (Invitrogen) according to the manufacturer’s guidelines. Primers were designed using the PrimerQuest Tool (IDT) to amplify ∼100 bp products within the target gene and the endogenous reference gene (Table S5). recA was used as the reference gene to normalize transcript levels between samples.⁶⁵ A five-point, 10-fold serial dilution of cDNA was used to generate standard curves for the msbA and recA primer sets, which was performed to validate the qPCR primer efficiencies using the Powertrack SYBR Green Master Mix kit (Thermo Fisher Scientific) according to the manufacturer’s guidelines. Reactions were set up in MicroAmp EnduraPlate Optical 96-Well Clear Reaction plates (Applied Biosystems) and amplified using the QuantStudio 3 Real-Time PCR System (Applied Biosystems). After amplification, a melt curve analysis was performed. The data were analyzed using the Design and Analysis Software from Thermo Fisher QuantiGene Plex Data Analysis (Version 2.6.2). The line of best fit for the msbA primer set was y = −3.4361x + 24.678 (R² = 0.996), with a primer efficiency calculated to be 95.4%. The line of best fit for the recA primer set was y = −3.4817x + 22.523 (R² = 0.998), with a primer efficiency calculated to be 93.6%. Upon satisfying the acceptable standard curve parameters (primer efficiency of 90–100%, slope between −3.58 and −3.10, and R² > 0.99),⁶⁴ the cDNA samples were diluted 1:100 before being used in RT-qPCR as described above. Relative quantification was carried out and relative fold gene expression was calculated using the ΔΔCt method.⁶⁶ A negative reverse transcriptase control and no template controls for both primer sets were included in each RT-qPCR plate to ensure the absence of extraneous and genomic DNA contamination. Three technical replicates were used for each sample, and the results are representative of three independent biological replicates.

CRISPRi-mediated msbA repression

The effect of msbA CRISPRi-mediated repression on bacterial fitness was assessed using the CLSI broth microdilution method,⁶³ as described above with modifications. Strains harboring pFD152:msbA were propagated on LB agar supplemented with spectinomycin and kanamycin. Cell inocula were prepared as described above. The anhydrotetracycline (aTc) inducer was serially titrated 2-fold in 50 μL of MHB II in a round-bottom 96-well microtiter plate (VWR). Then, 50 μL of the diluted cell suspensions were applied to the plate. The microtiter plates were incubated at 37°C with aeration (900 rpm) for 24 h, and the OD_600nm was measured using a BioTek Synergy H1 microplate reader.

MsbA ATPase assay

The MsbA protein from E. coli was purified and reconstituted into amphipols as previously described.¹⁸ MsbA ATPase activity was measured using a Transcreener ADP2 FP Assay (BellBrook Labs). To determine the IC₅₀, synthetic compound #20 was first incubated with 2× MsbA enzyme solution for 10 min and the reaction was initiated with 2× ATP solution. The final concentration of each reagent used in the assay was 9.2 nM of purified E. coli MsbA in amphipol, 50 μM ATP in 50 mM Tris (pH 7.5), 10 mM MgCl₂, 1% glycerol, 0.1% bovine gamma globulin, 1 mM dithiothreitol (DTT), 0.007% Brij-35 and 0.5% dimethyl sulfoxide (DMSO). Under these conditions, the typical and uninhibited ATPase activity for MsbA was 87 nmol min⁻¹ mg⁻¹. To quench the reaction, Transcreener Detection Buffer was added when 5% of the ATP was hydrolyzed to ADP. The IC₅₀ was determined by fitting the inhibition dose-response curve with a nonlinear four-parameter inhibition model (GraphPad Prism).

Thin-section transmission electron microscopy

A pellet of cultured bacteria was fixed in 1/2 Karnovsky’s fixative (2% paraformaldehyde, 2.5% glutaraldehyde in 0.1M sodium cacodylate buffer, pH 7.2) and then post-stained with 1% reduced osmium tetroxide (OsO4). “En block” staining of samples was in 0.5% uranyl acetate. The bacteria were then dehydrated through a series of increasing ethanol concentrations, washed in propylene oxide, and then processed into epoxy resin (Eponate 12). The infiltrated samples were then polymerized at 65°C for two days. Ultrathin sections were prepared with an Ultracut microtome (Leica) and transferred to 75 mesh copper grids. Sections were stained with 0.2% lead citrate. Images were taken with a JEOL JEM-1400 TEM (transmission electron microscope) at 80keV and a GATAN Ultrascan 1000 CCD camera.

