Evaluation of urinary limonene metabolites as biomarkers of exposure to greenness

Abstract

Exposure to plants is known to improve physical and mental health and living in areas of high vegetation is associated with better health. The addition of quantitative measures of greenness exposure at individual-level to other objective and subjective study measures will help establish cause-and-effect relationships between greenspaces and human health. Because limonene is one of the most abundant biogenic volatile organic compounds emitted by plants, we hypothesized that urinary metabolites of inhaled limonene can serve as biomarkers of exposure to greenness. To test our hypothesis, we analyzed urine samples collected from eight human volunteers after limonene inhalation or after greenness exposure using liquid chromatography-high resolution mass spectrometry-based profiling. Eighteen isomers of nine metabolites were detected in urine after limonene inhalation, and their kinetic parameters were estimated using nonlinear mixed effect models. Urinary levels of most abundant limonene metabolites were elevated after brief exposure to a forested area, and the ratio of urinary limonene metabolites provided evidence of recent exposure. The identities and structures of these metabolites were validated using stable isotope tracing and tandem mass spectral comparison. Together, these data suggest that urinary metabolites of limonene, especially uroterpenol glucuronide and dihydroperillic acid glucuronide, could be used as individualized biomarkers of greenness exposure.

1. Introduction

Several studies suggest that exposure to greenness is associated with beneficial effects on human health (WHO, 2016) and that living in areas of high surrounding greenness is associated with a decrease in the risk of cardiovascular disease (Yeager et al., 2020), and all-cause mortality (James et al., 2015; Yang et al., 2021). Exposure to green areas has positive effects on human health by promoting physical activity (de Keijzer et al., 2020), mitigating environmental hazards like air pollution and high temperature (Wolf et al., 2020), reducing stress and improving mental well-being (Dadvand et al., 2016; Ulrich et al., 1991), and potentially enhancing microbial diversity and human microbiota composition (Selway et al., 2020). Therefore, understanding the connection between green environments and health is crucial. Achieving this understanding requires a focus on individual biomonitoring.

Several metrics have been employed to measure greenspace exposure (Nguyen et al., 2021; Yang et al., 2021), including objective parameters like Normalized Difference Vegetation Index (NDVI) (James et al., 2015; Luo et al., 2020; Yeager et al., 2018), greenspace percentage (van den Berg et al., 2015), proximity to greenspaces (Luo et al., 2020), as well as viewing simulations (Jo et al., 2019) and physical activities within natural environments (Bowler et al., 2010). Additionally, subjective parameters such as self-reported exposure (Li et al., 2021) and perceived access to greenspace through window (de Keijzer et al., 2016) have been used to assess the impact of greenness on human health. Despite prior efforts to characterize individual-level exposure to greenness, there remains a need for the development of more rigorous quantitative methods, which will be essential for evaluating the potential health effects of greenness in future (Hart et al., 2022). This field is currently understudied, but it is critical for future quantitative assessments of direct health effects resulting from vegetation-related exposure.

Biogenic volatile organic compounds (BVOCs) are ubiquitously emitted by plants (Cai et al., 2021; Gibbs, 2019; Laothawornkitkul et al., 2009; Peñuelas and Staudt, 2010), and, therefore, estimating their levels in biofluids may offer a potential analytical solution to quantify greenness exposure. Although alone BVOCs may not provide a complete estimate of greenness exposure, they can be complementary to other objective and/or subjective measures, and better serve as means of describing health-effects of exposures to vegetation. Previous studies have shown that BVOCs are absorbed and metabolized by humans (Schmidt et al., 2013; Schmidt et al., 2015) and that they appear in blood after greenness exposure (Bach et al., 2021; Sumitomo et al., 2015). However, the levels of BVOC in the blood decrease quickly after inhalation (Falk-Filipsson et al., 1993; Falk et al., 1990). In comparison, urinary metabolites are more stable and their levels in urine are cumulative (Zhang et al., 2022). Thus, urinary levels can more reliably reflect inhaled BVOC exposure than the blood levels of these compounds. Currently, there is limited research on characterization of BVOC metabolites as potential biomarkers of greenness exposure (Li and Zhang, 2023). Although many BVOCs are generated by plants, limonene stands out as one of the most frequent and abundant BVOC emitted by a wide range of plants (Geron et al., 2000). Therefore, the urinary metabolites of limonene are promising candidates to measure exposure to greenness.

Limonene metabolism has been extensively studied (Crowell et al., 1994; de Alvarenga et al., 2021; Hardcastle et al., 1999; Igimi et al., 1974; Kodama et al., 1976; Regan and Bjeldanes, 1976; Schmidt and Göen, 2017b; Shimada et al., 2002; Vigushin et al., 1998; Watabe et al., 1981; Wishart et al., 2022) (Figure S1). During phase I metabolism, limonene undergoes oxidation of its double bonds, cyclohexenyl ring, or side chains. Further oxidation or reduction steps result in the production of alcohols, diols, and carboxylic acids, which then conjugate with glucuronic acid, glycine, or other endogenous molecules during phase II metabolism. These conjugates are typically excreted by kidneys, and only a negligible proportion is eliminated unchanged in urine (Falk-Filipsson et al., 1993). The metabolism of limonene is dependent on the dosage. Glucuronides are the primary phase II metabolites at lower doses, while other conjugates appear at higher doses (de Alvarenga et al., 2021). Numerous metabolites have been reported. However, many of them were identified in animal models (Igimi et al., 1974; Regan and Bjeldanes, 1976; Shimada et al., 2002; Watabe et al., 1981), through oral administration (Crowell et al., 1994; Hardcastle et al., 1999; Kodama et al., 1976; Schmidt and Göen, 2017b; Vigushin et al., 1998) (rather than inhalation), and with dosages exceeding typical levels found in green environments (Crowell et al., 1994; Vigushin et al., 1998). Additionally, limited knowledge of urinary metabolite kinetics following limonene inhalation complicates candidate selection and assessment of exposure-induced changes in green environments.

