Distribution and decline of human pathogenic bacteria in soil after ...

Journal of Applied Microbiology ISSN 1364-5072

ORIGINAL ARTICLE

Distribution and decline of human pathogenic bacteria in soil after application in irrigation water and the potential for soil-splash-mediated dispersal onto fresh produce J.M. Monaghan1 and M.L. Hutchison2 1 Fresh Produce Research Centre, Harper Adams University College, Newport, Shropshire, UK 2 Hutchison Scientific Ltd, Wolverhampton Science Park, Wolverhampton, West Midlands, UK

Keywords crops, decline, rain splash, soil distribution, zoonotic agents. Correspondence Mike Hutchison, Hutchison Scientific Ltd, Wolverhampton Science Park, Wolverhampton WV10 9RU, West Midlands, UK. E-mail: [email protected]

2011 ⁄ 1795: received 19 October 2011, revised 30 January 2012 and accepted 17 February 2012 doi:10.1111/j.1365-2672.2012.05269.x

Abstract Aims: To improve our understanding of the survival and splash-mediated transfer of zoonotic agents and faecal indicator bacteria introduced into soils used for crop cultivation via contaminated irrigation waters. Methods and Results: Zoonotic agents and an Escherichia coli marker bacterium were inoculated into borehole water, which was applied to two different soil types in early-, mid- and late summer. Decline of the zoonotic agents was influenced by soil type. Marker bacteria applied to columns of two soil types in irrigation water did not concentrate at the surface of the soils. Decline of zoonotic agents at the surface was influenced by soil type and environmental conditions. Typically, declines were rapid and bacteria were not detectable after 5 weeks. Selective agar strips were used to determine that the impact of water drops 24–87 ll could splash marker bacteria from soil surfaces horizontal distances of at least 25 cm and heights of 20 cm. Conclusions: Soil splash created by rain-sized water droplets can transfer enteric bacteria from soil to ready-to-eat crops. Persistence of zoonotic agents was reduced at the hottest part of the growing season when irrigation is most likely. Significance and Impact of the Study: Soil splash can cause crop contamination. We report the penetration depths and seasonally influenced declines of bacteria applied in irrigation water into two soil types.

Introduction In recent years, there have been a number of high profile outbreaks of human illness associated with the consumption of fresh fruits and vegetables (Ilic et al. 2008; Pezzoli et al. 2008; Soderstrom et al. 2008; Elviss et al. 2009; Muller et al. 2009). As fresh produce can be consumed without routine cooking that completely kills most human pathogenic micro-organisms, it is especially important to prevent contamination of these foods during production. In particular, special consideration needs to be given when irrigation is used to grow fresh, ready-to-eat foods such as lettuce, radish and strawberries (Brackett 1999). In the UK generally, irrigation water is not treated before

it is applied to crops and water for irrigation is commonly abstracted from rivers and other surface waters (Tyrrel et al. 2006). Potentially, human pathogenic microorganisms such as verotoxin-producing Escherichia coli (VTEC) and Cryptosporidium parvum can be washed into surface waters by rainfall (Collins et al. 2005) or more simply by direct fouling with animal manure. Not surprisingly, zoonotic agents have caused disease outbreaks that have been traced back to contaminated water in the UK (Smith et al. 2006). Human pathogenic micro-organisms have been shown to survive for extended periods of several months in river and storage reservoir waters (Tyrrel et al. 2006; Manning 2008; Jawahar and Ringler 2009). Thus, when ready-to-eat crops are irrigated, any human

ª 2012 The Authors Journal of Applied Microbiology 112, 1007–1019 ª 2012 The Society for Applied Microbiology

