Biocontrol of Escherichia coli O157:H7 by Enterobacter asburiae AEB30 on intact cantaloupe melons

Abstract

Escherichia coli O157:H7 causes >73,000 foodborne illnesses in the United States annually, many of which have been associated with fresh ready-to-eat produce including cantaloupe melons. In this study, we created a produce-associated bacterial (PAB) library containing >7500 isolates and screened them for the ability to inhibit the growth of E. coli O157:H7 using an in vitro fluorescence-based growth assay. One isolate, identified by 16S and whole-genome sequence analysis as Enterobacter asburiae, was able to inhibit the growth of E. coli by ~30-fold in vitro and produced zones of inhibition between 13 and 21 mm against 12 E. coli outbreak strains in an agar spot assay. We demonstrated that E. asburiae AEB30 was able to grow, persist and inhibit the growth of E. coli on cantaloupe melons under simulated pre- and post-harvest conditions. Analysis of the E. asburiae AEB30 genome revealed an operon encoding a contact-dependent growth inhibition (CDI) system that when mutated resulted in the loss of E. coli growth inhibition. These data suggest that E. asburiae AEB30 is a potential biocontrol agent to prevent E. coli contamination of cantaloupe melons in both pre- and post-harvest environments and that its mode of action is via a CDI system.

INTRODUCTION

Escherichia coli O157:H7 is a major foodborne pathogen that is associated with the majority of the E. coli outbreaks in the United States (Lim et al., 2010; Viazis & Diez-Gonzalez, 2011; Yang et al., 2017). The bacterium causes >73,000 foodborne illnesses, 2200 hospitalizations, and 60 deaths in the United States annually (Frenzen et al., 2005; Lim et al., 2010; Rangel et al., 2005); and its economic impact is estimated to be >$405 million (Heiman et al., 2015; Lim et al., 2010). From 1982 to 2012, there were 740 reported outbreaks of E. coli in the United States; among these 59% (438 outbreaks) were foodborne and 10% (44 outbreaks) were from produce including cantaloupe melons (Frenzen et al., 2005; Rangel et al., 2005). Fresh ready-to-eat produce can be contaminated with pathogenic bacteria, such as E. coli, during the pre- or post-harvest stages. Pre-harvest contamination can be caused by many factors including improperly composted manure, insect transmission, contaminated water, wild animals, contaminated harvesting equipment and farm workers (Frankel et al., 2009; Liu et al., 2013). In the post-harvest stage, contamination can result from unsanitary packing houses, transportation equipment, contaminated washing solutions and storage facilities (Castro-Rosas et al., 2012). Because many types of produce are consumed raw, little can be done to prevent illness once the produce becomes contaminated (Frankel et al., 2009; Rangel et al., 2005; Uhlig et al., 2021). Therefore, an effective method to prevent the contamination of produce in both the pre- and post-harvest environments is essential.

To prevent the contamination of produce in the pre-harvest environment, good agricultural practices must be strictly adhered to, including properly composting manure (if used), testing irrigation and other waters used (i.e. water used for pesticide and fertilizer application), ensuring wild animals are excluded from crop fields and minimizing the proximity of produce fields to confined animal feeding operations (Iwu & Okoh, 2019; Jung et al., 2014; Machado-Moreira et al., 2019; Park et al., 2015). In the post-harvest environment, several types of interventions are used to reduce/prevent contamination, most of which employ chemical or physical methods. The most common chemical interventions utilize aqueous chlorine sanitation solutions (Pao et al., 2009); however, chlorine-based compounds interact with organic compounds in the produce that reduce their effectiveness (Shen et al., 2012). Chlorinated compounds are also ineffective at reducing pathogens that become internalized into the produce due to tissue damage (e.g. wounds, stem scars, etc.) or by natural openings (e.g. stomata) (Erickson, 2012; Frankel et al., 2009). Lastly, chlorinated sanitizers have been shown to react with plant-associated compounds and form byproducts such as trichloromethanes that are known carcinogens (Fan & Sokorai, 2015). Physical methods, including ultraviolet light (Guan et al., 2012), hydrostatic pressure (Song et al., 2014) and gamma irradiation (Huang et al., 2016), have also been used. However, these methods can result in adverse sensory properties and suffer from limited consumer acceptance (i.e. irradiation). In addition, these methods usually do not completely eliminate the pathogens, and thus pathogen regrowth is possible.

In recent years, the use of biocontrol agents such as antagonistic bacteria and bacteriophages to control the growth and survival of pathogens on produce in both the pre- (Rabiey et al., 2020; Shahin et al., 2022; Zhang et al., 2023) and post-harvest (Johnston et al., 2009) environments has been explored. In this study, we constructed a produce-associated bacterial (PAB) library containing 7768 isolates and screened it for bacteria with the ability to inhibit the growth of E. coli O157:H7 in an in vitro fluorescence-based growth assay. We identified one isolate that could inhibit the growth of E. coli in vitro and on intact cantaloupe melons in simulated pre- and post-harvest environments and identified a contact-dependent growth inhibition (CDI) system responsible for the growth inhibition (Figure 1).

