Diurnal variability in urinary volatile organic compound metabolites and its association with oxidative stress biomarkers

Abstract

Volatile organic compounds (VOCs) are ubiquitous environmental pollutants that are associated with birth defects, leukemia, neurocognitive deficits, reproductive impairment and cancer in humans exposed to these compounds. Exposure to VOCs can be assessed by measuring their metabolites in urine. Little is known, however, about the temporal variability in urinary VOC metabolite (VOCM) concentrations within and between individuals. In this study, we determined the variability in the concentrations of 38 VOCMs in urine samples collected from 19 healthy individuals across a period of 44 days. We also measured seven biomarkers of oxidative stress (lipid, protein and DNA damage) in urine to assess the relationship of VOC exposure to oxidative stress. Seventeen VOCMs had detection frequencies (DFs) of >60% in urine, and we limited further data analysis to those compounds. The creatinine-adjusted geometric mean concentrations of VOCMs ranged from 2.70 μg/g to 327 μg/g in spot and 2.60 μg/g to 551 μg/g in first morning void (FMV) urine samples. Calculation of the intra-class correlation coefficients (ICCs) for 17 VOCM concentrations to assess their predictability and repeatability in urinary measurements showed ranges of 0.080–0.425 in spot and 0.050–0.749 in FMV urine samples, revealing notable within-individual variability. Our results suggest that taking only single measurements of VOCM concentrations in urine in epidemiological investigations may lead to exposure misclassification. In addition, VOCM concentrations were significantly and positively correlated with oxidative stress biomarkers. This study thus provides important information for formulating sampling strategies in the biomonitoring of VOC exposure in human populations.

Graphical Abstract

1. Introduction

Volatile organic compounds (VOCs) are organic chemicals that vaporize at ambient temperatures. Human exposure to VOCs is ubiquitous, with sources that vary widely from consumer products and industrial solvents to tobacco smoke and biomass burning (Li et al., 2021). The United States Environmental Protection Agency (EPA) categorizes VOCs such as 1,3-butadiene, propylene oxide, toluene, styrene, xylene, trichloroethylene, tetrachloroethylene, acrolein, acrylonitrile, acrylamide, carbon disulfide, 1-bromopropane, ethylbenzene, ethylene oxide and vinyl chloride as hazardous air pollutants (EPA, 2017). The International Agency for Research on Cancer (IARC) classifies benzene, trichloroethylene and 1,3-butadiene as known carcinogens; acrylamide, ethylbenzene, styrene and isoprene as probable carcinogens; and toluene, acrolein, crotonaldehyde and xylene as group 3 carcinogens (compounds whose carcinogenicity is unclassifiable due to lack of data) (ATSDR, 2019; Li et al., 2021; WHO, 2010). Human exposure to VOCs is associated with birth defects, leukemia, neurocognitive deficits, reproductive impairment and cancer (Li et al., 2021; McGraw et al., 2021; Jain, 2015a).

VOC exposure occurs through multiple routes, including ingestion, inhalation and dermal contact (Li et al., 2021). In humans, VOCs are then biotransformed by hepatic cytochrome P450 enzymes into water-soluble metabolites and excreted in urine. Thus, urinary VOCMs may be used as biomarkers of exposure (Alwis et al., 2012). The biological half-lives of VOCs are on the order of several hours (Li et al., 2021), and it is not known how well the measurement of VOCMs in urine collected at a single time point reflects integrated exposure to VOCs over time. Moreover, within- and between-individual variability in VOC exposures can arise from differences in environmental (traffic, residential location, diet, tobacco smoke, consumer product usage and occupation) or genetic factors (metabolic potential) (Siroux et al., 2016). In recent studies, we described the extent of within- and between-individual variability in urinary levels of polycyclic aromatic hydrocarbons, neonicotinoids, melamine and cyanuric acid (Zhu et al., 2021; Li et al., 2020, 2019; Zhu and Kannan, 2019), and similar studies of VOCs are needed to assess the suitability of using a single spot urine test in assessing VOC exposures.

Exposure to VOCs can result in the production of reactive oxygen species (ROS), which in turn can elicit oxidative damage to biological macromolecules such as lipids, proteins and DNA (Li et al., 2020). Oxidative stress can disrupt normal cellular signaling and therefore can be assessed through the analysis of oxidative stress biomarkers (OSBs) excreted in urine. Previous studies have examined the association between oxidative stress and VOC exposure (Kuang et al., 2021; Xu et al., 2021; Kwon et al., 2018) for a limited number of OSBs and VOCMs. In this study, we investigated within- and between-individual variability in VOCM concentrations in 515 urine samples collected from 19 healthy individuals across a period of 44 days. We determined 38 VOCMs and 7 OSBs in urine samples and examined the associations between VOCM concentrations and OSBs over time.

2. Materials and Methods

2.1. Participant characteristics and sample collection

Urine samples were collected from 19 healthy volunteers residing in Albany, New York, USA, across a period of 44 days from February to April 2018, as a part of a study that examined temporal variability in oxidative stress biomarkers (Martinez-Moral and Kannan, 2019). A total of 515 spot urine samples were collected which included 243 FMV urine samples. The samples were collected from non-smoking males (n = 11) and females (n = 8). None of the participants had smokers in the household and therefore, secondhand exposure to tobacco constituents was small. For FMV urine samples, volunteers provided the early morning void. However, due to logistical reasons, some volunteers did not provide FMV urine for all 44 days. Spot urine samples included those that were collected during anytime of the day. Available information on the participant characteristics is presented in the Supplementary Information (Table S1). Information regarding age, gender, ethnicity and BMI was collected at the beginning of the study. Other lifestyle parameters such as dietary supplement intake, medication intake, international travel and exercise frequency were self-reported and collected across the sampling period. Participants were given 50-mL polypropylene tubes for urine collection. After collection, urine samples were placed at −20 °C within 2 h for storage. Date and time of urine collection were recorded. The Institutional Review Board of the New York State Department of Health approved the analysis of urine samples.

