Water samples were collected from irrigation ponds and postharvest packing house dump tanks of two produce farms in southeastern Georgia between September 2020 and December 2021. Irrigation water sampling sites consisted of four surface-fed holding ponds used for crop irrigation using drip systems on each grower’s farm. Pond water samples were collected once a month during fallow periods and twice per month during harvest periods. In 2020, pond water samples were collected once a month during growing periods, but were collected twice a month in 2021. During harvest periods, spent water from one dump tank from each grower’s packing house was collected weekly. A total of 217 pond water samples and 46 dump tank water samples were collected during the study period.

Pond water samples were collected on – site by filtering 50 L of water using dead-end ultrafiltration (DEUF) (Kahler et al., 2021; Mull & Hill, 2012; Smith & Hill, 2009). Twenty liters of dump tank water was collected in 20-L cubitainers. Harvested produce was submerged in municipal water containing chlorine within the dump tank flume to cool the produce and remove debris. Dump tank water was sampled during produce washing before a water change to collect microorganisms that had been rinsed from the produce into the water. A 250-mL grab sample was collected alongside both pond and dump tank large-volume samples for quantification of total coliforms and E. coli. At the time of sampling, water temperature, pH, dissolved oxygen, turbidity, and conductivity were measured using a YSI Inc. ProDSS Multiparameter Digital Water Quality Meter (Yellow Springs, OH). Each grower used municipal tap water for dump tanks and an automated fertilizer dosing system for chlorinating the water during processing. Oxidation-reduction potential (ORP) was measured using the YSI meter to assess the oxidizing capacity of the dump tank water at the time of sample collection. Four replicate measurements were taken for each parameter to ensure accurate measurement in heterogeneous water matrices. Water samples were stored in insulated coolers with ice packs. Sodium thiosulfate was added to the dump tank (5 g) and grab water (2.5 g) samples immediately after collection to neutralize any chlorine residual. Grab water samples were analyzed for total coliforms and E. coli on the day of collection at the University of Georgia (2021) or Auburn University (2022) laboratories. Ultrafilters and dump tank samples were shipped overnight to the Centers for Disease Control and Prevention (CDC) laboratories in Atlanta, GA, for further processing.

At CDC, ultrafilters were backflushed by pumping 500 mL of a 0.5% Tween 80/0.01% sodium polyphosphate/0.001% Antifoam Y-30 solution through the dialysate port and recovering captured microbes from the blood port (Kahler et al., 2021; Smith & Hill, 2009). Due to their high organic load, dump tank water samples were concentrated by continuous flow centrifugation (CFC) as previously described (Kahler et al., 2021). Briefly, samples were pumped through a CFC Express unit at 750 mL/min using CFC 210 disposable bowls (Scientific Methods Inc., Granger, IN). The resulting concentrates were decanted, and the bowls were rinsed two times with 5 mL of a 0.5% Tween 80/0.01% sodium polyphosphate/0.001% Antifoam Y-30 solution. Samples were further concentrated by centrifugation at 4,000g for 15 min, the supernatants were removed by pipetting, and the pellets were resuspended with PBS. The resuspended concentrates were stored at −20 °C until nucleic acid extraction.

Municipal wastewater biosolids (sludge) were collected twice a month from the local wastewater treatment plant in Tifton, Georgia, from January 2020 through December 2021. This facility serves the county in which the two study farms were located and is located 10–25 miles from the irrigation ponds, depending on the pond. One liter grab samples were collected from two locations in the processing system, the sludge from the gravitational thickener (REC) used for land application and the return-activated sludge (RAS) from the aeration basin. Only the RAS sample was collected from November 2020 until July 2021 while routine maintenance was performed on the thickener system. Forty-six RAS sludge and 30 REC sludge samples were collected in total. Sludge samples were placed in an insulated cooler with cold reusable freezer packs and shipped overnight to CDC in Atlanta, GA, for further processing. Samples were frozen at −20 °C upon arrival until further processing. Thawed sludge samples were heat inactivated in a 60 °C water bath for 1 h. After cooling, approximately 500 mL was concentrated by centrifugation at 4,000g for 15 min, and the supernatants were removed by pipetting. The sludge concentrates were stored at −20 °C until nucleic acid extraction.

