Profiling how the Gut Microbiome Modulates Host Xenobiotic Metabolism in Response to Benzo[a]pyrene and 1-Nitropyrene Exposure

Abstract

The gut microbiome is a key contributor to xenobiotic metabolism. Polycyclic aromatic hydrocarbons (PAHs) are an abundant class of environmental contaminants that have varying levels of carcinogenicity depending on their individual structures. Little is known about how the gut microbiome affects the rates of PAH metabolism. This study sought to determine the role that the gut microbiome has in determining various aspects of metabolism in the liver, before and after exposure to two structurally different PAHs, benzo[a]pyrene and 1-nitropyrene. Following exposures, metabolic rates of PAH metabolism were measured, and activity-based protein profiling was performed. We observed differences in PAH metabolism rates between germ-free and conventional mice in both unexposed and exposed conditions. Our ABPP analysis showed that, in unexposed conditions, there were only minor differences in total P450 activity in germ-free mice relative to conventional mice. However, we observed distinct activity profiles in response to corn oil vehicle and PAH treatment, primarily in the case of 1-NP treatment. This study revealed that the repertoire of active P450s in the liver is impacted by the presence of the gut microbiome, which modifies PAH metabolism in a substrate-specific fashion.

Graphical Abstract

INTRODUCTION

Human exposure to environmental chemical contaminants is a severe societal health burden as exemplified by increased cancer rates, developmental delays, and birth defects.^1–3 While the phenotypes resulting from exposures are becoming more apparent, the biochemical mechanisms underlying their etiology is less clear. Mammalian metabolism of xenobiotics occurs primarily in the liver, but the fate of xenobiotics is also influenced by processes in other organs and the gut microbiome.⁴ The gut microbiome contributes to xenobiotic metabolism through direct and indirect mechanisms. Direct metabolism involves microbial enzymes that bio-transform compounds through a wide range of reactions including hydrolysis, reduction, dehyrodxylation, and nitrosamine formation.⁵ Indirect metabolism refers to alterations in drug processing genes in host tissues through the production of microbial metabolites, such as secondary bile acids, tryptophan, and short chain fatty acids that can engage in host receptor binding, and subsequent induction of signaling pathways.^6–7 The gut microbiome has significant reductive and hydrolytic reaction potential that impacts xenobiotic metabolism.⁸ However, the exact mechanisms through which the gut microbiome influences detoxification and bioactivation of environmental contaminants, such as polycyclic aromatic hydrocarbons (PAHs), remain incompletely understood.

PAHs, which are generated from the incomplete combustion of organic materials, are an environmental chemical contaminant of significant health concern. PAH exposure occurs from inhaling smoke and diesel exhaust, drinking from contaminated water sources, as well as ingesting grilled food.^9–11 PAHs can be cleared from the human body through phase 1 and 2 metabolism, but these otherwise helpful metabolic processes can sometimes result in the bioactivation of intermediates that are more harmful than the parent toxin.¹² One example is the conversion of benzo[a]pyrene into electrophilic epoxides that result from cytochrome P450 (P450) oxidation, that can readily form DNA adducts and potentially lead to tumor formation.^13–15 Another example is the metabolism of nitrated PAH’s resulting in the formation of protein adducts that accumulate at higher levels than DNA adducts and have higher stability.¹⁶ Therefore, liver metabolism must be tightly regulated to avoid these potentially harmful side process and understanding the inputs that lead these processes awry is crucial.

An intuitive place to look for metabolic regulators is the organs that “feed” the liver, such as the gut, or more specifically, the microbiome. The microbiome is a strong candidate for modifying exposures, especially in the case of PAHs, because smoke exposure can contaminate mucus, which is a major carbon source for the microbiome, or alternatively, PAHs can be ingested as charred food, which will pass through the gut. It is already known that perturbations in the gut microbiome can occur from exposure to environmental contaminants, and this has been linked to elevated occurrence of disease states such as obesity, cardiovascular disease, irritable bowel disease, and even neurological disorders.^17–20 There are already some convincing lines of evidence that point to specific links between the gut microbiome and PAH metabolism. Previous research has shown that chronic exposure to benzo[a]pyrene, a well-studied PAH, increases intestinal inflammation that may be involved in the onset of type 2 diabetes.²¹ Another study evaluating the effects of 1-nitropyrene exposure revealed in different metabolite profiles when comparing germ-free rats to conventional mice.²² Additionally, studies using germ-free mice showed that the absence of gut microbiota results in broadly altered expression of drug processing genes in the intestine and liver, including known PAH-metabolizing enzymes, indicating the importance of the gut microbiome on host metabolism.^23–24

Despite the clear links between the microbiome and metabolism of xenobiotics like PAHs, it is difficult to capture the functional consequences of the microbiome’s impact on hepatic drug processing genes in the liver using conventional approaches such as transcriptomic and proteomic techniques because abundance is quantified rather than activity. Assays for liver enzyme activity are available, but they do not report on individual enzyme isoforms, and in the case of P450 enzymes, do not include all families and subfamilies, which makes acquisition of a full picture impossible.²⁵ Demystifying these processes is particularly exigent because susceptibility to specific environmental toxins could be identified through microbiome profiling, and treatments and preventative approaches could be tailored with dietary or other therapeutic interventions.

To address these knowledge gaps, we leveraged a function-focused analysis to determine metabolic profiles for conventional and germ-free mice that were exposed to two structurally different PAHs, benzo[a]pyrene and 1-nitropyrene (Figure 1). These PAHs represent similar, but distinct chemical exposures, with related chemical structures that may provide insight on how gut microbiome-liver dynamics can be affected by small structural changes in the chemical exposure agent. Both PAH substrates have the same pyrene core, but benzo[a]pyrene has an additional fused hydrophobic aryl ring, whereas 1-nitropyrene has a strongly electron withdrawing and dipole-inducing substituent. Using two different PAHs allows for illumination of enzyme substrate specificities and determining links to rates of metabolism.

Figure 1.

