Net release and uptake of xenometabolites across intestinal, hepatic, muscle, and renal tissue beds in healthy conscious pigs

Abstract

Xenometabolites from microbial and plant sources are thought to confer beneficial as well as deleterious effects on host physiology. Studies determining absorption and tissue uptake of xenometabolites are limited. We utilized a conscious catheterized pig model to evaluate interorgan flux of annotated known and suspected xenometabolites, derivatives, and bile acids. Female pigs (n = 12, 2–3 mo old, 25.6 ± 2.2 kg) had surgically implanted catheters across portal-drained viscera (PDV), splanchnic compartment (SPL), liver, kidney, and hindquarter muscle. Overnight-fasted arterial and venous plasma was collected simultaneously in a conscious state and stored at −80°C. Thawed samples were analyzed by liquid chromatography-mass spectrometry. Plasma flow was determined with para-aminohippuric acid dilution technology and used to calculate net organ balance for each metabolite. Significant organ uptake or release was determined if net balance differed from zero. A total of 48 metabolites were identified in plasma, and 31 of these had at least one tissue with a significant net release or uptake. All bile acids, indole-3-acetic acid, indole-3-arylic acid, and hydrocinnamic acid were released from the intestine and taken up by the liver. Indole-3-carboxaldehyde, p-cresol glucuronide, 4-hydroxyphenyllactic acid, dodecanendioic acid, and phenylacetylglycine were also released from the intestines. Liver or kidney uptake was noted for indole-3-acetylglycine, p-cresol glucuronide, atrolactic acid, and dodecanedioic acid. Indole-3-carboxaldehyde, atrolactic acid, and dodecanedioic acids showed net release from skeletal muscle. The results confirm gastrointestinal origins for several known xenometabolites in an in vivo overnight-fasted conscious pig model, whereas nongut net release of other putative xenometabolites suggests a more complex metabolism.

NEW & NOTEWORTHY Xenometabolites from microbe origins influence host health and disease, but absorption and tissue uptake of these metabolites remain speculative. Results herein are the first to demonstrate in vivo organ uptake and release of these metabolites. We used a conscious catheterized pig model to confirm gastrointestinal origins for several xenometabolites (e.g., indolic compounds, 4-hydroxyphenyllactic acid, dodecanendioic acid, and phenylacetylgycine). Liver and kidney were major sites for xenometabolite uptake, likely highlighting liver conjugation metabolism and renal excretion.

INTRODUCTION

Although orally consumed xenometabolites (e.g., small molecules derived from foods, natural products, pharmaceuticals, industrial sources) have been extensively identified, microbe-derived xenometabolites remain poorly characterized. As it becomes more appreciated that the gut microbiota influences host health and disease (27), characterizing microbial production and bioavailabilities of xenometabolites and their derivatives becomes paramount to understanding how the microbiota exerts its influence on host physiology.

There is ample evidence that microbially derived xenometabolites, or microbial modification of host metabolites, contribute to host physiology (8, 10, 33, 37, 48). Microbial production of butyrate has been shown to induce β-oxidation in colonic epithelial cells and promote the expansion of colonic regulatory T cell, promoting a positive feedback loop to maintain a symbiotic relationship between the colonic epithelia and obligate anaerobes (7). Other studies have shown that the microbial breakdown product of tryptophan, indole, modulates intestinal tight-junction integrity (4) and glucagon-like peptide 1 secretion (11). In addition to the direct effects on intestinal tissues, several reports have suggested associations between microbe-derived xenometabolites and cardiovascular disease (41), kidney disease (24), obesity (9), type 2 diabetes (34), and several other diseases (20, 50).

Metabolomics has been a key tool to identify target tissues for some microbe-derived xenometabolites in various host tissues (2, 23, 29, 39); however, careful assessment of putative xenometabolite production, absorption, transportation, and interorgan flux has never been conducted. Several reasons why this area of research is lacking include the following. 1) The gut microbial ecology is notoriously difficult to recapitulate in an ex vivo system and still requires validation in vivo (38); 2) commonly used model systems, i.e., mouse and rat, have differences in physiology and function in the intestinal tract compared with humans (21); and 3) organ flux analyses requires ex vivo perfusion techniques or isotopic labeling administration. To address some of these limitations, we utilized an in vivo multiorgan metabolic flux model in conscious pigs (18). The pig gastrointestinal anatomy, physiology, and metabolism has been recognized as a translatable research model for humans (52), and the larger size of the animal allows for surgical interventions that are not feasible in smaller-animal models (e.g., mice/rats). This model allows for simultaneous measurement of plasma flow for the portally drained viscera (PDV), splanchnic compartment (SPL), kidney, and skeletal muscle, providing an estimate of organ metabolite uptake and release (40). The aim for the current study was to use a specialized metabolomics platform (XenoScan) enriched in xenometabolites, their derivatives, and gastrointenstinal (GI)-related compounds (bile acids) to characterize the origins and potential organ targets of putative xenometabolites and microbe-modified metabolites in healthy postabsorptive conscious pigs.

