Data from: Assessing foodborne pathogen survival in bird feces to co-manage farms for bird conservation, production, and food safety
Data files
Nov 13, 2024 version files 33.09 KB
-
README.md
1.06 KB
-
Spence_Dryad_Submission.zip
32.03 KB
Abstract
Because birds can carry foodborne pathogens, widespread concern exists that birds impose food-safety risks to farms. Growers are thus often encouraged to deter birds and forego harvesting crops near bird feces (e.g., no-harvest buffers). Developing a holistic understanding of the pathogen spillover process– from individual infection to pathogen persistence on crops– is essential to understand and manage food-safety risks associated with birds. Here, we coupled field and greenhouse experiments to understand the relative risks associated with different bird species on California farms. We first compared E. coli survival in bird feces on lettuce, soil, and plastic mulch to identify where pathogens are likely to persist. Next, we quantified pathogen survival in feces from 10 bird species to identify higher-risk species. Finally, we combined pathogen survival data with bird surveys and fecal transects to assess food-safety risks across 29 California lettuce farms. E. coli abundance rapidly declined in bird feces, but rates varied across substrates. E. coli survival was higher on lettuce compared to soil and plastic mulch, likely because of lettuce’s cooler and wetter microclimate. E. coli persistence was also much higher in feces from larger birds—which produce larger feces—than small songbirds. Importantly, minimal differences in persistence among species were observed after all feces were standardized to the same mass. Though bird feces were common on farms, most birds entering farms, contacting crops, and defecating on crops were small songbirds that defecate small feces. Coupled with our finding that ~90% of feces were deposited on soil, these results suggest that most birds on farms present relatively low food-safety risks. Synthesis and applications. Growers are often encouraged to deter all bird species and forego harvesting crops near all bird feces, but our work suggests not all birds and feces pose the same food-safety risks. If growers ignored small bird feces on soil, then we estimate that the area of California lettuce farms affected by no-harvest buffers could decrease from ~10.3% to ~2.7%. More broadly, our results suggest farmers could promote small, insect-eating birds by erecting nest-boxes or preserving habitat without necessarily compromising food safety.
https://doi.org/10.5061/dryad.nzs7h450x
Description of the data and file structure
All data for associated publication are included. Specific data include (1) experimental data (experiment_field.csv) from the field experiment; (2) experimental data (experimentlab.csv) from the laboratory/greenhouse experiment; (3) raw point count (pointcount.csv) data, including birds observed and associated information; (4) bird occurrence found from sequencing analysis (fecal_sequencing.csv); and (5) fecal mass (fecal_mass.csv) and associated bird mass data.
Descriptions of the procedures involved are included in the methods description and the resulting manuscript.
Files and variables
File: Spence_Dryad_Submission.zip
Description: All data are stored in CVS files. Each file has an associated README csv, that provides a verbal description of each column.
Missing data: NA
Our methods include two distinct field operations: first, we collected fresh fecal materials directly from birds in or near Davis, California for our experiments. Second, we collected point count and fecal density data directly on leafy-green farms in the Central Coast of California. The following methods and portions of the results have been included in a grant report for The Center for Produce Safety (Karp & McGarvey, 2024).
2.1 Field experiment:
To quantify E. coli persistence, we implemented a field experiment on the University of California Davis Student Farm – an organic vegetable farm in Davis, California, USA. The experiment focused on two species that are frequently observed on farms but differ in size: Wild Turkey (5,791g; Meleagris gallopavo) and Western Bluebird (26g; Sialis mexicanus) (Tobias et al., 2022). Throughout the spring and summer of 2022, we collected turkey feces by following individuals on the UC Davis campus and picking up samples after defecation with sterile forceps. We obtained bluebird feces by hand-capturing adult birds from nest boxes along Putah Creek in California’s Central Valley and placing birds in sterilized cotton bags, where they often defecated. We then collected feces from bags with sterilized tweezers.
After collection, sample containers were placed on ice and combined into composite samples by species. The composite samples were inoculated with E. coli RM14721NR – a non-pathogenic, spontaneous nalidixic acid and rifampicin resistant mutant derived from E. coli RM14721 (E. coli, hereafter) – which is an isolate from a lettuce field in Yuma, AZ (Carter & Pham, 2018). This mutant strain of E. coli is not naturally occurring in avian or soil fecal samples. The inoculum was prepared by growing E. coli in 5mL of tryptic soy broth (TSB) for 18h at 37°C with shaking (200 RPM) the day prior to fecal collection. The inoculum was mixed into Wild Turkey or Western Bluebird feces at a ratio of 1:20 (v/m) to achieve a final E. coli concentration of ~ 1×107 CFU/g feces, which is within the range of E. coli concentrations found in wild bird feces (Smith 2020). We quantified the inoculation broth’s E. coli concentration by plating and counting CFUs on TSA agar plates that were incubated at 37°C for 24 h. We then calculated the concentration of E. coli within each fecal sample (fecal E. coli concentration = inoculation concentration × total fecal mass).