FIGURE 1. Methodology used to identify produce-associated bacteria (PAB) that inhibit E. coli in vitro and on intact cantaloupe melons.

EXPERIMENTAL PROCEDURES Isolation of produce-associated bacteria (PAB)

The protocol to isolate PAB was previously described by Tran et al. (2020). Briefly, several different types of plant material including ginger, broccoli, cauliflower, etc. were purchased from grocery stores in Berkeley, CA, and washed in phosphate-buffered saline with 0.01% tween 80 (PBS + T) with 20 g of 5 mm sterile glass beads and shaken at 200 rpm at 25°C for 60 min. The liquid was decanted and plated onto different types of bacteriological culture media as described previously. The resulting colonies were transferred to 96-well plates containing tryptic soy broth (TSB) (Oxoid) using a QPix420 automated picking robot (Molecular Devices, San Jose, CA, USA) (Figure 1). In total, 7768 isolates were curated.

Inhibition of E. coli O157:H7 growth by PAB in vitro

The method to identify PAB that could inhibit the growth of E. coli was described previously (McGarvey et al., 2019; Tran et al., 2020) with minor modifications (Figure 1). Briefly, cultures of PAB in 96-well plates containing TSB were incubated at 37°C for 24 h without shaking. E. coli RM2323 pWM1029 that constitutively expresses the green fluorescent protein (GFP) (Cooley et al., 2003) was cultured in Luria-Bertani broth (LB) (Difco) supplemented with 50 μg mL⁻¹ kanamycin (LB + Kan) at 37°C for 24 h with shaking at 200 rpm, diluted to approximately 4 × 10³ CFU mL⁻¹ with LB + Kan and 50 μL was placed into each well of black 96-well plates (Thermo Fisher Scientific, Roskilde, Denmark). Twenty-five μL of the PAB 24 h cultures were added to each of the wells, incubated at 30°C for 24 h and analysed for fluorescence in a Victor3 Multilabel Counter (Perkin Elmer, Waltham, MA, USA). PAB isolates from the wells with the least amount of fluorescence, and thus the least amount of E. coli RM2323 pWM1029 growth was selected for further analysis.

The ability of the most inhibitory PAB isolate, later identified as Enterobacter asburiae, to inhibit the growth of 12 E. coli outbreak strains (Table 1) was assayed using a modified agar spot assay as described previously (Saltaji et al., 2020). Briefly, the E. coli outbreak strains were cultured in TSB broth at 37°C for 18 h and diluted to approximately 4 × 10⁷ CFU mL⁻¹ in PBS, and 100 μL was spread onto TSA agar plates and allowed to dry for 30 min. Ten μL of a 24 h E. asburiae AEB30 culture was then spotted onto them and incubated at 25°C for 18 h, and the size of the zones of inhibition was recorded. To determine if E. asburiae AEB30 cell-free culture supernatant was inhibitory to E. coli, an 18 h culture of E. asburiae AEB30 or an 18 h co-culture of E. asburiae AEB30 and E. coli RM2323 were centrifuged at 5000 g and the supernatants filtered through a 0.2 μm PES syringe filter (Corning, Corning, NY, USA), concentrated 20-fold in a Savant SpeedVac concentrator (Thermo Fisher Scientific, Waltham, MA, USA) without heat and 20 μL applied to a 6 mm filter paper disk and assayed for the ability to inhibit the growth of E. coli RM2323 as described previously (Tran et al., 2020). We also tested culture-free supernatants made the same way but without the filtration step.

TABLE 1 Inhibition of 12 Escherichia coli outbreak strains by Enterobacter asburiae AEB30.

Strain	Serotype	Isolate source	Zone size (mm)a
RM2323	O157:H7	NA	17.3 ± 1.2
RM10234	O157:H7	Apple juice, 1996	15.0 ± 0.0
RM12840	O103	Human, OSPHL	15.0 ± 0.0
RM13514	O145	Romaine lettuce	16.7 ± 4.0
RM13516	O145	Belgium ice cream, 2007	15.7 ± 0.6
RM15012	O104:H4	Sprout, 2011	15.3 ± 1.2
RM1713	O26:H11	Human, CADHS	14.7 ± 0.6
RM3647	O121:H19	Human, CPK201	16.7 ± 1.2
RM6069W	O157:H7	Bagged baby spinach, 2006	15.3 ± 1.2
RM6607W	O157:H7	Ground beef, 1993	19.3 ± 2.9
RM9427	O45	Human, CDPHE	13.3 ± 0.6
RM9722	O111	Human, SDDOH	13.7 ± 0.6

Abbreviations: CADHS, California Department of Health Services; CDPHE, Colorado Department of Public Health and Environment; CPK201, Canadian Paediatric Kidney Disease Research Center; NA, not applicable; OSPHL, Oregon State Public Health Laboratory; SDDOH, South Dakota Department of Health.

