Glucuronidation metabolomic fingerprinting to map host-microbe metabolism

Nina R Boyle; Josh J Sekela; Mingxun Wang; Helena Mannochio-Russo; Jeong Joo Pyo; Min Soo Kim; Shuchang Tian; Imhoi Koo; Elliot S Friedman; Ceylan Tanes; Mallappa Anitha; Yuan Tian; Ethan W Morgan; Iain A Murray; Joseph P Zackular; Kyle Bittinger; James D Lewis; Gary H Perdew; Gary D Wu; Babette S Zemel; Pieter C Dorrestein; Jordan E Bisanz; Matthew R Redinbo; Andrew D Patterson

doi:10.1038/s41467-026-73398-1

. 2026 May 20;17:6646. doi: 10.1038/s41467-026-73398-1

Glucuronidation metabolomic fingerprinting to map host-microbe metabolism

Nina R Boyle ^1,^#, Josh J Sekela ^2,^#, Mingxun Wang ³, Helena Mannochio-Russo ⁴, Jeong Joo Pyo ⁵, Min Soo Kim ⁶, Shuchang Tian ⁶, Imhoi Koo ⁷, Elliot S Friedman ^8,⁹, Ceylan Tanes ^9,¹⁰, Mallappa Anitha ¹¹, Yuan Tian ^1,¹¹, Ethan W Morgan ^7,¹¹, Iain A Murray ^7,¹¹, Joseph P Zackular ^9,¹², Kyle Bittinger ^9,¹⁰, James D Lewis ¹³, Gary H Perdew ^7,¹¹, Gary D Wu ⁸, Babette S Zemel ¹⁰, Pieter C Dorrestein ⁴, Jordan E Bisanz ^1,^6,¹⁴, Matthew R Redinbo ^2,⁵, Andrew D Patterson ^1,^6,^7,^11,^14,^✉

PMCID: PMC13381908 PMID: 42161945

Abstract

Glucuronidation is an important detoxification pathway that operates in balance with gastrointestinal microbial β-glucuronidase (GUS) activity, which can regenerate bioactive metabolites from their glucuronidated forms. How this host-microbe interaction shapes the distribution and pool of glucuronidated metabolites (i.e., the glucuronidome) remains poorly understood. In this study, we employed pattern-filtering data science approaches in conjunction with untargeted LC-MS/MS metabolomics to map the glucuronidome in urine, serum, and colon/fecal samples from gnotobiotic and conventional mice, and in humans. We find that microbial colonization and GUS activity compress the colonic glucuronidome and expand urinary glucuronidome diversity, revealing a compartmental redistribution of glucuronidated metabolites. Reverse metabolomics of known glucuronidated chemicals and glucuronidation pattern filtering searches in public metabolomics datasets exposed the diversity of glucuronidated metabolites in human and mouse ecosystems. In summary, we present a glucuronidation fingerprint resource that provides broader access to and analysis of the glucuronidome. Together, this work establishes a scalable analytical framework and provides mechanistic insight into how microbial activity reshapes systemic glucuronidation, with implications for drug metabolism, diet-microbe interactions, and biomarker discovery.

Subject terms: Microbiome, Microbial communities

Here, the authors provide a framework to map glucuronidated metabolites and show that gut microbes shape their distribution across the body, with findings in mice supported by human data, where colonization and diet influence glucuronidation patterns.

Introduction

Glucuronidation, the enzymatic transfer of a glucuronic acid moiety from uridine diphosphoglucuronic acid, is an essential biotransformation pathway for detoxifying and eliminating endobiotic and xenobiotic chemicals¹. Hormones, neurotransmitters, dietary chemicals, carcinogens, toxicants, and an estimated 40–70% of medications undergo this process^2,3. Glucuronidation occurs in the liver, intestines, and kidneys— tissues that possess important detoxification activities^4,5. Once glucuronidation occurs in the liver, the modified metabolites can be excreted into bile and transported to the gastrointestinal tract (GIT) through biliary excretion². Glucuronidated metabolites are generally less biologically active than their original forms and are more water-soluble, which prevents them from diffusing into cells and promotes their excretion in urine and feces^1,5.

Within the GIT, glucuronidated compounds can either be excreted in feces or de-glucuronidated by gut microbial β-glucuronidase (GUS) enzymes. These hydrolases cleave the bond between the original compound and the glucuronic acid moiety, regenerating the active parent compound (i.e., aglycone)^2,6–9. Regenerated compounds can then be reabsorbed, participate in chemical signaling, or undergo further metabolism by both the host and the gut microbiota^1,2. The reabsorption process, known as enterohepatic recirculation, extends chemical exposure and can lead to highly variable secondary absorption¹⁰. Elevated activity of gut microbial β-glucuronidases has been associated with several diseases, including colon, breast, and lung cancers, inflammatory bowel diseases, diabetes, obesity, polycystic ovary syndrome, and drug and hormone toxicities^{2,6,7,11–14}. These connections highlight the importance of understanding the interaction between GUS activity and glucuronidation in health and disease.

Recent research has extensively characterized glycoside hydrolase 2 (GH2) β-glucuronidase enzymes within the gut microbiome (the collection of all β-glucuronidase enzymes herein referred to as the GUSome). Studies have identified hundreds of unique GH2 β-glucuronidase variants in humans and mice, grouped into distinct structural classes^9,15,16. Some structural classes (loop 1, mini-loop 1, and flavin mononucleotide [FMN] binding C-terminal domain [CTD]) proteins appear to specialize in deconjugating small metabolites, while other classes (loop 2 and no loop) target larger polysaccharides¹⁷. In contrast to the microbiome, mammalian enzymes have limited intrinsic ability to hydrolyze glucuronidated compounds, with each species encoding only one gene for β-glucuronidase^18–20, a no-loop protein that processes polysaccharides⁹. Despite advancements in characterizing β-glucuronidase enzymes, the collection of all glucuronidated metabolites in a biological sample, herein referred to as the glucuronidome, remains undefined.

Untargeted tandem mass spectrometry enables comprehensive metabolite coverage; however, its effectiveness relies on the availability of reference spectral libraries, which are often incomplete. For example, public repositories, such as PubChem²¹, include over a hundred million unique chemical structures, approximately 2300 of which are glucuronidated metabolites. Only 661 of those are naturally produced molecules²² because drug-glucuronide conjugates dominate the structures. More importantly, only a small percentage have MS/MS information available, which is critical for enabling their annotation and discovery from LC-MS/MS data²³. The most widely used MS/MS spectral libraries used in LC-MS/MS-based untargeted metabolomics include NIST 20, which contains 53 glucuronidated chemicals, as well as GNPS2²⁴ and MoNA. These platforms combine data from public libraries and contain 43 unique glucuronidated metabolites, including many unannotated metabolites. In silico approaches that deconjugate experimental spectra²⁵ or conjugate existing library aglycone spectra²⁶ are promising but have limited scalability, constrained by high computational demands for complex datasets, such as urine, and the availability of reference spectra.

In this study, we developed a pattern-based analytical framework to systematically detect and annotate glucuronidated metabolites in LC-MS/MS datasets. We used this framework to build a glucuronidome reference resource and to uncover how microbial β-glucuronidase activity may reshape the distribution of glucuronides across host compartments. By integrating animal models, public datasets, and longitudinal human samples, we link specific GUS structural classes to compartment-level changes in the glucuronidome. This combined resource and mechanistic approach establishes glucuronidation as a powerful lens for decoding host-microbe metabolic interactions and their impact on human health.

Results

Glucuronidated metabolite annotation

We employed a two-tiered approach to define the glucuronidome (Supplementary Fig. 8). The first involved conventional MS/MS library development using 134 commercially available glucuronidated chemicals and 551 potential aglycone metabolites from the IROA phytochemical and gut microbial libraries. Of the 134 glucuronidated chemical standards (Supplementary Data 12), MS/MS spectra were obtained for 131 and 124 in negative and positive mode, respectively. The MS/MS query tool, Mass Spec Query Language (MassQL)²⁷ enabled us to leverage the known neutral loss (NL) of 176.0321 Da and less common 194.0425 Da^{23,25,26,28–30} observed in both positive and negative ionization modes, and the characteristic MS/MS product ions of 113.0244, 85.0295, 75.0087, and 71.0139 Da that are generated from the fragmentation of the glucuronate moiety and observed in negative ionization mode^29,30, to identify which metabolite standards were glucuronidated. A paired written and spectral visualization of a MassQL query for both the NLs and fragments at m/z 113.0244 and m/z 85.0295 is shown in Fig. 1A.

Fig. 1 — A Example and MS/MS visualization of a MassQL query for identifying glucuronidated features in negative mode MS/MS data. Font colors in the query correspond to the MS/MS feature colors. B Molecular network of glucuronidated and phytochemical standards. Diamond nodes represent MS/MS features with glucuronidation fragmentation patterns after filtering with MassQL. Circle nodes are metabolite features with similar MS/MS fragmentation patterns but without the glucuronidation MS/MS pattern. The gray, diamond nodes are glucuronidated isomers of the nodes (MS/MS features) connected with dashed edges. Tan edges are indicative of the mass difference between d-glucopyranoside and d-glucuronide conjugated metabolites (-CH2OH → -COOH, Δ*m/z* 13.979). Solid gray edges are indicative of hydroxyl group (-OH) differences (Δ*m/z* 15.995). Other nodes are colored by a common aglycone.

Of the 131 glucuronidated MS/MS collected in negative mode, a NL with m/z 176.0321 and/or 194.0425 was detected for 123 (94%) of the spectra, whereas 116 (89%) glucuronidated chemicals produced MS/MS spectra with at least 2 characteristic fragments of the glucuronate (Supplementary Fig. 1A). Due to MS/MS mass range detection or ionization limitations, a neutral loss of 176 or 194 was not observed in the MS/MS for 6 glucuronidated chemicals (ethyl β-D-glucuronide, cotinine N-ɑ-D-glucuronide, pregnanolone 3-β-D-glucuronide, etonogestrel β-D-glucuronide, daidzein-7-sulfate-4′-β-D-glucuronide, and 25-hydroxyvitamin D3 3-glucuronide). The MS/MS spectra of all 131 standards were distinguishable as glucuronidated by visual inspection and MassQL pattern filtering, with characteristic NL and/or fragmentation patterns.

The glucuronidated, phytochemical, and gut microbial metabolite libraries were networked using classical molecular networking, which uses MS/MS similarity to map chemical relationships within the metabolome³¹. Molecular networking connected MS/MS spectra with shared aglycone substructures, such as glucoside and glucuronide conjugates, as indicated by the connected, same-color nodes (Fig. 1B). This demonstrates that molecular networking can capture substructural MS/MS similarities between conjugated metabolites by predicting structural relatedness. Additional structural differences, such as the number of hydroxyl (-OH) attachments (visualized as network edges/connections with Δm/z 15.995), are captured in the molecular network (Fig. 1B). Furthermore, combined analysis of the phytochemical and gut microbial metabolite standards with the glucuronidated standards confirmed that our glucuronidation pattern MassQL query was specific to glucuronide-conjugated metabolites and was not generating false positive returns for other glycoside conjugated metabolites or substructures we might expect to see in our sample sets. Querying the above generated libraries for the glucuronidated neutral losses of m/z 176.0321 and m/z 194.0425 resulted in a false discovery rate of 6.5%, and adding the two most common glucuronide fragments (m/z 113.0244 and m/z 85.0295, Supplementary Fig. 1A) showed stronger specificity for spectra of glucuronidated metabolites with a reduced false discovery rate of 4.6%. All false returns for the combined neutral loss and fragments query contained galacturonic acid and/or galactopyranuronsosyl-substructures, which are isomers of glucuronic acid and distinguishable from glucuronidated MS/MS spectra by a high intensity fragment at m/z 115.0038 (Supplementary Fig. 1B). As a secondary quality control step, all MS/MS spectra returned by the MassQL pattern filtering (and included in the glucuronidated resource and glucuronidome analysis) were visually inspected to ensure the presence of the characteristic glucuronidation MS/MS pattern.

The second tier of our analysis employed the glucuronidation fragmentation pattern established during library development to query untargeted LC-MS/MS data from the mouse experiments (Fig. 1A) for all spectra exhibiting glucuronidation features (i.e., the sample glucuronidome). We then used molecular networking to tentatively annotate glucuronidated features not present in public libraries or in our glucuronidation (GlcA) library.

Validation of glucuronidated metabolite mapping

Our pattern filtering and the molecular networking approach to detect and identify glucuronidated features was validated using urine collected from mice treated with acetaminophen (N-acetyl-para-aminophenol, APAP) (experimental design Supplementary Fig. 2A). Urine samples were collected 24 h after a single intraperitoneal injection of acetaminophen for untargeted LC-MS/MS analysis for APAP metabolites³². Acetaminophen metabolism is well characterized, with several known urinary glucuronic acid and sulfate conjugated metabolites^32–37. Furthermore, reference spectra for acetaminophen are available from many public MS/MS libraries, including GNPS2, but MS/MS spectra for the main phase 2 conjugates (glucuronidated and sulfated) and less abundant metabolites/conjugates are less accessible; thus, the APAP metabolites are “known unknowns,” optimal for testing the power of the pattern filtering and molecular networking approach.

After the urine data was generated and aligned, the MS/MS of over 5000 features were imported into the GNPS2 environment for pattern filtering with MassQL and Feature-based Molecular Networking (FBMN) analysis. Pattern filtering returned 404 MS/MS spectra of acetaminophen, and metabolites of xenobiotic and endogenous origin with the glucuronidation pattern of a neutral loss at m/z 176.0321 or 194.0425 and fragments at m/z 113.0244 and 85.0295 (Supplementary Data 1). The treatment with acetaminophen resulted in significant differences in both the urine metabolome (Fig. 2A) and glucuronidome (Fig. 2B), compared to untreated controls. The most prominent drivers of the glucuronidome differences were acetaminophen metabolites: acetaminophen glucuronide, thiomethyl acetaminophen glucuronide, methoxy acetaminophen glucuronide, and thioacetaminophen glucuronide (Fig. 2C), which were identified by annotation propagation with molecular networking, specifically FBMN.

Fig. 2 — Principal components analyses of individual acetaminophen (N-acetyl-para-aminophenol, APAP) or control treated mice (n = 8 in each), plotted on the first two principal components (PC1 and PC2) with 95% confidence ellipses for A all MS/MS metabolite features from urine in negative ionization mode B all putative glucuronidated features, defined as having a neutral loss with *m/z*176.0321 or 194.0425 and fragments at *m/z*113.0244 and 85.0295. C Differential abundance of glucuronidated features from the urine of APAP compared to control-treated mice (red = increase in abundance with APAP, blue = decrease in abundance with APAP, gray = no difference). D Molecular network of APAP metabolites (negative ionization mode). Nodes are labeled with metabolite *m/z*. Edges connecting phase 2 metabolites are labeled with the mass difference between the two nodes and colored by the phase 2 pair: glucuronide:parent edges are green, sulfate:parent edges are yellow, and glucuronide:sulfate edges are orange. E MS/MS mirror plot of APAP glucuronide (top) and APAP (bottom) detected in mouse urine. Green features are the APAP fragments, common to both MS/MS spectra. The gray fragment at *m/z* 326.0882 is the precursor *m/z* for APAP glucuronide. MS/MS features used in the MassQL query for glucuronidated features are indicated with “@.” F MS/MS mirror plot comparing APAP glucuronide detected in mouse urine (top) to an analytic standard (bottom). C, D Metabolite labels: a - APAP glcA; b - thiomethyl APAP glcA; c - APAP glcA; d - methoxy APAP glcA; e - thio APAP glcA; f - APAP sulfate; g - APAP; h - thio APAP; i - cystein-S-yl APAP; j - cystein-S-yl APAP [2M-H]^–; k - APAP mercapturate glcA; m- APAP mercapturate; n - methyl-3-thioAPAP; o - methyl-3-thioAPAP sulphoxide; p - methyl-3-thioAPAP sulfoxide; s - S-methyl-3-thioAPAP. PCoA R² values were calculated by Adonis tests. Individual GlcAs were tested using Student’s T test and considered significant with a Benjamini and Hochberg FDR < 0.1 and log2FC > 1.0 or log2FC < −1.0, full statistics in Supplementary Data 1. All statistical tests are two-sided, as applicable.

