Apiaceous Vegetable Consumption Decreases PhIP-Induced DNA Adducts and Increases Methylated PhIP Metabolites in the Urine Metabolome in Rats^,,

Abstract

Background: Heterocyclic aromatic amines, such as 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), are carcinogenic compounds produced during heating of protein-containing foods. Apiaceous vegetables inhibit PhIP-activating enzymes, whereas cruciferous vegetables induce both PhIP-activating and -detoxifying enzymes.

Objective: We investigated the effects of these vegetables, either alone or combined, on PhIP metabolism and colonic DNA adduct formation in rats.

Methods: Male Wistar rats were fed cruciferous vegetables (21%, wt:wt), apiaceous vegetables (21%, wt:wt), or a combination of both vegetables (10.5% wt:wt of each). Negative and positive control groups were fed an AIN-93G diet. After 6 d, all groups received an intraperitoneal injection of PhIP (10 mg · kg body weight⁻¹) except for the negative control group, which received only vehicle. Urine was collected for 24 h after the injection for LC–tandem mass spectrometry metabolomic analyses. On day 7, rats were killed and tissues processed.

Results: Compared with the positive control, cruciferous vegetables increased the activity of hepatic PhIP-activating enzymes [39.5% and 45.1% for cytochrome P450 (CYP) 1A1 (P = 0.0006) and CYP1A2 (P < 0.0001), respectively] and of uridine 5′-diphospho-glucuronosyltransferase 1A (PhIP-detoxifying) by 24.5% (P = 0.0267). Apiaceous vegetables did not inhibit PhIP-activating enzymes, yet reduced colonic PhIP-DNA adducts by 20.4% (P = 0.0496). Metabolomic analyses indicated that apiaceous vegetables increased the relative abundance of urinary methylated PhIP metabolites. The sum of these methylated metabolites inversely correlated with colonic PhIP-DNA adducts (r = −0.43, P = 0.01). We detected a novel methylated urinary PhIP metabolite and demonstrated that methylated metabolites are produced in the human liver S9 fraction.

Conclusions: Apiaceous vegetables did not inhibit the activity of PhIP-activating enzymes in rats, suggesting that the reduction in PhIP-DNA adducts may involve other pathways. Further investigation of the importance of PhIP methylation in carcinogen metabolism is warranted, given the inverse correlation of methylated PhIP metabolites with a biomarker of carcinogenesis and the detection of a novel methylated PhIP metabolite.

Introduction

Heterocyclic aromatic amines (HAAs)⁴ are a group of carcinogenic compounds that are produced during heating of protein-containing foods. Exposure of humans to HAAs is mainly through consumption of overcooked meats and fish (1). Several epidemiologic studies have shown a positive association between HAA intake and colorectal cancer risk (2, 3). Of the identified HAAs, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine (PhIP), is the most abundant HAA in grilled meat (1). PhIP induces chromosomal aberrations and sister-chromatid exchanges (4), forms PhIP-DNA adducts and aberrant crypt foci in colon mucosa (5), and causes colon and mammary carcinomas in multiple animal models (6, 7).

PhIP itself is not genotoxic but requires activation in order to become carcinogenic. Cytochrome P450s (CYPs; primarily hepatic CYP1A2) are the enzymes that generate N²-hydroxy-PhIP, the metabolite responsible for DNA adduct formation (8). This CYP1A2-mediated hydroxylation of the exocyclic amine of PhIP is known as an initial activation step because the intermediate can be further esterified, mainly by N-acetyltransferase (NAT) 2 and sulfotransferase (SULT) 1A1 to N²-acetoxy-PhIP and N²-sulfonyloxy-PhIP, respectively. These esterified metabolites create PhIP-DNA adducts via electrophilic nitrenium ion formation (9–11). In contrast, conjugation by uridine 5′-diphospho-glucuronosyltransferase (UGT) 1A1 and glutathione S-transferase (GST) A1-1 usually leads to the production of unreactive metabolites and efficient excretion of PhIP (12–14).

Bioactive compounds present in various vegetables play an important role in reducing the risk of carcinogenesis (15). One of the proposed mechanisms is modulating the enzymes responsible for metabolizing carcinogens. Specifically, compounds present in apiaceous vegetables (e.g., celery and parsnips) inhibit CYP1As (16), whereas cruciferous vegetables (e.g., broccoli and cabbage) and the bioactive compounds they contain induce both activating (CYP1As) and detoxifying (GSTs and UGTs) enzymes (17–19). Of note, when these 2 vegetable families were provided together in a human feeding study, apiaceous vegetables prevented the induction of CYP1A2 activity by cruciferous vegetables, indicating that the inhibiting effects of apiaceous vegetables on CYP1A2 activity were more pronounced than the inducing effects of cruciferous vegetables (20). However, it has not been determined whether the combination of apiaceous and cruciferous vegetables will result in a synergistic detoxification of PhIP due to their complementary actions on activating and detoxifying biotransformation enzymes. Therefore, we investigated the effects of cruciferous and apiaceous vegetables and the combination of both vegetable families on PhIP metabolism by analyzing PhIP-metabolizing enzyme activities and their protein expression, PhIP urinary metabolite profiles, and PhIP-DNA adduct levels in the colon as a marker of carcinogenicity.

