To simulate potential sources of contamination, lettuce leaves were inoculated with a compost slurry containing ARB and ARGs originating from a biological amendment, in addition we included antibiotic-resistant strains of pathogenic E. coli O157:H7 and Pseudomonas aeruginosa as a representative of an important genus of concern for spoilage, as a known source of ARGs whose presence could be detected and quantified. The effects of post-harvest interventions on the corresponding lettuce leaf resistome (i.e., total ARG profile) was determined via shot-gun metagenomic sequencing. Corresponding changes to the relative abundance of taxonomic groups were assessed by 16S rRNA gene amplicon sequencing. Culture-based methods were used to quantify inoculated antibiotic-resistant E. coli O157:H7 and P. aeruginosa, with tet(A) carried by the latter strain further quantified by quantitative polymerase chain reaction (qPCR). Inoculated lettuce leaves were washed, packaged under MAP (98% nitrogen), subjected to gamma irradiation, and stored at 4°C for 14 days. Controls included: washed, non-irradiated lettuce (irradiation control), unwashed irradiated lettuce (wash control), washed, irradiated leaves inoculated with DNA of lysed P. aeruginosa containing 10⁵ copies of tet(A) gene (ARG DNA control). Experimental conditions included: sodium hypochlorite washed and non-irradiated lettuce stored for 1 day, sodium hypochlorite washed and non-irradiated lettuce and stored for 14 days, sodium hypochlorite washed and irradiated lettuce stored for 1 day, and sodium hypochlorite washed and irradiated lettuce stored for 14 days. Following these treatments, all lettuce was subject to MAP and stored at 4°C for 1 or 14 days.

Multi-drug resistant (MDR) P. aeruginosa strain PAO1 (ATCC 47085) and MDR E. coli O157:H7 strain SMS-3-5 (ATCC BAA-1743), both isolates from humans or the environment, and not produce associated, were revived from freezer stocks stored at -80°C by streaking onto Tryptic Soy Agar (TSA, Becton Dickinson, Franklin Lakes, NJ, United States) and incubating for 24 h at 37°C to obtain isolated colonies. An isolated colony of P. aeruginosa was streaked onto Pseudomonas Isolation Agar (PIA, Becton Dickinson, Franklin Lakes, NJ, United States) and an isolated colony of E. coli O157:H7 was streaked on to Eosin Methylene Blue Agar (EMB, Becton Dickinson, Franklin Lakes, NJ, United States) followed by incubation for 24 h at 37°C. Separate single colonies from PIA and EMB were incubated separately in Tryptic Soy Broth (TSB, Becton Dickinson, Franklin Lakes, NJ, United States) at 180 rpm for 24 h at 37°C. Cells were washed two times in 0.1% (wt/vol) peptone (Becton Dickinson, Franklin Lakes, NJ, United States) and were separately suspended in 9 ml of 0.1% (wt/vol) peptone to prepare the inoculation solution.

Compost generated from the manure of antibiotic-dosed cows, in a companion experiment (Ray et al., 2017), and was used to introduce compost-associated bacteria, including ARB and ARGs to lettuce. In brief, manure from dairy cattle was collected from cows that did not receive antibiotics (control dairy manure) or dairy cattle prophylactically administered cephapirin and therapeutic levels of pirlimycin (dairy manure with antibiotics). Both control dairy manure and dairy manure with antibiotics were collected and composted at 55°C for 3 days followed by storage at -20°C. Compost slurry was prepared by blending 50 g of DC or DCAB with sterile deionized water (450 ml for non-inoculated treatments or 440 ml for inoculated treatments) for 45 s using a blender (Oster^®, Boca Raton, FL, United States). Experiments were conducted with three replicates of each type of slurry inoculum (DC or DCAB (n = 3), but because there were no statistically significant differences in the parameters examined in this study (total aerobic heterotrophic bacteria, relative abundance of bacterial communities and relative abundance of ARGs), all compost treatments were combined (referred as DC, dairy compost) for subsequent analysis (n = 6).