FBMN³⁸ is an alignment-based tool for connecting known and unknown metabolic features, enabling the visualization of MS/MS data as an interconnected map based on molecular similarity and implied structural relatedness. Before networking, the only annotated acetaminophen metabolite was acetaminophen itself. Overlaying the MassQL pattern-filtering results onto the network revealed several glucuronidated features within the acetaminophen-related subnetworks (Fig. 2D). Within the network, the cosine similarity and Δm/z between two nodes provide valuable information for interpreting the structural relatedness of connected metabolites. Molecular networks link metabolites based on similarities in their MS/MS fragmentation patterns, but the approach is unsupervised and relies on filtering and sorting steps that simplify the network by reducing the number of edges. As a result, these networks do not necessarily mirror canonical metabolic pathways. Even so, incorporating additional chemical knowledge can help extend and refine metabolite annotations. For example, in the molecular network in Fig. 2D for APAP (g), APAP sulfate (f) is directly connected to APAP with Δm/z = 79.957 and APAP glucuronide (a/c) is connected to APAP-sulfate with Δm/z = 96.075 (glucuronate m/z 176.032–sulfate m/z 79.957), which allowed the identification of APAP glucuronide as a related glucuronide without the classical Δm/z = 176.0321 edge connection to APAP. The cosine similarity between two nodes, for example, acetaminophen-glucuronide and acetaminophen, is an indicator of fragmentation similarities (Fig. 2E) and potential shared substructures. Combining the edge information with shared MS/MS fragments and the presence/absence of the glucuronidation pattern, high-confidence “putative acetaminophen-related” annotations were assigned to several nodes. The putative annotations for acetaminophen glucuronide (Fig. 2F) and sulfate were confirmed using reference standards, and a literature search corroborated the annotations for the thio, mercapturic, methyl thio, and methoxy metabolites of acetaminophen³⁶. From the 5696 urinary metabolites, FBMN grouped the acetaminophen metabolites, tending to segregate by metabolism pathway (Fig. 2D). The metabolite nodes branch out as the metabolites are further changed from the core structure of acetaminophen. These results show that molecular similarity networking is a powerful tool for overcoming the limitations of MS/MS reference libraries, exploiting shared MS/MS fragmentation patterns to propagate annotations to connected metabolites.

Glucuronidome shifts with GUSome manipulation in mice

Gut microbial GUS enzymes are well documented to alter the flux and excretion of glucuronide-conjugated metabolites, including pharmaceuticals^12,39–42, endogenous compounds^17,43,44 and dietary xenobiotic chemicals^7,45,46. However, the GUSome effects on the larger glucuronidome are uncharacterized. To investigate the broader impact of the GUSome on glucuronide distribution, we applied MassQL pattern filtering to map changes in colonic content, serum, and urine glucuronidated features in germ-free (no GUSome) and fecal microbiota transplant (FMT) colonized mice (colonized with feces from conventional mice) that seeded a robust microbial GUSome in 7 days (experimental design Supplementary Fig. 2B).

FMT mice exhibited distinctly shifted colon (Fig. 3A), serum (Fig. 3B), and urine (Fig. 3C) glucuronidomes compared to uncolonized, germ-free mice, which included differences in both the richness (Fig. 3D–F) and distribution (Fig. 3G-I) of glucuronidated features. The greatest effect of introducing a diverse GUSome was observed for the colon glucuronidome, which exhibited reduced glucuronidome richness and diversity. Detected glucuronidated features dropped from 311 in germ-free colonic contents to 47 in the colonic contents of colonized mice (Fig. 3D). After colonization, the abundance of individual glucuronidated features was significantly reduced for 288 (90.9%) of the 317 total colon glucuronidated features (Fig. 3G and Supplementary Data 2). The reduction in the number and abundance of glucuronidated features supports the hypothesis of increased glucuronide deconjugation and potential metabolite reactivation, as reflected by increased microbial GUS activity.

Microbial colonization and the presence of GUS genes had opposite effects on the serum and urine glucuronidomes, increasing the diversity and abundance of detected microbial, dietary, and host glucuronidated features compared with germ-free controls. Glucuronides of microbially-produced compounds, including indoxyl-, phenyl-, and p-cresol-glucuronides, were more abundant in the serum and urine of FMT colonized mice compared to germ-free controls (Fig. 3H, I). Additionally, glucuronidated microbial metabolites from the soy-derived dietary compounds daidzein and genistein (e.g., equol glucuronides and dihydrogenistein glucuronide, respectively) were more abundant with FMT (Fig. 3H, I). Furthermore, with FMT, lower levels of glucuronidated, host-produced, β-muricholic acid were detected in the colon (Fig. 3G) and higher levels in the serum (Fig. 3H), suggesting increased GUS reactivation in the colon and subsequent recirculation and conjugation of β-muricholic acid.

With 44 glucuronidated features (Supplementary Data 3), the serum glucuronidome was the least complex of the three matrices. The serum glucuronidome of FMT mice was more robust, with 36 glucuronidated features compared to germ-free serum with 17 glucuronidated features (Fig. 3E). Of the 30 significantly shifted glucuronidated features, 29 were more abundant, and 1 was less abundant with GUS colonization (Fig. 3H).

The urine glucuronidome was the richest of the three matrices, with 390 glucuronidated features detected via MassQL pattern filtering (Supplementary Data 4). Seven days after vehicle or FMT colonization, the glucuronidome of germ-free control mice was relatively unchanged, compared to baseline, with 296 glucuronidated features. In contrast, the urinary glucuronidome of the colonized mice expanded to 372 glucuronidated features (Fig. 3F), a 22–26% increase compared to baseline and parallel controls. In addition to glucuronidated features detected above threshold only after FMT, 37 other glucuronidated features were more abundant after colonization, compared to germ-free controls.

Limited glucuronidome overlap was observed among the colon, serum, and urine, with most glucuronidated features detected in urine. Precursor mass, retention time, and MS/MS matching showed that the most remarkable glucuronidome overlap was of 89 glucuronidated features between the colon contents and urine of the GF/FMT mice. Of which, 17 glucuronidated features were common between the colon, serum, and urine glucuronidomes (Supplementary Fig. 3A).

The contrast between germ-free and colonized glucuronidomes provides vital information about the role of microbes in glucuronidated metabolite distribution, but is an imperfect comparison for translation to conventional mice or human populations. As a more translatable model, we investigated the colon, serum and urine glucuronidome effects induced by more moderate GUSome differences in conventional mice, achieved with low dose oral administration of two nonabsorbable antibiotics, gentamicin (35 mg/L) and vancomycin (45 mg/L) in the drinking water (experimental design Supplementary Fig. 2C). Vancomycin is well established to quickly change the microbiome, depleting Firmicutes and Bacteroidetes and enriching Proteobacteria^47–49. Gentamicin creates a different shift in microbial composition, tending to reduce the relative abundance of Proteobacteria and increase Bacteroidetes. Overall, both antibiotics are poorly absorbed in the GIT⁵⁰ and significantly alter, and even deplete, the gut microbiome upon oral administration. Given that we used antibiotics to shift the microbiome and GUSome, not deplete it, we selected lower doses and a shorter duration than commonly used in microbiome depletion studies⁵¹.

Compared to the effects observed between germ-free and FMT conditions, low-dose oral antibiotics produced more conservative shifts in the colon, serum, and urine glucuronidomes compared to untreated (unshifted) samples (Fig. 4A–C). Similar to the germ-free and colonized mice, the urine glucuronidome was the most complex with 334 glucuronidated features identified, followed by the colon contents with 73, and lastly, the serum glucuronidome was the least complex with 46 glucuronidated features (Supplementary Data S5–7). The urine and serum glucuronidomes from both experiments had comparable numbers of glucuronidated features. The most significant difference was observed between the colon glucuronidomes, with over 200 more glucuronidated chemicals detected in the colonic contents of germ-free mice compared to the more physiologically “normal” colonic glucuronidomes of the conventional and antibiotic-treated mice.

While there were similar numbers of urinary glucuronidated features between the control, gentamicin and vancomycin treated mice (Fig. 4A), distinct glucuronidome shifts were observed compared to control (unshifted) samples (Fig. 4D). The glucuronidome shifts without significant differences in the number of glucuronidated metabolites is attributed to the observation that many features were not differentially abundant compared to control but when there was differential abundance, glucuronidated features tended towards less abundance in urine following antibiotic treatment (Fig. 4D, E). Gentamicin produced a more conservative glucuronidome shift, with only 22 glucuronidated features with altered abundances (Fig. 4E), compared to the 110 altered glucuronidated chemicals with vancomycin treatment (Fig. 4F).

Despite the relatively low number of colonic glucuronidated features detected, antibiotic manipulation of the GUSome shifted the colon glucuronidomes (Fig. 4G). There were fewer glucuronidated features detected in the colon after gentamicin treatment (Fig. 4B, H), compared to controls. In contrast, the overall richness of the colon glucuronidome was relatively unchanged after 5 days of vancomycin treatment (Fig. 4B), despite the greater number of significantly altered glucuronidated features: 10 less abundant and 7 more abundant (Fig. 4I).

The serum glucuronidome showed a different shift in richness in response to treatment, with reduced richness observed only after vancomycin treatment (Fig. 4C). Despite this matrix difference, similar patterns were observed in the glucuronidome beta-diversity and response at the metabolite level between the colon and serum (Fig. 4J–L).

Less overlap was observed between the colon, serum, and urine glucuronidomes in the antibiotic-shifted mice than in the germ-free and colonized mice (Supplementary Fig. 3B). In conventional mice, the majority of glucuronidated features were observed in the urine glucuronidome, which had the highest degree of overlap with the serum glucuronidome; by contrast, in the GF/colonized mice, the most overlap was observed between the colon and urine glucuronidomes.

Molecular networking was employed to putatively annotate the glucuronidated features across the mouse glucuronidomes. To increase the chances of molecular similarity connections between glucuronidated features and aglycones, the colon contents, serum, and urine.mzML files from the antibiotic and colonization experiments were processed together through the classical molecular network workflow in GNPS2. The resultant networks contained 17,989 and 22,744 nodes for negative and positive ESI data, respectively. 405 of the negative ESI merged MS/MS spectra had the neutral loss and fragmentation fingerprint of glucuronidation, while 453 merged MS/MS spectra from the positive network contained the neutral loss indicative of glucuronidation. The classical molecular networking workflow calculates all the edges between nodes, then prunes the edges that satisfy the set parameters. In this case, edges with a modified cosine greater than 0.4 that were within the top 10 connections for the node were included, allowing network components of up to 100 nodes to be created. Since our analysis is focused on glucuronide:aglycone connections, pruned edges with Δm/z (±0.003 Da) = 176.032, 194.042, and 96.075 were re-added to the exported network in Cytoscape⁵². The re-addition of these edges maintained the network’s rigor while maximizing the ability to visualize glucuronide-aglycone pairs. For example, in positive mode, this step increased the number of Δm/z = 176.03 edges from 59 to 692, thus allowing for putative annotation of more glucuronidated features.

Molecular networking can be applied separately to positive and negative ESI MS/MS data. Bilirubin and biliverdin glucuronides were independently annotated through molecular networking of both negative (Fig. 5A) and positive (Fig. 5B) ESI data, branching off of bilirubin and biliverdin, respectively, with a Δm/z = 176.032, attributed to the glucuronide neutral loss. The MS/MS mirror plots illustrate the MS/MS similarities (green in Fig. 5C) that are used by the networking algorithms to form the edge connection between related nodes and allow for the putative annotation of nodes connected to library-matched nodes (Fig. 5C).

Fig. 5 — A Classical molecular networking subcluster of negative ESI heme-metabolite features. The structures are bilirubin glucuronide (left) and biliverdin glucuronide (right), with the glucuronic acid highlighted in yellow. B Classical molecular networking subcluster of positive ESI heme-metabolite features. C Classical molecular networking subcluster of soy-related flavonoid metabolites. The MassQL results for the glucuronidation pattern are shown as diamond-shaped nodes. Putative annotations were achieved by MS/MS comparison to the connected non-glucuronide features for shared fragmentation. Two examples of aglycone:glucuronide MS/MS comparisons are included for biochanin A and formononetin. Star (*) indicates the putative annotation was later confirmed with standards. C metabolite annotations: a - daidzein 7-β-D-glucuronide; b - daidzein; c - genistein; d - genistein sulfate; e - biochanin A sulfate; f - biochanin A; g - biochanin A glucuronide; h - daidzein 4’-β-D-glucuronide; i - genistein 7-β-D-glucuronide; j - hispidulin glucuronide (aka 6-O-methyl scutellarin); k - formononetin sulfate; m- formononetin; n - formononetin glucuronide.

Microbial GUS deconjugation of xenobiotic glucuronidated metabolites

The FMT mice were successfully colonized with a diverse GUSome containing all major GUS loop classes except for mini-loop1, 2 GUS, which is one of the least abundant orthologs¹⁶ (Supplementary Fig. 4A). The disparate conditions of a diverse GUSome after FMT and no GUSome in the GF mice produce a distinctive glucuronide and aglycone pattern between the colonic contents and urine. The aglycone metabolite abundance increases and the glucuronidated metabolite abundance decreases in the colonic contents after treatment with FMT, while an increase in the glucuronidated metabolite abundance is observed in the urine—exemplified by dihydrogenistein glucuronide and dihydrogenistein (Fig. 6A). From the GF to the FMT condition, the glucuronidome changes were relatively straightforward to follow. The introduction of gut microbial GUS allowed for the deconjugation of colonic dihydrogenistein glucuronide, regenerating colonic dihydrogenistein, and the paired increase in dihydrogenistein glucuronide abundance in the urine. By contrast, treatment with gentamicin and vancomycin antibiotics created more complex glucuronidated and aglycone metabolite patterns between the colonic contents and urine. Considering dihydrogenistein and dihydrogenistein glucuronide in the ABX mice, gentamicin treatment caused a paired increase in colonic dihydrogenistein (F(2) = 107.4, log₂FC = 3.126, p = 6.54e-5) and urine dihydrogenistein glucuronide (F(2) = 176.5, log₂FC = 2.44, p = 8.74e-6) however, the opposite was observed in the vancomycin treated mice which experienced a decrease in colonic dihydrogenistein (F(2) = 107.4, log₂FC = −4.019, p = 5.45e-6) and urine dihydrogenistein glucuronide (F(2) = 176.5, log₂FC = −3.32, p = 3.40e-7) (Supplementary Fig. 4B). The differential metabolite flux between gentamicin and vancomycin treatment can be attributed in part to differential abundance in the C-terminal domain (CTD) Schaedlerella_MGG38568_4 GUS gene. CTD GUS proteins have been shown to provide a C-terminal domain that stabilizes small molecule glucuronide substrates in the active sites of these enzymes¹⁷. The abundance of the Schaedlerella_MGG38568_4 CTD GUS gene is positively associated with the abundance of both colonic dihydrogenistein (Fig. 6B, t(23.77) = 3.418, Spearman’s ρ = 0.6837, p = 0.0046) and urine dihydrogenistein glucuronide (Fig. 6C, t(23.77) = 3.492,Spearman’s ρ = 0.6656, p = 0.0040). There are 20 differentially abundant GUS genes (13 less abundant and 7 more abundant) with gentamicin treatment (Fig. 6D), and 24 differentially abundant GUS genes (16 less abundant and 8 more abundant) with vancomycin treatment (compared to control, Fig. 6E). Of the 15 total GUS genes that were increased after antibiotic treatment, 9 are CTD (3 with gentamicin, 6 with vancomycin) and 2 are L1 (1 each with gentamicin and vancomycin). Thus, the majority of GUS genes differentially increased after antibiotics are CTD and L1 GUS, which are structural classifications documented to deconjugate small metabolite glucuronides^9,17,53. On the whole, the GUSome decreased with antibiotics (Supplementary Fig. 4C), but the increased abundance of the aforementioned GUS genes and the metabolite associations indicate that future work should consider both the larger GUSome and GUS classes as well as the individual GUS genes so as not to miss potentially impactful genes in larger GUSome changes.