Methods

Animals and diets

Male Wistar rats, 100–125 g body weight (4–5 wk old), were obtained from Harlan Laboratories and housed in wire-bottom stainless steel cages. Rats had free access to water and diets. Animal handling and housing followed NIH guidelines, and experimental procedures were approved by the University of Minnesota Animal Care and Use Committee.

The AIN-93G diet was fed to the negative and positive control groups. For the vegetable-supplemented diets, organically grown vegetables were purchased in one batch from a local market. A mixture of vegetables from each botanical family was chosen due to each vegetable differing in its predominant phytochemical and, in the case of cruciferous vegetables, because of evidence of equivalent biological response to different cruciferous vegetables in a colon carcinogenesis model (21). The cruciferous vegetable–supplemented diet (CRU diet) included fresh watercress, broccoli, and cabbage (21%, wet wt:wt; 7% per each vegetable); the apiaceous vegetable–supplemented diet (API diet) consisted of fresh celery and parsnips (21%, wet wt:wt; 10.5% per each vegetable); and the combined supplementation diet (CRU+API diet) included all of the aforementioned vegetables (10.5% cruciferous vegetables and 10.5% apiaceous vegetables, wet wt:wt). Fresh vegetables were used because of evidence that lyophilization potentially decreases the chemopreventive efficacy of cruciferous vegetables (21). Fresh vegetables were ground by using a food processor (Cuisinart Deluxe) and mixed into the powdered diet (i.e., AIN-93G) for 15 min with a mechanical mixer (model A-200-D; Hobart). We previously determined that this duration of diet mixing avoids heating the diet (thus minimizing risk of oxidation), yet achieves a homogenous incorporation of dietary components, including vegetables. Vegetable-supplemented diets were then manually mixed again and visually inspected to confirm homogeneous incorporation. Diets were divided into separate plastic bags for each day of the feeding period, stored at −80°C, and thawed before feeding. All vegetable-supplemented diets were balanced for macronutrients (Supplemental Table 1). Food intake was measured over 24 h on 3 separate days.

Experimental design

Initially, all rats were fed the AIN-93G purified diet for 5 d to adapt to a powdered diet and to allow elimination of any trace of plant bioactive compounds. Rats were randomly assigned into 5 groups, as 6 blocks of 8 rats and 1 block of 4 rats. The study groups were as follows: negative control (n = 11), positive control (n = 11), CRU diet (n = 10), API diet (n = 10), and CRU+API diet (n = 10). After 6 d of feeding, the positive control group and all 3 vegetable-supplemented groups received an intraperitoneal injection of PhIP (10 mg PhIP · kg body weight⁻¹; dissolved in dimethyl sulfoxide). PhIP was delivered through intraperitoneal injection because of differing estimates of bioavailability of PhIP when administered orally (22, 23). The negative control group received dimethyl sulfoxide only. Immediately after injection, the rats were placed in stainless steel metabolic cages and urine was collected. Rats were feed deprived for 12 h before being killed. Upon completion of a 24-h urine collection , the rats were anesthetized with isoflurane and then killed by exanguination. Liver and colon were harvested and stored at −80°C.

Analysis of bioactive compounds in cruciferous and apiaceous vegetables

Total glucosinolates in cruciferous vegetables were analyzed by using a spectrophotometric method, as previously described (24). For apiaceous vegetables, 3 furanocoumarins [5-methoxypsoralen (5-MOP), 8-methoxypsoralen (8-MOP), and isopimpinellin] were analyzed by using HPLC equipped with a UV diode array detector as described previously (25). The recovery was calculated by using one of the furanocoumarin standards (i.e., 8-MOP). The recovery of added 8-MOP from celery and parsnips was 83.3 ± 6.55% and 87.1 ± 2.72%, respectively. The final quantification of furanocoumarins was adjusted by using these recovery values.

Preparation of hepatic microsomes and cytosol

Hepatic microsomes and cytosol were isolated from fresh tissue as described elsewhere (26). Cytosolic and microsomal preparations were stored at −80°C until further analysis.

Measurement of biotransformation enzyme activity

CYP1A1 and CYP1A2 activity.

Stored microsomal fractions were washed by ultracentrifugation at 105,000 × g with 100 mM sodium pyrophosphate–10 mM EDTA buffer (pH 7.8) for 70 min. Pellets were resuspended in 58 mM Tris base buffer (pH 7.5) and homogenized in advance of measuring CYP1A1 and CYP1A2 activity. Microsomal CYP1A1 and CYP1A2 activity was measured by using 7-ethoxyresorfurin O-de-ethylation and 7-methoxyresorufin O-demethylation assays, respectively (27, 28). The reaction mixture was separated into aliquots into 96-well microtiter plates (4 wells/sample) and fluorescence measured by using a microplate reader (Biotek; excitation: 530/25 nm; emission: 590/35 nm).

SULT1A1 activity.

Cytosolic SULT1A1 activity was measured by a spectrophotometric method (29) that is effective for probing rat SULT1A1 (30).

UGT1A activity.