Experimental design overview. Germ-free and conventional mice were dosed with water, corn oil vehicle, benzo[a]pyrene, or 1-nitropyrene. Hepatic microsomes were isolated and metabolism was assessed by an HPLC-based kinetic assay and activity-based protein profiling.

For our selection of a robust model for determination of the gut microbiome’s influence on PAH liver metabolism, we decided to compare germ-free C57BL/6 mice to conventional C57BL/6 mice. Both strains are genetically identical, and hypothetically only differ by the presence of the microbiome. While germ-free mice are known to have developmental differences, we still saw this as a highly attractive model compared to other approaches, such as antibiotic treatment because it is very challenging to completely “knockdown” the microbiome and any remaining microbes, which could differ between mice, and confound our observations. We first wished to establish baseline metabolic differences between germ-free and conventional mice, so one group was simply treated with water. Second, treatment with PAHs requires a vehicle, so to account for changes from vehicle alone, we treated with corn oil. Lastly, we treated mice with a corn oil solution of benzo[a]pyrene or 1-nitropyrene using a dose based on the literature.²⁶

Differences in PAH metabolic capacity between germ-free and conventional mice were initially evaluated using conventional HPLC-based liver microsome assays. We examined the kinetics of PAH biotransformation by microsome extracts at baseline levels (untreated), post vehicle exposure (corn oil), and post PAH exposure (benzo[a]pyrene or 1-nitropyrene in corn oil). If the microsomes were from PAH pre-treated mice, we matched the substrate for our HPLC assay with the chemical exposure. Additionally, water-treated, and vehicle-treated microsomes were also screened against both assay substrates.

To elaborate on our observations with the HPLC-based activity assays, which narrowly analyze biotransformation of single substrates, we additionally broadly analyzed how the gut microbiome alters the activity of cytochrome P450 enzymes, which are highly relevant to PAH metabolism,^{13, 27} using activity-based protein profiling (ABPP), which employs small molecule probes that selectively label active enzymes of interest for downstream proteomic analysis.²⁸ We specifically used a suite of cytochrome P450 probes tailored to target P450 enzymes.²⁹ Studying the differences between conventional mice and germ-free mice, which have no intestinal microbiome, will reveal active P450 enzymes that are specifically regulated by the microbiome’s direct and indirect interactions. Collectively, this study highlights the role that the gut microbiome has on xenobiotic metabolism.

MATERIALS AND METHODS

Chemicals.

Benzo[a]pyrene, 1-nitropyrene, sucrose, phosphate buffered saline, trifluoroacetic acid, sodium ascorbate, dibenzo[def,p]chrysene, and copper sulfate were purchased from Sigma-Aldrich (St. Louis, MO, USA). Iodoacetamide was purchased from Acros Organics (New Jersey, USA). N-(3-azidopropyl)biotinamide (biotin-azide) was purchased from TCI (Tokyo, Japan). Tris(2-carboxyethyl)phosphine (TCEP) was purchased from ThermoScientific (Rockford, IL, USA). Pierce BCA Protein Assay Kit (Thermo Scientific), methanol, sodium sulfate, sulfuric acid, acetone, ethyl acetate, acetonitrile, potassium phosphate salts (dibasic and monobasic), dithiothreitol, reduced nicotinamide adenine dinucleotide phosphate (NADPH), were purchased from MilliporeSigma. TAMRA-azide and tris-hydroxypropyltriazolylmethylamine (THPTA) were purchased from Click Chemistry Tools (Scottsdale, AZ, USA). A cocktail of activity-based probes for targeting active P450 enzymes, consisting of 2-methylnaphthalene, ATW2 and ATW5, were synthesized as previously described.²⁹

Animals.

24 7-week-old female C57BL/6NTac and 20 7-week-old germ-free C57Bl/6GFTac mice were purchased from Taconic Laboratory (German Town, NY). All mice were fed autoclaved LabDiet PMI 5010 Autoclavable Diet ad libitum. Mice were housed in the PNNL Animal Resource Center. Germ-free mice (germ-free) were group housed in a designated germ-free facility room in a Tecniplast IsoP IVC caging system and the conventional mice were group housed in a separate room in Innovive IVC cages to maintain an environment like the germ-free mice. The rooms were kept at 22±1°C and at 50±20% humidity on a 12-hour light/dark cycle with free access to food and sterile water. All animal experiments were conducted in accordance with institutional guidelines for the care and use of laboratory animals. Mice were acclimated for 1 week prior to dosing. Mice were dosed daily with 180 mg/kg/day benzo[a]pyrene or 1-nitropyrene in sterile 180 mg/kg corn oil, by oral gavage for 72 hours. All handling of mice was conducted in a sterile biological safety cabinet. Mice were euthanized 48 hours post final dose. Four mice per mouse type and per treatment were dosed. To ensure that germ-free mice were not contaminated feces were collected each time the cage is opened, every other day, and at the end of the study just before sacrifice. The feces were collected using sterile technique inside a sterile BSC, fresh feces were collected both directly from the mouse and from the cage. A smear of the feces was then plated on TSAII plates with 5% sheep’s blood, brucella plates (to detect anaerobic bacteria), inoculated into thioglycolate broth, and a gram stain smear was conducted. Therefore, three different ways to check for contamination were used.

Preparation of liver microsomes.

Microsomes were prepared as previously described with minor modifications³⁰ Liver tissues from each treatment type were rinsed with cold 1.5% NaCl, minced with razor blade, and dounce homogenized with loose and tight pestles with 250 mM sucrose in EDTA-free phosphate buffered saline (PBS). Samples were centrifuged 10,000 × g for 30 minutes at 4 °C, separating the heavy membrane from the S9 fraction. The supernatant was centrifuged at 100,000 × g for 90 minutes at 4 °C, separating the microsome from the cytosolic fraction. Microsomes were reconstituted in 500 μl PBS and dounce homogenized. Protein concentrations were determined using Pierce BCA protein assay.

In vitro probe labeling and click chemistry.

