A host–gut microbial amino acid co-metabolite, p-cresol glucuronide, promotes blood–brain barrier integrity in vivo

ABSTRACT The sequential activity of gut microbial and host processes can exert a powerful modulatory influence on dietary components, as exemplified by the metabolism of the amino acids tyrosine and phenylalanine to p-cresol by gut microbes, and then to p-cresol glucuronide (pCG) by host enzymes. Although such glucuronide conjugates are classically thought to be biologically inert, there is accumulating evidence that this may not always be the case. We investigated the activity of pCG, studying its interactions with the cerebral vasculature and the brain in vitro and in vivo. Male C57Bl/6 J mice were used to assess blood–brain barrier (BBB) permeability and whole-brain transcriptomic changes in response to pCG treatment. Effects were then further explored using the human cerebromicrovascular endothelial cell line hCMEC/D3, assessing paracellular permeability, transendothelial electrical resistance and barrier protein expression. Mice exposed to pCG showed reduced BBB permeability and significant changes in whole-brain transcriptome expression. Surprisingly, treatment of hCMEC/D3 cells with pCG had no notable effects until co-administered with bacterial lipopolysaccharide, at which point it was able to prevent the permeabilizing effects of endotoxin. Further analysis suggested that pCG acts as an antagonist at the principal lipopolysaccharide receptor TLR4. The amino acid phase II metabolic product pCG is biologically active at the BBB, antagonizing the effects of constitutively circulating lipopolysaccharide. These data add to the growing literature showing glucuronide conjugates to be more than merely metabolic waste products and highlight the complexity of gut microbe to host communication pathways underlying the gut–brain axis.


Introduction
The gut microbiota can influence a wide range of physiological systems and pathological conditions, ranging from cardiovascular function 1 to metabolic disease 2 to cancer, 3 but communication between the gut microbiota and the brain is one of the most active areas of investigation. Interactions between the gut microbes and the brain have been identified in both cognitive functions such as anxiety 4 and sociability, 5 and in overt disease states including Parkinson's disease, 6 multiple sclerosis 7 and dementia. 8 Increasingly, evidence indicates a central role for microbe-derived metabolites in this dialogue, acting primarily through three routes: directly on enteric and gut extrinsic neural pathways, by modification of enteroendocrine signaling or, as we and others have shown, via the circulation and interactions with the blood-brain barrier (BBB). 9,10 Notably, while many microbe-derived metabolites circulate in their native form, many others are subjected to host metabolic enzymemediated biotransformation, thereby altering their biological activities.
A good example of this lies in the metabolism of the microbial product p-cresol (4-methylphenol). This molecule is produced by bacterial fermentation of dietary tyrosine and phenylalanine in the colon, 11 and passes through the gut epithelium into the portal vasculature. Notably, p-cresol undergoes extensive conjugation both in enterocytes 12 and by hepatic enzymes upon reaching the liver, 13 such that the parent compound is essentially not found within the systemic circulation, 14 existing rather as p-cresol sulfate (pCS) with a smaller component as p-cresol glucuronide (pCG). 15, 16 While pCS has been extensively studied in light of its role as a major uremic toxin, 15 the potential biological actions of pCG have received far less attention, although high concentrations of the metabolite have been shown to potentiate pCS-induced neutrophil oxidative burst activity in vitro 17 and to slightly raise fasting glycemia in mice. 18 What role the molecule might play at more physiologically representative concentrations is unclear.
Classically, glucuronidation is considered as part of the phase II metabolic pathways, with the actions of the numerous UDP-glucuronosyltransferases serving to enhance renal clearance of parent compounds. 19 More recent evidence suggests that this form of conjugation may not always be a neutralizing process however. Notably, glucuronide derivatives of morphine, ethanol and estradiol have been shown to act as agonists of the Toll-like receptor 4 (TLR4) complex, promoting allodynia and inflammation upon spinal cord administration, [20][21][22][23] actions quite distinct from those of their parent compounds, and highlighting the potential of glucuronide conjugates to gain potentially significant new pharmacological activity. Whether the same can be said for microbederived compounds present in the circulation is unclear, with studies into this question hindered by difficulties in obtaining pure molecules for study. We have recently established a novel pathway for chemical synthesis of pCG, 24 and here we employ a combined in vitro/in vivo approach to identify the actions of this compound on the cerebral vasculature and the brain.