Fig. 6 — A Log2 transformed peak areas (abundance) of colonic dihydrogenistein (aglycone) and dihydrogenistein glucuronide, as well as urine dihydrogenistein glucuronide in germ-free (GF) mice and mice 7 days post fecal microbiota transfer (FMT). *Hypothesis testing was performed by Student’s T test*. Spearman’s correlation of *Schaedlerella*_MGG38568_4_CTD GUS gene abundance and B colonic aglycone dihydrogenistein peak areas, fit line (y = 2.7451x + 22.1757) or C urine dihydrogenistein glucuronide peak areas, with fit line (y = 2.2304x + 25.5447). *Regression line and p-value generated using lm in R*. D, E Volcano plot of Log fold change (FC) by -Log(p-value) for GUS genes in gentamicin (gent) versus untreated control and vancomycin (vanc) versus control mice, respectively. The numbers in the lower corners D, E indicate the number of differentially abundant genes. The point shapes indicate the loop class of each gene, and color represents the statistical significance: blue = less abundant in treatment, gray = no statistical difference, and red = more abundant in treatment. *Hypothesis testing was performed with linear mixed effects models in R, using lmer for the main model and emmeans for pairwise comparisons*. Lowercase letters in (D&E) indicate CTD and loop 1 GUS genes; gene IDs are in Supplementary Data 8. F Proportional peak area of naringenin (Nar) to naringenin-7-O-β-D-glucuronide (Nar-7-GlcA) after incubating purified GUS enzymes with Nar-7-GlcA (grouped by loop classification: Fp-*Faecalibacterium prausnitzii*, Rg3-*Ruminococcus gnavus* 3, Rh2-*Roseburia hominis* 2, CP-*Clostridium perfringens*, Ec-*Escherichia coli*, Ee-*Eubacterium eligens*, Lr-*Lactobacillus rhamnosus*, Sa-*Streptococcus agalactiae*, AF25-*Bacteroides* sp. AF25-17LB, Bf-*Bacteroides fragilis* NCTC 9343, Bm-*Bacteroides massiliensis*, Rh1-*Roseburia hominis* 1, Bo-*Bacteroides ovatus*, Am-*Akkermansia muciniphila*, Bu3-*Bacteroides uniformis* 3, Pm-*Parabacteroides merdae*, Bd-*Bacteroides dorei*, Et-*Eisenbergiella tayi*, P-msp-*Parabacteroides*_multispecies, Bu1-*Bacteroides uniformis* 1, CTD-C-terminal domain, L1-Loop 1, L2-Loop 2, mL1-mini-Loop 1, mL1,2-mini-Loop 1,2, mL2-mini-Loop 2, NL-No loop and NTL-N-terminal domain. G Proportional peak area of Nar to Nar-7-GlcA after incubating increasing concentrations of purified *Schaederella*_MGG38568_4_CTD GUS with Nar-7-GlcA. HK is a heat-killed control. All statistical tests are two-sided, as applicable.

Dihydrogenistein glucuronide is not available commercially, so an available isomeric surrogate (naringenin-7-O-β-D-glucuronide), differing in the phenol attachment position on the aglycone (Supplementary Fig. 4D, E), was selected to confirm GUS deconjugation of the glucuronide by screening for aglycone generation with a panel of 24 representative purified GUS enzymes of all 8 loop classes. These enzymes were selected from human fecal microbiome studies to create a panel that samples the functional GUS categories characterized to date. The greatest deconjugation activity was observed with L1 GUS proteins, followed by N-terminal loop (NTL), mL1, CTD, and mL2 GUS enzymes (Fig. 6F). Schaedlerella_MGG38568_4 GUS also showed deconjugation activity with naringenin-7-O-β-glucuronide (Fig. 6G), although at levels lower than those observed in the GUS panel (none of which were detected in the samples collected from mice in this study). As noted in Fig. 6D, E, the genes for several CTD and L1 GUS are increased after antibiotics, including two additional taxa from Schaedlerella_MGG38568. Thus, in these mice, it is reasonable to conclude that CTD GUS, such as those from Schaedlerella taxa, would be capable of processing dihydrogenistein-glucuronides and related compounds in the colon. Wild-type K12 Escherichia coli (E. coli), which naturally expresses the L1 EcGUS used in the enzymatic assays, was selected to confirm both deconjugation activity in vitro and loss of deconjugation activity with uidA (GUS) knockout (Supplementary Fig. 4F). A greater proportion of naringenin to naringenin-7-glucuronide was detected in the E. coli samples compared to native degradation in the sterile controls, supporting the results from the purified enzyme assays.

Glucuronidome diversifies with microbiota colonization in infants

We next evaluated how this approach performed in a human setting by following the glucuronidome and GUSome of infants at 1, 4, and 12 M of age from the Infant Growth and Microbiome (IGRAM) Study⁵⁴. This design lets us test whether age, diet, and exposure history align with shifts in urinary glucuronidated metabolites and GUS gene content. Glucuronidated metabolite profiling of urine showed clear age-related structuring in the PCoA (Fig. 7A, all age comparisons p = 0.001 by PERMANOVA, detailed test values are in Supplementary Data 10) (Fig. 7A), and infants consuming other milk at later ages further separated, particularly at 12 M (formula vs. other milk PERMANOVA, statistic(1) = 2.63, R² = 0.07, p = 0.021, Supplementary Data 10). The number of glucuronidated features increased across infancy (Fig. 7B). At 1 M, breastmilk-fed infants had fewer glucuronidated features than formula-fed or mixed-fed infants, but this difference decreased as all groups showed steady age-related increases. The number of unique GUS genes rose sharply by 12 months (Fig. 7C), which we attribute to reduced breast milk consumption and related dietary diversification. It is worth noting that by 12 months, few individuals were receiving only breastmilk (n = 2) or mixed breast milk and formula/other milk (n = 2). However, GUS structural classes showed mixed patterns, with some domains and loop types increasing with age, while others remained stable (Fig. 7D). Globally, glucuronidome diversity increased in parallel to both microbial (data previously reported^54,55) and GUSome diversity, which is consistent with the outcomes observed in the mouse studies. Diet effects were limited to the glucuronidome, primarily early in life and appearing related to breast milk consumption (PERMANOVA, 1 M formula vs. breast milk statistic(1) = 9.26, R² = 0.26, p = 0.001 and mixed (mix of breast milk and formula) vs. breast milk statistic(1) = 4, R² = 0.16, p = 0.006, Supplementary Data 10). The pattern may reflect the relatively simple microbiome and exposures of the younger infants, with decreased diet effects observed over time, potentially due to confounding effects of increased exposures both within and beyond diet. Diet derived R,S-equol 7-O-glucuronide was detected more often, at all ages, in babies consuming any formula with even greater detection at 12 M after the diets diversified to include milks other than breast and formula (Fig. 7E). Reported acetaminophen (Tylenol) exposure corresponded to higher urinary acetaminophen glucuronide levels (Fig. 7F). These diet and drug related patterns align with reported diet and drug use, and support that the mass spectrometry-based glucuronidome captures biologically meaningful exposure signals. R,S equol 7-O-glucuronide showed a directionally consistent correlation with loop 1 GUS gene intensity (Fig. 7G), which fits with expectations that this domain contributes to glucuronide processing even if other factors also influence metabolite levels. Taken together, these results suggest that developmental changes dominate the shaping of these patterns while diet and other exposures add smaller context-dependent shifts.

Fig. 7 — A Principal coordinates analysis (PCoA) of the urinary glucuronidome from infants sampled at 1, 4, and 12 M, separated by feeding type, breast milk data are included in all panels. B Number of urinary glucuronidated features detected across age and feeding categories. C Number of unique GUS genes detected in infant samples, which rises sharply by 12 M. D Normalized abundances of GUS gene structural classes, over time and by feeding type. B–D Statistical testing was performed by linear mixed effects model in R, and post-hoc tests for pairwise differences were performed with the emmeans package or Benjamini–Hochberg false discovery rate correction, as appropriate. E Peak areas of R,S-equol 7-O-glucuronide across age and feeding categories. F Acetaminophen glucuronide peak areas in infants with and without reported acetaminophen intake. G Spearman’s correlation between Loop 1 GUS gene intensity and R,S-equol 7-O-glucuronide peak area at 12 M with linear regression fit line (y = 86771.85432x + 268967.7599). *Regression line and p-value generated using lm in R*. A–E, G Feeding type sample sizes in time order (1, 4, and 12 M): breast = 9, 3, 2; formula = 20, 32, 22; mixed = 14, 8, 2; other = 0, 0, 17; any formula = 34, 40, 24. Sample sizes for E: no APAP = 43, 36, 29; yes APAP = 0, 7, 14. All statistical tests are two-sided, as applicable.

Phenotypic associations of GlcAs

The Mass Spectrometry Search Tool (MASST)⁵⁶ was used to search for the glucuronidated standard MS/MS across the three major public metabolomics repositories Metabolights, Metabolomics Workbench, and GNPS2/MassIVE. The spectral matches from MASST were filtered to include matches with a minimum cosine similarity score of 0.7 to the glucuronidated standard spectra, considering raw (unfiltered) spectra. The controlled vocabulary metadata available through the paired Reanalysis of Data User (ReDU) interface and most recent Pan-ReDU implementation^57,58 allowed us to assess the distribution of the glucuronidated standards across tissues and biofluids in publicly available, metadata-harmonized, rodent and human data. We first looked at rodent samples, for which the most spectra matching our glucuronidated standards were returned in positive ESI samples from the lower and upper digestive tract (Supplementary Fig. 5A), followed by blood and liver samples in negative ESI (Supplementary Fig. 5B). In both positive and negative ESI rodent samples, feces was not a rich sample type for detecting the glucuronidated metabolites from our standards library, which is not unexpected due to suspected native fecal GUS activity and limited aglycone exposures in rodents under research conditions.

However, there are more human samples in the MASST and Pan-ReDU interfaces than rodent samples, and, as expected, we observed more matches to glucuronidated standards across human biosamples. In positive ESI, the greatest proportion of matches (normalized to sample type and frequency of detection abundance relative to the availability in ReDU) was observed for urine samples (Supplementary Fig. 5C), supporting urine as an important matrix for investigating the glucuronidome, which is consistent with glucuronidation being a conjugation reaction that facilitates excretion. More glucuronidated metabolites were detected in the human samples compared to the rodents. This is likely due to two factors. First, there were no urine samples available in Pan-ReDU for rodents and second, the human metabolome is more complex than the rodent metabolome due to diverse diets and microbiome communities. The matches from rodent samples are almost exclusively dietary glucuronides, whereas the human samples showed matches for not only dietary glucuronides but also for medications, such as acetaminophen and ibuprofen glucuronides, and other exposures, including phthalate and nicotine glucuronides. This data suggests that the human glucuronidome is more diverse than the mouse glucuronidome, which would be expected as humans have greater xenobiotic exposures (dietary, environmental, and personal products, etc) compared to the relatively limited exposome of a laboratory animal in a controlled environment.

Furthermore, the Pan-ReDU metadata allowed us to further map MS/MS matches to the glucuronidated standards library across studies with disease ontology information. There were 15 pathologies with matches to the glucuronidated standards. Of these, matches were concentrated in datasets with diabetes samples, with other notable conditions including obesity, osteoarthritis, Alzheimer’s disease, rheumatoid arthritis and, less notably, inflammatory bowel disease, which is spread across three disease ontologies: inflammatory bowel disease, Crohn’s disease and ulcerative colitis (Supplementary Fig. 6A). Furthermore, many of which have known or suspected associations with GUS^11,59–62. However, the presence of these compounds in these particular datasets does not necessarily mean they are correlated with the disease itself, and further investigation of the individual datasets is required. For example, the diabetes mellitus samples are predominantly urine samples, which we established as one of the richer matrices for detecting glucuronidated features, potentially skewing the returns to appear to favor glucuronidome differences in diabetes. Of particular interest to us were the diabetes mellitus and obesity samples from MSV000084112; however, these samples did not have paired microbiome data for a deeper investigation.

To validate our findings, we sought to replicate obesity conditions in a controlled animal model with paired gut microbial metagenomic sequencing and urine metabolomics. To investigate the GUSome and glucuronidome effects of a sustained high fat diet and obesity, we analyzed paired metabolomic samples and metagenomic sequencing from a 18-week study where mice were fed purified high fat or normal fat diets⁶³ (experimental design Supplementary Fig. 2D). After 18-weeks of high fat diet, 5 GUS genes (of 26 detected—Supplementary Data 9) were significantly different from the normal fat diet control group; of these, two were classified as CTD and two as Loop 1, important loop types for small molecular glucuronide deconjugation. In the mouse study, a high-fat diet and obesity were statistically associated with a CTD GUS gene (Supplementary Fig. 6B–D) and glucuronide:aglycone pair m/z 396.1684:220.1366 with a MS/MS cosine similarity score = 0.8734 (Supplementary Fig. 7A). Both the fecal aglycone (Supplementary Fig. 6C) and urine glucuronidated feature were positively associated with the unknown_unclassified_CTD GUS gene (Supplementary Fig. 6D), suggesting this GUS plays a role in the deconjugation and altered fecal:urinary excretion of the aglycone:glucuronide pair. Moreover, this glucuronidated feature was detected in the human obesity/diabetes mellitus dataset (MSV000084112) and was more abundant in obesity than in non-disease controls (Supplementary Fig. 6E and Supplementary Fig. 7B MS/MS cosine similarity score between mouse and human metabolite = 0.9892). Taken together, these data suggest there is a reciprocal relationship between the GUSome and glucuronidome in mice, and this relationship likely translates to humans.

Discussion

Annotation of untargeted tandem mass spectrometry metabolomics data depends on metabolite coverage in reference spectral libraries, which is limited by the availability of known compounds and chemical analogs, including glucuronidated chemicals. We introduced an integrative data filtering and molecular networking approach to map glucuronidated features, compiling a glucuronidated metabolite reference resource. This method enables systematic detection of dietary, endogenous, and xenobiotic metabolites from untargeted MS/MS data and complements existing in silico²⁶ and enzymatic approaches for identifying glucuronidated biomarkers⁶⁴. The approach also allows rapid identification of glucuronidated features without specialized sample preparation or extensive post-processing. In the datasets generated for this study, glucuronidated features represent a substantial portion of the chemical space that typically remains unannotated. These metabolites can provide a window into host and microbial metabolism^65–69, but their underrepresentation in public MS/MS libraries has limited their integration into biological studies. The resource and pattern-filtering workflow developed here offers a practical framework for incorporating these conjugated metabolites into broader analyses of host–microbiome interactions.

Application of this approach to gnotobiotic and antibiotic-treated mice shows that gut microbial β-glucuronidase activity strongly shapes the fate of glucuronidated metabolites. Microbial colonization of germ-free mice results in a redistribution of the glucuronidome from colonic contents to systemic circulation and urine, which is consistent with enhanced microbial β-glucuronidase activity, reabsorption of aglycones, and subsequent urinary excretion of reconjugated metabolites. These results are consistent, although on a much larger scale and representing a greater diversity of compartments, with previous efforts to understand the impact of GUS inhibitors¹⁷. Antibiotic treatment produces the opposite pattern with reduced urinary glucuronides, indicating altered downstream metabolism of liberated compounds. Metagenomic profiling reveals shifts in CTD and loop 1 GUS subclasses that favor small molecule glucuronides, demonstrating how microbial community composition and enzyme class together regulate enterohepatic cycling. These findings also have practical implications for study design. Urine provides broad coverage of glucuronidated metabolites when GUS activity is intact, whereas colonic and fecal matrices become more informative when microbial β-glucuronidase activity is reduced.