Normalized microsomes (1 mg/ml) were prepared as described above. For ABPP fluorescence characterization, 50 μg protein samples were treated with a mixture (20 μM) of 2EN, ATW8, and ATW12 activity-based probes to label active P450 enzymes.²⁹ The mixtures were respectively incubated with and without NADPH (2 mM) for 1 hour at 37 °C on a thermal shaker, shaking at 1000 rpm. NADPH is a required cofactor for P450 activity. Following incubation, copper catalyzed click chemistry was carried out as previously described.³¹ Samples were incubated in the dark at room temperature with rhodamine-azide (60 μM) followed by sodium ascorbate (5 mM), tris-hydroxypropyltriazolmethylamine (2 mM) and copper sulfate (4 mM). Samples were then mixed with one volume of cold MeOH and incubated on ice for 1 hour, centrifuged at 10,000 × g at 4 ^○C for 5 minutes, and the supernatant was discarded.

SDS-PAGE analysis.

Probe-labeled, rhodamine-clicked proteins were resolubilized in 25 μl SDS (1.2%) in PBS, heated 95 °C for 2 minutes followed by the addition of 25 μl 2X SDS running buffer, 5.5 μl 10X reducing agent (Invitrogen). Samples were heated for an additional 2 minutes prior to centrifugation at 10,000 × g for 5 minutes to remove any remaining insolubilized proteins. 15 μl of each sample were added to the wells of 4–12% Bis-Tris gels (Invitrogen). Samples were run at 150 V for 90 minutes. Samples were imaged with Typhoon FLA 9500 (General Electric). Gels were fixed for 30 minutes with a (50%), acetic acid (40%), and MilliQ water (10%) mixture. Fixed gels were rinsed with water and stained with Gelcode blue overnight. Prior to imaging, gels were washed 2X with water. Gels were imaged using GelDocEZ (BioRad Laboratories).

Samples used for mass spectrometry analysis.

1.0 mg protein samples were probe labeled with the P450 cocktail as described above. The mixtures were incubated with and without NADPH (2 mM) for 1 hour at 37 °C on a thermal shaker, shaking at 1000 rpm. Samples were incubated in the dark at room temperature with biotin-azide (60 μM), sodium ascorbate (5 mM), tris-hydroxypropyltriazolmethylamine (2 mM) and copper sulfate (4 mM). Excess biotin was removed by incubating proteins at −80 °C in 800 μl of cold methanol for 1 hour. Samples were centrifuged at 10,000 × g for 10 minutes and supernatant discarded. Proteins were resolubilized by adding SDS (1.2%) in PBS solution, heated at 95 °C for 2 minutes, and sonicated with 2 × 1 s pulses 60% amplitude.

Streptavidin enrichment.

Samples containing resolubilized proteins were normalized via BCA assay. Prior to incubation, 100 μl Streptavidin agarose resin was washed with SDS (0.5%) in PBS, urea (6 M) in NH₄HCO₃ (25 mM, pH 8.0), and PBS. Washed beads were incubated with normalized protein samples at 37 °C for 1 hour, rotating. Samples were subsequently washed with SDS (0.5%) in PBS, urea (6 M) in NH₄HCO₃, MilliQ water, 1X PBS, and NH₄HCO₃ (25 mM, pH 8.0). Protein-bound resin was transferred to 1.5 ml cryovials, where the protein was reduced with (tris(2-carboxyethyl) phosphine hydrochloride (5 mM)) and alkylated (iodoacetamide (10 mM)) at 37 °C and 50 °C, respectively. Samples were transferred to fritted columns and washed with PBS and NH₄HCO₃ (25 mM). Samples were transferred to tubes followed by the addition of 0.25 μg Trypsin (Promega). Trypsin digest was conducted overnight at 37 °C, shaking at 1000 rpm, in 200 μl NH₄HCO₃ (25 mM, pH 8.0). Following trypsin digestion, samples were centrifuged, supernatant collected, and dried under speedvac. Peptides were resolubilized in 40 μl NH₄HCO₃ (25 mM) for MS analysis.

LC-MS/MS analysis of probe enriched.

Peptide MS/MS spectra were searched using the mass spectra generating function plus (MSGF+) against the publicly available Mus musculus translated genome sequence. ^32–33 The peptides were filtered by a false discovery rate (FDR) of 5% that was manually calculated by MSGF+ using a decoy approach.³⁴ The redundancy within each peptide were removed by summing the peak area intensities. After reverse hits and contaminants are removed, peak areas were brought to the peptide level by summing within peptides. Peptides were filtered to only those measured in at least 3 of the 4 replicates in any treatment type. The resulting list of peptides were log₂ transformed and scaled to the protein level using Inferno RRollup software with the median peptide abundance value being used as the relative protein abundance value.³⁵ Only proteins with >3 peptides were included in downstream analysis.

In vitro benzo[a]pyrene and 1-nitropyrene metabolism in liver microsomes.

Incubations were carried out using a 500 μL solution of 0.299 mg/mL microsomal protein in phosphate buffer containing 3 mM MgCl₂ (pH 7.4). NADPH (1.5 mM) was added to experimental samples as the requisite P450 cofactor, and the samples were warmed to 37 °C. Immediately following, benzo[a]pyrene (0.625 μM – 20 μM) or (0.625 μM – 15 μM) 1-nitropyrene was added and allowed to incubate for 0- or 10-minutes. Incubations were stopped with H₂SO₄ (0.45 M) and placed on ice. A standard curve was generated (0.031 μM – 25.0 μM) in the same manner as the microsome without the addition of NADPH. An internal standard, dibenzochrysene (7 μM) was added to each quenched sample. Dibenzochrysene, benzo[a]pyrene, 1-nitropyrene, and metabolites were extracted using two 0.500 mL volumes of ethyl acetate. Samples were dried under a gentle N₂ stream and then resolubilized in 500 μL methanol. Reverse-phase HPLC using an Agilent 1100 HPLC system equipped with a fluorescence detector (Santa Clara, CA, USA was used for analyte quantification. Briefly, 10 μL was injected into an Ascentis 25 cm × 4.6mm, 5μm C18 column (Sigma-Aldrich, St. Louis, MO, USA) with a gradient of water and acetonitrile (55:45 to 0:100) from 0 to 10 minutes followed by 100% acetonitrile until 22 minutes at a flow rate of 0.95 mL/minute. Peaks were measured via fluorescence detection with excitation and emission wavelengths being 230 nm and 430 nm, respectively. Metabolic rates were calculated using an exponential regression of benzo[a]pyrene and 1-nitropyrene concentrations (μM) as a function of time (minutes). Nonparametric bootstrap method was used to generate confidence intervals of metabolic rates.