MATERIALS AND METHODS

Animals.

Twelve 2- to 3-mo-old female pigs (25.6 ± 2.2 kg) were used in this study. Pigs were purchased from a commercial breeder (Rosenbaum Farms, Brenham, TX) and then individually housed in galvanized bar runs (2 × 3 m) at Texas A & M University. Pens were enriched with straw and toys for acclimatization before surgical catheter placement. Pigs were maintained at 21–25°C during a 12-h light-dark cycle, with a radio turned on during the light period. Pigs were provided 1 kg/day of swine feed (7200 Harlan-Teklad Vegetarian Pig/Sow Grower; Envigo, Indianapolis, IN). Water was available ad libitum. All animal experiments were approved by the Institutional Animal Care and Use Committee of Texas A & M University.

Surgical procedure.

In-depth details of the surgical technique have been previously published (13, 17). Briefly, pigs were sedated using an intramuscular injection of tiletamine-zolazepam (3.3 mg/kg, Telazol; Zoetis, Kalamazoo, MI) after a 16-h fast. Animals were intubated, and anesthesia continued with isoflurane (2% in oxygen). A total of eight intravascular catheters were implanted (locations described below) in addition to a feeding tube inserted percutaneously into the stomach.

Tissue metabolite flux was measured using para-aminohippuric acid (PAH) infusion (17) by placing two catheters upstream of organs. For muscle flux measurements, a PAH infusion catheter was implanted into the abdominal aorta, with its tip 5 cm above the bifurcation. The other PAH infusion catheter was implanted into the splenic vein. Sampling catheters were inserted into the following sites for direct sampling: (arterial) in the abdominal aorta above the right renal artery for pre-organ compartment arterial plasma concentrations; (venous) into the inferior caval vein (iliac circumflex profunda vein), with its tip 5 cm above the bifurcation for muscle flux measurements and venous concentrations; (renal) into the left renal vein; (portal) into the portal vein with its tip in the liver hilus; (hepatic) into the hepatic vein by direct puncture of the liver for splanchnic measurements. One extra central vein catheter was implanted in the caval vein for administration of medicine. All catheters were secured in place and tunneled through the left abdominal wall. Pigs were dressed with a canvas harness to protect the catheters after abdominal closure. Catheters were filled with 0.5 mL of gentamycin (20 mg/mL) and α-chymotrypsin solution (225 U/mL) to keep the catheters patent.

Catheterized pigs were checked twice daily for body temperature, catheter patency, and overall behavior for 4 days after surgery. During these checkups, antibiotics (6.25 mg/kg lincomycin and 12.5 mg/kg spectinomycin) and analgetics (2 mg/kg flunizin meglumine) were administered via central vein catheter. Pigs were allowed to recover for 7–10 days while habituating to the experimental cage (0.9 × 0.5 × 0.3 m).

Experimental procedure.

Animals remained conscious during the whole experimental procedure. Starting at 0800, pigs were weighed and then given a continuous infusion of 25 mM PAH at 60 mL/h, reaching steady state before sampling (40). Samples were then taken in the following order: arterial, portal, hepatic, venous, and renal. Three samples were collected in Li-heparine tubes at each catheter. All collected blood samples were immediately placed on ice. Blood was centrifuged (8,000 g, 5 min, 4°C). For the PAH concentration measurements, 250 μL of plasma was transferred into a tube containing 25 μL of trichloroacetic acid (TCA) and then vortexed. The rest of the plasma were collected into clean tubes. All samples were snap-frozen in liquid nitrogen and then stored at −80°C until further analyses. Pigs were returned to their normal cages after the conclusion of the experimental procedure, with their food and water available.

PAH concentrations and flow calculations.

