Variability in urinary neonicotinoid concentrations in single-spot and first-morning void and its association with oxidative stress markers

Human exposure to neonicotinoid insecticides (hereafter “neonics”) is a concern. Spot urine samples have been widely used in the assessment of exposure to neonics. Urinary concentrations, however, can vary greatly over time due to variable exposure, potentially leading to exposure misclassification. In this study, within- and between-individual variability of urinary concentrations of 13 neonics and their metabolites collected consecutively for up to 44 days from 19 individuals were examined. We also measured seven oxidative stress biomarkers (OSBs) in repeated urine samples to elucidate their relationship with neonic exposure by mixed regression models. Intraclass correlation coefficients (ICCs, a ratio of between-individual variance to total variance) were used to assess the reproducibility of neonic/metabolite concentrations. Sensitivity and specificity were used to evaluate how well spot urine samples determined an individual’s average exposure over 44 days. A fair to good reproducibility was observed for N-desmethyl-acetamiprid (ICC = 0.42), whereas thiamethoxam, imidacloprid, clothianidin, imidaclothiz, 6-chloronicotinic acid, and sulfoxaflor showed poor reproducibility (ICC = 0.02–0.37). Use of single-spot urine samples to classify high (top 33%) exposure showed higher specificities (0.68– 0.92) than sensitivities (0.32–0.88). The minimum number of specimens (k) required to estimate participant-specific mean for neonic exposures within 20% of the “true” values ranged from 16 to 172. Significant positive correlations were found between some of neonic and OSB concentrations. The high variability found in the urinary concentrations of most neonics/ metabolites suggests that a single measurement can result in exposure misclassification.


Introduction
Neonicotinoid insecticides (hereafter "neonics"; Table S1) were developed to replace organophosphate and carbamate insecticides and are systemic in design for their neurotoxic action on the nicotinic acetylcholine receptors (Lu et al., 2018). Since the early 1990s, the use of neonics has increased dramatically worldwide, accounting currently for >25% of the global insecticide market (Bass et al., 2015;Cimino et al., 2017;Wang et al., 2018). The annual global production of neonic active ingredients was estimated at 40 million lbs in 2010 (Wang et al., 2018). In the United States, over 4 million lbs of neonics have been applied annually on 79% of the cropland (~200 million acres) (Cimino et al., 2017;Douglas and Tooker, 2015). In addition, neonics are used on lawns, home gardens, livestock, and pets (Jeschke et al., 2011). Neonic residues have been reported to occur in soil, house dust, aquatic environment, and foods (Anderson et al., 2015;Basley and Goulson, 2018;Bennett et al., 2019;Bishop et al., 2018;Hladik et al., 2018;Lu et al., 2018;Mitchell et al., 2017;Zhang et al., 2018a). Neonics have long biological half-lives in the environment (e.g., >1,000 days in soils) (Bonmatin et al., 2015). Exposure of humans to neonics has been associated with adverse developmental or neurological outcomes, including congenital heart defects, anencephaly, and autism spectrum disorders (Cimino et al., 2017;Zhang et al., 2018b).
Neonics are metabolized by human cytochrome P450 enzymes via two different pathways: hydroxylation and dehydrogenation, and nitroimine reduction and cleavage to yield nitrosoimine, guanidine, urea derivatives and hydroxylated metabolites, which are eliminated in urine (Wang et al., 2018). The half-lives of neonics and their metabolites in human bodies are in hours to days. For example, dinotefuran (DIN), clothianidin (CLO), imidacloprid (IMI), and N-desmethyl-acetamiprid (N-DMA) have reported half-lives of 0.17-1.45 days in humans . Single-spot or first-morning-void (FMV) urine specimens have been used in biomonitoring of human exposure to neonics (Table S2) (Kabata et al., 2016;Marfo et al., 2015;Nomura et al., 2013;Osaka et al., 2016;Taira et al., 2013;Ueyama et al., 2015;Wang et al., 2015). However, suitability of a single-spot or FMV urine to represent integrated exposure in individuals over time has not been investigated. Because exposure to neonics is expected to be episodic, and their toxico-kinetics vary in humans, appropriate exposure classification is important for epidemiologic data analysis Osaka et al., 2016). Our recent study that examined daily variability in urinary concentrations of organophosphate and pyrethroid insecticides, as well as phenoxy herbicides, found considerable within-individual variations, which suggested episodic nature of exposure to these pesticides (Li et al., 2019a). Studies that determine inter-and intraindividual variability in urinary neonic levels are needed to evaluate suitability of a singlespot or FMV urine sample in exposure assessments. To date, no studies have evaluated temporal variability of neonics and their metabolites in urine.
An increasing number of in vivo and in vitro studies indicated that oxidative stress plays a significant role in neurotoxicity, immunotoxicity, hepatotoxicity, nephrotoxicity, and reproductive toxicity of neonics (Ozsahin et al., 2014;Pandey and Mohanty, 2015;Sheets et al., 2016). Research on neonic toxic mechanisms also suggested the involvement of oxidative stress (Al-Sarar et al., 2015;Galdíková et al., 2015;Lopez-Antia et al., 2015;Wang et al., 2018;Yan et al., 2015). The generation of reactive oxygen/nitrogen species represents an important role in neonic-induced oxidative damage to lipids, proteins, and DNA (Wang et al., 2018). Earlier studies of oxidative stress induced by neonics were conducted using cell bioassays (Al-Sarar et al., 2015;Galdíková et al., 2015) or in vivo animal studies (Lopez-Antia et al., 2015;Yan et al., 2015). Nevertheless, associations between markers of oxidative stress and neonic exposure in humans have not been investigated.
To address the data gaps noted above, we performed a study with repeated analysis of neonic exposure in 19 healthy individuals, with the following aims: (1) to assess inter-and intraindividual variability in urinary concentrations of 13 neonics/metabolites in single-spot and FMV urine samples collected consecutively for up to 44 days; (2) to evaluate sensitivity and specificity of urinary measures of neonics/metabolites in individuals by comparing the "true" and "predicted" values; and (3) to examine the associations between urinary concentrations of 13 neonics/metabolites and biomarkers of oxidative damage to lipids, proteins and DNA.

