Unless otherwise stated, the E. coli strains used in this study (Table 1) were grown in LB -Lennox (LB broth) medium (Difco) or in M9 minimal defined medium (Na₂HPO₄ 6 g/L, KH₂PO₄ 3 g/L, NaCl 0.5 g/L, NH₄Cl 1 g/L, MgSO₄ 1 mM, CaCl₂ 0.1 mM, and 0.2% glucose) supplemented with thiamine (1 mg/L) and leucine (100 μg/mL; Eberhart et al., 2012; Zhao et al., 2015, 2017) at 37°C with shaking at 200 rpm. Antibiotics were added to media as needed (ampicillin, Amp, 100 μg/mL; tetracycline, Tet, 50 μg/mL; kanamycin, Kan, 50 μg/mL; nalidixic acid, Nal, 30 μg/mL; chloramphenicol, Cm, 32 μg/mL). Vibrio harveyi MM32 (ATCC BAA-1121) (Table 1) was grown in marine broth 2216 (Difco) or autoinducer bioassay (AB) medium [NaCl 17.5 g/L, MgSO₄ 12.3 g/L, casamino acids (vitamin-free) 2.0 g/L, KH₂PO₄ (pH 7.0) 1 M, L-arginine 0.1 M, and glycerol 10 mL/L; (ATCC)] at 30°C with shaking at 200 rpm and antibiotics were added as needed (Amp, 50 μg/mL; Kan, 25 μg/mL; Cm, 15 μg/mL).

All plasmids were extracted from E. coli by using a QIAprep Spin Miniprep kit (Qiagen). E. coli genomic DNA was extracted with the DNeasy Blood & Tissue kit (Qiagen). Platinum PCR Super Mix (Invitrogen) was used for preparative PCR when working with plasmid pBAD18-Cm, pDM4, pGEM-2, and pKD4. Complementation were performed using primers incorporating restriction sites (Supplemental Table 1) for PCR amplification, restriction digest (New England Biolabs Inc.), and ligation (T4 ligase, New England Biolabs Inc.) following standard cloning techniques. All conventional PCR for verification of constructs and gene detection used DreamTag Green PCR Master Mix (Thermo Scientific) and PCR products were confirmed by sequencing (Eurofins Genomics).

Gene-specific PCR-mediated gene deletion followed the methods of Datsenko and Wanner (Datsenko and Wanner, 2000). Briefly, primers (Supplemental Table 1) were designed to incorporate a 36- to 50-nucleotide segment that was complementary to the DNA sequence flanking the gene of interest. Primers were used to generate a PCR product that joined these flanking sequences to a Kan-resistance gene (kan^r) that originated from pKD4 (Table 1). PCR products were column purified by using a QIAquick PCR purification kit (Qiagen). Restriction enzyme (DpnI; New England Biolabs Inc.) was used to digest pKD4 plasmid for 4 h at 37°C before column purification was repeated. Processed PCR products (150 ng) were then suspended in 5 μL of 10 mM Tris (pH 8.0) and were electroporated into E. coli 25 with a Gene Pulser Xcell (Bio-Rad) as described previously (Zhao et al., 2015). Briefly, E. coli 25 carrying the λ Red plasmid pKD46 (Amp^r) was prepared for electroportation (1.8 kV, 25 μF, 200 Ω, 1 mm gap cuvette) by first growing culture to an optical density (OD_600nm) of ~0.6 in SOB medium (Fisher Scientific) (Table 1) with 1 mM _L-arabinose (30°C). Cells were then washed twice in ice-cold water and once in 10% glycerol. Cells were subsequently resuspended in 10% glycerol (50 μL) for electroporation. Immediately after electroporation, cells were resuspended in SOC recovery medium (Fisher Scientific) for 2 h at 30°C (200 rpm) before plating on Kan-containing LB agar and incubating overnight at 30°C. PCR was used to verify gene deletion of lsr, lsrK, and lsrR (Table S1). All mutants were generated utilizing this method with the exception of E. coli 25 ΔluxS, for which a splice-overlap-extension method was used (Heckman and Pease, 2007). Briefly, two 400- to 600-bp PCR fragments from sequences flanking luxS were joined and then cloned into a suicide plasmid (pDM4; Cm^r; Table 1) by using standard cloning procedures (Milton et al., 1996). Constructs were confirmed by DNA sequencing (Eurofins Genomics) prior to electroporating into electrocompetent E. coli S17-1 λpir (Table 1). Conjugation was performed with E. coli 25 to generate a mutant that was selected on LB agar plates (Tet and Cm antibiotics) followed by a 10% sucrose selection (Zhao et al., 2015). PCR was used to confirm the deletion of luxS.