Romaine lettuce heads were obtained from a retail grocery store (Blacksburg, VA, United States). Damaged and cut leaves were removed. Remaining intact lettuce leaves were separated from the basal part, divided into 100 g portions and dip inoculated using two treatments: (1) Slurry generated from composted dairy manure (DC, dairy compost) (n = 6), and (2) The same DC slurry further spiked with the above described cocktail of MDR P. aeruginosa and E. coli (n = 3). To assure that quantifiable populations of known pathogens and spoilage-associated genera were introduced to the compost slurry, 10 ml of culture cocktail (containing 5 ml of E. coli O157:H7 and 5 ml of P. aeruginosa) was used to inoculate the compost slurry as described above. Inoculated leaves were air dried in a biological safety cabinet for approximately 15 min until visibly dry.

Inoculated and dried lettuce leaves were washed with sodium hypochlorite (XY-12, EcoLab) per manufacturer’s recommendation (50 ppm free chlorine) with a 2 min contact time followed by a 30 s rinse with tap water. Free chlorine was measured for XY-12 washes and tap water rinse using free chlorine test strips (Chlorine Test Paper, Ecolab). Excess liquid was removed by drying using a salad spinner followed by air drying for 15 min in a biological safety cabinet. Whole leaves were then packaged under 98% nitrogen (Sandhya, 2010) in PD961 EZ film bags (Cryovac, Duncan, SC; oxygen transmission rate: 6000–8000 mL/m²/24 h) using the Koch Ultravac vacuum packaging machine. A set of unwashed lettuce was retained, which served as a wash control.

MAP lettuce was maintained at 4^oC and shipped to the USDA Food Safety and Interventions Technologies Research Unit (Wyndmoor, PA, United States) where it was treated with 1.0 kGy gamma irradiation at 4°C using a Cs-137 self-contained gamma radiation source (Lockheed-Georgia, Marietta, GA, United States) at 3.95 kGy/h. Irradiated lettuce was returned to Virginia Tech where it was stored at 4^oC for 14 days. Controls of non-irradiated washed and non-irradiated unwashed lettuce were also stored under the same conditions. Lettuce samples were prepared for analysis on day 1 (3 days after inoculation, including transit time) and on day 14 (17 days after inoculation, including transit time). Before opening the packages for microbiological analysis on days 1 and 14, oxygen and carbon dioxide levels of the packages were determined using Mocon Dansensor CheckPoint Handheld Gas Analyzer (Mocon Inc., Minneapolis, MN, United States).

16S rRNA gene amplicon sequences obtained from unwashed lettuce were assigned to 16 phyla. Over 92% of sequences were classified (in order of most abundant) within only three phyla: Proteobactetria, Firmicutes, and Actinobacteria (Supplementary Table S1). Within the classified phyla, 29 bacterial classes were identified and ∼97% of sequences (in order of most abundant) belonged to the classes Bacilli, Gammaproteobacteria, Actinobacteria, Clostridia, Betaproteobacteria, and Alphaproteobacteria (Supplementary Table S1). From the classified sequences, 46 bacterial orders were identified. Bacterial orders with about 77% of sequences (from most to least abundant) belonged to Bacillales, Pseudomonadales, Actinomycetales, Methylophilales, Enterobacteriales, and Clostridiales (Supplementary Table S1). Overall 90 bacterial families were identified and 53% sequences were classified in Pseudomonadaceae, Planococcaceae, Nocardiospaceae, Methylophilaceae, Enterobacteriaceae, and Bacillaceae. Most represented bacterial genera (in order of abundance) belonged to Pseudomonas, Ureibacillus, Escherichia, Bacillus, Symbiobacterium, and Thermoactinomycetes. Together they constituted about 37% of the classified sequences in the genera category (Supplementary Table S1).

