Authors: Gavin A. Kuziel (1Division of Infectious Diseases, Boston Children’s Hospital, Boston, MA, 02115 USA; 2Division of Gastroenterology, Boston Children’s Hospital, Boston, MA, 02115 USA; 6Department of Pediatrics, Harvard Medical School, Boston, MA, 02115 USA; 7Department of Microbiology, Harvard Medical School, Boston, MA, 02115 USA; 12These authors contributed equally), Gabriel L. Lozano (1Division of Infectious Diseases, Boston Children’s Hospital, Boston, MA, 02115 USA; 2Division of Gastroenterology, Boston Children’s Hospital, Boston, MA, 02115 USA; 6Department of Pediatrics, Harvard Medical School, Boston, MA, 02115 USA; 7Department of Microbiology, Harvard Medical School, Boston, MA, 02115 USA; 12These authors contributed equally), Corina Simian (8Whitehead Institute for Biomedical Research, Cambridge, MA, 02142 USA; 9Department of Chemistry and Chemical Biology & Department of Bioengineering, Northeastern University, Boston, MA, 02120 USA; 10Institute for Plant-Human Interface, Northeastern University, Boston, MA, 02120 USA), Long Li (1Division of Infectious Diseases, Boston Children’s Hospital, Boston, MA, 02115 USA; 2Division of Gastroenterology, Boston Children’s Hospital, Boston, MA, 02115 USA; 6Department of Pediatrics, Harvard Medical School, Boston, MA, 02115 USA; 7Department of Microbiology, Harvard Medical School, Boston, MA, 02115 USA), John Manion (4Department of Urology, Boston Children’s Hospital, Boston, MA, 02115 USA; 5Department of Surgery, Harvard Medical School, Boston, MA, 02115 USA; 6Department of Pediatrics, Harvard Medical School, Boston, MA, 02115 USA), Emmanuel Stephen-Victor (3Division of Immunology, Boston Children’s Hospital, Boston, MA, 02115 USA; 6Department of Pediatrics, Harvard Medical School, Boston, MA, 02115 USA), Talal Chatila (3Division of Immunology, Boston Children’s Hospital, Boston, MA, 02115 USA; 6Department of Pediatrics, Harvard Medical School, Boston, MA, 02115 USA), Min Dong (4Department of Urology, Boston Children’s Hospital, Boston, MA, 02115 USA; 5Department of Surgery, Harvard Medical School, Boston, MA, 02115 USA; 6Department of Pediatrics, Harvard Medical School, Boston, MA, 02115 USA), Jing-Ke Weng (8Whitehead Institute for Biomedical Research, Cambridge, MA, 02142 USA; 9Department of Chemistry and Chemical Biology & Department of Bioengineering, Northeastern University, Boston, MA, 02120 USA; 10Institute for Plant-Human Interface, Northeastern University, Boston, MA, 02120 USA), Seth Rakoff-Nahoum (1Division of Infectious Diseases, Boston Children’s Hospital, Boston, MA, 02115 USA; 2Division of Gastroenterology, Boston Children’s Hospital, Boston, MA, 02115 USA; 6Department of Pediatrics, Harvard Medical School, Boston, MA, 02115 USA; 7Department of Microbiology, Harvard Medical School, Boston, MA, 02115 USA; 11Broad Institute, Cambridge, MA, 02139 USA; 13Lead contact)
Categories: Article, Gut microbiome, diet, plant small molecules, phytochemicals, phenolic glycosides, polyphenols, glycoside hydrolase, colonization resistance, Clostridioides difficile, anti-inflammatory, colitis
Source: Cell
Authors: Gavin A. Kuziel, Gabriel L. Lozano, Corina Simian, Long Li, John Manion, Emmanuel Stephen-Victor, Talal Chatila, Min Dong, Jing-Ke Weng, Seth Rakoff-Nahoum
Plants are composed of diverse secondary metabolites (PSMs) which are widely associated with human health. Whether and how the gut microbiome mediates such impacts of PSMs are poorly understood. Here we show that discrete dietary and medicinal phenolic glycosides, abundant health-associated PSMs, are utilized by distinct members of the human gut microbiome. Within the Bacteroides, the predominant Gram-negative bacteria of the Western human gut, we reveal a specialized multi-enzyme system dedicated to the processing of distinct glycosides based on structural differences in phenolic moieties. This Bacteroides metabolic system liberates chemically distinct aglycones with diverse biological functions such as colonization resistance against the gut pathogen Clostridioides difficile via anti-microbial activation of polydatin to the stilbene resveratrol and intestinal homeostasis via activation of salicin to the immunoregulatory aglycone saligenin. Together our results demonstrate generation of biological diversity of phenolic aglycone “effector” functions by a distinct gut-microbiome-encoded PSM-processing system.
Diet is a critical determinant of human health^1^. In particular, plant-based diets have strong epidemiological associations with diverse health outcomes ranging from cardiovascular disease^2,3^, inflammatory disease and autoimmunity^4,5^ to the prevention of cancer and neurodegeneration^6–8^. In addition, hundreds of years of the use of medicinal plants across human cultures have provided both empirical support for the impact of plants on human health^9^ and inspired pharmaceutical and synthetic chemical approaches for prevention and treatment of disease^10^.
The gut microbiome has emerged as central to mediating the impact of diet, in particular, plant diets, across a broad spectrum of human disease^11–13^. Perhaps, the best appreciated mechanism by which the microbiome mediates the impact of plants on the host is the generation of short chain fatty acids, the products of gut microbiome fermentation of plant glycans^14–16^. In this process, plant polysaccharides are broken down into monosaccharides by a spectrum of microbiome-encoded metabolic enzymes such as glycoside hydrolases (GHs)^17–19^. These monosaccharides enter glycolytic pathways creating energy for the gut microbiome. As a consequence, short chain fatty acids (such as acetate, propionate, and butyrate) are generated as waste products. Such short chain fatty acids serve as the effectors of microbiome metabolism of plant polysaccharides on the host.
In addition to polysaccharides, plants contain a variety of secondary metabolites (PSMs). These PSMs are chemically and functionally diverse and abundant across dietary and medicinal plants^20^. PSMs play critical roles in plant survival, ranging from physiological roles in development and endocrine function^21^ to mediating plant-animal and plant-microbe interactions such as protection from herbivorous predators and bacterial and fungal pathogens^22^ via toxic^23^, neuroactive^24^ and anti-microbial^25^ activities. Importantly, many PSMs are glycosylated both to tightly regulate functions within the plant during storage and to enhance solubility^26^. These PSM-carbohydrate conjugates, known as glycosides, represent the extensive chemical diversity of PSMs in the form of the aglycone moiety of each glycoside, and encompass broad plant biosynthetic chemical diversity across monoaryl and polyphenols, terpenoids, and alkaloids^20^. Epidemiological studies support substantial benefits associated with the consumption of plant glycosides and human health and disease^27,28^. Phenolic glycosides demonstrate both chemical diversity ranging from monophenolic to polyphenolic aglycones, the chemical class of aglycone (‘simple’ aryls, coumarins, flavonoids [dihydrochalcones], non-flavonoids [stilbenes], etc.), linkage to monosaccharides, disaccharides, or oligosaccharides, and stereochemistry (α vs. β linkage) and site of linkage (α1, β2, etc.)^29,30^, are abundant in fruits, vegetables, and nuts^31,32^ consumed by humans over evolutionary time and used in ethnobotanical cultural practice^29,30,33^. Dietary interventions in human clinical trials and animal models have demonstrated roles of plant glycosides on health outcomes across inflammatory disease^34^, cancer^35^, cardiovascular disease,^36^ and neurodegenerative disorders^37^. However, the specific functions of plant glycosides, in particular, phenolic glycosides, on the host and both whether and how the gut microbiome mediates such impacts of PSMs in human health are poorly understood.
Here, we sought to define the capacity of the human gut microbiome to metabolize plant phenolic glycosides. Utilizing a multi-disciplinary approach spanning microbiology, bacterial genetics, biochemistry, enzymology, gnotobiotics, and mouse models of disease, we reveal the microbe-specific metabolism of distinct phenolic glycosides, the evolution of specialized enzymes among the human gut Bacteroides dedicated to processing distinct phenolic glycosides, and the functional diversification of plant glycosides by this system in vivo to mediate colonization resistance and intestinal immune homeostasis.
To define the capacity of the human gut microbiome to metabolize PSM glycosides, we assayed the ability of a taxonomically broad and representative panel of the membership of the human gut microbiome to utilize glycosides as energy sources serving as a scalable proxy for metabolic activity. To begin to capture the functional capacity of the gut microbiome to metabolize the broad chemical range of dietary and medicinal glycosides, we started with a panel of simple aryl, cyanogenic, or coumarin glycosides. Importantly, while each glycoside in this panel is bound to the same sugar (D-glucose) via the same linkage (β1) (Figure 1A), the aglycones of each compound have subtle yet distinct chemical modifications, such as carbon length of the alcohol group (arbutin vs. gastrodin), functionalization by aldehyde vs. alcohol group (salicin vs. helicin), substitution of aromatic methanol modification (ortho-salicin vs. para-gastrodin), or aglycone linker length (arbutin vs. salidroside) (Figure 1A). Members of the Enterobacteriaceae such as Escherichia coli, Klebsiella pneumoniae, Citrobacter portucalensis (C. freundii complex) and Salmonella enterica, and other facultative gut anaerobes such as Enterococcus faecalis and faecium, demonstrated either broad utilization of each glycoside as “generalists” (K. pneumoniae, E. faecalis, and E. faecium; Figure 1B, orange color) or unable to use any (E. coli, C. portucalensis, and S. enterica; Figure 1B, navy color), consistent with previous studies reporting absent or cryptic metabolism of glycosides in these taxa^38,39^.
Gut Clostridia, such as Lactobacilli, Hungatella hathewayi, Clostridium symbiosum, C. scindens, P. hiranononis, C. cadaveris, C. sporogenes and Coprobacillus cateniformis, and the gut Actinobacteria (Bifidobacterium longum) demonstrated generalist (used all or most) or inability/cryptic utilization across the panel of glycosides (Figure 1B, green color). However, among other gut Clostridia and Erysipelotrichia members representative of taxonomic diversity across Lachnospiraceae, Clostridiaceae, and Erysipelotrichales families, we observed unique patterns of glycoside utilization. Ruminococcus gnavus demonstrated comparable growth on specific glycosides (arbutin, gastrodin, salicin, salidroside) but not amygdalin and demonstrated strain-level variation (R. gnavus 2_1_58FAA vs. CC55_001C) in esculin utilization (pink color text; Figure 1B). Other members of gut Clostridia demonstrated unexpected specialization for specific glycosides (Figure 1B, teal color). Longicatena innocuum specifically utilized arbutin, esculin, helicin and salicin but demonstrated absence of growth in salidroside or amygdalin (Figure 1B). Blautia massiliensis demonstrated unique specialization (enhanced growth compared to other aryl glycosides) for arbutin, while Faecalimonas umbilicata utilized arbutin and salicin, but not other aryl glycosides.