This study aimed to evaluate urinary limonene metabolites (LMs) as potential biomarkers of exposure to greenness. We used liquid chromatography-high resolution mass spectrometry to analyze urine samples from human volunteers and mice. The study consisted of four steps (Figure 1). The initial step (Step 1) aimed to identify the LMs in the urine collected after a 10-minute limonene inhalation and select candidates for subsequent exposures. Secondly (Step 2), a stable isotope tracing mouse experiment was performed to confirm LM assignments and rule out possible misassignments from isomeric and/or isobaric species. Step 3 determined the elimination rates of identified LMs. Final step (Step 4), aimed to assess whether the most abundant urinary LMs were found after exposure to the green environment of a forest for 4 hours. This investigation served as an independent validation sample set, to confirm the presence of LMs’ metabolites in urine and their increase after exposure.

Figure 1.

Experimental design for discovery and evaluation of limonene metabolites as biomarkers for greenness exposure.

2. Materials and methods

2.1. Chemicals and reagents

R-Limonene (CAS# 5989-27-5, ⩾99.0%) for mouse exposure was purchased from Sigma-Aldrich, Inc. (St. Louis, MO). Limonene for human exposure was purchased from True Terpenes (Hillsboro, OR). R-Limonene 2,3,3,5,5,-d₅ (limonene-d₅) was obtained from Toronto Research Chemicals, Ontario Canada. UHPLC-MS grade acetonitrile, UHPLC-MS grade water, LC-MS grade formic acid, and Infinity^™ Creatinine Liquid Stable Reagent were purchased from Thermo Fisher Scientific Inc., Waltham MA.

2.2. Human urine sample collection

Self-reported healthy participants without any chronic disease diagnosis were recruited from the University of Louisville and surrounding areas through electronic advertising. Demographic information (Table S1) including sex, age and race was recorded. For 24 h prior to the study, participants were asked to abstain from outdoor activities as well as consuming grilled foods, foods high in BVOCs such as apples, mangoes, citrus fruit, broccoli and beer, and avoid using personal hygiene products or household cleaning products with essential oils or other fragrances. Study visits took place in a special exposure chamber at the University of Louisville Envirome Institute with increased air turnover rates and minimized exposure to common sources of BVOCs. Eight participants (4 female and 4 male), age 25–50 years (Table S1), were asked to take 5 deep breaths every min for a total of 10 min from a 15 ml conical tube with ~ 100 μl limonene. Urine samples were collected before the exposure, and at 0.5, 1, 2, 3, and 4 h after exposure, centrifuged and stored at −80°C until analysis. This study was conducted by one research assistant (RA). Only one participant was in the exposure chamber during the study visit. The RA explained the study protocol to the participants and obtained their consent. The RA also collected all the samples and provided them to the bio-banking team. The bio-banking coordinator provided the deidentified samples to the analytical chemists. The study was approved by the University of Louisville Institutional Review Board (IRB 22.0543) and no study related measures were completed prior to informed consent.

For greenness exposure urine collection, the field experiment was performed in Charlestown State Park (Charlestown, IN, USA) on September 17, 2017. The forest was dominated by deciduous broad-leaved trees and shrubs including sugar maple (Acer saccharum), red maple (Acer rubrum), Black cherry (Prunus serotina), and White ash (Fraxinus americana) (Lynch, 2016; Maxwell and Thomas, 2003). Before the experiment, 8 healthy volunteers (5 males and 3 females, age 25–70 years; Table S1) were asked to abstain from potential limonene containing products and atmosphere for 24h. During the experiment, participants stayed in a forested environment of Charlestown State Park. Urine samples were collected before and after 4h of exposure. This was a continuous exposure experiment and the 4h period was chosen to allow for accumulation of limonene in the body and to enhance the response of limonene metabolites in the urine. The urine samples were centrifuged immediately and stored at −80°C until analysis. The study was conducted by two RAs. The exposure started and ended at the same time for all 8 participants. The RAs stayed with the study subjects throughout the duration of the experiment. Both RAs were involved in explaining the study protocol to the participants and obtaining their consent, and in the sample collection. The study was approved by the University of Louisville Institutional Review Board (IRB 16.0944).

2.3. UPLC-QTOF-MS DIA analysis of urine samples

The urine samples (50 μL) were thawed on ice and mixed with 0.1% formic acid (450 μL) in water. An aliquot of 2 μL of the mixture was analyzed using an Acquity I-Class UPLC system (Waters, MA). The separation was performed using a 2.1mm × 150mm Acquity Premier CSH C18 1.7 μm UPLC column (Waters, MA) maintained at 60 °C at a flow rate of 0.45 mL/min. The columns were eluted with a gradient composed of 0.1% formic acid in water (solvent A) and 0.1% formic acid in acetonitrile (solvent B). The gradient profile started at 0% of solvent B, increased to 23% of solvent B over 11 min, and then increased to 95% of solvent B over 3.6 min. The gradient was held at 95 % of solvent B for 2.4 min before returning to the initial conditions over 0.05 min and re-equilibrated for 2.95 min before the next injection. To monitor the performance of LC-MS method (i.e., retention time shifts (ΔRT) and mass errors (Δm/z), one injection of QC sample prepared by pooling equal amounts of each urine sample, was added for every ten injections of urine samples. The acceptable ΔRT was ±0.05 min and Δm/z was ±7.5 mDa. The list of ΔRT and Δm/z of LMs in QC samples can be found in Table S2.

