Microbial and chemical risk from reclaimed water use for residential irrigation

The availability of water and its sustainable management has become a critical issue for many drought-stricken or water resource-limited communities in the United States. Diminished water supplies have resulted in an increased use of treated municipal wastewater, commonly known as reclaimed wastewater, for reuse applications. Reclaimed wastewater is a valuable resource for crop irrigation, in particular in semi-arid and arid climates not only as a water source, but also as a source of nutrients. State regulatory authorities typically specify a certain level of treatment and monitoring of reclaimed wastewater to protect public health and the environment, in a fit-for-purpose approach (Chhipi-Shrestha et al. 2017). Water quality regulations for utilizing reclaimed wastewater for irrigation typically specify acceptable log reductions of microorganisms, physicochemical parameters (e.g., turbidity, oxygen demanding material, chlorine residual), agronomic parameters (e.g., salinity, sodium adsorption ratio, pH), and the number and types of treatment technologies required for the particular reuse scenario (Shoushtarian & Negahban-Azar 2020). Yet depending on the level of treatment there may still be waterborne pathogens or pharmaceuticals and personal care products (PPCPs) present in reclaimed wastewater used for irrigation. In the fit-for-purpose treatment approach, various engineered and agricultural controls can be utilized to reduce the potential risk to users irrigating with reclaimed wastewater (Mohr et al. 2020). For example, irrigation methods can be selected to limit risk to consumers by either limiting the edible portion of the crops exposed to irrigation water (i.e., drip or subsurface irrigation of lettuce) or only irrigating crops that are cooked prior to eating (i.e., potatoes) (van Ginneken & Oron 2000).

Depending on the level of treatment of reclaimed wastewater, numerous waterborne pathogens have been reported to be present including various bacteria (e.g., Escherichia coli, enterococci, Staphylococci, Salmonella) (Goldstein et al. 2014b; Santiago et al. 2018), protozoans (e.g., Giardia and Cryptosporidium) (Ryu et al. 2007; Domenech et al. 2018) and viruses (e.g., adenovirus, norovirus) (Jjemba et al. 2010; Gonzales-Gustavson et al. 2019). Yet the risk to human health from exposure to these pathogens in reclaimed wastewater used for irrigation is not always clear. This uncertainty arises from (1) lack of analysis of pathogens themselves and reliance instead on fecal coliforms which do not always correlate with pathogen abundance (Wu et al. 2011); (2) lack of information on the viability of the pathogens due to reliance on nucleic acid-based techniques rather than culturing (Whiley & Taylor 2016); (3) lack of studies on persistence of pathogens after irrigation and (4) left-censored datasets of pathogen concentrations below detection limits (Kato et al. 2013). Furthermore, there are a limited number of risk assessments of vegetable crops irrigated with reclaimed wastewater where the crops themselves were evaluated for pathogens. Instead, many quantitative microbial risk assessments (QMRAs) for irrigation of foods are based on measuring coliforms or pathogens in water that are presumed to be retained on crops (van Ginneken & Oron 2000; Hamilton et al. 2006; Al-Sa'ed 2007; Ryu et al. 2007; Verbyla et al. 2016; Gonzales-Gustavson et al. 2019).

In addition to risks posed by pathogens, PPCPs in reclaimed wastewater used for irrigation may also present human health risks. There are several review articles on the impacts of PCPPs on vegetable crops irrigated with reclaimed wastewater (Qin et al. 2015; Wu et al. 2015; Miller et al. 2016; Fu et al. 2019) and human risk assessment. The emphasis of these studies has been the uptake and translocation of chemicals of concern into edible roots and above ground tissue of fruit or vegetable crops. Studies have been conducted using hydroponic systems (Dodgen et al. 2013, 2015; Wu et al. 2013) and greenhouse (Wu et al. 2010; Goldstein et al. 2014a) or lysimeter studies (Malchi et al. 2014; Mordechay et al. 2018). Few studies have been conducted under field-irrigation conditions. Field conditions that were studied included drip irrigation (Riemenschneider et al. 2016; Christou et al. 2017b) or buried drip tape (Wu et al. 2014) for water delivery to the plants in commercial agriculture practices. The use of sprinkler irrigation of home vegetable gardens has not been reported. Studies evaluating risks for humans consuming fruits and vegetables irrigated with reclaimed wastewater report the risk is de minimis (Prosser & Sibley 2015; Christou et al. 2017b; Liu et al. 2020). Two exceptions are notable. Carbamazepine and its metabolites were detected in urine of individuals who consumed produce irrigated with reclaimed water (Paltiel et al. 2016) and the metabolites of carbamazepine, as genotoxic compounds, were reported to pose a risk through the consumption of the leaves of carrots and sweet potato and the roots of carrots (Malchi et al. 2014).