1007

Soil distribution and splash of bacteria

J.M. Monaghan and M.L. Hutchison

pathogenic micro-organisms present in irrigation water have a potential to cause a food poisoning outbreak (Soderstrom et al. 2008). However, despite these widely reported risks, there are no universally accepted guidelines and microbiological test criteria for irrigation water used for ready-to-eat crops (Tyrrel et al. 2006). To produce to the high-quality standards demanded by consumers for leafy vegetable crops such as lettuce, growers of these plants are increasingly looking to have irrigation water available during the entire temperate growing season (Knox and Weatherhead 2005). As a consequence, an increasing proportion of salad and field vegetable crops grown in the UK are being irrigated during their production (Knox et al. 2010). Between 1982 and 2005, the irrigated area of field vegetables in England and Wales increased by 210% and the volume of water applied increased by 360% (Knox et al. 2010). Nearly threequarters of vegetable crops were irrigated using overhead methods, the remainder receiving drip-irrigation (Knox and Weatherhead 2005). Within field vegetables, the most commonly irrigated crops include those previously implicated as vectors of food-borne illness such as lettuce (Soderstrom et al. 2008) and baby leaf salads (Brackett 1999; Knox et al. 2010). Given there is increasing use of untreated water for the irrigation of fresh produce, it is important that we have an understanding of the key behaviours of waterborne human pathogenic micro-organisms. There were a sufficient number of publications describing bacterial survival on crop phylloplanes to justify a review (Heaton and Jones 2008). However, there are very few publications that describe water-splash-mediated human pathogen transfer from soils. A single, seminal publication used strategically placed agar dishes to show that splash-based transfer can occur for pathogen surrogates (Girardin et al. 2005). However, there are still a number of gaps in our fundamental understanding of water-mediated bacterial transfer from soils and the fate of micro-organisms present in water applied to soils used to produce crops. It is apparent from visible inspection of soil residue on crops such as lettuce that rain and irrigation splash can contaminate fresh produce with soil (Hutchison et al. 2008). However, information of the type required to model the importance of soil splash for the transfer of human pathogenic micro-organisms from the soil onto crops has not yet been reported. Assuming that soil splash redistributes surface soil, the factors that may influence the extent of risk are (i) the location of bacteria in the soil following contamination, (ii) the subsequent persistence of bacteria at the soil surface and (iii) the distance that bacteria can be distributed by soil splash. Consequently, the aim of this work is to improve our understanding of the initial soil distribution, subsequent 1008

survival and splash-mediated transfer of zoonotic agents and faecal indicator bacteria introduced into two soils via contaminated irrigation water. A series of experiments using two soils were completed to study (i) the initial distribution of marker bacteria applied to cultivated soil, (ii) the persistence of zoonotic agents in the surface layer of soil and (iii) the basic dynamics of bacterial transfer from the soil surface by water splash caused by the impact of rain- or irrigation-sized water drops. Materials and methods Characterization of soils Two soil types sourced from Harper Adams University College farm were used for the soil splash experiments. A representative sample of each soil type was sent to an external testing laboratory (Eurofins, Wolverhampton, UK) for physical and chemical characterization as has been previously described (Hutchison et al. 2004). Marker micro-organism and culture conditions A natural methionine auxotrophic strain of E. coli (isolate EQ1) that was resistant to nalidixic acid was used as a benign marker for enteric human pathogens. The strain has been used previously to determine the spread of enteric human pathogens in commercial food processing premises (Buncic et al. 2002; Collis et al. 2004). The EQ1 strain was stored on Protect beads (Technical Services Consultants, Heywood, UK) at )70C. The bacteria were resuscitated by removing one bead and inoculated by streaking on a Columbia Blood Agar (Oxoid, Basingstoke, UK) plate to obtain isolated colonies. Luria–Bertani broth (30 ml) supplemented with 200 lg ml)1 nalidixic acid was equilibrated to 37C, then inoculated with one E. coli colony of and incubated at 37C overnight without shaking. The strain was used experimentally at a concentration of 1 · 105 CFU ml)1 (soil distribution experiments) or 1 · 106 CFU ml)1 (rain splash experiments) with the bacterial numbers estimated spectrophotometrically immediately before use. Zoonotic agents and culture conditions The zoonotic agents used for these studies were originally isolated from cattle manures and have been previously used for similar studies (Hutchison et al. 2004). The bacteria used were Salmonella enterica serotype Typhimurium DT104 (strain S8118 ⁄ 99), Campylobacter jejuni (strain 20001424) and a nonverotoxin-producing E. coli O157 (strain 20001383). E. coli O157 (E) and Salm. typhimurium (S) were propagated in Luria–Bertani broth