^aZone of inhibition size shown as the mean value ± standard deviation from three independent experiments.

Growth and persistence of E. asburiae AEB30 on cantaloupe melon rinds

E. asburiae AEB30 was grown for 24 h in TSB at 37°C shaking (200 rpm) and diluted to approximately 1 × 10⁶ CFU mL⁻¹ in PBS. Five cm² sections of cantaloupe rinds were removed from the melons using a sterile scalpel and placed into sterile 100 × 25-mm-deep petri plates (Falcon, Corning, NY, USA). Ten l-μL drops of the diluted culture were deposited onto the surface of the rinds and incubated at 30°C, and growth on the rinds was quantified daily for 3 days by homogenizing the sections in 100 mL of PBS in a sterilized blender jar with an Osterizer Beehive blender (Oster, Neosho, MO, USA) for 1 min. The resulting solutions were serially diluted with PBS, plated onto Sorbitol-Mac Conkey (SMAC) (Difco) agar plates supplemented with 5 μg mL⁻¹ vancomycin and 25 μg mL⁻¹ lincomycin, incubated at 30°C for 24 h and counted.

The ability of E. asburiae AEB30 to persist on cantaloupe rinds after a simulated post-harvest dunk-tank inoculation and subsequent cold storage was determined by growing E. asburiae AEB30 in 600 mL of TSB for 24 h at 37°C with shaking (200 rpm). The resulting culture was placed into a sterile 3.8-L plastic bag containing a cantaloupe melon to mimic dunk-tank inoculation. The air from the plastic bag was carefully removed by hand so that the culture was in constant contact with the cantaloupe and incubated at 25°C for 30 min. The melons were blotted dry with paper towels and air-dried for 18 h at 25°C. Five cm² sections of the melon rinds were removed using a sterile scalpel and placed into sterile 100 × 25-mm-deep petri plates and incubated at 4°C. The persistence of E. asburiae AEB30 on the melon rinds was quantified on days 0, 3, 6 and 9 as described above.

The persistence of E. asburiae AEB30 on the surfaces of immature pre-harvest melons grown in a greenhouse (30°C, 16 h light) and still attached to the plant was evaluated by growing cantaloupe plants (Cucumis melo var. reticulatus) from seeds (Park Seed, SC, USA) in a greenhouse and pollinated by hand. When the melons were fully formed but still unripe (~30 days post-emergence), they were sprayed with 100 mL of a 24 h culture of E. asburiae AEB30, and the number of E. asburiae AEB30 was quantified at days 0, 3, 6 and 9 as described above.

Inhibition of E. coli O157:H7 growth on cantaloupe melon rinds by E. asburiae AEB30

For E. coli growth inhibition assays, we generated a spontaneous nalidixic acid and rifampicin-resistant mutant of E. coli RM10234, an isolate associated with the Odwalla apple juice outbreak in California in 1996 (CDC, 1996), as described previously (Devarajan et al., 2021) and was designated E. coli RM10234NR. The ability of E. asburiae AEB30 to inhibit the growth of E. coli RM10234NR on cantaloupe rinds after a simulated post-harvest dunk-tank inoculation was evaluated by coating melons with E. asburiae AEB30 as described above. Five cm² sections of the melon rinds were removed using a sterile scalpel and placed into sterile 100 × 25-mm-deep petri plates, and 10–1 μL drops of a 24 h culture of E. coli RM10234NR grown in TSB, diluted to ~1 × 10⁶ CFU mL⁻¹ in PBS, were placed onto the cantaloupe rinds. The rinds were incubated at 30°C for 24 h and quantified for E. coli by homogenizing the rind sections in 100 mL of PBS in a sterilized blender jar with an Osterizer Beehive blender (Oster, Neosho, MO, USA) for 1 min. The resulting solutions were serially diluted with PBS and plated onto Sorbitol-MacConkey (SMAC) (Difco) agar plates supplemented with 50 μg mL⁻¹ nalidixic acid and 100 μg mL⁻¹ rifampicin. As a control we treated melons with PBS using the same protocol. To evaluate the ability of E. asburiae AEB30 to inhibit the growth of E. coli RM10234NR on cantaloupe rinds under simulated post-harvest cold storage conditions, we coated cantaloupe melons as described above with E. asburiae AEB30, removed 5 cm² sections and inoculated them with E. coli RM10234NR as described above. The rinds were incubated for 3 or 6 days at 4°C, followed by 1 day at 30°C, and the number of E. coli RM10234NR CFU on the rinds was quantified as described above.