Portable toilet samples were collected every week during harvest seasons, as available, for a total of 37 samples. Samples were collected by inserting a SteriWare ViscoThief cream sampler into the toilet seat opening (Sampling Systems, Ltd., UK). Approximately 200 mL of portable toilet material was drawn into the sampler and then transferred into a sterile 250-mL Nalgene bottle. Samples were placed in an insulated cooler with cold reusable freezer packs and shipped overnight to CDC for further processing. At CDC, samples were concentrated by centrifugation at 4,000g for 15 min. The supernatant was removed by pipetting, and the sample concentrate was stored at 4 °C until further processing. Sample concentrates that exceeded a volume of 10 mL were placed in a filtered stomacher bag. Parasites were then eluted off the concentrate in the stomacher bag by adding 100 mL of 0.1% Alconox and stomaching at 230 rpm for 1 min using a Stomacher^® 400 Circulator Lab Blender (Seward Ltd., UK). The eluate was removed from the bag, and the stomacher bag was rinsed manually with an additional 100 mL of 0.1% Alconox. The sample eluates were combined and further concentrated by centrifugation at 4,000g for 15 min. The supernatant was removed, and the sample concentrates were stored at −20 °C until nucleic acid extraction.

Nucleic acid from pond and dump tank water concentrates was extracted using the Qiagen AllPrep^® PowerViral^® DNA/RNA Kit (Germantown, MD) with the following modifications. Dithiothreitol (Thermo Fisher Scientific, Waltham, MA) at a concentration of 0.01 μM was used in place of β-mercaptoethanol in the initial lysing buffer, which acts to prevent RNase activity for total nucleic acid extractions. Bead beating was performed on a FastPrep-24^™ Classic bead beating grinder and lysis system (MP Biomedicals, Santa Ana, CaA) at 6 m/s for 60 s. The final elution was completed with 100 μL of Tris-EDTA (TE) buffer. Nucleic acid was extracted from the municipal wastewater sludge concentrates using the Qiagen AllPrep^® PowerViral^® DNA/RNA Kit following the same procedure as water samples. DNA was extracted from the frozen portable toilet sample concentrates using the Qiagen DNeasy^® PowerMax^® Soil Kit and the extracts were further concentrated by ethanol precipitation and resuspension in 0.1 mL TE buffer as described in the PowerMax^® Soil Kit Handbook. An extraction blank was included with each batch of extracted samples (N = 11). Nucleic acid that was not immediately subjected to molecular testing was stored at −20 °C.

Quantitative real-time PCR (qPCR) was performed for the detection of a C. cayetanensis 18S rRNA gene target for all samples (Qvarnstrom et al., 2018). Samples were analyzed in triplicate using a 5 μL template in 50-μL reaction volumes. Bovine serum albumin (BSA, Millipore Sigma, Burlington, MA) and T4 Gene 32 protein (gp32, New England Biolabs, Ipswich, MA) were added to the reaction mixture at concentrations of 0.5 mg/mL and 25 μg/mL, respectively, to improve amplification efficiency and reduce inhibition. Primer and probe concentrations were 0.5 μM and 0.1 μM, respectively. An internal amplification control (IAC) consisting of a synthetic 200-bp ultramer DNA sequence was included in the C. cayetanensis assay at a concentration of 500 copies per reaction (U.S. Food and Drug Administration, 2017). IAC primer and probe concentrations were 0.5 μM and 0.1 μM, respectively.

qPCR was performed for the detection of Bacteroides HF183 (HF183) and crAssphage (056 assay) in pond and dump tank water samples (Stachler et al., 2017; U.S. Environmental Protection Agency, 2019). Samples were analyzed in triplicate using a 2 μL template in 25-μL reaction volumes. BSA was added to the reaction mixtures at a concentration of 0.2 mg/mL. Primer and probe concentrations for the assays were 1 μM and 0.08 μM, respectively. An internal amplification control (IAC) consisting of a linearized synthetic plasmid was included in the HF183 assay at a concentration of 100 copies per reaction to evaluate assay interference. IAC primer and probe concentrations were 1 μM and 0.08 μM, respectively.

qPCR amplification was performed in a 7500 Real-Time PCR System with software version 2.3 using TaqMan^™ Environmental Master Mix 2.0 (Applied Biosystems, Waltham, MA). Forty-five cycles of annealing and extension were performed for all assays to visualize amplification curves for samples with late-stage amplification (C_q 35–40). For C. cayetanensis, the annealing and extension temperature was 67 °C, and for HF183 and crAssphage, the annealing and extension temperature was 60 °C. Three no-template controls with nuclease-free water as template were included with each instrument run.

All gene targets were enumerated using a standard curve consisting of at least six, 10-fold serial dilutions of a synthetic standard (Integrated DNA Technologies, Coralville, IA) analyzed in triplicate during each instrument run (Table S1). Gene copy number per reaction was calculated by inputting the average quantification cycle (C_q) of the detected replicates into a pooled standard curve equation. This value was multiplied by the proportion of the original sample that was concentrated, extracted, and tested by qPCR to obtain log₁₀ gene copies detected per 100 mL of the original sample. A sample was considered to have a detection for HF183 or crAssphage if amplification was observed for at least 2 of 3 replicates, each having a C_q value less than 40.