QTOF-MS data were collected using a Synapt XS MS (Waters, MA) with Masslynx 4.2 software and an electrospray ion source operated in a negative mode. The capillary voltage was 2.25 kV, the source temperature was 120 °C, the desolvation gas flow was 700L/h at a temperature of 650 °C, and the cone gas flow was 150 L/h. The MS^E data-independent acquisition (DIA) was performed over the mass-to-charge ratio (m/z) range of 40−930 Da with low collision at 2 V (function 1) and high collision energy ramping from 10 to 40 V (function 2). Each of these functions employed a scan time of 0.2 s. Sodium formate was used for the mass calibration before the sample run, and leucine enkephalin with m/z ~ 554.2620 was used as the lock mass solution during the acquisition. After the acquisition, raw files were analyzed using UNIFI 1.9 software package (Waters, MA).

2.4. Curation of limonene metabolite scouting library

An integrated library-guided analysis (ILGA) workflow (Xie et al., 2023) was established to annotate LC-MS peaks of urinary LMs. Figure 1 shows the general workflow developed for the assignment and analysis of LMs. Our approach started with curating a scouting library containing known and proposed LMs. After literature review (Crowell et al., 1994; Hardcastle et al., 1999; Igimi et al., 1974; Kodama et al., 1976; Regan and Bjeldanes, 1976; Schmidt and Göen, 2017a; Shimada et al., 2002; Vigushin et al., 1998; Watabe et al., 1981; Wishart et al., 2022), structures of reported LMs amenable to ESI⁻ MS detection were found by searching of SciFinderⁿ database, downloading structure files in .mol format, and importing them into a UNIFI library. Additional possible metabolites (previously unreported) were proposed and included in the library. Stereoisomers and conjugation isomers (isomers with different conjugation positions) were not considered as different entries of the scouting library. The scouting library was then used to process the MS^E data collected from human limonene inhalation samples (Table S3).

2.5. Candidate limonene metabolite peak selection

The UPLC-QTOF MS data of limonene inhalation urine samples were processed using UNIFI. After peak picking, the extracted features were queried against the scouting library. The peaks with m/z within 20 ppm of the theoretical value of LM negative ions ([M-H]⁻) were selected for further analysis. To be included in the annotation, the response of candidate peaks should increase after limonene inhalation. To test that, the peak responses were normalized to the urinary creatinine levels and then compared across all time points. Only significantly increased peaks (p < 0.05) were considered. Any features identified as in-source fragments and peaks with a maximum response lower than 1000 were removed.

2.6. Peak assignment

Peak assignment based solely on m/z matching is not feasible due to the presence of isomers. To overcome this problem, five rules were created to assign the most likely isomers to an LC-MS peak (see Figure S2). During the assignment, the rules prioritize reported structures whenever possible, so if a reported structure can explain the peak, a proposed one should not be assigned. The rules are as follows:

1. One to one: A peak with one matching item in the scouting library was assigned as is.

2. Many to one: For multiple peaks with one matching item, each peak represented an isomer that is indistinguishable in the library. To differentiate these isomers in the assignment, annotation for each peak was supplemented with its respective retention time (Rt) value.

3. One to many (reported and proposed items): For a peak with two or more matching entries that include both reported and proposed items, the item previously reported was chosen.

4. Many to many: For multiple peaks with two or more matching entries that include both reported and proposed items, the previously reported item was assigned to the more abundant one, while the proposed item was assigned to the less abundant one.

5. One to many (all reported items): For a peak with two or more possible matching entries, all previously reported, these entries were merged into one label.

After assignment, the peak information, including name, formula, and Rt were used for analysis of urine samples from greenness exposure experiment. Isomers were analyzed separately unless specifically indicated otherwise.

2.7. Stable isotope tracing in mouse for metabolite confirmation

To confirm that metabolite candidates originate from limonene, stable isotope tracing experiments were performed in mice. Specifically, 5 mg of limonene or limonene-d₅ was dissolved in sterile saline with 1% Tween 80. Two adult male mice (WT, C57BL/6) were injected intraperitoneally with 66.8 mg/kg body weight of limonene or limonene-d₅ in 100 μl of solution. Sterile saline with 1% Tween 80 was used as control. After the injection, urine was collected overnight in graduated cylinders surrounded by 4 °C water-jacketed organ baths. Mice were provided with 3% glucose and 0.125% saccharin in drinking water and food (Lynch et al., 2020). Urine samples were centrifuged (1800g, 5 min; to pellet feces or food) before being decanted and stored at −80 °C. After collection, the supernatant was analyzed by the same UPLC- Q TOF MS method as the human urine. The LC-MS result was searched for presence of corresponding unlabeled / labeled LM peaks using UNIFI and matched with the corresponding results for human exposure experiments. The protocol was approved by the Institutional Animal Care and Use Committee of the University of Louisville (IACUC 22204).

2.8. Measurement of urinary creatinine

Urinary creatinine levels were measured on an Ace Axcel Clinical Chemistry System (Alfa Wassermann, West Caldwell, NJ) using ACE Creatinine reagent (Alfa Wassermann, West Caldwell, NJ).

