Seed inoculation with antagonistic bacteria limits occurrence of E. coli O157:H7gfp + on baby spinach leaves

During the last decades, outbreaks of foodborne illnesses have increasingly been linked to fresh and/or minimally processed fruit and vegetables. Enterohemorrhagic Escherichia coli was the causal agent for major outbreaks in Europe with leafy green vegetables and sprouts. To improve food safety, microbial antagonism has received attention during recent years and could be one of the solution to prevent contamination of food borne pathogens on fresh produce. Here we investigate the antagonistic effect of three bacterial strains (Pseudomonas orientalis, P. flavescens and Rhodococcus sp.) isolated from spinach leaves against E. coli O157:H7gfp + under laboratory and greenhouse conditions. Our results shows that significantly less culturable E.coli O157:H7gfp + were retrieved from the spinach canopy subjected to antagonist seed treatment than canopy inoculation. Seeds inoculated with Rhodococcus sp. significantly reduced growth of E. coli O157:H7gfp + compared with the other antagonists. The result from the in vitro study shows a significant reduction of growth of E. coli O157:H7gfp+, but only after 44 h when E. coli O157:H7gfp + was propagated in a mixture of spent media from all three antagonists. The antagonistic effect on phyllospheric E.coli O157:H7gfp + observed after seed inoculation with Rhodococcus sp. might be an indication of induced resistance mechanism in the crop. In addition, there was a small reduction of culturable E.coli O157:H7gfp + when propagated in spent media from all three antagonists. Nutritional conditions rather than metabolites formed by the three chosen organisms appear to be critical for controlling E. coli O157:H7gfp+.


Backround
Consumption of fresh fruit and vegetables is associated with a healthy diet and lifestyle and has increased in popularity during the past decade. However, fresh produce is also recognised as a possible vehicle for transmission of food-borne human pathogens and the number of reported outbreaks is continually increasing [1][2][3]. Among these pathogens, Escherichia coli O157:H7 is the most frequently identified serotype [4]. To mitigate the occurrence and risks associated with shigatoxigenic E. coli (STEC) in horticultural value networks a hurdle approach could be used [5]. The hurdle technology is combining several methods that is not effective by its own but together can serve as a tool to reduce or even eliminate pathogens in food products [5]. Introduction into cultivation systems of microorganisms with Open Access *Correspondence: maria.e.karlsson@slu.se 1 Microbial Horticulture Division, Department of Biosystems and Technology, Swedish University of Agricultural Sciences, PO Box 190, 234 22 Lomma, SE, Sweden Full list of author information is available at the end of the article antagonistic characteristics to food-borne pathogens is one such hurdle.
The phyllosphere as a habitat varies in structure and is constantly exposed to fluctuations in environmental factors such as temperature, humidity, nutrient availability and light [6][7][8]. Thus, its inhabitants need to possess a certain degree of plasticity to adapt to these changes [9]. The success of a microbe in its community is dependent how well it can compete with other community members [1]. For a human pathogen seeking to invade a habitat outside its normal niche, its fate is dependent on resource availability and the structure of the microbial community [10]. To successfully outcompete microbial invaders, such as shigatoxigenic E. coli, antagonists must: (i) share the same niche, (ii) have similar nutrition preferences to the undesired microbe, but a faster growth rate under ambient conditions, and (iii) fit into the existing microbial aggregate on the leaf surface. Formation of metabolites deleterious to the intruder under ambient conditions offers an additional mode to exclude microbial invaders.
This study investigated the potential for counteracting culturable E. coli O157:H7gfp + on spinach leaves by enrichment with three endemic phyllosphere bacteria. All three bacterial strains are known to display in vitro antagonistic properties against E. coli [11]. The objectives of this study was to investigate bacterial antagonism against E. coli O157:H7gfp + on spinach baby leaves using two different inoculation methods and also investigate in vitro and whether the metabolites of the antagonists could limit growth of E. coli O157:H7gfp+.
The hypotheses tested were that: (i) Canopy spray, but not seed inoculation, limits the occurrence of culturable E. coli O157:H7gfp + on spinach leaves. (ii) Metabolites from antagonists reduce the growth of E. coli O157:H7gfp + when propagated in broth with metabolites from the antagonists.
Likewise, E. coli O157:H7gfp+ (log CFU) were significantly lower when E. coli O157:H7gfp + was exposed to plants that were seed-inoculated with multiple bacterial strains (Estimate (95% CI) -1.07 (-0.78 --0.35) ( Fig. 1). No significant impact on culturable E. coli O157:H7gfp + was found after seed inoculation with either of the two Pseudomonas strains. Interestingly, despite two canopy sprays with the individual or multiple strain treatment, E. coli colonised spinach leaves to the same extent in the antagonist and the control treatment. There was a tendency for higher E. coli O157:H7gfp + counts in the single-strain antagonist treatments than in the treatment with a mixture of all three antagonists, but the differences were not significant.