Few studies include a comprehensive evaluation of both microbial and chemical risk (Munoz et al. 2010) from irrigation with reclaimed wastewater in home gardens and evaluate exposure risk to both irrigation water and produce. There is also evidence to suggest that individuals may use reclaimed wastewater in a variety of ways beyond irrigation. In a survey of community members in the study area of Hyrum, UT, it was reported that 11.8% of the citizens who had access to reclaimed wastewater used the water to fill swimming pools, 29.4% for children's water play, and the 3.7% for infrequent drinks out of the hose; exposure routes not considered in previous studies (Flint & Koci 2021). Therefore, the overall goal of this project was to evaluate chemical and microbial risks to residents, primarily using reclaimed wastewater for sprinkler irrigation in home gardens and infrequently as drinking water. Microbial risk was evaluated using traditional coliform measurements, quantitative polymerase chain reaction (qPCR) for specific pathogens and antibiotic resistance genes in irrigation water and associated with fruits and vegetables.

All water samples were collected in separate sterile 4-L amber glass bottles for microbial and PPCP analysis. Details of the samples collected and dates of collection are shown in Supplementary Table S1. After aseptic collection of samples for microbial analysis, samples were stored on ice during transport to the laboratory. All water samples for microbial analysis were stored at 4 °C upon arrival and were processed within 12 h of arrival at the laboratory. The water samples for PPCP analysis were returned to the laboratory within 4 h and stored at 4 °C.

Fourteen types of vegetables and fruits (Supplementary Table S1) were collected from five household gardens that were located close to the WWTP (Figure 1, locations G1 to G5). Vegetables and fruits that are eaten raw, with the exception of potatoes, were sampled and analyzed along with soil co-located with the plants. Vegetation and soil collection and processing methods are detailed in the Supplementary material and Table S1.

Pathogens in water, soil and vegetables were quantified by both culture and qPCR. Culturing of E. coli and enterococci from samples was performed following EPA Methods 1603 and 1600, respectively (USEPA 2002a, 2002b). Modified MTEC and MEI agar (Hardy Diagnostics, USA) were prepared based on the manufacturer's instructions. Water samples were also filtered through sterile mixed cellulose ester 0.45 and 0.22 μm filters (Fisher Scientific Inc., USA), as necessary. Nucleic acids were extracted from the filters using the methods described in the Supplementary material. Quantitative PCR was used to determine the abundance of the uidA gene of E. coli, the 23S rRNA gene of Enterococcus spp. (enterococci), the invA gene of Salmonella enterica, the mip gene of Legionella pneumophila, the β-giardin gene of Giardia intestinalis, the polymerase-capsid junction (ORF1-ORF2) of norovirus genotype I and the hexon gene for serotypes human adenovirus serotype 40 and 41. Standard curves were made using a nine-fold serial dilution of plasmids containing uidA, invA or 18S genes or stabilized Ultramer^® DNA oligonucleotides (Integrated DNA Technologies, USA) for L. pneumophila, G. intestinalis, Norovirus GI and Human Adenovirus 40/41. The thermocycler conditions, primer and probe concentrations and mastermix used for qPCR are presented in Supplementary Table S2. The abundance of pathogens in samples determined by qPCR in gene copies/mL water or gene copies/g vegetable or soil were converted to cells per mL or g by dividing by the number of gene loci per organism. It was assumed that E. coli had five copies of uidA (Metcalf & Wanner 1993); enterococci have nine copies of 23S rRNA (Chakravorty et al. 2007); Salmonella spp. have five copies of invA (González-Escalona et al. 2009) and Giardia intestinalis have two copies of β-Giardin P241 (Alonso et al. 2010). Norovirus, adenovirus and Legionella were assumed to have one copy of the gene targeted by qPCR.