2.9. Relative quantification of limonene metabolites and statistical analysis

After acquisition, the LC-MS^E data were processed with UNIFI. Each LC-MS peak corresponding to LMs was identified based on the Rt and m/z and the peak response was integrated. For the relative quantification, the instrumentation responses (defined as 3-dimensional chromatographic peak volume with a cutoff value of 50) were used to measure the relative abundance of LMs in study subjects after normalization to urinary creatinine to adjust for dilution. Response values for undetected LMs or below cutoff were imputed by dividing the cutoff value (50 counts) by the square root of 2 per established protocols (Richardson and Ciampi, 2003; Succop et al., 2004; Xie et al., 2023). All LM response values were normalized to urinary creatinine levels which were measured on an Ace Axcel Clinical Chemistry System (Alfa Wassermann, West Caldwell, NJ) using ACE Creatinine reagent (Alfa Wassermann, West Caldwell, NJ).

ANOVA and Student’s t-test analyses were conducted to examine the differences in LM at various time points using SAS, version 9.4 (SAS Institute, Inc., NC). Since the distributions of the normalized response were right-skewed, they were log-transformed to improve the normality for statistical analysis. The statistical significance was set at the p-value <0.05.

2.10. Kinetic modeling

A nonlinear mixed effect model (NLME) was built with R 4.2.2 using “saemix” function from “saemix” package (Comets et al., 2017; Comets et al., 2021) to describe the generation and elimination of each LM. A pseudo one-compartment model was assumed and adopted from previous reports (Barbeau et al., 2018; Lou et al., 2009; Lutier et al., 2016), as follows:

$$C_{ij}=\frac{K{a}_i\times F_i\times D}{V\left(K{a}_i-K{e}_i\right)}\left(e^{-K{e}_i\times t_j}-e^{-K{a}_i\times t_j}\right)+{\epsilon }_{ij}$$ (1)

where i is an index for the participant and j for the urine sample. C_ij is the creatinine normalized peak response for urine j of participant i, and t_j is the urine sample collection time (h). Ka_i and Ke_i were the first-order urinary generation or elimination rate constant (h⁻¹), respectively, for participant i. F_i is the fraction of specific LM over absorbed limonene for participant i. D is the amount of parent compound (limonene) absorbed and V is volume of distribution. Since relative quantification results were used as input data, absolute values of D, V, and F cannot be acquired. For convenience, D and V were given fixed values of 100 and 1, respectively. ε_ij is the error term. Ka, Ke, and F were assumed to be independent and normally distributed among participants.

2.11. Structure confirmation with MS/MS spectra

MS/MS spectra of LMs were collected using product ion scan in human urine with high LM level. Sample preparation and UPLC conditions were identical to those described in Section 2.3. The product ion scans (40–500 m/z) used parameters similar to MS^E function 2. The resulting MS/MS spectra were averaged around their respective Rt. To confirm LM structures, the MS/MS spectra obtained were compared with those reported in the literature.

3. Results

3.1. Curation of a scouting library for LM annotation (Step 1)

The first step of the ILGA workflow was to construct a scouting library based on prior knowledge of limonene metabolism (Table S3, Figure S1). The library comprised 31 LMs, of which 25 had been previously reported. Among these, 17 were phase I metabolites primarily consisting of limonene oxidation products such as epoxides, alcohols, aldehydes, ketones, and carboxylic acids (Crowell et al., 1994; Hardcastle et al., 1999; Igimi et al., 1974; Kodama et al., 1976; Regan and Bjeldanes, 1976; Schmidt and Göen, 2017a; Shimada et al., 2002; Vigushin et al., 1998; Watabe et al., 1981; Wishart et al., 2022). The remaining 8 are phase II metabolites, which included one glycine conjugate, two methyl esters, and five glucuronides. In addition, given that glucuronidation is a major pathway in the phase II metabolism of limonene (Rinaldi de Alvarenga et al., 2022), six predicted glucuronide structures derived from known phase I metabolites were added to the library.

3.2. Candidate LM peak selection and assignment (Step 1)

Human urine samples collected after limonene inhalation were used for candidate LM peak selection (Figure 1). Typically, peak picking resulted in approximately 10,000 features (m/z-Rt pairs) per LC-MS sample. After being queried against the scouting library, 407 LC-MS peaks matched the initial criteria (within 20 ppm of the theoretical m/z of [LM-H]⁻). The creatinine-normalized responses were then compared across different time points. Thirty-two of them that significantly increased after inhalation were further inspected to exclude in-source fragments and low intensity (response < 1000). Finally, as listed in Table S4, 18 common LM peaks were found, with each of them detected in more than two third of the samples.

As seen in Table 1, 18 candidate LM peaks were assigned with 9 corresponding metabolite IDs (LM1 – LM9) in the scouting library. LM1 and LM4 had one peak per ID, while the rest had multiple peaks per ID, indicating the presence of more than one isomer, which were labeled with a letter following the ID number. Among the 9 metabolite IDs, 3 of them (LM7, LM8, and LM9; corresponding to LMN-1,2-O-GlcA, DHPA-2-O-GlcA, and PA-8,9-O-GlcA, respectively) were proposed structures, and the rest 6 have been reported previously. Details of the peak assignments can be found in the supplementary section.

Table 1.

Peak assignment for limonene metabolites in human urine collected after limonene inhalation.