Among members of the human gut Bacteroidales, the predominant Gram-negative bacteria of the human gut microbiome^40,41^, we observed unexpected variation in glycoside utilization. B. fragilis S36 L11, B. eggerthii DSM, B. vulgatus ATCC, and B. dorei 9_1_42FAA demonstrated inability/cryptic utilization of each glycoside (Figure 1B, navy color). B. ovatus CL03T12C18 and each member of Parabacteroides species, P. goldsteinii CL02T12C30, P. distasonis 31_2, P. johnsonii CL02T12C29, and P. merdae CL03T12C32, exhibited broad utilization of each glycoside (except for helicin, for which we identified only H. hathewayi, C. tertium, C. innocuum, and Enterococci as able to utilize this glycoside; Figure 1B, orange color). Comparatively, we find Bacteroides members exhibiting unique isolate-specific specialization for specific glycosides (Figure 1B, teal color). B. salyersiae DSM and B. nordii CL02T12C05 utilized only esculin and amygdalin, while B. finegoldii DSM exclusively utilized esculin (Figure 1B). In contrast, B. uniformis ATCC demonstrated robust growth in the aryl glycosides (arbutin, salicin, gastrodin and salidroside) but limited or no growth for those of greater complexity (esculin or amygdalin). B. caccae ATCC utilized all simple glycosides (excluding helicin, as above) but not arbutin. Together, our data demonstrate unexpected variation in utilization of glycosides among the Clostridia, Erysipelotrichia, and Bacteroidales of the human gut microbiome whereby certain members demonstrate “all-or-none” utilization of glycosides while others differentially utilize structurally similar PSMs varying by distinct modifications of phenolic aglycones.
Given the abundance of the gut Bacteroidales in the human gut microbiome^40,41^, the links between plant diets and the Bacteroides in health^13,42,43^, and our findings demonstrating that the Bacteroidales differentially utilize specific health-relevant plant-derived glycosides, we sought to more deeply investigate phenolic glycoside utilization by the gut Bacteriodales. Defining the functional capacity of 52 Bacteroides and Parabacteroides strains across 18 species to utilize our panel of simple glycosides, we reveal extensive strain-level and aglycone modification-level variation in glycoside utilization (Figure 2A). In general, we observe similar patterns of glycoside utilization among strains of the same species. For example, each strain of the species B. ovatus, B. xylanisolvens and across Parabacteroides species were generalists utilizing each of the panel of simple glycosides. Similarly, we find that inability to utilize any of the glycosides was a species-wide trait of B. vulgatus and B. dorei (Figure 2A). However, among other Bacteroides species we find strain-level variation for distinct glycosides. For example, while each strain of B. thetaiotaomicron specialized in esculin and amygdalin utilization, we observed strain-level variation in the dynamics of growth during salicin utilization (Bt 1_1_6 vs. Bt VPI) (Figure 2B). All strains of B. fragilis were universally unable to utilize phenolic glycosides with the exception of B. fragilis CL05T00C42, which specifically utilized esculin (Figure 2C). Both B. uniformis ATCC and B. uniformis D20 specialized in utilization of aryl glycosides, but B. uniformis CL03T00C23 did not grow on any phenolic glycoside (Figure 2D). Similarly, B. caccae ATCC grew on most glycosides, but B. caccae CL03T12C61 was unable to utilize any (Figure 2E). Together our results reveal extensive strain-level and aglycone modification-level variation in glycoside utilization among the human gut Bacteroides.
Across the Bacteroides, we reveal unique patterns of aryl glycosides utilization, suggesting differential mechanisms for the metabolism of health-related plant small molecules. We thus next sought to define the mechanistic basis for aryl glycoside utilization among the Bacteroides. We began with a forward genetic approach performing transposon insertion mutagenesis coupled with next generational sequencing (Tn-Seq) to identify the genetic basis for aryl glycoside utilization in B. ovatus ATCC, which we identify as demonstrating broad generalist utilization of these substrates. Identifying Tn insertions resulting in loss-of-function in arbutin compared with glucose identified 67 genes with log2fold-change >= 1 and p-value < 0.05. Among these we find insertions in each gene of a five gene operon conferring specific fitness disadvantage in arbutin, Bovatus_02231–Bovatus_02227 (Figure 3A). This operon is composed of a transcriptional regulator (Bovatus_02231), two nicotinamide-dependent oxidoreductases (Bovatus_02230 and Bovatus_02229), a protein with structural homology^44^ to the class of family-16-like GH (Bovatus_02228), and a sugar-phosphate isomerase (Bovatus_02227) (Figure 3B). We generated isogenic strains with non-polar deletions of each gene of this operon. Deletion of each gene in this operon was defective in utilization of each aryl glycoside (Figure 3C) and restored by complementation (Figure S1A). To date, GHs responsible for hydrolysis of monosaccharide-aglycone linkages within the human gut microbiome have been demonstrated to hydrolyze monosaccharide-monosaccharide linkages for disaccharide or polysaccharide utilization^45^, the latter thought to be the primary evolved substrate of these enzymes^46^. Indeed, we find that Bovatus_02228 and each gene in the Bovatus_02231–02227 operon (except Bovatus_02227, the sugar-phosphate isomerase) is required for utilization of the disaccharides trehalose (α1,1 glucose-glucose homodimer) and palatinose (α1,6 glucose-fructose heterodimer) (Figures 3D, S1B,C), consistent with a recent report demonstrating activity of a homolog of the GH16 family member Bovatus_02228 for both glucosinolates and disaccharides^47^. Together, we conclude that the aryl glycoside generalist of the gut Bacteroides, B. ovatus, employs a single operonic generalist system (we name the operon glycoside generalist hydrolase, ggh, composed of gghR, Bovatus_02231; gghA, Bovatus_02230; gghB, Bovatus_02229; gghC, Bo_02228; gghD, Bo_02227) to utilize both β1-linked aryl glycosides and α1-linked disaccharides.
We next sought to determine the genetic basis for aryl glycoside utilization in the glycoside specialist B. uniformis. Compared to B. ovatus, which utilizes aryl, coumarin, and cyanogenic glycosides, B. uniformis both specializes in aryl glycosides and notably demonstrates greater growth yield in aryl glycosides than in glucose (Figures 2A,2D). Leveraging the same approach as in B. ovatus, we constructed a genome-wide transposon mutagenesis library of B uniformis and performed Tn-seq after selection in glucose vs. arbutin (Figure 3E). Transposon insertions in each gene of an operon composed of four genes (BACUNI_00922 – BACUNI_00919), distinct from that in B. ovatus, conferred loss of fitness in arbutin, for which BACUNI_00919, encoding a GH3 family member, was the most statistically significant loss-of-fitness Tn insertion (Figure 3E). This operon, which we name glycoside specialist hydrolase gsh, is composed of esterases (gshA, BACUNI_00922; gshB, BACUNI_00921; gshC, BACUNI_00920) and a GH3 family member (gshD, BACUNI_00919) (Figure 3F). To determine the requirement of each gene in the B. uniformis gsh operon for aryl glycoside utilization, we generated non-polar deletions for each gene of the operon. As opposed to the ggh generalist utilization locus of B. ovatus, deletion of only the GH3 gshD in the B. uniformis gsh operon resulted in impaired glycoside utilization (Figures S2A,B). Importantly, compared to the complete defect in aryl glycoside utilization upon deletion of the GH16 gghC in B. ovatus, deletion of the GH3 gshD in B. uniformis resulted in only partially defective growth in aryl glycosides (Figure S2A), suggesting that as opposed to the generalist B. ovatus, B. uniformis employs multiple GHs to utilize these substrates.
To identify genes acting in concert with gshD to utilize aryl glycosides in the specialist B. uniformis, we performed a synthetic TnSeq on a Bu ∆gshD background under arbutin selection. This screen identified Tn insertions in two genes, BACUNI_01042 (gshG, GH3 family member) and BACUNI_0952 (gghB, a homolog of the nicotinamide-dependent oxidoreductases identified in B. ovatus ATCC directly upstream of a homolog of Bo GH16 gghC) (Figure 3G, Table S2). We next generated a series of single-, double- and triple-gene deletion strains lacking combinations of gshD, gshG, and gghC. Deletion of Bu GH3 gshG or Bu GH16 gghC alone showed no defects in utilization of aryl glycosides, demonstrating the synergistic requirement of the GH3 gshD for aryl glycoside utilization (Figure S2C).
Comparing the utilization of each aryl glycoside of combinations of gsh GH3 gene deletions in B. uniformis revealed striking findings. Deletion of both Bu GH3 gshD with Bu GH3 gshG (Bu ∆gshD∆gshG) demonstrated impaired growth on arbutin compared to that of Bu ∆gshD alone (Figure 3H) but notably exhibited residual utilization of this aryl glycoside. In comparison, deletion of Bu GH3 gshD and the Bu GH16 gghC (Bu ∆gshD∆gghC) demonstrated a complete defect in arbutin utilization demonstrating the requirement of a three GH enzymatic system for the utilization of arbutin in B. uniformis, with a greater contribution of Bu GH16 gghC than Bu GH3 gshG for arbutin utilization (Figure 3H). Similar to arbutin, this three GH enzyme system was required for utilization of gastrodin, which is functionalized with a para-hydroxymethylation vs. para-hydroxylation as in arbutin. Compared with the three GH enzyme requirement for the utilization of arbutin and gastrodin, Bu GH16 gghC but not Bu GH3 gshG was required to utilize salicin. However, for B. uniformis utilization of salidroside, we see the opposite, as Bu GH3 gshG but not Bu GH16 gghC is required with Bu GH3 gshD for utilization of this aryl glycoside. Together, our findings demonstrate that while B. ovatus uses one GH to utilize aryl glycosides, B. uniformis employs unique combinations of GH3 and GH16 to utilize specific PSM glycosides.
The requirement for pairs of GH enzymes to utilize distinct aryl glycosides in the glycoside specialist B. uniformis suggested that Bu GH3 enzymes may be specific and/or dedicated for distinct phenolic glycosides. The requirement of Bu GH3 gshG for salidroside but not salicin utilization suggested specialization of this enzyme for specific glycosides. To test this hypothesis, we first over-expressed Bu gshG in B. vulgatus, which is unable to grow on either salidroside or salicin (Figures 2A and S3A). Over-expression of Bu GH3 gshG resulted in enhanced growth rate and yield on salidroside compared with salicin, that latter used as a substrate but to a lesser degree (Figure 4A). This suggested that specificity for hydrolysis of salidroside by Bu GH3 gshG may be encoded via the unique aglycone of this phenolic glycoside vs. that of salicin. To definitively test the substrate specificity of Bu GH3 gshG, we purified recombinant Bu GH3 GshG (Figure S3B) and measured hydrolytic activity for salicin and salidroside. Bu GH3 GshG had 10-fold greater catalytic turnover of salidroside compared with salicin (Figure 4B). These data reveal that the GH3 GshG of B. uniformis demonstrates substrate specificity based on discrete structural differences of the aglycone moiety of the glycosides.