The nucleic acids extracted from the water distribution system lines (n=12), irrigation reservoir (n=2), treatment plant effluent (n=2) and crops (n=3 each of cherry tomato, cucumber, lettuce and zucchini) were analyzed for antibiotic resistance genes at the University of Minnesota. Details of the microfluidic qPCR assay targeting 47 antibiotic resistance genes were described previously (Ahmed et al. 2018). The resulting concentrations of ARGs were normalized by the mass of DNA extracted from each sample as measured by a Qubit fluorometer.

All pathogen concentrations for vegetables and water were aggregated and a Bayesian analysis of the left-centered data was performed to generate distributions for each exposure route. Details of the Bayesian analysis, R code, distributions of raw data and model fit parameters are included in Supplementary Figures S1 and S2. The two exposure routes considered for home gardeners were: (1) direct ingestion through drinking from a hose and (2) eating raw produce irrigated with the reclaimed water with and without rinsing. The exposure routes were selected based on surveys of the community where 3.7% of respondents indicated drinking from the hose containing reclaimed wastewater and 92% of respondents used the reclaimed wastewater to irrigate vegetables (Flint & Koci 2020). The daily dose of pathogen ingested through eating vegetables was estimated by Dose_vegetables = C_i*M_veg*M_body. Furthermore, as many QMRA studies evaluate contaminated water retained on vegetables as a surrogate for testing vegetables themselves, we also estimated the daily dose of pathogen ingested through reclaimed wastewater retained on vegetables by Dose_{water on vegetables} = C_i*M_veg*M_body*V_{water retained}. The daily dose of pathogen ingested during drinking from a reclaimed wastewater hose was estimated by Dose_{drinking water ingestion} = C_i*V_{drinking water}. Where C_i is drawn from 10,000 resamplings of the Bayesian transformed distribution for each pathogen in reclaimed wastewater or on vegetables measured by culture and/or qPCR methods; M_veg is the mass of vegetables consumer per unit body weight per day; the M_body is the body weight of an average adult in the United States; V_{water retained} is the volume of reclaimed wastewater retained on lettuce after irrigation and V_{drinking water} is the volume of water consumed from a hose. Furthermore, the V_{drinking water} from a hose was assumed to be the volume of cold tap water consumed at home by pregnant women employed part time or less. While this assumption is less conservative than assuming the total volume of water consumed in a day comes from the garden hose, it is likely still over estimating the volume of reclaimed wastewater that is consumed per exposure event from a hose (USEPA 2019).

The annual probability of infection after eating vegetables or drinking water contaminated with E. coli (Powell et al. 2000) or Salmonella (Ahmed et al. 2010) were estimated using the beta-Poisson model, with enterococci (Tseng & Jiang 2012) and adenovirus (Teunis et al. 2016) by two different exponential models, and with norovirus (Messner et al. 2014) by the fractional Poisson model. Details of the exposure models, parameters and assumptions for each organism are shown in Supplementary Tables S3 and S4. The probability of illness, P_ill, was determined by modifying the probability of infection, P_inf, by the model P_ill=P_inf×P_ill|inf. The probability of illness given infection, P_ill|inf, for each organism is shown in Supplementary Table S4. The annual risk of disease, P_{ann ill}, was calculated by P_{ann ill}=1−(1−P_ill)ⁿ, where n represents the number of exposure events per year (Seidu et al. 2013). In this study, it was assumed that people ate vegetables irrigated with reclaimed wastewater either 15, 30 or 90 days a year and drank reclaimed wastewater from the garden hose once or twice per week during the growing season or 12–36 days per year. The models did not account for the decay of pathogens on produce after deposition.