Metabolite ID	LC-MS peak (m/z_Rt)	Number of peaks with same m/z	Number of library items with matching m/z	Assignment rules applied	Assigned peak identity
LM1	167.1078_9.95	1	1	(1)	DHPA
LM2a	183.1027_11.88	2	1	(2)	DHPA-2-OH_Rt11.88
LM2b	183.1027_12.69	2	1	(2)	DHPA-2-OH_Rt12.69
LM3a	327.1449_12.73	2	4 (2 proposed)	(2) (3) (5)	LMN-O-GlcA_Rt12.73
LM3b	327.1449_12.86	2	4 (2 proposed)	(2) (3) (5)	LMN-O-GlcA_Rt12.86
LM4	341.1242_12.7	1	2 (1 proposed)	(3)	PA-GlcA
LM5a	343.1398_12.75	2	1	(2)	DHPA-GlcA_Rt12.75
LM5b	343.1398_12.85	2	1	(2)	DHPA-GlcA_Rt12.85
LM6a	345.1555_11.28	4	2 (1 proposed)	(4) (2)	LMN-8,9-O-GlcA_Rt11.28
LM6b	345.1555_11.5	4	2 (1 proposed)	(4) (2)	LMN-8,9-O-GlcA_Rt11.5
LM7a	345.1555_11.65	4	2 (1 proposed)	(4) (2)	LMN-1,2-O-GlcA_Rt11.65
LM7b	345.1555_8.23	4	2 (1 proposed)	(4) (2)	LMN-1,2-O-GlcA_Rt8.23
LM8a	359.1348_10.53	4	1 (1 proposed)	(2)	DHPA-2-O-GlcA_Rt10.53
LM8b	359.1348_11.11	4	1 (1 proposed)	(2)	DHPA-2-O-GlcA_Rt11.11
LM8c	359.1348_7.14	4	1 (1 proposed)	(2)	DHPA-2-O-GlcA_Rt7.14
LM8d	359.1348_7.24	4	1 (1 proposed)	(2)	DHPA-2-O-GlcA_Rt7.24
LM9a	375.1297_4.53	2	1 (1 proposed)	(2)	PA-8,9-O-GlcA_Rt4.53
LM9b	375.1297_5.91	2	1 (1 proposed)	(2)	PA-8,9-O-GlcA_Rt5.91

3.3. Parent compound confirmation (Step 2)

After matching the theoretical m/z and Rt, 17 out of 18 human LMs were also found in mouse urine after exposure to unlabeled limonene (Table S5), which indicates similarity in LM metabolites in both species. In the LC-MS results from urine collected after limonene-d₅ exposure, labelled peaks (mass error < 5 mDa) were found for all 17 detected LMs, indicating the integration of limonene-d₅ into its downstream metabolites and thus validating the assignments. Human urinary metabolite DHPA (LM1) was absent from mouse urine, but its conjugate, DHPA-GlcA (LM5a and LM5b), was detected in both non-labeled and labeled forms in mouse urine.

3.4. Kinetics of LMs after limonene inhalation (Step 3)

Figure 2 displays the time courses for urinary levels of the 18 LMs after limonene inhalation. These levels were relatively quantified based on peak intensity normalized to creatinine. Generally, all LMs exhibited a similar trend: starting at relatively low levels before exposure (at 0 h), increasing immediately after exposure, reaching a peak at 0.5 or 1 h, and then gradually decreasing in time. However, peak levels varied dramatically, with the maximum intensity ranging from approximately 2,000 counts (LM2a) to 2,000,000 counts (LM6a) (Table S4). The three most abundant LMs, LM6a, LM6b and LM5b, contributed over 75% of the total response, consistent with the report of their aglycones (LMN-8,9-OH and DHPA) as major human urinary LM after enzymatic hydrolysis (Schmidt and Göen, 2017b).

Figure 2.

Time courses of 18 limonene metabolites in human urine collected after limonene inhalation. A. LM1 (DHPA); B. LM2a (DHPA-2-OH_Rt11.88); C. LM2b (DHPA-2-OH_Rt12.69); D. LM3a (LMN-O-GlcA_Rt12.73); E. LM3b (LMN-O-GlcA_Rt12.86); F. LM4 (PA-GlcA); G. LM5a (DHPA-GlcA_Rt12.75); H. LM5b (DHPA-GlcA_Rt12.85); I. LM6a (LMN-8,9-O-GlcA_Rt11.28); J. LM6b (LMN-8,9-O-GlcA_Rt11.5); K. LM7a (LMN-1,2-O-GlcA_Rt11.65); L. LM7b (LMN-1,2-O-GlcA_Rt8.23); M. LM8a (DHPA-2-O-GlcA_Rt10.53); N. LM8b (DHPA-2-O-GlcA_Rt11.11); O. LM8c (DHPA-2-O-GlcA_Rt7.14); P. LM8d (DHPA-2-O-GlcA_Rt7.24); Q. LM9a (PA-8,9-O-GlcA_Rt4.53); R. LM9b (PA-8,9-O-GlcA_Rt5.91).

Human urine samples were collected before and after limonene inhalation, and analyzed using UPLC-Q TOF MS. For each limonene metabolite, peak responses were normalized to urinary creatinine levels and plotted. Filled circles and error bars are mean and SE respectively. Student’s t-test was performed after log-transformation. (*: p< 0.05).