2.2. Analytical standards

Thirty-eight native standards were used. N-Acetyl-S-(2-carboxyethyl)-L-cysteine, CEMA; N-acetyl-S-(2-hydroxypropyl)-L-cysteine, 2HPMA; N-acetyl-S-(benzyl)-L-cysteine, SBMA; N-acetyl-S-(2-carbamoylethyl)-L-cysteine-sulfoxide, AAMA Sul; N-acetyl-S-(2-carbamoyl-2-hydroxyethyl)-L-cysteine, GAMA; N-acetyl-S-(2-cyanoethyl)-L-cysteine, CYMA; N-acetyl-S-(1-cyano-2-hydroxyethyl)-L-cysteine, CHEMA; N-acetyl-S-(2-hydroxyethyl)-L-cysteine, HEMA; N-acetyl-S-(1,2-dichlorovinyl)-L-cysteine, 1,2DCVMA; N-acetyl-S-(2,2-dichlorovinyl)-L-cysteine, 2,2DCVMA; N-acetyl-S-(2,4-dimethylphenyl)-L-cysteine, 2,4DPMA; N-acetyl-S-(2,5-dimethylphenyl)-L-cysteine, 2,5DPMA; N-acetyl-S-(3,4-dimethylphenyl)-L-cysteine, 3,4DPMA; N-acetyl-S-(3,4-dihydroxybutyl)-L-cysteine, DHBMA; (R,S)-N-acetyl-S-(1-hydroxymethyl-2-propenyl)-L-cysteine + (R,S)-N-acetyl-S-(2-hydroxy-3-butenyl)-L-cysteine, MHBMA1+2; N-acetyl-S-(4-hydroxy-2-butenyl)-L-cysteine, MHBMA3; N-acetyl-S-(3-carboxy-2-propyl)-L-cysteine, CMEMA; 2-aminothiazoline-4-carboxylic acid, ATCA; N-acetyl-S-(N-methylcarbamoyl)-L-cysteine, AMCC; N-acetyl-S-(1-phenyl-2-hydroxyethyl)-L-cysteine + N-acetyl-S-(2-phenyl-2-hydroxyethyl)-L-cysteine, 1,2PHEMA+2,2PHEMA; N-acetyl-S-(phenyl)-L-cysteine, SPMA; trans,trans-muconic acid, MU; N-acetyl-S-(n-propyl)-L-cysteine, BPMA; 2-thioxothiazolidine-4-carboxylic acid, TTCA; N-acetyl-S-(trichlorovinyl)-L-cysteine, TCVMA; N-acetyl-S-ethyl-L-cysteine, EMA; N-acetyl-S-methyl-L-cysteine, MMA; N-acetyl-S-(4-nitrophenyl)mercapturic acid, NANPC; N-acetyl-S-(2-hydroxy-3-methyl-3-buten-1-yl)-L-cysteine + N-acetyl-S-(1-hydroxymethyl)-2-methyl-2-propen-1-yl)-L-cysteine, IPM1; N-acetyl-S-(4-hydroxy)-2-methyl-2-trans-buten-1-yl)-L-cysteine, IPM3; and N-acetyl-S-(3-hydroxypropyl-1-methyl)-L-cysteine, HPMMA, were purchased from Toronto Research Chemicals (Toronto, Ontario, Canada). 2-Methylhippuric acid, 2MHA; 3-methylhippuric acid, 3MHA; 4-methylhippuric acid, 4MHA; mandelic acid, MA; and phenylglyoxylic acid, PGA, were purchased from Sigma Chemicals (St. Louis, MO, USA). N-Acetyl-S-(3-hydroxypropyl)-L-cysteine, 3HPMA, and N-acetyl-S-(2-carbamoylethyl)-L-cysteine, AAMA, were purchased from C/D/N Isotopes (Pointe-Claire, Quebec, Canada).

Thirty-five corresponding isotopically labelled internal standards (ISs) were used. CEMA-¹³C₃; 2HPMA-D₃; SBMA-D₅; GAMA-D₃; CYMA-D₃; CHEMA-D₃; HEMA-D₄; 1,2DCVMA-¹³C-D₃; 2,2DCVMA-¹³C,D₃; 2,4DPMA-D₃; 2,5DPMA-D₃; 3,4DPMA-D₃; DHBMA-¹³C₄; MHBMA3-D₃; CMEMA-D₃; ATCA-¹³C,¹⁵N2; AMCC-¹³C₃,¹⁵N; PGA-D₅; 1,2PHEMA-¹³C₆ + 2,2PHEMA-¹³C₆ SPMA-¹³C₆; BPMA-D₇; TTCA-¹³C₃; TCVMA-D₃; EMA-D₅; MMA-D₃; IPM1-D₃; IPM3-D₃; and HPMMA-D₃ were purchased from Toronto Research Chemicals (Toronto, Ontario, Canada). 3HPMA-D₆; AAMA-D₄; 2MHA-D₇; 3MHA-D₇; 4MHA-D₇; and MA-D₅ were purchased from C/D/N Isotopes (Pointe-Claire, Quebec, Canada). MU-¹³C₆ was purchased from Sigma-Aldrich (St. Louis, MO, USA).