Identification of E. asburiae AEB30 by use of 16S rRNA gene and genomic DNA sequence analysis

Enterobacter asburiae AEB30 was inoculated into 2 mL of TSB broth and incubated for 24 h at 37°C with shaking (200 rpm). The genomic DNA was extracted using the Wizard Genomic DNA Purification Kit (Promega, Madison, WI, USA) as per the manufacturer's instructions. The 16S rRNA gene was PCR amplified in 50 μL reactions containing 25 μL High-Fidelity PCR Master Mix (Roche, Nutley, NJ, USA), 10 ng DNA and 10 μM of the primers 27F (AGAGTTTGATCMTGGCTCAG) and 1492R (TACGGYTACCTTGTTACGACTT) (Fredriksson et al., 2013) in a C1000/S1000 Touch Thermocycler (Bio-Rad, Hercules, CA, USA) under the following conditions: one cycle of 95°C for 5 min, 30 cycles of 94°C for 30 s, 55°C for 30 s, 72°C for 1.5 min, and one cycle of 10 min at 72°C. The resulting PCR product was sequenced using the 27F and 1492R primers and the BigDye Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, Foster City, CA, USA). Sequencing reactions were purified using the BigDye XTerminator Purification Kit (Applied Biosystems, Foster City, CA, USA); electrophoresis and readout were performed using an Applied Biosystems 3730XL Genetic Analyser (Applied Biosystems, Foster City, CA, USA). The forward and reverse sequences were aligned using the SeqManII software (DNASTAR Inc., Madison, WI, USA). Sequences were analysed using the BLAST software and the NCBI nr/nt database set to exclude models and uncultured/environmental sample sequences and limited to sequences from type material (available at http://blast.ncbi.nlm.nih.gov/).

To resolve E. asburiae AEB30 to the species level and identify possible secondary metabolite encoding operons, we sequenced the genome using a combination of Pacific Biosciences (PacBio) RS II and MiSeq Illumina platforms as described in detail by Tran et al. (2021). The genome was analysed for possible secondary metabolite-producing gene clusters using antiSMASH software version 7.0 (Blin et al., 2013; Blin et al., 2017; Blin et al., 2019; Blin et al., 2021; Blin et al., 2023; Medema et al., 2011; Weber et al., 2015) and for possible type Vb secretion systems using TXSScan software (Abby et al., 2014; Abby & Rocha, 2017; Denise et al., 2019) available at https://research.pasteur.fr/en/tool/txsscan-models-and-profiles-for-protein-secretion-systems/.

Mutagenesis of cdiA

A transposon was inserted into cdiA using the EZ-Tn5 <R6Kγori/KAN-2> Insertion kit (Lucigen, Madison, WI, USA) following the manufactures' instructions. Briefly, a 2000 bp fragment of cdiA, starting at the start codon, was PCR amplified in a 50 μL reaction containing 100 ng of AEB30 genomic DNA, 25 μL of 2 × KAPA HiFi HotStart ReadyMix and 0.3 μM of the primers 5′-GGTGGTGAGCTCATGGATACCCGTCACCCACC-3′ and 5′-GGTGGTCCCGGGCGCGGTAATCTGAACATTGC-3′ in a C1000/S1000 Touch Thermocycler (Bio-Rad, Hercules, CA, USA) under the following conditions: 1 cycle of 95°C for 3 min, 25 cycles of 98°C for 20 s, 67°C for 15 s, 72°C for 2 min and one cycle of 3 min at 72°C. The PCR product was digested with SacI and XmaI and ligated into a SacI and XmaI digested pUC19 vector, generating pUC19-cdiA. Transposon mutagenesis of pUC19-cdiA was carried out in a 10 μL reaction containing 200 ng of pUC19-cdiA, 85 ng of EZ-Tn5 transposon, 1 μL of EZ-Tn5 10 × reaction buffer and 1 μL of EZ-Tn5 transposase. The reaction was incubated at 37°C for 2 h, stopped by adding 1 μL EZ-Tn5 stop solution and heating at 70°C for 10 min. One μl of the reaction mixture was electroporated into 50 μL of One Shot TOP10 Electrocomp E. coli cells (Invitrogen, Carlsbad, CA, USA) using a Bio-Rad genepulser eletroporator with a 0.1-cm cuvette (Bio-Rad, Hercules, CA, USA) at 1.8 kV for 4.6 ms, and selected on LB + Kan agar plates. The resulting construct (pUC19-cdiA::Tn5) was confirmed by sequence analysis using the KAN-2 FP-1 primer (Lucigen, Madison, WI, USA). Electrocompetent E. asburiae AEB30 cells were prepared by inoculating 100 mL of TSB with 1 mL of a 24 h AEB30 culture and incubating at 37°C with shaking (200 rpm) until an OD600 of 0.6–0.7 was reached. Cells were placed on ice for 20 min, harvested by centrifugation, washed three times with ice-cold 10% glycerol, resuspended in 400 μL of ice-cold 10% glycerol, frozen on dry ice and stored at −80°C until needed. The cdiA::Tn5 fragment from pUC19-cdiA::Tn5 was PCR amplified as described above, purified using the QIAquick PCR purification kit (Qiagen, Red Wood City, CA, USA), and electroporated into E. asburiae AEB30 containing the pMJH46 helper plasmid (Addgene, Watertown, MA, USA). The resultant E. asburiae AEB30 cdiA::Tn5 mutant was selected on LB + Kan and confirmed by sequence analysis.