Results of kinetic analysis of the LMs are presented in Table 2. Fitting curves for the 3 most abundant LMs for each participant were shown at Figure S3. The fraction of specific metabolite over absorbed limonene (F) varied greatly, with values ranging from 0.3 to over 100. The F value was generally consistent with proportion of peak response for each LM (Figure S4A). LM6a, LM6b and LM5b had the largest F values (accounting for >75% of total F value), confirming that they are major urinary metabolites following limonene inhalation. The parameter Ka, which represents the rate at which the metabolite is generated, was associated with the observed peak time of each LM. As expected, LM with a higher Ka (> 4 h⁻¹) peaked earlier, at 0.5 h, while those with a lower Ka (< 4h⁻¹) peaked later, at 1 hour (Figure S4B). Conversely, the parameter Ke, representing the rate at which the metabolite is eliminated, exhibited minimal variation among different LMs. The elimination t_1/2 of each LM ranged from 0.7 to 2.7 hours.

Table 2.

Urinary limonene metabolite kinetics parameter estimation based on a pseudo one-compartment model.

Metabolite ID	F	Ka (h−1)	Observed Tmax (h)	Ke (h−1)	Elimination t1/2(h)
LM1	0.4 ± 0.1	0.93 ± 0.37	1	0.68 ± 0.27	1.01 ± 0.40
LM2a	0.2 ± 0.1	0.86 ± 0.33	1	0.80 ± 0.31	0.87 ± 0.33
LM2b	0.2 ± 0.1	2.54 ± 0.80	1	0.44 ± 0.10	1.57 ± 0.34
LM3a	1.0 ± 0.2	9.38 ± 9.10	0.5	0.99 ± 0.20	0.70 ± 0.14
LM3b	1.0 ± 0.2	9.83 ± 31.25	0.5	0.62 ± 0.12	1.12 ± 0.23
LM4	7.1 ± 1.2	15.03 ± 23.82	0.5	0.91 ± 0.08	0.76 ± 0.07
LM5a	7.0 ± 2.6	0.70 ± 0.23	1	0.86 ± 0.29	0.80 ± 0.27
LM5b	20.8 ± 3.6	1.36 ± 0.33	1	0.62 ± 0.09	1.12 ± 0.17
LM6a	109.4 ± 16.5	6.01 ± 1.79	0.5	0.78 ± 0.09	0.89 ± 0.10
LM6b	39.9 ± 9.6	8.91 ± 3.46	0.5	0.79 ± 0.08	0.87 ± 0.09
LM7a	4.9 ± 1.1	4.97 ± 1.49	0.5	0.81 ± 0.07	0.86 ± 0.07
LM7b	10.4 ± 3.8	0.90 ± 0.31	1	0.51 ± 0.18	1.35 ± 0.47
LM8a	2.0 ± 0.6	1.00 ± 0.36	1	0.53 ± 0.18	1.31 ± 0.44
LM8b	0.3 ± 0.1	2.53 ± 0.78	1	0.70 ± 0.17	0.99 ± 0.24
LM8c	0.6 ± 0.1	1.37 ± 0.33	1	0.54 ± 0.10	1.29 ± 0.24
LM8d	0.8 ± 0.2	1.49 ± 0.58	1	0.25 ± 0.12	2.73 ± 1.24
LM9a	0.5 ± 0.1	1.84 ± 0.50	1	0.44 ± 0.10	1.58 ± 0.36
LM9b	6.1 ± 1.5	1.30 ± 0.36	1	0.50 ± 0.13	1.38 ± 0.36

3.5. LM levels after exposure to greenness (Step 4)

Among the 18 LMs tested, 14 were detected in urine samples after exposure to greenness (Table S6). The remaining 4 LMs, which were minor metabolites constituting less than 0.5% of the overall LM response when inhaling limonene, could not be detected. Notably, despite the 24hr restriction from limonene-rich foods, personal, and household care products, and avoiding other sources of limonene, there were still measurable and highly variable baseline LM levels in the urine of all participants (Figures 3 and S5). Five LMs, including all three most abundant LMs in the inhalation urine, exhibited significant differences in their levels over the baseline following exposure to greenness. As depicted in Figure 3, the levels of these three LMs significantly increased (>3x fold-increase in the response) after 4 h of greenness exposure, suggesting they could be biomarkers for exposure to greenness. Figure S5 displays representative chromatograms of these 3 LMs in human urine samples obtained before and after exposure to either limonene inhalation or greenness.

Figure 3.

Levels of 3 abundant limonene metabolites in human urine collected before and after exposure to greenness. A. LM5b (DHPA-GlcA_Rt12.85); B. LM6a (LMN-8,9-O-GlcA_Rt11.28); C. LM6b (LMN-8,9-O-GlcA_Rt11.5).

Human urine samples were collected before and after 4 hours of forest bathing, and analyzed using UPLC-Q TOF MS. For each limonene metabolite, peak responses were normalized to urinary creatinine levels, and plotted. The data points from each individual participant were connected with a line to show the trend over time. Filled circles and error bars: mean and SE; empty circles with lines: individual values (subject IDs labeled aside); solid lines: male subject; dashed lines: female subject. Student’s t-test was performed after log-transformation. (*: p< 0.05).

3.6. Exposure impact on LM ratio (Step 4)

The changes in the LM6/LM5b ratio (LM6 = LM6a + LM6b) following limonene inhalation are presented in Figure 4A. The ratio exhibited a time-dependent pattern similar to the urinary levels of LM6a (Figure 2I) and LM6b (Figure 2J): starting at 3.2 before exposure (0 h), reaching a peak at 15.6 immediately after exposure (0.5 h), and gradually decreasing to 3.8 within 4 h. However, a significant distinction was observed: while the urinary levels of LM6a and LM6b exhibited significant increases at each time point following inhalation, the LM6/LM5b ratio displayed a significant difference only immediately after exposure at 0.5, 1, and 2 h. This suggests that the LM6/LM5b ratio increase only with recent limonene exposure. The increase in the LM6/LM5b ratio after exposure to greenness is shown in Figure 4B. The average ratio increased from 6.9 before the exposure to 11.1 after 4 h of greenness exposure, validating usefulness of this ratio as a more accurate time tag for recent exposure to greenness.