Animals
Wild-type male C57Bl/6 J mice aged between 7 and 8 weeks (Charles River UK Ltd., Margate, UK) were used for all experiments. Mice were kept under a 12 h:12 h light:dark regime, with ad libitum access to standard chow (Envigo Ltd., UK) and drinking water; all animals were acclimatized to the holding facility environment for 1 week prior to experimentation. Animals were treated as described below and killed by transcardial perfusion with ice-cold saline under pentobarbitone anesthesia. All experiments were approved by the QMUL Animal Welfare and Ethical Review Board and were performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments and with the UK Animals (Scientific Procedures) Act, 1986, under Project License PFA5C4F4F.

In vivo BBB permeability analysis
Mice (n = 5-6 per group) were injected i.p. with 1 mg/kg body weight pCG in 100 µl saline vehicle, a dose calculated to approximately double circulating concentrations, 18 followed 2 h or 6 h later by assessment of Evans blue extravasation. One hour before assessment animals were injected i.p. with 100 µl of a 2% (w/v) solution of Evans blue dye in 0.9% saline. Dye was permitted to circulate for 1 h before animals were transcardially perfused with 0.9% saline at 4°C to remove dye remaining in the vasculature. Blood samples were allowed to coagulate at 37°C for 15 minutes prior to centrifugation at 800 g for 10 minutes to separate serum. Brains were removed and homogenized by maceration in 0.1 M phosphate-buffered saline. Suspended macromolecules were precipitated by incubation with 60% trichloroacetic acid, and dye content of resulting supernatants was detected using a CLARIOstar spectrophotometer (BMG Labtech GmbH, Germany) alongside a standard curve of defined concentrations of Evans blue in the same buffer. Brain Evans' blue content was expressed as µg of dye per mg of brain tissue, normalized to circulating serum concentrations.

RNAseq data analyses
Processing of mouse brain samples (taken at 2 h) and RNA extraction were performed as described previously. 10 RNA samples (n = 6 pCG, n = 6 control) were sent to Macrogen Inc. (Seoul, Republic of Korea) where they were subject to quality checks (RIN analysis); libraries were prepared (TruSeq Stranded mRNA LT Sample Prep Kit) for pairedend (2× 100 nt) sequencing on an Illumina HiSeq 4000 apparatus. Three pCG-treated and three saline controls produced data of poor quality so were excluded from analyses after quality checks and consideration of Macrogen quality reports; as such, samples SG1, SG3, SG5, CG2, CG5 and CG6 were used for all analyses described from hereon. Raw RNAseq (fastq) sequence data were processed in house as described previously. 10 Entrez gene identifiers were converted to gene symbols using Mus musculus annotations downloaded from NCBI on 4 January 2021; only those genes with valid Entrez gene identifiers were retained in analyses. Significantly differentially expressed genes (FDR P < .05) identified using DESeq2 v1.22.1 25 were analyzed by mouse KEGG pathway over-representation analysis using Enrichr 26,27 and manual curation. Signaling Pathway Impact Analysis (SPIA v v1.22.1) 28 was used to determine whether Kyoto Encyclopedia of Genes and Genomes (KEGG) Mus musculus pathways (downloaded on 22 December 2021) were activated or inhibited in mouse brain cells exposed to pCG. RNAseq data have been deposited in ArrayExpress under accession number E-MTAB-11340. Normalized and log 2 -transformed RNAseq data are available as Supplementary Material (Suppl. Table 1).

