High-Throughput Suspect Screening Reveals Widespread Pesticide Exposure in Early Childhood

Abstract

Characterization of the human exposome, the totality of exposure, provides unprecedented insights into the role of broad environmental factors in human diseases. Environmental exposures during development periods are increasingly recognized as key determinants of long-term health outcomes, yet full-scale characterization of early life exposure remains largely underexplored. In the present study, we established a nontarget data acquisition for target analysis-based suspect screening workflow capable of screening over 1000 pesticides in a 14 min run. It was next applied to the analysis of 438 urine samples collected from 187 children to comprehensively assess their pesticide exposure landscapes. A total of 215 pesticide residues were identified, with 51 detected in at least 20% of the samples. On average, 31 pesticides were identified per child, with individual exposures ranging from 21 to 50 compounds. Significant associations were observed for detection frequencies between child age, sex, and gestational age at birth and substantial pesticide co-occurrence patterns, suggesting potential coexposure risks. In total, this study constructs a high-throughput suspect screening pipeline and reveals widespread pesticide exposure in a vulnerable subpopulation, representing significant advancements toward characterizing, deciphering, and prioritizing the human exposome.

Introduction

The chemical exposome aims to capture the diversity, scope, and biological effects of various chemical exposures. Historically, exposomic studies have identified thousands of anthropogenic chemical compounds with diverse human exposures and distinct toxicological outcomes. Environmental exposures have been linked to numerous diseases and health-related conditions, − highlighting the critical role of nongenetic factors in modulating biological function. , As an important chemical class contributing to the human exposome, pesticides pose significant health risks due to their widespread application and well-documented toxicity. According to the Food and Agriculture Organization, between 1990 and 2021, the United States accounted for a substantial share of global pesticide use, including 42.9% of total pesticides, 55.9% of herbicides, 30.7% of insecticides, and 25.6% of fungicides. In California alone, 13,092 pesticide products with 1059 different active ingredients are currently registered for use. Subsequently, the ubiquitous presence of pesticide residues in human biofluids and tissues across all age groups reflects a broad population-level exposure burden. Accumulating evidence has associated chronic low-dose pesticide exposure with a range of adverse health outcomes, including cancer, diabetes, reproductive disorders, and impaired neurodevelopment. However, current methodologies often fall short in fully characterizing the chemical space in human biomonitoring (HBM) campaigns, due to selective coverage, methodological biases, and the constrained scope of authentic standards. These limitations often restrained the pesticide screening coverage, sensitivity, confidence of identification, and longitudinal comparability, thereby challenging the comprehensive characterization of pesticide exposome (i.e., the totality of human exposure to pesticides and their metabolites). ,

The early life exposome is of critical importance due to the heightened vulnerability of young populations and the potential for long-term adverse impacts on later-life health outcomes. Targeted analysis revealed that children worldwide are exposed to various classes of pesticides, including organophosphates, pyrethroids, carbamate, amide, and urea, with notable spatiotemporal and regional variations. , Epidemiological and toxicological evidence suggest associations between early life pesticide exposures and a range of adverse health outcomes, such as preterm birth, neurodevelopment disorders, cardiovascular dysfunctions, and childhood leukemia. However, the comprehensive assessment of early life pesticide exposure remains largely underexplored, particularly regarding the full-spectrum exposure profiling, demographic associations, and coexposure patterns. The vast chemical diversity and constraints in current analytical strategies continue to pose emerging challenges.

High-throughput chemical screening along with integrated biological interpretation constitutes two central objectives in exposomic research. Advances in high-resolution mass spectrometry (HRMS) have facilitated initial efforts to meet these challenges by assessing exposure to complex chemical mixtures under realistic exposure scenarios. Supported by bioinformatics algorithms and specialized databases, HRMS offers sensitive, accurate, and cost-effective identification of a wide range of compounds with high confidence and throughput. The concurrent measurements of both exogenous contaminants and endogenous metabolites provide integrated insights into the mode of action and molecular interactions of toxicants, bridging the exposome and metabolome. Nevertheless, conventional HRMS approaches face certain limitations in nontargeted analysis, including heavy reliance on authentic reference standards, the absence of streamlined and standardized analytical pipelines, and the challenge of identifying transformation products arising from environmental degradation or biological metabolism. The development of advanced HRMS-based analytical methods and the expansion of comprehensive chemical databases are urgently needed and are essential for large-scale pesticide exposure characterization, facilitating more effective chemical hazard identification and human health risk assessment.