Microsomal fractions were washed as described above and then microsomal UGT1A activity was determined by using a previously described method (31) with slight modifications. The reaction mixture (200 μL/sample) included 600 μM 4-methylumbelliferone, 5 mM uridine 5′-diphospho-glucuronic acid, and 4 mM magnesium chloride in 100 mM phosphate buffer (pH 7.4) with 50 μg of microsomal protein. The reaction was conducted for 90 min at 37°C and was terminated by the addition of 24% perchloric acid. The mixture was then centrifuged at 5000 × g for 10 min. To analyze the final product of the reaction (i.e., 4-methylumbelliferone glucuronide), 20 μL of the supernatant fraction was injected into an HPLC system equipped with a UV detector (Gilson) and 4-methylumbelliferone glucuronide was separated on an Ultrasphere ODS C18 column (4.6 × 250 mm, 5 μm; Hichrom). The initial mobile phase composition was 15% acetonitrile in 25 mM phosphate buffer (pH 3.1) and gradually increased to 50% acetonitrile over 5 min at a flow rate of 1.0 mL · min⁻¹. Column eluent was monitored at 316 nm, and 4-methylumbelliferone glucuronide was quantitated by comparison to an authentic standard. Herein, we use the less specific term of UGT1A activity because, in rats, 4-methylumbelliferone is metabolized by UGT1A1 and other UGT subfamilies [e.g., UGT1A6 (32, 33)] despite 4-methylumbelliferone being specific for UGT1A1 activity in humans (31).

Protein quantification.

Protein concentrations of liver microsomes and cytosol preparations were measured by the Bradford method (34) by using a commercial kit (Sigma-Aldrich).

Measurement of biotransformation enzyme protein expression

Protein expression of CYP1A1, CYP1A2, SULT1A1, and UGT1A1 in liver was determined by ELISA with the use of commercial kits (Uscn Life Science) according to the manufacturer’s instructions. Washed microsomes were used to measure CYP1A1, CYP1A2, and UGT1A1 expression, and the cytosolic fraction was used for measuring SULT1A1 expression.

LC-electrospray ionization-tandem MS analysis of urinary PhIP metabolites

Urinary PhIP metabolites were identified as previously described (35) by LC-tandem MS (MS/MS) analysis by using an ultra-performance LC quadrupole-time-of-flight system (Waters). The structural identity of metabolites was elucidated on the basis of the fragmentation pattern from MS/MS analysis and confirmed by comparison with reported fragmentation patterns (35). The relative proportion of urinary PhIP metabolites was calculated by using the percentage of the peak area of each metabolite per total peak area (i.e., PhIP and all identified PhIP metabolites).

Methylation of PhIP in human liver S9 fraction

The final reaction mixture contained 34 μM of S-adenosyl methionine, 1 mM of PhIP, and 60 μg human liver S9 fraction protein (XenoTech) in 50 mM Tris base buffer (pH 7.5). After 60 min of incubation at 37°C, the reaction was terminated by adding 50% acetonitrile and then centrifuged at 18,000 × g for 12 min to remove proteins and particles. The supernatant was analyzed for methylated PhIP metabolites by using the quadrupole-time-of-flight system as described above.

Measurement of DNA adducts

Preparation of colon tissue.

After harvest, colon tissue was dissected and flushed with ice-cold PBS (pH 7.4). The tissue was trimmed to remove the mesentery and opened lengthwise. Harvested colons were frozen in liquid nitrogen and stored at −80°C.

Isolation and enzymatic digestion of DNA.

DNA was isolated from rat colon by using the Qiagen Gentra Puregene DNA Purification Kit following the manufacturer’s instructions. Isotopically labeled internal standard, [¹³C]-dG-C8-PhIP, was added to the DNA extracts at a level of 10 adducts per 10⁷ nucleotides before enzymatic digestion. Isolated DNA (50 μg) was treated with DNase I, followed by nuclease P1, and then further digested by using alkaline phosphatase and phosphodiesterase, under conditions described elsewhere (36). The efficacy of enzymatic digestion was confirmed by HPLC of the hydrolyzed DNA (4 μg), monitored at 260 nm (36). Detailed experimental conditions and a representative chromatogram of nucleosides are shown in Supplemental Figure 1.

Quantification of DNA adducts.

PhIP-DNA adduct levels in colon were determined by using an online column-switching LC-MS/MS method as described previously (36).The mass spectral parameters were optimized as described elsewhere (37).

Calibration curves and method validation.

Calibration curves were constructed by using the ratio between unlabeled PhIP-DNA adducts (ranging from 0 to 30 adducts per 10⁷ nucleotides) and isotopically labeled internal standard (i.e., [¹³C]-dG-C8-PhIP; set at 10 adducts per 10⁷ nucleotides). The R² value of the calibration curve was >0.995. The intra-assay CVs were 5.54%, 7.33%, and 6.55% for 5, 10, and 15 adducts per 10⁷ nucleotides, respectively. Detailed experimental conditions for method validation and results are shown in Supplemental Table 2.

Statistical analysis

All results are expressed as least-squares means ± SEMs. Differences between the positive control group and vegetable-supplemented groups were tested by using 1-factor ANOVA followed by multiple comparisons using the difference matrix of the least-squares mean (SAS 9.2; SAS Institute). The positive and negative control groups were compared separately by using the difference matrix of the least-squares mean. In addition to analyses by individual metabolite, urinary PhIP metabolites were also categorized into the following 5 groups for statistical analysis by overall metabolite group, depending on their metabolic pathways as well as types of biotransformation enzymes involved: glucuronidated metabolites (XI, XII, XIII, XIV, XV, XVI, and XVII), 5-hydroxylated metabolites (VI, IX, X, and XIV), sulfated metabolites (VIII, IX, and X), CYP1A-hydroxylated metabolites (IV, V, VII, VIII, X, XIII, XV, XVI, and XVII), and methylated metabolites (II, III, and VII). A P value <0.05 was considered significant. The association between PhIP and its metabolites and PhIP-DNA adducts was analyzed by using Pearson correlation coefficients.

