Occurrence of Escherichia coli, Campylobcter, Salmonella and Shiga-Toxin Producing E. coli in Norwegian Primary Strawberry Production

The aim of this study was to investigate the bacteriological quality of strawberries at harvest and to study risk factors such as irrigation water, soil and picker’s hand cleanliness. Four farms were visited during the harvest season in 2012. Samples of strawberries, irrigation water, soil and hand swabs were collected and analyzed for E. coli, Campylobacter, Salmonella and STEC Although fecal indicators and pathogens were found in environmental samples, only one of 80 samples of strawberries was positive for E. coli (1.0 log10 cfu/g) and pathogens were not detected in any of the strawberry samples. The water samples from all irrigation sources were contaminated with E. coli in numbers ranging from 0 to 3.3 log10 cfu/g. Campylobacter (8/16 samples) and Salmonella (1/16 samples) were isolated from samples with high numbers of E. coli. The water samples collected from a lake had lower numbers of E. coli than the samples from rivers and a stream. The present study indicated continuous background contamination in the primary production environment. Although the background contamination was not reflected on the strawberries tested here, the results must be interpreted with caution due to the limited number of samples.

hazards and climate change, and consumption patterns and trends. There is a large global trade of strawberries, and in countries like Norway with a very short domestic harvesting season, there is a considerable import of strawberries. The aim of this study was to investigate the bacteriological quality of strawberries in primary production and to study risk factors such as irrigation water, soil and picker's hands in the primary production environment.

Strawberry Producers and Sampling Period
Four strawberry producers were selected for participating in the described study. Details of the farms are listed in Table 1. The sampling period was from 18 June 2012 (first visit on farm 1) to 8 August 2012 (fourth sampling on farm 4). All farms used surface water for irrigation. However, farm 3 used water from a large lake where the water was pumped from a few meters depth about 20 m from the shore.

Climatic Conditions during Sampling
The summer of 2012 was generally cold and wet, and the strawberry harvesting season was later and shorter than normal. A normal open field harvesting season is usually around 4-5 weeks, but in 2012 it was only three weeks. Data on mean daily temperature and cumulative day precipitation were collected from VIPS (Forecasts of plant diseases, pest and weeds) [13] for farms 1-3 and from eKlima (database of weather and climate data from Norwegian Meteorological Institute) [14], for farm 4. At the 'Lier' weather station (farm 1-3) the mean daily temperature from 15 June 2012-31 July 2012 was 14.9 °C and there was a cumulative precipitation in the same period of 196.4 mm. For the Buran weather station (farm 4), only precipitation was recorded. During the period 15 July 2012-15 August 2012, the precipitation was 73.5 mm. For this farm, the temperature was recorded at sampling by the farmer and was between 13 and 15 °C at the four sampling occasions.

Sampling of Strawberries, Soil, Water and Hands
Samples were collected four times at each farm during the harvesting season. Farms 1-3 were visited by the authors for sample collection, while farm 4 collected the samples themselves and sent the samples refrigerated by courier service to the laboratory within 24 h. Each time, five samples of strawberries in the field, five samples of soil, one sample of irrigation water, and five swabs of hands were collected. The strawberry samples were collected from the field being harvested at the time, and consisted of 10 strawberries each. Soil samples (50-100 g) were taken from around the area the plants from which the strawberries were picked. Water samples consisted of 5 L collected either from the water source (farms 1 and 4), at pump houses (farms 2 and 3), or from the irrigation equipment when irrigation took place during sampling. Since harvesting of strawberries is manual work, it was considered important to sample the hands of the pickers. This was done by swabbing one hand of five different pickers. Each swab was kept in 2.5 mL of peptone saline water. All samples were transported under chilled conditions to the laboratory and analyses were commenced within 24 h of sample collection.