Overall, β-diversity of irradiated and non-irradiated lettuce was similar, immediately after irradiation (ANOSIM, R = 0.171, p = 0.002; Figure 1). However, after 14 days storage at 4°C the bacterial community profiles of irradiated and non-irradiated samples were significantly distinct, ANOSIM, R = 0.673, p = 0.002, Figure 1), indicating that re-growth of bacteria surviving the 1.0 kGy dose alters the community structure during subsequent storage. Bacterial community richness and evenness (Chao1, Shannon) were unchanged immediately after treatment, but significantly declined during the 14 days refrigerated storage period (Supplementary Table S2). Analysis of relative abundance of bacterial taxa indicated significant shifts in bacterial community composition of irradiated lettuce after 14 days of storage compared to non-irradiated lettuce (Table 1). There were no statistical differences in relative abundance of bacterial taxa between non-irradiated and irradiated day 1 lettuce samples, therefore the averaged day 1 samples were compared to day 14 non-irradiated and irradiated lettuce samples (Table 1). Irradiation resulted in a 30% decrease in Proteobacteria and a 26% increase in Firmicutes compared to non-irradiated lettuce stored for 14 days (p < 0.020, Wilcoxon, Table 1). Relative abundance of Gammaproteobacteria, which includes the inoculated antibiotic resistant pathogen and spoilage bacteria in this study, decreased significantly (30%) on irradiated lettuce compared to the non-irradiated lettuce after 14 days of storage (p < 0.001, Wilcoxon). A comparison of the relative abundance of the lettuce phyllosphere communities for Day 1 and irradiated Day 14 lettuce indicated a 40.2% increase in Bacilli and 17% decrease in Gammaproteobacteria after 14 days of storage (p < 0.026, Wilcoxon). No significant changes in Bacilli and Gammaproteobacteria were observed between Day 1 and non-irradiated Day 14 lettuce samples, indicating that irradiation was responsible for altering the bacterial communities that regrew during the 14 days period (Table 1). Storage for 14 days was associated with an increase in relative abundance of Carnobacteriaceae by 50.3% on non-irradiated lettuce and 62.2% on irradiated lettuce (p < 0.009, Wilcoxon). Pseudomonadaceae showed a 12.5% decline from days 1 to 14 irradiated samples (p = 0.020, Wilcoxon), while no significant differences were noted between day 1 and non-irradiated day 14 samples, indicating that irradiation was effective in preventing regrowth of part of the bacterial community (Table 1). Bacterial genera that decreased significantly (p < 0.050, Wilcoxon) in relative abundance on irradiated lettuce compared to non-irradiated lettuce on 14 days included: Pseudomonas (reduced by 20%), Yersinia (reduced by 6%), and genera of unclassified Enterobacteriaceae (reduced by 4%). In comparing samples immediately after irradiation, it is evident that the relative abundance of Pseudomonas decreased further during storage for 14 days (p = 0.021, Wilcoxon), indicating that the combination of post-harvest interventions was effective in reducing these potential spoilage bacteria. Yersinia was not detected on 1 day lettuce, yet its relative abundances climbed to 6.3% on non-irradiated lettuce at 14 days. Relative abundance of unclassified Enterobacteriaceae increased from 0.25% on 1 day samples to 4.3% in non-irradiated lettuce stored for 14 days. In contrast, the relative abundance of Yersinia and genera of unclassified Enterobacteriaceae did not increase in the irradiated lettuce stored for 14 days lettuce (Table 1).

The effect of irradiation on the ARG tet(A) that was contained within an intact microorganism (inoculated P. aeruginosa) or in naked DNA of the non-intact microorganism, was quantified using qPCR. No detectable copies of tet(A) were detected on samples on non-inoculated lettuce. Relative abundance of tet(A) genes normalized for the 16S rRNA genes measured in washed non-irradiated lettuce stored for 1 day was 1.45 × 10^-4± 1.35 × 10^-4 tet(A) gene copies/16S rRNA gene copies and did not change significantly after irradiation or storage of irradiated lettuce for 14 days (Supplementary Table S3). The absolute abundance of tet(A) (total tet(A) gene copies) in washed non-irradiated lettuce stored for 1 day was 1.56 × 10⁴ ± 6.04 × 10³. Thus, irradiation did not result in measureable changes to absolute abundance of tet(A) within intact cells or naked DNA on lettuce stored for 1 or 14 days (Supplementary Table S4). Storage of irradiated and washed lettuce over 2 weeks at 4^oC resulted in what appeared to be a small, but not statistically significant, reduction in absolute and relative abundance of tet(A) (Supplementary Tables S3, S4). There was no significant difference in relative abundance of tet(A) detected from treatments featuring intact cells versus naked DNA.