Results

Food intake, tissue weight, and weight gain.

Visual inspection of the food cups indicated that the vegetables remained well mixed within the diet throughout each day. Overall, food intake for each of the vegetable-supplemented diet groups (i.e., CRU, API, and CRU+API diets) was slightly greater than that of the positive control group (Supplemental Table 3). Because food intake was calculated on the basis of wet weight, the greater food intake for the vegetable-supplemented diets was likely due to the water content of the fresh vegetables. There were no differences in body weight gain or tissue weights between any of the groups.

Amounts of bioactive compounds in vegetables.

The total glucosinolate content of the CRU diet was 183 mg · kg⁻¹ diet, and thus half that for the CRU+API diet. The total glucosinolate concentration was highest in watercress followed by cabbage and broccoli. The respective concentrations of 5-MOP, 8-MOP, and isopimpinellin by fresh weight were 0.551, 0.514, and 0.688 μg · g celery⁻¹ and 11.0, 5.19, and 23.0 μg · g parsnips⁻¹. Hence, amounts of these furanocoumarins combined in the API and CRU+API diet were 4.30 mg · kg diet⁻¹ and 2.15 mg · kg diet⁻¹, respectively (Table 1).

TABLE 1.

Total glucosinolates in cruciferous vegetables and furanocoumarins in apiaceous vegetables¹

	Cruciferous vegetables			Apiaceous vegetables
	Broccoli	Cabbage	Watercress	Celery	Parsnip
Total glucosinolates, mg · g−1	0.172 ± 0.100	0.460 ± 0.0765	4.58 ± 0.166	—	—
5-MOP, μg · g−1	—	—	—	0.551 ± 0.0146	11.0 ± 0.347
8-MOP, μg · g−1	—	—	—	0.514 ± 0.0250	5.19 ± 0.184
Isopimpinellin, μg · g−1	—	—	—	0.688 ± 0.0482	23.0 ± 0.698

¹Values are means ± SDs, n = 3. 5-MOP, 5-methoxypsoralen; 8-MOP, 8-methoxypsoralen.

Effects of vegetable intake on CYP1A1 and CYP1A2 activity and expression.

There was no difference noted between the positive control and negative control in CYP1A2 enzyme activity, although the positive control group showed 31.1% higher CYP1A1 activity compared with the negative control group. Compared with the positive control, the CRU diet increased CYP1A1 activity by 39.5% (P = 0.006) and CYP1A2 by 45.1% (P < 0.001; Table 2). The CRU+API diet also increased CYP1A1 (32.7%; P = 0.008) and CYP1A2 (58.8%; P < 0.001) activity compared with the positive control. The API diet increased CYP1A2 activity (37.9%; P = 0.0037) but not CYP1A1 activity. There was no effect on enzyme protein expression by any vegetable-supplemented diet compared with the positive control; however, CYP1A1 expression was higher in the positive control than in the negative control group (31.3%; P = 0.0411; Table 2).

TABLE 2.

Activity and protein expression of PhIP-metabolizing hepatic enzymes (CYP1As, UGT1A, and SULT1A1) after a 7-d diet intervention in Wistar rats¹

	CYP1A1		CYP1A2		UGT1A		UGT1A1		SULT1A1
	Activity, pmol resorufin · min−1 · mg protein−1	Expression, ng protein · mg total protein−1	Activity, pmol resorufin · min−1 · mg protein−1	Expression, ng protein · mg total protein−1	Activity, nmol 4-methylumbelliferone glucuronide · min−1 · mg protein−1	Expression, ng protein · mg total protein−1	Activity, nmol p-nitrophenol · min−1 · mg protein−1	Expression, ng protein · mg total protein−1
Negative control + no PhIP	7.01 ± 0.840*	0.838 ± 0.0784*	5.67 ± 0.622	4.60 ± 1.06	20.3 ± 0.833	1.55 ± 0.177*	11.7 ± 0.455	0.206 ± 0.0367
Positive control + PhIP	9.19 ± 0.689c	1.10 ± 0.0912	6.36 ± 0. 527b	5.92 ± 0.645	16.3 ± 1.27b	1.79 ± 0.224	11.7 ± 0.441	0.171 ± 0.0190b
CRU diet + PhIP	12.8 ± 0.698a	1.13 ± 0.0924	9.23 ± 0.534a	5.93 ± 0.654	20.3 ± 1.28a	1.36 ± 0.227	11.7 ± 0.451	0.261 ± 0.0192a
API diet + PhIP	10.6 ± 0.681b,c	1.16 ± 0.0902	8.77 ± 0.521a	7.72 ± 0.639	21.7 ± 1.23a	1.93 ± 0.221	11.0 ± 0.441	0.225 ± 0.0188a,b
CRU+API diet + PhIP	12.2 ± 0.738a,b	1.25 ± 0.0902	10.1 ± 0.565a	7.39 ± 0.639	18.4 ± 1.25a,b	1.84 ± 0.221	11.7 ± 0.441	0.210 ± 0.0188a,b

¹Values are least-squares means ± SEMs, n = 10–11. *Different from positive control, P < 0.05. Labeled means in a column that do not share a letter differ, P < 0.05 (excluding the negative control from analyses). API, apiaceous vegetable–supplemented; CRU, cruciferous vegetable–supplemented; CRU+API, combined cruciferous and apiaceous vegetable–supplemented; CYP, cytochrome P450; PhIP, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine; SULT, sulfotransferase; UGT, uridine 5′-diphospho-glucuronosyltransferase.