The IGRAM infant cohort⁵⁴ extends these mechanisms to early human life and is, to our knowledge, the first longitudinal evidence that the urinary glucuronidome richness expands in step with microbiome diversification^54,55 and fecal GUS gene abundance over the first year of life, supporting the relevance of these processes in humans. This pattern likely reflects both microbial maturation and increasing exposure to dietary, pharmaceutical, and environmental compounds⁷⁰. The transition from milk to a more varied diet introduces plant-derived polyphenols and other dietary xenobiotics. Early-life exposure to medications, such as antipyretics, expands the range of substrates entering glucuronidation and β-glucuronidase pathways. Environmental exposures may add further complexity through additional conjugation substrates. These observations resonate with prior work showing large interindividual differences in microbial drug metabolism, and suggest that age-linked changes in GUS capacity may contribute to variability in exposure to bioactive aglycones and their glucuronidated forms during infancy^71,72. Together, these changes provide a physiological and ecological basis for the expanding urinary glucuronidome observed during infancy.

Searches across public metabolomics repositories highlight how these dynamics are shaped by tissue distribution. In rodents, glucuronidated metabolites are most enriched in blood and digestive tract tissues (Supplementary Fig. 5A, B). In humans, enrichment is highest in urine, with lower levels in feces, saliva, oral cavity, milk, cerebrospinal fluid, brain, and blood (Supplementary Fig. 5C). This pattern suggests that glucuronidation and β-glucuronidase activity ripple across multiple compartments and reflect both local and systemic processes that evolve over time and in response to diet, medication use, and environmental exposures.

This work combines computational resource development with biological discovery. The glucuronidation library and pattern filtering workflow provide a scalable and reproducible strategy for uncovering an underexplored chemical space. The mouse and human data demonstrate the metabolic significance of the GUSome and how microbial deconjugation can redirect in vivo metabolic flux from fecal to urinary excretion. These processes have broad implications for understanding how diet, environmental exposures, and pharmaceuticals interact with the microbiome to influence chemical fates in the body. Microbial GUS activity can shape exposure to dietary xenobiotics, modify host responses to bioactive compounds, and alter the pharmacokinetics of drugs^{2,6,7,11–14}. Remaining challenges include integrating sulfation and other phase II conjugation pathways, resolving isomeric diversity with higher accuracy, and dissecting the overlapping functions of GUS subclasses. Expanding spectral libraries and linking them with defined microbiome manipulations in controlled models will be essential for addressing these questions.

In summary, this study establishes a scalable approach to explore glucuronidated metabolites and reveals how microbial GUS activity determines their distribution across tissues, developmental stages, and species. These findings provide both methodological and biological insight into host–microbiome metabolic interactions and offer a framework for future studies on nutrition, environmental exposures, and drug metabolism.

Materials and methods

Chemicals and reagents

LC-MS grade methanol (A456-4, Fisher Scientific, Fair Lawn, NJ), acetonitrile (BJLC015-4, West Chester, PA), water (W6-4, Fisher Scientific, Fair Lawn, NJ). Formic acid (A117-50, Fisher Scientific, Fair Lawn, NJ). Chlorpropamide (SC-234350, Santa Cruz Biotechnology, Dallas, TX) was used as an internal standard for untargeted LC-MS/MS. Phytochemical Metabolite Library of standards (PHYTOMLS, Sigma-Aldrich, St. Louis, MO) and Microbiome Metabolite Library of standards (GUTMLS, Sigma-Aldrich, St. Louis, MO). See Methods Supplementary Data 12 for the list of 134 glucuronidated standards and chemical information.

Spectral library generation

Glucuronidated chemical standards were individually reconstituted in the preferred solvent(s) from the manufacturer’s documentation. Individual chemicals were combined into pools of up to 20 chemicals and brought to 5 μM in 50% methanol with 0.1% formic acid and 1 μM chlorpropamide for LC-MS/MS analysis. IROA library plates (96-well) were resuspended as suggested by IROA. Briefly, plates were centrifuged at 3000 rpm for 3 min, then each well was resuspended in 100 μL of the suggested solvent(s) (water, methanol and/or ethanol). Row-wise pools were made by combining 10 μL from each well across a row—pooling up to 12 compounds for LC-MS/MS analysis.

Pooled standards were injected at 2 μL or 5 μL and fragmented in positive and negative mode with stepped normalized collision energies (NCE) at 20, 35, and 50. For more details, see “Standards and mouse LC-MS/MS data acquisition” and Methods Supplementary Data 13.

GNPS2 spectral libraries were generated from the.mzML files using the MSMS chooser workflow⁷³ in GNPS2 and are publicly available for analysis integration or download on GNPS2.

Classical molecular networking (chemical standards)

The classical molecular networking workflow^24,74 in GNPS2 was used to create molecular networks from the library MGF files. In brief, the precursor and fragment mass tolerances were set to 0.002 and 0.005, respectively, and clustering was turned off (“0”). The minimum cosine value for networking was 0.25, and a minimum of three matched fragments was required for networking. The maximum node component size was 25, and only the top 5 node pairs were considered. The same library mgf files were used for library matching with a minimum cosine of 0.7 and at least 4 matched fragment peaks. Only the top library match was reported. The generated network was exported to Cytoscape for visualization and overlay of the MassQL query for neutral losses at 176.03 Da or 194.04 Da. The network can be found on GNPS2 with https://gnps2.org/status?task=bf9e78402ce04ef39980605e94869fb6 and the massQL with https://gnps2.org/status?task=df017bedaaef47ee96c0b4f1e1c31a73.

Mass query language (MassQL) query development

The MassQL queries for glucuronidation pattern filtering were developed and tested against the glucuronidated standards library. The queries were developed to target the characteristic glucuronide neutral loss (m/z 176.0321) and expanded to include the benzylic glucuronide neutral loss of m/z 194.0425. To increase query specificity, the two most common glucuronide fragments, m/z 113.0244 and 85.0295, were added to the negative ESI query (Fig. 1B and Supplementary Fig. 1A).

Negative ESI query: QUERY scaninfo(MS2DATA) WHERE MS2NL = (176.0321 OR 194.0425):TOLERANCEMZ = 0.003 AND MS2PROD = 113.0244:TOLERANCEMZ = 0.003 AND MS2PROD = 85.0295:TOLERANCEMZ = 0.003

Positive ESI query: QUERY scaninfo(MS2DATA) WHERE MS2NL = (176.0321 OR 194.0425):TOLERANCEMZ = 0.003

False discovery analysis

To assess for false discovery in the MassQL pattern filtering for glucuronidated features, quality checks were included at two points in the analysis pipeline. The first quality assessment analysis was performed during initial query development, where the performance of potential glucuronidated pattern queries was tested against our glucuronidated, gut microbial and phytochemical libraries. This step assured high fidelity of the pattern filtering to glucuronidated features with minimal crossmatches to other glycosylations and common substructures. Secondary quality assessment was performed during the analysis of biological samples. The MS/MS returned as glucuronidated from the mouse untargeted data was visually inspected to confirm MS/MS pattern conformation with the pattern expected of glucuronidation.

Animals

All animals were housed in the Pennsylvania State University Animal Care Facilities and received ad libitum food and water, unless otherwise specified. All mouse handling protocols and procedures were approved by The Pennsylvania State University Institutional Animal Care and Use Committee (IACUC) under protocols PROTO202302579 (APAP study), PROTO202101826 (ABX and FMT studies) and PROTO202001416 (HFD study) and conducted in accordance with state and federal regulations. Mice were maintained on the standard 12 h light/dark cycle.

APAP study

Ten-week-old, C57BL/6J male mice were fed pelleted AIN93G Purified Rodent Diet (Dyets Inc, Bethlehem, PA) for 7 days and received an intraperitoneal injection of vehicle (1× PBS, pH = 7.4) or APAP (300 mg/kg), after 12 h of fasting (n = 8/group). Urine samples were collected 24 h after APAP administration.

ABX study

Male, C57BL/6J mice (Strain:000664, The Jackson Laboratory) were singly housed with a 12 h light schedule (7 am–7 pm) and fed pelleted 5053 mouse chow (LabDiet, Gray Summit, MO). At 12 weeks of age, animals were treated with autoclaved water (vehicle), water with 45 mg/L vancomycin hydrochloride (0990-1G, VWR) or water with 35 mg/L gentamicin sulfate (102987-214, VWR), refreshed daily for 5 days (n = 5/group). Urine and fecal samples were collected daily, including day 0 (baseline sample prior to antibiotic administration), and stored at −80 °C until processing.

Germ-free and FMT study

Twelve to thirteen-week-old germ-free mice (C57BL/6J background) were randomly divided into 2 gnotobiotic isolators (n = 5/group) and individually housed. All mice received ad libitum food (LabDiet 5021, Gray Summit, MO) and water for the duration of the study. FMT were prepared by isolating the gut microbial community from fresh fecal samples collected from conventional C57BL/6J, male mice. Under anaerobic conditions, 2 mL (10% weight/volume) of 0.2 μM syringe filtered BHI-CHV media w/ 20% glycerol (vehicle) was added to 200 mg of conventional mouse feces, vortexed until mixed. 100 μL of vehicle or FMT was administered to each mouse via oral gavage. Vehicle and FMT-treated mice were maintained in separate isolators for 7 days. Fecal and urine samples were collected prior to FMT and 7 days after administration and stored at −80 °C until analysis.

High-fat diet study

Mice were treated as previously described⁶³. In brief, 4-week-old, male, C57BL/6J, wild-type mice (Jackson Laboratories, Bar Harbor, MN), were fed either F4031 normal fat diet (NFD, control) or F3282 high fat diet (HFD) (Bio-Serv, Flemington, NJ) continuously for 16-weeks. Urine and feces were collected before sacrifice at 21 weeks old and stored at −80 °C until processing.

All mice were euthanized by CO2 asphyxiation promptly followed by cardiac puncture for blood collection. The blood was collected in BD Microtainer® SST tubes (365967, BD) and placed on ice for serum separation. Colonic contents were collected by excising the colon and scraping the contents out of the ends of the colon. The colon contents were snap frozen in liquid nitrogen and stored at −80 °C until analysis. The blood was centrifuged at 5000 × g and 4 °C for 5 min for serum separation. The top layer (serum) was collected and stored at −80 °C.

Metabolite extraction and sample preparation

Urine samples were diluted 1:1 with cold methanol with 0.1% formic acid and 2 μM chlorpropamide and incubated for 20 min on ice. Next, samples were centrifuged at 18,213 × g for 15 min at 4 °C. The supernatant was transferred to injection vials for analysis.

Serum glucuronides were extracted from 20 μL of serum diluted with 4 volumes of ice cold 80% MeOH with 0.1% formic acid, which was quickly vortexed and incubated on ice for 20 min. After a second quick vortex, samples were spun in a 4 °C centrifuge for 15 min at max speed (18,213 × g). The supernatant was transferred to fresh tubes and dried down under nitrogen in a Labconco RapidVap Vertex Dry Evaporator (89203-118, Avantor, Radnor, PA) at increasing pressures from 20 to 45 psi, without heat. The pellets were resuspended in 50 μL 50% MeOH with 0.05% formic acid and 1 μM chlorpropamide (internal standard) by vortexing for 30 s, sonicating for 10 min at room temperature and vortexing for 30 s. Next samples were incubated on ice for 20 min and centrifuged at 4 °C and max speed for 15 min. The supernatant was collected into injection vials for LC-MS analysis.

Approximately 20 mg of fecal/GI contents were extracted by adding 1 mL of ice cold 80% methanol with 1% formic acid and homogenized (D2400, Benchmark Scientific, Sayreville, NJ) with 1 mm beads at max speed for 3 rounds of 30 s intervals with 5 s intermission. Samples were then centrifuged for 10 min at 14,000 × g and 4 °C. The supernatant was transferred and dried down under nitrogen in a Labconco RapidVap Vertex Dry Evaporator (89203-118, Avantor, Radnor, PA) at pressures ranging from 20-45 psi, without heat. The samples were reconstituted in 100 μL of 50% methanol with 1 μM chlorpropamide. Pooled samples were made by combining equal volumes of each sample by matrix/tissue.

Standards and LC-MS/MS data acquisition

Each sample was analyzed once by LC-MS/MS. Two to five microlitres of extract was analyzed depending on the sample type and standard concentration. Metabolite separation was completed using a Vanquish Horizon Ultra High-Performance Liquid Chromatography (UHPLC) System (ThermoFisher Scientific) with an ACQUITY UPLC BEH C18 Column, 130 Å, 1.7 µm, 2.1 mm × 100 mm (Waters). The column compartment and in-line preheater temperature were 55 °C, while samples were maintained at 5 °C. Mobile phase solvent A was water with 0.1% formic acid, and solvent B was acetonitrile (ACN) with 0.1% formic acid. The mobile phase gradient started at 3% solvent B at time zero, increasing to 45% solvent B by 10 min and 75.00% solvent B at 12 min. The solvent concentration remained at 75% solvent B until 17.50 min, decreasing to 3% solvent B by 18 min and remaining at this concentration until the method ended at 20 min. Alternatively, solvent B was 90% ACN with 0.1% formic acid with a gradient increasing from 3%B to 50%B over the first 10 min, then increasing to 83%B at 14 min and holding until 17.5 min, then decreasing to 3%B at 18 min until the method ended at 20 min. Both LC methods were performed at a singular flow rate of 250 μL/min.

An Orbitrap Exploris 120 (Thermo Scientific) was used to generate all MS/MS data. The default ion source global settings for the flow rate of 250 μL/min were used for all injections. Briefly, MS1 scans were collected at 120 K resolution within a m/z range of 90–1000. MS/MS fragmentation was generated with stepped normalized collision energies (NCE) at 20, 35, 50 and collected at 30 K resolution. The intensity threshold for MS/MS fragmentation was 20,000, and a maximum of 4 ddMSMS scans were collected between MS1 scans. The full list of MS settings are in Methods Supplementary Data 13.

File conversion

All.RAW (Thermo Scientific data format) and.wiff (Sciex data format) MS/MS data files were centroided and converted to.mzML using the vendor Peak Picking algorithm in MSconvertGUI from ProteoWizard⁷⁵ version 3.0.25035.

Glucuronidome analyses

The converted MS/MS files from experimental samples were processed with MS-DIAL (version 5.3.240719) for spectral alignment, peak picking and metabolite identification. A combination of the in-house generated libraries and publicly available MS/MS libraries were used for metabolite identification. The aligned data was transferred into the Global Natural Products Social Molecular Networking (GNPS2) ecosystem for pattern filtering and molecular networking. MS-DIAL peak areas were imported into R⁷⁶ version 4.4.2(2024 October 31) for statistical analysis and graph generation. All graphs were generated with the “ggplot2” package⁷⁷. Bray–Curtis distances were computed with the “vegdist” function from the “vegan” package⁷⁸, and the “pcoa” function from the “ape” package⁷⁹ was used to generate the PCoA vectors. PCoA hypothesis testing was performed by Permutational Multivariate Analysis of Variance (PERMANOVA) (“vegan::adonis2,” with permutations = 9999). 3D PCoA plots were generated with gg3D⁸⁰. Glucuronidated feature counts were generated by summing the number of glucuronidated peak features with areas greater than the noise threshold (250 K colon contents and GF/FMT urine, 50 K for serum and 300 K for ABX urine). Statistical significance was determined by Poisson regression (“glm” function, family = poisson, R stats package⁷⁶). Individual glucuronidated features were log2 transformed before statistical testing with DescTools::DunnetTest⁸¹ or stats::t.test⁷⁶ as appropriate, and p-values were Benjamini and Hochberg FDR corrected for multiple hypothesis testing using the “stats” package, “p.adjust” function.