Statistical data analysis.

Metabolic rates were determined based on substrate disappearance. For each substrate concentration, metabolic rates (as initial rates of substrate disappearance; nmol/min/mg microsome) were calculated using an exponential regression of benzo[a]pyrene on 1-nitropyrene concentrations (μM) as a function of time (min). Confidence intervals of metabolic rates were calculated using a nonparametric bootstrap, where metabolic rates were resampled with replacement at each substrate concentration and time, and the regression was repeated on the resampled dataset (n=1000). Metabolic rates as a function of substrate concentration were evaluated using a Michaelis-Menten equation for saturable kinetics. Intrinsic cleared values (CL_int) were estimated by dividing the V_max by the K_M. Fit parameter confidence intervals were calculated using a bootstrap method using the best fit model with n=1000.

RESULTS

Germ-free mice demonstrate different trends in PAH metabolism when compared to conventional mice

We evaluated in vitro metabolism of benzo[a]pyrene and 1-nitropyrene in hepatic microsomes from unexposed germ-free and conventional mice using an HPLC assay, in which the biotransformation kinetics of purified BaP or 1-NP was analyzed through tracking disappearance of a spiked-in PAH substrate (Figure 2). Microsomes from unexposed and corn oil-treated mice were respectively tested in both the BaP and 1-NP assays. If the microsomes were from mice post-PAH-treatment, we matched the substrate for our HPLC assay with the chemical exposure. We considered any kinetic parameters that did not have any overlap of the respective 95% confidence intervals (CI) as being significantly different. From our analysis, we observed vastly different patterns of benzo[a]pyrene metabolism in germ-free and conventional mice.

Figure 2.

Rates of benzo[a]pyrene metabolism in hepatic microsomes from A) germ-free mice and B) conventional mice incubated with benzo[a]pyrene (0.62 μM–20 μM). Rates of 1-nitropyrene metabolism in hepatic microsomes from C) germ-free mice and D) conventional mice incubated with 1-nitropyrene (0.62 μM–10 μM).

Biotransformation of BaP of hepatic microsomes from conventional and germ-free mice primarily differ by rates observed at high substrate concentrations.

In hepatic microsomes from unexposed CV mice, we observed baseline kinetic parameters for BaP biotransformation of V_max = 0.47 nmol/min/mg, K_M = 1.29 μM, CL_int = 0.36 ml/min/mg. Corn oil treatment did not have any statistically significant impact on any of the kinetic parameters for BaP biotransformation by CV liver microsomes relative to water-treated controls. As expected, microsomes from CV mice that were treated with BaP had increased biotransformation rates of BaP at saturating concentrations (i.e. BaP V_max) and at low substrate concentrations (i.e. BaP CL_int) relative to vehicle-treated controls (1.3-fold and 1.6-fold, respectively). However, BaP treatment had no impact on the substrate concentration required for a microsome extraction to reach the maximal BaP biotransformation velocity (i.e. BaP K_M) relative to vehicle-treated controls.

In unexposed GF mice, we observed baseline kinetic parameters for biotransformation of BaP of V_max = 5.36 nmol/min/mg, K_M = 29.63 μM, CL_int = 0.18 ml/min/mg. Corn oil treatment had a substantial effect on BaP biotransformation kinetics in GF microsomes relative to water-treated controls, resulting in microsomes with a decreased maximal velocity (9.4-fold) and decreased BaP K_M (14.8-fold), but with no effect on BaP CL_int. Similar to CV mice, microsomes from post-BaP-treated GF mice had increased BaP V_max (1.6-fold) and BaP CL_int rates (1.9-fold) relative to vehicle treated controls, and there was no substantial effect on BaP K_M.

Comparison of kinetic parameters for CV and GF microsomes reveals that, under unexposed conditions, microsomes from GF mice have substantially higher (>10 fold) rates of BaP biotransformation at high substrate concentrations (BaP V_max (nmol/min/mg): GF_unexposed = 5.36, CV_unexposed = 0.47) relative to CV mice. Microsomes from animals treated with corn oil also had different kinetic profiles between the two strains. Significant inhibition effects from vehicle treatment were observed in GF mice, but not CV mice, which resulted in GF BaP V_max values that were significantly lower than their CO-treated CV counterparts (BaP V_max (nmol/min/mg): CV_CO = 1.21, GF_CO = 0.57), resulting in an inversion of rank ordering relative to water-treated animals. BaP treatment had roughly proportional activating effects in both strains relative to vehicle treated animals. Notably, most differences between GF and CV mice were in the velocity reached at saturating substrate concentrations.

Hepatic microsomes from 1-NP-treated GF mice have decreased biotransformation rates of 1-NP relative to vehicle-treated controls, but no effect is observed in CV mice.

Liver microsomes from water-treated CV mice had qualitatively lower 1-NP biotransformation rates at low and saturating conditions relative to the rates observed or water-treated CV mice in the BaP biotransformation assay, although the differences were not statistically significant. Also, unlike the BaP biotransformation assay, a significant effect on 1-NP biotransformation was observed in CV mice that were treated with corn oil, with significant elevation of intrinsic clearance rates observed relative to water-treated control animals (1-NP CL_int (ml/min/mg): CV_unexposed = 0.04, CV_CO = 1.06). Treating CV mice with 1-NP appears to have no effect on 1-NP biotransformation in our HPLC assay, as all kinetic parameters for liver microsomes from 1-NP-treated CV mice were statistically indistinguishable from vehicle-treated CV mice.