2.3. Determination of VOCMs in urine

Thirty-eight VOCMs were determined in urine using a method described elsewhere with slight modification (Alwis et al., 2012). Briefly, 50 μL of a 500 ng/mL IS mixture (in methanol at final IS concentrations of 50 ng/mL for each analyte) was spiked into a mixture of 50 μL of urine sample and 400 μL of 15 mM ammonium acetate buffer (pH ~6.8) in a 2.0-mL snap-cap polypropylene microcentrifuge tube (Costar^®, Corning Incorporated, Corning, NY, USA). The sample mixture was centrifuged in a mini-centrifuge (MiniSpin^®, Eppendorf, Enfield, CT, USA) at 10,000 rpm for 3 min. The supernatant was transferred into a 300-μL glass insert contained in a 2-mL HPLC vial for high-performance chromatography–tandem mass spectrometry (HPLCMS/MS) analysis. Identification and quantification of target analytes was accomplished using a Waters ACQUITY Class I UPLC system (Waters, Milford, MA, USA) coupled with an AB SCIEX 5500 electrospray triple-quadrupole mass spectrometer (ESI-MS/MS; Applied Biosystems, Foster City, CA, USA). The HPLC-MS/MS instrument was operated in negative ionization mode with multiple reaction monitoring (MRM) (Tables S2A and S2B). Chromatographic separations were performed using ab Acquity UPLC^® HSS T3 column (150 mm length, 2.1 mm internal diameter, 1.8 μm particle size; Waters, Milford, MA, USA) serially connected to a VanGuard^™ pre-column (5 mm length, 2.1 mm internal diameter, 1.8 μm particle size; Waters, Milford, MA, USA) with acetonitrile (A) and 15 mM ammonium acetate in water (pH ~6.8, B) as mobile phases. The mobile-phase flow rate was kept at 250 μL/min for the initial 2 min and then at 300 μL/min for the remaining 8 min. The sample injection volume was 3.0 μL. The mobile-phase gradient used for separation was 97% B from 0.0 to 2.0 min, 95% B from 2.0 to 3.0 min, 90% B from 3.0 to 5.0 min, 70% B from 5.0 to 6.5 min, 60% B from 6.5 to 7 min, 85% B from 7.0 to 7.5 min and 90% B from 7.5 min to 8 min, and was then returned to the initial condition (97% B) for 2 min to allow the column to equilibrate. The total run time was 10 min. The column was maintained at a constant temperature of 40 °C using an HPLC column heater (Cole-Palmer Instrument Company, Vernon Hills, IL, USA).

Quantification of analytes was carried out using an isotope dilution method, by taking the ratio of the absolute response of each native analyte to that of the corresponding isotope-labeled internal standard. Peak integration, calibration and quantification of analytes were performed using Analyst^® software (version 1.7.2, AB Sciex, Framingham, MA, USA). A linear regression model was used (1/x weighting, where x is the standard concentration) for fitting the calibration curve except for IPM3 and HPMMA, for which quadratic regression was used.

2.4. Quality assurance (QA) and quality control (QC)

A mixture of acetonitrile and water (50:50, v/v) was used to rinse the HPLC syringe twice before and after each injection. Acetonitrile was used in place of urine for procedural blanks, which were analyzed with every batch of 10 samples. None of the target analytes was found in blanks (procedural or matrix) at concentrations above the limit of detection (LOD). A midpoint calibration standard (50 ng/mL native:50 ng/mL IS) was injected after every 50 samples to monitor instrumental drift in sensitivity. A 16-point standard calibration curve prepared in 15 mM ammonium acetate buffer (at a concentration range of 0.01–1000 ng/mL and a regression coefficient of >0.990 for each analyte) was used to quantify target analytes. The recoveries of VOCMs were determined by spiking a known amount of target analytes into a sample of synthetic urine (Surine® Negative Urine Control; Cerilliant Corporation, TX, USA) at three different concentrations (low, 10 ng/mL; medium, 50 ng/mL; high, 100 ng/mL). The relative recoveries of VOCMs were in the range of 80%–130%, except for 3HPMA, which had recovery of 64% at a fortified concentration of 10 ng/mL (Table S3). However, analysis of 3HPMA without the centrifugation step in sample preparation yielded recoveries of 64% at 10 ng/mL, 81% at 50 ng/mL and 81% at 100 ng/mL. Some VOCMs were not chromatographically separated, and those were quantified as summed concentrations of two or three coeluting metabolites (e.g., 1,2 DCVMA and 2,2DCVMA; 2,4DPMA, 2,5DPMA and 3,4DPMA; 3MHA and 4MHA; MHBMA1+2 and MHBMA3). The relative standard deviation (RSD%) of repeated analysis of the QC samples was < 20% for all VOCMs. The method limits of detection (LODs) were in the range of 0.210–16.2 ng/mL, except for MMA (20.0 ng/mL), DHBMA (24.0 ng/mL) and MA (25.3 ng/mL). The recoveries of VOCMs measured in National Institute of Standards and Technology (NIST) smokers’ (3672) and non-smokers’ urine (3673) Standard Reference Materials (SRMs) were in the range of 70–130%.

2.5. Determination of OSBs and creatinine in urine

OSBs of proteins (o,o′-dityrosine [diY]); DNA (8-hydroxy-2′-deoxyguanosine [8-OHdG]); and lipids (malondialdehyde [MDA] and four isoprostane isomers, F₂-prostaglandin F_2α (8-isoprostaglandin F_2α [8-PGF_2α], 11β-prostaglandin F_2α [11-PGF_2α], 15(R)-prostaglandins F_2α [15-PGF_2α] and 8-iso,15(R)-prostaglandin F_2α[ 8,15-PGF_2α]), as well as creatinine, were measured in the same set of urine samples as reported in our previous study (Martinez-Moral and Kannan, 2019). The details are presented in the Supplementary Information.

2.6. Data analysis

Statistical analyses were performed using SPSS 19.0 (SPSS Inc., Chicago, IL, USA) and GraphPad Prism 9.1.1 (GraphPad Software, San Diego, CA, USA). Concentrations of analytes below the LOD were replaced with a value of LOD divided by the square root of 2. Creatinine-adjusted VOCM and OSB concentrations (micrograms analyte per gram creatinine) were calculated by dividing the analyte concentration (microgram analyte per liter of urine) by the corresponding creatinine concentration (gram creatinine per liter urine) (Barr et al., 2005). VOCMs concentrations and OSB levels were log-transformed (x + 1) prior to statistical analysis.