After inoculation, Western Bluebird aggregated fecal matter was subdivided into two size classes (0.03g and 0.06 g; N = 32 and 41, respectively), and Wild Turkey aggregated fecal matter was subdivided into four classes (0.03g, 0.06g, 2.00g, and 4.75g; N = 52, 48, 52, and 52, respectively). Masses were chosen to correspond to the average fecal mass of (1) a small songbird (e.g., Yellow-rumped Warbler, Setophaga coronata), (2) a Western Bluebird, (3) a juvenile Wild Turkey, and (4) an adult Wild Turkey, based on samples obtained from 30 warblers, 15 bluebirds, and 27 turkeys. Feces were then transported to the UC Davis Student Farm – an organic, experimental, and teaching farm— and placed in the open environment on one of three substrates within four hours of collection: lettuce (Lactuca sativa; butterhead variety), organic soil, or black plastic mulch (commonly used to manage weeds and decrease soil water loss in produce fields). Three temperature and humidity loggers were placed on the lettuce, soil, and plastic mulch (HOBO Data Logger, Onset, MA, USA) to measure abiotic conditions. We collected samples after 1–30 days (64 samples after 1–2 days; 63 after 3–5 days; 66 after 6–8 days; 32 after 14–15 days; and 52 after 21–30 days; total= 277 samples). If fecal samples had not disintegrated, we collected the sample and surrounding substrate with sterile tweezers and transported them on ice to the USDA Western Regional Research Center in Albany, California.
To quantify the amount of E. coli remaining, we combined ~10g of feces and substrate (lettuce leaves, plastic, or soil) with 90ml phosphate buffered saline (PBS) in a blender jar and blended them with an Osterizer Beehive blender on high speed for 60 seconds. The solution was serially diluted in PBS and plate counted on MacConkey agar plates containing 50mg/l nalidixic acid and 100mg/l rifampicin after incubation at 30°C for 24h. This assay only allowed growth of the selected for the nalidixic acid and rifampicin resistant specific mutant strain of E. coli added at the point of fecal inoculation, but inhibited the growth of any indigenous E. coli present in the fecal samples. Any other naturally-occurring E. coli that may have been initially present in bird feces would not be able to grow on the nalidixic acid/rifampicin media and thus was not included in final E. coli counts. We enriched samples that showed no growth by adding 1ml of the blended sample into 9ml TSB containing 50mg/l nalidixic acid and 100mg/l rifampicin. Samples were then incubated at 30˚C for 24h (shaking 200 rpm), plated onto MacConkey agar plates containing 50mg/l nalidixic acid and 100mg/l rifampicin, and incubated for 18h and examined for growth. If the sample showed E. coli growth after enrichment, it was assigned a value of 1x103 CFUs (i.e., the limit of detection). Otherwise, it was designated as E. coli negative.
2.2 Greenhouse experiments
We implemented greenhouse experiments to quantify E. coli survival in feces from different bird species. Greenhouse experiments were necessary as feces from different bird species were available in different seasons and abiotic conditions needed to be standardized as much as possible. Specifically, we used feces from 10 bird species collected from December 2022 through May 2023 (Table S1; See supplementary methods for fecal collection protocols). Species varied in sample sizes due to the unpredictable nature of sample collection. Together, our species have geographic wide ranges, represent six taxonomic orders, vary in body size and thus fecal deposit size, are found in farming environments, and exhibit a variety of migratory patterns. The average fecal mass produced naturally by each species was determined by placing feces in airtight containers (sterile vials or plastic bags) within 30 minutes of defecation and weighing feces within 5 hours of collection. Captured birds were also weighed to relate fecal mass to bird mass. For uncaptured species (e.g., Wild Turkeys), we obtained an estimate of bird mass from the literature (Tobias et al., 2022).
Samples were transported to the USDA Western Regional Research Center on ice. As above, we aggregated samples by species and then inoculated feces with E. coli, using the same methods described above. To disentangle whether fecal mass or species identity per se influenced E. coli survival, we subdivided feces into the average natural size for the species (0.03 – 9.8g) or a standardized mass (0.08g; or the average mass of a common songbird: the White-crowned Sparrow; Zonotrichia leucophrys). Within six hours of fecal collection, we placed the samples on lettuce from the same seed stock in a greenhouse, alongside a temperature/humidity logger. After three days, we collected the samples and the amount of E. coli remaining was quantified as described above. All methods were approved under University of California, Davis Institutional Animal Care and Use Committee Protocol #22562, California Permit S-212650004-21266-001, and Federal permit 24033.