For the GH16 family member gghC (conserved between Bo and Bu), we demonstrated roles in utilization by B. ovatus for both aryl glycosides and disaccharides (trehalose and palatinose), demonstrating the broad hydrolytic activity of this GH16 member for PSM glycosides and disaccharides (Figures 3C and 3D). Therefore, we next sought to determine whether B. uniformis GH3 (required for aryl glycoside utilization) were also required to utilize disaccharides. We find Bu ∆gshG partially defective in utilization of gentiobiose, demonstrating the requirement of Bu GH3 gshG for utilization of this β1,6 glucose-glucose homodimeric disaccharide (Figures 4C and S3C). Purified recombinant Bu GH3 GshG protein hydrolyzed both gentiobiose and the β1,4 glucose-glucose homodimeric disaccharide cellobiose, with comparable efficiency to hydrolysis of salidroside (Figure 4D). These data demonstrate that Bu GH3 gshG enables the utilization of both the aryl glycosides salidroside, arbutin and gastrodin (and to a lesser extent salicin) and the disaccharide gentiobiose and thus hydrolyzes both aryl glycosides and disaccharides.
We next sought to determine the specificity of Bu GH3 GshD, which is required for utilization of each aryl glycoside by B. uniformis (Figure 3H). Overexpression of Bu GH3 gshD in B. vulgatus conferred similar capacity to utilize each of these phenolic glycosides (Figure 4E), compared to the salidroside specificity of Bu GH3 gshG (Figure 4A). As opposed to Bu GH3 gshG, absence of Bu GH3 gshD did not affect utilization of disaccharides by B. uniformis (Figure S3C). This suggested that Bu GH3 gshD may be dedicated to the metabolism of PSM glycosides. Indeed, we find Bu GH3 gshD is a part of a unique clade of GH3 enzymes among gut bacteria, both distinct from Bu GshG (Figure S4D) and present in B. uniformis and P. distasonis among the Bacteroidales (Figure S3E). To definitively test the substrate specificity of Bu GH3 gshD, we purified Bu GH3 GshD (Figure S3F) and measured hydrolytic activity for salicin and salidroside and a range of disaccharides. Bu GH3 GshD had comparable hydrolytic activity for both salicin and salidroside (Figure 4F) but did not hydrolyze any disaccharide (Figure 4G). Together, we demonstrate that B. uniformis contains GH3 enzymes with substrate specificity dependent on discrete structural differences of each substrate’s aglycone moiety (Bu GH3 GshG), and potentially dedicated to the hydrolysis of PSM glycosides (Bu GH3 GshD).
Our data have demonstrated the utilization and hydrolysis of aryl glycosides by a specialized metabolic system in the predominant human gut microbiome member B. uniformis^48,49^. B. uniformis GH enzymes responsible for aryl glycoside metabolism are predicted to localize to the outer membrane^50^. Thus, we next asked if hydrolysis of aryl glycosides by B. uniformis resulted in the liberation and extracellular availability of the range of aryl aglycones. To test this, we measured the presence of distinct aglycones by liquid chromatography mass spectrometry (LC-MS) in supernatant after growth of B. uniformis in media supplemented with each aryl glycoside. For each aryl glycoside, arbutin, salidroside, gastrodin, helicin and salicin, B. uniformis liberated each of the corresponding aglycones, hydroquinone, gastrodigenin, tyrosol, salicylaldehyde and saligenin, respectively (Figure 5A).
We next asked both whether B. uniformis hydrolyzes and liberates aglycones from a broader range of PSMs extending our studies to polyphenolic glycosides, both abundant in diet and associated with wide-ranging impacts across health and disease^51,52^. To address this, we generated a panel of polyphenolic glycosides which, similar to our panel of aryl glycosides, aglycones are bound to glucose via β1-linkages, but for which for each aglycone represents broad chemical diversity spanning flavonoids (flavonols, rutin; flavanones, naringin; flavones, cynaroside; isoflavones, genistein and daidzein), dihydrochalcones (phloridizin), stilbenes (polydatin) and lignans (pinoresinol diglucoside) (Figure 5B). Using a targeted LCMS approach, we found that across each polyphenolic glycoside, B. uniformis liberated the predicted aglycone deglycosylation product from each polyphenolic glycoside substrate (Figure 5C). Together, these data demonstrate that B. uniformis metabolism of aryl glycosides result in the liberation of diverse mono- and polyphenolic aglycones.
Having defined the broad capacity of the plant glycoside specialist B. uniformis to both metabolize a diverse array of phenolic glycosides and liberate chemically diverse aglycones, we next sought to determine potential functions of such plant small molecules on microbe-microbe and microbe-host functions in the mammalian gut. In plants, deglycosylation of phenolic glycosides by plant exo-glycosidases results in the liberation of aglycones, which play diverse and important functions in defense from microbial pathogens^25,53,54^. Thus, we asked whether B. uniformis liberation of diverse aglycones from distinct glycosides activated anti-microbial functions against human gut pathogens. To test this, we determined the fitness effects of aryl or polyphenolic glycosides and each corresponding aglycone liberated by B. uniformis (Figure 6A) on a panel of human gut pathogens representing taxa of multidrug resistant (MDR) and opportunistic pathogens. This screen revealed distinct fitness effects on gut pathogens for specific polyphenolic glycosides such as the pro-microbial (growth enhancement) of the B. uniformis liberated aglycone of the isoflavone daidzin, daidzein across gut pathogens and anti-microbial effects of E. faecium by the liberated aglycone of the flavone cynaroside, luteolin (Figure 6A). Most notably, we reveal the polyphenolic aglycones phloretin and resveratrol liberated from the dihydrochalcone glycoside, phloridizin (abundant in apples^31^) and the stilbene glycoside, polydatin (abundant in grapes and in plants used in traditional medicine notably Polygonum cuspidatum; Japanese knotweed^55^), respectively, as potent inhibitors of Clostridioides difficile strain M7404 (Figure 6A). Neither the parent polyphenolic glycosides phloridzin or polydatin were anti-microbial against C. difficile (Figure 6A).
B. uniformis cultivated with polydatin liberated resveratrol (Figure 5C) and generated potent anti-C. difficile activity (Figure S4A), both functions ablated in B. uniformis ∆gshD (Figures S4B and S4C), demonstrating bioactivation of polydatin and liberation of resveratrol by B. uniformis dependent on the glycoside-specific GH3 Bu gshD. C. difficile infection (CDI) is a major cause of mortality and morbidity, most associated with defects in colonization resistance by the gut microbiome in the setting of use of antibiotics^56^. Across 20 unique C. difficile strains representing toxin production and hypervirulence, we find the B. uniformis-liberated polyphenolic aglycone resveratrol, but not its parent glycoside polydatin, as universally anti-microbial against C. difficile (Figure 6B). In a mouse model of CDI, the B. uniformis-generated aglycones resveratrol and phloretin demonstrated potent anti-microbial effects on C. difficile in vivo (Figure 6C) occurring within the context of a restored gut microbiome after antibiotics (Figures S4D and S4E). In addition, we do not find evidence of degradation of resveratrol (Figure S4F) nor metabolism to dihydroresveratrol by the mouse gut microbiome (Figure S4G), which can occur via reduction of resveratrol by specific members of the human gut microbiome^57,58^. Furthermore, we find that dihydroresveratrol maintains anti-microbial activity against C. difficile (Figure S4H). Together, our data reveal both a novel function of polyphenolic aglycones against the human gut pathogen C. difficile and the role of bioactivation of dietary plant glycosides by members of the gut microbiome in mediating a novel mechanism of colonization resistance to human gut pathogens mediated by microbiome bioactivation of specific phytochemicals.
Plant glycosides, which function in diverse roles in plants in homeostasis to stress^53,59^, have been postulated to play both beneficial and pathological role in human health, notably in inflammation and cancer^60,61^. However, whether and how the gut microbiome mediates such effects of plant glycosides on the host is poorly understood. Given the abundance of the Bacteroides in the human gut, in particular B. uniformis^48,49^, and our identification of functionalization of plant glycosides by liberation of diverse aglycones by this prominent member of the human gut microbiome, we next sought to identify potential functions of the biotransformation of PSM glycosides by the B. uniformis on inflammation. To test this, we assayed the function of aryl and polyphenolic glycosides and the corresponding B. uniformis-liberated aglycones of each glycoside to regulate the production of tumor necrosis factor (TNF) and interleukin-6 (IL-6) by lipopolysaccharide (LPS)-stimulated macrophages. The aryl glycosides salicin and salidroside and the polyphenolic glycosides rutin and naringin did not affect cytokine production by macrophages. In contrast, each of the B. uniformis-liberated saligenin, tyrosol, quercetin and naringenin, respectively, demonstrated broad anti-inflammatory function for both TNF and IL-6 (Figures 7A and 7B). The aryl glycoside arbutin was found to selectively negatively regulate TNF, but not IL-6, production, while the arbutin aglycone hydroquinone was cytotoxic to macrophages (Figure S5A). The polyphenolic glycoside polydatin enhanced IL-6, but not TNF, production from macrophages, whereby this specific pro-inflammatory function was converted to potent anti-inflammatory function when deglycosylated to the aglycone resveratrol (Figure 7B). While the aryl glycoside gastrodin did not impact macrophage inflammatory cytokine production, the B. uniformis-liberated aglycone of gastrodin, gastrodigenin, demonstrated opposing functions on macrophages, repressing TNF while enhancing IL-6 production (Figure 7A). Both the polyphenolic glycosides naringin and gastrodin had no effect on cytokine production. However, the B. uniformis-liberated aglycone, naringenin repressed IL-6 and enhanced TNF production by macrophages, while we observed the opposite effect (IL-6 enhancement and TNF repression by the B. uniformis-liberated aglycone of gastrodin, gastrodigenin (Figure 7B). We found that the polyphenolic glycoside phloridzin is inert to macrophages while the B. uniformis-liberated aglycone, phloretin, similarly does not affect TNF but suppresses IL-6 production from macrophages. Together, our findings reveal generation of remarkable diversity of inflammatory function of B. uniformis-liberated aglycones across diverse PSM glycosides.
The bioactivation of the anti-inflammatory effect of PSM glycosides such as salicin, salidroside, rutin, and naringin by B. uniformis suggested that anti-inflammatory functions of these dietary and medicinal plant glycosides^62^ may be mediated by metabolism and liberation of their respective bioactive aglycones by the gut microbiome. For example, the aryl glycoside salicin has long been known to have broad analgesic and anti-inflammatory effects across health and disease and is the basis for development of salicylic acid and acetylsalicylic acid as aspirin^63^. We found that the liberated aglycone saligenin, the product of salicin metabolism by B. uniformis, demonstrated potent inhibition of both TNF and IL-6 production by LPS-stimulated macrophages (Figures 7A and S5B). Saligenin liberation was dependent on the Bu glycoside metabolic system as conversion of salicin to saligenin in vitro was absent in Bu deficient in all three GH (GH∆∆∆; ∆gshD∆gshG∆gghC) mutants compared to Bu wild type (WT) (Figure S6A). Colonization of germ-free mice with either Bu WT or Bu GH∆∆∆ followed by oral administration of salicin demonstrated both Bu-dependent liberation of saligenin in the gut (as detected in feces) and dependence on the Bu glycoside metabolic system (Figure 7C).