Three definitions were used for determining qPCR detection of the C. cayetanensis 18S rRNA gene target in a sample. For Definition 1, the target was considered detected in a sample if amplification was observed for at least 2 of 3 qPCR replicates each having a C_q value less than 40 (Mattioli et al., 2021; Rao et al., 2022). Definition 2 corresponds to the criteria outlined in FDA’s Bacteriological Analytical Manual for the detection of C. cayetanensis in agricultural water, in which the target was considered detected in a sample if amplification was observed for at least one qPCR replicate having a C_q value less than 40 (U.S. Food and Drug Administration, 2020a). Definition 3 is a proposed approach for standardization of detection limits for environmental qPCR assays, in which the detection limit for an assay is the lowest concentration of a standard that produces amplification in at least 95% of qPCR replicates (Klymus et al., 2020). For the C. cayetanensis 18S rRNA gene target assay, the qPCR detection limit was 35 gene target copies per reaction (C_q ≤ 33).

A workflow algorithm was used to confirm qPCR detections of the C. cayetanensis 18S rRNA gene target in irrigation water or sludge (Fig. 1). Samples that met the criteria for sequence characterization (C_q < 37) were subjected to testing using a C. cayetanensis PCR genotyping panel, consisting of six assays targeting regions from the C. cayetanensis nuclear genome and two assays targeting regions from the mitochondrial genome (Nascimento et al., 2020). Amplicons from all eight assays were deep sequenced on the Illumina MiSeq (Illumina, San Diego, CA) to identify haplotypes. The haplotypes were compared against known clinical C. cayetanensis haplotype sequences in a reference database compiled from U.S. Cyclospora outbreak strains, and a sample with amplified haplotypes matching C. cayetanensis haplotypes was considered a confirmed detection of C. cayetanensis (Barratt et al., 2021). For samples with confirmed C. cayetanensis haplotypes, genotyping could be attempted for samples with either (a) amplification of at least 5 of the eight markers or (b) amplification of 4 markers, if at least 3 of those 4 markers were any combination of the following: MSR, Mt-Junction, HC360i2, and HC378 (Barratt et al., 2021). A pairwise distance matrix was generated for samples in which these criteria were met, and samples were clustered based on genetic relatedness as previously described (Barratt et al., 2021).

A sample with haplotypes possessing low homology to known C. cayetanensis haplotypes was considered a false detection caused by suspected cross-reaction in the qPCR assay. Amplicon sequences were BLAST searched against the NCBI nucleotide database to identify closely related organisms (percent identity > 95%) that may have been present in the sample, leading to cross-reaction with the qPCR assay (Camacho et al., 2009). A phylogenetic tree was created that included a subset of samples in which the mitochondrial MSR marker amplified (Barratt et al., 2021; Nascimento et al., 2020). For comparison, mitochondrial DNA sequences homologous to the MSR locus were obtained from GenBank for several other coccidian parasites; these sequences were also included in our phylogenetic analysis. Sequences were aligned using MUSCLE v3.8.425 in Geneious Prime (version 2022.1.1) with default parameters (https://www.geneious.com) (Edgar, 2004; Kearse et al., 2012). The resulting alignment was analyzed using the phangorn package in R (version 4.0.4) to generate a neighbor-joining tree using the Jukes-Cantor substitution model, with 1,000 bootstraps (Saitou & Nei, 1987; Schliep, 2011).

Samples that did not produce any PCR amplification by the typing assays were considered unconfirmed detections, meaning the qPCR detection may have been the result of a cross-reaction or detection of C. cayetanensis genomic material that was below the detection limit of the typing assay. Finally, samples that were not able to be further characterized due to high qPCR C_q values (C_q ≥ 37) were also considered unconfirmed detections.

Total coliforms and E. coli were enumerated from 100 mL of the grab samples within 6 h of sample collection by the IDEXX Quanti-Tray 2000^® method using Colilert-18 media (IDEXX, Westbrook, ME) (American Public Health Association, 2017).

Statistical analyses were performed in R version 4.2.2 (R Core Team, 2017). Figures were created using the ggplot2 package (Wickham, 2016). Cochran-Mantel-Haenszel Chi-Squared test was used to evaluate the association between qPCR detections of the C. cayetanensis 18S rRNA gene target and the human fecal indicator HF183 while controlling for pond using the mantelhaen.test function (R Core Team, 2017). P values ≤0.05 were considered statistically significant.