To address these knowledge gaps, in this study, we established a nontarget data acquisition for target analysis (nDATA)-based suspect screening workflow and constructed an extensive compound database (CDB) enabling the screening of 1072 pesticides within a 14 min run, with high accuracy and robustness. nDATA employs nontargeted data acquisition strategies such as mDIA to comprehensively capture MS¹ and MS² information with minimal bias, which is subsequently analyzed using a library-based targeted suspect screening workflow, thereby enabling retrospective targeted analysis. We applied the pipeline to evaluate the full-scale pesticide exposure landscapes in early life populations by analyzing 438 urine samples from 187 young children. High-coverage compound identification and relative quantification of pesticide residues were performed, along with assessment of their associations with demographic variables and embedded coexposure patterns.

Materials and Methods

Chemicals and Reagents

Chemical standards of pesticides were kindly provided by Dr. Jon Wong from the U.S. Food and Drug Administration (FDA). A total of 1,138 pesticides were prepared as 16 mixed standard stock solutions at a concentration of 20 μg/mL individually in methanol/acetonitrile (50:50, v/v) with 44 to 80 pesticides contained in each standard mixture, respectively. Subsequently, the Pesticide Action Network (PAN; https://www.pesticideinfo.org/) database was applied to designate and link the chemical class, use type, and toxicity level for each available pesticide. Detailed information on the investigated pesticides is provided in Table S1. All analytical reagents are of high-performance liquid chromatograph grade or higher unless otherwise stated.

Human Urine Collection

Urine samples were collected as part of the UNC Baby Connectome Project (BCP), which is an ongoing longitudinal study of early life brain development. Eligible children had a term delivery (38 ± 4 weeks) of normal birthweight (3,500 ± 1,000 g), no significant complications during pregnancy (e.g., eclampsia, placental abruption, fetal distress, etc.) or delivery (e.g., dystocia, umbilical cord prolapse, neonatal asphyxia, etc.), and no first-degree relatives with a history of autism, intellectual disability, schizophrenia, or bipolar disorder. Urine was chosen as the primary biospecimen due to the noninvasive collection and its ability to reflect a broad spectrum of xenobiotic exposures. Spot urine samples were collected from infants and toddlers aged 10 days to 7 years using either hospital-grade cotton balls placed in a new diaper or a sterile toilet hat and further transferred into sterile polypropylene collection cups. In total, 438 urine samples were obtained from 187 children with one to seven samples collected per individual at different visits (Table S2). Urine was kept at 4 °C for up to 24 h until processing and then aliquoted and stored at −80 °C to preserve sample stability and pesticide integrity. Urine samples were collected between June 2017 and March 2020. Participant characteristics including age of children at sample collection, sex, ethnicity, race, birthweight, gestational age at birth, maternal race, maternal ethnicity, and maternal education as well as total annual income were collected at study visits (Table S3). This study was approved by the UNC institutional review board (IRB 21-1923).

Urine Sample Preparation

Sample preparation was performed following protocols in our previous study. Briefly, 0.2 mL of each collected urine was adjusted to pH 6.5 with sodium acetate buffer (1 mol/L, pH 5.0) and incubated with 1 mL of β-glucuronidase (1,000 units/mL in sodium acetate buffer) at 37 °C for 2 h to remove glucuronic acid conjugates. β-Glucuronidase deconjugation ensured compatibility with existing compound databases and minimized potential underestimation of internal exposure. An ISOLUTE C18 solid phase extraction cartridge (100 mg) was prewashed with 2 mL of 5% NH₃·H₂O in methanol (MeOH) and 2 mL of sodium acetate buffer. Following the loading of urine samples, the column was washed with 3 mL of 5% MeOH in sodium acetate buffer and subsequently eluted by adding 3 mL of 5% NH₃·H₂O in MeOH. The addition of NH₃·H₂O facilitates the recovery of acidic as well as certain polar compounds. The extracts were evaporated to dryness and then dissolved in 50 μL of ACN/water (20:80, v/v). The reconstituted samples were stored in a −80 °C fridge until further instrumental analysis.