The estimated annual probability of illness was compared against two benchmarks. First, residents drinking reclaimed wastewater from the hoses were compared against the acceptable probability of illness benchmark from the USEPA of a one-time infection per 10,000 individuals in a given year (abbreviated as 10⁻⁴ per person per year or pppy) (Rose & Gerba 1991). Previous QMRA studies for vegetables irrigated with reclaimed wastewater have commonly applied the drinking water 10⁻⁴ probability of infection per year benchmark (Hamilton et al. 2006). This benchmark is highly conservative. Second, data from the CDC on annual illness originating from pathogen contaminated vegetables reports 9,388,075 annual illnesses from 1998 to 2008 for an average population of 298 million US citizens in 2006 (Scallan et al. 2011). This results in a 3.15×10⁻² pppy or 315 in 10,000 individuals in a given year falling ill from pathogens on produce.

The QMRA distribution fittings, calculations and random sampling for the risk study were conducted with SAS (ver. 9.4; SAS Institute, Inc., Cary, NC) using PROC UNIVARIATE, PROC COOR and the RAND function. All input parameters were drawn from 10,000 random samplings of their probability distribution functions (Supplementary Tables S3 and S4). The sensitivity of the estimated probability of infection due to variability in the input parameters was assessed by evaluating the Pearson correlation between the P_{ann inf} and input parameters (e.g., pathogen concentration, volume of water retained). This method was chosen due to its ease of implementation and ability to show nonlinear correlations between parameters as reported previously (Hamilton et al. 2006; Haas et al. 2014; Verbyla et al. 2016).

The analytes used in this study included pharmaceuticals (acetaminophen, carbamazepine, gemfibrozil, sulfamethoxazole and fluoxetine), hormones (estrone, progesterone and ß-estradiol), personal care products (N,N-diethyl-metatoluamide (DEET) and triclosan), a fire-retardant (tris-2-chloroethyl phosphate) and caffeine. These compounds were selected due to their wide usage and varying physical/chemical properties providing representative chemicals of the many the PPCPs commonly used. Samples were analyzed for PPCPs using a modified USEPA Method 1694. The selected PPCPs were quantified using an Agilent 1290 Infinity LC system with an Agilent 6490 Triple Quadrupole MS. Details of sample processing and analytical methods are given in the Supplementary material.

The two PPCP exposure routes considered for home gardeners were the same as those considered for the microbial risk. None of the PPCPs evaluated in this study are reported carcinogens, and quantitative risks associated with chemical PPCP exposure due to the use of reclaimed wastewater was estimated by using the following clinical dose or toxicity endpoints: clinical dose of pharmaceuticals, mg/d; toxic dose low (TD_LO), mg/kg and average daily intake (ADI), mg/kg, for 70 kg adult and/or 7 kg infant; no-observed-adverse-effect level (NOAEL), mg/d and the margin of exposure (MOE)=NOAEL/Calculated Dose based on 1 L/d water or 336 g/d vegetable intake, which is recommended to be >10 (Coordinators 2017). The values of these clinical doses or toxicity endpoints available for the PPCPs monitored in this study are shown in Supplementary Table S5. If a clinical dose was available for a compound, the volume of water, Vol_water, or mass of vegetable matter, Mass_vegetable, that would need to be consumed to reach this clinical dose was calculated by Vol_water=Clinical Dose/C_w or Mass_vegetable=Clinical Dose/C_v, where C_w and C_v are the highest concentration of the compound measured in the paired garden spigot water (mg/L) and the household garden vegetable samples (mg/g), respectively, from samples collected in 2018. If a TD_LO or an ADI were available, the lowest Vol_water or Mass_vegetable required to reach these doses were calculated by Vol_water=TD_LO or ADI*70 kg/(C_w) for adult exposure [or 7 kg/(C_w) for infant exposure] or Mass_vegetable=TD_LO or ADI*70 kg/C_v for adult exposure [or 7 kg/(C_v) for infant exposure]. If a NOAEL for a compound was available, an MOE was calculated based on the ratio of the calculated dose expected from exposure to the highest concentrations for that PPCP detected in a spigot or vegetable sample collected during the study, and standard 1 L/d water or 336 g/d vegetable intake values. The MOE was then calculated by MOE=NOAEL/[(C_w)*(1 L/d)] or NOAEL/[(C_v)*(336 g/d)].