Impact of bacterial metabolites on growth of E. coli O157:H7gfp+
The initial nutrient concentration in the overnight culture medium was decisive for the growth of E. coli O157:H7gfp + in the spent media. The nutritional contribution from the metabolites seemed to promote rather than reduce the growth of E. coli O157:H7gfp+. This was also confirmed when calculating the generation time for E. coli O157:H7gfp + in all treatments except that with the spent media from Rhodococcus, where the generation time was shortened compared with the two control treatments (Table 1). Interestingly, the fastest generation time was seen in the multiple-strain treatment (Table 1). Treatment with spent media filtrate from P. flavescens, P. orientalis and the multiple-strain mixture reached the maximum number of E. coli O157:H7gfp + after only 22 h, whereas in the control this took 44 h (Table 1). When E. coli O157:H7gfp + was grown in full-strength TSB together with the multiple-strain spent media filtrate, a significant reduction in E. coli O157:H7gfp + numbers was observed after 44 h (Estimate (95% CI )-1.22 (-1.59 --0.85), p < 0.001, Fig. 2 A). There was also a significant reduction of growth when E. coli O157:H7gfp + was propagated in the spent media filtrate of Rhodococcus sp (Estimate (95% CI)-1.73 (-2.10 --1.36), p < 0.001, Fig. 2 A). When E. coli O157:H7gfp + was grown in lowconcentration TSB (10%) with spent media filtrate, a significant decrease in growth was seen for the filtrate from

Discussion
Bacterial antagonism is not a new phenomenon. Lactobacilli have also been shown to have inhibitory effects on different pathovars of E. coli in vitro [12]. Both Pseudomonas strains and the Rhodococcus sp. strain used in the study originated from spinach leaves. Antagonistic potential of both genera is frequently mentioned in the literature with respect to various plant pathogens producing antibacterial substances [13][14][15]. In the present study, seed inoculation, but not canopy inoculation, with  indicate that induced resistance rather than niche exclusion might be mechanisms to further follow up on. The 1 log reduction of E. coli O157:H7gfp + achieved by Rhodococcus sp. is encouraging and the differences observed are of biological relevance, but not of food safety relevance. Rhodococcus belongs to the phylum Actinobacteria, a Gram-positive bacteria commonly present in the environment. Rhodococcus also produces secondary metabolites that have an antibiotic effect on both Gram-positive and Gram-negative bacteria [16,17]. However, the Rhodococcus strain used in the seed inoculation treatment does not substantially colonise the phyllosphere, so other mechanisms must affect the capacity of E. coli O157:H7gfp + to establish on spinach leaves, for example microbial community distortion towards a composition unfavourable to the invader or induced resistance plant mechanisms. In future studies, it would be highly interesting to follow the dynamics of Rhodococcus and to examine plant-related processes in the rhizosphere and canopy post seed inoculation. As the genus Rhodococcus is classified as a BSL2-organism, it is imperative to remove all possible doubts regarding its genetic make-up before it is tested against other serovars of shigatoxigenic E. coli.
The present study did not examine the capacity of the organisms to form secondary metabolites under chosen conditions (nutrient composition, temperature, length of pre-culture) or characterise such potential metabolites. To shed light on potential causes of decreased occurrence of the model strain, the impact of spent media extracts collected during the stationary phase on the model strain was examined. The spent media were collected from broth with different nutrient concentrations and a potential effect of metabolites was expected to emerge under nutrient-restricted conditions. However, production of secondary metabolites may not be produced in planktonic culture since it is strongly linked to biofilm formation mechanisms [18] and therefore it is doubtful that the antagonist in our study had enough time to produce biofilm and secondary metabolites only after 50 h. The results showed that nutritional conditions, rather than metabolite formation, were crucial for the growth of E. coli O157:H7gfp+ (Fig. 2 A-C). Growth of E. coli O157:H7gfp + was influenced only to a very limited extent by secondary metabolites potentially formed by the three antagonists in the beginning of the stationary phase.

Conclusions
Seed inoculation rather than canopy spray has the potential to mitigate the occurrence of culturable E. coli O157:H7gfp+, however, the choice of antagonist strain or multiple strain composition is essential. To explain the impact of seed inoculation with Rhodococcus sp. on E. coli O157:H7gfp+, induced resistance mechanisms should receive attention in future experiments. Nutritional conditions rather than metabolites formed by the three chosen organisms appear to be critical for controlling E. coli O157:H7gfp+.

Materials and methods
A two-factor greenhouse experiment was set up, where factor 1 was treatmentwith antagonistic bacteria (Pseudomonas orientalis, P. flavescens, Rhodococcus sp.) and factor 2 was mode of treatment (seed or canopy inoculation). The model pathogen strain used was E. coli O157:H7 gfp+ (verotoxin-1 negative, verotoxin-2 negative, eae gene positive) obtained from the Swedish Institute for Communicable Disease Control (Solna, Sweden), which was labelled with green fluorescent protein plasmid [19]. The three bacterial strains were isolated from spinach leaves in a previous study and identified with PCR and Sanger sequencing. The antagonistic activity against E.coli CCUG 29300T was determined with the perpendicular streak method [11].