Bacteriological Analyses
For the analysis of E. coli in strawberries, soil and swab samples, Petrifilm Select E. coli (3M™ Petrifilm™ Select E. coli, St. Paul, MN, USA) was used as described by the manufacturer. Briefly, five berries (approximately 10 g) was mixed by shaking by hand for 30 s with 9× volume of peptone saline (0.1% peptone (Difco,Sparks, MD, USA), 0.85% NaCl (Merck,Darmstadt, Germany)), and serially diluted. A volume of 1 mL of the appropriate dilutions was placed on the Petrifilm prior to incubation. The detection limit was 1.0 log10 cfu/g. For analysis of water for E. coli and enterococci, Colilert-18 (Colilert ® -18/Quanti-Tray/2000, IDEXX Laboratories, Wilmington, DE, USA) were used according to the manufacturer's instructions. A total of 100 mL of water was used for each analysis.
For the detection of pathogens, NMKL No. 71 [15] was used for Salmonella, NMKL no 119 [16] was used for Campylobacter and the then draft method of ISO TS 13136 [17] was used for the detection of STEC. Briefly, 25 g of strawberries or soil were mixed with the appropriate enrichment broth (Buffered Peptone Water, BPW (Oxoid Ltd., Basingstoke, Hampshire, UK), for Salmonella and STEC, and Bolton-broth (Oxoid) for Campylobacter) and incubated at 37 °C for Salmonella and STEC and 41.5 °C for Campylobacter prior to mixing with glycerol and freezing at −80 °C. For analysis of water, 1 L was filtered through a 0.45 µm filter (Millipore, Billerica, MA, USA). If the filter clogged, more than one filter was used per sample. The filter(s) was (were) transferred to 225 mL of the appropriate enrichment broths and incubated as described above. For all analyses of the pathogens, 6.55 mL of the primary enrichment were frozen with 3.45 mL 85% glycerol (Merck) at −80 °C awaiting further analysis. When ready to start the appropriate analysis, the frozen enrichment broths were placed immediately in a water bath at 50 °C and thawed for approximately 2.5 min until the ice disappeared. The tubes were removed and inverted 2-3 times during thawing. After thawing, the tubes were inverted five times. For Salmonella and STEC, one mL of the primary enrichment broth was transferred to 9 mL of BPW and incubated at 37 °C for three h. For further analysis of Salmonella, 0.1 mL of the BPW was transferred to 10 mL Rappaport-Vassiliadis-Soya Peptone Broth (Oxoid) and incubated at 41.5 °C for 24 ± 3 h, followed by plating on XLD (Oxoid) and Brilliance™Salmonella agar (Oxoid). The plates were incubated at 37 °C for 24 h and examined for typical colonies. The presence of Salmonella was confirmed by testing the colonies on Triple Sugar Iron agar (Difco) and Urea agar (agar base: Oxoid, with 40% urea (Sigma-Aldrich, St. Louis, MO, USA), and subsequently using API 20E for biochemical characterization (bioMerieux, Marcy l'Etoile, France).
For further analysis of STEC, DNA extractions, using a gradient-centrifugation method [18] using 1 mL of sample were performed prior to real time PCR as described in ISO TS 13136 [17]. If a sample was positive for stx1 and/or stx2 and eae, serotype specific PCR analyses for O26, O103, O111, O145 and O157 were carried out. Subsequently, if the sample was positive for one or more serotypes by PCR, isolation was attempted using automated immunomagnetic separation (AIMS) for the specific serotypes detected by the PCR followed by plating on CHROMagar™ O157 (CHROMagar, Paris, France) and sorbitol-MacConkey agar (SMAC, Oxoid) for serotypes O103, O111 and O145, CHROMagar™ O157 and R-Mac (MacConkey with rhamnose) (MacConkey Agar base, Difco, Rhamnose, Sigma-Aldrich) for O26 and CHROMagar™ O157 and CT-SMAC (SMAC with cefixime and tellurite, Oxoid) for O157. A total of 10 colonies from each plate were pre-agglutinated with the appropriate antiserum. If a colony agglutinated, it was isolated as a pure culture on blood agar, checked if it was presumptive E. coli by indole testing, re-agglutinated and tested for auto-agglutination. A presumptive E. coli of the aforementioned serotypes was then submitted to real-time PCR in order to confirm serotype and determine virulence factors.
For Campylobacter, 1 mL of of the frozen enriched Bolton broth was after thawing, transferred to 9 mL of Bolton broth containing 5% Laked horse blood (Oxoid), but without the selective antibiotic supplement (Oxoid), and incubated at 41.5 °C for 4-5 h. GENbox microaer (bioMerieux) was used to achieve the correct micro-aerobic atmosphere. After incubation, the samples were plated on mCCDA (Oxoid) and CampyFood Agar (bioMerieux) and incubated at 41.5 °C for 48 h using jars with GENbox microaer (bioMerieux). Typical presumptive positive colonies were isolated on blood agar and determined to Campylobacter spp. by testing motility and microaerophilic growth at 25 °C. Campylobacter spp. were further identified by a multiplex PCR [19].
Pure cultures of Campylobacter, Salmonella and STEC as positive controls were inoculated into their respective enrichment broths and submitted to the same treatment (enrichment and freezing) as the samples and analyzed in parallel after the samples were thawed and analyses continued.

Statistical Analyses
Proportions of positive samples were compared using Fisher's exact test. Counts of bacteria in different groups were compared using Wilcoxon/Kruskal-Wallis tests (Rank-Sums). Significance level was set at 0.05.

Results
The overall results (Table 2) indicated that the bacteriological quality of the strawberries sampled and tested at harvest was good, whereas the results for water, soil and hands were more variable.