Pathogen genomic material was frequently detected on home garden produce and in soils (Figure 3). There was a wide variability in the pathogen detection frequency with the uidA gene of E. coli (87%) and the 23S gene of enterococci (88%) being found most frequently, while invA gene of Salmonella (40%), adenovirus 40/41 (42%) and norovirus GI (17%) were detected less frequently. The highest pathogen concentrations observed were for enterococci ranging from 3.4±2.1 log gene copies/g, then the uidA gene of E. coli at 2.1±1.2 log gene copies/g, invA gene of Salmonella spp. at 1.3±1.7 log gene copies/g, adenovirus 40/41 at 23±69 gene copies/g and norovirus at 15±66 gene copies/g. In all cases, the pathogen genomic material was in higher concentration in stomached samples as compared to the corresponding vegetable rinse water (Figure 3). Culturable enterococci were found more frequently in the stomached vegetable samples (23 of 37 samples) than in the rinse water (20 of 40 samples) from the vegetables. In contrast, fewer culturable E. coli were detected in stomached vegetable samples (12 of 37 samples) than in the rinse water (17 of 40 samples). Culturable enterococci (5 of 12 composite samples) and E. coli (1 of 12 composite samples) were detected in the garden soils, although at low levels (3.4±6.3 enterococci CFU/g and 0.08±0.29 E. coli CFU/g). Pathogen genomic material was in higher concentration in soils compared to culture-based results (Figure 3) and ranged from 20 gene copies/g soil to 4 log gene copies/g soil.

Eight antibiotic resistance families were detected in lettuce, cucumber, zucchini and cherry tomatoes (Supplementary Figure S3). In general, the cherry tomatoes and lettuce were found to have higher abundances of antibiotic resistance genes than the cucumbers and zucchini. These findings support the studies by others that suggested antibiotic resistance genes decreased in concentration along the continuum from soil, rhizosphere, roots, leaves and fruits (Cerqueira et al. 2019). In this study, sulfonamide resistance genes made up 7–28% of the antibiotic resistance genes detected in the crops in which sulfamethoxazole was quantifiable (Table 1 and Supplementary Figure S3), i.e., lettuce, zucchini and cucumber. Relatively few studies have reported on field trials evaluating the plant uptake of antibiotics from reclaimed water irrigation (Malchi et al. 2014; Wu et al. 2014; Prosser & Sibley 2015; Christou et al. 2017a). In contrast to the results herein, some studies did not detect the accumulation of sulfamethoxazole in tissues of vegetables irrigated with water containing 0.28–250 ng/L of the antibiotic (Goldstein et al. 2014a; Wu et al. 2014), yet others reported accumulation in select crops (Malchi et al. 2014; Franklin et al. 2016).

The frequency of detection of PPCPs, as well as the maximum concentration detected in various samples collected during the field study, is shown in Table 1. Samples included home garden spigots, home garden vegetables and soils that were irrigated with reclaimed wastewater and corresponding samples from the control locations. The spigot water from the Salt Lake City control site had detectable concentrations of carbamazepine and DEET in all samples, while progesterone and sulfamethoxazole were detected in one of the three samples collected (Supplementary Table S6).