Study population and sample collection
Thirteen neonics and their metabolites were analyzed in repeated urine samples from a study designed to examine the variability in urinary levels of oxidative stress biomarkers (OSBs) (Martinez-Moral and Kannan, 2019). From February to April 2018, 19 volunteers (58% men; 42% women; Table S3) who resided in Albany, New York, USA, were recruited by convenience. The study volunteers were healthy, nonsmoking, with no history of diabetes or renal disease that may affect metabolism, and no documented occupational exposure to neonics. The mean age of participants was 33.8 ± 12.4 years, and the average body mass index (BMI) was 23.6 kg/m 2 . Majority of the participants were Asians (68%) followed by Caucasians (32%). None of the participants was a habitual alcohol consumer. The Institutional Review Board of the New York State Department of Health approved this study.
The urine samples from 19 participants included spot and FMV samples, which were collected in 50-mL polypropylene (PP) tubes, consecutively, for up to 44 days. The samples were devoid of personal identifiers and were stored at −20° C until analysis. A total of 515 urine samples, with an average of 27 samples per person, were available (Table S3). Among the 515 urine samples, 243 samples were FMVs, and the rest were spot urine. FMV refers to those urine samples that were collected immediately after waking up in the morning. For the purpose of data analysis, all 515 samples were considered in the category of spot urine, whereas 243 samples that were exclusively identified as FMVs were grouped in that category.
The method for the extraction of neonics from urine by a solid-phase extraction (SPE) method was described earlier (Honda et al., 2019). Briefly, urine samples (500 μL) were spiked with stable isotopically labeled internal standards and mixed with 2% formic acid (1.5 mL). The samples were then passed through an SPE cartridge (Bond Elut Plexa 3cc, 60 mg, Agilent, MA, USA). The elutes (3 mL of methanol) were dried under a gentle stream of nitrogen and reconstituted with 250 μL of methanol/water (1:1, v/v).
Quantification of neonics/metabolites was accomplished by the use of native standards prepared in synthetic urine by the addition of isotopically labeled internal standards. The recoveries of target analytes through the analytical method were examined by spiking a known amount of target neonics/metabolites (1.0, 10, and 20 ng/mL) into a synthetic urine sample. The recovery of DIN in the spiked urine sample was <10%, and, therefore, this compound was excluded from further analysis. The recoveries of the other 13 analytes ranged from 70% to 120%, with relative standard deviations (RSDs) of ≤17%. The method limits of detection (LODs) of the 13 analytes were in the range of 0.002 to 0.099 ng/mL. The concentrations of all target compounds in procedural blanks were below the LOD (Table  S4). A total of 25 urine samples were analyzed in duplicate to evaluate method precision.