Intra-class correlation coefficients (ICCs), the ratios of between-individual variance to the sum of within- and between-individual variance, were calculated to assess the variability of VOCM concentrations over time (Pleil et al., 2018). ICCs values were classified as excellent (>0.75), fair to good (0.40–0.75) and poor (ICC <0.40); the higher the ICC values, the better the predictability/repeatability of the measure (Wang et al., 2016). The within- and between-individual variances were calculated using a linear mixed-effects model with maximum likelihood estimation, as implemented by SPSS. Alkaike information criterion (AIC) values were calculated to evaluate the fitness of models of urinary VOCM concentrations reported as unadjusted values (i.e., ng/mL), creatinine-corrected values (μg/g creatinine) and creatinine as a covariate. The lower the AIC values, the better the fitness of the measure (Wang et al., 2016). Pearson correlation analysis was performed and heat maps were generated using GraphPad Prism to evaluate the correlations among VOCMs and between individual VOCM and OSBs. The significance level was set at p < 0.05. Principal component analysis (PCA) was applied to predict the directionality of variations in the datasets (Chen et al., 2020). PCA was performed on VOCM concentrations and OSBs to calculate the eigen decomposition of the covariance matrix by means of factor analysis. For the calculation, FMV urine samples were analyzed separately (n = 243) as well as in combination with spot urine samples (n = 515).

3. Results and Discussion

3.1. Urinary concentrations and profiles of VOCMs

The detection frequency (DF, %) and the unadjusted (ng/mL) and creatinine-adjusted geometric mean concentrations of 38 VOCMs (μg/g) in spot (n = 515) and first morning void (FMV) urine samples (n = 243) are presented in Table 1. The DFs of VOCMs in urine varied from 1.2% to 99.2%. DHBMA, 3+4MHA, 2MHA, AAMA, CMEMA, HPMMA, CEMA, SBMA, MU and 3HPMA were frequently detected (DF >90%), suggesting that human exposure to their parent VOCs (1,3-butadiene, xylene, acrylamide, crotonaldehyde, acrolein, toluene, benzene) is widespread. TCVMA, EMA, 1,2+2,2DCVMA and 2,4+2,5+3,4DPMA were not detected in any of the urine samples. Previous studies (Kuang et al., 2021; Mcgraw et al., 2021; De Jesús et al., 2020) reported similar findings in DF for urinary VOCMs in non-smoker samples; high DFs were found for CEMA, 3HPMA, 2HPMA, AAMA, PGA, AMCC, ATCA, 2MHA, 3+4MHA, DHBMA, BPMA and HPMMA, whereas low DFs were found for HEMA, 1,2DCVMA, 2,2DCVMA, 2,4+2,5+3,4DPMA, 1,2+2,2PHEMA, SPMA, TTCA and TCVMA (Table S4). In our study, we identified AAMA-Sul, CYMA, CHEMA, HEMA, 1,2+2,2DCVMA, 2,4+2,5+3,4DPMA, MHBMA1+2+3, 1,PHEMA+2,2PHEMA, SPMA, TTCA, MMA, NANPC, EMA, MMA, NANPC, IPM1 and IPM3 at DFs <60%. These metabolites arise from exposure to acrylamide, acrylonitrile, vinyl chloride, ethylene chloride, 1,3-butadiene, xylene, styrene, benzene, carbon disulfide, tetrachloroethylene, ethylating agents, 4-chloronitrobenzene and isoprene. It is worth noting that VOCMs with low detection frequency may not imply absence of exposure. It could arise from inappropriateness of the biomarker selected or the inadequate sensitivity of the method to detect those exposures. Furthermore, VOCs have short half-lives in the body (few hours) and excrete quickly and therefore exposure may not have been captured. The analytes with low detection frequencies were excluded from further analysis in order to eliminate any possible bias that could arise from substitution of values below LODs (Li et al., 2019).

Table 1.

Detection frequency (DF %), unadjusted geometric mean concentrations (ng/mL) and creatinine-adjusted geometric mean volatile organic compound metabolite (VOCM) concentrations (μg/g) in 19 healthy volunteers collected across a period of 44 days.

Parent VOC	VOCM	DF%	Spot urine (n = 515)		FMV urine (n = 243)
			Unadjusted concentrations	Creatinine-adjusted concentrations	Unadjusted concentrations	Creatinine-adjusted concentrations
Acrolein	CEMA	94.4	73.0	57.0	126	76.0
Acrolein	3HPMA	92.2	184	143	310	188
Propylene oxide	2HPMA	84.5	32.2	25.1	55.2	33.4
Toluene	SBMA	92.2	3.50	2.70	4.30	2.60
Acrylamide	AAMA	98.6	26.1	20.3	41.0	24.8
Acrylamide	AAMA-Sul	54.8	<LOD	<LOD	<LOD	<LOD
Acrylamide	GAMA	88.7	11.9	9.30	15.3	9.20
Acrylonitrile	CYMA	8.00	<LOD	<LOD	<LOD	<LOD
Acrylonitrile	CHEMA	6.40	<LOD	<LOD	<LOD	<LOD
Acrylonitrile, vinyl chloride, ethylene oxide	HEMA	32.0	<LOD	<LOD	<LOD	<LOD
Trichloroethylene	1,2+2,2 DCVMA	0	<LOD	<LOD	<LOD	<LOD
Xylene	2,4+2,5+3,4 DPMA	0	<LOD	<LOD	<LOD	<LOD
Xylene	2 MHA	98.4	10.7	8.20	16.7	10.7
Xylene	3+4 MHA	98.8	68.7	53.5	106	64.2
1,3-Butadiene	DHBMA	99.2	232	181	351	212
1,3-Butadiene	MHBMA1+2+3	49.7	<LOD	<LOD	<LOD	<LOD
Crotonaldehyde	CMEMA	97.1	420	327	911	551
Cyanide	ATCA	80.2	52.3	40.7	61.2	37.0
N,N-Dimethylformamide	AMCC	60.4	32.4	25.2	41.0	24.8
Ethylbenzene, styrene	PGA	73.0	28.6	22.2	87.3	52.8
Styrene	1,2 PHEMA+2,2 PHEMA	16.1	<LOD	<LOD	<LOD	<LOD
Ethylbenzene, styrene	MA	76.5	62.1	48.4	82.1	49.6
Benzene	SPMA	16.9	<LOD	<LOD	<LOD	<LOD
Benzene	MU	94.0	161	126	255	154
1-Bromopropane	BPMA	83.5	6.20	4.80	11.60	7.00
Carbon disulfide	TTCA	51.5	<LOD	<LOD	<LOD	<LOD
Tetrachloroethylene	TCVMA	0	<LOD	<LOD	<LOD	<LOD
Ethylating agents	EMA	0	<LOD	<LOD	<LOD	<LOD
Methylating agents	MMA	1.20	<LOD	<LOD	<LOD	<LOD
4-Chloronitrobenzene	NANPC	7.00	<LOD	<LOD	<LOD	<LOD
Isoprene	IPM1	22.5	<LOD	<LOD	<LOD	<LOD
Isoprene	IPM3	46.4	<LOD	<LOD	<LOD	<LOD
Crotonaldehyde	HPMMA	97.7	338	264	586	354