We next sought to determine if B. uniformis metabolism of salicin into the anti-inflammatory aglycone saligenin impacted intestinal homeostasis in vivo. To address this, we employed a mouse model of intestinal inflammation, dextran sodium sulfate (DSS) in drinking water in specific pathogen-free (SPF) mice colonized with either Bu WT or Bu GH∆∆∆. Mice colonized with Bu WT and fed salicin were protected from weight loss, colon shortening, and leukocyte infiltration, compared to mice colonized with Bu WT without salicin or mice colonized with Bu GH∆∆∆ fed salicin (Figures 7D-7F) (in which Bu WT and GH∆∆∆ equally colonized mice; Figure S6B). Notably, colons from mice colonized with Bu WT fed salicin, compared to Bu GH∆∆∆ fed salicin were protected from DSS-induced accumulation of interferon (IFN)-γ-producing CD4^+^ T-cells and Tbet^+^ TH1 cells at the lamina propria, suggesting a role of liberation of saligenin from salicin by the Bu glycoside metabolic system in regulating pathogenic Th1 mediated intestinal inflammation (Figure S6C). Protection from colitis occurred in the setting of a restored microbiome after antibiotics (allowing B. uniformis administration) (Figures S6D and S6E) and without evidence of saligenin degradation by the mouse gut microbiome (Figure S6F). Administration of saligenin in antibiotic-naïve mice with intact microbiome diversity and composition (Figures S6D and S6G) protected mice from colitis (Figures S6H-S6J), demonstrating the direct role of saligenin in protection from colitis by B. uniformis metabolism of salicin. Similar to salicin, the closely related aryl glycoside arbutin is metabolized by the combination of Bu GH3 gshD and Bu GH16 gghC (Figure 3H). Importantly, however, liberation of the aglycone hydroquinone by Bu (Figure 5A) results in cell toxicity (Figure S5A) compared to the anti-inflammatory properties of the Bu-liberated saligenin from salicin (Figures 7A and S5B). Compared with mice fed salicin, dietary arbutin did not protect Bu WT-colonized mice from DSS-mediated weight loss, colon shortening or leucocyte inflammation (Figures 7E-7G and S6K). Together, our data demonstrate both the critical role of microbiome bioactivation of salicin to mediate the anti-inflammatory and intestinal immune homeostatic role in vivo of this ancient medicinal factor by the gut microbiome and the role of functional diversification of dietary glycosides mediated by the same gut microbial member and glycoside metabolic system on differential host outcomes.
In their capacity to generate energy to replicate and survive within the intestine, members of the human gut microbiome employ hundreds of diverse GHs ultimately transforming a multitude of chemically diverse polysaccharides into a limited number of host-active fermentation products (e.g., formate, acetate, propionate, lactate, succinate, butyrate). As such, the biosynthetic diversity and chemical complexity of plant metabolism to produce glycans is “funneled” whereby plant chemical diversity via a matched diversity of GHs is reduced to a limited set of effectors metabolites serving as the functional output of gut microbiome metabolism of plants on the host. Here we show that deglycosylation of diverse glycosides by a defined set of GHs encoded within the gut microbiome results in the bioactivation of diverse host-active functions mediated by liberated aglycone products of microbiome enzymatic transformation. Thus, in contrast to such reduction of the biosynthetic diversity of plant glycans, deglycosylation of glycosides by the gut microbiome maintains the biosynthetic diversity of plant small molecules acting in bowtie architecture^64,65^ to generate diverse host effector functions (Figure S7A).
GHs within the human gut microbiome have been demonstrated to have substrate specificity dependent on the type of monosaccharide (e.g., glucose, fructose, arabinose, xylose), stereochemistry (α vs. β linkage) and site of linkage (α1, β2, etc.) between these two sugars^17,66–69^. Here we reveal two distinct metabolic systems within the human gut Bacteroides to hydrolyze glycoside PSMs (Figure S7B). In B. ovatus, a single GH16 family member is required to utilize both a range of phenolic glycosides as well as disaccharides. Such substrate specificity is consistent among members of human associated microbiomes, notably that of B. thetaiotaomicron (hydrolyzing both glucosinolates and disaccharides^47^), Enterococcus^39,70^, Lactobacillus^45^, Streptococcus^46^ and E. coli^38,71,72^ in which the GH genes/enzymes are capable of hydrolyzing both glycosides and disaccharides. Such enzyme systems may have evolved primarily for sugar-sugar hydrolysis with broad specificity for glycosides or later expanded functional capacity to include glycoside hydrolysis. In contrast, in B. uniformis, one of the most prevalent members of the human gut microbiota^48,49^, we have revealed a glycoside utilization system composed of GH3 family members dedicated to phenolic glycosides and unable to hydrolyze disaccharides (Bu gshD) and demonstrating specificity for distinct phenolic glycosides depending on the chemical structure of the aglycone phenol group (for example the position of hydroxyl or methanol group) (Bu gshG). Interestingly, these two GH3 of Bacteroides uniformis are present within common clades of GH3 shared with members of plant and soil associated bacteria that hydrolyze plant glycosides^73–76^ (Figure S3G). Together, this suggests that our demonstration of glycoside-dedicated and -specific GH3 enzymes within the human gut Bacteroidales evolved specifically for plant glycosides. It will be interesting to identify the plant glycoside utilization systems and substrate specificity for members of the gut microbiome across diverse taxa (Lachnospiraceae, Clostridiaceae, and Erysipelotrichales) for which we have identified unique glycoside utilization patterns (Figure 1B) and of herbivores (birds, non-human primates) and humans subsisting on plant diets abundant in PSMs such as in gathering cultures^77,78^.
PSM glycosides have been associated with a range of effects on human health and disease. Here, we demonstrate the role of the microbiome in mediating such impacts via the bioactivation and liberation of PSM aglycones in the gut. For salicin, long associated with anti-inflammatory effects, we show that immunoregulation in the gut is mediated by microbiome deglycosylation. In comparison, our screen of Bacteroides metabolism and liberation of aglycones across a panel of dietary phenolic glycosides revealed both the novel pro-inflammatory functions of specific liberated aglycones and in the case of resveratrol, the aglycone product of microbiome deglycosylation of polydatin, as a specific inhibitor of the gut pathogen C. difficile, adding to the benefits of resveratrol on human health^79,80^. Our studies both reveal the impact of the microbiome on known connections between plant dietary glycosides and the host (intestinal inflammation) and of new functions and mechanisms of colonization resistance based on the combination of specific dietary substrates and microbiome bioactivation conferring anti-microbial function. By revealing the microbiome functional activation of specific PSM glycosides across a range of known and as-of-yet unknown host outcomes across physiology and disease, our work presents a novel approach based on pharmacologic dosing and delivery of specific glycosides targeting specific members (endogenous, probiotic, or engineered) to convert and deliver specific functions in prevention and treatment of disease. More broadly, given the diminishing quantities of PSMs in human diets^81,82^, elucidation of the conversion of PSMs into immunoregulatory products by the microbiome presents a novel mechanism of diet-microbiome interactions contributing to the increasing incidence of diseases of industrialization such as autoimmunity and atopy.
There are certain limitations of our work that will be the focus of future research. First, as our work focused on plant phenolic glycoside utilization and bioactivation by a panel of members of the human gut microbiome in vitro and in the context of the murine gut microbiome in vivo, future work will seek to define PSM glycoside metabolism across a broader membership of the human gut microbiome and at the community-level toward defining the role of inter-individual variation in the human gut microbiome on PSM glycoside metabolism utilizing in vivo (humanized mice) and ex vivo (human fecal culture) approaches. Relatedly, given the expansive metabolic capacity of the human gut microbiome, future studies will focus on defining how liberated and functional aglycones may be transformed by members of the gut microbiome resulting in increased or decreased bioactivity. Targeting such members and/or aglycone metabolism can be harnessed to optimize the functional effects of microbiome-liberated aglycones in vivo. While our work has revealed functional roles of specific microbiome-liberated aglycones, future studies will elucidate the molecular mechanisms, such as via specific host chemoreceptors or metabolic pathways, mediating the immunoregulatory functions the salicin aglycone saligenin and the mechanism of action of resveratrol on inhibition of C. difficile. Finally, our work focused on metabolism and functionalization of PSM phenolic glycosides administered individually, future work will determine gut microbiome bioactivation of PSM within the natural matrix of dietary plants.
Requests for further information and resources may be directed to and will be fulfilled by the lead contact, Seth Rakoff-Nahoum (seth.rakoff-nahoum@childrens.harvard.edu).
Plasmids and recombinant strains generated in this study will be distributed on request.
Bacterial strains used in this study were listed in Key Resource Table.
Six-week-old female germ-free C57BL/6 mice were used for gnotobiotic mouse experiments. Brigham and Women’s Hospital Massachusetts Host-Microbiome Center approved under protocol 2020N000054 by the Brigham and Women’s Hospital Institutional Animal Care and Use Committee (IACUC). Mice were housed in Class II Biological Isolator in a temperature-controlled (~21^o^C) facility on a 12 h light/dark cycle. Mice were fed a standard chow (Laboratory Rodent Diet 5025, LabDiet, St. Louis, MO, USA) unless otherwise indicated.
All in vivo disease models were carried out under protocols approved by the Boston Children’s Hospital IACUC (00001465, 00001278). Six-week-old male C57BL/6J Foxp3^YFPCre^ mice were used for experimental colitis experiments. Six to eight-week-old male and female C57BL/6J WT mice were used for C. difficile infection experiments (bred internally, from mice obtained from Jackson Laboratory). For C. difficile infections, for biosafety considerations experiments took place in a biosafety level 2 facility that contains specific pathogens.
All human gut bacterial strains were purchased from the American Type Culture Collection (ATCC), Deutsche Sammlung von Mikroorganismen und Zellkulturen GmbH (DSMZ), the Biodefense and Emerging Infections Research Resources Repository (BEI resources), or from collaborators specified above, except for those strains generated in this work. Frozen stock cultures of bacteria were stored at −80^o^C in cryogenic vials. Individual strains were derived from the same glycerol stocks throughout this study.
All Bacteroidetes, Firmicutes, and Proteobacteria isolates were grown on brain-heart-infusion supplemented with hemin (50 mg/L) and vitamin K1 (0.25 mg/L) (BHIS) agar plates or in YBHIS (brain-heart-infusion powder (37 g/L), yeast (5 g/L), D-(+)-cellobiose (1 g/L), D-(+)-maltose monohydrate (1 g/L), cysteine (0.5 g/L), hemin (50 mg/L), and vitamin K1 (0.25 mg/L). Bifidobacteria isolates were grown on commercial Brucella Blood Agar (Thermo Fisher) or in YBHIS. All isolates were recovered on and grown in pre-reduced media in a Coy anaerobic chamber (Coy Labs). To determine the ability of diverse gut bacteria to utilize glucose (Sigma) or phytochemical glycosides (salicin (Sigma), arbutin (Sigma), salidroside (ChemImpex), gastrodin (TCI), helicin (TCI), esculin (Sigma), amygdalin (Sigma)) as a sole carbon source, each isolate was recovered on a BHIS or BBA agar plate and incubated anaerobically at 37^o^C for 48 h. A starter culture (1 mL) of YBHIS was inoculated with each isolate and grown for 15 h. This starter culture was then used to inoculate a YBHIS sub-culture (1 mL) at a final dilution of 10. This sub-culture was grown for 3 h to an OD6000.5 before being used to inoculate experimental cultures at a final dilution of 50. The experimental media consisted of a modified YCFA media (mYCFA; casitone (2.5 g/L), yeast extract (0.625 g/L), cysteine (0.5 g/L), magnesium sulfate heptahydrate (f.c. 365 µM), calcium chloride dihydrate (f.c. 610 µM), sodium bicarbonate (4 g/L), dipotassium phosphate (0.45 g/L), monopotassium phosphate (0.45 g/L), sodium chloride (0.9 g/L), sodium acetate (2.71 g/L), Hemin (50 mg/L), Vitamin K1 (0.25 mg/L), ferrous sulfate (4 µg/mL), ATCC vitamin mix (1% v/v)) supplemented with glucose (15 mM, Sigma) or glycoside (15 mM). Bacterial growth was monitored by measuring the optical density at 600 nm (OD600~) of each culture in a 384 well plate with an Epoch2 spectrophotometer (BioTek Instruments) and MicroPlate stacker (Agilent BioTek) for 24 hours. Growth was calculated as the area under each 48-hour bacterial growth curve (AUC units). Background growth due to catabolism of media components was subtracted from overall growth of each bacterium in glucose of glycoside to normalize growth across bacteria and substrates. Maximal growth is represented by a deep blue color. Growth was considered for when max OD600 greater then 0.100 above max OD600 in baseline media without carbon source (Figure 1) or greater then 0.100 from starting OD (Figure 2).
