Correlation of in vitro biofilm formation capacity with persistence of antibiotic-resistant Escherichia coli on gnotobiotic lamb's lettuce

Abstract

Bacterial contamination of fresh produce is a growing concern for food safety, as apart from human pathogens, antibiotic-resistant bacteria (ARB) can persist on fresh leafy produce. A prominent persistence trait in bacteria is biofilm formation, as it provides increased tolerance to stressful conditions. We screened a comprehensive collection of 174 antibiotic-susceptible and -resistant Escherichia coli originating from fresh leafy produce and its production environment. We tested the ability of these strains to produce biofilms, ranging from none or weak to extreme biofilm-forming bacteria. Next, we tested the ability of selected antibiotic-resistant isolates to colonize gnotobiotic lamb's lettuce (Valerianella locusta) plants. We hypothesized that a higher in vitro biofilm formation capacity correlates with increased colonization of gnotobiotic plant leaves. Despite a marked difference in the ability to form in vitro biofilms for a number of E. coli strains, in vitro biofilm formation was not associated with increased survival on gnotobiotic V. locusta leaf surfaces. However, all tested strains persisted for at least 21 days, highlighting potential food safety risks through unwanted ingestion of resistant bacteria. Population densities of biofilm-forming E. coli exhibited a complex pattern, with subpopulations more successful in colonizing gnotobiotic V. locusta leaves. These findings emphasize the complex behavior of ARB on leaf surfaces and their implications for human safety.IMPORTANCEEach raw food contains a collection of microorganisms, including bacteria. This is of special importance for fresh produce such as leafy salads or herbs, as these foods are usually consumed raw or after minimal processing, whereby higher loads of living bacteria are ingested than with a food that is heated before consumption. A common bacterial lifestyle involves living in large groups embedded in secreted protective substances. Such bacterial assemblies, so-called biofilms, confer high persistence and resistance of bacteria to external harsh conditions. In our research, we investigated whether stronger in vitro biofilm formation by antibiotic-resistant Escherichia coli correlates with better survival on lamb's lettuce leaves. Although no clear correlation was observed between biofilm formation capacity and population density on the salad, all tested isolates could survive for at least 3 weeks with no significant decline over time, highlighting a potential food safety risk independently of in vitro biofilm formation.

Keywords: Escherichia coli; antibiotic resistance; biofilms; fresh leafy produce; lamb's lettuce; plant-microbe interactions.

Fig 1

In vitro biofilm formation capacity as determined by crystal violet (CV) staining and quantification of CV stain by OD measurement at 600 nm. (A) Biofilm formation capacity by medium. Changes in biofilm type of individual strains depending on the medium are highlighted with colored lines (green: change from ABTCAA [AB minimal medium with casamino acids] to LB-NaCl(0) [nutrient rich medium without NaCl] to a stronger biofilm type, red: change from ABTCAA to LB-NaCl(0) to a weaker biofilm type, black: no change in biofilm type between media). (B) Biofilm formation capacity by biofilm type (none/weak, moderate, strong, and extreme) in the two media. In (A and B), asterisks indicate statistically significant differences between with 𝛼 = 0.05. (C) biofilm formation capacity by source of isolation and by medium. Letters indicate statistically significant differences between groups (source and media) with 𝛼 = 0.05. (D) biofilm formation capacity by a representative number of phylotype isolates relative to their phylogroup and growth medium. Letters indicate statistically significant differences between phylogroups within a medium, with 𝛼 = 0.05, from linear mixed-effect regression models. In all cases, violin plots are depicted to show the distribution of the data and boxplots indicating the minimum, the interquartile range, and the maximum value. A strain-specific representation of the data can be found on Zenodo.

Fig 2

Correlation between bacterial biofilm formation capacity as determined by crystal violet (CV) staining and quantification of CV stain by OD measurement at 600 nm, in ABTCAA (AB minimal medium with casamino acids) and LB-NaCl(0) (nutrient rich medium without NaCl). Pearson’s correlation, r = 0.71. Highlighted strains designated with letter and number (e.g., H2) were selected for bacterial persistence screening on gnotobiotic V. locusta plantlets. As a reference, the identity function is shown as a continuous line: strains below or above the line showed stronger in vitro biofilm formation in ABTCAA or LB-NaCl(0), respectively.