FMV, first morning void; LOD, limit of detection.

Previous studies (Kuang et al., 2021; Mcgraw et al., 2021; De Jesús et al., 2020) have reported the concentrations of the majority of VOCMs in urine similar to our study (Table S5). Unadjusted geometric mean concentrations of VOCMs in spot urine samples ranged from 3.50 ng/mL (for SBMA) to 420 ng/mL (for CMEMA), whereas those in FMV urine samples ranged from 4.30 ng/mL (for SBMA) to 911 ng/mL (for CMEMA). After adjustment for urinary creatinine, the geometric mean concentrations of VOCMs in spot urine varied from 2.70 μg/g (for SBMA) to 327 μg/g (for CMEMA). HPMMA (264 μg/g), DHBMA (181 μg/g), 3HPMA (143 μg/g), MU (126 μg/g) and CEMA (114 μg/g) were the other VOC metabolites found at notable concentrations in spot urine samples.

The coefficient of variation (CV%) of unadjusted VOCM concentrations in 515 spot urine samples ranged from 77.0% to 394%, and that in 243 FMV urine ranged from 64.4% to 375% (Tables S6A and S6B). In contrast, the CV of creatinine-adjusted concentrations ranged from 64.2% to 344% in spot and from 44.5% to 396% in FMV urine samples (Tables S7A and S7B). Adjustment of VOCM concentrations for urinary creatinine decreased the overall variability for majority of the analytes, and therefore creatinine-adjusted values were used in subsequent data analysis.

The composition profiles of VOCMs in creatinine-adjusted geometric mean concentrations (Figure 1), gender, age, ethnicity, body mass index (BMI), dietary supplement intake, alcohol intake and exercise categories (Figure S1) were determined. DHBMA, CMEMA, 3HPMA, CEMA and HPMMA collectively accounted for >70% of the total VOCM concentrations in urine, suggesting that exposure to 1,3-butadiene, acrolein and crotonaldehyde is prevalent among human populations. In this study, concentrations of VOCMs were significantly higher in females than in males (p < 0.05) (Figure S2). Jain (2015a) reported similar findings from the United States National Health and Nutrition Examination Survey (NHANES). However, a few studies have reported a lack of significant gender-related differences in urinary VOCM concentrations (Kuang et al., 2021; McGraw et al., 2021; Jain, 2015b). Participants of ages <30 years had higher VOCM concentrations than those aged 30–40 years (p < 0.05). Our results are thus similar to those reported earlier showing higher exposures in teenagers than adults (Kuang et al., 2021; De Jesús et al., 2020; Jain, 2015; Cakmak et al., 2014). A small subset of urine samples from European-descended American individuals analyzed in this study had higher VOCM concentrations than were found in samples from Asians (living in the USA), following a pattern similar to those reported earlier (Jain 2015; McGraw et al., 2021). The results indicate that FMV samples contained higher VOCM concentrations than spot urine samples (p < 0.05). BMI did not have any effect on VOCM concentrations (p > 0.05). Dietary supplement intake tended to increase urinary VOCM concentrations (p < 0.05).

Fig. 1.

Composition profile of volatile organic compound metabolite (VOCM) concentrations (creatinine-adjusted, μg/g) in spot urine samples (n = 515) from 19 volunteers collected across a period of 44 days.

The creatinine-adjusted urinary VOCM concentrations varied across the 19 participants (Table S8). The variability in VOCM concentrations across participants could be due to differences in personal exposure (source, route, duration and frequency), metabolism (genetic variations and metabolic capacity), demographic traits (sex, age, ethnicity) and lifestyle characteristics (exercise frequency, diet and alcohol intake).

A significant positive correlation existed among the concentrations of 17 VOCMs in urine samples (r values varying from 0.093 to 0.900; Table S9) collected from 19 healthy individuals. Urinary concentrations of CEMA showed the strongest correlation with HPMMA in spot urine (r = 0.900, p < 0.001) and FMV samples (r = 0.883, p < 0.001), indicative of concurrent exposure to acrolein and crotonaldehyde. Acrolein and crotonaldehyde are two α,β-unsaturated aldehydes formed during the incomplete combustion of wood and fossil fuel (Dwivedi et al., 2018; Boyle et al., 2016).