Instrumental Analysis

The workflow was operated on a Thermo Fisher Scientific Vanquish Ultra High Performance Liquid Chromatography (UHPLC) system coupled to a Q Exactive mass spectrometer, following column, mobile phase gradient, and instrument parameters adapted from previous research. Only the HESI positive mode was used. A Thermo Hypersil GOLD C18 HPLC Column (100 × 2.1 mm, 1.9 μm) was applied with mobile phases consisting of 4 mmol/L ammonium formate with 0.1% formic acid in water (A) and 4 mmol/L ammonium formate with 0.1% formic acid in MeOH (B), flowing at a rate of 0.3 mL/min under 40 °C with a 14 min gradient: 12% B, 0 min; 12–95% B, 0–8 min; 95–100% B, 8–9 min; 100% B, 9–11 min; 100–12% B, 11–11.1 min; 12% B, 11.1–14 min. Five injections of background blank samples (ACN/water, 20:80, v/v) were performed prior to each batch analysis of urine samples to monitor and ensure no contamination was present originating from industrial usage such as an LC-HRMS system, sample tube, mobile phase, LC column, etc. For the analysis, a pooled mix quality control sample (i.e., combining equal aliquots from all extracted urine samples to provide a generalized representation of overall exposure patterns and serving as a quality control for analytical reproducibility), a method blank sample (i.e., water substituted for urine and processed in parallel with urine samples), and a 10 ng/mL standard sample were injected every 12 injections. Six isotopic-labeled pesticide residues (Carbendazim-d ₄, Metoxuron-d ₆, Carbofuran-d ₃, Diuron-d ₆, Metolachlor-d ₆, Picolinafen-d ₄) along with all unlabeled analytical standards were contained in the standard control samples. Due to the wide coverage and structural diversity of pesticides studied, isotope-labeled internal standards were spiked in standard controls instead of each sample, serving to monitor instrumental stability and ensure data consistency for relative quantification analysis. For the CDB establishment, MS¹ (m/z, isotopic pattern, and RT) and MS² information (fragmentation-derived product ions) was obtained through 16 separate full-MS/PRM runs using pesticide standard mixtures. For urine analysis, a full MS-coupled multiplexing data-independent acquisition (mDIA) was applied (Figure , Text S1). This method enabled the simultaneous collection of both MS¹ (m/z, isotopic pattern, and RT) and MS² (fragmentation pattern) information in a single 14 min run, which is further utilized for building the pesticide CDB and suspect screening in human urine using TraceFinder 3.3 software (ThermoFisher Scientific, Germany). The parameter settings for suspect screening include a mass accuracy of ±5 ppm for both precursor and fragment ions, a signal-to-noise ratio of 3, an RT window of 60 s, matched isotopic patterns, and at least one fragment ion detected. To further validate accuracy and robustness of the mDIA result, parallel reaction monitoring (PRM) was performed (Text S1) to manually verify the identity of pesticides with high or moderate detection rates (>20%) (Figure ). Pesticide annotations with a spectral similarity greater than 0.5 were assigned a confidence level of 1. SPE recovery was assessed using fortified urine with pre/postspiked standard mixtures. Matrix effects were assessed by comparing the mean peak areas of SPE-treated fortified urine samples (spiked postextraction at 10 ppb) with those of solvent standard samples. Retention time (RT) shifts were evaluated using fortified urine and solvent blank samples spiked with standard mixtures.

1.

Schematic framework of the nDATA workflow to achieve high-throughput pesticide suspect screening in the urine of infants and toddlers.

Statistical Analysis

For statistical analysis, an initial raw intensity matrix for the measured pesticides was obtained through MS¹ signal area integration. Blank subtraction was performed by subtracting the average background signal intensities which were consistently detected in method blank samples. Following prefiltering and deduplication, the data was consolidated with demographic information using unique sample IDs. Qualitative (a 0/1 value was assigned for detection/no detection) (Table S4) and relative quantitative (integrated area for mass spectrometry signals) (Table S5) assessments were performed separately. For qualitative analysis, a two-sided Chi-squared test was performed (SPSS ver. 30.0.) to examine whether two categorical variables are correlated (e.g., sex vs pesticide qualitative results), and a Kruskal–Wallis test (one-way ANOVA on ranks) was used to assess whether there are significant differences between the medians of two or more independent groups on a continuous variable (e.g., age vs pesticide qualitative outcomes). For relative quantitative analysis, ordinary least-squares linear regression analysis was applied to measure the linear relationship between two quantitative variables (e.g., age vs pesticide quantitative results), and the Spearman correlation coefficient was calculated to investigate pesticide pairwise coexposure patterns. P values from multiple tests were adjusted for false discovery rate (FDR) using the Benjamini–Hochberg (BH) method. A significance threshold of p _adj < 0.05 was applied for all statistical tests. Missing values were treated as nondetects and replaced with zero intensity to allow comparability across samples without introducing artificial bias from imputation during relative quantification analysis. Standard descriptive statistics (frequency and mean ± standard deviation) were calculated for all other variables.