In unexposed GF mice, we observed baseline kinetic parameters for biotransformation of 1-NP of V_max = 1.1 nmol/min/mg, K_M = 3.5 μM, CL_int = 0.04 ml/min/mg. Corn oil treatment did not have an inhibitory effect as it did for GF mice in the BaP biotransformation assay, and slightly increased 1-NP intrinsic clearance rates were observed (CL_int (ml/min/mg): GF_unexposed = 0.31, GF_CO = 0.54). Liver microsomes from GF mice post 1-NP treatment displayed markedly reduced 1-NP biotransformation rates at saturating concentrations (1-NP V_max (nmol/min/mg): GF_CO = 0.7, GF_1-NP = 0.09) and at low concentrations (1-NP CL_int (ml/min/mg): GF_CO = 0.54, GF_1-NP = 0.08) relative to corn oil-treated controls, but there was no effect on the concentration required to reach maximal biotransformation rates.

Comparison of the respective conditions between strains reveals that unexposed CV mice have significantly reduced 1-NP intrinsic clearance rates (1-NP CL_int (ml/min/mg): CV_unexposed = 0.04, GF_unexposed = 0.31) relative to unexposed GF mice. Corn oil treatment resulted in statistically significant increase in intrinsic clearance of 1-NP by both CV and GF microsomes relative to water-treated controls, however, the impact was disproportionately higher in the CV microsomes, which raised CL_int to the same level as GF, and resulted in statistically indistinguishable CL_int values between strains post-corn oil treatment. Microsomes from post-1-NP-treated GF animals had a > 7-fold drop in biotransformation of 1-NP relative to their vehicle-treated GF counterparts (1-NP V_max (nmol/min/mg): GF_1-NP = 0.09, GF_CO = 0.7), but 1-NP treatment did not have any impact on the activity of CV microsomes relative to vehicle treatment. Overall, unlike the BaP analysis, in the 1-NP analysis, we found differences at low substrate concentrations between CV and GF mice, in addition to differences at saturating concentrations.

1-nitropyrene exposure induces higher P450 labeling relative to BaP with activity-based protein profiling in both germ-free and conventional mice.

To more broadly understand how the microbiome influences liver metabolism, evaluation of P450 enzymes by ABPP via SDS-PAGE was conducted. To conduct the most complete analysis of P450 metabolism, we used a cocktail of three different probes (2EN, ATW8, and ATW12; Figure 2), with different hydrocarbon structures, that collectively label the majority of P450 enzymes (xxcite), and therefore enable maximum coverage for profiling of P450 activity. Probe labeling was conducted with and without NADPH, which is a cofactor is required for P450 activity, to account for any non-specific labeling. To preliminarily visualize the extent of probe labeling, azido-rhodamine was appended to probe-labeled protein conjugates, and fluorescence imaging after gel electrophoresis was conducted. Fluorescence imaging after SDS-PAGE revealed a major P450 band at roughly 52 kDa, and this band was the focus of our preliminary analysis (Figure 3).

Figure 3.

A) Activity-dependent labeling of Cytochrome P450s, and B) chemical structures for activity-based probes. C) SDS-PAGE-based analysis of cytochrome P450 activity using activity-based probes. Protein bands for conventional mice had higher fluorescence intensity in benzo[a]pyrene and corn oil treatment conditions than corresponding protein bands for germ-free mice, though the 1-nitropyrene treatment condition had the highest band intensity for both germ-free and conventional mice.

Our analysis revealed that, in vehicle-treated microsomes, higher activity was observed in CV samples than GF samples based on band intensity. In BaP-exposed microsomes, we observed an increase in activity in both CV and GF conditions relative to respective vehicle-treated controls. We observed an even higher increase in microsomes from 1-NP-treated CV mice. We were somewhat surprised to observe a similar elevation with GF microsomes given that decreased activity was observed post 1-NP treatment in our HPLC assay. However, the HPLC assay only reports specifically on 1-NP metabolizing enzymes, whereas the ABPP data provide a global view of P450 activity. While the SDS-PAGE-based analysis suggests differences in P450 activity between CV and GF mice, both in PAH-exposed and in unexposed conditions, it does not inform on the specific contributions of individual P450s and is furthermore biased toward high abundance enzymes. Therefore, as a next step to further investigate the microbiome’s role in regulation of P450 activity, we implemented mass spectrometry-based ABPP.

Differences in P450 isoform activity between unexposed and vehicle-treated GF and CV mice.

Liver microsomes were co-incubated with the ABP cocktail, and then labeled P450s were enriched via affinity chromatography. After trypsinization of the enriched proteins, the resulting peptide mixture was analyzed by LC-MS. For our analysis, we excluded any proteins that were not significantly enriched relative to control samples that were not treated with ABPs (student’s t-test, p < 0.05, and log2 Fold-Change(probe/no-probe) > 2). For all quantitative comparisons and statistical analyses, a measurement must have been made; no imputed values were used to calculate fold-changes. If peptides were detected in the probe-labeled condition, but no measurable quantity of peptide was detected in the no-probe sample, values were not imputed, and these enrichments were simply designated as “significant” and were not assigned fold-change values. In heatmap visualizations, missing values are indicated by blank white space. For comparisons of probe-labeled conditions, we call any difference significant that has a p-value less than 0.05 as determined by a student’s t-test analysis. We did not impose a fold-change criterion for significance when comparing probe-labeled samples because all samples analyzed at this stage had already passed a fold-change test for the respective probe vs. no-probe comparisons.

We began our analysis by examining the differences in P450 activity of control mice that were treated with water or corn oil vehicle (Figure 4). First, when comparing water-treated CV vs GF mice, we observed few differences in active P450 levels (Figure 4A). There were two P450s (Cyp27a1 and Cyp2d11) that were uniquely detected in the CV strain and two P450s (Cyp2c29 and Cyp2f2) that were uniquely detected in the GF strain (SI Figure 1). Of the 8 total P450s that were detected in both strains, three of these P450s were found to have significantly different levels between the strains, but the magnitude of the fold-change (FC) difference was modest (log₂FC_(CV/GF): Cyp2c37 = −0.6, Cyp2c70 = −0.8, Cyp3a11 = 1.8).

Figure 4.