2.3 Statistical analyses of pathogen survival experiments
First, we modeled temperature and humidity as a function of substrate to determine whether microclimates differed between lettuce, soil, and plastic mulch. Both variables were modeled with Gaussian distributions, including day of year as a covariate.
Next, we implemented Generalized Linear Mixed-effect Models (GLMMs) to analyze E. coli survival. For the field experiment, we modeled E. coli survival (presence or absence of E. coli; binomial distribution) and the percent of original E. coli remaining (final divided by initial E. coli concentrations; Gaussian distribution with a log-link) as response variables. Both response variables were modeled as a function of the number of days in the field, fecal mass category, substrate (lettuce, soil, or plastic), and species (Western Bluebird or Wild Turkey). We did not include temperature or humidity in the field models to avoid collinearity (as a result, all models had variance inflation factors less than 1.3). Interactions between mass and substrate as well as mass and days in the field were initially tested, but then omitted as interactions were not significant. Higher E. coli survival could reflect variation in initial E. coli abundances. We thus additionally explored including either initial E. coli concentration or total initial E. coli abundance as covariates in survival models (i.e., E. coli presence/absence models). Finally, we included ‘sampling batch’ as a random intercept.
For the greenhouse experiment, we modeled the percent of original E. coli remaining as a function of fecal mass, fecal mass squared (i.e., a quadratic term), species identity, temperature (°C), and relative humidity, using a Gaussian distribution with a log-link to meet assumptions of normality. To test for the effect of explanatory power of species identity versus fecal mass, we created two additional models: one modeling the percent of original E. coli remaining as a function of fecal mass, temperature, and relative humidity only and another as a function of species, temperature, and relative humidity. We included batch as a random intercept. We compared models using deviance information criterion (DIC) and considered models better if ∆DIC > 2.
All models were implemented in a Bayesian framework using JAGS in R (R Core Team, 2019). For each model, three chains of 50,000 iterations were thinned by 10 with a burn-in of 10,000, resulting in a posterior sample of 12,000. Parameter convergence was confirmed via visual inspection of traceplots as well as requiring a Gelman–Rubin statistic <1.1 (Gelman et al., 2013). Parameters were considered strong predictors (i.e., significant) if the 95% Bayesian Credible Intervals (BCI) did not overlap zero.
2.4 Avian point counts and fecal transects
We conducted avian censuses and fecal contamination surveys to quantify and assess bird intrusion and fecal contamination across 29 organic farms growing leafy greens. We completed the point counts in the California Central Coast as it is one of the largest leafy-green producing regions in the U.S. (Karp et al., 2015). In total, 18, 18, and 11 farms were visited three times in summer (5/26/2022-7/23/2022), three times in fall (9/14/2022-11/18/2022), and three times in winter (2/15/2023-4/20/2023), respectively. At each farm, we established point-count sites in fields growing lettuce (N= 274 total), with 3-8 sites per farm, depending on farm size (mean 5.8 sites). Within farms, point-count sites were established a median of 220m apart from one another (inter-quartile range: 120 – 447m).
The same expert observer conducted all points counts within the season. Surveys occurred near harvest (i.e., the highest risk period for fecal contamination), except in winter (when surveys occurred at the seedling stage). Surveys began at sunrise, with 1-2 farms surveyed per day. Observers waited 5 minutes before beginning the count, and then, during each 10-minute point count, recorded all birds heard or seen within the 50 m survey radius, noting whether birds were physically contacting crops.
To quantify fecal contamination, observers established three 25m transects on each farm in the same fields as the point counts. Transects were located at a field edge, in a field interior (up to 500 m into a crop field), and halfway in between. Observers visited the transects once per season (N= 141 transects). Transects were divided into 25, 1m2 adjacent quadrats, and the number and location (i.e., on crop or soil) of all encountered bird feces were noted for each quadrat. To estimate the amount of area that would fall within 1m2 no-harvest buffers, we calculated the fraction of quadrats with fecal samples. Observers also collected up to 10 feces per transect. We then used molecular methods to identify birds that defecated on lettuce (see supplementary methods). If 10 feces could not be located, then observers scoured adjacent areas until they collected 10 samples. In total, 1075 samples were collected (N= 373 from fecal transects and 702 collected from other areas), 277 of which were identified to the originating wild bird species.