Acetaminophen metabolite feature-based molecular networking (FBMN)

The MS-DIAL alignment output, MGF and feature table were imported into the GNPS2 for feature-based molecular networking³⁸. The integrated GNPS2 spectral libraries and our generated phytochemical, gut microbial and glucuronidated metabolite libraries were applied in the networking workflow. Precursor and fragment ion tolerances were set to 0.02 Da and the minimum parameters for networking were cosine = 0.25 and 3 matched fragments. Connections were pruned to the top 10 pairs, and the maximum mode cluster size was 100. Library matches were set to cosine > 0.7 with a minimum of 4 shared peaks between the library and experimental spectra. The GNPS2 FBMN job can be found and explored at: https://gnps2.org/status?task=63503ceaa42c4d0797599a0d23ad141c.

Classical molecular networking analysis of urine, colonic contents, and serum

Urine, colonic and serum.mzML files were subjected to the Classical Molecular Networking workflow in GNPS2. The integrated GNPS2 spectral libraries and our generated phytochemical, gut microbial and glucuronidated metabolite libraries were applied in the networking workflow. Precursor and fragment ion tolerances were set to 0.01 Da and the minimum parameters for networking were cosine = 0.4 and 3 matched fragments. MS/MS spectra had to be detected in at minimum 2 files, for inclusion in MS/MS clustering and molecular networking. Connections were pruned to the top 10 pairs, and the maximum mode cluster size was 100. Library matches were set to cosine > 0.7 with a minimum of 4 shared peaks between the library and experimental spectra. The positive ESI networking job can be found at: https://gnps2.org/status?task=661c1b2cfbb944d69dcd66a131eb2127, and the negative ESI network at https://gnps2.org/status?task=f856a02122db4b2e956d24aa746cef4e. The molecular networks with singletons were downloaded and uploaded into Cytoscape⁵², where pruned edges with Δm/z (±0.003 Da) = 176.032, 194.042, and 96.075 were re-added to the networks.

MassQL pattern filtering of glucuronidated features

The MassQL query used for negative ESI pattern filtering in the biologic samples was:

QUERY scaninfo(MS2DATA) WHERE MS2NL = (176.0321 OR 194.0425):INTENSITYPERCENT = 5:TOLERANCEMZ = 0.003 AND MS2PROD = 113.0244:INTENSITYPERCENT = 3:TOLERANCEMZ = 0.003 AND MS2PROD = 85.0295:INTENSITYPERCENT = 3:TOLERANCEMZ = 0.003.

The query corresponds to the glucuronide neutral losses and common fragments detected in the glucuronidated standards (Fig. 1B and Supplementary Fig. 1A). Intensity percent tolerances were added to the queries generated for the standards to account for increased MS/MS noise in the biologic samples. Positive ESI data was queried only for the glucuronide neutral losses:

QUERY scaninfo(MS2DATA) WHERE MS2NL = (176.0321 OR 194.0425):INTENSITYPERCENT = 5:TOLERANCEMZ = 0.003.

Microbial drop plates (GF/FMT)

Germ-free/colonized status was confirmed via anaerobic culture of microbial drop plates for day 0 and day 7 feces under anaerobic conditions (20% CO₂, 5% H₂, 75% N₂) using brain heart infusion agar (BHI) supplemented with 0.05% w/v cysteine, 5 μg/mL hemin, and 1 μg/mL vitamin K3. Feces were stored at −80 °C prior to weighing out ~25 mg over dry ice and transferring to the anaerobic chamber. All culture steps were completed inside the anaerobic chamber. Fecal samples were resuspended in 1 mL of 1× phosphate buffered saline (PBS) and vortexed to mix. Serial dilutions were performed in 96-well plates, pre-loaded with 90 μL 1× PBS by adding 10 μL of sample mixture to the first column and transferring 10 μL down the plate for 7 dilutions. Dilutions 10e-6 to 10e-9 were plated at 5 μL in quadruplicate. Sample droplets were allowed to dry on the plates for 5–10 min (uncovered and upright), then covered and inverted to incubate at 37 °C for 72 h. Colony growth was counted at 24, 48, and 72 h. No growth was considered confirmation of germ-free status.

Gut microbiome DNA extraction and sequencing

Approximately 25 mg of fecal material was weighed out into 2 mL homogenizer tubes (10025-754, VWR, Radnor, PA) containing 300–350 mg of 1 mm silica beads (NC9847287, Fisher Scientific, Fair Lawn, NJ). Microbial DNA was extracted using SOP 6 from the International Human Microbiome Standards (IHMS) and the QIAamp Fast DNA Stool Mini Kit (51604, Qiagen, Hilden, Germany), as per the protocol, with the following deviations. Samples were homogenized at 5000 rpm for 5 min in 60 s intervals, with 60 s pause in between each interval. Samples were eluted in 100 μL of ATE buffer from the kit. DNA purity and quality, respectively, were tested via NanoDrop One (Thermo Scientific, Waltham MA) and Qubit 4 Fluorometer (Invitrogen, Thermo Scientific, Waltham, MA). Ethanol precipitation and purification was performed as necessary. In brief, 0.1 volume (10 μL) of 3 M sodium acetate and 2.5 volumes (250 μL) or ice cold 100% ethanol was added and vortexed to mix. Sample(s) were incubated at −80 °C for 1 hour to re-precipitate the DNA. Following a 1-h incubation at −80 °C, the sample(s) were centrifuged at max speed (18,213 × g) and 4 °C for 20 min. The supernatant was removed, and the pellet was washed with 300 μL of ice cold 75% ethanol 3× by adding the ethanol, centrifuging at max speed and 4 °C for 5 min, then removing the ethanol. After the last wash, the pellet was air dried for 5 min and resuspended in 100 μL of Buffer ATE.

Library prep and shotgun metagenomic sequencing was performed on 25 μL of DNA product by Novogene (Sacramento, CA) on a NovaSeq PE150 (Illumina, San Diego, CA).

The microbial genomic DNA extraction and sequencing of the HFD/NFD mouse samples was performed as described previously⁶³.

Microbial metagenomic analysis

Initial processing of mouse metagenomic samples

Raw metagenomic reads were trimmed, filtered, and assembled de novo into ORFs (open reading frames), using NGLess (v1.4.2)⁸² with default parameters. Generic feature format (GFF) files and predicted protein sequences were generated from the ORFs using Prodigal (v2.6.3)⁸³.

Identification and characterization of GUS gene sequences from mouse gut metagenomic data

Metagenomic amino acid sequences were clustered to remove redundancies using CD-HIT (v4.8.1)⁸⁴ at a sequence identity threshold of 100%. These representative protein sequences were then used to identify putative GUS enzymes as reported previously^9,17. Accepted sequences were clustered again at 95% using CD-HIT, forming a representative set of GUS sequences for downstream analysis. Representative sequences were aligned to reference sequences from each loop class in a multiple sequence alignment using Clustal Omega⁸⁵, and the GUS class was assigned according to parameters reported previously^9,17. Taxonomy was assigned to representative GUS sequences with a lowest common ancestor algorithm, using Diamond (v2.0.15.153)⁸⁶ to map GUS representatives to proteins in the CMMG⁸⁷.

Metagenomic gene intensity quantitation

For each sample, an alignment index was created from the ORFs using bowtie2⁸⁸. A SAM file was then created using Bowtie2 by aligning the trimmed and filtered reads to the alignment index. The SAM and GFF files were used to generate gene read counts in each sample, using featureCounts from the package SUBREAD (v2.0.1)⁸⁹. Gene read counts for each sample were normalized based on the total reads detected in each sample and in all samples, yielding normalized read counts. These normalized read counts were then filtered to only the representative GUS sequences, and these GUS normalized read counts were further normalized to correct for the association between gene length and normalized read counts. These methods were implemented as reported in previous work^17,90.

GUS gene intensity analysis

Differential abundance analysis was performed on metagenomic GUS genes summed at 95% identity. For the antibiotics study, analysis was performed with linear mixed effects models using lmer and emmeans. GUS gene abundance was predicted as a function of the fixed effect “specific treatment” (control, gentamicin, vancomycin) as well as a random effect of the individual “mouse ID.” For the high/normal fat diet study, differences in GUS gene intensity were assessed using Kruskal–Wallis tests. For endpoint samples in both the ABX and HF/NF study, simple linear regression analysis was performed for all GUS genes against all glucuronide/aglycone metabolites of interest using lm.

Microbial GUS assays and LC-MS/MS

All GUS enzymes were expressed in E. coli and purified as described previously (Simpson et al.¹⁷). Naringenin-7-O-β-D-glucuronide standard was purchased from Cayman Chemical, and naringenin standard was purchased from Sigma Aldrich. Lyophilized powder was resuspended in DMSO to a concentration of 40 mM. Assays with 24 representative purified GUS enzymes were conducted in Costar half-area 96-well plates at a total volume of 50 µL. The reaction consisted of 5 µL assay buffer (250 mM HEPES, 250 mM NaCl, pH 6.5), 5 µL of purified GUS enzyme (2 nM), 5 µL of Naringenin-7-O-β-D-glucuronide (500 µM stock), and 35 µL of ddH20. The reaction condition for Schaedlerella_MGG38568_4_CTD GUS assays were 5 µL assay buffer (250 mM HEPES, 250 mM NaCl, pH 6.5), 5 µL of purified Schaedlerella_MGG38568_4_CTD GUS (10 nM, 100 nM, 250 nM, 750 nM, and 1 µM), 5 µL of Naringenin-7-O-β-D-glucuronide (500 µM stock), and 35 µL of ddH20. The reaction was incubated for 30 min at 37 °C, then quenched by adding 50 µL of 25% TCA solution. The samples were transferred to 1.5 mL tubes and microcentrifuged for 10 min at 17,000 × g and 4 °C to precipitate proteins. The supernatants were carefully transferred to LC vials for downstream LC-MS/MS analysis. Naringenin-7-O-β-D-glucuronide and naringenin concentrations after 30 min of enzyme reaction were measured on an Agilent 1260 Infinity II liquid chromatography system. Sample injection volume was 5 µL, and samples were separated on an Agilent InfinityLab Poroshell 120 EC-C18 column (4.6 × 150 mm, 2.7-µm particle size) at 40 °C with a flow rate of 1 mL/min. Mobile phase solvent A was 95% water, 5% ACN, 0.1% formic acid, and solvent B was 95% ACN, 5% water, 0.1% formic acid. The gradient started at 5% solvent B at time zero and kept for 3 min, then gradually increased to 90% solvent B by 10 min and maintained for 2 min. At 12.1 min, the gradient was switched back to 5% solvent B and maintained until the method ended at 15 min. After separation, samples were introduced to an Agilent 6460 Triple Quad Mass Spectrometer. Naringenin-7-O-β-D-glucuronide (446.9 > 271) eluted at 7.8 min and naringenin (271 > 150.9, 118.9) eluted at 9.2 min were detected using multiple reaction monitoring in negative polarity mode. The source gas temperature was set to 300 °C at 5 L/min flow rate. The sheath gas temperature was 250 °C at 11 L/min. The capillary voltage was -3500V. Nebulizer pressure was 45 psi.

Microbial deconjugation assays and LC-MS/MS

Wild type and uidA (GUS) knock out K12 Escherichia coli (E. coli) were grown up in LB broth and seeded at 1% into 1 mL of fresh LB broth with 0, 1, or 10 μM of Naringenin-7-O-β-D-glucuronide and then kept at 37 °C and ambient air, shaking at 200 rpm for 24 h. Sterile controls were performed for all treatment doses. All treatment conditions were performed in triplicate. At termination, the whole culture was collected and stored at −80 °C. Two hundred microlitres were extracted for LC-MS/MS by the following protocol. Four volumes of ice cold 100% LC-MS grade methanol with 0.1% FA and 1 μM chlorpropamide was added, followed by three freeze-thaw cycles in liquid nitrogen. After sonication for 10 min and centrifugation at max speed and 4 °C for 15 min, 250 μL of the supernatant was transferred to fresh tubes and dried down under liquid nitrogen. Dried samples were stored in −80 °C until they were resuspended in 50 μL of 50% LC-MS methanol, vortexed, sonicated for 10 min and incubated on ice for 20 min before centrifugation at 4 °C and max speed for 15 min. The supernatant was transferred to 250 μL injection vials and 5 μL of sample or 10 μM naringenin and naringenin-7-O-β-D-glucuronide reference standards in 50% MeoH were injected for LC-MS/MS analysis. Samples were analyzed with the LC-MS/MS conditions described above for the standards and mouse samples.

IGRAM cohort, LC-MS/MS and metagenomics

The Infant Growth and Microbiome (IGRAM) Study, previously reported⁵⁴, was a prospective, longitudinal, observational maternal-infant cohort study that enrolled women in their third trimester of pregnancy and followed them through the first 12 months of life. Stool and urine samples were collected from infants at 1 month, 4 months, and 12 months of age. Information regarding feeding types (e.g., exclusively breastfed, mixed, exclusively formula, and other milk) and xenobiotic exposures (e.g., acetaminophen) were collected via questionnaire at the time of sample collection (Supplementary Data 11). The study was approved by the Committee for the Protection of Human Subjects (Internal Review Board) at Children’s Hospital of Philadelphia (#14-010833), and informed consent was obtained from all subjects.

Urine from the infants with paired samples from all three timepoints (n = 43) were analyzed. Urine samples (50 μL) were extracted as reported for urine in the above “Metabolite extraction and sample preparation” methods section, and LC-MS/MS data generated as described in the above “Standards and LC-MS/MS data acquisition” methods section and Methods Supplementary Data 13.

Microbial GUS gene abundance from the same subjects and time points for which metabolomic data was available was analyzed as described in the “Microbial metagenomic analysis” section.

Principal coordinated analysis was performed with Bray–Curtis distances and hypothesis testing by PERMANOVA. The glucuronidome and GUSome comparisons were made with a linear mixed effects model with feeding type and visit as fixed effects and subjects as random effects. Post-hoc tests were conducted using the emmeans package to identify the pairwise differences between visits. A separate set of linear models were then conducted for each visit to establish the differences between feeding types. Multiple tests were corrected for false discovery rate using the Benjamini–Hochberg method. Simple linear models were used to compare GUS gene intensities with glucuronidome features at the 12 M timepoint, using lm in R.

Glucuronidome repository searches

The MS/MS spectra of the glucuronide standards were searched against the indexed datasets part of the GNPS2/MassIVE, Metabolights, and Metabolomics Workbench repositories using fast MASST (FASST)⁹¹, which is an updated version of MASST⁵⁶. The parameters of the searches were specified as follows: minimum cosine similarity of 0.7, precursor and fragment ion tolerance of 0.02 Da, minimum number of matching fragments of 4, and database “metabolomicspanrepo_index_latest.” These searches were performed on February 10th (2025) using the RESTful web API (https://github.com/robinschmid/microbe_masst)^92,93. Since the FASST searches are performed on indexed and filtered spectra from the public domain (which allows the searches to be performed in seconds), we performed an additional filter of these results to increase its confidence level. This filter consisted of calculating the cosine similarity between the queried spectra and the pre-indexed unfiltered spectra and further removing any MS/MS spectra that resulted in a cosine similarity below 0.7. The outputs were uploaded to Zenodo at 10.5281/zenodo.1944288493. The code used for filtering the FASST results to ensure a higher confidence in the matches are available at https://github.com/helenamrusso/cosine_calculation_FASST_search. To further validate the spectral matches identified through FASST, cosine similarity scores were independently recalculated using the original, unfiltered MS/MS spectra. The spectra were retrieved via the Metabolomics Spectrum Resolver API (10.1101/2020.05.09.086066) and similarity scores were computed using the matchms (10.21105/joss.0241194) Python package.