Cultures (5 mL) were grown overnight in M9 medium and total RNA was extracted from an aliquot (1.5 mL) with the RiboPure^TM-Bacteria kit (Ambion) per manufacturer's instruction with an additional DNase treatment with a RQ-1 RNase-Free DNase (Promega). RNA was quantified by using a NanoDrop^TM 2000 spectrophotometer (ThermoFisher Scientific). Complementary DNA was generated from DNase-treated total RNA (500 ng) with iScript Reverse Transcription Supermix (Bio-Rad) per manufacturer's instruction. Quantitative RT-PCR was completed in triplicate using the SsoAdvanced SYBR Green Supermix (Bio-Rad) per manufacturer's instruction with indicated primers (Supplemental Table 1). A CFX98 Real-Time System (Bio-Rad) was used to perform the thermal cycling parameters: one cycle at 95°C for 30 s; 39 cycles of 95°C for 5 s, 55°C for 15 s with plate read and 72°C for 30 s; 65°C for 5 s and plate read every 0.5°C/cycle to 95°C. The relative gene expression level was calculated with wild-type E. coli 25 serving as the control for calculations using the ΔΔCt method (Livak and Schmittgen, 2001). To detect potential DNA contamination, before the reverse transcription reaction an aliquot of each RNA extraction was subjected to conventional qPCR with rpoD primers (Cq-values >37, signified low level of DNA contamination).

Co-culture competition assays were performed with a modified competition assay protocol (Chen et al., 2003; Zhao et al., 2015). Briefly, strains to be competed were grown individually in LB broth overnight. The next day the individual overnight cultures were combined (1:1) and inoculated into fresh M9 medium at a ratio of 1:100 for competition for 4, 8, 12, and 24 h at 37°C with aeration. Individual strains were also inoculated (monoculture) under the same conditions as controls. When appropriate, antibiotics and/or 0.2% (w/v) _L-arabinose was added to pBAD18-Cm constructs or 0.5 mM Isopropyl β-D-1-thiogalactopyranoside (IPTG) to pGEM-2 constructs unless otherwise noted. Colony forming unit (CFU) were quantified by using serial dilution and a 6X6 drop-plate technique (Chen et al., 2003).

Measurement of AI-2 production by E. coli 25 and complemented luxS strains was done by using an autoinducer bioassay (AB) as previously described (Surette and Bassler, 1998, 1999). Strains of interest were grown overnight at 30°C with aeration in LB medium supplemented with 0.5% glucose. V. harveyi MM32 was grown overnight at 30°C with aeration in AB medium supplemented with 0.5% glucose. On the following day, bacterial cultures were inoculated (1:100) into fresh media (as described, respectively) and were grown for 8 h at 30°C with aeration. E. coli 25 ΔluxS/pBAD18-Cm and E. coli 25 ΔluxS/pBAD18-Cm::luxS were grown in LB without glucose to avoid araC inhibition during _L-arabinose induction of pBAD18-Cm::luxS (Simcikova et al., 2014). Samples were centrifuged at 18,000 × g for 10 min and filtered (0.22 μm) to obtain cell-free supernatants that were stored at −20°C. The autoinducer bioassay (AB) medium (Bassler et al., 1993) was used to grow reporter strain V. harveyi MM32 (autoinducer 1⁻, autoinducer 2⁻; Bassler et al., 1993). Previously prepared cell-free supernatants were tested for the presence of AI-2 by adding to V. harveyi culture followed by detection of luminesce. Briefly, reporter strain V. harveyi MM32 was grown overnight in AB medium (30°C for 16 h) and was then diluted (1:5,000) in fresh AB medium. An aliquot (90 μL) was added to each well of a 96-well plate with 10 μL supernatant sample (from above). A positive-control well contained cell-free supernatant from E. coli 25 wild-type, while a negative-control well contained V. harveyi MM32 with no supernatant added. Plates were sealed with breathable sealing film (Axygen) and luminescence was measured every hour using an Infiniti M1000 PRO microplate reader (Tecan Systems). Each assay was repeated for three independent replicates.