The AIC values were calculated for unadjusted log-transformed (x + 1) and creatinine-adjusted log-transformed (x + 1) urinary VOCM concentrations, as well as for log-transformed (x + 1) creatinine as a covariate to assess the fitness of the measures (Table 2). The AIC values were lowest when VOCM concentrations were adjusted for creatinine, except for GAMA, AMCC and PGA, for which the lowest AIC values were obtained with creatinine as a covariate. Since the AIC values were lowest in the creatinine-adjusted concentrations for most VOCMs, creatine-adjusted value was deemed a better measure for linear mixed-effect modeling.

Table 2.

Akaike information criterion (AIC) values for estimating fitness of log-transformed (x + 1) volatile organic compound metabolites (VOCM) concentrations in urine samples collected longitudinally for 44 days (n = 515).

VOCM	AIC values (Unadjusted)	AIC values (Creatinine-adjusted)	AIC values (Creatinine as a covariate)
CEMA	492	217	291
3HPMA	710	522	590
2HPMA	673	587	629
SBMA	209	−21.7	49.2
AAMA	346	104	153
GAMA	−289	−293	−474
2 MHA	284	17.0	94.2
3+4 MHA	356	18.2	112
DHBMA	440	69.3	176
CMEMA	657	484	537
ATCA	504	433	462
AMCC	236	−34.5	−66.4
PGA	771	602	591
MA	175	−26.3	−19.2
MU	537	319	400
BPMA	560	463	532
HPMMA	474	172	254

3.2. Within and between-individual variability in VOCM concentrations

The apportionment of variances in log-transformed (x + 1) creatinine-adjusted VOCM concentrations in spot and FMV urine samples from 19 healthy individuals collected across a period of 44 days is shown in Table 3 and Table 4, respectively. The ICC values for VOCM concentrations in the current study ranged from 0.080 to 0.425 for spot and from 0.050 to 0.749 for FMV urine samples. In both sets of samples, within-individual variance was higher than between-individual variance for most of the VOCMs except for SBMA, 2MHA, 3+4MHA and AMCC in FMV urine samples in which the between-individual variance was higher. This is in line with reports in previous studies of higher within-individual variability than between-individual variability for urinary neonicotinoid insecticides (Li et al., 2019, 2020).

Table 3.

The apportionment^a of variance in creatinine-adjusted log-transformed (x + 1) volatile organic compound metabolites (VOCM) concentrations in urine samples (n = 515).

VOCM	ICC valuesb	Within-individualc		Between-individuale
		Variance (%)	Variance (95% CI)d	Variance (%)	Variance (95% CI)
CEMA	0.263	0.086 (73.7)	(0.076, 0.098)	0.031 (26.3)	(0.015, 0.062)
3HPMA	0.142	0.152 (85.8)	(0.135, 0.173)	0.025 (14.2)	(0.011,0.054)
2HPMA	0.201	0.177 (79.9)	(0.156, 0.201)	0.044 (20.1)	(0.021, 0.091)
SBMA	0.425	0.054 (57.5)	(0.048, 0.062)	0.040 (42.4)	(0.020, 0.078)
AAMA	0.324	0.068 (67.6)	(0.060, 0.077)	0.032 (32.4)	(0.016, 0.064)
GAMA	0.217	0.032 (78.3)	(0.028, 0.036)	0.008 (21.7)	(0.004,0.018)
2 MHA	0.277	0.055 (72.3)	(0.049,0.063)	0.021 (27.7)	(0.010, 0.042)
3+4 MHA	0.384	0.055 (61.6)	(0.049, 0.063)	0.034 (38.4)	(0.017, 0.068)
DHBMA	0.124	0.063 (87.6)	(0.056, 0.071)	0.008 (12.4)	(0.004, 0.019)
CMEMA	0.392	0.142 (60.8)	(0.125, 0.161)	0.092 (39.2)	(0.047, 0.179)
ATCA	0.080	0.131 (92.0)	(0.115, 0.148)	0.011 (8.0)	(0.004, 0.028)
AMCC	0.387	0.053 (61.3)	(0.047, 0.060)	0.033 (38.7)	(0.017,0.066)
PGA	0.347	0.189 (65.3)	(0.166, 0.214)	0.100 (34.7)	(0.051, 0.198)
MA	0.094	0.052 (90.6)	(0.046, 0.059)	0.005 (9.4)	(0.002, 0.012)
MU	0.162	0.101 (83.8)	(0.089, 0.115)	0.019 (16.2)	(0.009, 0.042)
BPMA	0.208	0.139 (79.2)	(0.123, 0.158)	0.036 (20.8)	(0.017, 0.075)
HPMMA	0.322	0.078 (67.8)	(0.069, 0.089)	0.037 (32.2)	(0.018, 0.074)

^aAge, gender, body mass index and ethnicity were included as covariates.

^bICC: intra-class correlation coefficient.

^ccontribution of within-individual variance to total variance.

^dCI: confidence interval within 95% range.

^eThe proportion of between-individual variance to total variance.

Table 4.

The apportionment^a of variance in creatinine-adjusted log-transformed (x + 1) volatile organic compound metabolite (VOCM) concentrations in first morning void urine samples (n = 243).