Cell culture
The human cerebromicrovascular endothelial cell line hCMEC/D3 was maintained and treated as described previously. 29 Cells were cultured to confluency in complete endothelial cell growth medium MV2 (PromoCell GmbH, Germany), whereupon VEGF was removed and cells were further cultured for a minimum of 4 d to enable intercellular tight junction formation prior to experimentation. All cell cultures were used between passages 28 and 32 to ensure retention of appropriate endothelial characteristics. 30

Cell survival analysis
The potential for pCG-induced cytotoxicity was assessed using the MTT assay. Briefly, cells were treated with pCG for 24 h (0.1, 1, 10, 100 µM), prior to administration of MTT at 500 μg/ml. Cells were incubated at 37°C for 2 h, medium was removed and resulting crystals were solubilized by incubation for 2 minutes in dimethyl sulfoxide. Absorbance was read at 540 nm using a CLARIOstar spectrophotometer (BMG Labtech, Ortenberg, Germany), with a reference wavelength at 570 nm.

In vitro barrier function assessments
Paracellular permeability and transendothelial electrical resistance were measured on 100% confluent cultures polarized by growth on 24-well plate polyethylene terephthalate (PET) transwell inserts (surface area: 0.33 cm 2 , pore size: 0.4 μm; Appleton Woods, UK) previously coated with calf-skin collagen (15 µg/cm 2 and fibronectin 3 µg/cm 2 ; both Merck Life Science UK Ltd.). The permeability of hCMEC/D3 cell monolayers to 70 kDa FITC-dextran (2 mg/ml) was measured as described previously. 31,32 Transendothelial electrical resistance (TEER) measurements were performed using a Millicell ERS-2 Voltohmmeter (Millipore, Watford, UK) and were expressed as Ω.cm 2 . In all cases, values obtained from cell-free inserts similarly coated with collagen and fibronectin were subtracted from the total values.

Immunofluorescence
Confluent hCMEC/D3 monolayers grown on transwell inserts as described above were fixed by immersion in 2% formaldehyde in 0.1 M PBS for 10 minutes at room temperature. Cells were immunostained according to standard protocols 33 using a primary rabbit anti-human antibody directed against zona occludens-1 (ZO-1; 1:100, Thermo Fisher Scientific, UK) and a AF488-conjugated secondary goat anti-rabbit antibody (1:500, Thermo Fisher Scientific, UK) or AF488-conjugated phalloidin (100 nM; Cytoskeleton Inc., Denver, USA). Nuclei were counterstained with DAPI (50 ng/ml; Merck Life Science UK Ltd., UK). Images were captured using an LSM880 confocal laser scanning microscope (Carl Zeiss Ltd., Cambridge, UK) fitted with 405 nm and 488 nm lasers and a 63x oil immersion objective lens (NA, 1.4 mm, working distance, 0.17 mm). Images were captured with ZEN imaging software (Carl Zeiss Ltd., UK) and analyzed using ImageJ 1.53c (National Institutes of Health, USA).

Efflux transporter assays
Activity of the major efflux transporters P-glycoprotein and BCRP was determined using commercially available assays (PREDEASY™ ATPase Assay Kits, Solvo Biotechnology Inc., Budapest, Hungary), performed according to the manufacturer's instructions. Stepwise dose-response curves centered around reported physiological circulating concentrations of pCG (12.3 nM-27 µM) were constructed (n = 4) to investigate stimulatory and inhibitory effects upon transporter activity.

Statistical analysis
Sample sizes were calculated to detect differences of 15% or more with a power of 0.85 and α set at 5%, calculations being informed by previously published data. 10,29 Experimental data are expressed as mean ± SEM, with n = 6-9 independent experiments for all studies. In all cases, normality of distribution was established using the Shapiro-Wilk test, followed by analysis with two-tailed Student's t-tests to compare two groups or, for multiple comparison analysis, one-or two-way ANOVA followed by post hoc analysis by either Dunnett's test (for dose-response experiments) or Tukey's HSD test (all other comparisons). A P value of less than or equal to 5% was considered significant.