All Bacteroidetes isolates were grown on brain-heart-infusion supplemented with hemin (50 mg/L) and vitamin K1 (0.25 mg/L) (BHIS) agar plates or in basal media (BSG; proteose peptone (20 g/L), yeast (5 g/L), NaCl (5 g/L), glucose (5 g/L), potassium phosphate dibasic (5 g/L), cysteine (0.5 g/L), hemin (50 mg/L), and vitamin K1 (0.25 mg/L). All isolates were recovered on and grown in pre-reduced media in a Coy anaerobic chamber (Coy Labs). To determine the ability of Bacteroidetes strains to utilize glucose (Sigma), phytochemical glycosides (salicin (Sigma), arbutin (Sigma), salidroside (ChemImpex), gastrodin (TCI), helicin (TCI), esculin (Sigma), amygdalin (Sigma)), or disaccharides (D-(+)-cellobiose (Sigma), lactose (Sigma), D-(+)-maltose monohydrate (Sigma), melibiose (Sigma), sucrose (Sigma), trehalose (Sigma), palatinose (TCI), gentiobiose (TCI)) as a sole carbon source, each isolate was recovered on a BHIS agar plate and incubated anaerobically at 37^o^C for 48 h. A starter culture (1 mL) of BSG was inoculated with each isolate and grown for 15 h. This starter culture was then used to inoculate a BSG sub-culture (1 mL) at a final dilution of 10. This sub-culture was grown for 3 h to an OD600~~0.5 before being used to inoculate experimental cultures at a final dilution of 50. The experimental media consisted of Bacteroides minimal media (BMM; ammonium sulfate (1 g/L), sodium carbonate (1 g/L), potassium phosphate monobasic (0.9 g/L), sodium chloride (0.9 g/L), calcium chloride dihydrate (26.5 mg/L), magnesium chloride hexahydrate (2 mg/L), manganese(II) chloride tetrahydrate (1 mg/L), cobalt(II) chloride hexahydrate (1 mg/L), hemin (50 mg/L), vitamin K1~ (0.25 mg/L), ferrous sulfate heptahydrate (4 mg/L), vitamin B12 (5 mg/L)) supplemented with glucose (15 mM, Sigma), glycoside (15 mM) or disaccharide (15 mM). Bacterial growth was monitored by measuring the optical density at 600 nm (OD600) of each culture in a 384 well plate with an Epoch2 spectrophotometer (BioTek Instruments) and MicroPlate stacker (Agilent BioTek). Growth was calculated as the area under each 48-hour bacterial growth curve (AUC units). Maximal growth is represented by a deep blue color.
Bacteroides strains (Bacteroides ovatus ATCC 8483 and B. uniformis ATCC 8492, and B. uniformis ATCC 8492 ∆gshD) were grown in 20 mL of BSG to the late-lag phase (OD600 0.1). The E. coli S17–1 lambda pir donor, which harbors the transposon plasmid pWH2-Term6^87^, was grown in 1 mL of LB media supplemented with ampicillin to exponential phase. Both Bacteroides and E. coli donor cultures were centrifuged (4000 rpm, 5 min, 21C), re-suspended in equal volumes of phosphate buffered saline (PBS), mixed in equal proportions, and spotted onto BHIS agar plates. These plates were incubated aerobically at 37^o^C for less than 18 hours. The conjugation mix was then selected on BHIS agar plates supplemented with gentamicin (200 µg/mL) and erythromycin (5 µg/mL) anaerobically at 37^o^C for 48 hours. Over 100,000 transconjugant colonies were scrapped from the plates and pooled in 25% glycerol stocks stored at −80^o^C. An aliquot of the transposon mutant bank was thawed, recovered overnight in BSG media (starting OD600~ 0.1), back-diluted 10-fold and grown to mid-exponential phase (OD600~ ~0.5). This culture was then diluted 50-fold into BMM containing glucose (0.25%, Sigma) or arbutin (0.25%, Sigma). An aliquot of this culture was also saved as an input control. All experimental cultures were grown to late-exponential phase and harvested by spinning down (8000 rpm, 5 min, 21C) and storing at −80C for library preparation and sequencing.
To construct TnSeq DNA libraries for sequencing, the collected samples were first subjected to DNA extraction using Invitrogen PureLink Genomic DNA mini kit (Thermo Fisher Scientific, Waltham, MA). An aliquot of 5 ug of DNA was sheared in 100 µL ultrapure water to 300500 base pairs using an M220 focused ultrasonicator (Covaris Inc). The sheared DNA was subjected to end repair, A-tailing and ligated with prepared adapters (generated by annealing two 5’-TACCACGACCA-NH2–3’ and 5’-GTGACTGGAGTTCAGACGTGTGCTCTTCCGATCTGGTCGTGGTAT-3’) using NEBNext Ultra DNA Library Prep Kit (New England Biolabs). Next, transposon junction fragments containing the insertion sites were enriched by two PCR amplifications using Platinum SuperFi PCR Master Mix (Thermo Fisher Scientific, Waltham, MA). In the first PCR, the fragments were amplified (98^o^C for 10s, 60^o^C for 10s, 72^o^C for 30s with 25 cycles) with universal primers matching the transposon and adaptor 5’-CCCATTGGGAATAATAACCTTTATACCTG-3’ and 5’-GTGACTGGAGTTCAGACGTGTG-3’. In the second PCR, the PCR products were amplified (98^o^C for 10s, 60^o^C for 10s, 72^o^C for 30s with 18 cycles) with Truseq-compatible indexed primers containing specific annealing sequences. The PCR fragments corresponding to 200600 bp were selected using AMPure XP beads (Beckman Coulter Inc). These libraries were normalized using NEB- Next Library Quant kit (New England Biolabs) and subjected to HiSeq 2500 High Output sequencing (Illumina Inc, San Diego, CA) with a read depth of 5 million reads per sample.
Clean deletion mutants were created by allelic replacement whereby the following genes were Bacteroides uniformis ATCC 8492 gshA (BACUNI_00922), gshB (BACUNI_00920), gshC (BACUNI_00922), gshD (BACUNI_00919), gshG (BACUNI_01042), gghC (BACUNI_00951) and B. ovatus ATCC 8483 gghR (Bovatus_02231), gghA (Bovatus_02230), gghB (Bovatus_02229), gghC (Bovatus_02228), gghD (Bovatus_02227). We leveraged a modified pKNOCK-bla-ermGb suicide vector, pKNOCK-bla-ermGb_pheS*, we previously engineered with a counter-selection system mediated by a variant phenylalanyl-tRNA synthetase (PheSA303G), lethal in the presence of 4-chloro-phenylalanine^87^. To generate deletion constructs, 1kb regions flanking the genomic region-of-interest were amplified by PCR (Platinum SuperFi PCR Master Mix), purified, and cloned into the pKNOCK-bla-ermGb_pheS* plasmid in a three-piece ligation reaction^88^.
To mobilize resulting plasmids into Bacteroides, a triparental mating strategy was implemented. Overnight cultures of Bacteroides grown in BSG were diluted 50-fold into fresh pre-reduced BSG and grown to early exponential phase (OD600 ~0.1). The recipient strain was centrifuged at 4,000 rpm for 5 minutes and resuspended in 200 µL of 1x PBS. One-day-old E. coli DH5α lambda pir donor strain and E. coli DH5α helper strain with pRK231 were scrapped from LB agar plates into 200 µL of 1x PBS. Donor, helper, and recipient strains were combined at a 1:1 helper:recipient culture volume ratio at a final volume of 90 µL and spotted on non-selective BHIS agar plate for 18 hours at 37^o^C under aerobic conditions to allow for conjugation. Mating lawns were scrapped onto BHIS agar plates containing gentamicin (200 µg/mL) and erythromycin (5 µg/mL) to select for transconjugants (merodiploids). Single colonies were isolated by re-streaking colonies onto BHIS agar plates containing erythromycin (5 µg/mL). A single colony was re-streaked onto BMM agar plates supplemented with 4-Chloro-DL-phenylalanine (2 mg/mL, Acros Organics) to select for the loss of the pKNOCK-bla-tetQ-pheS* vector. Individual clones were replica-plated onto BHIS agar and BHIS agar supplemented with erythromycin (5 µg/mL) and those clones that did not grow on erythromycin were confirmed for genetic manipulation by PCR and sequencing.
Expression of Bacteroides uniformis ATCC 8492 genes gshD and gshG into B. vulgatus ATCC 8482, B. uniformis ATCC 8492 gene gshD into B. uniformis ATCC 8492 ∆gshD, and B. ovatus ATCC 8483 genes gghR, gghA, gghB, gghC, and gghD into B. ovatus ATCC 8483 ∆gghR, ∆gghA, ∆gghB, ∆gghC, and ∆gghD, respectively, was performed using the extra-chromosomal Bacteroides high-copy expression vector pFD340^86^. Target genes plus 50 base pairs upstream of the gene-of-interest containing the native RBS were amplified (Platinum SuperFi PCR Master Mix) using primers listed elsewhere (Table S1), purified, and cloned into pFD340 directly downstream of the vector-borne promoter in a two-piece ligation reaction^88^. A similar triparental conjugation strategy was performed as described above to mobilize the pFD340 constructs into heterologous hosts. Mating lawns were scrapped onto BHIS agar plates supplemented with gentamicin (200 µg/mL) and erythromycin (5 µg/mL) to select for Bacteroides containing the constructs. Single colonies were isolated by re-streaking colonies onto BHIS agar plates supplemented with erythromycin (5 µg/mL).