PPCP detection was generally more frequent and at higher concentrations in the Hyrum household samples compared to the control location samples (Table 1 and Supplementary Table S6), indicating that the Hyrum WWTP effluent did impact PPCP concentrations in home garden samples collected in this study (Table 1). All PPCPs that were found in household spigot water, except triclosan and tris-(2-chloroethyl), were found in fruits and vegetables irrigated with this reclaimed wastewater. Triclosan (Wu et al. 2013) has been reported to accumulate in plant roots although roots were not analyzed in this study. Triclosan also biodegrades in soil (Durán-Álvarez et al. 2015) supporting the lack of detection of this compound in both the irrigated and control soils. In contrast to the results herein, the flame retardant tris-(2-chloroethyl) phosphate was reported by others to accumulate in strawberries and lettuce up to 200 ng/g dry weight (Hyland et al. 2015a). It was shown that the accumulation and translocation of tris-(2-chlorethyl) phosphate was primarily due to transport from the roots to shoots (Hyland et al. 2015b). Therefore, although this flame retardant was detected in 30% of the spigot water samples, its lack of detection in soils, likely due to biological transformation of the compound in soils (Zhang et al. 2021), resulted in a lack of accumulation of the compound in vegetables in this study.

DEET and sulfamethoxazole accumulated in the edible portion of plants but were not detected in the soil (Table 1). Acetaminophen, caffeine, carbamazepine, fluoxetine and progesterone also accumulated in the edible portion of plants in concentrations 2–13 times greater than found in the adjacent soil (Table 1). Fluoxetine as a cation, sorbs to soil limiting plant uptake (Wu et al. 2010, 2014) and has limited translocation within plants (Wu et al. 2013). But in this study fluoxetine was one of the more frequently detected PPCP in vegetable samples (52%) and was found in all vegetable types (cucumber, tomato, squash, peppers, apples, potato, strawberry) (Supplementary Figure S4). Progesterone was also present in 50% of the tested vegetables (mean of 10.16±4.57 ng/g) as well as 53% of the spigot water (mean of 4.07±5.84 ng/L) and in 20% of the tested soil samples (mean of 3.84±1.51 ng/g) (Supplementary Figure S4). Progesterone, however, has been reported to be naturally present in plant tissue including tomato, apples, potato, peas and beans and functions as a natural plant growth regulator (Janeczko 2012). Detection of progesterone in control plant samples in this study, including tomatoes, onion and potatoes, occurred at a lower frequency (33%) and at a lower average (8.48±3.22 ng/g) but not statistically different concentration than plants irrigated with reclaimed wastewater. Gemfibrozil was not detected in spigot, plant nor soil samples despite being consistently detected in the WWTP effluent and distribution line samples (Figure 2 and Table 1).

As with all risk assessments, there are uncertainties in the input variables that should be considered when interpreting the results. The QMRA likely overestimated the risk to consumers as the dose of pathogens was estimated from qPCR results which include both live and dead organisms. Furthermore, the risk may be overestimated as the decay of pathogens on the vegetables after irrigation was not considered. Other inputs into the QMRA that could result in over or underestimation of risk include: the mass of the vegetable consumed, the volume of reclaimed wastewater retained on the vegetable, the individual bodyweight, exposure frequency, the volume of reclaimed wastewater consumed from a garden hose and the potential for population immunity. The role of population immunity in the risk of illness was not considered herein, but it is likely that sustained exposure to the pathogens in the reclaimed wastewater used for irrigation may result in fewer annual illnesses in this exposed population.

The chemical risk assessment based on clinical dose or toxicity endpoints available for the PPCPs monitored in this study is shown in Table 1 and is considered negligible for all PPCPs evaluated. The calculated MOE for water or vegetable consumption values were based on the highest concentrations of a given PPCP detected in a spigot or vegetable sample collected during the study (Table 1), and the volume of water or mass of vegetable matter that would have to be consumed to reach the NOAEL for that PPCP. The ADI via ingestion by exposure pathway (contaminated food and water) and route (drinking or eating contaminated food or water) is listed as the water volume (gallons per day) or mass (tons per day of produce) that would be required to be consumed to reach either the clinical dose, the lowest toxic threshold dose TD_LO or the recommended ADI, whichever was lower. The lowest MOE was determined to be 238 based on a daily consumption of 336 g/d of green beans grown in the garden location with the maximum progesterone concentration of 25 ng/g (Table 1), compared to the recommended value for the MOE of 100 or more. Risks associated with clinical dose, toxic dose or ADI-related impacts were found to be even lower based on the excessive daily quantities of water (444 gal/d for infants exposed to tris-(2-chloroethyl phosphate) to 31.3 MGD for adults exposed to carbamazepine) or vegetables (0.85 T for adult DEET exposure to 118 T/d for adult exposure to carbamazepine) that are required to be consumed to reach these levels of human health concern.