(LB; Oxoid), and Camp. jejuni (C) was grown in modified Park Sanders broth (PSB; Oxoid) supplemented with 2% (v ⁄ v) water-lysed fresh human blood. No media supplements inhibitory to other bacteria were used for PSB. All culture broths were supplemented with 3% (w ⁄ v) NH4Cl and 1% NaCl. Cultures were grown without agitation at 37C (E, S and C) or 25C (L). Headspace in Campylobacter culture vessels was filled with 8% (v ⁄ v) carbon dioxide, 7% (v ⁄ v) oxygen and 85% (v ⁄ v) nitrogen (British Oxygen Company, Guilford, UK). After incubation, broth cultures were transported to the field site during which time (c. 2 h) they cooled slowly to reach ambient temperature. Distribution of bacteria applied to soil in irrigation water A series of free-draining plastic tubs (60 cm diameter, 50 cm depth) were filled with one of the soil types and a series of sampling holes drilled at 10, 20 and 30 cm down the walls of the tub. The sampling holes were covered in duct tape and the tubs lowered back into the ground where they were left undisturbed and weed free for 2Æ5 years to allow the soil structure to reform. The tubs were removed from the ground when the soils were saturated following heavy rain, transferred into an unheated, windowless building and left to stand on a free-draining surface for 48 h to allow soils to reach field capacity. Prior to application of inoculum, the soil surface was broken up using a hoe to a depth of c. 10– 15 cm to prevent soil capping and mimic commercial production practices. A volume of borehole water (1 l) was inoculated to a final concentration 1 · 105 CFU ml)1 with E. coli marker, chosen to be broadly comparable with the concentration of E. coli used for the decline studies. The contaminated borehole water was slowly emptied into the soil at the centre of the tub. Particular care was taken to apply the water to a central area, approximately 30 cm diameter, ensuring that the water did not drain down the inner side of the tub wall. After 1 h, samples were collected from the surface of the soil and, horizontally, using the predrilled holes, from depths of 10, 20 and 30 cm using a sterile soil auger. The samples closest to the soil surface were taken first. For the samples taken below the depth of the soil surface, a soil auger was pushed horizontally into the soil columns so that it passed through the centre of the column. The auger penetrated approximately 5 cm beyond the centre before being withdrawn. The soil sample was taken from the 10 cm segment of soil at the end of the corer, which corresponded to a central zone approximately 5 cm either side of the centre of the column at each depth. The soil samples were tested to determine E. coli num-


bers as follows. In brief, 10 g of soil were added to a similar volume of maximum recovery diluent (MRD; Oxoid), vortexed to homogenize and a series of decimal dilutions prepared using MRD. Dilutions were plated in duplicate onto MacConkey No. 3 agar supplemented with 200 lg ml)1 nalidixic acid and incubated at 37C ± 1C for 40 h. Bacterial numbers on decimally diluted plates were calculated according to criteria defined by ISO 7218 (ISO 7218 2007). Decline of zoonotic agents in soils Cultures of zoonotic agents were introduced into irrigation water sourced from the untreated Harper Adams farm deep borehole. A water sample was tested to exclude the presence of the zoonotic agents used in the study. Each of the bacteria were introduced into the borehole water at concentrations of either 1 · 105 CFU ml)1 (high application) or 1 · 102 CFU ml)1 (low application). Bacterial concentrations were chosen to be low dilutions of the highest and average concentrations of zoonotic agents found in a survey of UK livestock manures (Hutchison et al. 2004). Bacteria were distributed through the water by gentle agitation taking care not to excessively oxygenate the liquid. Negative control plots were watered with unchlorinated borehole water pretested to ensure it did not contain any zoonotic agents. The plots were contained within a bio-secure electrically fenced area (Hutchison et al. 2008) used previously for similar studies. The mass of water used to irrigate each 0.2 m2 field plot was 2l. The contaminated water was applied as a single application at the beginning of each experiment. Three replicate field plots were generated for each treatment and control. Declines in the numbers of each of the zoonotic agents were followed by collecting weekly samples over 6-week period. No crops were grown, and samples were taken only from the soils. Three sets of 6-week trials (a typical growth period for a batch of lettuce) were run between May and October (the typical for the lettuce growing season in the UK). Weekly soil collections (250–300 g) were generated from a minimum of 20 combined subsamples collected to a depth of 5 cm using sterile soil augers. Samples were packed on ice and shipped from the field site to the testing laboratory where analyses were commenced within 4 h of sample collection. Escherichia coli O157, Salmonella and Campylobacter were enumerated from the soil samples using a general filter-based resuscitation method, followed by culture using different selective media as appropriate for each organism. The methods and the media used have been described in some detail previously (Hutchison et al. 2004).