Effects of vegetable intake on UGT1A and SULT1A1 activity and expression.

There was no difference between the positive control and negative control in UGT1A and SULT1A1 enzyme activity (Table 2). Compared with the positive control, both the CRU and API diets increased UGT1A activity [24.5% (P = 0.0267) and 33.1% (P = 0.006), respectively], whereas there was a trend for increased activity by the combination group (P = 0.06). There was no difference in UGT1A1 protein expression among the vegetable groups. SULT1A1 protein expression increased by 52.6% with the CRU diet (P = 0.0015) compared with the positive control, yet no effect on SULT1A1 activity was observed. The API and CRU+API diets had no effect on SULT1A1 activity or protein expression.

Urinary PhIP metabolites.

We identified PhIP (parent compound) and 16 PhIP metabolites in urine samples. Structures of PhIP urinary metabolites and their proposed metabolic pathways are summarized in Figure 1. Metabolites and their retention times are summarized in Supplemental Table 4. Of the 16 metabolites, 7 metabolites were glucuronidated (XI, XII, XIII, XIV, XV, XVI, and XVII), 3 metabolites were sulfated (VIII, IX, and X), and 3 metabolites were methylated (II, III, and VII). Interestingly, metabolite II did not match any previously reported fragmentation pattern. This newly detected metabolite II as well as metabolite III were also produced when PhIP was incubated with human liver S9 fraction (Supplemental Figure 2). The API diet increased the relative proportion of 2 metabolites (VII and XII) and decreased 1 metabolite (IV) compared with the positive control group. For the CRU diet group compared with the positive control, 2 of the 3 methylated metabolites (II and VII) had a lower relative proportion, whereas the proportions of 5 glucuronidated (XII, XIV, XV, XVI, and XVII) metabolites and a 5-hydroxylated metabolite (X) were higher. PhIP (I) and 5 metabolites (III, V, VI, XI, and XIII) did not differ from the positive control regardless of which vegetables were supplemented (Table 3).

FIGURE 1.

Structures of urinary PhIP metabolites and metabolic pathways. Solid arrows indicate either CYP-dependent oxidation reactions or phase II–dependent conjugation reactions. The dashed arrow represents a CYP-independent oxidation reaction. The structure of metabolite II has yet to be elucidated. CYP, cytochrome P450; NAT, N-acetyltransferase; NMT, N-methyltransferase; PhIP, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine; SULT, sulfotransferase; UGT, uridine 5′-diphospho-glucuronosyltransferase. Reproduced from reference 35 with permission.

TABLE 3.

UPLC-QTOF-MS/MS results of relative proportions of urinary PhIP metabolites in Wistar rats fed CRU, API, and CRU+API diets¹

	Relative abundance, %			Fold of positive control
	Positive control	CRU	API	CRU + API
Individual metabolites
I (PhIP)	16.7 ± 1.11	0.886 ± 0.0671	0.968 ± 0.0656	1.04 ± 0.0656
II	9.79 ± 0.383a	0.865 ± 0.0397b	1.07 ± 0.0387a	0.860 ± 0.0387b
III	13.1 ± 0.764	0.887 ± 0.0590	0.984 ± 0.0576	0.850 ± 0.0576
IV	16.4 ± 0.778a	0.999 ± 0.0480a	0.737 ± 0.0469b	0.935 ± 0.0469a
V	0.823 ± 0.240	0.608 ± 0.291	0.576 ± 0.288	0.473 ± 0.288
VI	2.41 ± 0.134a,b	1.08 ± 0.0562a	0.836 ± 0.0549b	1.08 ± 0.0549a
VII	22.7 ± 0.826b	0.764 ± 0.0368c	1.15 ± 0.0359a	0.761 ± 0.0359c
VIII	3.65 ± 0.347b,c	1.26 ± 0.0961a,b	0.955 ± 0.0939c	1.50 ± 0.0939a
IX	0.0833 ± 0.0336b,c	2.10 ± 0.408a,b	0.674 ± 0.399c	2.37 ± 0.399a
X	0.0134 ± 0.0270b	5.25 ± 2.04a	1.99 ± 2.00a,b	3.97 ± 1.99a,b
XI	3.07 ± 0.197	1.01 ± 0.0649	1.15 ± 0.0634	1.17 ± 0.0634
XII	0.426 ± 0.110b	2.11 ± 0.262a	2.33 ± 0.256a	2.56 ± 0.256a
XIII	0.370 ± 0.0628a,b	1.15 ± 0.172a	0.546 ± 0.168b	1.15 ± 0.168a
XIV	0.0311 ± 0.0240b	3.26 ± 0.780a	0.725 ± 0.762b	3.75 ± 0.762a
XV	0.0920 ± 0.0351b	2.46 ± 0.387a	0.50 ± 0.378b	2.28 ± 0.378a
XVI	0.980 ± 0.158b	1.69 ± 0.164a	0.951 ± 0.160b	1.24 ± 0.160a,b
XVII	9.28 ± 0.937b	1.83 ± 0.102a	1.11 ± 0.100b	1.62 ± 0.100a
Grouped metabolites
Glucuronidated2	14.2 ± 1.12b	1.64 ± 0.0794a	1.12 ± 0.0776b	1.52 ± 0.0776a
5-Hydroxylated3	2.52 ± 0.126b	1.18 ± 0.0509a	0.837 ± 0.0497b	1.18 ± 0.0497a
Sulfated4	3.72 ± 0.372b	1.30 ± 0.101a	0.954 ± 0.100b	1.54 ± 0.100a
CYP1A-hydroxylated5	54.3 ± 1.18b	1.07 ± 0.0220a	0.990 ± 0.0215b	1.02 ± 0.0215a,b
Methylated6	45.7 ± 1.31b	0.821 ± 0.0290c	1.09 ± 0.0284a	0.808 ± 0.0284c