PAH concentrations were compared with PAH standards and assessed using a microplate spectrophotometer (Spectramax; Molecular Devices, Sunnyvale, CA) and SoftmaxPro software (Molecular Devices). Plasma flow through organs was calculated using the dilution of PAH over the organ compartment (13, 17). Samples were collected after PAH reached steady state, so PAH_IN = PAH_OUT. Plasma flow through the hindquarter muscle (HQ), the intestine (PDV; portal drained viscera), and the splanchnic compartment (SPL; PDV + liver) was calculated using the following formula:

$$\text{flow}=\frac{\text{infusion\hspace{0.17em}rate}\hspace{0.17em}\times {\left[\text{PAH}\right]}_{\text{infused}}}{{\left[\text{PAH}\right]}_{\text{post}}-{\left[\text{PAH}\right]}_{\text{pre}}}\text{.}$$

The PAH infusion site for the HQ is the catheter in the abdominal aorta, and the site for PDV and SPL is the catheter into the splenic vein. [PAH]_pre is the PAH concentration in the main bloodstream, sampled through line “arterial.” [PAH]_post is the PAH concentration in the efferent vein of the target organ: line portal for PDV, line hepatic for SPL, and line venous for HQ. Plasma flow through the kidneys can be calculated as follows:

$${\text{flow}}_{\text{kidneys}}=\frac{\text{infusion\hspace{0.17em}rate}\hspace{0.17em}\times {\left[\text{PAH}\right]}_{\text{infused}}}{{\left[\text{PAH}\right]}_{\text{pre}}-{\left[\text{PAH}\right]}_{\text{post}}}\text{,}$$

where the extracted PAH is the infused [PAH] from both infusion sites.

Xenometabolomics.

Metabolomics assessment was conducted by liquid chromatography-mass spectrometry using previously described methods (28). Subsequent analysis focused on known and putative xenometabolites and their derivatives, using the XenoScan platform based on the Arkansas Children’s Nutrition Center’s library of authentic standards (28). Briefly, duplicate serum samples (100 µL) were extracted in methanol (2:1). Pooled quality control (QC) samples (n = 12) were then prepared by pooling equal volumes of each sample extract. Samples and QC extracts were evaporated to dryness under a nitrogen stream and reconstituted in 300 µL of 5% aqueous methanol containing a mix of internal standards [Lorazepam, D6-trans-cinnamic acid, and D4-glychocholic acid; 5.1 μM final concentration (Sigma Aldrich, St. Louis, MO)]. Chromatography was performed on a Dionex Ultimate 3000 UHPLC using an XSelect CSH C18 reversed phase column (2.1 × 100 mm, 2.5 μm) kept at 49°C, as previously described (28). Sequence order of all samples, including duplicates, was randomized before analysis. Metabolites were detected with a Q-Exactive Hybrid Quadrupole-Orbitrap mass spectrometer. All samples were analyzed by positive and negative electrospray ionization (ESI+/−) full-MS scan mode. Xcalibur 4.0 software was used for data acquisition (28).

Acquired data (full MS and data-dependent MS²) were processed using Compound Discoverer 3.0 using an untargeted metabolomics workflow. In-depth details of parameters associated with this workflow have been published previously (28). Metabolites were identified by using MassList, (accurate mass ± 5 ppm, RT ± 15 s to a known standard), mzCloud (online ddMS₂ database), and mzVault (in house ddMS₂ database) and given the following confidence levels: level 1, identification if accurate mass, retention time, and MS² spectra matching to mzVault or mzCloud; and level 2, identification if accurate mass, retention time matching to known standard, and no ddMS₂ information.

Peak area values were generated on all metabolites available in our in-house library (i.e., not limited to only xenometabolites and microbial comodified metabolites). All available data were in data preprocessing steps. Relative standard deviation (RSD) of QC samples averaged 14.5% and 11.4% in negative and positive modes, respectively. Cross-contribution compensating multiple internal standard normalization (CCMN) was used to lower between-sample variation (36). All data, including QC samples, were assessed by principal component analysis (PCA) to ensure proper clustering of QCs and to check for sample outliers. Four samples from positive-mode data had PCA scores that were >3 SD from the average PCA score along either PC1 or PC2; these samples were removed from further analyses, resulting in n = 10 and 11 for portal and hepatic sites, respectively. Metabolite distributions within each sample were also plotted and checked for deviations. Univariate outliers were iteratively screened using Grubb’s Test, removed, and then imputed with the K-Nearest Neighbor algorithm (46); <1% of data were affected by the outlier removal procedure in both positive and negative modes. Median RSD among duplicate samples was 13.9% and 17.5% for positive- and negative-mode data, respectively. Average of duplicate samples were calculated and used in further analyses.