Isolated colonies were inoculated into 5-mL LB media with appropriate antibiotic and grown as described. Overnight culture was diluted (1:100) into fresh M9 media (10 mL) and grown overnight at 37°C with 200 rpm shaking until OD600 ~ 0.6 at which point 0.02% (w/v) _L-arabinose was added for 24 h at room temperature with shaking at 200 rpm. Total proteins were collected by centrifugation at 18,000 × g at 4°C for 5 min. Cell pellets were resuspended in 1x laemmli sample buffer (Bio-Rad) and boiled for 10 min. Any kD Tris-glycine precast gels (Bio-Rad) were used for SDS-PAGE protein separation. A Trans-Blot turbo transfer starter system (Bio-Rad) was used to transfer proteins onto a low-fluorescence polyvinylidene fluoride membrane (Bio-Rad) and Ponceau S stain was used to verify protein transfer prior to addition of antibodies for specific protein detection. Primary antibody anti-His-tag (1:1,000; Thermo Scientific) was used with secondary goat anti-mouse antibody (1:5,000; DyLight 650, conjugate). A ChemiDoc MP Imaging System (Bio-Rad) was used to detect fluorescent signal and band intensity was quantified with ImageJ software (Schneider et al., 2012). A ratio of McpM value to DnaK value served to normalize and quantify and are represented by arbitrary unit (AU).

Where appropriate, a one-way analysis of variance (ANOVA) was used to compare experimental results with a Dunnett's one-way multiple pairwise comparison test. Depending on the experimental design, a two-way ANOVA was used in conjunction with a Tukey's all pairwise multiple comparison test (SigmaPlot version 12.5; Systat Software, Inc., San Jose, CA).

We conducted co-culture competition assays with luxS deletion strains for both the microcin-PDI positive (E. coli 25) and susceptible strains (E. coli K-12 BW25113; Figure 1). Differences in inhibition were clearly evident for the mid-to-late log growth phase (8 h), which is the same time that there was a 14-fold increase in the abundance of mcpM mRNA relative to the 4-h culture of wild-type E. coli 25 (Figure S1; Eberhart et al., 2012). Compared to inhibition of BW25113 by the wild-type positive control (at 8 h, Figure 1), eliminating luxS from E. coli 25 was 1.6-log less effective while eliminating luxS from BW25113 reduced the mccPDI phenotype by 2.7-log. When co-culture involved both luxS deletion strains, the total reduction in mccPDI phenotype was ~3.5 log relative to the wild-type strain; a finding that was consistent with AI-2 from both strains contributing to a signal for upregulation of mccPDI. After 8 h the effect of luxS deletion was no longer observed (Figure 1). As expected, co-culture with the susceptible E. coli BW25113 had no negative effects on E. coli 25 growth with or without a luxS (Figure S2).

AI-2 deficient mutant strain E. coli 25 ΔluxS was complemented by using in trans expression of luxS under the control of an _L-arabinose inducible promoter, araBAD (pBAD18-Cm). A pBAD18-Cm plasmid with no cloned insert was used as a negative control while E. coli 25 was used as positive control. Complementation restored the ability of luxS deletion strain to inhibit BW25113 compared to the respective un-induced strain (Figure 2, compare first and last bars under E. coli 25 ΔluxS/pBAD18-Cm::luxS). Adding arabinose to the culture regardless of the presence or absence of the pBAD18-Cm plasmid produced some growth advantage for the E. coli 25 strains relative to the susceptible strain (Figure 2, compare the open and filled bars), although this effect did not exceed 0.5 log on average. A western blot confirmed synthesis of the complemented LuxS protein (8 h culture; Figure S3A), and an autoinducer bioassay was consistent with increased production of AI-2 (Figure S3B).

To further validate the contribution of AI-2 to the regulation of the mccPDI phenotype, we constructed an E. coli 25 ΔlsrACDBFG (E. coli Δlsr) mutant (Wang et al., 2005). The lsr operon consists of six genes of which lsrACDB encodes the ABC transporter, and lsrF and lsrG are involved in the degradation of AI-2. A separate lsrR/K operon encodes an uptake repressor and kinase to phosphorylate AI-2, respectively (Li et al., 2007). After 8 h the reduction in inhibition for the Δlsr strain was statistically indistinguishable from the reduction for the ΔluxS strain (Figure 3). We further confirmed that deletion of the lsr operon or the lsrR/K operon does not affect production of AI-2 itself (Figure S4).