E. asburiae AEB30 cdiA::Tn5 mutant complementation

To complement E. asburiae AEB30 cdiA::Tn5, two overlapping (8000 bp) fragments of the cdiBAI operon were PCR amplified using the primers: 5′-GGTGGTTCTAGAAGGTGGTTTACGTGGGTAAC-3′ and 5′-TATTCACCACCTGGCTGTTC-3′ for fragment 1; 5′-CATCGATAACCGCAGCGT-3′ and 5′-GGTGGTTTAATTAACCAGGGCTCGTTGTTCTAAT-3′ for fragment 2. The PCR was carried out in a 50-μl reaction containing 25 μL of PrimeStar GXL Premix Fast 2 × (Takara Bio Inc., San Jose, CA, USA), 0.2 μM of each primer and 100 ng of AEB30 genomic DNA in a C1000/S1000 Touch Thermocycler (Bio-Rad, Hercules, CA, USA) under the following conditions: 1 cycle of 98°C for 3 min, 30 cycles of 98°C for 10 s, 60°C for 5 s, 68°C for 2 min and 1 cycle of 5 min at 68°C. Fragments 1 and 2 were cloned into pUC119-MCS-GmR via the StuI and SalI sites, and the SalI and PacI sites, sequentially. To construct pUC119-MCS-GmR, the gentamycin resistance gene was excised from pRGD-GmR (Addgene, Watertown, MA, USA) and ligated into pUC119-MCS (Addgene, Watertown, MA, USA) via the BtgI and HindIII sites. The recombinant plasmid, pUC119-MCS-GmR-cdiBAI, was electroporated into electrocompetent E. asburiae AEB30 cdiA::Tn5 as described above, plated onto LB + gentamycin 10 μg mL⁻¹ agar plates and confirmed by sequence analysis.

Statistical analysis

All experiments in this study were performed using a completely randomized design. Experimental data were analysed with one-way anova analysis using Sigma Plot (SSPS, Version 12). The results are presented as the mean value ± SD of at least three independent experiments in which each rind was a replicate.

RESULTS Identification of PAB capable of inhibiting the growth of E. coli in vitro

We constructed a PAB library containing 7768 isolates and screened it for the ability to inhibit the growth of E. coli RM2323 pWM1029 in an in vitro fluorescence-based growth assay described previously (McGarvey et al., 2019). Thirty-seven PAB isolates, designated AEB1-AEB37 (for Anti E. coli Bacterium), were identified that inhibited the fluorescence of E. coli RM2323 pWM1029 by ~30-fold after 24 h, as compared to a PBS-treated control. The isolates were also tested for the inhibition of E. coli RM2323 pWM1029 growth using an agar spot assay; however, only three of the isolates formed zones of inhibition. Among these isolates, AEB30 (isolated from ginger) produced the greatest fluorescence inhibition of E. coli RM2323 pWM1029 in vitro (Table 2) and the largest zone of inhibition in the agar spot assay (18 mm). We also performed the agar spot assay against 11 additional E. coli outbreak strains, all of which produced zones of inhibition of 13–21 mm (Table 1). Because zones of inhibition are often the result of a diffusible compound (e.g. antibiotics), we tested AEB30 cell-free culture supernatants for the ability to inhibit E. coli growth in the agar spot assay. However, cell-free culture supernatants made from 18 h cultures that were centrifuged and then filtered or only centrifuged, did not produce zones of inhibition nor did cell-free supernatants of AEB30 co-cultured with E. coli RM2323, indicating that the growth inhibition was cell contact dependent.

TABLE 2 Fluorescence inhibition of Escherichia coli RM2323 pWM1029 by Enterobacter asburiae AEB30.

Treatment	Avg. RFU (SD)	Fold difference	p-value
E. asburiae AEB30	6703 (2279)	29.6	<0.05
TSB control	198,663 (45,303)	NA	NA

Abbreviations: E. asburiae AEB30, E. coli RM2323 pWM1029 fluorescence after 24 h co-culture with E. asburiae AEB30; NA, not applicable; RFU, relative fluorescent units; TSB control, E. coli RM2323 pWM1029 fluorescence after 24 h culture in tryptic soy broth.

To identify AEB30, we sequenced its 16S rRNA gene (GenBank accession number MW826273) and compared it to sequences in GenBank via BLAST and found that it was >99% similar to several species within the genus Enterobacter. We also sequenced the AEB30 genome (accession number: CP046618.1; genome size: 4,748,641 bp; total genes: 4522) and compared it to other genomes in GenBank and found that it was 94.3% identical to E. asburiae L1 (GenBank accession number CP007546), 94.4% identical to E. asburiae FDAARGOS 1432 (GenBank accession number CP077411) and 94.3% identical to E. asburiae L1 delta-T1RM (GenBank accession number CP074584) (Tran et al., 2021). Because of these similarities, we designated this isolate E. asburiae AEB30.