Bacteroides uniformis ATCC 8492 genes gshD and gshG were amplified with primers (Table S1) and genomic DNA from B. uniformis ATCC 8492 as template. Each gene was cloned into pET-28a (Novagen) by Gibson Assembly^88^. The recombinant plasmids were transformed into E. coli BL21 (DE3) (Novagen) strain, grown to OD6000.5 in LB plus kanamycin at 37^o^C and induced with IPTG (f.c. 0.1 mM) at 18^o^C overnight. After harvesting (10,000 g, 10 min, 4^o^C), cells were suspended in lysis buffer (20 mM K2HPO4~, 20 mM KH2PO4, 500 mM NaCl, 50 mM Imidazole) with 1x Halt Protease Inhibitor Cocktail (Thermo Fisher Scientific) and lysed by sonication. The cell extracts were spun down, filtered through a 0.45 µm filter, mixed with 2 mL of pre-washed Ni-NTA resin (HisPur Ni-NTA Resin, Thermo Fisher Scientific), and incubated on a rotator at 4^o^C overnight. The resin was loaded onto a column and washed with lysis buffer containing Imidazole (50 mM). The recombinant proteins were eluted using elution buffer (20 mM K2HPO4, 20 mM KH2PO4, 500 mM NaCl, 500 mM Imidazole) for both gshD and gshG. After concentration in a spin concentrator pre-dialysis, fractions of each protein were dialyzed overnight at 4^o^C in 1 L of storage buffer (pH 8.0, 25 mM HEPES, 50 mM NaCl) three times. The size and purity of the recombinant proteins were verified by SDS-PAGE gel (NuPAGE 4 to 12%, Bis-Tris Mini Protein Gel, Thermo Fisher Scientific) and the concentration of each protein was measured by absorbance at 280 nm using Nanodrop (Thermo Fisher Scientific).
All B. uniformis strains were grown on brain-heart-infusion supplemented with hemin (50 mg/L) and vitamin K1 (0.25 mg/L) (BHIS) agar plates or in basal media (BSG; proteose peptone (20 g/L), yeast (5 g/L), NaCl (5 g/L), glucose (5 g/L), potassium phosphate dibasic (5 g/L), cysteine (0.5 g/L), hemin (50 mg/L), and vitamin K1 (0.25 mg/L). All isolates were recovered on and grown in pre-reduced media in a Coy anaerobic chamber (Coy Labs).
To assess the capacity of Bacteroides to liberate aglycones from their parent aryl, coumarin, and cyanogenic glycosides, wildtype B. uniformis ATCC 8492 was recovered on a BHIS agar plate and incubated anaerobically at 37^o^C for 48 h. A starter culture (1 mL) of BSG was inoculated with each isolate and grown for 15 h. This starter culture was then used to inoculate a BSG sub-culture (1 mL) at a final dilution of 10. This sub-culture was grown for 3 h to an OD600~0.5 before being used to inoculate experimental cultures at a final dilution of 50. The experimental media consisted of Bacteroides minimal media (BMM; ammonium sulfate (1 g/L), sodium carbonate (1 g/L), potassium phosphate monobasic (0.9 g/L), sodium chloride (0.9 g/L), calcium chloride dihydrate (26.5 mg/L), magnesium chloride hexahydrate (2 mg/L), manganese(II) chloride tetrahydrate (1 mg/L), cobalt(II) chloride hexahydrate (1 mg/L), hemin (50 mg/L), vitamin K1 (0.25 mg/L), ferrous sulfate heptahydrate (4 mg/L), vitamin B12 (5 mg/L)) supplemented with glucose (0.1%, Sigma) and any one G1G (500 µM, Sigma) or glucose (0.1%, Sigma) and DMSO (0.5%, Sigma). At late log, each experimental culture was isolated, diluted (1:50) into 80% methanol-water spiked with tyrosol-D4 (40 nM, TRC Canada), vortexed (2 min), centrifuged (5000 rpm, 5 min), and the supernatant was recovered for down-stream LCMS analyses.
To assess the capacity of Bacteroides to liberate aglycones from their parent polyphenolic glycosides, B. uniformis ATCC 8492 WT, ∆gshD, and ∆gshD∆gshG∆gghC were recovered on a BHIS agar plate and incubated anaerobically at 37^o^C for 48 h. A starter culture (1 mL) of BSG was inoculated with each isolate and grown for 15 h. This starter culture was then used to inoculate a BSG sub-culture (1 mL) at a final dilution of 10. This sub-culture was grown for 3 h to an OD600~0.5 before being used to inoculate experimental cultures at a final dilution of 50. The experimental media consisted of Bacteroides minimal media (BMM; ammonium sulfate (1 g/L), sodium carbonate (1 g/L), potassium phosphate monobasic (0.9 g/L), sodium chloride (0.9 g/L), calcium chloride dihydrate (26.5 mg/L), magnesium chloride hexahydrate (2 mg/L), manganese(II) chloride tetrahydrate (1 mg/L), cobalt(II) chloride hexahydrate (1 mg/L), hemin (50 mg/L), vitamin K1 (0.25 mg/L), ferrous sulfate heptahydrate (4 mg/L), vitamin B12 (5 mg/L)) supplemented with glucose (0.1%, Sigma) and any one G3G (500 µM, Sigma) or glucose (0.1%, Sigma) and DMSO (0.5%, Sigma). At late log, each experimental culture was isolated, diluted (1:50) into 80% methanol-water spiked with C^13^-Resveratrol (250 nM, Sigma), vortexed (2 min), centrifuged (5000 rpm, 5 min), and the supernatant was recovered for down-stream LCMS analyses.
Six-week-old female GF C57BL/6J mice were colonized with 10^9^ CFUs of Bacteroides uniformis ATCC 8492 or Bacteroides uniformis ∆gshD∆gshG∆gghC. At two weeks post-colonization, mice were treated intra-gastrically (i.g.) gavaged with salicin (100 mg/kg) dissolved in sterile 1x PBS. The stool of these mice was collected immediately prior to gavage (T0) and then three (T3) and five (T5) hours post gavage. Stool was immediately stored at −80^o^C for confirmation of Bacteroides colonization and targeted LCMS detection of Bacteroides-liberated saligenin.
To extract metabolites from fecal samples, 200 µL of 0.1 mm zirconia/silica beads (BioSpec Products) and 1 mL of organic solvent (methanol:water, 1) supplemented with tyrosol-D4 (40 nM, TRC Canada) were added to 5–50 mg of pre-weighed fecal matter. Material was homogenized by vortexing at maximum speed for 30 seconds before mechanical disruption with a bead beater (BioSpec Products) for 5 minutes on high setting at room temperature. Samples were incubated in an ultrasonicator for 5 minutes before incubation at −20^o^C overnight. Samples were subsequently thawed at room temperature before centrifugation (15,000 rpm, RT) for 5 minutes. 800 µL of supernatant was isolated from each sample for analysis by LCMS.
For aryl glycoside and aglycone-containing samples, samples were dried under nitrogen flow and resuspended in 50 µL of methanol 50% in water. A standard curve was prepared using the same extraction solution and volumes as for the samples. The curve was prepared as a 10 points 1/5 dilution series with 1 mM as the highest concentration. Samples were quantified on a QEplus mass spectrometer coupled to an Ultimate 3000 LC (Thermo fisher). Five microliters were injected on a Luna Omega Polar C18 column (2mm x 150 mm, Phenomenex) maintained at 40°C. The mobile phases were A : water and B: Methanol. The gradient was as 1% B for 4 min, then to 100% B in 6 min. the mobile phases were then maintained at 100% B for 15 min, followed by 5 min re-equilibration at 1% B. The flow rate was 0.15 mL min^-1^. Ionization for the mass spectrometer was achieved with atmospheric pressure chemical ionization (APCI) with a probe at 350 °C, in switching polarity mode. Quantification was performed using Tracefinder (Thermo fisher), using the ratio of the area under the peak for each compound and of the internal standard, using the accurate mass of the [M-H]^-^ ions mainly, except for hydroquinone and benzoquinone where the [M+e]^-^ and the [M+H]^+^ were used respectively.
For polyphenolic glycoside and aglycone-containing samples, samples were dried under nitrogen flow and resuspended in 50ul of methanol 50% in water. A standard curve was prepared using the same extraction solution as for the samples. The curve was prepared as a 10 points 1/5 dilution series with 100 μM as the highest concentration. Samples were quantified on a QEplus mass spectrometer coupled to an Ultimate 3000 LC (Thermo fisher). Five microliters were injected on a C18Evo column (2mm x 150 mm, Phenomenex) maintained at 30°C. The mobile phases were A : water, 5mM ammonium formate, with pH adjusted to 7 and B: Acetonitrile, 0.1% ammonium hydroxide. The gradient was as 0% B for 5 min, then to 40% B in 9 min, then to 50% B in 6min, and finally to 100% B in 1 min. The mobile phases were then maintained at 100% B for 7 min, followed by 5 min re-equilibration at 0% B. The flow rate was 0.15 mL min^-1^. Ionization was achieved by heated electrospray ionization (HESI) in negative mode. Quantification was performed using Tracefinder (Thermo fisher), using the ratio of the area under the peak for each compound and of the internal standard, using the accurate mass of the [M-H]^-^ ions.
Recombinant gshD (10 nM) was combined with salidroside or salicin (substrate concentration 0, 1, 10, 20, 40, 60, 80, 100, 250, 500, 750 µM, 1000 µM) in reaction buffer (pH 6.5, 25 mM HEPES, 25 mM NaCl). Reactions were set to incubate at 37^o^C for 10 minutes before being quenched with sodium carbonate (200 mM, f.c. 100 mM). Each reaction was further diluted in methanol (100%, f.c. 33%) in preparation for down-stream LCMS analyses. Each reaction was performed in triplicate, twice on separate days.
Recombinant gshG (500 nM) was combined with salidroside or salicin (substrate concentration 100, 250, 500, 750, 1000, 1250, 1500, 1750, 2000, 2250, 2500, 2750, 3000, 3250, 3500, 3750, 4000 µM) in reaction buffer (pH 6.5, 25 mM HEPES, 25 mM NaCl). Reactions were set to incubate at 37^o^C for 18 hours before being quenched with sodium carbonate (200 mM, f.c. 100 mM). Each reaction was further diluted in methanol (100%, f.c. 33%) in preparation for down-stream LC-MS analyses. Each reaction was performed in triplicate, twice on separate days.
Recombinant gshD (10 nM) or gshG (500 nM) were combined with salidroside, tyrosol, salicin, saligenin, cellobiose, lactose, maltose, melibiose, sucrose, trehalose, palatinose, or gentiobiose (substrate concentration of 900 µM) in reaction buffer (pH 6.5, 5 mM HEPES, 5 mM NaCl). Reactions were set to incubate at 37^o^C for 1 hour (gshD) or 18 hours (gshG) before being quenched with sodium carbonate (200 mM, f.c. 100 mM). Each reaction was further diluted in methanol (100%, f.c. 33%) in preparation for down-stream LCMS analyses. Each reaction was performed in triplicate, twice on separate days.
Data was collected with a Q-Exactive Orbitrap mass spectrometer (Thermo Fisher Scientific, San Jose, CA) for MS scans in both positive and negative polarity modes and additionally for targeted SIM mode. The samples were run through a SeQuant^®^ ZIC^®^-pHILIC 150 × 2.1 mm analytical column equipped with a 2.1 × 20 mm guard column (both 5 mm particle size; EMD Millipore). Buffer A was 20 mM ammonium carbonate, 0.1% ammonium hydroxide; Buffer B was acetonitrile. The chromatographic gradient was run at a flow rate of 0.150 mL/min as 0–20 linear gradient from 80–20% B; 20–20.5 linear gradient form 20–80% B; 20.5–28 hold at 80% B.