To explore the distribution of the glucuronidated molecules in different organisms and body parts, we merged the FASST output with the ReDU controlled vocabulary metadata^57,58. This merged table was further filtered to only contain rows relative to humans (“9606|Homo sapiens”) or rodents (“10088|Mus,” “10090|Mus musculus,” “10105|Mus minutoides,” “10114|Rattus,” “10116|Rattus norvegicus”) in the NCBITaxonomy column. The results of body part distribution (UBERONBodyPartName column) and disease ontology (DOIDCommonName column) in rodents and humans were visualized with heatmaps that were obtained using the “seaborn.clustermap” package (version 0.12.2) in Python (version 3.7.6). The body parts and biofluids heatmaps show the percentages of how many files of that body part had a match to glucuronidated compounds, considering repository availability for each body part in each ionization mode separately. On the other hand, the disease ontology heatmaps show in which diseases a glucuronidated compound was matched to (i.e., if a compound was only observed in one disease, it will be shown as 100%).

Statistics and reproducibility

Study design and statistical analysis information for each dataset is described in the respective methods sections above. The metagenomic and glucuronidated chemical analysis of the mouse studies was exploratory (HFD analysis was secondary), so no power calculation was performed to determine sample size. The glucuronidated chemical analysis of the human datasets were secondary and exploratory, so no power calculation was performed to determine sample size. The gnotobiotic experiment sample size was determined based on access to germ-free mice due to housing and breeding limitations. Mice were randomized to treatment groups. Samples were randomized prior to LC-MS/MS analysis. The human study was prospective, so there was no subject randomization. Blinding was not part of the study protocol and was not used. No data were excluded from this analysis. Studies were not repeated.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Supplementary information

Supplementary Information^{(1.5MB, pdf)}

41467_2026_73398_MOESM2_ESM.pdf^{(15KB, pdf)}

Description of Additional Supplementary Files

Supplemental Data 1–13^{(378KB, xlsx)}

Reporting Summary^{(92.1KB, pdf)}

Transparent Peer Review File^{(666.7KB, pdf)}

Acknowledgements

The coauthors acknowledge the Huck Institutes of the Life Sciences Metabolomics Core Facility (RRID:SCR_023864) and the Host Microbial Analytic and Repository Core (H-MARC) of the Center for Molecular Studies in Digestive and Liver Diseases (NIH P30DK050306). A.D.P. discloses support from National Institutes of Health (NIH) grant S10OD021750, NIH/National Institute of Environmental Sciences grant R35 ES035027, and the US Department of Agriculture National Institute of Food and Federal Appropriations under project PEN047702 and accession number 7006412. G.H.P. was supported by the National Institutes of Health Grants ES028244 and by the USDA National Institute of Food and Federal Appropriations under Project PEN04607 and Accession number 7000371. J.E.B. was supported by the NIH/National Institute of General Medical Sciences R35 GM151045. A.D.P., G.D.W., and M.R.R. were supported by the Helmsley Charitable Trust. N.R.B. was supported by NIH/National Institutes of Diabetes and Digestive Disease grant F31 DK134090 and the One Health Microbiome Center seed grant. M.R.R. was supported by NIH grants GM152079 and DK139249. H.M.-R., M.W. and P.C.D. are supported by NIH 5U24DK133658-02. M.S.K. was supported by NIH/ National Institute of Allergy and Infectious Disease F31 AI183815 and National Institute of Diabetes and Digestive Disease T32 DK120509. E.S.F., C.T., K.B. and J.P.Z. are supported by the Center for Microbial Medicine at the Children’s Hospital of Philadelphia. The Infant Growth and Microbiome Study (B.S.Z., G.D.W.) was supported by an unrestricted donation from the American Beverage Foundation for a Healthy America to the Children’s Hospital of Philadelphia to support the Healthy Weight Program, the Research Institute of the Children’s Hospital of Philadelphia, The PennCHOP Microbiome Program, the NIH National Center for Research Resources Clinical and Translational Science Program (grant no. UL1TR001878), the National Institute of Digestive Diseases and Disorders of the Kidney (grant no. R01DK107565). J.P.Z. was supported by the National Institutes of Health grant R35GM138369.

Author contributions

N.B., J.E.B. and A.D.P. conceived of the project. J.J.S. performed GUS analyses and GUS*metabolite correlations for all datasets. HMR performed the MASST searches, generated heatmaps and conducted primary analysis of MSV000084112. J.J.P. performed the GUS activity assays with naringenin-7-O-glucuronide and generated Schaedlerella_MGG38568_4 CTD gus. M.W. adapted GNPS tools for better integration into the analysis pipeline and provided guidance for creating the glucuronidome resource. M.S.K. helped with LC-MS method development and provided LC-MS and MS/MS analysis guidance. I.K. provided guidance for metabolomics and statistical analyses. E.M., I.A.M. performed the APAP experiment and collected urine samples. S.T. provided guidance and assisted with ABX and FMT experiments and anaerobic procedures. M.A. provided assistance with mouse experiments, sample processing and troubleshooting. Y.T. performed the HFD/NFD mouse experiment and provided assistance with mouse sample collection and guidance for the E. coli deconjugation assay. I.K., J.E.B., M.R.R., J.P.Z., K.B., J.D.L., G.D.W. and B.S.Z. provided data analysis and interpretation. G.H.P., G.D.W., P.C.D., M.W., J.E.B., M.R.R. and A.D.P. provided feedback and guidance for the duration of the project. N.B., J.J.S., M.R.R., P.C.D. and A.D.P. wrote the manuscript. All authors edited and approved the final manuscript.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Data availability

Mouse untargeted metabolomics raw data generated from the above studies are deposited on GNPS2/MassIVE (https://massive.ucsd.edu) under accession number: MSV00097426. The glucuronidated standards libraries are available as part of the GNPS2 public libraries (https://external.gnps2.org/gnpslibrary). Mouse metagenomic sequencing files are available at SRA accession number: PRJNA1241896. IGRAM MS/MS data are deposited at MSV000098859, and sequencing data is available under SRA accessions PRJNA1379156, PRJNA1173239, PRJNA1042647, PRJNA1106565, PRJNA1145027 and PRJNA557731 (https://www.ncbi.nlm.nih.gov/bioproject/557731) (more details in Supplementary Data 11).

Code availability

All original code used to perform the data analyses and generate the figures is deposited at https: https://github.com/NinaRBoyle/Glucuronidation and is publicly available as of the date of publication.

Competing interests

P.C.D. is an advisor and holds equity in Cybele, Sirenas, and BileOmix, and he is a scientific co-founder, advisor, holds equity and/or receives income from Ometa, Enveda, and Arome with prior approval by UC San Diego. P.C.D. consulted for DSM Animal Health in 2023. M.R.R. is a founder of Symberix, Inc., and has received funding from Merck and Lilly. M.W. is a co-founder of Ometa Labs LLC. G.D.W. is a member of advisory boards for Danone, BioCodex, and the Institute for the Advancement of Food and Nutrition Sciences. He has also received research funding from Nestle. J.P.Z. has previously consulted for Vedanta Biosciences, Inc., and AstraZeneca. The remaining authors declare no competing interests.

Footnotes

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

These authors contributed equally: Nina R. Boyle, Josh J. Sekela.

Supplementary information

The online version contains Supplementary material available at 10.1038/s41467-026-73398-1.