A) Heatmap comparing p450 labeling in hepatic microsomes from water-treated CV and GF mice. Individual replicates for each condition are represented with a single box in each condition column. Boxes are colored according to abundance values. B) Volcano plot illustrating differences in P450 labeling levels in hepatic microsomes from corn oil-treated CV and GF mice. Horizontal dashed line indicates the point at which all points above have a statistically significant difference between conditions (p ≤ 0.05). Vertical line indicates that all points to the right are at least one standard deviation above the mean.

Despite observing few differences in untreated mice, we observed substantial differences in how many P450s were identified as significantly different between the strains upon treatment with corn oil vehicle. Upon corn oil treatment, 11 new P450s were detected for GF mice and 27 new P450s were detected in CV mice that were not detected in the respective water-treated controls (SI Figure 2). In total, when we compared active P450s that were detected in both strains, we observed 29 P450s with significantly different labeling levels between CV and GF mice, which represents a majority of the 36 total shared P450s that were analyzed (Figure 4B). Strikingly, all P450s that were designated as significantly different were higher in CV mice, with a range of 0.8–3.4 log₂FC_(CV/GF) and a mean log₂FC_(CV/GF) of 1.6±0.6. So, not only were more total unique P450s designated as active in CV mice upon CO treatment, of those that were shared with GF, all were more active in CV mice.

There was significant diversity among the P450s that were found to be higher in the CV corn oil treatment condition relative to the GF corn oil treatment condition, with examples present from the Cyp1, Cyp2, Cyp3, Cyp4, Cyp8, Cyp20, and Cyp51 families (SI Figure 2). The Cyp2 family had the greatest number of proteins that were different between the strains, with 16 members represented, although it should be noted that is also the family with the most P450 genes. Five P450s had activity equal to, or higher, than one standard deviation above the mean in CV mice relative to GF mice (log₂ FC_(CV/GF): Cyp3a16 = 2.2, Cyp4f14 = 2.3, Cyp3a25 = 2.4, Cyp3a11 = 2.6, and Cyp4f3 = 3.4). An over-representation of the Cyp3a and Cyp4f sub-families, which are involved in cholesterol, steroid, and lipid synthesis, and arachidonic acid metabolism, was apparent in the highest fold-change group.

We next broadly analyzed the effects of BaP and 1-NP treatment on P450 activity in both strains. When we analyzed the collective data at a family level, we noticed only a few discernable differences between unexposed mice and the two respective PAH exposures (Figure 5). Upon BaP treatment, we observed a significant increase in labeling of members of the Cyp1 family relative to corn oil for both strains (log₂FC: CV_BaP/CO = 3.6, GF_BaP/CO = 3.3), which is already known to be inducible by PAH treatment. For CV mice, the only other significant change at the isoform level that was observed upon BaP treatment when compared to corn oil-treated CV mice was a decrease in Cyp27a1 (log₂ FC: CV_(BaP/CO) = −1.5; SI Figure 3). For GF mice, the only other significant change upon BaP treatment relative to corn oil treatment was an elevation of Cyp7b1 (log₂ FC: GF_(BaP/CO) = 2; SI Figure 4).

Figure 5.

Heatmap comparing family-level differences in activity-based labeling of P450s from hepatic microsomes from either A) conventional or B) germ-free mice that were respectively treated with corn-oil vehicle, benzo[a]pyrene and 1-nitropyrene. Individual replicates for each condition are represented with a single box in each condition column. Boxes are colored according to abundance values. White space indicates P450s that were not detected in the mass spectrometry analysis.

Both strains had a different P450 response post-1-NP treatment relative to BaP treatment. Most notably, in contrast to BaP treatment, activation of the Cyp1 family by 1-NP treatment relative to corn oil-treated controls was either not observed (GF) or substantially attenuated relative to BaP treatment (CV). Several other differences of note were observed upon analyzing the changes from 1-NP treatment relative to corn oil for each strain at the isoform level. CV mice had altered labeling of 20 P450s post 1-NP treatment relative to corn oil, whereas GF mice had only 5 P450s that had modified activity relative to corn oil (Figures 5 and 6 respectively). Overall, there were no directional changes relative to vehicle-treatment that were unique to the 1-NP-treated GF group, although there were differences in the magnitude of changes relative to the corresponding CV conditions. Many differences in the P450 response to 1-NP treatment between the strains were in the Cyp2 family, and several members had increased in activity relative to vehicle-treated controls in CV mice including Cyp2a5, Cyp2b9, Cyp2c29, Cyp2c37, Cyp2c38, Cyp2d10, and Cyp2d26 (log₂ FC_(1-NP/CO) = 1.3, 3.2, 0.9, 5.6, 4.6, 1.8 and 1.8, respectively; SI Figures 5 and 6 respectively).

Figure 6.

Heatmap for isoform-level comparison of differences in activity-based labeling of P450 enzymes from hepatic microsomes of conventional or germ-free mice treated with corn oil vehicle or 1-nitropyrene. Individual replicates for each condition are represented with a single box in each condition column. Boxes are colored according to abundance values. Only proteins that were detected in all four conditions are shown. Any proteins that were not detected in one condition or more were not included in this analysis but can be found in SI Figure 7.

We conducted a fold-change analysis for the CYPs that were detected in CO-treatment and 1-NP treatment for both strains to determine differences in levels of ABP labeling among the shared proteins (Figure 6). Any proteins not detected in one condition, or more, were not considered for the analysis. Relative to vehicle, in 1-NP treated mice, Cyp2b10 was found to be increased in both strains (log₂ FC: CV_(1-NP/CO) = 4.7, GF_(1-NP/CO) = 1.5), and Cyp2c54 was found to be decreased in both strains (log₂ FC: CV_(1-NP/CO) = −4.5, GF_(1-NP/CO) = −6.2). Similar changes among the Cyp3 family occurred between both strains including an increase in both strains for Cyp3a11 (log₂ FC: CV_(1-NP/CO) = 3.6, GF_(1-NP/CO) = 2.1) and Cyp3a41a (log₂ FC: CV_(1-NP/CO) = 4.1, GF_(1-NP/CO) = 1.5), and both strains decreased in Cyp3a16 (log₂ FC: CV_(1-NP/CO) = −0.6, GF_(1-NP/CO) = −2.4).