Results

In-House Pesticide Database

To provide comprehensive insight into characterizing and deciphering early life pesticide exposure, we constructed a comprehensive in-house pesticide database (Figure ). The pesticides involved were selected based on their widespread agricultural use, high relevance to human exposure, prior biomonitoring evidence, recognized toxicity concerns, and availability of standard chemicals. A total of 1,072 pesticides were investigated in this study, representing, to the best of our knowledge, one of the largest-scale pesticide CDB in HBM campaigns. The constructed CDB contained information about two domains: agricultural applications and LC-MS analysis. In terms of agricultural applications, we assigned chemical class, use type, and toxicity levels to each available pesticide, with interconnected linkages established (Figure a). Analysis of the Sankey plot clearly revealed that most highly hazardous pesticides belonged to the organophosphates, which are widely recognized as contributors to multiple adverse health outcomes including oxidative stress, immunotoxicity, endocrine disruption, neurodegenerative effects, and cancer. Urea pesticides were primarily utilized as herbicides, displaying relatively low toxicity risks (e.g., cycluron, fenuron, and metobromuron). Urea herbicides function by inhibiting acetolactate synthase in plants rather than mammals. Still, it is crucial to include urea herbicides in our study due to their high production volume and widespread application, along with their potential antiandrogenic and neurotoxic effects. Additionally, although xenobiotic metabolism is typically considered a detoxification process, bioactivation-induced toxicities have been frequently identified with emerging research interest. − It was noteworthy that the toxicity of pesticide metabolites remains largely unknown (e.g., acetamiprid-metabolite-IM-1-2, amitraz metabolite DMF, and metolachlor oxanilic acid). Given their short biological half-life and phenotype-related properties, the significance of pesticide metabolites cannot be overlooked in evaluating human pesticide exposome. The established high-coverage, generalized CDB includes pesticides with varying solubility, which is expected to be used not only for urine screening but also for other applications, such as xenobiotic profiling in food, soil, indoor dust, etc., contributing to a comprehensive understanding of human exposome landscapes. Therefore, including low-solubility pesticides with potential exposure and health risks is highly practical and human relevant.

2.

Detailed information on the self-constructed pesticide database. (a) Interrelation of toxicity, chemical class, and use of the pesticides investigated in this study. (b) Distribution of pesticide m/z with median value indicated by the dotted line. (c) Distribution of pesticide RT with the median value indicated by the dotted line.

For the established pesticide CDB, a total of 1,072 pesticides were manually assigned with m/z and RT, with collected nonprecursor fragments ranging from 1 to 5. 66 of the 1,138 pesticides included in standard stock solutions were excluded due to incompatibility with the current analytical conditions. The distribution of m/z and RT is presented in Figure b,c, revealing a median m/z of 309.0556 and a median RT of 6.97 min, indicating adequate chromatographic retention and reliable mass detection performance. To validate the accuracy of the generated pesticide CDB and evaluate the sensitivity and selectivity of the nDATA workflow, 1 μg/mL and 10 ng/mL total standard mixtures were tested, resulting in true discovery rates of 89.5% (±1.2%) and 77.8% (±2.6%), respectively (Table S6). The method blank sample was tested following the same analytical workflow, yielding a false discovery rate of 2.3% (±0.4%). Additionally, an average SPE extraction recovery of 51.89% and a relative standard deviation (RSD) of 12.96% were achieved (Table S7). The average matrix effect was determined to be 85.42%, indicating minimal ion suppression or enhancement (Table S7). Besides, an average RT shift of 0.12 s and a standard deviation of 0.102 s were observed (Table S8). These results indicated high RT consistency and stability across the data set with minimal matrix effect, supporting reliable compound annotations and true positive identifications. While the 14 min LC-MS runtime indicates high screening throughput, subsequent data processing (e.g., peak picking, database matching, blank subtraction, association analysis) remained the most time-consuming step of our workflow. Given the 12 loop counts in the established mDIA method, the number of detected data points per chromatographic peak is limited; still, for an average fwhm of 15 s, at least 6–7 points are typically collected, which is generally sufficient for reliable peak picking, database matching, and relative quantification. Compared with conventional suspect screening methods that typically trigger MS/MS acquisition only for predefined precursor ions, the use of DIA enabled comprehensive collection of fragmentation information without biased precursor selection while maintaining sensitivity and selectivity, allowing retrospective and flexible targeted analysis. Compared with data-dependent acquisition (DDA), which often produces incomplete and abundance-biased fragmentation spectra, DIA provides more consistent MS/MS information for low-abundance compounds, making it well-suited for large-scale, trace-level pesticide screening. In addition, DIA offers greater analytical reproducibility by systematically acquiring fragmentation data across predefined m/z windows in each run. This strategy provided high-confidence identifications while retaining broad compound coverage, forming the basis of the high-throughput DIA-driven suspect screening approach. Besides, the established pesticide CDB includes compounds with limited exposure information or previously unreported spectra, thereby expanding the current resources available for future pesticide exposure analysis. In total, we established a highly comprehensive pesticide database incorporating reliable liquid chromatography and mass spectrometry information. The robust and validated nDATA pipeline enabled us to screen over a thousand pesticides in a simple 14 min run, representing one of the largest-scale characterizations of pesticide exposure in early life populations.