¹Values are least-squares means ± SEMs, n = 10–11. Statistical analysis was performed by using least-squares mean of relative abundance. Labeled means in a row (positive control = 1.0) that do not share a letter differ, P < 0.05. API, apiaceous vegetable–supplemented; CRU, cruciferous vegetable–supplemented; CRU+API, combined cruciferous and apiaceous vegetable–supplemented; CYP, cytochrome P450; MS/MS, tandem mass spectrometry; PhIP, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine; QTOF, quadrupole-time-of-flight; UPLC, ultra-performance liquid chromatography.

²Glucuronidated metabolites: XI + XII + XIII + XIV + XV + XVI + XVII.

³5-Hydroxylated metabolites: VI + IX + X + XIV.

⁴Sulfated metabolites: VIII + IX + X.

⁵CYP1A-hydroxylated metabolites: IV + V + VII + VIII + X + XIII + XV + XVI + XVII.

⁶Methylated metabolites: II + III + VII.

In statistical analyses by metabolite groups based on metabolic pathways, the CRU diet increased the proportion of glucuronidated, 5-hydroxylated, sulfated, and CYP1A-hydroxylated metabolites, but decreased methylated metabolites compared with the positive control. With the exception of no change in CYP1A-hydroxylated metabolites, the CRU+API diet produced the same changes in the metabolite profile as the CRU diet. The relative abundance of methylated metabolites was increased by the API diet.

PhIP-DNA adduct levels in colon.

There was a 20.4% reduction in colonic PhIP-DNA adducts by the API diet compared with the positive control group (P = 0.049). The CRU diet trended toward a reduction in PhIP-DNA adducts (17.4% reduction; P = 0.07). No effect was noted by the CRU+API diet (Figure 2). We also assessed the correlation between groups of urinary PhIP metabolites and PhIP-DNA adducts. The relative proportions of sulfated metabolites and CYP1A-hydroxylated metabolites were positively correlated with PhIP-DNA adducts (r = 0.41 and 0.33, P < 0.05 for both), whereas the proportion of methylated metabolites was negatively correlated with PhIP-DNA adducts (r = −0.43, P < 0.05; Table 4). Generally, the correlations between the abundance of each individual urinary metabolite and PhIP-DNA adducts were in the same direction as the group of PhIP metabolites into which it had been placed (Table 4). Subsequent metabolites of 4′-hydroxy-PhIP (i.e., VIII and XV) were positively correlated with PhIP-DNA adducts (r = 0.42 and 0.36 for VIII and XV, P < 0.05 for both). The parent compound [i.e., PhIP (I)] and 4′-hydroxy-PhIP (IV), which is one of the CYP1A-hydroxylated metabolites, were also positively correlated with PhIP-DNA adducts (Table 4). Two of 3 methylated metabolites were negatively correlated with PhIP-DNA adducts (r = −0.54 and −0.61 for metabolites II and III, respectively; P < 0.001 for both; Table 4).

FIGURE 2.

Colonic PhIP-DNA adducts after a 7-d diet intervention in Wistar rats. Values are least-squares means ± SEMs, n = 10–11. Groups not sharing a common letter differ, P < 0.05 (negative control excluded from analyses). API, apiaceous vegetable–supplemented; CRU, cruciferous vegetable–supplemented; CRU+API, combined cruciferous and apiaceous vegetable–supplemented; PhIP, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine.

TABLE 4.

Correlation coefficients of PhIP-DNA adducts with individual PhIP metabolites by metabolic pathways in Wistar rats fed CRU, API, and CRU+API diets¹

Metabolite	Not metabolized	Glucuronidated metabolites2	5-Hydroxylated metabolites3	Sulfated metabolites4	CYP1A-hydroxylated metabolites5	Methylated metabolites6
I	0.32*
II						−0.54**
III						−0.61**
IV					0.42*
V					0.22
VI			0.04
VII					−0.15	−0.15
VIII				0.42*	0.42*
IX			0.29	0.29
X			0.19	0.19	0.19
XI		0.16
XII		0.14
XIII		−0.01			−0.01
XIV		0.14	0.14
XV		0.36*			0.36*
XVI		−0.13			−0.13
XVII		−0.01			−0.01
Correlation coefficients for overall metabolic groups		0.02	0.15	0.41*	0.33*	−0.43*

¹Correlations between PhIP-DNA adducts and urinary PhIP metabolites were analyzed by using Pearson correlation coefficients. *P < 0.05, **P < 0.001 (the negative control group was excluded). API, apiaceous vegetable–supplemented; CRU, cruciferous vegetable–supplemented; CRU+API, combined cruciferous and apiaceous vegetable–supplemented; CYP, cytochrome P450; PhIP, 2-amino-1-methyl-6-phenylimidazo[4,5-b]pyridine.