DISCUSSION

Environmental contaminant exposure has been linked to elevated occurrence of disease states and is directly related to higher morbidity and mortality rates with metabolism greatly influencing their toxicity.^36–38 The gut microbiome has been shown to be a significant contributor in xenobiotic metabolism, though the mechanisms through which the gut microbiome contributes remains unclear. Identifying the gut microbiome’s contribution to the rate of metabolism, in the host, will increase our understanding of contaminant-gut-host interactions. This study employed a unique function-focused approach to understanding how the gut microbiome impacts liver P450 activity under baseline and exposure conditions. Furthermore, we have rarely seen examples of these two analytical approaches being combined to provide a richer understanding of a biological problem. Our kinetic and ABPP data suggest that there are differences in P450 activity in untreated CV and GF mice.

We began with a kinetic analysis of liver microsomes from CV and GF mice, which provided a narrow, but in-depth, comparison of the biotransformation of single, purified PAH substrates. It was found that BaP biotransformation rates at saturating substrate concentrations were different between CV mice and GF mice in all treatment conditions. Microsomes from both strains post-BaP treatment had increased maximal velocity rates relative to vehicle-treated controls. Interestingly, vehicle treatment had a significant inhibitory effect on maximal BaP biotransformation velocities in GF microsomes, but not in CV microsomes.

In our 1-NP biotransformation assay, we found examples of differences in both intrinsic clearance rates and biotransformation rates at saturating substrate concentrations across conditions. In the 1-NP biotransformation assay, the most notable effects were much lower 1-NP intrinsic clearance rates in microsomes from unexposed CV mice relative to unexposed GF mice, and a much stronger activating effect from corn oil treatment found in CV than GF microsomes (25-fold vs. <2-fold), and significant inhibition of both intrinsic clearance and maximal velocity rates subsequent to 1-NP treatment found in GF, but not in CV mice.

In total, our kinetic data show that some specific responses are shared between CV and GF mice (e.g. elevation of maximal BaP biotransformation rates post BaP treatment relative to respective CO-treated controls), but there are some notable differences (e.g. strong inhibition of maximal 1-NP biotransformation rates subsequent to 1-NP treatment in GF mice but not CV mice). Although these data strongly argue for important functional links of the microbiome and P450 metabolism, both in unexposed and treated conditions, this assay exclusively focuses on enzymes that directly biotransform BaP or 1-NP.

To more broadly understand the differences in P450 metabolism between CV and GF mice, we employed ABPP, which can detect and assign activity levels to the vast majority of P450s. ABPP functions through using a cocktail of tagged substrate probes that covalently label P450s after the enzymatic oxidation reaction, which then enables affinity enrichment of active P450s for mass spectrometry analysis. In contrast to the kinetics analysis, because we used a substrate cocktail, which interacts with the majority of P450s, ABPP can provide information on global P450 changes and not just enzymes that act on 1-NP or BaP.

Both strains of untreated mice had very few detected P450s. Each strain respectively had two uniquely detected P450s. Of the eight isoforms of the P450s that were detected in both strains, there was only a modest difference in magnitude of activity for three. Our ABPP data for unexposed mice are in good agreement with global proteomics analyses for the same strains which showed fold-change differences of generally less than 2-fold for detected P450s. Specifically, all three significant differences detected by our ABPP analysis agreed with the magnitude and direction of change for the global proteomics analysis. The most abundant protein in our ABPP dataset, Cyp3a11, was also reported as the most differentially expressed P450 in the global proteomics analysis. Our analysis detected substantially less P450s in total compared to the global proteomic analysis, but this is to be expected from a process that specifically enriches only active proteins. In addition, our ABPP analysis revealed germ-free mice had significantly increased activity for cytochromes 2d10 which is a known xenobiotic metabolizing P450.^39–40 Previous research evaluating germ-free mice and conventional mice have shown that, at baseline conditions, germ-free mice and conventional mice have markedly different mRNA profiles with germ-free mice having increased expression of 2d10,²⁴ although our data suggests there are few differences in the level of activity for enzymes detectable by our ABPP analysis.

In the entirety of our analysis, both kinetics and ABPP, the corn oil vehicle treatment unexpectedly leaves the most unanswered questions. Our ABPP data provocatively imply that corn oil treatment results in P450 activity profiles in which activity is disproportionately elevated in CV mice relative to GF mice. This could be attributed to host receptor landscape differences that therefore respond differently to lipid signaling, or the microbiota changes the nature of lipid metabolism to impact how they act as signaling molecules, or most likely a combination effect. It is noteworthy that the highest labeled P450s were the 3Cyp, and 4Cyp family, which are involved in cholesterol, steroid, and lipid synthesis, and arachidonic acid metabolism, which are all fatty substances. It is tempting to speculate, since corn oil is a fatty liquid, that the observed increase in P450 activity, primarily in CV mice, may have a connection to the observation that GF mice are resistant to weight gain from high fat diets.⁴¹ However, we would like to point out that the obesity resistance of GF mice has recently been called into question by several studies.⁴² Previous research has shown that dietary polyunsaturated fatty acids have a dramatic effect on gene expression including peroxisome proliferator-activated receptor (PPAR) and the pregnane X receptor (PXR) which regulates the cytochrome 3 and 4 families. ^43–45

Our ABPP data showed shared and divergent P450 features among the profiles of the PAH-treated animals. The major discernable difference in levels of P450 upon treatment with BaP, relative to corn oil controls, in both strains was the upregulation in activity for the Cyp1 family, which has been shown to increase upon benzo[a]pyrene exposure and it is worthy to note that it is one of the most important enzymes in the formation of DNA adducts.⁴⁶ Our ABPP findings show that 1a1, which is already known to be the key enzyme in BaP metabolism,⁴⁷ and 1a2 activity was significantly higher than the control in conventional mice. While 1a2 was detected, 1a1 was not detected in germ-free mice. While it is tempting to speculate about the lack of 1a1, we will refrain on making comparisons where a measurement was not recorded. Altogether, these data seem to suggest that the microbiome does impact BaP metabolism in terms of absolute magnitude, but it does not impact the elevation in P450 activity post-treatment.