VOCM	ICC valuesb	Within-individualc		Between-individuale
		Variance (%)	Variance (95% CI)d	Variance (%)	Variance (95% CI)
CEMA	0.085	0.075 (91.5)	(0.062, 0.090)	0.007 (8.5)	(0.002, 0.026)
3HPMA	0.163	0.145 (83.7)	(0.121, 0.174)	0.028 (16.3)	(0.010, 0.077)
2HPMA	0.354	0.194 (64.6)	(0.162, 0.233)	0.106 (35.4)	(0.043, 0.264)
SBMA	0.564	0.058 (43.6)	(0.048, 0.070)	0.075 (56.4)	(0.032, 0.177)
AAMA	0.092	0.073 (90.8)	(0.061, 0.087)	0.007 (9.2)	(0.002, 0.029)
GAMA	0.159	0.023 (84.1)	(0.019, 0.028)	0.004 (15.9)	(0.002, 0.013)
2 MHA	0.537	0.045 (46.3)	(0.037, 0.054)	0.052 (53.7)	(0.022, 0.122)
3+4 MHA	0.615	0.055 (38.5)	(0.046, 0.066)	0.087 (61.5)	(0.038, 0.200)
DHBMA	0.107	0.071 (89.3)	(0.059, 0.085)	0.008 (10.7)	(0.003, 0.026)
CMEMA	0.213	0.120 (78.7)	(0.100, 0.144)	0.033 (21.3)	(0.013, 0.084)
ATCA	0.468	0.110 (53.2)	(0.092, 0.132)	0.097 (46.8)	(0.043, 0.219)
AMCC	0.749	0.049 (25.1)	(0.041, 0.059)	0.147 (74.9)	(0.064, 0.334)
PGA	0.050	0.136 (95.0)	(0.113, 0.164)	0.007 (5.0)	(0.001, 0.040)
MA	0.237	0.047 (76.3)	(0.039, 0.056)	0.014 (23.7)	(0.005, 0.039)
MU	0.136	0.086 (86.4)	(0.072, 0.104)	0.014 (13.6)	(0.004, 0.042)
BPMA	0.370	0.144 (63.0)	(0.120, 0.172)	0.084 (37.0)	(0.034, 0.208)
HPMMA	0.119	0.058 (88.1)	(0.049, 0.070)	0.008 (11.9)	(0.002, 0.026)

^aAge, gender, body mass index and ethnicity were included as covariates.

^bICC: intra-class correlation coefficient.

^cThe contribution of within-individual variance to total variance.

^dCI: confidence interval within 95% range.

^eThe proportion of between-individual variance to total variance.

In spot urine samples, all VOCMs showed poor predictability (low ICC values) except SBMA (ICC = 0.425; within-individual variability at 57.5%), which showed fair to good predictability. ATCA showed the lowest ICC value (ICC = 0.080), followed by MA (ICC = 0.094). Our results reveal poor predictability and high within-individual variability for urinary VOCM concentrations. Our findingsindicate that limiting data collection to a single spot urine analysis of VOCMs in epidemiological investigations may lead to exposure misclassification.

In comparison to spot urine, VOCM concentrations in FMV samples showed fair to good predictability for SBMA, 2MHA, 3+4MHA, ATCA, AMCC and PGA (ICC value > 0.40). The parent VOCs for these metabolites are toluene, xylene, cyanide, N,N-dimethylformamide, ethylbenzene and styrene. It is likely that exposure to these parent VOCs occurs consistently from day-to-day items such as paints, inks, adhesives, plastics, food and water (Li et al., 2021). Our results indicate that FMV urine constitutes a better sample than spot urine for evaluating individual exposure to certain VOCs. Similarly, previous studies have shown that FMV urine samples provide more predictable measurements of exposures to phthalates, bisphenol A, triclosan, parabens, pyrethroid pesticides and organophosphate pesticides (Kim et al., 2020; Egeghy et al., 2011; Kissel et al., 2005).

To the best of our knowledge, this is the first study to assess the predictability of urinary VOCM concentrations in spot and FMV urine samples from non-smokers. One previous study assessed variability in urinary VOCM concentrations among smokers, and that study reported fair to good predictability for 2MHA, 3+4MHA, AAMA, GAMA, CEMA, 3HPMA, 2HPMA, MA, PGA, HEMA, HPMMA and ATCA and excellent predictability for AMCC (Lorenz et al., 2021). It is plausible that VOC exposures among non-smokers are more variable diurnally than those of smokers.

3.3. Correlation between urinary VOCMs and biomarkers of oxidative stress

We calculated Pearson correlation coefficients to elucidate the relationships between VOCMs and OSBs in spot (Figure 2) and FMV (Figure S3) urine samples collected from 19 healthy individuals. Urinary concentrations of most VOCMs were significantly positively correlated with seven biomarkers of oxidative stress in both spot (r values varying from 0.103 to 0.346; Table S10) and FMV (r values varying from 0.131 to 0.301; Table S11) samples. 3+4MHA (a metabolite of xylene) showed positive correlation with 8OHdG (biomarker for DNA damage) in both spot (r = 0.346, p < 0.001) and FMV urine (r = 0.301, p < 0.001) samples. Exposure to xylene induces DNA damage in human cells (Chen et al., 2008; ATSDR, 2007a). Urinary concentrations of 2MHA, ATCA and MA were also significantly correlated with 8OHdG levels. Earlier studies associated xylene (Sisto et al., 2020; Wang et al., 2013; ATSDR, 2007a), cyanide (ATSDR, 2006; Bhattacharya and Lakshmana Rao, 1997), ethylbenzene and styrene (Costa et al., 2012; Wongvijitsuk et al., 2011; Liu et al., 2010) exposures with DNA damage. Urinary concentrations of CEMA, 3HPMA, 2HPMA, ATCA BPMA and HPMMA were significantly correlated with MDA (a biomarker for lipid damage). Exposure to acrolein (ATSDR, 2007b), propylene oxide (Erukova et al., 2000), cyanide (Kadiri and Asagba, 2019; Bhattacharya and Lakshmana Rao, 1997), 1-bromopropane (Lee et al., 2005) and crotonaldehyde (Zhang et al., 2019) was associated with lipid damage. Likewise, urinary concentrations of ATCA were significantly correlated with DiY, a biomarker for protein damage. Exposure to cyanide is associated with cellular protein damage (ATSDR, 2006; Bhattacharya and Lakshmana Rao, 1997). Urinary concentrations of VOCMs showed varying levels of association with four isoprostane isomers concentrations. 3HPMA, 2HPMA, SBMA, 2MHA, DHBMA and ATCA concentrations were significantly correlated with that of 8,15-PGF_2α whereas 2MHA, 3+4MHA and ATCA concentrations were significantly correlated with that of 8-PGF_2α, and ATCA concentration with that of 11-PGF_2α. Isoprostanes are prostaglandin-like inflammatory biomarkers formed from peroxidation of arachidonic acid (Roberts and Milne, 2009). Acrolein (Park et al., 2007), toluene (Mögel et al., 2011), xylene (Wang et al., 2020) and cyanide (Sakaida et al., 1992) have been shown to upregulate oxidation of arachidonic acid.