Pesticide Screening in the Urine of Children

Further, we applied the pesticide CDB and established an nDATA workflow to 438 urine samples collected from 187 infants and toddlers to assess their pesticide exposure landscapes (Figure ). A total of 215 pesticides were identified based on the aforementioned criteria, corresponding to 160 distinct active substances (Figure S1a). On average, each subject exhibited 31 (±5) identifications, with detection coverage ranging from 21 to 50 pesticides (Figure S1b). Based on the detection frequency, pesticides were categorized into three groups: high detection (>60%), moderate detection (20–60%) and low detection (<20%). Sixteen pesticides were identified as highly detected ones, whereas 35 and 164 were classified as moderate and low detection, respectively. The distribution of m/z and RT is depicted in Figure a with varied chromatographic behaviors potentially attributed to the pesticide-specific polarity and metabolic efficiency. Regarding the agricultural uses of the detected pesticides, the categories were as follows 40% as metabolites, 22.8% as insecticides, 13.5% as herbicides, 13% as fungicides, 6% as other, and 4.7% as unclassified (Figure b). The large proportion taken up by pesticide metabolites highlighted the importance of assessing xenobiotic metabolism in characterizing the human exposome. In terms of chemical classes, the most highly detected categories were carbamates (15.8%) and organophosphates (8.4%), which is consistent with previous targeted research suggesting these as two major pesticide classes detected in children from agricultural communities. ,, Figure c shows the detection frequency and intensity of the top 51 most frequently detected pesticides (i.e., pesticides in both the high and moderate detection groups). The intensity distribution exhibited substantial variations (Figure c), suggesting that early life pesticide exposure is highly dynamic with spatiotemporal and individual determinants. For QA/QC, peak intensities and retention time deviations of six representative pesticides in pooled QC samples were evaluated, showing that most RT deviations were within ±0.1 min, and intensity RSDs ranged from 8% to 14% (Figure S2), indicating promising analytical consistency. Although low to moderate intrabatch variations were observed for some pesticide intensities, potentially attributed to factors such as low signal intensity and poor peak integration, these fluctuations remained within acceptable limits and hardly compromised overall analytical reliability.

3.

Pesticide screening in the urine of young children. (a) The distribution of m/z and RT of the detected 215 pesticides in human children urine. (b) Agricultural use and chemical class of the identified 215 pesticides. (c) Detection frequency and intensity of the 51 most frequently detected pesticides (detection frequency ≥ 20%).

To validate the findings from pesticide screening, two separate PRM runs were applied to pooled QC samples to verify the identified pesticides with moderate and high detection frequencies (Figure ). N,N-Diethyl-meta-toluamide (DEET), one of the most effective and commonly used insect repellents, was identified with a detection frequency of 95.4% (Figure b–d). Extensive HBM studies have confirmed the presence of DEET and its metabolites in populations, − while toxicological evidence indicated a relatively low human-relevant toxicity associated with DEET itself. − Still, given DEET is rapidly degraded into its oxidative metabolites, this notably high detection frequency of the parent compound in children urine raises safety concerns about its metabolites’ effects and potential synergistic toxicity. Glyodin, a fungicide used to protect foliage from general fungal attack, was detected at a frequency of 71.5% (Figure e–g). First introduced as a pesticide active ingredient in 1950s–1980s, glyodin was reported with a concentration of 148 ng/L in surface water samples from the Potomac River watershed, United States; however, little is known about its contemporary human exposure and other environmental distributions. MGK-11 (Figure h–j) and verbenone (Figure k–m), two insect repellents with very limited human exposure-related knowledge, were detected with detection frequencies of 57.1% and 77.2%, respectively. Notably, in this study, spot urine samples were analyzed as they are easy to obtain and thus feasible for large-scale exposome characterization. While they provide practical advantages, spot urine samples are subject to substantial temporal variability from short-term fluctuations in diet, activity, and hydration. In contrast, 24-h urine provides a more generalized and average measure of daily exposure, while its collections are often costly, labor intensive, and also realistically impractical in large-scale infant and toddler populations. Future studies are encouraged to combine spot urine with 24-h samples to better capture temporal exposure patterns, thereby improving the comprehensive characterization of the early life exposure. In summary, our nDATA workflow uncovered novel pesticides with population-wide exposure of high analytical confidence. Considering the vulnerability of developing children, further risk assessments and hazard identification work of these widespread xenobiotics are warranted to better understand their health impact on humans.