Treatment with 1-NP had a much more dramatic effect relative to corn oil controls, and the two strains responded differently. All directional changes in P450 activity that were found in GF were also found in CV, but CV had additional examples that were not found in GF. However, even though all directional 1-NP-induced changes in GF were found in CV, the magnitude of the change was often different between the strains. It is interesting to note that for each mouse strain, cytochrome 1a1 was absent when exposed to 1-nitropyrene. As mentioned previously Cyp1a1 is the major metabolic enzyme induced when exposed to benzo[a]pyrene, yet it is absent when exposed to a structurally different PAH. Previous research has shown that 1-nitropyrene weekly induced the mRNA profiles of 1a1 but was shown to be inactive when evaluated with western blotting which is consistent with our ABPP analysis. ⁴⁸

When the ABPP data is looked at in total, there are only few detectable differences in unexposed mice, but there are clearly differences in how the strains respond to both vehicle and PAH treatment. This has important implications. With respect to the vehicle, for one, in terms of laboratory pharmacology studies, it will be crucial to understand liver versus microbiome impacts. In terms of drug administration, high fat consumption, possibly in the form of diet in humans, has a substantial gut microbiome-driven impact on the repertoire of active P450s in the liver. There are several notable examples in the Cyp2 family activity, which is linked to drug metabolism, being disproportionately upregulated in CV relative to GF mice in our study.

While there were a few significant microbiome-specific effects in response to BaP treatment, we noticed several microbiome-linked differences in the 1-NP treatment condition. The nitro group 1-NP is known to be reduced by the microbiome, and therefore, a more substantial response observed in CV mice could be due to the creation of structurally different metabolites. The magnitude of microbiome-specific differences between BaP and 1-NP metabolism highlight the need for a more thorough understanding of the specific molecular features that are within the purview of microbiome-modified liver metabolism because it does not uniformly affect PAH metabolism.

We foresee our study having several important impacts. Mammalian metabolism and general health cannot be solely attributed to the functions encoded by human genome, but also by the diverse contributions of the gut microbiome.⁴⁹ Following oral exposure, the gut microbiome is thought to be first to interact with xenobiotics with the highest exposure. ^50–52 It has been shown that the microbiome can affect hepatic drug processing genes and modify the adsorption, distribution, and metabolism of a variety of xenobiotics, however, there is a lack of literature quantifying these functional changes, specifically for the rate of metabolism and active enzymes.^{8, 53–54} Here we show that the gut microbiome contributes to the rates of metabolism and P450 profiles in unexposed conditions, vehicle-treated and upon exposure to two structurally different polycyclic aromatic hydrocarbons. To gain a more complete picture, it is important to determine why and how the lack of gut microbiome alters active enzymes and rates of metabolism, to do this we will incorporate metabolomics and metagenomic approaches in future experiments. Furthermore, utilizing individual PAHs resulted in impactful information, but environmental exposures are complex and further research is needed to determine the microbiome’s effect on environmentally relevant PAH mixtures. Understanding the role that the gut microbiome plays in xenobiotic metabolism will not only aid in demystifying host-microbial interactions, but it will aid in determining overall risk across populations and could elucidate individual susceptibility. This knowledge can pave the way for designing personalized treatment plans, better informed dose response, and aid in risk assessment decisions for contaminated sites. Finally, we hypothesize that the microbiome’s modulation of host enzyme function is not limited to P450s and that there is ample opportunity in applying function-focused approaches, such as ABPP, to understand the scope and magnitude of these links.

Table 1.

Metabolism parameters for hepatic microsomes from conventional and germ-free mice unexposed, corn oil vehicle, and with 180 mg/kg/day benzo[a]pyrene for 72 hrs. Statistically significant differences between CV and GF mice for each respective treatment are denoted with an asterisk.

Exposure	Mouse	BaP Vmax (nmol/min/mg)	95% CI	BaP KM (μM)	95% CI	BaP CLint (ml/min/mg)	95% CI
Water	CV	0.47*	0.29–1.19	1.29	0.44–15.16	0.36	0.07–0.68
	GF	5.36*	3.07–16.1	29.63	13.40–114.53	0.18	0.14–0.29
Corn oil	CV	1.21*	1.13–1.31	4.38*	3.43–6.28	0.28	0.20–0.34
	GF	0.57*	0.48–0.69	2*	1.47–3.39	0.29	0.20–0.33
BaP	CV	1.52*	1.34–1.75	2.64	1.91–4.15	0.58	0.41–0.72
	GF	0.91*	0.74–1.16	1.67	1.04–3.27	0.54	0.33–0.72

Table 2.

Metabolism parameters for hepatic microsomes from conventional and germ-free mice unexposed, corn oil vehicle, and with 180 mg/kg/day 1-nitropyrene for 72 hrs. Statistically significant differences between CV and GF mice for each respective treatment are bolded with an asterisk.

Exposure	Mouse	1-NP Vmax (nmol/min/mg)	95% CI	1-NP KM (μM)	95% CI	1-NP CLint (ml/min/mg)	95% CI
Water	CV	0.11	0.07–4.76	2.7	1.09–4.84	0.04*	0.01–0.08
	GF	1.1	0.88–1.41	3.5	2.34–5.56	0.31*	0.25–0.38
Corn oil	CV	1.36	0.56–1.98	1.28	0.76–3.24	1.06	0.10–2.44
	GF	0.7	0.46–0.99	1.29	0.61–2.15	0.54	0.43–0.76
1-NP	CV	0.7*	0.67–0.73	0.94	0.85–1.04	0.74*	0.68–0.80
	GF	0.09*	0.06–0.16	1.16	0.57–3.45	0.08*	0.04–0.11

ABBREVIATIONS

PAH

Polycyclic aromatic hydrocarbon

GF

Germ-free mice

CV

Conventional mice

1-NP

1-nitropyrene

P450

Cytochrome P450