pCG modulates BBB integrity and the whole-brain transcriptome in vivo
A defining property of the cerebral vasculature is the existence of a tight barrier function limiting passage of soluble molecules into the brain parenchyma, the BBB. We examined whether exposure to increased levels of pCG could affect BBB integrity in vivo, assessed by monitoring extravasation of administered Evans blue dye into the brain tissue. Treatment of mice with 1 mg/kg pCG i.p. (a dose known to approximately double baseline serum concentrations 18 ) caused a significant reduction in entry of Evans blue to the brain tissue, by approximately 50% within 6 h of treatment (Figure 2a).
To investigate the mechanism(s) underlying this action of pCG, we performed bulk RNAseq analysis of brain tissue from animals treated for 2 h with 1 mg/kg pCG i.p., identifying 7702 significantly differentially expressed genes (Figure 2b; Suppl. Table 2), of which 1658 and 1433 showed greater than twofold up-or downregulation respectively following correction for  (d) Biological processes associated with genes found to be significantly and ≥ 2-fold upregulated (n = 1433) or downregulated (n = 1658) upon exposure of mice to pCG. Images are based on Enrichr P value ranking from GO analysis, the lighter the color and longer the bar, the more is significant is the result, as determined by rank-based ranking; only the top ten results are shown in each case. (e) Volcano plot showing significantly (P FDR < 0.05) differentially expressed BBB-relevant genes (red dots). (f) Biological processes associated with BBB-relevant genes found to be significantly upregulated (n = 78) or downregulated (n = 24) upon exposure of mice to pCG. Images are based on Enrichr P value ranking from GO analysis, the lighter the color and longer the bar, the more is significant is the result, as determined by rank-based ranking; only the top ten results are shown in each case. multiple testing (Figure 2c). Analysis of gene ontology categories over-represented within these gene sets using Enrichr 26,27,34 identified a number of different biological process ontologies exhibiting significant changes (Figure 2d), with ontologies relating to axon generation and extracellular matrix organization being notably upregulated, while pathways associated with protein synthesis and ribosomal activity were downregulated. SPIA of all differentially expressed genes revealed several significantly over-represented KEGG pathways (Suppl. Figure 2a), notably indicating pathways associated with growth factor/transcription factor signaling and the response to infection as being activated, while pathways associated with cellular degradation and metabolism were inhibited (Suppl. Figure 2b).
To specifically examine the interactions of pCG with the BBB, we further interrogated transcriptomic changes induced by pCG treatment by comparison with a defined list of 203 known BBBrelevant genes, 29 identifying a total of 78 upregulated and 24 downregulated number of genes exhibiting statistically significant regulation (Figure 2e; Suppl. Table 3). Examination of associated biological process gene ontologies here identified clear upregulation in multiple transport pathways and suppression of inflammatory processes (Figure 2f). Individual gene-level analysis of differentially expressed transporter systems identified enhanced expression of a wide range of nutrient uptake transporters, whereas in contrast only the transporters for myo-inositol and transferrin and aquaporin-4 were significantly downregulated (Table 1).

pCG has limited direct effects upon an in vitro model of the BBB
Following these in silico analyses, we sought to investigate the biological pathway(s) through which pCG affected the BBB, using a well-established model of the human brain capillary endothelium, the hCMEC/D3 cell line. 35 Initial assessment of potential pCG toxicity using the MTT assay showed no effects on cell survival following 24 h treatment of hCMEC/D3 cells with concentrations of up to 100 µM pCG (Figure 3a). Similarly, as βglucuronidase is known to be present in the cerebral endothelium, albeit at low levels, 36 it is plausible that the effects of pCG may be caused by reversion to its parent p-cresol molecule. However, exposure of hCMEC/D3 cells to p-cresol itself (5 µM, 24 h) caused a significant increase in paracellular permeability to a 70 kDa FITC-dextran tracer and accompanying reduction in transendothelial electrical resistance (Suppl. Figure 1), indicating an abrogation of BBB integrity, essentially the opposite of our in vivo findings.
We then examined the ability of pCG itself to affect hCMEC/D3 monolayer barrier integrity. Exposure of hCMEC/D3 cells for 24 h to pCG caused a dose-dependent increase in transendothelial electrical resistance, becoming statistically significant with 10 µM and 100 µM concentrations (Figure 3b), but this was not accompanied by any change in permeability to the 70 kDa FITC-dextran tracer (Figure 3c). Microscopic examination of the tight junction component ZO-1 and the actin cytoskeleton to which it binds similarly revealed little effect of pCG upon the endothelial cells (Figure 3d). Similarly, pCG treatment had no effect on the expression of the key tight junction molecule, occludin (Figure 3e-f).
As our transcriptomic data indicated upregulation of multiple nutrient uptake transporter genes, we investigated whether pCG could also affect two of the principal efflux transport systems of the BBB, namely P-glycoprotein and Breast Cancer Resistance Protein (BCRP). While pCG had no effect on cell surface P-glycoprotein expression at any dose tested (Figure 3g), exposure of cells to 100 μM pCG did cause a slight, but significant reduction in BCRP expression (Figure 3j). Neither transporter was activated or inhibited by the presence of pCG at any concentrations tested ( Figure  3h-i, k-l).