Determination of oxidative stress biomarkers
, and 15(R)-prostaglandin F 2α (15-PGF 2α ), were analyzed in urine samples by following the method described in detail elsewhere (Martinez-Moral and Kannan, 2019). Quantification of OSBs was based on an isotope-dilution method and the details of which are provided in the Supplementary Data.

Determination of urinary creatinine
Urinary creatinine concentrations were determined by Acquity HPLC (Waters, Milford, MA, USA), coupled to Acquity TQD MS/MS (Waters, Milford, MA, USA) by following the method described earlier (Martinez-Moral and Kannan, 2019).

Data analysis
Data were analyzed using SPSS 19.0 (SPSS Inc., Chicago, IL, USA). Concentrations below the LOD were replaced with a value of LOD divided by square root of 2. Data were logtransformed (χ + 1) to normalize their distributions for statistical analyses. Mixed regression models were developed to explore the associations between urinary concentrations of neonics/metabolites and OSBs.
Intraclass correlation coefficients (ICCs, a ratio of between-individual variance to total variance) were computed as a measure of the reproducibility of urinary concentrations of neonics/metabolites within individuals over time. The ICC values were categorized as excellent (ICC ≥ 0.75), fair to good (0.75 > ICC ≥ 0.40), and poor (ICC < 0.40) for the determination of reproducibility (Attfield et al., 2014;Wang et al., 2016). Between-and within-individual variances were calculated by a linear mixed-effects model, using maximum likelihood estimation. Akaike Information Criterion (AIC) values were used to evaluate the fitness of models of urinary concentrations reported as uncorrected values, creatinine-corrected values, and creatinine as a covariate. A lower AIC value indicates a better fitness of the model. By comparing the distribution of "true" and "predicted" levels, sensitivity (a true positive rate) and specificity (a true negative rate) were evaluated on one, two, or three randomly selected spot/FMV urine sample(s) from each participant as predictors of high (top 33%) exposure based on their 44-day average concentration (Wang et al., 2016). FMV urine samples (n = 243) were analyzed separately to test whether this represents exposure better than do spot urine samples. For the "true" levels, arithmetic means of log-transformed (χ + 1) concentrations of neonics/metabolites from each participant (n = 19) were calculated for samples collected during the 44-day period. The values were scaled into tertiles, and the "true" top 33% exposure levels were identified. For the "predicted" levels, ten data sets, with each as containing one randomly selected spot or FMV sample per participant, were produced (i.e., 19 observations per data set). Within each randomly selected data set, concentrations of the 19 observations were identified for the "predicted" top 33% exposure levels. The average sensitivity and specificity observed across the ten separately and randomly selected data against the "true" exposure were calculated for spot or FMV urine samples. To examine whether a collection of multiple urine samples would improve the sensitivity, we repeated the analysis, using arithmetic means of two or three urine samples that were randomly selected from each participant and collected on different days. The minimum number of spot or FMV urine samples (k) required to predict the participant-specific mean concentration of neonics/metabolites, to be within 20% of the "true" values with a probability of 95%, was calculated after log-transformation (χ + 1) of measured concentrations, using the following equation: where CV is the within-individual coefficient of variation (Kim et al., 2014;Wang et al., 2016).