4.

MS/MS validation of the identified pesticides. (a) Pesticide-wide distribution of the identified 215 pesticides. Pesticides with low, moderate, and high detection frequency were marked as blue, orange, and red, respectively. Liquid chromatogram and MS² spectra compared with standard chemicals, mDIA spectra, and proposed fragmentation mechanisms were presented for four pesticides individually: (b–d) DEET, (e–g) glyodin, (h–j) MGK-11, and (k–m) verbenone.

Qualitative Correlational Analysis between the Pesticide Exposure and Participant Characteristics

To deepen our understanding of the relationship between pesticide exposures and participant demographic features, qualitative correlation statistical analysis was performed (Figure a). The significance matrix showed BH-adjusted significance for pesticides detected with high and moderate detection frequencies, with characteristic clustering patterns established. Notably, age emerged as the variable with the most substantial and significant relationships (33.3%) (Figure b,c). The pivotal role of age in the pesticide exposure of children is unsurprising, given existing studies demonstrating clear links between age and factors such as toxicokinetic properties, constant ingestion, and unhygienic habits (for example, hand-to-mouth activity). Seventeen pesticides were identified with age-dependent patterns, indicating a dynamic and individualized risk map of pesticide exposure across different stages of young children development. Specifically, the observed significantly negative correlation between DEET and age (Figure b) was consistent with previous studies, indicating that younger children are subjected to heavier dermal uptake of DEET. Similarly, propoxur was negatively correlated with gestational age at birth (Figure d), indicating a potential relevance to preterm birth. Besides, several pesticides such as Acetamiprid-metabolite-IM-2-1 exhibited sex-specific exposure patterns (Figure e). The observed sex-related differences may be explained by variations in exposure patterns, lifestyle factors, and sex-specific physiology and metabolism, as previously reported. − Other factors including ethnicity, race, maternal ethnicity, and maternal race also demonstrated significant correlations with varying statistical significance (Figure f–i), suggesting a subtle correlation between complex lifestyle and dynamic pesticide exposure patterns. In total, our qualitative analysis between the early life pesticide exposure and demographic variables revealed unique correlation patterns.

5.

Qualitative correlation analysis of early life pesticide exposure in relation to demographic variables. (a) Significance matrix for fifty-one qualitative detected pesticides across ten demographic variables. Columns in the upper panel represented the detection frequency of each pesticide; grids in the lower panel displayed the Benjamini-Hochberg adjusted p values derived from a two-sided Chi-squared test (for categorical variables) or Kruskal–Wallis test (for continuous variables). Kruskal–Wallis test outcomes of (b, c) age; (d) gestational age at birth; Chi-squared test results of (e) sex, (f) ethnicity, (g) maternal ethnicity, (h) race, and (i) maternal race.