Figure 4.

Ratio of LMN-8,9-O-GlcA (LM6a and LM6b) over DHPA-GlcA (LM5b) after limonene inhalation or greenness exposure. A. LM ratio after limonene inhalation (*: p< 0.05); B. LM ratio after exposure to greenness. Filled circles and error bars: mean and SE; empty circles with lines: individual values (subject IDs labelled aside); solid lines: male subject; dashed lines: female subject (*: p< 0.05).; C. LM Levels after Limonene inhalation in log scale. Red line: LM5b; blue line: LM6 (LM6a + LM6b); The lengths of dashed lines show the difference of log values of the two metabolites at the same time points and correspond to the ratios of the two metabolites [log(LM6/LM5b) = log(LM6)-Log(LM5b)].

3.7. Structure confirmation (Step 4)

To validate structures of the three potential biomarkers (LM6a, LM6b, and LM 5b), their MS/MS spectra were compared with those previously reported. Figure S6 demonstrates that LM6a (A) and LM6b (B) exhibited nearly identical MS/MS spectra. They encompassed all the reported fragments of LMN-8,9-O-GlcA, including 9 from Andersen et al. (Andersen et al., 2014) and 12 from de Alvarenga et al. (Rinaldi de Alvarenga et al., 2022) The highest peaks in the spectra, namely m/z 113.0225, 85.0278, and 75.0075, were also previously reported as the most prominent peaks (Rinaldi de Alvarenga et al., 2022). Likewise, the MS/MS of LM5b (C) contained all the fragments reported in prior studies, comprising 18 from Andersen et al. (Andersen et al., 2014) and 7 from de Alvarenga et al (Rinaldi de Alvarenga et al., 2022). Taken together, the structure assignments of the three potential biomarkers were supported by the MS/MS results reported in earlier studies.

4. Discussion

To the best of our knowledge this study is the first investigation on human phase II metabolites after limonene inhalation or greenness exposure. Using the ILGA workflow, a scouting library of LMs was created, and LC-MS analysis of human urine identified 18 LM isomers after limonene inhalation. Their kinetic parameters were estimated based on model fitting. The most abundant LMs found in urine were LM5b, LM6a, and LM6b. Greenness exposure significantly increased the levels and ratios of these LMs. Uroterpenol glucuronides (LM6a and LM6b) and dihydroperillic acid glucuronide (LM5b) can potentially be used as biomarkers for greenness exposure, with a higher LM6/LM5b ratio indicating recent exposure.

Limonene was selected since it is a ubiquitous BVOC emitted by several plants, including many species of trees in our chosen study site (Clement et al., 1990; Geron et al., 2000). Limonene has a very short half-life in blood (t_1/2α ~ 3 min) (Falk-Filipsson et al., 1993), making it unsuitable as a biomarker outside of a laboratory setting. In fact, previous result did not show an increase of limonene blood level after exposure to greenness (Bach et al., 2021). On the other hand, LM levels are much higher in urine than in blood after limonene administration (Schmidt and Göen, 2017b). Many LMs, especially glucuronides, are readily amenable for highly sensitive LC-ESI-MS detection (Rinaldi de Alvarenga et al., 2022). Therefore, urine was selected as the biofluid of choice and UPLC-Q TOF-MS was used to provide enhanced coverage, sensitivity, and specificity (Xiao et al., 2012; Zhang et al., 2012). Indeed, with the highly sensitive analytical method, we detected 18 LM isomers in limonene inhalation urine, and 14 of them were found in urine after exposure to greenness. As expected, majority of them were glucuronides (Table 1).

Kinetics play a critical role in biomarker research and should be carefully considered (Verhagen et al., 2004; Zare Jeddi et al., 2021). The results of our kinetic analysis were consistent with those obtained from relative quantification. As shown in Figure S4A, there was a strong linear correlation (R² = 0.993) between the F values (obtained from kinetic modeling) and the average values of observed proportion (obtained from relative quantification). Similarly, Figure S4B demonstrates consistent results between the Ka values (from kinetic modeling) and observed peak time (Observed T_max, from relative quantification), both of which are indicative of the generation rate of the metabolite. These findings validate the reliability of the parameters derived from kinetic modeling.

Notably, the Ka values were found to be associated with the structures of LMs. Specifically, similar Ka values were often observed among isomers, such as the pairs of LMN-O-GlcA (LM3a and LM3b), LMN-8,9-O-GlcA (LM6a and LM6b), and DHPA-2-O-GlcA (LM8c and LM8d) (Table 2). These associations likely stem from the shared metabolic pathway responsible for the formation of these metabolites. In addition, all metabolites related to DHPA (LM1, LM2a, LM2b, LM5a, LM5b, LM8a, LM8b, LM8c and LM8d), exhibited relatively low Ka values, indicating slower generation compared to other metabolites (Figure S4B). These metabolites are formed via oxidation of the methyl sidechain and reduction of the cyclohexenyl ring, potentially catalyzed by cytochrome P450, aldo-keto reductase, and other oxidoreductases (Figure S1) (Kanehisa et al., 2023; Shimada et al., 2002). Further investigation is warranted to elucidate the specific step that contributes to the slower generation of this group of metabolites.