Profiles of urinary neonics and their metabolites
Neonics and their metabolites were measured in 515 urine samples collected consecutively from 19 participants (13-40 samples per person) during February-April 2018. The available demographic information of the study participants is presented in Table S3. The detection frequency (DF), distribution, and volume-and creatinine-adjusted concentrations of neonics/ metabolites in spot and FMV urine are presented in Tables 1, S5 and S6. The DF of neonics in urine varied widely among individuals. The most frequently detected neonics/metabolites in urine were in the decreasing order of: N-DMA > 6-CN > IMI -CLO > SUF > IMZ > THX (DFs, 75-100%). Other neonics/metabolites (FLO > NIT > N-DMT > ACE ≈ THI > TA) were detected in 7-62% of the samples analyzed. Further analyses of the data were restricted to those analytes that had a DF of >60% in all samples.
The creatinine-corrected concentrations of neonics/metabolites (micrograms per gram creatinine) in spot urine samples collected from the 19 participants over the 44-day period are shown in Fig. 2. There were no consistent patterns in neonics/metabolites concentrations among the 19 participants. The urinary concentrations of THX and SUF varied by one to three orders of magnitude within individuals during the study period. The urinary concentrations of IMI, CLO, IMZ, N-DMA, and 6-CN varied by an order of magnitude for all participants, with the exception of IMI in three participants (P5, P13, and P16) and CLO in two participants (P5 and P12). All neonics/metabolites varied within an order of magnitude for two Asian Indian participants (P2 and P3) and a Caucasian (P8).
To the best of our knowledge, seven studies have reported urinary concentrations of neonics and their metabolites in general populations, with four of the seven studies for a Japanese population (Table S2). The urinary concentrations of THX, IMI, CLO, 6-CN, IMZ, and N-DMA with DFs of ≥55% were also reported in a US population (Honda et al., 2019), and the urinary concentrations of these compounds (median, 0.07-0.43 ng/mL) were similar to those reported in our study. In comparison to the US population, Japanese adults had one to two orders of magnitude higher urinary concentrations of NIT, THX, IMI, ACE, THI, CLO, and N-DMA, which suggested the prevalence of neonic exposure in Japan (Marfo et al., 2015;Ueyama et al., 2014Ueyama et al., , 2015. In a study of Japanese children, urinary neonic levels were <LODs of 0.03-1.07 ng/mL and DFs were <25%, with the exception of DIN (median, 0.44 ng/mL; DF, 58%) (Osaka et al., 2016). The median concentration of IMI measured in a general population in China  fell within the range reported in our study. A nationwide survey of urinary concentrations of neonics in China showed higher median concentrations (ng/mL) of THX (0.15), IMI (0.21), ACE (0.01) and CLO (0.24) than did our study (Zhang et al., 2019). All of these studies, however, used single-spot urine samples collected from a small sample size. None of the studies investigated temporal variability in urinary levels of neonics/metabolites within an individual.

Creatinine correction of urinary concentrations
Three different models were used to evaluate the influence of concentration corrections on the reproducibility of neonic measurements in spot urine samples. We examined betweenand within-individual variance in urinary concentrations of neonics/metabolites by three different models, namely, uncorrected concentrations (i.e., volume-based), creatininecorrected concentrations, and creatinine as a covariate ( Table 2). The highest AIC values for the multilevel random-effects models were achieved with the uncorrected concentrations, which indicated the worst fit of the models. The AIC values decreased when urine dilution was taken into account. The lowest AIC values were found for creatinine-corrected concentrations of neonics. Urine dilution explained some of the observed variations in neonic concentrations, as the model fitness improved when urine dilution was accounted for (Li et al., 2019a;Wang et al., 2016). The apportionment of between-and within-individual variances in target chemical concentrations was similar for creatinine-corrected values and when creatinine was used as a covariate. Because creatinine-corrected concentrations of neonics presented the lowest AIC values, we used creatinine-corrected values in subsequent analyses.