Growth and persistence of E. asburiae AEB30 on cantaloupe melon rinds

To determine if E. asburiae AEB30 could grow and persist on cantaloupe melon rinds, we inoculated cantaloupe rinds with approximately 1 × 10⁴ CFU rind⁻¹ of E. asburiae AEB30 and observed that the bacterium was able to multiply to >1 × 10⁷ CFU rind⁻¹ after 24 h at 30°C and was able to persist at this level throughout the 3-day experiment (Figure 2A). To determine if E. asburiae AEB30 was able to persist under refrigeration temperatures, that is, post-harvest storage conditions, we inoculated intact cantaloupe melons using a simulated dunk-tank method and incubated the melons at 4°C for 9 days. At day 0, we observed approximately 1 × 10⁸ CFU rind⁻¹, after which the number of CFU rind⁻¹ declined significantly (p < 0.05) until day 9 when approximately 5 × 10⁴ CFU rind⁻¹ remained (Figure 2B). To determine if E. asburiae AEB30 could persist under simulated pre-harvest conditions, we sprayed a 24 h culture of E. asburiae AEB30 onto the surfaces of immature, greenhouse-grown cantaloupe melons that were still attached to the plant. At day 0, we removed 5 × 5 cm² sections of the rinds and observed approximately 1.8 × 10⁸ E. asburiae AEB30 CFU rind⁻¹, and the level did not change significantly (p > 0.05) throughout the 9-day experiment (Figure 2C).

FIGURE 2. Enterobacter asburiae AEB30 persistence on commercial melons at 30°C (A), 4°C (B), or under greenhouse conditions (30°C, 16 h light) (C). Data represent the mean values and standard deviations from at least three independent experiments.

Inhibition of E. coli growth on cantaloupe melon rinds

We examined the ability of E. asburiae AEB30 to inhibit the growth of E. coli RM10234NR, an isolate from the 1996 apple juice outbreak (CDC, 1996), under simulated pre- and post-harvest conditions. To simulate pre-harvest conditions, we inoculated 5 × 5 cm² sections of cantaloupe melon rinds treated with E. asburiae AEB30 or a PBS control with approximately 2 × 10⁴ CFU E. coli RM10234NR and incubated them at 30°C for 24 h. On the PBS-treated rinds, the levels of E. coli reached >1 × 10⁶ CFU rind⁻¹ after 24 h; however, on the E. asburiae AEB30-treated rinds, E. coli was unable to grow (Figure 3A). To simulate post-harvest conditions, we inoculated melon rinds treated as above and incubated them at 4°C for 3 or 6 days and then shifted the temperature to 30°C for 1 day to allow for E. coli growth. On the PBS-treated rinds, E. coli grew to approximately 3.2 × 10⁵ and 5.9 × 10⁵ CFU rind⁻¹ after 3 and 6 days, respectively. However, on the E. asburiae AEB30-treated rinds, the levels of E. coli decreased significantly (p < 0.05) (Figure 3B).

FIGURE 3. Escherichia coli growth on cantaloupe rinds at 30°C (A) or 4°C followed by 1 day at 30°C (B). Data represent the mean values and standard deviations from at least three independent experiments. PBS = phosphate-buffered saline-treated melons, AEB30 = Enterobacter asburiae AEB30-treated melons.

Identification of genes involved in E. coli growth inhibition

We performed in silico analysis of the E. asburiae AEB30 genome using the antiSMASH 7.0 software (Blin et al., 2023) that identifies gene clusters that are predicted to be involved in secondary metabolite biosynthesis, to identify genes responsible for E. coli growth inhibition. However, the only gene clusters identified were predicted to encode the siderophores amonabactin (57% similar) (Telford & Raymond, 1997) and aerobactin (66% similar) (Barghouthi et al., 1989) and the carotenoid-like pigment aryl polyene (100% similar) (Schöner et al., 2016), none of which have been shown to have antimicrobial activity.

Because the inhibition of E. coli growth was contact dependent, we examined the E. asburiae AEB30 genome for the presence of a contact-dependent growth inhibition (CDI) system using the TXSScan software (Abby et al., 2014; Denise et al., 2019). We identified three genes that were predicted to encode type Vb secretion system (T5bSS) translocators in the E. asburiae AEB30 genome (EAAEB30_00265, EAAEB30_00715 and EAAEB30_18470) (Figure 4). Analysis of the potential T5bSS operons revealed that EAAEB30_00265 was an intact T5bSS translocator, but it was followed by a partial cdiA and there was no cdiI present. EAAEB30_00715 was also an intact T5bSS translocator, but it had neither a cdiA nor a cdiI in its operon. However, EAAEB30_18470 had both a cdiA and a cdiI in its operon (Figure 4). To determine if this operon was responsible for the E. asburiae AEB30 growth inhibition of E. coli, we made a site-specific mutation in cdiA and found that the resultant E. asburiae AEB30 mutant (E. asburiae AEB30 cdiA::Tn5) lost the ability to inhibit the growth of E. coli in both the in vitro fluorescence inhibition assay and the agar spot assay (Table 3). However, when E. asburiae AEB30 cdiA::Tn5 was complemented with the cdiBAI operon in trans, it was able to inhibit the growth of E. coli in both assays (Table 3).