Net organ balance calculation.

$$\text{Net\hspace{0.17em}balance}=\text{plasma\hspace{0.17em}flow}\times ({\left[\text{metabolite}\right]}_{\text{pre}}-{\left[\text{metabolite}\right]}_{\text{post}})\text{,}$$

where “plasma flow” is the mean plasma flows of all pigs, as described (40), and [metabolite] is normalized MS peak areas. Net balance was expressed in peak area·kg body wt⁻¹·min⁻¹.

Statistical analysis.

All statistical analyses were conducted in the R Statistical Language (version 3.6.2; The R Foundation for Statistical Computing, 2019); α was set at 0.05 unless otherwise noted. Values provided in text are means ± SE natural log-transformed peak area. PCA was used to visualize whether variance in absolute metabolomics data could be explained by port location or individual pigs using the first two PCA components. Linear mixed models were developed for each metabolite to assess the main effect of port location, with pig ID as a random effect. Log likelihood ratio tests comparing the models including the main effect of port location against an intercept only model were conducted to determine statistical significance of the main effect. Net organ balance rate data were tested against zero (H₀: net organ flux is equal to 0) using a one-sample Wilcoxon signed-rank test. Every net balance comparison is graphically displayed in Supplemental Fig. S1 (Supplemental Material for this article is available online at https://doi.org/10.15482/USDA.ADC/1518604). All results from hypothesis-based statistical tests were adjusted for multiple comparisons using the false discovery rate procedure by Benjamini and Hochberg (6).

RESULTS

Absolute differences in port collections.

A total of 48 xenometabolites were structurally identified in pig plasma across all tissue ports (Supplemental Table S1). Average peak heights of all metabolites were greatest in hepatic drainage compared with all other sampling sites (hepatic: 15.7 ± 0.08; arterial: 15.3 ± 0.07; venous: 15.1 ± 0.10; portal: 15.4 ± 0.06; renal: 15.4 ± 0.06), whereas venous drainage was significantly lower than portal and renal port collections. PCA analysis identified significant between-pig variation, with ports clustering closely within each pig (Fig. 1A). No major variation was associated with individual ports along PC1; however, metabolites in renal collection differed from portal, hepatic, and muscle collections along PC2 (Fig. 1B).

Fig. 1.

Principal component analysis (PCA) of catheter port of xenometabolites (ion peak area) in conscious female pigs (n = 12). PCA scores are colored by pigs (A) and port (B). Pigs were overnight fasted (postabsorptive). Sampling sites included arterial, venous, renal, portal, and hepatic placed catheters. PCA scores (i.e., variance associated with individual pig) are overlayed with pig ID (A) and port classifiers (B). Data were scaled to unit variance before PCA. Confidence ellipses are based on Hotelling’s T statistic. Axis scaling differs between A and B to facilitate interpretation of confidence ellipses. Box plots in B describe distribution between scores along PCA component 2. Lines connecting 2 boxes indicate differences between port location along component 2 using Tukey’s test at P < 0.05.

Raw differences in xenometabolite and bile acid abundances between portal collections are described in Supplemental Table S1. A total of 30 xenometabolites had at least one collection site that differed from the other sites (Supplemental Fig. S2). Two major patterns emerged for the 30 metabolites that displayed one or more port differences: 1) metabolites with greater hepatic plasma concentrations relative to the other sites or 2) lower venous concentrations relative to all other sites. Metabolites patterned against the former 1) included bile acids, several indole derivatives, hydrocinnamic acid, and stercobilin; the latter pattern 2) included p-cresol glucuronide, phenylacetylglycine, o-hydroxyhippuric acid, N-cinnamylglycine, indoleacetylglycine, indoxyl-β-d-glucuronide, hippuric acid, gluconic acid, phenol, dodecanedoic acid, 4-hydroxyphenyllactic acid, and trigonelline.

Bile acids are excreted from intestines and taken up by the liver.

Bile acids are synthesized in the liver, excreted into the small intestine via chime, modified by gut bacteria, reabsorbed in the colon, and then recycled by the liver. Thus, their inclusion in this analysis provides a set of positive control metabolites for the flux analysis. As expected, we found that all primary and conjugated bile acids had a similar pattern in that there was a net release of bile acids from the portal vein and a net uptake only in the liver (Fig. 2). Because there was no net release from the splanchnic pool, this pattern suggests that little to no bile acids escape the liver-intestinal circulation.