pCG antagonizes the BBB-permeabilizing actions of bacterial LPS
In light of the contrast between the limited effects of pCG seen in our in vitro BBB model, we sought alternative explanations for the more pronounced effects of the metabolite seen in vivo, taking a lead from the indicated suppression of inflammatory process ontologies. Several structurally dissimilar glucuronidated molecules interact with the bacterial LPS receptor TLR4 and its heterodimeric partner MD-2, including morphine-3glucuronide, 22 ethyl glucuronide 20 and a range of steroid hormone glucuronide conjugates, 21 leading us to hypothesize that this may also be the case for pCG. LPS is known to circulate at low, but nonzero, levels in normal mice and humans, 37,38 and is known to enhance BBB permeability in vitro and in vivo 10 ; hence, we investigated the interaction between it and pCG in our model system.
We initially confirmed that hCMEC/D3 cells express TLR4 and its accessory proteins MD-2 and CD14 (Suppl. Figure 3a-c). Treatment of hCMEC/D3 cells with LPS (Porphyromonas gingivalis, 10 ng/ml, 24 h) significantly enhanced paracellular permeability to a 70 kDa FITC-dextran tracer (Figure 4a) and reduced transendothelial electrical resistance (Figure 4b), effects that were both prevented by 30 minutes pre-treatment with pCG (1 µM). Similar treatment of endothelial monolayers with LPS disrupted circumferential localization of the key tight junction molecule ZO-1 (Figure 4c) and induced the appearance of large numbers of cytosolic actin fibers (Figure 4d), both of which features were prevented by 30 minutes pre-treatment with pCG (1 µM). This effect did not appear to be due to down-regulation of TLR4 or its accessory molecules CD14 or MD-2 on the surface of the endothelial cells (Figure 4e-g), suggesting pCG may be acting as an antagonist at this receptor.
To provide further support for this hypothesis, we investigated whether pCG could functionally antagonize an alternative and unrelated effect of LPS treatment, upregulation of surface expression of the integrin CD11b on the human monocyte cell line THP-1 (Suppl. Figure 3). Treatment of THP-1 cells with LPS (20 ng/ml, 24 h) significantly upregulated surface CD11b expression, an effect prevented by 30 minutes pre-treatment with 1 µM pCG (Figure 4g), confirming the ability of pCG to antagonize LPS-induced signaling responses in distinct circumstances.

