Files
Abstract
Agricultural water plays an important role in the production of produce, which is of increasing focus as a vehicle for foodborne illnesses including Shiga toxin-producing Escherichia coli (STEC) and Salmonella. Understanding the risk posed by surface water sources is important for preharvest food safety. Quantification of E. coli has been used to indicate if fecal-oral pathogens are likely to be present, but it is imperfect. This study began by examining how levels of E. coli differ spatially across three ponds using mean relative difference (MRD) to show the average difference in E. coli level at a location from the overall. The log MPN E. coli 100 mL-1 (EC MRD) values ranged from -0.25 to 0.33 in Pond 1, -1.5 to 0.65 in Pond 2, and -1.25 to 0.65 in Pond 3. EC MRD correlated positively with MRDs for chlorophyll, and turbidity, and negatively with dissolved organic matter, dissolved oxygen (DO), specific conductance and pH at Pond 1, positively with chlorophyll, DO, phycocyanin, pH, and temperature at Pond 2, and positively with nitrate at Pond 3. Explanatory models using regression and machine learning methods were created for E. coli levels, and the presence of pathogenic E. coli and Salmonella. Random Forest (RF) models performed best for E. coli levels (Root Mean Square Error (RMSE)=0.45) in the overall dataset), Pond 1 (RMSE=0.44), Pond 2 (RMSE=0.71), while generalized regression performed best at Pond 3 (RMSE=0.41, n=198). Pathogenic E. coli presence for all ponds (n=717) was best explained by RF (Receiver Operating Characteristic Area Under the Curve (ROC AUC)=0.94). Salmonella presence for ponds 1 and 2 (n = 284) was best explained by multinomial regression (ROC AUC=0.83). Finally whole-genome sequencing was used to characterize E. coli and Salmonella isolates to examine their possible sources and changes over time, and the presence of virulence and antimicrobial resistance genes among them. Phylogroups of E. coli identified were diverse and showed seasonality. Phylogroup B1 was dominant among the samples, indicating wildlife and domestic animals were the most likely sources. Antimicrobial resistance and pathogenicity genes were common in the E. coli isolates but limited among Salmonella.