Fig. 2.

Interorgan flux of bile acids in conscious overnight-fasted female pigs (n = 12). A negative flux reflects net uptake (flux < 0; indicated by downward arrow) by the organ, whereas a positive flux reflects a net release (flux > 0; indicated by upward arrow) by the organ. Plasma flow measurements were determined with the para-aminohippuric acid dilution method and were used along with metabolite concentrations to calculate flux (see materials and methods). *Organ that has a statistically significant net flux greater or lower than zero (1-sample Wilcoxon signed-rank test; P < 0.05). bMCA, β-muricholic acid; CDCA, chenodeoxycholic acid; GHDCA, glycohyodeoxycholic acid; GLCA, glycolithocholic acid; HDCA, hyodeoxycholic acid; HQ, hindquarter compartment; LCA, lithocholic acid; PDV, portal drained viscera; SPL, splanchnic compartment.

Conjugated indole molecules are released mainly from liver and kidney and taken up by both gut and liver.

Of the seven indole derivatives identified in pig plasma, only four were found to have a significant flux among surveyed tissues (Fig. 3). Indole-3-acetic acid, indole-3-acrylic acid, and indole-3-carboxaldehyde all showed a net release from the intestines; however, only indole-3-acetic acid and indole-3-acrylic acid were released from the gut. Indole-3-carboxaldehyde was also released from skeletal muscle tissue. Similarly, uptake of both indole-3-acetic acid and indole-3-acrylic acid occurred in the liver, but the uptake of indole-3-carboxaldehyde could not be determined. Contrary to the other indole derivatives, no tissue release of indole-acetylglycine could be determined, and concentrations were highly variable in blood from ports interrogating SPL and liver tissue beds; net uptake occurred in the kidneys. p-Cresol glucuronide, another microbe-derived product of amino acid catabolism, is released by either the gut or liver, as suggested by the combined net release from the portal and splanchnic drainage. Based on the flux analysis, kidney is the main uptake site of p-cresol glucuronide.

Fig. 3.

Interorgan flux of microbial breakdown products of aromatic amino acids in conscious overnight-fasted female pigs (n = 12). A negative flux reflects net uptake (flux < 0; indicated by downward arrow) by the organ, whereas a positive flux reflects a net release (flux > 0; indicated by upward arrow) by the organ. Plasma flow measurements were determined with the para-aminohippuric acid dilution method and were used along with metabolite concentrations to calculate flux (see materials and methods). HQ, hindquarter compartment; PDV, portal drained viscera; SPL, splanchnic compartment. *Organ that has a statistically significant net flux greater or lower than zero (1-sample Wilcoxon signed-rank test; P < 0.05).

Net organ flux of phenols or phenolic derivatives is widely variable among tissues.

Hydrocinnamic acid, a phenylpropanoic acid, was only released from the gut and taken up by the liver (Fig. 4A). No release or uptake of hydrocinnamic acid was found in other tissues. No clear release of 4-hydroxybenzoic acid was determined from surveyed tissues, and variability was high for the PDV; however, net uptake of this metabolite was identified in kidney and liver (Fig. 4A). In contrast, 4-hydroxyphenyllactic acid was released from intestines and skeletal muscle, but no tissues had a clear, significant net uptake (Fig. 4A).

Fig. 4.

Interorgan flux of phenolic and other xenometablites in conscious overnight-fasted female pigs (n = 12). A negative flux reflects net uptake (flux < 0; indicated by downward arrow) by the organ, whereas a positive flux reflects a net release (flux > 0; indicated by upward arrow) by the organ. Plasma flow measurements were determined with the para-aminohippuric acid dilution method and were used along with metabolite concentrations to calculate flux (see materials and methods). HQ, hindquarter compartment; PDV, portal drained viscera; SPL, splanchnic compartment. *Organ that has a statistically significant net flux greater or lower than zero (1-sample Wilcoxon signed-rank test; P < 0.05).

Other metabolites with no clear chemical class-cluster relationship that had significant tissue flux included atrolactic acid, dodecanedioic acid, and phenylacetylglycine (Fig. 4B). Both atrolactic acid and dodecanedioic acid were released from the intestines and skeletal muscle and had no defined tissue of uptake. Release of phenylacetylglycine, a glycine conjugate of phenylacetic acid, from both the portal and splanchnic drainage is consistent with a net release by liver (a well-known primary site of glycine conjugation). Significant kidney uptake for this metabolite is suggested.