References

1.Yang, G. et al. Glucuronidation: driving factors and their impact on glucuronide disposition. Drug Metab. Rev.49, 105 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
2.Pellock, S. J. & Redinbo, M. R. Glucuronides in the gut: sugar-driven symbioses between microbe and host. J. Biol. Chem.292, 8569–8576 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
3.Mackenzie, P. I. et al. Nomenclature update for the mammalian UDP glycosyltransferase (UGT) gene superfamily. Pharmacogenet. Genom.15, 677–685 (2005). [DOI] [PubMed]
4.Janov, P. & Iller, M. Phase II drug metabolism. in Topics on Drug Metabolism (ed. Paxton, J.) 35–60 (InTech, 2012).
5.Meech, R. et al. The UDP-glycosyltransferase (UGT) superfamily: new members, new functions, and novel paradigms. Physiol. Rev.99, 1153–1222 (2019). [DOI] [PubMed] [Google Scholar]
6.Collins, S. L. & Patterson, A. D. The gut microbiome: an orchestrator of xenobiotic metabolism. Acta Pharm. Sin. B10, 19–32 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
7.Dashnyam, P. et al. β-glucuronidases of opportunistic bacteria are the major contributors to xenobiotic-induced toxicity in the gut. Sci. Rep.8, 16372 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
8.Javdan, B. et al. Personalized mapping of drug metabolism by the human gut microbiome. Cell181, 1661–1679 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
9.Pollet, R. M. et al. An atlas of β-glucuronidases in the human intestinal microbiome. Structure25, 967–977 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]
10.Ibarra, M., Trocóniz, I. F. & Fagiolino, P. Enteric reabsorption processes and their impact on drug pharmacokinetics. Sci. Rep.11, 5794 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
11.Gloux, K. & Anba-Mondoloni, J. Unique β-glucuronidase locus in gut microbiomes of Crohn’s disease patients and unaffected first-degree relatives. PLoS ONE11, e0148291 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
12.LoGuidice, A., Wallace, B. D., Bendel, L., Redinbo, M. R. & Boelsterli, U. A. Pharmacologic targeting of bacterial β-glucuronidase alleviates nonsteroidal anti-inflammatory drug-induced enteropathy in mice. J. Pharmacol. Exp. Ther.341, 447–454 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
13.Zimmermann, M., Zimmermann-Kogadeeva, M., Wegmann, R. & Goodman, A. L. Mapping human microbiome drug metabolism by gut bacteria and their genes. Nature570, 462–467 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
14.Patel, J., Chaudhary, H., Rajput, K., Parekh, B. & Joshi, R. Assessment of gut microbial β-glucuronidase and β-glucosidase activity in women with polycystic ovary syndrome. Sci. Rep.13, 11967 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Creekmore, B. C. et al. Mouse gut microbiome-encoded β-glucuronidases identified using metagenome analysis guided by protein structure. mSystems4, 10–1128 (2019). [DOI] [PMC free article] [PubMed]
16.Walker, M. E., Simpson, J. B. & Redinbo, M. R. A structural metagenomics pipeline for examining the gut microbiome. Curr. Opin. Struct. Biol.75, 102416 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
17.Simpson, J. B. et al. Gut microbial β-glucuronidases influence endobiotic homeostasis and are modulated by diverse therapeutics. Cell Host Microbe32, 925–944 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
18.GUSB protein expression summary—The Human Protein Atlas, https://www.proteinatlas.org/ENSG00000169919-GUSB.
19.Thul, P. J. et al. A subcellular map of the human proteome. Science356, eaal3321 (2017). [DOI] [PubMed]
20.National Library of Medicine (US). Gusb glucuronidase, beta [Mus musculus]. National Center for Biotechnology Information,https://www.ncbi.nlm.nih.gov/gene/110006 (2004).
21.Kim, S. et al. PubChem 2025 update. Nucleic Acids Res.53, D1516–D1525 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
22.Sorokina, M., Merseburger, P., Rajan, K., Yirik, M. A. & Steinbeck, C. COCONUT online: collection of open natural products database. J. Cheminform.13, 2 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
23.Ma, Y., Kind, T., Yang, D., Leon, C. & Fiehn, O. MS2Analyzer: a software for small molecule substructure annotations from accurate tandem mass spectra. Anal. Chem.86, 10724–10731 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Wang, M. et al. Sharing and community curation of mass spectrometry data with Global Natural Products Social Molecular Networking. Nat. Biotechnol.34, 828–837 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]
25.Huber, C. et al. In silico deconjugation of glucuronide conjugates enhances tandem mass spectra library annotation of human samples. Anal. Bioanal. Chem.414, 2629–2640 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
26.Rodríguez, E. P., Li, Y., Vaniya, A., Shih, P. M. & Fiehn, O. Alternative identification of glycosides using MS/MS matching with an in silico-modified aglycone mass spectra library. Anal. Chem. 10.1021/ACS.ANALCHEM.3C00957/SUPPL_FILE/AC3C00957_SI_002.XLSX (2023). [DOI] [PMC free article] [PubMed]
27.Jarmusch, A. K. et al. A universal language for finding mass spectrometry data patterns. Nat. Methods22, 1247–1254 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Kuuranne, T., Vahermo, M., Leinonen, A. & Kostianen, R. Electrospray and atmospheric pressure chemical ionization tandem mass spectrometric behavior of eight anabolic steroid glucuronides. J. Am. Soc. Mass Spectrom.11, 722–730 (2000). [DOI] [PubMed] [Google Scholar]
29.Fabregat, A., Pozo, O. J., Marcos, J., Segura, J. & Ventura, R. Use of LC-MS/MS for the open detection of steroid metabolites conjugated with glucuronic acid. Anal. Chem.85, 5005–5014 (2013). [DOI] [PubMed] [Google Scholar]
30.Murray, K. J., Villalta, P. W., Griffin, T. J. & Balbo, S. Discovery of modified metabolites, secondary metabolites, and xenobiotics by structure-oriented LC-MS/MS. Chem. Res. Toxicol. 10.1021/acs.chemrestox.3c00209 (2023). [DOI] [PubMed]
31.Watrous, J. et al. Mass spectral molecular networking of living microbial colonies. Proc. Natl. Acad. Sci. USA109, E1743–E1752 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
32.Chen, C., Krausz, K. W., Idle, J. R. & Gonzalez, F. J. Identification of novel toxicity-associated metabolites by metabolomics and mass isotopomer analysis of acetaminophen metabolism in wild-type and Cyp2e1-null mice. J. Biol. Chem.283, 4543–4559 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]
33.Betowski, L. D., Korfmacher, W. A., Lay, J. O., Potter, D. W. & Hinson, J. A. Direct analysis of rat bile for acetaminophen and two of its conjugated metabolites via thermospray liquid chromatography/mass spectrometry. Biomed. Environ. Mass Spectrom.14, 705–709 (1987). [DOI] [PubMed] [Google Scholar]
34.Cooper, A. J. L. et al. Cysteine S-conjugate β-lyases: important roles in the metabolism of naturally occurring sulfur and selenium-containing compounds, xenobiotics and anticancer agents. Amino Acids41, 7–27 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]
35.David, A. et al. Acetaminophen metabolism revisited using non-targeted analyses: implications for human biomonitoring. Environ. Int.149, 106388 (2021). [DOI] [PubMed] [Google Scholar]
36.Gorrochategui, E. et al. High-resolution mass spectrometry identifies delayed biomarkers for improved precision in acetaminophen/paracetamol human biomonitoring. Environ. Int.181, 108299 (2023). [DOI] [PubMed] [Google Scholar]
37.McGill, M. R. & Jaeschke, H. Metabolism and disposition of acetaminophen: recent advances in relation to hepatotoxicity and diagnosis. Pharm. Res.30, 2174–2187 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Nothias, L.-F. et al. Feature-based molecular networking in the GNPS analysis environment. Nat. Methods17, 905–908 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
39.Malfatti, M. A. et al. Manipulation of the gut microbiome alters acetaminophen biodisposition in mice. Sci. Rep.10, 4571 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
40.Takasuna, K. et al. Inhibition of intestinal microflora β-glucuronidase modifies the distribution of the active metabolite of the antitumor agent, irinotecan hydrochloride (CPT-11) in rats. Cancer Chemother. Pharmacol.42, 280–286 (1998). [DOI] [PubMed] [Google Scholar]
41.Taylor, M. R. et al. Vancomycin relieves mycophenolate mofetil–induced gastrointestinal toxicity by eliminating gut bacterial β-glucuronidase activity. Sci. Adv.5, eaax2358 (2019). [DOI] [PMC free article] [PubMed]
42.Bhatt, A. P. et al. Targeted inhibition of gut bacterial β-glucuronidase activity enhances anticancer drug efficacy. Proc. Natl. Acad. Sci. USA117, 7374–7381 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
43.Ervin, S. M. et al. Gut microbial β-glucuronidases reactivate estrogens as components of the estrobolome that reactivate estrogens. J. Biol. Chem.294, 18586–18599 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
44.Asano, Y. et al. Critical role of gut microbiota in the production of biologically active, free catecholamines in the gut lumen of mice. Am. J. Physiol. Gastrointest. Liver Physiol. 303, G1288–G1295 (2012). [DOI] [PubMed] [Google Scholar]
45.Kim, D. H. et al. Intestinal bacterial metabolism of flavonoids and its relation to some biological activities. Arch. Pharm. Res.21, 17–23 (1998). [DOI] [PubMed] [Google Scholar]
46.Zhang, J. et al. Microbial enzymes induce colitis by reactivating triclosan in the mouse gastrointestinal tract. Nat. Commun. 13, 136 (2022). [DOI] [PMC free article] [PubMed]
47.Aggarwal, H. et al. Vancomycin-induced modulation of gram-positive gut bacteria and metabolites remediates insulin resistance in iNOS knockout mice. Front. Cell. Infect. Microbiol.11, 795333 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
48.Huang, C., Feng, S., Huo, F. & Liu, H. Effects of four antibiotics on the diversity of the intestinal microbiota. Microbiol. Spectr.10, e0190421 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
49.Ray, P., Chakraborty, S., Ghosh, A. & Aich, P. Effects of treatment with three antibiotics, vancomycin, neomycin, and AVNM on gut microbiome in C57BL/6 mice. Preprint at bioRxiv10.1101/2021.02.08.430372 (2021).
50.Gheorghe, C. E. et al. Investigating causality with fecal microbiota transplantation in rodents: applications, recommendations and pitfalls. 13, 10.1080/19490976.2021.1941711 (2021). [DOI] [PMC free article] [PubMed]
51.Kennedy, E. A., King, K. Y. & Baldridge, M. T. Mouse microbiota models: comparing germ-free mice and antibiotics treatment as tools for modifying gut bacteria. Front. Physiol.9, 1534 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]
52.Shannon, P. et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res.13, 2498–2504 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar]
53.Biernat, K. A. et al. Structure, function, and inhibition of drug reactivating human gut microbial β-glucuronidases. Sci. Rep. 9, 825 (2019). [DOI] [PMC free article] [PubMed]
54.Bittinger, K. et al. Bacterial colonization reprograms the neonatal gut metabolome. Nat. Microbiol.5, 838–847 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
55.Rimal, B. et al. Bile salt hydrolase catalyses formation of amine-conjugated bile acids. Nature626, 859–863 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
56.Wang, M. et al. Mass spectrometry searches using MASST. Nat. Biotechnol.38, 23–26 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
57.Jarmusch, A. K. et al. ReDU: a framework to find and reanalyze public mass spectrometry data. Nat. Methods17, 901–904 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]
58.El Abiead, Y. et al. Enabling pan-repository reanalysis for big data science of public metabolomics data. Nat. Commun.16, 4838 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]
59.Śliżewska, K. et al. Comparison of the activity of fecal enzymes and concentration of SCFA in healthy and overweight children. Nutrients15, 987 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
60.Connell, E. et al. Microbial-derived metabolites as a risk factor of age-related cognitive decline and dementia. Mol. Neurodegener.17, 43 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
61.Li, P. et al. Vancomycin relieves tacrolimus-induced hyperglycemia by eliminating gut bacterial beta-glucuronidase enzyme activity. Gut Microbes16, 2310277 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
62.Zhang, L. & Chu, C.-Q. Gut microbiota-medication interaction in rheumatic diseases. Front. Immunol.12, 796865 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
63.Tian, Y. et al. Early life polychlorinated biphenyl 126 exposure disrupts gut Microbiota and metabolic homeostasis in mice fed with high-fat diet in adulthood. Metabolites12, 894 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
64.Tsiara, I. et al. Characterizing the sulfated and glucuronidated (poly)phenol metabolome for dietary biomarker discovery. J. Agric. Food Chem. 10.1021/acs.jafc.4c12596 (2025). [DOI] [PMC free article] [PubMed]
65.Ho, A. et al. Circulating glucuronic acid predicts healthspan and longevity in humans and mice. Aging11, 7694 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
66.Hu, S. et al. Gut microbial beta-glucuronidase: a vital regulator in female estrogen metabolism. Gut Microbes15, 2236749 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]
67.Leonard, L. M., Simpson, A. M. R., Li, S., Reddivari, L. & Cross, T.-W. L. A Gnotobiotic mouse model with divergent equol-producing phenotypes: potential for determining microbial-driven health impacts of soy isoflavone daidzein. Nutrients16, 1079 (2024). [DOI] [PMC free article] [PubMed]
68.Letertre, M. P. M. et al. Characterizing the metabolic effects of the selective inhibition of gut microbial β-glucuronidases in mice. Sci. Rep.12, 17435 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
69.Oto, J. et al. Validation of urine p-cresol glucuronide as renal cell carcinoma non-invasive biomarker. J. Proteom.311, 105357 (2024). [DOI] [PubMed] [Google Scholar]
70.Mykkänen, H., Tikka, J., Pitkänen, T. & Hänninen, O. Fecal bacterial enzyme activities in infants increase with age and adoption of adult-type diet. J. Pediatr. Gastroenterol. Nutr.25, 312–316 (1997). [DOI] [PubMed] [Google Scholar]
71.Elmassry, M. M., Kim, S. & Busby, B. Predicting drug-metagenome interactions: Variation in the microbial β-glucuronidase level in the human gut metagenomes. PLoS ONE16, e0244876 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]
72.Candeliere, F. et al. -Glucuronidase pattern predicted from gut metagenomes indicates potentially diversified pharmacomicrobiomics. Front. Microbiol.13, 826994 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
73.Vargas, F. et al. Protocol for community-created public MS/MS reference spectra within the Global Natural Products Social Molecular Networking infrastructure. Rapid Commun. Mass Spectrom.34, e8725 (2020). [DOI] [PubMed] [Google Scholar]
74.Aron, A. T. et al. Reproducible molecular networking of untargeted mass spectrometry data using GNPS. Nat. Protoc.15, 1954–1991 (2020). [DOI] [PubMed] [Google Scholar]
75.Chambers, M. C. et al. A cross-platform toolkit for mass spectrometry and proteomics. Nat. Biotechnol.30, 918–920 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
76.R Core Team. R: A Language and Environment for Statistical Computing, https://www.R-project.org/ (2025).
77.Wickham, H. ggplot2: Elegant Graphics for Data Analysis, https://ggplot2.tidyverse.org (2016).
78.Oksanen, J. et al. vegan: Community Ecology Package, https://CRAN.R-project.org/package=vegan (2025).
79.Paradis, E. & Schliep, K. ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics35, 526–528 (2019). [DOI] [PubMed]
80.Acker, D. gg3D: 3D perspective plots for ggplot2, https://github.com/AckerDWM/gg3D (2024).
81.Signorell, A. DescTools: Tools for Descriptive Statistics, https://CRAN.R-project.org/package=DescTools (2025).
82.Coelho, L. P. et al. NG-meta-profiler: fast processing of metagenomes using NGLess, a domain-specific language. Microbiome7, 84 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]
83.Hyatt, D. et al. Prodigal: prokaryotic gene recognition and translation initiation site identification. BMC Bioinform.11, 119 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]
84.Li, W. & Godzik, A. Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences. Bioinformatics22, 1658–1659 (2006). [DOI] [PubMed] [Google Scholar]
85.Sievers, F. & Higgins, D. G. Clustal Omega, accurate alignment of very large numbers of sequences. Methods Mol. Biol.1079, 105–116 (2014). [DOI] [PubMed] [Google Scholar]
86.Buchfink, B., Xie, C. & Huson, D. H. Fast and sensitive protein alignment using DIAMOND. Nat. Methods12, 59–60 (2015). [DOI] [PubMed] [Google Scholar]
87.Kieser, S., Zdobnov, E. M. & Trajkovski, M. Comprehensive mouse microbiota genome catalog reveals major difference to its human counterpart. PLoS Comput. Biol.18, e1009947 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]
88.Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods9, 357–359 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]
89.Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics30, 923–930 (2014). [DOI] [PubMed] [Google Scholar]
90.Simpson, J. B. et al. Metagenomics combined with activity-based proteomics point to gut bacterial enzymes that reactivate mycophenolate. Gut Microbes14, 2107289 (2022). [DOI] [PMC free article] [PubMed]
91.Batsoyol, N., Pullman, B., Wang, M., Bandeira, N. & Swanson, S. P-massive: a real-time search engine for a multi-terabyte mass spectrometry database. In Proc. SC22: International Conference for High Performance Computing, Networking, Storage and Analysis, 1–15 (IEEE, 2022).
92.Zuffa, S. et al. microbeMASST: a taxonomically informed mass spectrometry search tool for microbial metabolomics data. Nat. Microbiol.9, 336–345 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]
93.Huber, F. et al. Matchms - processing and similarity evaluation of mass spectrometry data. J. Open Source Softw.5, 2411 (2020). [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Information^{(1.5MB, pdf)}

41467_2026_73398_MOESM2_ESM.pdf^{(15KB, pdf)}

Description of Additional Supplementary Files

Supplemental Data 1–13^{(378KB, xlsx)}

Reporting Summary^{(92.1KB, pdf)}

Transparent Peer Review File^{(666.7KB, pdf)}

Data Availability Statement

[CR1] 1.Yang, G. et al. Glucuronidation: driving factors and their impact on glucuronide disposition. Drug Metab. Rev.49, 105 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR2] 2.Pellock, S. J. & Redinbo, M. R. Glucuronides in the gut: sugar-driven symbioses between microbe and host. J. Biol. Chem.292, 8569–8576 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR3] 3.Mackenzie, P. I. et al. Nomenclature update for the mammalian UDP glycosyltransferase (UGT) gene superfamily. Pharmacogenet. Genom.15, 677–685 (2005). [DOI] [PubMed]

[CR4] 4.Janov, P. & Iller, M. Phase II drug metabolism. in Topics on Drug Metabolism (ed. Paxton, J.) 35–60 (InTech, 2012).

[CR5] 5.Meech, R. et al. The UDP-glycosyltransferase (UGT) superfamily: new members, new functions, and novel paradigms. Physiol. Rev.99, 1153–1222 (2019). [DOI] [PubMed] [Google Scholar]

[CR6] 6.Collins, S. L. & Patterson, A. D. The gut microbiome: an orchestrator of xenobiotic metabolism. Acta Pharm. Sin. B10, 19–32 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR7] 7.Dashnyam, P. et al. β-glucuronidases of opportunistic bacteria are the major contributors to xenobiotic-induced toxicity in the gut. Sci. Rep.8, 16372 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR8] 8.Javdan, B. et al. Personalized mapping of drug metabolism by the human gut microbiome. Cell181, 1661–1679 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR9] 9.Pollet, R. M. et al. An atlas of β-glucuronidases in the human intestinal microbiome. Structure25, 967–977 (2017). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR10] 10.Ibarra, M., Trocóniz, I. F. & Fagiolino, P. Enteric reabsorption processes and their impact on drug pharmacokinetics. Sci. Rep.11, 5794 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR11] 11.Gloux, K. & Anba-Mondoloni, J. Unique β-glucuronidase locus in gut microbiomes of Crohn’s disease patients and unaffected first-degree relatives. PLoS ONE11, e0148291 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR12] 12.LoGuidice, A., Wallace, B. D., Bendel, L., Redinbo, M. R. & Boelsterli, U. A. Pharmacologic targeting of bacterial β-glucuronidase alleviates nonsteroidal anti-inflammatory drug-induced enteropathy in mice. J. Pharmacol. Exp. Ther.341, 447–454 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR13] 13.Zimmermann, M., Zimmermann-Kogadeeva, M., Wegmann, R. & Goodman, A. L. Mapping human microbiome drug metabolism by gut bacteria and their genes. Nature570, 462–467 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR14] 14.Patel, J., Chaudhary, H., Rajput, K., Parekh, B. & Joshi, R. Assessment of gut microbial β-glucuronidase and β-glucosidase activity in women with polycystic ovary syndrome. Sci. Rep.13, 11967 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR15] 15.Creekmore, B. C. et al. Mouse gut microbiome-encoded β-glucuronidases identified using metagenome analysis guided by protein structure. mSystems4, 10–1128 (2019). [DOI] [PMC free article] [PubMed]

[CR16] 16.Walker, M. E., Simpson, J. B. & Redinbo, M. R. A structural metagenomics pipeline for examining the gut microbiome. Curr. Opin. Struct. Biol.75, 102416 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR17] 17.Simpson, J. B. et al. Gut microbial β-glucuronidases influence endobiotic homeostasis and are modulated by diverse therapeutics. Cell Host Microbe32, 925–944 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR18] 18.GUSB protein expression summary—The Human Protein Atlas, https://www.proteinatlas.org/ENSG00000169919-GUSB.

[CR19] 19.Thul, P. J. et al. A subcellular map of the human proteome. Science356, eaal3321 (2017). [DOI] [PubMed]

[CR20] 20.National Library of Medicine (US). Gusb glucuronidase, beta [Mus musculus]. National Center for Biotechnology Information,https://www.ncbi.nlm.nih.gov/gene/110006 (2004).