The PPCPs in the irrigation water all represent insignificant risks to the public under worst-case exposure assumptions of direct ingestion of either reclaimed wastewater itself, or raw foods irrigated with reclaimed wastewater containing low concentrations of these PPCPs. The required harmful human threshold consumption rates of water and produce from the private gardens sampled in this study are not physically possible, which confirms findings from other researchers (Prosser & Sibley 2015; Christou et al. 2017b) that human health risks associated with PPCPs in reclaimed wastewater appear to be de minimis, although this study did not evaluate degradation products of various PPCPs that may pose higher risk than the parent compound. Risk from PPCPs in reclaimed water via direct water ingestion or ingestion of raw foods irrigated with reclaimed wastewater does not contribute directly to the overall risk of reclaimed wastewater use in Hyrum, the bulk of which comes from health risks associated with pathogen exposure.

Limitations in the PPCP risk assessment primarily arise from the small subset of the potential organic chemicals present in the wastewater treatment system that were identified and quantified in this study. The compounds that were studied were selected based on their presence in other reclaimed wastewater sources, the range of physical/chemical/biological properties they represent and the authors' experience with these compounds in prior and on-going studies. There may be other compounds, i.e., PFAS compounds for example, that were not quantified in this study that could represent a greater human health risk than those that were studied. In addition, as Malchi et al. (2014) have suggested, metabolites could potentially represent greater risk than their parent compounds, and these issues warrant further study. Furthermore, lifetime exposure to PPCPs in home garden vegetables and reclaimed wastewater at low-doses observed in this study could present potential risks not captured by the clinical dose values available for these compounds.

The results presented herein are one of the few studies looking at uptake of PPCPs, pathogens and antibiotic resistance genes in food crops at a field scale. Overall, the results suggest that after irrigation with reclaimed wastewater: (1) the risks from pathogens on crops eaten raw exceed benchmark levels of gastroenteritis in the US (Scallan et al. 2011), (2) the risks from exposure to PPCPs taken into crops or through direct ingestion of reclaimed wastewater is de minimis, but (3) multiple families of antibiotic resistance genes are present on food crops which may be associated with the presence of antibiotics in this reclaimed wastewater. Therefore, while the direct chemical risk from exposure to PPCPs is low, antibiotics in reclaimed water may exert selective pressure on microorganisms on food crops to acquire or maintain antibiotic resistance. When partnered with surveys of community members using the reclaimed wastewater (Flint & Koci 2021), this work suggests that increased outreach to the public is required to ensure community members are not drinking the nonpotable water and that they are properly washing harvested fruits and vegetables before consumption. Furthermore, the inclusion of a post-chlorination step should be considered for the reclaimed wastewater that is distributed for irrigation to minimize the growth and accumulation of biofilms and subsequent antibiotic resistance genes within the secondary water distribution system. Finally, QMRAs evaluating risk to consumers from irrigation with reclaimed wastewater should consider also sampling vegetation rather than just water to more accurately reflect the true health risk associated with reclaimed wastewater reuse.

The authors acknowledge funding from the US Department of Agriculture (NIFA 2017-69007-26310). The authors are grateful to Stephen Cavanaugh and Rubayat Jamal for aid in processing the samples, Kevin Maughn for allowing access to the Hyrum WWTP and the Food Science laboratory at USU for access to the stomacher. The authors also thank the residents of Hyrum who allowed sampling of their irrigation water and garden crops.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.