The mRNA for mcpM peaks at the mid-to-late log phase growth and declines when cultures enter stationary phase (Figure S1). Under monoculture (1:500 initial dilution) mcpM expression differed at 8 h was reduced for ΔluxS strains compared to the isogenic wild-type (Figure 4). To verify ΔluxS monoculture results, we repeated the experiment from co-culture samples with reduced inoculant (1:1,000 instead of 1:500 to normalize with co-culture experiments) and observed a similar mcpM expression pattern, but at later point of 12 h (Figure S5). The pattern of up and down-regulation of mcpM expression matches what has been reported previously (Eberhart et al., 2012; Zhao et al., 2017). The AI-2 deficient mutant (ΔluxS) clearly exhibits reduction of mcpM (Figure 4 and Figure S5) with an overall 5-fold reduction in mcpM transcription, consistent with phenotype differences (Figure 1). Furthermore, the strain deficient in AI-2 (ΔluxS) has a greater reduction of mcpM expression compared to ΔlsrR (Figure 4). This suggests that the deletion of the AI-2 uptake regulation gene (ΔlsrR) or uptake mechanism Δlsr (Figure 3) can be mitigated through another means of cell entry such as passive diffusion of AI-2 through porins (Galloway et al., 2011).

To examine the kinetics of McpM protein synthesis, we used a vector (pCR2.1) with the mcpM endogenous promotor (P_{mic−10/−210}) coupled with mcpM (Zhao et al., 2017). Normalized densitometry of western blot results showed a delay in E. coli 25 ΔluxS/pCR2.1::P_{mic−10/−210}mcpM recombinant McpM production compared to the strain E. coli 25 ΔmcpM/pCR2.1::P_{mic−10/−210}mcpM that retained an intact luxS (Figure 5). The kinetics of McpM synthesis for both strains mirrored the typical mcpM transcription except with a 2-h delay for the ΔluxS strain (Figure 5B). The lack of luxS does not inhibit the production of McpM because EnvZ/OmpR is still the primary regulator of mcpM (Zhao et al., 2017) as confirmed by loss of McpM synthesis with the deletion of the ompR (Figure S6A). Deletion of luxS also does not affect ompR expression, which remains constant through different growth phases (Figure S6B).

Published work demonstrates that AI-2 QS and LsrR influence the synthesis of the sRNA micC (Li et al., 2007), which in turn regulates outer membrane porins (OmpC and OmpF) in a manner similar to the EnvZ/OmpR two-component system (Mizuno et al., 1988). To examine the effects of micC and micF (another sRNA known to regulate outer membrane porin OmpF in E. coli; Delihas and Forst, 2001) in PDI-producer strain, we overexpressed micC and micF in E. coli 25 during competition with strain BW25113. After 8-h co-culture competition it was readily apparent that overexpression of micC and micF reduced the PDI phenotype significantly (Figure 6). Compared to positive control competition culture (with empty vector pGEM-2; 5-log loss in susceptible BW25113), micC overexpression resulted in a 1-log reduction in BW25113 while micF overexpression resulted in a complete loss of the PDI phenotype. There was evidence that a “leaky” pGEM-2 vector permitted sufficient micF and micC expression to reduce the PDI phenotype by 5-log and 2-log, respectively, in the absence of IPTG induction (Figure 6).

From a broad perspective, QS-regulated bacteriocin production should provide a competitive advantage when resources become limited in the presence of large population of competitors (Blanchard et al., 2016). The PDI-positive strain (E. coli 25) used in this study was originally isolated from a cow (Sawant et al., 2011), and by using a neonatal calf model Eberhart et al. showed that the wild-type E. coli 25 out competed an isogenic PDI-defective strain (E. coli 25 ΔmcpM ΔmcpI; Eberhart et al., 2014). The “growth phase” of bacteria in the gastrointestinal (GI) tract is likely variable depending on conditions at any given time, but the size of the bacterial population (E. coli >10⁶/g feces in cattle) is likely to be within a range that is conducive to QS (Maki and Picard, 1965; Alberghini et al., 2009).

LuxS is necessary for AI-2 synthesis and E. coli uses AI-2 for interspecies communication and global gene regulation (Sperandio et al., 2001). E. coli can also sense AI-1, AI-3, epinephrine/norepinephrine and other QS molecules (Sperandio et al., 2003; Smith et al., 2004; Walters and Sperandio, 2006; Walters et al., 2006; Connolly et al., 2015; Moreira and Sperandio, 2016) even though it does not produce these signal molecules with exception of AI-3-producing EHEC (Michael et al., 2001; Dyszel et al., 2010; Soares and Ahmer, 2011; Sabag-Daigle et al., 2012). Loss of mcpM regulation with deletion of luxS (Figure 5) and the combined effect of AI-2 when both E. coli 25 and BW25113 are co-cultured (Figure 1) are consistent with AI-2 influencing McpM synthesis. It is presumed that the lower GI tract of a cattle experiences relatively low osmolarity (Brouwer and Van Weerden, 1956) that is conducive to EnvZ/OmpR-mediated upregulation of mcpM. In this environment, AI-2 concentration likely provides “fine-tuned” control of expression so that even with permissive osmolarity, McpM is only synthesized when high-density bacterial populations experience conditions conducive to further population growth (e.g., after the host animal ingests a meal).