[CR21] 21.Kim, S. et al. PubChem 2025 update. Nucleic Acids Res.53, D1516–D1525 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR22] 22.Sorokina, M., Merseburger, P., Rajan, K., Yirik, M. A. & Steinbeck, C. COCONUT online: collection of open natural products database. J. Cheminform.13, 2 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR23] 23.Ma, Y., Kind, T., Yang, D., Leon, C. & Fiehn, O. MS2Analyzer: a software for small molecule substructure annotations from accurate tandem mass spectra. Anal. Chem.86, 10724–10731 (2014). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR24] 24.Wang, M. et al. Sharing and community curation of mass spectrometry data with Global Natural Products Social Molecular Networking. Nat. Biotechnol.34, 828–837 (2016). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR25] 25.Huber, C. et al. In silico deconjugation of glucuronide conjugates enhances tandem mass spectra library annotation of human samples. Anal. Bioanal. Chem.414, 2629–2640 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR26] 26.Rodríguez, E. P., Li, Y., Vaniya, A., Shih, P. M. & Fiehn, O. Alternative identification of glycosides using MS/MS matching with an in silico-modified aglycone mass spectra library. Anal. Chem. 10.1021/ACS.ANALCHEM.3C00957/SUPPL_FILE/AC3C00957_SI_002.XLSX (2023). [DOI] [PMC free article] [PubMed]

[CR27] 27.Jarmusch, A. K. et al. A universal language for finding mass spectrometry data patterns. Nat. Methods22, 1247–1254 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR28] 28.Kuuranne, T., Vahermo, M., Leinonen, A. & Kostianen, R. Electrospray and atmospheric pressure chemical ionization tandem mass spectrometric behavior of eight anabolic steroid glucuronides. J. Am. Soc. Mass Spectrom.11, 722–730 (2000). [DOI] [PubMed] [Google Scholar]

[CR29] 29.Fabregat, A., Pozo, O. J., Marcos, J., Segura, J. & Ventura, R. Use of LC-MS/MS for the open detection of steroid metabolites conjugated with glucuronic acid. Anal. Chem.85, 5005–5014 (2013). [DOI] [PubMed] [Google Scholar]

[CR30] 30.Murray, K. J., Villalta, P. W., Griffin, T. J. & Balbo, S. Discovery of modified metabolites, secondary metabolites, and xenobiotics by structure-oriented LC-MS/MS. Chem. Res. Toxicol. 10.1021/acs.chemrestox.3c00209 (2023). [DOI] [PubMed]

[CR31] 31.Watrous, J. et al. Mass spectral molecular networking of living microbial colonies. Proc. Natl. Acad. Sci. USA109, E1743–E1752 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR32] 32.Chen, C., Krausz, K. W., Idle, J. R. & Gonzalez, F. J. Identification of novel toxicity-associated metabolites by metabolomics and mass isotopomer analysis of acetaminophen metabolism in wild-type and Cyp2e1-null mice. J. Biol. Chem.283, 4543–4559 (2008). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR33] 33.Betowski, L. D., Korfmacher, W. A., Lay, J. O., Potter, D. W. & Hinson, J. A. Direct analysis of rat bile for acetaminophen and two of its conjugated metabolites via thermospray liquid chromatography/mass spectrometry. Biomed. Environ. Mass Spectrom.14, 705–709 (1987). [DOI] [PubMed] [Google Scholar]

[CR34] 34.Cooper, A. J. L. et al. Cysteine S-conjugate β-lyases: important roles in the metabolism of naturally occurring sulfur and selenium-containing compounds, xenobiotics and anticancer agents. Amino Acids41, 7–27 (2011). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR35] 35.David, A. et al. Acetaminophen metabolism revisited using non-targeted analyses: implications for human biomonitoring. Environ. Int.149, 106388 (2021). [DOI] [PubMed] [Google Scholar]

[CR36] 36.Gorrochategui, E. et al. High-resolution mass spectrometry identifies delayed biomarkers for improved precision in acetaminophen/paracetamol human biomonitoring. Environ. Int.181, 108299 (2023). [DOI] [PubMed] [Google Scholar]

[CR37] 37.McGill, M. R. & Jaeschke, H. Metabolism and disposition of acetaminophen: recent advances in relation to hepatotoxicity and diagnosis. Pharm. Res.30, 2174–2187 (2013). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR38] 38.Nothias, L.-F. et al. Feature-based molecular networking in the GNPS analysis environment. Nat. Methods17, 905–908 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR39] 39.Malfatti, M. A. et al. Manipulation of the gut microbiome alters acetaminophen biodisposition in mice. Sci. Rep.10, 4571 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR40] 40.Takasuna, K. et al. Inhibition of intestinal microflora β-glucuronidase modifies the distribution of the active metabolite of the antitumor agent, irinotecan hydrochloride (CPT-11) in rats. Cancer Chemother. Pharmacol.42, 280–286 (1998). [DOI] [PubMed] [Google Scholar]

[CR41] 41.Taylor, M. R. et al. Vancomycin relieves mycophenolate mofetil–induced gastrointestinal toxicity by eliminating gut bacterial β-glucuronidase activity. Sci. Adv.5, eaax2358 (2019). [DOI] [PMC free article] [PubMed]

[CR42] 42.Bhatt, A. P. et al. Targeted inhibition of gut bacterial β-glucuronidase activity enhances anticancer drug efficacy. Proc. Natl. Acad. Sci. USA117, 7374–7381 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR43] 43.Ervin, S. M. et al. Gut microbial β-glucuronidases reactivate estrogens as components of the estrobolome that reactivate estrogens. J. Biol. Chem.294, 18586–18599 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR44] 44.Asano, Y. et al. Critical role of gut microbiota in the production of biologically active, free catecholamines in the gut lumen of mice. Am. J. Physiol. Gastrointest. Liver Physiol. 303, G1288–G1295 (2012). [DOI] [PubMed] [Google Scholar]

[CR45] 45.Kim, D. H. et al. Intestinal bacterial metabolism of flavonoids and its relation to some biological activities. Arch. Pharm. Res.21, 17–23 (1998). [DOI] [PubMed] [Google Scholar]

[CR46] 46.Zhang, J. et al. Microbial enzymes induce colitis by reactivating triclosan in the mouse gastrointestinal tract. Nat. Commun. 13, 136 (2022). [DOI] [PMC free article] [PubMed]

[CR47] 47.Aggarwal, H. et al. Vancomycin-induced modulation of gram-positive gut bacteria and metabolites remediates insulin resistance in iNOS knockout mice. Front. Cell. Infect. Microbiol.11, 795333 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR48] 48.Huang, C., Feng, S., Huo, F. & Liu, H. Effects of four antibiotics on the diversity of the intestinal microbiota. Microbiol. Spectr.10, e0190421 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR49] 49.Ray, P., Chakraborty, S., Ghosh, A. & Aich, P. Effects of treatment with three antibiotics, vancomycin, neomycin, and AVNM on gut microbiome in C57BL/6 mice. Preprint at bioRxiv10.1101/2021.02.08.430372 (2021).

[CR50] 50.Gheorghe, C. E. et al. Investigating causality with fecal microbiota transplantation in rodents: applications, recommendations and pitfalls. 13, 10.1080/19490976.2021.1941711 (2021). [DOI] [PMC free article] [PubMed]

[CR51] 51.Kennedy, E. A., King, K. Y. & Baldridge, M. T. Mouse microbiota models: comparing germ-free mice and antibiotics treatment as tools for modifying gut bacteria. Front. Physiol.9, 1534 (2018). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR52] 52.Shannon, P. et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res.13, 2498–2504 (2003). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR53] 53.Biernat, K. A. et al. Structure, function, and inhibition of drug reactivating human gut microbial β-glucuronidases. Sci. Rep. 9, 825 (2019). [DOI] [PMC free article] [PubMed]

[CR54] 54.Bittinger, K. et al. Bacterial colonization reprograms the neonatal gut metabolome. Nat. Microbiol.5, 838–847 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR55] 55.Rimal, B. et al. Bile salt hydrolase catalyses formation of amine-conjugated bile acids. Nature626, 859–863 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR56] 56.Wang, M. et al. Mass spectrometry searches using MASST. Nat. Biotechnol.38, 23–26 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR57] 57.Jarmusch, A. K. et al. ReDU: a framework to find and reanalyze public mass spectrometry data. Nat. Methods17, 901–904 (2020). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR58] 58.El Abiead, Y. et al. Enabling pan-repository reanalysis for big data science of public metabolomics data. Nat. Commun.16, 4838 (2025). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR59] 59.Śliżewska, K. et al. Comparison of the activity of fecal enzymes and concentration of SCFA in healthy and overweight children. Nutrients15, 987 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR60] 60.Connell, E. et al. Microbial-derived metabolites as a risk factor of age-related cognitive decline and dementia. Mol. Neurodegener.17, 43 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR61] 61.Li, P. et al. Vancomycin relieves tacrolimus-induced hyperglycemia by eliminating gut bacterial beta-glucuronidase enzyme activity. Gut Microbes16, 2310277 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR62] 62.Zhang, L. & Chu, C.-Q. Gut microbiota-medication interaction in rheumatic diseases. Front. Immunol.12, 796865 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR63] 63.Tian, Y. et al. Early life polychlorinated biphenyl 126 exposure disrupts gut Microbiota and metabolic homeostasis in mice fed with high-fat diet in adulthood. Metabolites12, 894 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR64] 64.Tsiara, I. et al. Characterizing the sulfated and glucuronidated (poly)phenol metabolome for dietary biomarker discovery. J. Agric. Food Chem. 10.1021/acs.jafc.4c12596 (2025). [DOI] [PMC free article] [PubMed]

[CR65] 65.Ho, A. et al. Circulating glucuronic acid predicts healthspan and longevity in humans and mice. Aging11, 7694 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR66] 66.Hu, S. et al. Gut microbial beta-glucuronidase: a vital regulator in female estrogen metabolism. Gut Microbes15, 2236749 (2023). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR67] 67.Leonard, L. M., Simpson, A. M. R., Li, S., Reddivari, L. & Cross, T.-W. L. A Gnotobiotic mouse model with divergent equol-producing phenotypes: potential for determining microbial-driven health impacts of soy isoflavone daidzein. Nutrients16, 1079 (2024). [DOI] [PMC free article] [PubMed]

[CR68] 68.Letertre, M. P. M. et al. Characterizing the metabolic effects of the selective inhibition of gut microbial β-glucuronidases in mice. Sci. Rep.12, 17435 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR69] 69.Oto, J. et al. Validation of urine p-cresol glucuronide as renal cell carcinoma non-invasive biomarker. J. Proteom.311, 105357 (2024). [DOI] [PubMed] [Google Scholar]

[CR70] 70.Mykkänen, H., Tikka, J., Pitkänen, T. & Hänninen, O. Fecal bacterial enzyme activities in infants increase with age and adoption of adult-type diet. J. Pediatr. Gastroenterol. Nutr.25, 312–316 (1997). [DOI] [PubMed] [Google Scholar]

[CR71] 71.Elmassry, M. M., Kim, S. & Busby, B. Predicting drug-metagenome interactions: Variation in the microbial β-glucuronidase level in the human gut metagenomes. PLoS ONE16, e0244876 (2021). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR72] 72.Candeliere, F. et al. -Glucuronidase pattern predicted from gut metagenomes indicates potentially diversified pharmacomicrobiomics. Front. Microbiol.13, 826994 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR73] 73.Vargas, F. et al. Protocol for community-created public MS/MS reference spectra within the Global Natural Products Social Molecular Networking infrastructure. Rapid Commun. Mass Spectrom.34, e8725 (2020). [DOI] [PubMed] [Google Scholar]

[CR74] 74.Aron, A. T. et al. Reproducible molecular networking of untargeted mass spectrometry data using GNPS. Nat. Protoc.15, 1954–1991 (2020). [DOI] [PubMed] [Google Scholar]

[CR75] 75.Chambers, M. C. et al. A cross-platform toolkit for mass spectrometry and proteomics. Nat. Biotechnol.30, 918–920 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR76] 76.R Core Team. R: A Language and Environment for Statistical Computing, https://www.R-project.org/ (2025).

[CR77] 77.Wickham, H. ggplot2: Elegant Graphics for Data Analysis, https://ggplot2.tidyverse.org (2016).

[CR78] 78.Oksanen, J. et al. vegan: Community Ecology Package, https://CRAN.R-project.org/package=vegan (2025).

[CR79] 79.Paradis, E. & Schliep, K. ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics35, 526–528 (2019). [DOI] [PubMed]

[CR80] 80.Acker, D. gg3D: 3D perspective plots for ggplot2, https://github.com/AckerDWM/gg3D (2024).

[CR81] 81.Signorell, A. DescTools: Tools for Descriptive Statistics, https://CRAN.R-project.org/package=DescTools (2025).

[CR82] 82.Coelho, L. P. et al. NG-meta-profiler: fast processing of metagenomes using NGLess, a domain-specific language. Microbiome7, 84 (2019). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR83] 83.Hyatt, D. et al. Prodigal: prokaryotic gene recognition and translation initiation site identification. BMC Bioinform.11, 119 (2010). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR84] 84.Li, W. & Godzik, A. Cd-hit: a fast program for clustering and comparing large sets of protein or nucleotide sequences. Bioinformatics22, 1658–1659 (2006). [DOI] [PubMed] [Google Scholar]

[CR85] 85.Sievers, F. & Higgins, D. G. Clustal Omega, accurate alignment of very large numbers of sequences. Methods Mol. Biol.1079, 105–116 (2014). [DOI] [PubMed] [Google Scholar]

[CR86] 86.Buchfink, B., Xie, C. & Huson, D. H. Fast and sensitive protein alignment using DIAMOND. Nat. Methods12, 59–60 (2015). [DOI] [PubMed] [Google Scholar]

[CR87] 87.Kieser, S., Zdobnov, E. M. & Trajkovski, M. Comprehensive mouse microbiota genome catalog reveals major difference to its human counterpart. PLoS Comput. Biol.18, e1009947 (2022). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR88] 88.Langmead, B. & Salzberg, S. L. Fast gapped-read alignment with Bowtie 2. Nat. Methods9, 357–359 (2012). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR89] 89.Liao, Y., Smyth, G. K. & Shi, W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics30, 923–930 (2014). [DOI] [PubMed] [Google Scholar]

[CR90] 90.Simpson, J. B. et al. Metagenomics combined with activity-based proteomics point to gut bacterial enzymes that reactivate mycophenolate. Gut Microbes14, 2107289 (2022). [DOI] [PMC free article] [PubMed]

[CR91] 91.Batsoyol, N., Pullman, B., Wang, M., Bandeira, N. & Swanson, S. P-massive: a real-time search engine for a multi-terabyte mass spectrometry database. In Proc. SC22: International Conference for High Performance Computing, Networking, Storage and Analysis, 1–15 (IEEE, 2022).

[CR92] 92.Zuffa, S. et al. microbeMASST: a taxonomically informed mass spectrometry search tool for microbial metabolomics data. Nat. Microbiol.9, 336–345 (2024). [DOI] [PMC free article] [PubMed] [Google Scholar]

[CR93] 93.Huber, F. et al. Matchms - processing and similarity evaluation of mass spectrometry data. J. Open Source Softw.5, 2411 (2020). [Google Scholar]

PERMALINK

Glucuronidation metabolomic fingerprinting to map host-microbe metabolism

Nina R Boyle

Josh J Sekela

Mingxun Wang

Helena Mannochio-Russo

Jeong Joo Pyo

Min Soo Kim

Shuchang Tian

Imhoi Koo

Elliot S Friedman

Ceylan Tanes

Mallappa Anitha

Yuan Tian

Ethan W Morgan

Iain A Murray

Joseph P Zackular

Kyle Bittinger

James D Lewis

Gary H Perdew

Gary D Wu

Babette S Zemel

Pieter C Dorrestein

Jordan E Bisanz

Matthew R Redinbo

Andrew D Patterson

Abstract

Introduction

Results

Glucuronidated metabolite annotation

Fig. 1. Detecting and identifying glucuronidated features from untargeted LC-MS/MS.

Validation of glucuronidated metabolite mapping

Fig. 2. Acetaminophen effects on the glucuronidome of mice.

Glucuronidome shifts with GUSome manipulation in mice

Fig. 3. Conventional FMT shifts the glucuronidome from the colon to the serum and urine.

Fig. 4. Low-dose oral antibiotics shift the urine, colonic, and serum glucuronidomes.

Fig. 5. Annotating glucuronidated features through molecular networking.

Microbial GUS deconjugation of xenobiotic glucuronidated metabolites

Fig. 6. Microbial GUS-driven glucuronidated and aglycone metabolite shifts.

Glucuronidome diversifies with microbiota colonization in infants

Fig. 7. Urinary glucuronidated metabolites and GUS gene features across infancy.

Phenotypic associations of GlcAs

Discussion

Materials and methods

Chemicals and reagents

Spectral library generation

Classical molecular networking (chemical standards)

Mass query language (MassQL) query development

False discovery analysis

Animals

APAP study

ABX study

Germ-free and FMT study

High-fat diet study

Metabolite extraction and sample preparation

Standards and LC-MS/MS data acquisition

File conversion

Glucuronidome analyses

Acetaminophen metabolite feature-based molecular networking (FBMN)

Classical molecular networking analysis of urine, colonic contents, and serum

MassQL pattern filtering of glucuronidated features

Microbial drop plates (GF/FMT)

Gut microbiome DNA extraction and sequencing

Microbial metagenomic analysis

Initial processing of mouse metagenomic samples

Identification and characterization of GUS gene sequences from mouse gut metagenomic data

Metagenomic gene intensity quantitation

GUS gene intensity analysis

Microbial GUS assays and LC-MS/MS

Microbial deconjugation assays and LC-MS/MS

IGRAM cohort, LC-MS/MS and metagenomics

Glucuronidome repository searches

Statistics and reproducibility

Reporting summary

Supplementary information

Acknowledgements

Author contributions

Peer review

Peer review information

Data availability