Relative Quantitative Analysis and Coexposure Pesticide Exposure Landscapes

Next, an in-depth assessment of the relative quantitation outcomes and potential applications derived from the nDATA workflow was conducted. Originally devised as a qualitative screening strategy, our method was restrained with limited scan rate of the mDIA-coupled full scan, as well as the pesticide-specific distinct ionization efficiency and matrix effect. Absolute quantification of each detected pesticide proved challenging; ,, nonetheless, the application of relative quantification offered an appealing description of the characterized pesticide exposome. The evaluation began with the six serial-diluted isotope-labeled pesticides, demonstrating high linearity and fitting goodness (R ²: 0.994–0.999) (Figure S3). Subsequently, the quantification outcomes between precursors and their fragments were investigated, using indole-3-acetic acid (IAA) as an example due to its high detection frequency (Figure b). IAA is both an endogenous metabolite and a xenobiotic plant growth regulator, making its sources and exposure contributions indistinguishable. Both product ions m/z 130.0654 (R ² = 0.84) and m/z 103.0547 (R ² = 0.73) were positively correlated with precursor m/z 176.0706. The proportional, concentration-dependent associations between precursor and fragment ions suggest that fragment signals not only confirm compound identification but also support reproducible and robust relative quantification outcomes, minimizing the risk of false positives from stochastic fragmentation. Additionally, the relative quantification outcomes suggested significant correlations with age: Amitraz metabolite DMF (R _spearman = 0.37, 95% CI: (0.24, 0.48)) and DEET (R _spearman = −0.44, 95% CI: (−0.51, −0.36)) either showed relatively modest positive or subtle negative correlations with age, respectively (Figure c,d). The average RSD for the standard compounds across the analytical sequence was 21.27%, which is typical for high-throughput human biomonitoring studies and generally acceptable for relative quantification purposes (Table S9).

6.

Quantitative Spearman correlation analysis of pesticide exposure outcomes. (a) Spearman correlation coefficient matrix of pesticide co-occurrence pattern. Legend marks the Spearman correlation coefficient with Benjamini-Hochberg adjusted p values labeled (“*” for p < 0.05; “**” for p < 0.01; “***” for p < 0.001, respectively). (b) Linear regression analysis on the precursor-fragment relationship for indole-3-acetic acid. (c) Linear regression analysis between Amitraz metabolite DMF with age. (d) Linear regression analysis between DEET with age. (e) Linear regression analysis between Tepraloxydim metabolite 620M075 and Tepraloxydim metabolite BH 620-DML. (f) Linear regression analysis between Aminocarb and Metolcarb. (g) Linear regression analysis between MGK-11 with Carbofuran phenol.

With the verified quantification capacity of the nDATA method, we redirected our attention on the pesticide exposure correlations, revealing intriguing co-occurrence/coexposure landscapes (Figure a). Three major perspectives were summarized to explain the observed coexposure features. First, the presence of conjugated metabolites, where correlated pesticides share a common parent compound, was explored (Figure e). Tepraloxydim metabolite 620M075 and Tepraloxydim metabolite BH 620-DML are both phase I biotransformed products from Tepraloxydim, only distinguished by a distinct unsaturated bond. Their positive correlation (R _spearman = 0.65, 95% CI: (0.57, 0.71)) suggested an inherent linkage from biological activities (i.e., metabolic biodegradation). The role of chemical class was also highlighted in the context of coexposure, as illustrated in Figure f. Aminocarb and Metolcarb, both belonging to the N-Methyl Carbamate class of insecticides, have been concurrently detected in American field workers. The observed positive correlation (R _spearman = 0.61, 95% CI: (0.49, 0.71)) indicates a potential connection regarding their structurally dependent absorption, distribution, metabolism, and excretion properties. Additionally, pesticide use-type emerged as another crucial factor in coexposure landscape analysis (Figure g). MGK-11 and Carbofuran Phenol, representing two distinct insecticides, had a poorly documented coapplication history. The statistically significant association (R _spearman = 0.46, 95% CI: (0.34, 0.56)) we observed implies potential coexistence in products where they served as active ingredients. In summary, quantitative exploration facilitated the discovery of novel pesticide coexposure landscapes. These observations may further contribute to a comprehensive characterization of coexposure patterns, offering valuable insights into the mapping of the early life human pesticide exposure.

In summary, in this study, we established a HRMS-based nDATA workflow which facilitated high-throughput screening of 1,072 pesticide residues in a 14 min run, with high confidence in both MS¹ and MS² information identified (level 1). Building on prior developments, we expanded the analyte scope, optimized the pipeline in human urine context, and extrapolated it to an early life cohort, enabling the large-scale suspect screening and revealing the widespread presence of 215 pesticide residues. We identified correlations between specific pesticide exposures and demographic variables and also observed unique coexposure patterns. The observed co-occurrence patterns may reflect common exposure sources and behavioral factors, such as dietary intake and residential pesticide use and environmental background exposure. For instance, coexposure patterns involving organophosphates, pyrethroids, and neonicotinoids may be associated with food consumption and household pest control, whereas clusters dominated by herbicides such as glyphosate and atrazine may reflect lawn care practices and agricultural activities. These observations are broadly consistent with previous studies, highlighting the importance of investigating co-occurrence in exposome characterization and prioritization. To our knowledge, these data sets represent one of the largest pesticide exposure characterizations, contributing valuable insights into the understanding of early life pesticide exposome.