FIGURE 4. Two partial and one complete type Vb secretion system (T5bSS) operons in Enterobacter asburiae AEB30. Operon 1 spans nucleotide positions 53,101 to 66,365 and contains an intact cdiB (EAAEB30_00265) but a truncated cdiA (EAAEB30_00270) and no cdiI; Operon 2 spans nucleotide positions 158,737 to 160,455 and contains an intact cdiB (EAAEB30_00715) but no cdiA or cdiI; Operon 3 spans nucleotide positions 3,898,886 to 3,914,318 and contains an intact cdiB (EAAEB30_18470), an intact cdiA (EAAEB30_18465), and an intact cdiI (EAAEB30_18460). [Image omitted. See PDF.] cdiB, [Image omitted. See PDF.] cdiA, [Image omitted. See PDF.] cdiI.

TABLE 3 Enterobacter asburiae AEB30 strains' growth inhibition of Escherichia coli RM2323 pWM1029.

Treatment	Avg. RFU (SD)	Fold differencea	Zone sizeb (mm)
E. asburiae AEB30	6703 (2279)	29.6 (p < 0.05)	17.3 ± 1.2
E. asburiae AEB30 cdiA::Tn5	213,217 (58,921)	0.9	0 ± 0
E. asburiae AEB30 cdiA::Tn5 pcdiBAI	5713 (1999)	34.8 (p < 0.05)	17.7 ± 0.6
TSB control	198,663 (45,303)	NA	NA

Abbreviation: RFU, relative fluorescent units of E. coli RM2323 pWM1029 after 24 h co-culture with E. asburiae AEB30, AEB30 cdiA::Tn5, AEB30 cdiA::Tn5 pcdiBAI, or in TSB.

^aFold difference between the fluorescence of E. coli RM2323 pWM1029 after growth in TSB for 24 h or in co-culture with E. asburiae AEB30, AEB30 cdiA::Tn5 or AEB30 cdiA::Tn5 pcdiBAI.

^bSized of the zone of inhibition formed by E. asburiae AEB30, AEB30 cdiA::Tn5 or AEB30 cdiA::Tn5 pcdiBAI against E. coli RM2323 pWM1029 shown as the mean value ± standard deviation from three independent experiments.

DISCUSSION

At a minimum, a bacterial biocontrol agent should be able to grow and persist on the target produce in both the pre- and post-harvest environments and prevent the target pathogen's growth (Linares-Morales et al., 2018). In this study, we demonstrated that E. asburiae AEB30 was able to satisfy all these requirements. We demonstrated that E. asburiae AEB30 was able to grow and persist on cantaloupes in the pre-harvest environment (i.e. greenhouse-grown immature cantaloupes still attached to the plant) as well as in the post-harvest environment (i.e. mature melons stored at 30 and 4°C). We also demonstrated that E. asburiae AEB30 could inhibit the growth of E. coli RM2323 on cantaloupe melons under a simulated pre-harvest environment (i.e. 30°C) as well as under post-harvest cold storage (4°C) followed by a temperature shift to 30°C to simulate temperature abuse. In addition to these requirements, a biocontrol agent should also not cause disease in the host plant, impart undesirable smell or colour to the fruit, or cause early rot. E. asburiae ASB30 also satisfied these requirements as we did not observe any deleterious effects on cantaloupe seedlings inoculated with E. asburiae AEB30, nor did we observe any differences in the colour, smell or shelf life of melons sprayed or dunk-tank inoculated with E. asburiae AEB30 (data not shown).