Fig 3

Characteristics of bacterial strains selected for in planta assays. Antibiotic resistance profile, multiple antibiotic resistance (MAR) index, biofilm type in ABTCAA(AB minimal medium with casamino acids) and LB-NACl(0) (nutrient rich medium without NaCl), phylogroup (PG), and isolation source are indicated for each strain. Black: resistant; light gray: intermediate; white: sensitive. AM10: ampicillin 10 µg; FEP: cefepime; FOX: cefoxitin; CPD: cefpodoxime; AMC: amoxicillin-clavulanic acid; CRO: ceftriaxone; TPZ: piperacillin-tazobactam; CXM: cefuroxime; TOB: tobramycin; CN: gentamicin; NA: nalidixic acid; NOR: norfloxacin; SXT: trimethoprim-sulfamethoxazole; MI: minocycline; K: kanamycin; AK: amikacin; CIP: ciprofloxacin; LEV: levofloxacin; TMP: trimethoprim; SMZ: sulfonamide; TE: tetracycline; TEMO: temocillin; KF: cefalotin; CAZ: ceftazidime; CTX: cefotaxime. All strains were sensitive towards tigecycline, ertapenem, meropenem, imipenem, nitrofurantoin, fosfomycin, and colistin; n.d.: not determined. MAR index was calculated based on the screening of 32 antibiotics.

Fig 4

Correlation between bacterial culture data (CFU per g fresh weight [gFW]) and qPCR data (yccT copies per gFW) for each AR E. coli strain tested on gnotobiotic V. locusta, (A) all the data combined or (B) grouped by biofilm type (none/weak, moderate, strong, and extreme) determined in ABTCAA medium (AB minimal medium with casamino acids), at every sampling time point and from all conducted experiments (log₁₀ scale). Pearson’s correlation (r), P values (p), and linear equations are displayed. In (B), the continuous line represents the linear relationship between observed values, while the dashed line represents a reference line of perfect equivalency (y = x).

Fig 5

Bacterial persistence over time on leaves of gnotobiotic V. locusta. Each point depicts the AR E. coli abundance as copies of the yccT gene on each replicate plant at a given sampling point (n = 6 plants per strain and time point; N = 3 independent experiments). The outlined black point indicates the estimated marginal mean for a group from a generalized least squares model. Strains were grouped according to biofilm type determined in ABTCAA medium (AB minimal medium with casamino acids) (A), or shown as individual strains (B). Groups with different letters indicate significant differences from adjusted multiple pairwise comparisons (P < 0.05). gFW, gram fresh weight; dpi, days post-inoculation.

Fig 6

Distribution of biofilm-forming E. coli populations on gnotobiotic V. locusta. Histograms and probability density function of the abundance of E. coli population as yccT copies per gram of fresh weight (gFW⁻¹) over time (dpi, days post-inoculation) for strains grouped by biofilm type determined in ABTCAA medium (AB minimal medium with casamino acids): none/weak, moderate, strong, and extreme. Total population distribution is depicted as a black line, while subpopulations derived from mixture models are shown as green and magenta lines.

Fig 7

Live/dead stain micrographs of representative AR E. coli of different biofilm types determined in ABTCAA medium (AB minimal medium with casamino acids), colonizing the phyllosphere of gnotobiotic V. locusta. Images were taken at 7, 14, and 21 days post-inoculation (dpi) of the plantlets. Images were the result of overlaying the live/dead fluorescence channels: yellow corresponds to viable cells; magenta corresponds to not viable cells and plant autofluorescence. Biofilm types and corresponding representative strains: none/weak: strain H2; moderate: strain C13; strong: strain C30; extreme: strain B466. Scale bar = 50 µm. Triangles indicate plant leaf stomata.