1009



Water drop splash-mediated bacterial transfer from soil A modified protocol of the one described originally by Jenkinson and Parry (1994) was used to determine effect of droplet size on splash-mediated bacterial transfer from soil. Separately, the two soils were layered to a depth of 4 cm inside seed trays (30 · 30 cm). The soil surfaces were inoculated by evenly misting with 5 ml of E. coli K12 marker using a trigger sprayer. Perspex strips (45 · 2Æ5 cm) were overlaid with agar (MacConkey No. 3, Oxoid) supplemented with 200 lg ml)1 nalidixic acid, and excessive surface moisture was removed from the agar by drying (42C, 4 h) before use. Glass pipettes were drawn out in a Bunsen burner to provide a series of three different-sized nozzles. The drop volumes generated by each nozzle were determined by weighing 50 water drops from each nozzle. Assuming the drops to be spherical, the mean drop volume and mean drop diameter were calculated. A few drops of sterilized deionized water were dripped through each of the nozzles to determine where the drops landed on the floor. The soil trays were then positioned over the landing spot, and three Perspex strips were clamped to extend in the X-, Y- and Z-axes commencing 1cm from the landing spot. Two ml of sterile deionized water were dripped through each of the nozzles from a height of 6 m to allow drops to reach terminal velocity. The splash-inoculated Perspex strips were incubated at 37C ± 1C for 40 h with periodic removal of any evaporated water from the incubator internal surfaces. Each determination was replicated three times on different days. These studies were undertaken indoors in air which was as still as possible. Estimation of the areas colonized by marker The media-coated Perspex strips were photographed with a digital camera after incubation (Nikon D50; Nikon, Kingston upon Thames, UK) using the camera’s highest resolution (3008 · 2000 pixels). The photographs were imported into Adobe Photoshop (Adobe Systems, Stockley Park, UK) and electronically sectioned into 2Æ5 · 2Æ5 cm squares for processing. The Photoshop magic wand tool, which allows areas of image of similar colour to be selected, was used to select regions of the picture with bacterial growth. The wand tolerance values allowed for use by the programme were between 0 and 255. By trial and error, a value of 43 was found to select contiguous areas of media covered with bacteria and exclude uncolonized areas. When required, further areas of bacterial growth were shift-clicked until the entire growth area was selected. The Photoshop histogram function was used to determine the total number of pixels that were selected. The total area of the media not colo1010