Populations of aerobic heterotrophic bacteria reduced from 4.70 ± 0.26 to 2.75 ± 0.88 log CFU/g following irradiation and 14 days of storage, indicating effectiveness of irradiation in reducing the total aerobic populations on romaine lettuce and that the populations did not return to initial levels during the 14 days of storage. In contrast, the total aerobic heterotrophic bacteria count for the non-irradiated control had bacterial counts 4.89 ± 0.08 log CFU/g. In particular, irradiation combined with washing and refrigerated storage resulted in significant reductions in the culturable log CFU/g of inoculated E. coli and Pseudomonas (Figure 2, ANOVA, p < 0.05). E. coli O157:H7 and Pseudomonas populations were decreased from ∼4.5 log CFU/g on non-irradiated lettuce to below the limit of detection (BLD; 2 log CFU/g) on irradiated lettuce (ANOVA, p < 0.05). No declines of inoculated E. coli O157:H7 or Pseudomonas sp. were observed on non-irradiated lettuce stored for 14 days (Figure 2). No significant regrowth of either inoculated bacteria occurred on irradiated lettuce during the 14 days of storage (Figure 2). This is consistent with 16S rRNA sequencing results, where no significant changes to the relative abundance of OTUs classified as Escherichia were observed between samples 1 day versus 14 days after irradiation. Due to the presence of substantial numbers of native Pseudomonas sp. on the lettuce phyllosphere, the use of antibiotic in culturing assays was necessary to facilitate detection of the inoculated strain. While irradiation did result in a loss in culturability of the inoculated Pseudomonas strain, other native members of the Pseudomonas population increased during the 14 days of storage post-irradiation (Table 1). While the antibiotic resistance statuses of the native strains cannot be confirmed, there were no colonies detected on PIA tetracycline plates from non-inoculated lettuce, indicating irradiation decreased the inoculated strain of Pseudomonas.

Inclusion of sanitizer lettuce wash water significantly decreased the bacterial community richness and evenness (Chao1, Shannon) compared to unwashed lettuce (Supplementary Table S2). Relative abundance of some bacterial taxa on the washed irradiated lettuce phyllosphere stored for 14 days were significantly different compared to the unwashed lettuce (Table 2). Most notably, there was a 36.4% decline in relative abundance of Proteobacteria and a 47% increase in relative abundance of Firmicutes when washing with sanitizer was employed prior to irradiation (Wilcoxon, p = 0.045, Table 2). Bacilli, a representative class of Firmicutes, increased by 46.0% and Proteobacteria classes- Gammaproteobacteria, Betaproteobacteria, and Alphaproteobacteria, were reduced by 27.9, 8.83, and 3.98%, respectively, after washing (Wilcoxon, p < 0.05, Table 2). Washing and storage for 14 days were associated with the increase in relative abundance of Carnobacterium by 63.6%, while the relative abundance of Pseudomonas and Escherichia decreased by 15.3 and 1.9%, respectively, compared to unwashed irradiated lettuce (Wilcoxon, p < 0.05, Table 2).

The populations of aerobic heterotrophic bacteria decreased from 6.53 ± 0.39 to 4.60 ± 0.13 log CFU/g following washing with sanitizer (sodium hypochlorite, 50 ppm free chlorine), packaging under modified atmosphere, and storage at 4°C for 1 or 14 days (ANOVA, P < 0.05, Figure 3). No significant regrowth of aerobic heterotrophic bacteria was observed on sanitizer-washed lettuce in MAP after 14 days of storage at 4°C (Figure 3). Sanitizer washing combined with MAP and storage at 4°C resulted in significant reductions in the culturable log CFU/g of inoculated antibiotic-resistant pathogenic and spoilage bacteria (Figure 3; ANOVA, P < 0.05). The E. coli O157:H7 population decreased by ∼ 2 log CFU/g on sanitizer-washed lettuce compared to unwashed lettuce (Figure 3; ANOVA, p < 0.05). Sanitizer washing also reduced Pseudomonas sp. resistant to tetracycline on lettuce from ∼5.9 log CFU/g on unwashed lettuce to ∼4.4 log CFU/g on sanitizer-washed lettuce. No significant regrowth of either inoculated bacteria occurred on unwashed or sanitizer-washed lettuce during the 14 days of storage indicating MAP and storage at 4°C was effective in suppressing the bacterial growth on sanitizer-washed lettuce over time (Figure 3).