Intraclass correlation coefficients and reproducibility
The within-and between-individual variances in creatinine-corrected concentrations of neonics/metabolites in spot and FMV samples from the 19 individuals are shown in Table 3. Fair to good reproducibility/reliability was observed for N-DMA (ICC = 0.42) during the 44-day period. Urinary concentrations of THX, IMI, CLO, IMZ, 6-CN, and SUF showed poor reproducibility (ICCs, 0.02-0.37) in both spot and FMV urine samples. The withinindividual variance (58-98%) was much higher than the between-individual variance (2-42%). The FMV urine samples (ICCs, 0.02-0.37) did not show a better reproducibility in comparison to the spot urine samples (ICCs, 0.06-0.42).
Most of the variance in repeated measures of neonics was attributable to within-individual variability in both spot and FMV urine samples. The low ICC values of THX, IMI, CLO, IMZ, 6-CN, and SUF imply that measurements of these neonics/metabolites in a single-spot urine sample may not adequately represent exposure over a long period of time. High within-individual variability in urinary neonic concentrations is related to short half-lives of these pesticides in humans . Reproducibility of IMI and N-DMA concentrations (ICCs, 0.28-0.42) in urine was greater than that of the other target compounds (ICCs, 0.12-0.19), which could be related to the frequency of exposures and doses. IMI was the most frequently detected neonics in watersheds of both urban and agricultural areas in the USA (Hladik et al., 2018;Hladik and Kolpin, 2016). ACE, the parent compound of N-DMA, is one of the most widely used insecticides in the U.S. agriculture (Douglas and Tooker, 2015;Englert et al., 2017). The high within-individual variance in FMV urinary neonic concentrations suggests that FMV did not provide an accurate estimate of exposure for over a month. No earlier study has reported the reproducibility of neonics in urine. Our recent study on variability in temporal urinary concentrations of organophosphate, pyrethroid insecticides, and phenoxy herbicides also found considerable within-individual variability in both spot and FMV samples (Li et al., 2019a). This is probably due to the short half-lives of these pesticides, episodic nature of pesticide exposure, characteristics of the route(s) and frequency of exposure, and physiological characteristics of the biomonitoring matrix (Aylward et al., 2014;Harada et al, 2016;Osaka et al., 2016). Table 4 shows sensitivity and specificity results, to accurately categorize a high-exposure group (top 33% of 44-day average concentrations) based on one, two, or three randomly selected urine samples collected on different days. The proportion of participants who were correctly classified in the top 33% of average exposure from a single sample was 0.32-0.82 and 0.40-0.88 from spot and FMV samples, respectively. Analysis of two spot urine samples collected several days apart offered an increase in sensitivity for N-DMA (72%), whereas for other analytes, the increment was ≤12% in comparison to a single-spot urine sample. The analysis of two FMV urine samples increased the sensitivity by 6-40% for all compounds, except IMZ, in comparison to a single FMV sample. Analysis of three urine samples collected several days apart yielded moderate-to-high sensitivities in the ranges of 0.55-0.95 and 0.48-0.98 for spot and FMV samples, respectively. The specificity (0.68-0.99) was uniformly higher than the sensitivity (0.32-0.98) for both spot and FMV urine samples. This suggests that the ability to predict a false value (i.e., specificity) was higher than that of a true value (i.e., sensitivity). The minimum number of urine specimens (k) required from a single person to estimate the participant-specific mean for the 13 neonics/metabolites within 20% of the "true" values was 16-172 and 17-127 for spot and FMV samples, respectively (Table 4).

Sensitivity and specificity analysis for predicting concentrations
In our sensitivity and specificity analyses, we observed trends similar to those found for ICCs. For example, use of single-spot urine samples to classify high (top 33%) exposure from the 44-day average concentrations showed high sensitivities for IMI (0.82) and N-DMA (0.73) but low sensitivities for other compounds (0.32-0.65) in spot urine samples. This finding implies that data from a single urine sample can result in exposure misclassification of neonics/metabolites. It is worth noting, however, that sensitivity and specificity values may be overestimated due to the contribution of "predicted" values in the calculation (Li et al., 2019a;Wang et al., 2016). A fraction of the increased sensitivity and specificity observed when selecting two or more samples (per participant) instead of a single sample may be due to the increased dependence between the errors of the "true" and "predicted" values (Hauser et al., 2004). The minimum number of urine samples required to predict neonic exposures was in the order of 10s to 100s, which is comparable to that reported in our previous study for other classes of pesticides (Li et al., 2019a). Our results suggest that adequate caution should be exercised when extrapolating neonic exposures based on a single urine measurement.