Fig. 2.

Heat map showing correlation distribution among creatinine-adjusted concentrations of log-transformed (x + 1) volatile organic compound metabolites and log-transformed (x + 1) oxidative stress biomarkers in spot urine samples (n = 515).

To investigate the similarities in the distribution patterns of VOCMs and OSBs in spot urine, we performed PCA. Six principal components (PC1, PC2, PC3, PC4, PC5 and PC6) with eigenvalues of >1.0 were extracted for each component (Table S12A). PC1, PC2 and PC3 accounted for 29%, 12% and 8% of the total variance, respectively (Table S12B). PC plots of VOCM (PC1, 0.772 to 0.337; PC2, 0.466 to −0.466; PC3, 0.371 to −0.366) and OSBs (PC1, 0.230 to −0.068; PC2, 0.649 to 0.164; PC3, 0.630 to 0.081) showed distinctive clustering, which could be explained by the differences in the sources of exposure to these compounds. CEMA and 3HPMA (metabolites of acrolein), HPMMA and CMEMA (metabolites of crotonaldehyde) and PGA (a metabolite of ethylbenzene and styrene) clustered together (PC1, 0.700 to 0.648; PC2, −0.311 to −0.503; PC3, 0.371 to 0.108) (Figure 3). Acrolein, crotonaldehyde, ethylbenzene and styrene share a common exposure source, namely fossil fuel combustion and vehicular exhaust (ATSDR, 2021; NTP, 2021). 2MHA and 3+4 MHA (metabolites of xylene), MA (a metabolite of styrene) and AMCC (a metabolite of N,N-dimethylformamide) clustered together (PC1, 0.766 to 0.542; PC2, 0.368 to 0.222; PC3, −0.241 to −0.366), which can be explained by their derivation from building materials, varnishes and paints (xylene, styrene and N,N-dimethylformamide; ATSDR, 2021; NTP, 2021). AAMA (a metabolite of acrylamide), BPMA (a metabolite of 1-bromopropane), DHBMA (a metabolite of 1,3-butadiene) and MU (a metabolite of benzene) clustered together (PC1, 0.705 to 0.487; PC2, −0.026 to −0.503; PC3, 0.109 to −0.123); personal care products are major sources of these metabolites are (Li et al., 2021; ATSDR, 2021; NTP, 2021). ATCA (a metabolite of cyanide), GAMA (a metabolite of acrylamide) and SBMA (a metabolite of toluene) clustered together (PC1, 0.503 to 0.337; PC2, 0.466 to 0.264; PC3, −0.023 to −0.148) and are derived from industrial or manufacturing releases (Li et al., 2021; Boyle et al., 2016; ATSDR, 2021; NTP, 2021). Thus, PCA provided information regarding the various sources of exposure to particular VOCs.

Fig. 3.

Principal component analysis (PCA) of creatinine-adjusted urinary concentrations of log-transformed (x + 1) volatile organic compound metabolites (circles) and log-transformed (x + 1) oxidative stress biomarkers (squares) in spot urine samples (n = 515). Three major components, PC1 (29%), PC2 (12%) and PC3 (8%), are depicted. The factors consisted of CEMA, 3HPMA (ThHPMA), 2HPMA (TwHPMA), SBMA, AAMA, GAMA, 2MHA (TwMHA), 3+4MHA (ThFMHA), DHBMA, CMEMA, ATCA, AMCC, PGA, MA, MU, BPMA, HPMMA, diY, 8-OHdG (EightOHdG), MDA, 8-PGF2α (EightPGF2α), 11-PGF2α (ElevenPGF2α), 15-PGF2α (FifteenPGF2α) and 8,15-PGF2α (EightFifteenPGF2α).

4. Conclusions

We found that longitudinal measurement of urinary VOCMs showed high intra-individual variability and low temporal predictability in VOC exposures. Therefore, analysis of repeated samples of urine at multiple time points is a better approach for characterizing average exposure to VOCs over time. Longitudinal measurement would reduce bias in exposure misclassification in human biomonitoring studies of VOCs. Our study also provides evidence that VOC exposures are positively correlated with several biomarkers of oxidative stress longitudinally.

The strengths of this study include relatively large sample size per individual and the measurements of 38 VOCMs and 7 OSBs in spot and FMV urine samples collected for 44 days. Concurrent analysis of OSBs added valuable information linking VOC exposure to oxidative stress. The profiles of VOCMs in urine also provided cues to exposure sources. However, we would like to reiterate that the results of our study are based on a small convenience sample collected from 19 individuals from a single geographic location and further work is required to generalize our findings to a larger population. Other limitations of the this study include variable number of spot and FMV urine samples; potential bias from cross-sectional study design, lack of information on passive smoking and occupational VOCs exposure.The half-lives of VOCMs range widely, which can result in plausible exposure measurement error. Other environmental contaminants can elicit oxidative damage, and those covariates were not included in our models..Nevertheless, our study provides valuable information for formulating sampling strategies to assess VOC exposures in epidemiological investigations.

Highlights.

• 38 VOC metabolites were measured in urine collected from 19 individuals for 44 days.

• Urinary VOC metabolite concentrations varied temporally within- and between-individuals.

• ICC values of VOC metabolites in urine varied from 0.05 to 0.749.

• Urinary VOC metabolites were significantly correlated with oxidative stress markers.