MS data acquisition was performed in a range of m/z = 70–1000, with the resolution set at 70,000, the AGC target at 1×10^6^, and the maximum injection time (Max IT) at 20 msec. For detection of salidroside, tyrosol, salicin, saligenin, cellobiose, lactose, maltose, melibiose, sucrose, trehalose, palatinose, or gentiobiose, targeted selected ion monitoring (tSIM) scans in negative mode were included. The isolation window was set at 1.0 m/z and tSIM scans were centered at m/z = 299.1136 (salidroside, RT = 3.88 min), m/z = 137.0608 (tyrosol, RT = 2.67 min), m/z = 285.0980 (salicin, RT = 4.08 min), m/z = 123.0452 (saligenin, RT = 2.88 min), m/z = 341.1089 (palatinose, RT =10.04 min; gentiobiose, RT = 11.10 min; sucrose, RT = 9.99 min; melibiose, RT = 11.14 min; maltose, RT = 10.65 min; lactose, RT = 10.85 min; cellobiose, RT = 10.60 min; trehalose, RT = 10.67 min) and m/z = 179.0061 (glucose, RT = 9.63 min; fructose, RT = 8.59 min; galactose, RT = 9.78 min). For all tSIM scans, the resolution was set at 70,000, the AGC target was 1×10^5^, and the max IT was 200 msec. The raw data was processed using Xcalibur (Thermo Fisher Scientific) and the peaks’ area and height were calculated by setting the peak algorithm detection as Genesis for all the samples.
All gut pathogen isolates were grown on brain-heart-infusion supplemented with hemin (50 mg/L) and vitamin K1 (0.25 mg/L) (BHIS) agar plates or in basal media (BSG; proteose peptone (20 g/L), yeast (5 g/L), NaCl (5 g/L), glucose (5 g/L), potassium phosphate dibasic (5 g/L), cysteine (0.5 g/L), hemin (50 mg/L), and vitamin K1 (0.25 mg/L). Isolates were recovered on and grown in pre-reduced media in a Coy anaerobic chamber (Coy Labs). To determine the fitness effects of salicin (Sigma), saligenin (Sigma), arbutin (Sigma), hydroquinone (Sigma), salidroside (ChemImpex), tyrosol (Sigma), gastrodin (TCI), gastrodigenin (Sigma), helicin (TCI), salicylaldehyde (Sigma), rutin (TCI), quercetin (TCI), naringin (TCI), naringenin (TCI), cynaroside (MedChemExpress), luteolin (TCI), phloridzin hydrate (TCI), phloretin (TCI), polydatin (TCI), resveratrol (TCI), genistin (Thermo Scientific Chemicals), genistein (TCI), daidzin (TCI), daidzein (TCI), pinoresinol diglucoside (Sigma) and pinoresinol (Sigma) on enteric pathogen fitness, each pathogen isolate was recovered on a BHIS agar plate and incubated anaerobically at 37^o^C for 48 h. A starter culture (1 mL) of BSG was inoculated with C. difficile and grown for 15 h. This starter culture was then used to inoculate a BSG sub-culture (1 mL) at a final dilution of 10. This sub-culture was grown for 3 h to an OD6000.5 before being used to inoculate experimental cultures at a final dilution of 50. The experimental media consisted of a modified YCFA media (mYCFA; casitone (2.5 g/L), yeast extract (0.625 g/L), cysteine (0.5 g/L), magnesium sulfate heptahydrate (f.c. 365 µM), calcium chloride dihydrate (f.c. 610 µM), sodium bicarbonate (4 g/L), dipotassium phosphate (0.45 g/L), monopotassium phosphate (0.45 g/L), sodium chloride (0.9 g/L), sodium acetate (2.71 g/L), Hemin (50 mg/L), Vitamin K1 (0.25 mg/L), ferrous sulfate (4 µg/mL), ATCC vitamin mix (1% v/v)) supplemented with glucose (15 mM, Sigma) plus each compound (150 µM) or glucose (15 mM, Sigma) plus DMSO (0.15%, Sigma). Bacterial growth was monitored by measuring the optical density at 600 nm (OD600~) of each culture in a 384 well plate with an Epoch2 spectrophotometer (BioTek Instruments) and MicroPlate stacker (Agilent BioTek) for 24 hours.
B. uniformis strains were grown on brain-heart-infusion supplemented with hemin (50 mg/L) and vitamin K1 (0.25 mg/L) (BHIS) agar plates or in basal media (BSG; proteose peptone (20 g/L), yeast (5 g/L), NaCl (5 g/L), glucose (5 g/L), potassium phosphate dibasic (5 g/L), cysteine (0.5 g/L), hemin (50 mg/L), and vitamin K1 (0.25 mg/L). Isolates were recovered on and grown in pre-reduced media in a Coy anaerobic chamber (Coy Labs). To assess the effects of resveratrol liberation from polydatin on the fitness of C. difficile M7404, wildtype B. uniformis ATCC 8492 and B. uniformis ATCC 8492 ∆gshD were recovered on BHIS agar plates and incubated anaerobically at 37^o^C for 48 h. A starter culture (1 mL) of BSG was inoculated with each isolate and grown for 15 h. This starter culture was then used to inoculate a BSG sub-culture (1 mL) at a final dilution of 10. This sub-culture was grown for 3 h to an OD600~0.5 before being used to inoculate experimental cultures at a final dilution of 50. The experimental media consisted of Bacteroides minimal media (BMM; ammonium sulfate (1 g/L), sodium carbonate (1 g/L), potassium phosphate monobasic (0.9 g/L), sodium chloride (0.9 g/L), calcium chloride dihydrate (26.5 mg/L), magnesium chloride hexahydrate (2 mg/L), manganese(II) chloride tetrahydrate (1 mg/L), cobalt(II) chloride hexahydrate (1 mg/L), hemin (50 mg/L), vitamin K1 (0.25 mg/L), ferrous sulfate heptahydrate (4 mg/L), vitamin B12 (5 mg/L)) supplemented with glucose (0.1%, Sigma) and polydatin (500 µM, Sigma) or glucose (0.1%, Sigma) and DMSO (0.5%, Sigma). After 8 hours, conditioned media was prepared from experimental cultures by centrifugation (5000 rpm, 5 min), and the supernatant was recovered and filter-sterilized (0.22 μm).
To assess the capacity of B. uniformis metabolism of polydatin to antagonize the fitness of C. difficile, C. difficile M7404 was recovered on a pre-reduced BHIS agar plate and incubated anaerobically at 37^o^C for 48 h. A starter culture (1 mL) of BSG was inoculated with C. difficile and grown for 15 h. This starter culture was then used to inoculate a BSG sub-culture (1 mL) at a final dilution of 10. This sub-culture was grown for 3 h to an OD6000.5 before being used to inoculate experimental cultures at a final dilution of 50. The experimental media consisted of a modified YCFA media (mYCFA; casitone (2.5 g/L), yeast extract (0.625 g/L), cysteine (0.5 g/L), magnesium sulfate heptahydrate (f.c. 365 µM), calcium chloride dihydrate (f.c. 610 µM), sodium bicarbonate (4 g/L), dipotassium phosphate (0.45 g/L), monopotassium phosphate (0.45 g/L), sodium chloride (0.9 g/L), sodium acetate (2.71 g/L), Hemin (50 mg/L), Vitamin K1 (0.25 mg/L), ferrous sulfate (4 µg/mL), and ATCC vitamin mix (1% v/v)) supplemented with glucose (15 mM, Sigma) plus ½-diluted B. uniformis WT or ∆gshD conditioned media from growth with or without polydatin. C. difficile growth was monitored by measuring the optical density at 600 nm (OD600~) of each culture in a 384 well plate with an Epoch2 spectrophotometer (BioTek Instruments) for 12 hours.
We utilized an antibiotic-treatment-induced CDI mouse model developed previously^89^. Briefly, six- to eight-week-old male and female SPF C57BL/6J WT mice were treated for three days with a mixture of antibiotics including vancomycin (0.4 mg/mL, Sigma) colistin (850 U/mL, Sigma), metronidazole (0.215 mg/mL, Sigma), gentamicin (0.035 mg/mL, RPI), and kanamycin (0.045 mg/mL, Sigma) in their drinking water. Mice were then treated with regular water for two days and then injected intraperitoneal (i.p.) with clindamycin (10 mg/kg, Mylan Pharmaceuticals) in normal saline. After one day, mice were gavaged with 10^4^ spores of C. difficile 630 diluted in PBS. Resveratrol, phloretin, or vehicle (0.1% Tween) were provided in the drinking water at 500 µM the day of infection (day 0) and then each subsequent day for two days (day 1 and day 2). Stool was collected from mice at one day and two days post infection, and immediately taken for assessment of C. difficile infection load.
Mouse feces were collected from mice on day one and day two of C. difficile infection. Feces were weighed and resuspended to 50 mg/mL in pre-reduced 1x PBS. CFUs were enumerated by plating multiple dilutions on pre-reduced commercial ChromID C. difficile-selective plates (Biomerieux) after incubation for 24 hours at 37^o^C in an anaerobic chamber (Coy Labs).
Immortalized mouse bone marrow-derived macrophages (iBMDM) were plated in a 96-well plate (100,000 cells/well) and individually incubated with all aryl glycosides and their respective aglycones at 5 mM, 2.5 mM, 1.25 mM, and 0.625 mM: salicin (Sigma), saligenin (Sigma), arbutin (Sigma), hydroquinone (Sigma), salidroside (ChemImpex), tyrosol (Sigma), gastrodin (TCI), gastrodigenin (Sigma), helicin (TCI), and salicylaldehyde (Sigma), as well as all polyphenolic glycosides and their respective aglycones at 75 µM: rutin (TCI), quercetin (TCI), naringin (TCI), naringenin (TCI), cynaroside (MedChemExpress), luteolin (TCI), phloridzin hydrate (TCI), phloretin (TCI), polydatin (TCI), resveratrol (TCI), genistin (Thermo Scientific Chemicals), genistein (TCI), daidzin (TCI), daidzein (TCI), pinoresinol diglucoside (Sigma) and pinoresinol (Sigma). Cells were incubated for 16 hours at 37^o^C and 5% CO2 and then stimulated with LPS (1 µg/mL, E. coli, Serotype O111:B4, Enzo Life Sciences) or vehicle for 24 hours. Cells were spun down for five minutes at 400 rpm and then the supernatant was recovered and stored at −20^o^C for future analyses. For in vitro cytokine measurements, TNF-α (Thermo Fisher, 88–7324-88) and IL-6 (Thermo Fisher, 88–7064-88) were measured by ELISA in the supernatant of LPS- or vehicle-stimulated, glycoside- or aglycone-treated iBMDM. Cellular viability was measured at the end-point of each experiment involving salicin, saligenin, arbutin, and hydroquinone via an MTT assay (ATCC, 30–1010K).