Across all samples, 81,845 reads aligned to 577 ARGs were identified on romaine lettuce. Based on the mechanisms of resistance, detected ARGs were categorized into 22 classes. Genes conferring resistance to antibiotic drug classes glycopeptide, macrolide-lincosamide-streptogramin, multidrug, polymyxin, rifamycin, tetracycline, triclosan, and trimethoprim were detected in all samples. Spearman’s rank correlation analysis revealed significant correlations among bacterial families and ARGs found on the lettuce phyllosphere (Figure 4 and Supplementary Table S5, p < 0.05). A positive correlation in occurrence of Pseudomonadaceae OTUs and presence of ARGs conferring resistance to multidrug, triclosan, polymyxin, quinolone, bacitracin, aminoglyocide, peptide, β-lactam, and fosfomycin were noted (Supplementary Table S5; p < 0.05). Enterobacteriaceae OTU occurrence was positively correlated with detection of peptide, multidrug, β-lactam, polymyxin, quinolone, triclosan, and aminoglycoside genes (Supplementary Table S5; p < 0.05). In addition, compost-associated bacterial families – Bacillaceae, Thermoactinomycetaceae, Streptomycetaceae, Paenibacillaceae, Nocardiopsaceae, and Clostridiaceae exhibited significant positive correlation with some ARGs detected on lettuce phyllosphere (Supplementary Table S5, p < 0.05). Interestingly, detection of Carnobacteriaceae on the lettuce phyllosphere increased with cold storage and was significantly correlated with ARGs conferring resistance to several antibiotics (Supplementary Table S5, p < 0.05).

The β-diversity of resistome, as represented by the Bray-Curtis similarity of ARG occurrence, was significantly affected by irradiation (Figure 5). Non-irradiated lettuce carried distinct ARG profiles compared to irradiated lettuce (ANOSIM, R = 0.406, p = 0.008, Figure 5). Irradiation of lettuce resulted in significant decreases in relative abundance of six ARG classes compared to non-irradiated lettuce (Wilcoxon, p < 0.015, Table 3). The largest decreases in relative abundance of ARG classes (fold change in parentheses after class) on irradiated lettuce compared to non-irradiated lettuce were triclosan (4x), quinolones (3x), multidrug (2x), polymyxin (2x), and β-lactam (2x). ARG class conferring resistance to fosfomycin, which was detected at low relative abundance on non-irradiated lettuce, was not detected on irradiated lettuce (Table 3). Washing prior to irradiation was associated with further reductions to relative abundance of nine ARG classes encoding resistance (listed in order of largest change between unwashed and washed samples) to aminoglycoside, β-lactam, peptide, bacitracin, multidrug, sulfonamide, polymyxin, phenicol, and glycopeptide (p < 0.05, Wilcoxon; Table 3). In absence of irradiation, washing with sodium hypochlorite (50 ppm free chlorine) resulted in a 1.5 log reduction in total ARGs compared to unwashed lettuce. Significant decreases in relative abundance of only five ARG classes occurred, however, no significant decline was noted with the most abundant multidrug and triclosan classes, where irradiation was an effective strategy for reduction (Table 3, Wilcoxon, p < 0.05).

Further exploration of ARGs of elevated clinical concern revealed the presence of β-lactamases (OXA-50, TEM-17, TEM-91) and vancomycin-resistance (vanA) in the lettuce resistome. OXA-50 and TEM-17 were detected on all lettuce samples dipped in ARB-inoculated compost, TEM-91 was detected in 70% of lettuce samples treated with ARB-inoculated compost, and vanA was detected in 88% of all lettuce samples. Irradiation and washing did not significantly reduce the relative abundance of OXA-50, TEM-17, TEM-91, or vanA (Supplementary Table S6).