Limitations of this study are further discussed as follows. For the sample preparation, since the current extraction protocol utilizes a polar solvent (95% methanol) with a C18 column, which enables broad-spectrum pesticide analysis, it may underperform for extremely nonpolar or strongly conjugated compounds. Additional sample preparation (e.g., alternative sorbents, pH adjustments), separation (e.g., hydrophilic interaction liquid chromatography), and analytical strategies (e.g., gas chromatography) may aid in expanding the chemical space coverage of exposome studies in the future. A moderate true discovery rate of 77.8% was achieved at 10 ng/mL, which was relatively low yet comparable with other suspect screen analysis. Several factors can limit the scope and sensitivity/specificity of the current analysis, such as high coverage of the pesticide database, high complexity of the suspect screen workflow, and intrinsic chemical properties (e.g., low ionization efficiency, difficulty for fragmentation, and degradation to in-source fragment). For DIA analysis, cofragmentation within wide isolation windows and reliance on computational deconvolution may complicate spectral interpretation and increase the risk of misassignment in complex biological matrices. This study focused on qualification/relative quantification analysis for broad-spectrum pesticide screening, which precluded absolute quantification using internal standards or calibration curves and subsequent risk assessments, such as daily intake estimation and hazard quotient calculations. The reported associations were derived from detection frequencies and relative signal intensities and therefore should be interpreted conservatively as exploratory due to the absence of absolute quantification. Collecting repeated urine samples from the same child may introduce within-child variability, which was not fully explored in the present work. While urine samples were treated with β-glucuronidase, the absence of arylsulfatase may have limited detection of sulfate conjugates from an alternative Phase II pathway, and investigations into the pesticide phase I metabolites were largely neglected. Due to the limited availability of reference standards and lack of well-curated spectral libraries for pesticide metabolites, our analysis was largely restricted to parent compounds, which may lead to an underestimation of total exposure given their rapid metabolism and excretion; future studies are needed to more comprehensively characterize their metabolites to better reflect actual internal exposure patterns. In total, this study establishes a high-throughput pesticide screening workflow and reveals widespread pesticide exposure landscapes in a vulnerable population, which makes important advancements toward characterizing, deciphering, and prioritizing the early life exposome.

All data are available in the main text or the Supporting Information.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/envhealth.5c00684.

• Detailed information on instrumental analysis (Text S1); Pesticide-wide exposome screening in children’s urine (Figure S1); QA/QC evaluation of retention time (RT) and signal intensity stability across different analytical batches (Figure S2); Quantitative assessment of nDATA workflow for 6 isotope-labeled pesticide standards (Figure S3) (PDF)

• In-house pesticide database (Table S1); Demographic information summary of urine-collected children and their mothers involved in this study (Table S2); Urine sample demographic information (Table S3); Qualitative pesticide screening result (Table S4); Relative quantitative pesticide screening result (Table S5); Screening of pesticides at 10-ppb level (Table S6); Assessments of SPE recovery, matrix effect, and process efficiency (Table S7); RT shift assessment (Table S8); Standard repeatability assessment (Table S9) (XLSX)

Conceptualization: Haoduo Zhao, Stephanie M. Engel, Kun Lu. Methodology: Haoduo Zhao, Yun-Chung Hsiao, Chih-Wei Liu, Emily Werder, Jake Thistle. Visualization: Haoduo Zhao. Supervision: Stephanie M. Engel, Kun Lu. Writingoriginal draft: Haoduo Zhao. Writingreview and editing: Stephanie M. Engel, Kun Lu, Yifei Yang, Jiahao Feng, Xueying Wang, Jingya Peng, Taylor Teitelbaum.

UNC Superfund Research Program (P42ES031007); UNC Center for Environmental Health and Susceptibility grant (P30-ES779 010126); EPA STAR (R840219); NIEHS RO1 grant (ES033518).

The study was approved by the Institutional Review Boards at the University of North Carolina at Chapel Hill (IRB 21-1923). This study was conducted in full compliance with the ethical principles outlined in the Declaration of Helsinki, ensuring adherence to standards for research involving human subjects. Informed consent was obtained from all parents of participants prior to their involvement in the study.

The authors declare no competing financial interest.

Published as part of Environment & Health special issue “Transforming Environmental Exposure and Health Research through Emerging Technologies”.