nized by the bacteria was determined using similar approach and a wand tolerance of 10, again determined as optimal by trial and error. The percentage of each media square that was colonized was calculated. Data were analysed using anova (SPSS 11.5, SPSS Inc., Chicago, IL, USA) to determine whether there were significant (P < 0Æ05) differences between different drop volumes, dimension of splash and soil types. Environmental measurements A portable weather station (Mini-Met; Skye Instruments, Llandrindod Wells, UK) was installed in the buffer strip surrounding the field plots. Precipitation was continuously collected in a rain gauge (Skye Instruments). All other parameters were recorded every 10 min. Soil temperature was recorded at a depth of 5 cm. Air temperature and air humidity at 20 cm above ground, daylight hours and the intensity of solar radiation were recorded for the duration of the experiments at 35 cm above ground. All data were saved on a DataHog 2 device (Skye Instruments), which was downloaded weekly. Statistical analyses Log averages and associated standard deviations (SD) from each set of three replicates were calculated for each sample time using Excel 2010 (Microsoft, Redmond, WA, USA). Statistical comparison was undertaken using Mann–Whitney U-tests for nonparametric data and t-tests, homoscedastic t-tests and paired t-tests for normally distributed data as appropriate (SPSS 11.5; SPSS, Inc.). Decimal reduction times (DRT; the number of days required for a 1-log decline in bacterial numbers) were calculated as the reciprocals of initial gradients from plots of bacterial numbers over time and scaled to represent a single log reduction. R2 values were determined by the least-squares method. Groups of DRT were compared using one-way analysis of variance (anova) with Tukey’s post hoc test (SPSS). Results The physicochemical characteristics of Soil A and Soil B used for this study are shown in Table 1. Soil A was classified as Sandy Loam, and Soil B was classified as Loamy Sand (Brady and Weil 2007). Soil A had a volumetric water content of 29% when at field capacity and a Saturated Hydraulic Conductivity of 48Æ5 mm h)1; Soil B had a volumetric water content of 11% when at field capacity and a Saturated Hydraulic Conductivity of 85Æ5 mm h)1 (Brady and Weil 2007). Both were typical of the soil types used to grow fresh produce crops on a commercial scale in the region.




Table 1 Physicochemical profile of the two soil types used for these studies

Analysis pH Organic matter (%) Total nitrogen (%) Extractable phosphorus (mg l)1) Extractable potassium (mg l)1) Extractable magnesium (mg l)1) Particle size distribution 2000–600 lm – coarse sand 600–212 lm – med sand 212–63 lm – fine sand 63–20 lm – coarse silt 20–2 lm – fine silt 6 (ND)

Value in brackets = SE (n = 3); ND, not determined.

persistence of E. coli O157 was observed in late summer with complete decline below the test limits of detection after 3Æ3 and 1Æ3 weeks in soil A at the high and low inoculations, respectively (Table 4). For the lower inoculation into soil B, slower initial decline and increased persistence compared with the mid-season was also observed. As was observed for Salm. Enteriditis, E. coli O157 persisted longest in soil B at the higher inoculation. At the end of the late summer experiment, E. coli O157 was still recovered from the soil at a mean log of 1 log CFU g)1 (Table 4). No Camp. jejuni was recovered from the soil in the early summer experiment although phase-contrast microscopic inspection of the culture prior to application revealed viable cells with characteristic energetic corkscrew motility. In midsummer and for both soil types,

initial declines of Camp. jejuni were of the order of 0Æ5 logs per day (Table 3) and, consequently, the numbers bacteria declined to below quantifiable numbers in 10 cm from the drop impact, bacterial growth was smallest on the Z-axis (Fig. 4). Droplet size had significant effect from 7Æ5 to 30Æ0 cm from the point of impact with significantly (significance ranging from P = 0Æ03 to P < 0Æ001) greater bacterial growth from the large droplet size compared with the small and medium droplets 1014




Table 5 Area of colonization along the strips, averaged across soil types and dimensions. Significance shows the level of significance of the effect of droplet size and dimension of splash obtained from the ANOVA. There was no significance of soil or interactions between soil, droplet size or dimension of spread at any location Distance from origin (cm)

0Æ0

2Æ5

5Æ0

7Æ5

10Æ0

12Æ5

15Æ0

17Æ5

20Æ0

22Æ5

Area of colonization (%) Significance Droplet size Dimension

16Æ6

17Æ3

17Æ3

17Æ6

16Æ0

10Æ5

8Æ3

6Æ9

4Æ4

2Æ7

0Æ09