Strawberries
No pathogens were detected in any of the 80 strawberry samples, and only one sample was positive for E. coli in very low numbers (1 log10 cfu/g).

Irrigation Water
E. coli was enumerated in all the 16 water samples in varying numbers, ranging from 0 to 3.3 log10 MPN/100 mL. There was a clear difference between the numbers of E. coli enumerated from the different farms, where farm 3, which used a large lake as irrigation water source, had significantly lower numbers of E. coli compared to the other three farms that used water originating from rivers and streams (Figure 1). Salmonella Newport was detected in one sample of water from the irrigation source (river) at farm 1. Campylobacter spp. was isolated from the water source (river and stream) from three out of four producers. Fifty percent of the water samples (8/16) contained Campylobacter, which was significantly higher than the frequency of occurrence of Salmonella, 6% (1/16). There was significantly higher numbers of E.coli in Campylobacter-positive than in Campylobacter-negative samples (Figure 2). The sample, from which Salmonella was isolated, also harbored a high number of E. coli (>3.0 log10 MPN/100 mL). The Campylobacter isolates were identified as C. jejuni and C. lari, while one isolate was not identified by the method used. STEC was not isolated from any sample. However, 10 out of 12 water samples from farms 1, 2 and 4 were presumptive positive for STEC by PCR (stx1 and/or stx2+, eae+, one or more serotype+), although no STEC isolates were obtained.

Soil
E. coli was detected in soil samples from all the farms visited ( Table 2). The numbers in the countable samples (detection level was 1 log10 cfu/g) ranged from 1 to 3.3 log10 cfu/g, with a median of 1.6 log10 cfu/g, indicating that the levels of E. coli in the soil samples were usually low. Farms 2 and 3 had the greatest proportion of E. coli positive soil samples, but not the highest level of E. coli in the samples. Salmonella was not detected in any sample. One soil sample was presumptive positive for STEC by having positive PCR results for stx, eae and one or more serotypes, but no isolates were obtained. It came from farm 4, which was the only farm using manure as fertilizer. There was no statistically significant difference in E. coli levels in soil samples from farm 4 compared to the other farms ( Figure 3).

Hands
The median and average levels of hand contamination in the different farms were similar. Farms 1 and 3 had higher proportions of samples from which E. coli was enumerated (8/20 and 7/20, respectively), although the difference was significant only compared to farm 2 (1/20). Maximal contamination level found in these farms was 2.2 log10 cfu/hand.