²Glucuronidated metabolites: XI + XII + XIII + XIV + XV + XVI + XVII.

³5-Hydroxylated metabolites: VI + IX + X + XIV.

⁴Sulfated metabolites: VIII + IX + X.

⁵CYP1A-hydroxylated metabolites: IV + V + VII + VIII + X + XIII + XV + XVI + XVII.

⁶Methylated metabolites: II + III + VII.

Discussion

To our knowledge, this is the first study to report that apiaceous vegetables decrease PhIP-DNA adducts in the colon. Vegetables may modify colon cancer risk by modifying biotransformation enzyme activity. Although CYP1As activate carcinogens such as PhIP, the increases we observed in CYP1A1 and CYP1A2 activity, as well as in UGT1A activity, with feeding of cruciferous vegetables were shown previously in humans (18) and animal models (38). However, CYP1A1, CYP1A2, and UGT1A1 protein expression did not change. Although uncommon, this lack of correlation between CYP and UGT protein expression and enzyme activity has been previously reported (39–41). There was no inhibition of CYP1A1 and CYP1A2 activity by apiaceous vegetables; indeed, CYP1A2 activity increased with the API diet. This was unexpected, given in vitro studies indicating that constituents in apiaceous vegetables inhibited human CYP1A2 activity (16, 42). Furthermore, Gwang (43) reported that CYP1A1 and CYP1A2 activities decreased after Wistar rats received 25 mg of 8-MOP · kg body weight⁻¹ intraperitoneally (greater than the ∼0.420 mg of 8-MOP · kg body weight⁻¹ and 2.94 mg of total furanocoumarins · kg body weight⁻¹ consumed in our study). CYP1A2 inhibition was also observed in humans who consumed 266 g · d⁻¹ of apiaceous vegetables (20, 44). In contrast, we fed an amount of apiaceous vegetables equivalent to the human consumption of 101 g · d⁻¹, based on an 8368-kJ intake (∼2000 kcal).Furthermore, although we used the same animal model as Gwang, we fed fresh vegetables instead of administering intraperitoneal injections of purified furanocoumarins. The bioavailability of phytochemicals can be influenced by the chemical structure within the food (e.g., glycosylation) and the food matrix (45). Moreover, intraperitoneal injections generally yield higher systemic availability by bypassing intestinal first-pass metabolism (46). It is possible that our rats received fewer bioactive compounds than in previous studies. Nevertheless, we suggest that our approach is more physiologically relevant.

The induction of UGT1A activity by the API diet was also unexpected. Few studies investigated the effects of apiaceous vegetables on UGT1A. Navarro et al. (18) observed in humans that combined consumption of cruciferous and apiaceous vegetables trended toward increased UGT1A1 activity more than eating cruciferous vegetables only, implying that apiaceous vegetables induce UGT1A1 activity. Although furanocoumarins act as mechanistic inhibitors of CYPs, they also modulate CYP protein expression via the aryl hydrocarbon receptor–mediated pathway (47, 48). Because gene expression of several UGT1As is partly regulated by the same pathway, it is possible that UGT1A was induced by furanocoumarins (49).

Urinary metabolomics showed that approximately half of the PhIP metabolites were glucuronidated, confirming previous indications that UGTs play an important role in facilitating excretion of PhIP and thereby decrease genotoxicity (50). Surprisingly, glucuronidated metabolites, as a group, had no correlation with adducts; individually, only metabolite XV was correlated with adducts but, paradoxically, in a positive direction.

We looked at the response of the other metabolites to vegetable supplementation and their correlations with adducts for an explanation as to why metabolite XV positively correlated with adducts. The CRU diet increased the sulfated and CYP1A-hydroxylated groups of metabolites and both metabolite groups positively correlated with PhIP-DNA adducts (Table 4). These positive correlations are notably driven by metabolite IV, a CYP1A-hydroxylated metabolite, and its subsequent sulfated or glucuronidated metabolites (VIII and XV), each of which individually correlated strongly with PhIP-DNA adducts. These positive correlations are biologically plausible: 1) 4′ hydroxylation and the generation of metabolite IV is the major metabolic pathway of PhIP in rodents (51, 52) and 2) as the quantity of PhIP increases, more PhIP is metabolized through activation pathways that lead to adduct formation (53). Hence, we suggest that metabolite IV is proportional to the amount of PhIP and thus reflects the quantity of parent PhIP available in the system for activation pathways and subsequent adduct formation.