DISCUSSION

We used liquid chromatography-mass spectrometry to identify a broad range of xenometabolites from our in-house standard library that has been curated for known or putative xenometabolites from food or microbial sources. This XenoScan library was constructed based on known literature and not through exometabolomics (42); thus, many of the microbial xenometabolites are considered putative xenometabolites that will require validation. However, it should be noted that we do not know of a validated model to conclusively recapitulate the complex intestinal dynamics related to host-microorganism and microorganism-microorganism metabolite exchange in addition to host absorption. For example, in silico assessment of the human metagenome suggests that the majority of bacterially synthesized secondary bile acids require multiple bacteria (19). If done correctly under anaerobic conditions, culturing still remains an effective technique to identify bacterial metabolism under well-defined conditions; however, the translatability in vivo is still questionable in addition to whether products seen in vitro are absorbed in the host organism. Thus, the ability to identify xenometabolites in PDV in a conscious pig model provides a unique system to validate absorption and trafficking of xenometabolites.

The ability of this model to capture gut-derived absorbed molecules was highlighted by the gut/liver exchange of bile acids. Because all bile acids showed the expected enterohepatic pattern (i.e., release from intestines and uptake in the liver), we feel that they act as a positive control for the organ flux model. Two xenometabolites, hydrocinnamic acid and indole-3-acetate, had the same flux pattern observed with bile acids. Hydrocinnamic acid, a phenylpropanoic acid commonly used as a flavoring agent, has also been shown to be synthesized from cultured Clostridium spp (31) in addition to industrial production by Escherichia coli (44). Interestingly, Elsden et al. identified (14) both phenylpropanoic acid and indole acetate as products formed from Clostridia cultures provided phenylalanine, tyrosine, and tryptophan, further supporting a microbial source for hydroxycinnamic acid. Although there is little literature related to health indications of hydrocinnamic acid, there has been increasing interest in the microbe-derived indole compounds. Indole-3-acetic acid in particular was recently shown to modulate tumor necrosis factor-α (TNFα)-stimulated lipid accumulation in multiple liver cell lines in addition to reducing TFNα-stimulated gene expression of fatty acid synthase and sterol regulatory element-element binding protein (26). This further illustrates the likely importance of bioavailable xenometabolites to host physiology.

Our results are purely observational and do not speak to mechanistic implications of any identified metabolites. However, these data do indicate that many of the microbial products of aromatic amino acid breakdown are readily absorbed by the gut. Of all the identified indoles and cresol, only indole-acetylglycine did not show a net release from the intestines, which was to be expected since the liver is a major site of glycine conjugation (3). Absorption of indolic compounds have been well documented in humans and pigs (16, 25); however, data regarding tissue distribution and organ flux of these compounds is limited. Several pig studies have studied the relationships between intestinal skatole and indole production and their accumulation in fat and muscle (1, 25, 49), suggesting that these compounds are distributed more widely throughout the host organism. Our methods do not provide adipose flux balance, but we did observe renal uptake for p-cresol glucuronide and indole-acetylglycine. The former has been widely suggested as a uremic solute (15, 35) and has been shown to be an independent predictor of adverse outcomes related to chronic kidney disease (35). Interestingly, indole-3-carboxaldehyde and p-cresol-glucuronide were also released from muscle and liver, respectively, suggesting nonintestinal origins for these metabolites. Because these are likely host-modified derivatives of the parent compound, the release from nonintestinal tissue may reflect the release of organ-specific metabolism. This is likely the case for metabolites released from the liver due to hepatic conjugation metabolism. Release from other organs will require future investigations to determine the kinetics associated with product uptake, metabolism, and net release.