It is unclear how the concentration of AI-2 regulates mcpM expression. We know that decreased AI-2 concentration increases ompC expression and represses ompF expression during stationary-phase growth (Ren et al., 2004). OmpF is an outer membrane that must be present on susceptible cells before McpM is able to inhibit these cells (Zhao et al., 2015), and as a consequence Zhao et al. speculated that mcpM expression should mirror ompF expression (Zhao et al., 2017). For E. coli LsrR serves as an autoregulatory repressor protein that also regulates lsrACDB (AI-2 ATP-binding cassette transporter; Xue et al., 2009). Furthermore, when AI-2 is phosphorylated by LsrK, it subsequently binds to LsrR to regulate other genes associated to biofilm, membrane porins, and sRNA production (Li et al., 2007; Xue et al., 2009). A functional lsr AI-2 transport system is AI-2 (luxS) dependent (Taga et al., 2001).

Deletion of lsrR is associated with the up-regulation of sRNA micC through AI-2 signaling (Li et al., 2007). sRNAs micC and micF bind the mRNA of ompC and ompF to form MicC-ompC and MicF-ompF complexes that prevent translation of these mRNAs (Schmidt et al., 1995; Chen et al., 2004; Vogel and Papenfort, 2006). In M9 defined medium, conditions favoring OmpF expression in E. coli also favor the synthesis of McpM in E. coli 25 at late-log growth phase (Zhao et al., 2017). We speculate that during exponential growth phase, both micF and micC expression are kept at a base level similar to ompR expression (Figure S6B). When the PDI-producer strain reaches stationary growth phase, micF and micC are up-regulated to reduce synthesis of OmpF and OmpC. We surmise that the sRNA micF and micC also interact and regulate synthesis of McpM. IntaRNA prediction of pmic_−500/0mcpM (mcpM−500 to 0 bp promotor region) sequence against micC and micF suggests a potential interaction between mcpM mRNA (189–241 nt) and micC (7–66 nt); mcpM (290–345 nt) and micF (1–64 nt; Wright et al., 2014). As a result, sRNA micC and micF could potentially mediate the translation of McpM as suggested in our overexpression experiment (Figure 6), and this would provide a mechanism for down-regulating mcpM as the population enters a stationary growth phase.

Disruption of the QS AI-2 synthesis and uptake system in the microcin-PDI producer strain (E. coli 25) does not result in complete repression of McpM. Presumably, this is because OmpR interacts directly with the mcpM promoter as reported earlier (Figure S6A; Zhao et al., 2017). Herein we propose a McpM regulation mechanism model that incorporates both the EnvZ/OmpR two-component regulatory system and QS AI-2 (Figure 7).

The AI-2 molecule is derived from 4, 5-dihydroxy-2,3-pentadione (DPD), which is catalytically transformed by the LuxS from S-ribosylhomocysteine (Schauder et al., 2001). As cellular density increases, AI-2 molecules accumulate in the extracellular milieu. Via the Lsr ABC transporter (comprised of lsrACDB), AI-2 in medium is actively transported into permissible cells (Li et al., 2007) although passive diffusion of AI-2 across the cellular membrane is possible (Galloway et al., 2011). The Lsr transporter moves AI-2 into the bacterial cytoplasm where it is phosphorylated by LsrK (Xue et al., 2009). Phospho-AI-2 binds LsrR thereby blocking further repression of lsr-transporter genes, which leads to additional AI-2 uptake (Li et al., 2007). At this stage LsrR may bind to a “factor X” that interacts directly with the EnvZ/OmpR two-component system to activate transcription of micF [via OmpR which binds to the promoter of micF (Coyer et al., 1990; Delihas and Forst, 2001)] and/or directly regulates transcription of micC (Chen et al., 2004). sRNA micC and/or micF in turn block translation of mcpM mRNA. Because neither E. coli 25 ΔluxS nor E. coli 25 Δlsr mutants completely or continuously repress the mccPDI phenotype (Figure 1), it is likely that another pathway further contributes to regulation of micC and/or micF transcription.

The antibacterial activities of mccPDI are described under US Patent No. 9,492,500 for which DC is an author. The other authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.