Interestingly, methylated PhIP metabolites inversely correlated with PhIP-DNA adducts (Table 4). Individually, metabolites II and III both had strong inverse correlations with adducts (r = −0.54 and −0.61, P < 0.001 for both). The methylated group of metabolites overall was present in greater relative abundance with the API diet than with the positive control, the relation being driven primarily by metabolite VII, which is dependent on the availability of III, a methylation product. In contrast, the abundance of the methylated group of metabolites (specifically metabolites VII and II) was lower in the CRU and CRU+API diets than in the positive control. The decrease in metabolite VII occurred despite the CRU diet inducing CYP1A enzyme activity, suggesting a decrease in availability of metabolite III due to decreased methylation of PhIP. Because oral administration of phenethyl isothiocyanate (a constituent of cruciferous vegetables) downregulated mRNA of N-methyltransferase (NMT) in the liver of rats (54), the decreased abundance of methylated PhIP metabolites by the CRU diet may be due to downregulation of NMTs. Little is known regarding the importance of methylation in PhIP metabolism; it is uncertain how PhIP methylation could lead to decreased PhIP-DNA adducts. However, methylation of the exocyclic amine of PhIP might suppress activation to the adduct-forming metabolite by preventing hydroxylation of the exocyclic amine.

Altogether, our discovery of an additional methylated metabolite, the demonstration of PhIP-methylating capability in human liver S9 fraction, and the negative correlation of methylated metabolites with adducts imply a protective role for NMTs in PhIP metabolism and carcinogenesis. Consequently, we postulate that NMTs represent a major metabolic pathway leading to decreased PhIP genotoxicity. We speculate that increases in metabolites IV, VIII, and XV reflect a channeling of PhIP (I) away from methylation by NMT as evidenced by the concurrent decreases we observed in metabolites II and VII. We further speculate that increases in metabolites IV, VIII, and XV are reflective of more PhIP moving through activation pathways, resulting in increased PhIP-DNA adduct formation. Conversely, channeling of PhIP toward methylation may thereby lower the likelihood for PhIP activation (i.e., hydroxylation of N² position of PhIP), as evidenced by the decrease in adducts in the API group when there was a simultaneous increase in metabolite VII and decrease in metabolite IV. Although we believe this scenario is highly plausible, apiaceous constituents may potentially reduce adducts by other mechanisms. For example, PhIP can be extruded from cells via multidrug resistance–associated protein 2 (MRP2) in the colon (55), and importantly, several flavonoids present in apiaceous vegetables can either induce MRP2 mRNA or protein expression or inhibit the activity of MRP2, thereby modulating biliary or renal excretion of MRP2 substrates (56). In addition, the apiaceous constituents chlorogenic acid and quercetin, respectively, increase DNA repair (57) and induce apoptosis (58).

A strength of our approach is the provision of vegetables in amounts that are achievable in the human diet. We provided an equivalent of ∼1 cup/d based on an 8368-kJ intake (∼2000 kcal) using allometric scaling based on energy intake, a more appropriate scaler than body weight (59, 60). For example, 21% wet wt:wt of fresh apiaceous vegetables only contributed 1.85% of total calories in the diet (equivalent to 37 kcal of 2000 kcal for humans) because most of the vegetable weight was water [79.5–95.4% water per the USDA National Nutrient Database (http://www.ars.usda.gov/ba/bhnrc/ndl)]. We also comprehensively assessed initiation-related endpoints after an intervention of vegetables known to modulate carcinogen-metabolizing enzymes. The metabolomics approach allowed the assessment of multiple known PhIP metabolites as well as conferring the ability to discover new PhIP metabolites, as we did here. However, PhIP metabolites may exhibit different ionization properties for mass detection in the LC-MS/MS system. Therefore, comparisons should only be made between diets for a given metabolite or group of metabolites, not between metabolites within a particular diet. Although it is presumed that the formation of PhIP-DNA adducts in target tissues increases the risk of carcinogenesis, there are scant and inconsistent data demonstrating a direct correlation between PhIP-DNA adduct levels and precancerous lesions (aberrant crypt foci) or colon tumor incidence (5, 53, 61). Nonetheless, genotoxic carcinogens do not induce cancers without DNA adduct formation (62), and the dose of PhIP and resulting PhIP-DNA adduct levels are proportional (53). Last, although we used a well-established animal model of colon carcinogenesis, there are inherent interspecies differences in microbiota and biotransformation (52).

We hypothesized that the combination of cruciferous and apiaceous vegetable feeding would synergistically protect against PhIP-DNA adduct formation in the colon via complementary actions on PhIP-metabolizing enzymes. Our results indicate that apiaceous vegetables were most effective in reducing PhIP-DNA adducts. However, this reduction by the API diet was unexpectedly not due to inhibiting CYP1A activation of PhIP. Because the API diet increased the relative abundance of a methylated PhIP metabolite, we propose that enhancement of PhIP-methylating pathways may be important in PhIP detoxification. In addition, we detected a novel methylated PhIP metabolite and demonstrated for the first time PhIP-methylating capability by human liver S9 fraction. Considering that the novel methylated metabolite was negatively correlated with PhIP-DNA adducts, structural elucidation of the metabolite is warranted. Further investigation seems warranted of the importance of methylated PhIP metabolites, NMTs, and carcinogen metabolism (e.g., PhIP) and carcinogenesis and the effects of apiaceous vegetables on NMTs. Interactions of apiaceous vegetables with mechanisms beyond biotransformation enzymes (e.g., phase III excretion and DNA repair) also warrant additional study.