Measured water quality parameters from the study ponds and dump tanks are presented in Table S2. Total coliforms and E. coli were routinely cultured in the pond water samples throughout the study period (100%; 85%, N = 217), with median E. coli concentrations ranging from 6 to 63 MPN/100 mL (Table 1). While total coliforms were also routinely cultured from dump tank water (84%, N = 46), E. coli was rarely detected in the samples (20%). Time series values of total coliform and E. coli concentrations are presented in the research project report (Mattioli et al., 2022). HF183 and crAssphage were detected in 71 (33%) and 14 (6%) of the pond water samples, respectively (Table 1 and Figs. S1–S8). HF183 was detected in 2 (4%) dump tank samples, and crAssphage was not detected in any dump tank sample.

The C. cayetanensis 18S rRNA gene target was detected by qPCR in 59 (27%) pond water samples using Definition 1 detection criteria, in 91 (42%) pond water samples using Definition 2, and in none of the water samples using Definition 3 (Table 2 and Figs. S1–S8). The C. cayetanensis gene target was detected by qPCR in 1 (2%) dump tank sample using Definitions 1 and 2. There was no association between incidence of detection of HF183 and incidence of C. cayetanensis qPCR detection using either Definition 1 or Definition 2 in water {Definition 1: χ²(1) = [0.006], 95% CI [0.39, 2.37], p = 0.94; Definition 2: χ²(1) = [0.04], 95% CI [0.39, 1.86], p = 0.84; Definition 3 had no C. cayetanensis detections}. The C. cayetanensis gene target was detected by qPCR in 9 (20%) RAS samples using Definition 1, in 16 (35%) samples using Definition 2, and in 2 (4%) samples using Definition 3 (Fig. 2). Finally, the C. cayetanensis gene target was detected in 9 (30%) REC samples using Definition 1, 12 (40%) samples using Definition 2, and 1 (3%) sample using Definition 3. There were no C. cayetanensis qPCR detections in portable toilet samples.

Of pond samples in which the C. cayetanensis 18S rRNA gene target was detected by qPCR, 45 (76%) samples had at least one replicate with a C_q value <37 and were further tested using the C. cayetanensis genotyping panel PCR assays (Nascimento et al., 2020). To evaluate the C_q value cut-off for sequencing confirmation, eight additional samples with qPCR C_q values ≥37 were also tested (N = 53 total water samples tested by PCR panel). The single qPCR detection in the dump tank water did not meet inclusion criteria, with all replicate C_q values >37. All sludge samples with detections by qPCR were tested by the PCR genotyping panel, including five additional RAS samples and three additional REC samples for which only one replicate amplified at a C_q value <37 (N = 26 total sludge samples tested by PCR panel).

No amplification was observed for any genotyping marker for 15 pond water samples, and only the mitochondrial 16S rRNA gene (MSR) target amplified for 38 irrigation pond water samples. The MSR haplotype sequences for these pond water samples did not match any known C. cayetanensis haplotypes but were closely related to segments of the mitochondrial target region conserved amongst other Apicomplexan protozoa. MSR amplicon sequences were analyzed on a neighbor-joining tree, demonstrating that the sequences from the irrigation pond samples belonged to coccidia outside the clades occupied by clinically isolated C. cayetanensis sequences (Fig. S9). BLASTN searches revealed that the MSR amplicon sequences did not match any known organism, but mitochondrial genome sequences from Eimeria, Isospora, and Caryospora were the closest matches. Therefore, these 38 samples were determined to be false detections caused by suspected cross-reactions from closely related coccidian parasites in the irrigation pond water. Of the remaining samples, 15 were considered unconfirmed detections for C. cayetanensis due to the lack of amplification of any of the genotyping markers and 12 were considered unconfirmed detections as all three qPCR replicates had C_q ≥ 37.

Of the 26 sludge samples tested by the C. cayetanensis genotyping panel, no amplification was observed for 18 samples. One sample amplified five of the 8 typing markers and matched C. cayetanensis haplotypes from clinical specimens from 2018 to 2021; the MSR locus obtained from this sample is included in our phylogeny (Table S3 and Fig. S9). Six additional samples produced amplification in one to three of the typing panel targets, and all the amplified haplotypes matched C. cayetanensis haplotypes from clinical specimens. However, the number of markers amplified from each of these six samples was too few to meet the criteria for clustering via the previously described genotyping procedure (Nascimento et al., 2020). One sample had only one mitochondrial target amplify, and the haplotype sequence did not match any C. cayetanensis haplotypes but did match a conserved segment of the Apicomplexan mitochondrial target and thus was considered a suspected cross-reaction.