Previous studies have demonstrated that strains of E. asburiae are able to inhabit the rhizosphere of several plant species including cotton, bean and cucumber (Hallmann et al., 1997; Hallmann et al., 1998; McInroy & Kloepper, 1995) and can become endophytic (Quadt-Hallmann et al., 1997). It has also been shown that E. asburiae can inhibit the colonization of plants by fungal and viral plant pathogens. For example, Kim et al. (2020) demonstrated that E. asburiae could inhibit root rot in Gom-chwi (Ligularia fisheri) caused by the plant pathogen Phytophthora drechsleri by suppressing the formation of zoosporangia and zoospore germination. In addition, Li et al. (2016) and Bharathkumar et al. (2009) demonstrated that E. asburiae-treated tomato and cucumber plants were protected from tomato yellow leaf curl virus and Colletotrichum orbiculare by the elicitation of induced systemic resistance in the plants. E. asburiae has also been shown to inhibit the colonization of the plant rhizosphere by the human pathogens E. coli and Salmonella enterica. Cooley et al. (2003) demonstrated that when gnotobiotic Arabidopsis thaliana roots were inoculated with a combination of E. coli and E. asburiae or with S. enterica and E. asburiae, the pathogen's growth was inhibited by >10-fold. They also demonstrated that E. asburiae could inhibit E. coli growth on the roots of lettuce grown in soil. However, they reported that E. asburiae could not suppress E. coli growth in complete or minimal media which is contradictory to our observation. Ye et al. (2009) reported similar results when they co-inoculated tomato plant roots with S. enterica and E. asburiae, but interestingly when they inoculated tomato fruits with both organisms it was not effective at reducing the level of S. enterica after incubation at 15°C for 7 days. In another set of experiments, they demonstrated that E. asburiae was able to reduce the growth of S. enterica on mung bean sprouts (Ye et al., 2010) and that spent culture filtrate was ineffective at inhibiting S. enterica growth, which is consistent with our study but contradictory to that of Cooley et al. (2003). It is likely that strain-to-strain differences among E. asburiae isolates account for the phenotypic differences observed in these studies.

We demonstrated that the growth inhibition of E. coli by E. asburiae AEB30 was cell contact dependent and identified a type Vb secretion system (T5bSS) that when mutated eliminated the ability of E. asburiae to inhibit E. coli growth. T5bSS were first described in E. coli by Aoki et al. (2005) who demonstrated that some strains of E. coli could inhibit the growth of other strains by a contact-dependent growth inhibition (CDI) system. CDI systems in the Enteriobacteriaceae are composed of three genes: cdiB, that encodes a β-barrel export protein that resides on the outer membrane of the cell; cdiA, that encodes a large stick-like toxic effector containing protein that is secreted via the Sec pathway to the periplasm and then displayed on the outer membrane by CdiB; and cdiI that encodes a small immunity protein that binds the toxic effector on CdiA, neutralizing its toxic activity in the attacking cell before export (Aoki et al., 2005; Leo et al., 2012; Morse et al., 2012; Ruhe et al., 2013). Upon contact with target cell, the receptor-binding domain of CdiA interacts with a receptor on the target cell, resulting in the cleavage and release of the CdiA C-terminal toxic effector domain (CdiA-CT). The CdiA-CT are tRNA, rRNA or DNA-specific nucleases or ionophores that bind to target cell receptors and are taken up by the cell (Aoki et al., 2010; Ruhe et al., 2018; Webb et al., 2013) resulting in the degradation of nucleic acids or the collapse of the protonmotive force, causing cell growth inhibition and death (Aoki et al., 2009; Aoki et al., 2010; Morse et al., 2012). Although T5bSS have been proposed to be involved in biocontrol (Barret et al., 2011; Marian et al., 2021) to our knowledge this is the first report of a biological control agent employing a T5bSS to inhibit the growth of its target organism.

CONCLUSIONS

This study describes the use of the bacterium E. asburiae AEB30 as a biocontrol agent for the prevention of pathogenic E. coli growth on cantaloupe melons and possibly other types of produce. We demonstrated that E. asburiae AEB30 can grow and persist on cantaloupes and prevent the growth of pathogenic E. coli in simulated pre- and post-harvest environments. We also demonstrated that the growth inhibition of E. coli is contact dependent via a CDI system as E. asburiae AEB30 cdiA mutants are not able to inhibit E. coli growth in the in vitro fluorescent assay or produce a zone of inhibition in the agar spot assay. However, when this mutation is complemented these traits are restored. Interestingly, the production of a zone of inhibition in an agar spot assay is usually the result of the bacterium producing a diffusible compound (e.g. antibiotic). However, in silica analysis of the E. asburiae AEB30 genome failed to identify any predicted antibiotic gene clusters and 18 h E. asburiae cell-free culture supernatants failed to produce zones of inhibition. It is possible that E. asburiae AEB30 cells, which are known to be mobile, travel away from the bacterial spot and kill E. coli in the near proximity. Future studies will need to be performed to examine this phenomenon.

AUTHOR CONTRIBUTIONS

Thao D. Tran: Data curation; methodology; writing – original draft. Sang In Lee: Data curation; formal analysis. Robert Hnasko: Conceptualization; methodology; writing – review and editing. Jeffery A. McGarvey: Conceptualization; investigation; project administration; resources; supervision; writing – review and editing.

FUNDING INFORMATION

This work was supported by the USDA, Agricultural Research Service, National Program 108: Food Safety (Animal and Plant Products).

CONFLICT OF INTEREST STATEMENT

The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT

The sequence of the E. asburiae AEB30 16S rRNA gene was deposited into GenBank (accession number MW826273) available at https://www.ncbi.nlm.nih.gov/nuccore/MW826273. The genome sequence of E. asburiae AEB30 was deposited into GenBank (accession number CP046618) available at https://www.ncbi.nlm.nih.gov/nuccore/CP046618.1/.