Discussion
The results from this study indicate that levels of E. coli and occurrence of the pathogens Campylobacter, Salmonella and STEC in the strawberries sampled and tested are low as no pathogens were detected and only one of 80 samples harbored E. coli with 1.0 log10 cfu/g. However, these results should be interpreted with caution as the number of samples tested is limited and samples were collected during only one season. Nevertheless, these results are in agreement with other studies carried out on strawberries pre-harvest. Delbeke et al. [10] enumerated E. coli in two of 72 samples in a study similar to ours, whereas in Yoon et al. [9] and Muherjee et al. [8] the levels of E. coli were below the detection limit.
For pathogens such as Salmonella and STEC or E. coli O157, the results were also comparable. However, the present study indicates that there is a background contamination in the primary production environment, shown by the continuous presence of E. coli in irrigation water and soil or growth substrate, as well as the isolation of pathogens such as Campylobacter and Salmonella. In our study E. coli was enumerated from 39% (31/80) of soil samples and in all the samples of irrigation water. Delbeke et al. [10] detected E. coli in approximately 30% of the substrate samples and 56% of the water samples. The higher prevalence of E.coli in water samples in our study can be explained by the fact that the Norwegian producers used surface water without treatment, while in Belgium borehole water and collected rainfall water stored in ponds with or without barriers were used. In contrast, Yoon et al. [9] studied greenhouse production of strawberries and did not recover E. coli from irrigation water or growth substrate.
The results show large variations of E. coli numbers depending on water source, with the lowest levels observed in the samples from the lake. We expected that the lake had lowest numbers of E. coli. This is due to the volume of water and the dilution effect. In addition, the water intake for farm 3, who used lake water, was at a few meters depth about 20 m from the shore. Water intakes from the rivers and stream were neither at such depth nor distance from the shore. The high numbers of E. coli in the water from the rivers and stream was most likely due to the wet conditions in the 2012 growth season with subsequently little irrigation. Therefore, the majority of the water samples were collected directly from the water source (river, stream, lake) and not the water that was actually used for irrigation. Previous results from one of the rivers indicate that the numbers of E. coli are much more variable during periods with much precipitation, while in drier periods the levels are less variable, but still rather high [20]. Since the water is of poor bacteriological quality, measures, such as changing of irrigation method or water treatment, should be considered in order to reduce the risk for contamination. In our study there were eight Campylobacter-positives and one Salmonella-positive sample from water used in strawberry production. These samples also had high levels of E. coli. When pathogens were identified in water, irrigation was not in use due to rainfall. It is well known that the use of contaminated irrigation water is a risk in the production of fresh produce. We did expect the possibility of isolating both Campylobacter and Salmonella from water. We thought it likely that we would find more Campylobacter positive samples than Salmonella, which was the case. In Norway, untreated water (surface water) is a known risk factor for campylobacteriosis [21]. The prevalence of Salmonella in Norwegian livestock is low, but Salmonella Typhimurium is endemic in wild birds [22] and in hedgehogs in only particular areas in the country [23]. The impact of the irrigation water on the quality of the product depends on the irrigation method and how close to harvest irrigation is used. Among the producers visited in this project, three (farms 1, 2 and 3) used overhead spray irrigation while the fourth (farm 4) used drip irrigation. Farmer 4 also informed that he knew that the irrigation water was contaminated and had, in cooperation with the local food safety authority, decided that the use of drip irrigation was the best option for his farm. In Norway, the farmers are required to test at least one sample of irrigation water and the sample should be collected close to the first harvest regardless of which product that is irrigated.
Approximately 6% of the samples analyzed for STEC were presumptive positive by PCR (stx1 and/or stx2+, eae+, positive for one or more serotypes), but no isolates were recovered. Several studies have shown similar results and challenges when screening similar environmental samples; the samples are positive by PCR, but only a very few isolates are obtained [9,24]. In our study, the only presumptive positive soil sample originated from the farm using manure as a fertilizer (farm 4). In accordance with common practice in Norway, manure was only applied on the field before planting. Furthermore, the farmer used small, fresh plants as starting material in field, and such plants do not start yielding strawberries until the following year. This practice suggests that it is unlikely that the manure applied to the field is the source of the presumptive STEC-positive soil sample. Previous results from Norway indicated that E. coli O157:H7 survived for up to 12 weeks from soil used for production of lettuce when temperatures varied between 12 °C and 15 °C in a climate room [25]. Another source for STEC would be the irrigation water, but the farmer informed us that no irrigation had been applied this season. Delbeke et al. [10] isolated STEC O26 from strawberry irrigation water and substrate on one farm that also had cattle. In a follow-up study on the particular farm one year later, fecal samples from the cattle were PCR positive for STEC O26, but no isolates were obtained for further comparison. The authors conclude that it was likely that the cattle were the source [10]. The presence or proximity to livestock is a known risk factor for STEC in fresh produce and in our study, farm 4 was also the only farm that had livestock (sheep). Another interesting result was the presence of the eae-gene in the soil samples. Out of 80 soil samples, 60% (48 samples) were positive for eae, with only one being presumptive positive for STEC (eae+/stx+, positive for one or more serotypes). The eae gene is generally not widespread, but Citrobacter rodentium may harbor this gene. It is also possible that atypical enteropathogenic E. coli (atypical EPEC) was present in the samples, but isolation was not attempted.
About 22% (18 out of 80) of the swab samples of workers' hands were positive for E. coli. This may suggest poor hand hygiene, but may also be a result of the wet harvesting season. None of the workers used gloves when picking strawberries as this is not required on Norway. Differences between farms were observed. All the farmers gave hygiene training to the workers in the beginning of the season, but it was variable how hygiene practices were followed up. All four farms had requirements to personal hygiene and toilets with washing stations, and two out of four farms had regular meetings with the workers. There was no effect of farm size on swab findings. Farm 1 and farm 4 were smaller farms with less than 50 workers during the season, while farms 2 and 3 were bigger with around 200 workers in the high season. When comparing our results to those from Delbeke et al. [9], we find a higher proportion of positive samples in Norway (22%, 18/80) compared to the Belgian results (7%, 4/57). This contrast can probably be explained by the fact that the samples in Norway were collected from open field operations in a wet harvesting season and inevitably, splashes from soil and dirt water may be transferred to the hands during picking. In Belgium the majority of the producers used either greenhouses or tunnel systems.
Although the results from the present study and others indicate that the bacteriological quality is generally good, it is important to keep in mind the limited number of samples that have been tested. A heterogeneous spread of contamination is expected in unprocessed fruits and vegetables, and no sampling scheme can guarantee the absence of pathogens. Therefore, focus on good agricultural practices and hygiene is recommended. The few samples positive for pathogens made it difficult to draw comparisons or to make inferences related to safe practices in primary production.

Conclusions
The results from the present study suggest that in spite of continuous background contamination in the primary production environment, the levels of E. coli and occurrence of Campylobacter, Salmonella and STEC on the berries is low. However, the results should be interpreted with caution as only a limited number of samples have been tested.