We utilized an experimental model of colitis leveraging dextran sodium sulfate (DSS, MP Biomedicals). Briefly, six-week-old male SPF C57BL/6J Foxp3^YFPCre^ mice were pre-treated IG via gavage with a mixture of antibiotics including vancomycin (500 mg/L, Sigma), metronidazole (1 g/L, Sigma), ampicillin (1 g/L, Sigma) and neomycin (1 g/L, Sigma) once a day for five days. Mice were then treated with regular water for three days to wash out residual antibiotics. After 24 hours mice were colonized with 10^9^ CFUs of wildtype Bacteroides uniformis ATCC 8492 or Bacteroides uniformis ∆gshD∆gshG∆gghC via i.g. administration. After 24 hours, mice were massed and IG administered salicin (100 mg/kg, Sigma) dissolved in sterile 1x PBS, arbutin (100 mg/kg, Sigma) dissolved in sterile 1x PBS, or vehicle and concomitantly treated with 2.5% DSS (MP Biomedicals) in their drinking water. For six subsequent days mice were massed and treated with salicin and DSS, arbutin and DSS, or vehicle and DSS. On the eighth day post-colonization, stool was collected and stored at −80^o^C. Mice were then sacked, massed, the length of each colon from caecum to rectum was measured, and 1 cm descending colon tissue was isolated into neutral buffered formalin (10%) for histological analyses. After 24 hours, tissue was stored in 70% ethanol for long-term storage.
For aglycone protection determination, six-week-old male SPF C57BL/6J Foxp3^YFPCre^ mice were pre-treated i.g. with a mixture of antibiotics including vancomycin (500 mg/L, Sigma), metronidazole (1 g/L, Sigma), ampicillin (1 g/L, Sigma) and neomycin (1 g/L, Sigma) once a day for five days. Mice were then treated with regular water for three days to wash out residual antibiotics. After 24 hours, mice were IG administered saligenin (100 mg/kg, Sigma) dissolved in sterile 1x PBS or vehicle and concomitantly treated with 2.5% DSS (MP Biomedicals) in their drinking water. For six subsequent days mice were treated with saligenin and DSS or vehicle and DSS. On the eighth day post-colonization, stool was collected and stored at −80^o^C. At sacrifice, the length of each colon from caecum to rectum was measured, and 1 cm descending colon tissue was isolated into neutral buffered formalin (10%) for histological analyses. After 24 hours, tissue was transferred to 70% ethanol for long-term storage.
For histological analyses, tissue was delivered to the Beth Israel Deaconess Medical Center (BIDMC) core facility where it was embedded. Briefly, blocks were cut at 6 µm and subjected to Hematoxylin and eosin (H&E) staining. Slides were scored blind, and scores are representative of the entire sample (0 – leukocyte infiltration is normal, 1 – mild localized mild leukocyte infiltrate, 2 – moderate generalized mild leukocyte infiltrate with areas of heavy leukocyte infiltrate, 3 – severe generalized moderate leukocyte infiltrate with areas of high leukocyte infiltrate). Representative images were taken later.
To confirm Bacteroides uniformis colonization in vivo we performed a modification of MK-SpikeSeq which we previously developed^90^. Stool from experimental colitis model mice (Figure 5) treated with (1) vehicle, (2) 2.5% DSS, (3) 2.5% DSS plus wildtype B. uniformis ATCC 8492, (4) 2.5% DSS plus B. uniformis WT plus salicin, or (5) 2.5% DSS plus B. uniformis ∆gshD∆gshG∆gghC plus salicin spiked with 0.01 OD of Salinibacter ruber DSM 13855 per 10 mg of mouse stool such that the final amount of S. ruber per sample would be approximately 1% of the total bacterial 16S abundance. Next, total gDNA was isolated from each sample (ZymoBiomics DNA Miniprep Kit). Relative quantification of B. uniformis using B. uniformis-specific primers was performed by real-time PCR (QuantStudio 3) using Fast SYBR Green Master Mix (Applied Biosystems). The abundance of Bacteroides colonization was calculated versus PBS-treated mice using S. ruber as a standard wherein the resulting -∆∆Ct value was a measure of bacterial colonization in B. uniformis-treated versus non-B. uniformis-treated mice.
Fecal samples were collected and stored at −80^o^C for long-term storage. Raw genomic DNA for downstream 16S amplicon next generational sequencing was isolated using the ZymoBIOMICS™ – 96 DNA Kit (Zymo Research). The 16S amplicon library was prepared using dual-index barcodes in a 96-well format and cleaned with the DNA Clean and Concentrator ™ – 5 (Zymo Research) before quantification by qPCR (NEBNext Library Quant Kit). 20 pM of cDNA were loaded on an Illumina MiSeq and sequenced (v3, 600-cycle, 300nt paired-end). To generate the OTU table for downstream analyses of murine gut microbiome composition and diversity, the obtained Illumina raw reads were de-multiplexed, paired end joined, adapter trimmed, quality filtered, dereplicated, and denoised. Sequences were mapped against the publicly available 16S rRNA DNA databases SILVA and UNITE and clustered into OTUs at greater than or equal to 97% nucleotide sequence identity. OTU-based microbial community diversity was estimated by calculating the alpha diversity (Simpson Index).
Freshly collected adult SPF B6 mouse fecal pellets transported on ice were taken into a Coy anaerobic chamber (Coy Labs). Fecal pellets (0.5 grams) were suspended in pre-reduced YBHIS (15 mL) with 20% glycerol in a 50 ml sterile conical. The suspension was mixed and filtered with a 70 μm cell strainer (Corning). One milliliter of the filtered suspension was stored in sterile cryogenic vials at −80 °C. A mouse fecal glycerol stock was used to inoculate, with a final dilution of 1000, pre-reduced modified YCFA media (mYCFA; casitone (2.5 g/L), yeast extract (0.625 g/L), cysteine (0.5 g/L), magnesium sulfate heptahydrate (f.c. 365 µM), calcium chloride dihydrate (f.c. 610 µM), sodium bicarbonate (4 g/L), dipotassium phosphate (0.45 g/L), monopotassium phosphate (0.45 g/L), sodium chloride (0.9 g/L), sodium acetate (2.71 g/L), Hemin (50 mg/L), Vitamin K1 (0.25 mg/L), ferrous sulfate (4 µg/mL), ATCC vitamin mix (1% v/v)), glucose (0.1 g/L, Sigma), maltose (0.1 g/L, Sigma), cellobiose (0.1 g/L, Sigma), GlcNAc (0.1 g/L, Sigma) and arginine (1 g/L)) supplemented with resveratrol (100 μM, TCI) or saligenin (100 μM, Sigma). Two hundred microliters of the cultures with the corresponding aglycone were grown in a 96-well plate (Corning) for 24 hours at 37 °C in the anaerobic chamber. Cultures at times 0 (start) and 24 (end, stationary phase) hours were stored at −20 °C.
Samples were thawed to room temperature. Each experimental culture was isolated, diluted (1:20) into 80% methanol-water spiked with resveratrol-^13^C (250 nM, Cambridge Isotopes), vortexed (2 min), centrifuged (5000 rpm, 5 min), and the supernatant was used for down-stream LCMS analyses. A standard curve was prepared using the same extraction solution as for the samples. The curve was prepared as a 10-point 1/5 dilution series with 100 μM as the highest concentration. Dihydroresveratrol standard (Sigma) was used to determine metabolism of resveratrol in mouse fecal cultivation. Samples were quantified on an Agilent 6530 Q-TOF mass spectrometer coupled to an Agilent 1290 Infinity II LC System. One microliter was injected into a ZORBAX Eclipse C18 column (2.1 mm x 50 mm, 1.8-Micron) maintained at 40 °C. The mobile phases were A: water, 5mM ammonium acetate, and B: methanol, 5mM ammonium acetate. The gradient was as 13% B for 1 min, then to 100% B in 4 min, the mobile phases were then maintained at 100% B for 3 min, followed by 1 min re-equilibration at 13% B. The flow rate was 0.3 mL min^-1^. Ionization was achieved by heated electrospray ionization (HESI) in negative mode. Quantification was performed using Q-TOF Quantitative Analysis (Agilent), using the area under the peak for each compound, using the accurate mass of the [M-H]^-^ ions.
16S rRNA gene sequences within genomes were located using BLAST, and a multiple gene alignment was carried out using MAFFT version 7. A 16S rRNA tree was reconstructed using the neighbor-joining method implemented in MAFFT version 7, with Jukes-Cantor as the substitution model and a bootstrap value of 1000 replications^91^.
Strains without a complete taxonomic classification were classified at the genus level by comparing the full 16S rRNA genes against curated 16S rRNA databases such as SILVA, and at the species level by comparing the nucleotide identity of several housekeeping genes against several species on the JGI integrated microbial genomes and microbiomes portal.
The phylogenic tree of Bacteroidales strains was generated (Figure 2A), based on the core genes (526 for Bacteroides and 1114 for Parabacteroides) which have 80% identity in amino acid sequence levels, using the concatenated alignment for a neighbor-joining tree in MEGA (https://www.megasoftware.net/).
Proteins belonging to the GH3 family were identified by BLAST alignment tools^92^ using the gshD protein sequence on 126 genomes from diverse gut bacteria (Table S3). In addition, 41 previously characterized bacterial GH3 were identified and included in the analysis (http://www.cazy.org/GH3_characterized.html). Protein sequence alignments of GH03 were performed with MAFFT, version 7^93^, and were manually adjusted using as a guide the residue-wise confidence scores generated by GUIDANCE2^94^. ProtTest 3 was used to select the best-fit model of amino acid replacement^95^. Phylogenetic relationships were inferred by maximum likelihood using RAxML-HPC2 on XSEDE conducted on the CIPRES project^96^ cluster at the San Diego Supercomputer Center. Phylogenetic trees were visualized using the interactive tree of life (iTOL)^97^.
Amino acid sequence alignment of Bacteroides uniformis ATCC 8492 gshD as reference was performed using NCBI BLAST service against 42 Bacteroides strains (Figure S3E).
To analyze the fitness of transposon mutants in the library, Illumina raw reads were subjected to the following bioinformatics the de-multiplexed reads were first trimmed using cutadapt v1.17^98^ to remove transposon and adapter sequences (“-g GACTTATCATCCAACCTGT -O 17 -e 0.2” and “-a ATACCACGAC -O 5 -e 0.1 -m 15”). These trimmed reads were then mapped to the corresponding Bacteroides reference genome using Bowtie v1.2.2^99^ with the setting “-n 3 -l 28 -e 120-best”. The derived insertion tallies were subjected to comparative analyses between glucose and arbutin using Transit v2.1.0^100^ under the “resampling” mode with the setting “-n TTR -iN 0.05 -iC 0.05”. As each experiment was performed in two biological replicates, to combine fold changes and p-values from two replicates, we used the smaller absolute value fold change and the higher p-value for each gene, which is more conservative than other methods such as Fisher’s method. As the Transit resampling method had a default p-value detection limit of 10e-4, for more significant hits that were marked as p-value = 0 we randomly assigned p-values <10e-4 for the sake of volcano plots.
All in vitro bacterial growth experiments (chemical utilization, chemical killing, enzyme activity, in vitro metabolism) encompass three technical triplicates and are representative of two biological duplicates. All in vivo disease model (experimental colitis and C. difficile infection) data is derived from at least two different experiments. For pairwise and two independent group comparison Student’s t test was used, while for multiple group comparison one-way and two-way ANOVA were performed. Statistical significance is indicated as non-significant (ns) > 0.05, *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Error bars represent the standard deviation (SD) or error (SE) of the mean as indicated per experiment.
SUPPLEMENTAL INFORMATION
Supplemental information can be found online at …