Association between neonics and oxidative stress
Pearson's correlation analysis was used to examine the relationships among the urinary concentrations (log-transformed) of seven neonics (Table S7) or seven OSBs (Table S8), in spot (n = 515) and FMV (n = 243) urine samples of the 19 participants. Significant positive correlations were found between the pairs of urinary CLO vs. THX and IMI, N-DMA vs. THX and CLO, and 6-CN vs. IMZ and N-DMA (r, 0.096-0.594, p < 0.05) in both spot and FMV urine samples. Significant correlations were only found for SUF vs. THX, IMI and CLO (r, 0.205-0.266, p < 0.01) in FMV samples. These results suggest that the sources of human exposure to THX, IMI, CLO, SUF and ACE are common in the USA (Douglas and Tooker, 2015;Englert et al., 2017). It is worth noting that the correlations between urinary concentrations of THX and CLO were the strongest (r, 0.537-0.594, p < 0.01), which agrees with a recent report from China (Zhang et al., 2019). Besides direct exposure from food and water consumption (Chen et al., 2019;Klarich et al., 2017), CLO is formed from the metabolism of THX (Nauen et al., 2003;Zhang et al., 2019). Pearson's correlations suggested significant positive relationships among most pairs of the seven OSBs with the correlation coefficients ranging from 0.086 to 0.394 (Table S8). Similar results have been reported previously (Li et al., 2019a(Li et al., , 2019b. We found significant positive associations between repeated measures of neonics/ metabolites and OSBs, particularly 8-OHdG and 15-PGF 2α in spot and FMV samples collected during the 44-day study period (Table 5). Creatinine-corrected concentrations of diY, 8-OHdG, 11-PGF 2α , and 15-PGF 2α were significantly correlated with IMI, IMZ, N-DMA, and 6-CN in spot samples, whereas diY, MDA, and 15-PGF 2α concentrations were significantly correlated with IMI, IMZ, N-DMA, and 6-CN in FMV urine samples. The significant positive association between repeated measures of neonics and OSBs support the previous finding of neonics-induced oxidative damage to lipids, proteins, and DNA (Wang et al., 2018), even at the low concentrations measured in this study. The low strength of the positive associations (β, 0.03-0.20), however, may be related to the low concentrations found in urine (median concentrations: <0.5 ng/mL). A recent review showed oxidative stress induced by neonicsin non-target organisms (Wang et al., 2018). Significant negative associations between creatinine-corrected concentrations of SUF and three OSBs (i.e., diY, 11-PGF 2α and 15-PGF 2α ) were found in FMV urine samples. When the sum concentrations (log-transformed) of the seven neonics was considered for cumulative exposure assessment, a significant positive association was found with 8-OHdG, MDA, 8-PGF 2α , 11-PGF 2α , and 15-PGF 2α in either spot or FMV urine samples. This indicates concurrent effect of neonics on the oxidative damage to DNA (i.e., biomarker of 8-OHdG) and lipids (i.e., biomarkers of MDA, 8-PGF 2α , 11-PGF 2α , and 15-PGF 2α ) in humans. For the first time, we report urinary levels of neonics in repeated samples and their associations with oxidative stress in humans.

Conclusions
This study provides evidence that the US population is exposed to low levels of neonics, with the median urinary concentrations ≤0.5 ng/mL (including their metabolites). Repeated measurements of 13 neonics and their metabolites in urine displayed considerable withinindividual variability in either spot or FMV samples. Low ICC values of urinary neonic concentrations indicated poor reproducibility of overall exposures from a single measurement. To better characterize average/integrated exposure over time, collection and analysis of repeated samples are recommended. Furthermore, we found positive associations between urinary concentrations of neonics and oxidative stress biomarkers of lipid, proteins and DNA damage.
The strengths of this study lie in its large sample size per individual and measurement of 13 neonics/metabolites in two types of urine samples collected over a 44-day period. In addition, for the first time, we report temporal variability of urinary levels of neonics in repeated samples and their associations with oxidative stress using mixed regression models. This study, however, was conducted on a convenience sample of participants from a single geographic location (n = 19). The participants for this study is limited to the age range of 11 and 56 years. In addition, our study did not exclude the influence of other environmental contaminants that may induce oxidative damage to lipids, proteins, and DNA. Therefore, caution is warranted in terms of the generalizability of neonic exposures to larger populations.
c The minimum number of urine samples required to estimate participant-specific mean within 20% of the "true" value.
Environ Int. Author manuscript; available in PMC 2021 February 01.  Table 5 Mixed regression models a of log-transformed (χ + 1) creatinine-corrected concentration of urinary neonicotinoid insecticides and their metabolites and oxidative stress biomarkers.