Inoculum preparation
The model pathogen E. coli O157:H7gfp + was propagated in LB agar supplemented with ampicillin and arabinose and incubated at 37℃ according to procedure described earlier in [20]. The three antagonists were propagated on TSA (Tryptic soy agar) plates and incubated overnight at 25 °C. A single colony was then transferred to 5 ml of TSB (Tryptic soy broth) and incubated overnight at 25 °C. The overnight cultures of the three antagonists and the pathogen were repeatedly washed with 0.85% NaCl and centrifuged at 3000 x g at 4 °C for 10 min (AvantiTM J-20 Centrifuge, Beckman Coulter Corporation, Brea CA, USA) and re-suspended in sterile NaCl (0.85% followed by 0.085%),Cell density was adjusted to OD 0.2 with NaCl (0.085%) which correspond to log 8.9 CFU mL − 1 for P. orientalis and P. flavescens and log 8.2 CFU mL − 1 for Rhodococcus ssp. Cell density of E. coli O157:H7gfp + was adjusted to OD 1 which correspond tolog 9.7 CFU mL − 1 . For canopy inoculation, the final inoculum density of single-or multiple-strain suspensions was adjusted to log 7.7 with 0.085% NaCl. The volume of the spray inoculum was 150 ml. Multiple strain treatments consisted of equal numbers (CFU mL − 1 ) of the individual strains. For seed inoculation, 3 g of spinach seeds were imbibed overnight in 50 mL of the antagonist inoculum suspension at 25 °C.
Canopy inoculation of the antagonistic strains occurred on day 23 after sowing and on day 28 after sowing. Canopy spraying with E. coli O157:H7gfp + was performed one day before harvest of the spinach leaves. Spinach leaves were harvested aseptically and leaf-associated microbes were extracted as previously described by [21]. Briefly, leaves were transferred to a sterile plastic bag with filter (Separator 400, Grade products Ltd., Coalville, UK) and 50 ml of TRIS-buffer (0.01 M Trishydroxymethyl, pH 8,) was added. The bag was placed in a smasher lab blender (Biomérieux Industry, Durnham NC, USA) for 30 s normal mode. The extraction suspension was 10-fold serially diluted in NaCl solution (0.85%). For viable counts, E. coli O157:H7gfp + was cultured on LB agar supplemented with ampicillin (100 µg mL − 1 ) and L-arabinose (0.2 g mL − 1 ) at 37 ℃ and fluorescent colonies were enumerated after 18 h under UV light. The Pseudomonas and Rhodococcus isolates were re-isolated on King Agar B (Merck) and full-strength TSA, respectively, supplemented with rifampicin (100 µg mL − 1 ) at 25℃ for 24 h.

Impact of metabolites of antagonistic bacterial strains on growth of E. coli O157:H7gfp+.
The potential impact of metabolites of P. flavescens, P. orientalis and Rhodococcus sp. on E. coli O157:H7gfp + was tested in in vitro assays in a two-factor experiment, using the supernatants from overnight cultures of the three antagonists (factor 1) and three concentrations of the propagation broth (0.01x, 0.1x, 1x; factor 2) and with six replicates. The antagonists and E. coli O157:H7gfp + was propagated in three concentrations of TSB overnight. The supernatant was membrane-filtered (syringe filter 0.2 μm, Thermo Fisher) after centrifugation for 10 min at 3200 x g to remove remaining cells that were not collected by centrifugation. This treatment will not lyse the cells but removes them from the supernatant. This filtrate is referred to as spent media [22] Cells from an overnight culture of E. coli O157:H7gfp + were washed in 0.85% NaCl through repeated twice centrifugation (3200 xg for 10 min), finally suspended in LB and cell density was adjusted to 5.3 log CFU mL − 1 (OD 620 ). Aliquots of 5 mL of membrane-filtered antagonist suspension were inoculated with 100 µL of E. coli O157:H7gfp+. Spent media from an overnight culture of E. coli O157:H7gfp + was used as a first control and E. coli O157:H7gfp + in LB media without any spent media filtrate as a second control. All tubes were incubated at 25 °C. Culturable E. coli O157:H7gfp + was enumerated after 6, 10, 22, 26, 30, 44 and 50 h. Plates were incubated at 25 °C and analysed after 24 h.

Statistical analysis
Generalised linear model was used to investigate the effect of each antagonist and the mode of application on the survival of E. coli O157:H7gfp + on spinach leaves. Log-transformed number of culturable E. coli O157:H7gfp+ (log CFU + 1) was used as the response variable and treatment and inoculation method as co-variates. The analysis were performed in R 3.5.3 (R Core Team, 2019). In the in vitro study, E. coli O157:H7gfp + propagated in its own metabolites was used as the response variable and treatment with the metabolites from the antagonists as co-variate.