Of the remaining identified metabolites, 4-hydroxyphenyllactic acid, atrolactic acid, and dodecanedioic acid were released from the intestines and skeletal muscle. In addition to being a product of phenylalanine catabolism and a key marker of phenylketonuria, 4-hydroxyphenyllactic acid is also derived from bacterial breakdown of phenylalanine and tyrosine (5). Although it has been speculated that gut bacterially derived hydroxyphenyllactic acid is absorbed by the colon (43), our study is the first to demonstrate this in vivo. Atrolactic acid, also known as 2-phenyllactic acid, is commonly found in fermented food (47) but is also an intermediate in the conversion of phenylalanine (12); thus, its release from intestines and muscles appears to be justified in this model. Dodecanedioic acid is a C12 dicarboxylic acid, which is thought to be produced by plants and animals, the latter due to ω-oxidation of longer dicarboxylic acids or lauric acid (30). Paradoxically, ω-oxidation is not thought to occur in skeletal muscle but could be more apparent in brain (30). A phenolic derivative of benzoic acid, 4-hydroxybenzoic acid, is commonly found in plants and certain foods but also in different lines of bacteria (22, 32). This particular model does not confirm intestinal absorption, but it does indicate hepatic and renal uptake, likely for conjugation and/or excretion. Finally, phenylacetylglycine, a glycine conjugate of phenylacetate, was released from the splanchnic drainage, indicating net flux of this metabolite from both the intestines and liver. This particular metabolite has been shown to be elevated with inborn errors of metabolism as well as through microbial synthesis of phenylacetate, which is then glycine-conjugated (45). Uptake by the kidneys likely indicates renal clearance in this case.

Our study is the first report to use a conscious animal model to interrogate xenometabolites and host metabolites altered by microbes, providing an in vivo assessment of the distribution of these metabolites. The study also utilized a porcine model, which is arguably the closest nonprimate gastrointestinal tract resembling the physiology of the human gastrointestinal tract. Finally, by measuring arterial and venous blood flow, we were able to determine metabolite flux from surveyed organs. This confirmed the flux of several known flux pathways (e.g., enterohepatic bile acid exchange) and provided direct evidence of intestinal xenometabolite absorption and tissue uptake. Still, we would be remiss not to acknowledge the limitations in this study. Although this is the first in vivo study to track xenometabolites in a conscious animal model, the results are still observational and do not prove that absorption occurred, nor do the data provide any mechanistic information. We do not have any paired fecal samples to prove that particular microbially derived xenometabolites are produced in the gastrointestinal lumen. Furthermore, the model currently allows us to survey only a few organs, but other organs may take up or release xenometabolites. Our study was conducted only in female pigs due to commercial and surgical limitations, so we cannot rule out sex differences in the interorgan flux of these metabolites. Finally, the XenoScan platform is limited to in-house standards of currently known or putative xenometabolites, but certainly the xenometabolome is composed of many hundreds to thousands of compounds that await characterization. Nevertheless, the current study provides the most comprehensive overview of xenometabolite interorgan kinetics to date.

This model provides a unique in vivo system to investigate how gut microbial metabolism has the potential to influence host physiology in multiple organs. Future short- and long-term studies will investigate the effects of how dietary foods influence the metabolite signals emanating from the gut microbiome. For example, nondigestible carbohydrate consumption is known to promote bacteria that specialize in carbohydrate fermentation (51). Although short-chain fatty acids are typically the primary product of bacterial carbohydrate fermentation, other xenometabolites have been identified (23). Thus, this model will allow a longitudinal assessment of postprandial absorption and trafficking of xenometabolites that can be directly related to changes in the metagenome or metatranscriptome. Other potential utilities of this model could include seeding of microbes with known enzymatic capabilities to track specific xenometabolites in isolation, thus providing an in vivo model to investigate the efficacy of probiotics with targeted outcomes.

In conclusion, the results from this study confirm gastrointestinal origins for a number of known xenometabolites and microbe-modified host metabolites under postabsorptive conditions in conscious female pigs; however, other putative xenometabolites may be released from other organs, suggesting nonmicrobial origins or tracking of host-modified compounds. Principle organs for xenometabolomics uptake were the liver and kidney, likely suggesting hepatic conjugation and/or renal excretion.

GRANTS

This work was funded by United Stated Department of Agriculture/Agricultural Research Service Project 6026-51000-012-06S.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

G.A.T.H., N.E.D., S.H.A., and B.D.P. conceived and designed research; K.E.M., G.A.T.H., L.P., and N.E.D. performed experiments; B.D.P. analyzed data; R.L., N.E.D., S.H.A., and B.D.P. interpreted results of experiments; B.D.P. prepared figures; B.D.P. drafted manuscript; K.E.M., G.A.T.H., L.P., R.L., N.E.D., S.H.A., and B.D.P. edited and revised manuscript; K.E.M., G.A.T.H., L.P., R.L., N.E.D., S.H.A., and B.D.P. approved final version of manuscript.