Outer membrane permeability assay

Strain permeability was evaluated using 1-N-phenylnaphthylamine (NPN) (Sigma-Aldrich) uptake assays. Overnight cultures grown in MHB II at 37°C for 18 h were subcultured to an OD_600nm of 0.1 in MHB II, then grown to an OD_600nm of 0.6 at 37°C. The cells were harvested and washed three times with phosphate-buffered saline (PBS) and resuspended in PBS to a final OD_600nm of 0.6. To disrupt efflux activity, 0.781 μg/mL, 6.25 μg/mL, and 12.5 μg/mL of carbonyl cyanide m-chlorophenyl hydrazone (CCCP), equivalent to 1x the MIC of each strain, was incubated with the cell suspensions of ΔtolC, EKO-35, and the TE mutant strains, respectively, at room temperature for 30 min. Solid black bottom 96-well microtiter plates (VWR) were blocked with 2% (w/v) bovine serum albumin (HyClone) for 15 min at room temperature. 180 μL of the cell inocula were then applied to the microtiter plates, followed by the addition of NPN at a final concentration of 1 μg/mL, resulting in a final well volume of 200 μL. Fluorescence was measured using a BioTek Synergy H1 microplate reader for 30 min using excitation and emission wavelengths of 350 nm and 420 nm, respectively.

Scanning electron microscopy

Saturated overnight cultures were diluted 1:100 in LB, supplemented with antibiotics when necessary, and grown at 37°C with aeration (220 rpm) until an OD600 nm of ∼0.5 was achieved. When required, the subcultures were supplemented with 0.5x the MIC of synthetic compound #20 or aTc to determine the effects of MsbA inhibition and msbA repression, respectively, on cell morphology. For ΔtolC, 0.1563 μg/mL of synthetic compound #20, equivalent to 0.5x the MIC, was added. 1.5 mL of cultures in the mid-exponential growth phase were harvested by centrifugation, washed three times with sterile 0.85% (w/v) NaCl, and resuspended in 400 μL of 0.85% (w/v) NaCl. 200 μL of the cell suspension was applied to a carbon planchette (Ted Pella) and incubated at room temperature for 30 min. The cell suspension was removed and the planchettes with adhered cells were submerged in 2% glutaraldehyde (v/v) for 30 min, then washed three times with 0.85% (w/v) NaCl. The planchettes were then submerged in 1% (v/v) osmium tetraoxide for 30 min and then washed once with 0.85% (w/v) NaCl. The cells were dehydrated in a graded ethanol series, and then fully dried using a Denton DCP-1 critical point dryer. Samples were coated immediately with 15 nm of gold using a sputter coater (Denton Desk V TSC). Images were acquired using an FEI Quanta FEG 250 scanning electron microscope (Thermo Fisher Scientific) operated at 20.0 kV under high vacuum at the Molecular and Cellular Imaging Facility (University of Guelph). ImageJ (Fiji)⁵⁸ was used to measure cell lengths.

Atomic force microscopy

AFM experiments were performed in PBS at room temperature using a JPK NanoWizard 4 NanoScience Instrument. Overnight cultures were diluted 1:100 in LB and grown at 37°C with aeration (220 rpm) until the mid-exponential phase of growth (OD_600nm = ∼0.5) was achieved. The cells were then harvested by centrifugation and washed three times in PBS. A 100x diluted bacterial suspension was left to adhere to polyethylenimine (PEI)-coated glass bottom dishes for 1 h, then gently rinsed three times with PBS before AFM measurements. AFM cantilevers were calibrated by the thermal noise method.⁶⁷ Imaging was performed in quantitative (QI) mode using SNL-10 cantilevers. Height images were recorded at a constant approach/retract speed of 25 μm/s, using an applied force of 0.25 nN and a ramp size of 500 nm. Cell surface morphology parameters (height, roughness, length, and volume) were extracted from height images acquired at low resolution (128 × 128 pixels² on 8 × 8 μm² areas) or high-resolution (256 x 256 pixels² on 200 × 200 nm² areas on top of the bacteria). For the mechanical characterization of strains, force-indentation curves were recorded in force-volume mode (16 curves × 16 curves) in 250 × 250 nm² regions on top of the cells. Curves were recorded using MSCT-E cantilevers, using an applied force of 1 nN, a constant approach/retract speed of 1 μm/s, and a ramp size of 1 μm. The approach segment of the force-distance curves was fitted with the Hertz/Sneddon model over a distance of 20 nm, using a Poisson ratio of 0.5 and considering a conical tip of 17.5° half-cone angle.⁶⁸ Young’s Modulus (E) was extracted from the nonlinear region of the curves and the spring constant (k) was calculated through the slope of their linear region. For each cell, these values were plotted as histograms and the average values were extracted by fitting with a Gaussian function. Statistical analysis was carried out using Origin software (OriginPro 2021), applying a statistical Tukey’s multiple-comparison test. Data were analyzed with JPK Data Processing software.

Quantification and statistical analysis

Statistical analyses were conducted using GraphPad Prism 9.4.1, Excel 16.36, and OriginPro 2021. Data obtained from qRT-PCR experiments were analyzed using the Thermo Fisher QuantiGene Plex Data Analysis (Version 2.6.2). Detailed information regarding statistical methods for each experiment can be found in the figure legends and the corresponding figures.

Published: March 27, 2024