Discussion
Glucuronidation is a key stage in phase II metabolism and clearance of endogenous and exogenous molecules and has long been investigated in this regard. Much is now known about the various UDP-glucuronosyltransferases responsible for glucuronidation at different sites in the body, 19 but the   biological actions of glucuronide compounds once they have been formed are rather less understood. In most cases, glucuronide conjugates have been considered as biologically inactive and simply destined for renal elimination, but our data add to the steadily building picture that this may not be universally true. Notably, glucuronide derivatives of morphine, ethanol and estradiol have been shown to act as agonists of the TLR4 complex, promoting allodynia and inflammation upon spinal cord administration. [20][21][22] Our data add the tyrosine/phenylalanine metabolite pCG to the list of glucuronide conjugates that can interact with TLR4 signaling, but with the marked difference that, in contrast to the other known activating agents, pCG is a functional antagonist and prevents the permeabilizing effects of bacterial endotoxin exposure upon the BBB. While pCG has long been known to circulate in the blood, its physiological and potentially pathological actions have remained somewhat elusive. Our description of an antagonistic action of pCG upon the principal LPS receptor, the TLR4 complex, indicates an anti-inflammatory effect of the molecule and suggests that it may, at least at physiological concentrations, aid cerebrovascular resilience to the damaging effects of LPS exposure, 33,39 thereby protecting against the development of sickness behaviors. 40 However, pCG is also well known as a potential uremic toxin. 41 Exposure at levels seen in patients undergoing hemodialysis has been reported to directly evoke a low level of endothelial reactive oxygen species release, 42 to impair endothelial succinate dehydrogenase function 43 and to potentiate some of the inflammatory effects of pCS upon leukocytes 17 and the endothelium. 44 Notably, individuals with renal dysfunction have increased susceptibility to bacterial infection [45][46][47] despite the presence of chronic low-grade leukocyte activation. 48 Given that the majority of circulating pCG in such patients is freely available 49 and thus presumably able to interact with TLR4, the potential contribution that such antagonism by pCG makes to masking signs of bacterial infection bears further investigation.
Beyond emphasizing the need to look again at glucuronide conjugates as potential biological actors, our data also highlight the position of the cerebral vasculature and the BBB as targets for the actions of microbial metabolites and an important aspect of the gut-brain axis. A range of gut microbe-derived metabolites, including shortchain fatty acids, methylamines and, here, cresols, have now been shown to regulate BBB integrity-9,10,29 in vivo. That such structurally diverse molecules can modulate BBB function epitomizes the complexity of the gut microbiome-brain axis, but also emphasizes the importance of systematic investigation of this communication pathway. Moreover, as pCG is a product of both gut microbial and host enzymatic co-metabolism of aromatic amino acids, our data emphasize the need to consider both microbial and host systems in regulating gut microbe-brain communication.
With over 200 known microbe-derived metabolites present in the human circulation, 50 there is clearly much still to learn about how they might influence the cerebral vasculature and their implications for health, aging and disease.
A notable feature of the gut microbiota is its exquisite sensitivity to dietary change, 51 with the makeup of the gut microbiome changing in a matter of weeks of exposure to a novel diet. 52 As diet is also known to be a major risk factor for cerebrovascular and neurological health, 53 studying the links between diet, gut microbe-host co-metabolites and the BBB may be instructive in understanding the pathogenesis of and, potentially, treatment for neurovascular disease. In particular, our study of the simple phenolic glucuronide pCG may be of relevance when it comes to understanding the actions of its more chemically complex relatives, the dietary polyphenol glucuronides. Diets supplemented with foods containing polyphenols have been shown to improve cerebral blood flow and neurovascular coupling in humans, [54][55][56][57][58] and rodent studies have revealed polyphenols to protect against ischemia or trauma-induced BBB integrity damage. [59][60][61][62] Notably, however, such dietary polyphenols are primarily found in the circulation as conjugates: sulfates, methylates, and, conspicuously, glucuronides. 63 At the least, the presence of high levels of glucuronide conjugates suggest that these agents should be investigated as potential mediators of the beneficial effects of dietary polyphenols upon the cerebral vasculature.

Conclusion
Here, we show that pCG, thought to be a relatively inert product of gut microbe-host enzyme co-metabolism, can influence the BBB and potentially immune cell activity through functional antagonism at the TLR4 complex. Notably, pCG administered within its physiological range was active in vivo, enhancing permeability barrier function of the cerebromicrovascular endothelium and further inducing significant transcriptional change throughout the brain. These findings extend our understanding of the role of glucuronide conjugates as not only targets for renal elimination but also as potent biological actors in their own right. Moreover, our data emphasize the importance of considering both microbial and host metabolic processes in understanding the mechanism(s) of communication that underlie the gut microbiota-brain axis.

Statements and declarations
The authors have no competing interests to declare that are relevant to the content of this article. Author Contributions: AVS prepared and purified pCG, TBAK, SNS and SM performed cell culture and animal experiments, LH performed bioinformatic analyses. AVS, LH and SM wrote the manuscript. All authors have read and approved the final version of the manuscript.