The kinetic variations among the three abundant LMs resulted in time-dependent changes in their ratio, making it a time tag for recent exposure. Despite similar Ke values (elimination rate), these metabolites exhibited significantly different Ka values (generation rate). LM5b, a metabolite structurally related to DHPA, displayed a lower Ka value, peaking later (observed T_max ~1 h), whereas LM6a and LM6b, had higher Ka values, peaking earlier (observed T_max ~0.5 h) (Table 2). Combining LM6a and LM6b as LM6, Figure 4C illustrates the comparison of the ratio of LM6/LM5b and their levels after limonene inhalation. Immediately after the exposure, LM6 was generated faster than LM5b, leading to dramatical increases of both the ratio and their levels at 30 min. However, at the later time points (after 2 h), their levels declined at a similar rate (attributed to similar Ke values), resulting in little change in the LM6/LM5b ratio.

For assessment of limonene exposure, we considered two characteristics: metabolite levels and their ratio. Metabolite levels are related to the overall extent of exposure, encompassing factors such as limonene concentration. In contrast, the ratio provides a time-related marker independent from exposure intensity. It serves as a temporal overview of an individual’s exposure to greenness. This is particularly valuable in distinguishing between short-term and long-term effects of exposure. Combining abundant metabolite levels (representing exposure extent) with their ratio (representing exposure timing) provides a more accurate depiction of an individual’s exposure. For example, high levels (>3x fold-change) and high ratio above 5 of abundant LMs indicate a recent (within 2 h) exposure; conversely high levels and low ratio of abundant LMs indicate an earlier exposure past 2 h (Figure 4A). The values are in good agreement with findings from the greenness exposure experiment, suggesting their potential utility as cutoff points for distinguishing between time-points. However, at the current stage, we cannot determine how non-greenness sources of limonene, which are responsible for variable baseline LM levels, influence these ratios. Therefore, further research with a larger cohort is necessary to solidify the anticipated fold-changes and to examine effects of limonene co-exposures.

This study has certain limitations that need to be addressed in future research. The dose response of limonene inhalation was not tested and the sample size for greenness exposure was relatively small, which may limit the generalizability of the findings. Additionally, more data points are required for accurate kinetic modeling. Furthermore, absolute quantification of phase II metabolites of limonene could not be performed due to the unavailability of authentic standards. In this study, the untargeted analysis, validated with QC pool controls, provided relative-quantitative results only. Method’s accuracy was not assessed due to a lack of standards. While this may affect absolute values, data trends and ratios remained consistent. Precision and accuracy, evaluated through multiple urine measurements, will be evaluated in future analyses. Further, authentic LM standards should be synthesized and utilized for precise structure confirmation and absolute quantitation. With the absolute quantitation and larger sample size, baseline LM levels can be accurately established for proper exposure assessments. Finally, supplementary information from tools like questionnaires and interviews will complement these findings by providing subjective assessment of individual exposures.

Sources of limonene other than greenness exposure were not investigated in this study. To minimize their influence, we controlled and limited exposure of this type before conducting the experiment (see Section 2.2). Despite these efforts, measurable amounts of LMs were present before exposure to greenness (Figure 3). Thus, when used alone, the LMs probably will not offer sufficient specificity as standalone biomarkers of exposure to greenness, especially in human biomonitoring population studies, where sources of limonene such as food and personal care and household products, or live setting (urban, suburban, rural etc.) cannot be eliminated. Interview-based greenness exposure assessment is useful for identifying potential sources (e.g., diet, use of personal care and household products etc.), but is not quantitative for personal exposure assessment. Our method provides objective quantitative measures of limonene exposure which will be invaluable in determining its effect on health outcomes at individual level. Combining these with subjective questionnaire data (e.g., self-reported walking times to the nearest green space) and objective tools, including NDVI or GPS data indicating the physical activity around the greenspace, would provide a multi-dimensional metric to accurately describe the extent and timing of one’s individual exposures to greenness and establishing cause-and-effect relationships in human studies. Moreover, the LMs together with other BVOC metabolites can lead to better description of the plant species within the exposure area. Hence, this study provides a basis for future investigations of additional BVOC metabolite biomarkers.

5. Conclusions

In this study, human urinary metabolites were profiled after limonene inhalation, and their kinetics were investigated. In addition, the study reported here reveals the relationships between greenness exposure and urinary levels of LMs, and between the temporal proximity of the exposure and the LM ratio. Based on these findings, we proposed that DHPA-GlcA and LMN-8,9-O-GlcA could serve as individual-level measures of exposure to greenness, especially in combination with other study tools. The framework presented in this study could be expanded to examine relationship between greenness exposure and metabolites of other BVOCs. The discovery of new exposure biomarkers will enhance our understanding of benefits of greenness at individual level and facilitate our effort to improve the environment in the future.

• UPLC-QTOF-MS was used to measure urinary metabolites of limonene

• Eighteen human urinary metabolites were found and characterized

• Top three compounds elevated after exposure; their ratio indicates recent exposure

• Those 3 metabolites can serve as biomarkers of individual exposure to greenness

• They can complement other study measures to describe effects of vegetation exposure

Abbreviations:

BVOC

biogenic volatile organic compounds

ILGA

integrated library-guided analysis

UPLC-QTOF-MS

ultra-performance liquid chromatography-quadrupole time-of-flight mass spectrometry

DIA

data-independent acquisition

LMs

limonene metabolites

NDVI

Normalized Difference Vegetation Index

Rt

retention time

NLME

nonlinear mixed effect model