Chapter 8: Phase IV Metabolism: Gut Microbial Metabolism of Common GI Drugs

INTRODUCTION

The gastrointestinal tract is home to trillions of microorganisms including viruses, fungi and bacteria, that functionally impact nearly every aspect of host physiology. Bacteria have been the focus of a great deal of research ever since the advent of next generation sequencing technologies, which are culture-independent methods. The rapid development of sequencing technologies coupled with their decreasing cost has precipitated an explosion of research in the past three decades. Many discoveries have associated disease states with compositional differences of gut microbiota. Compositional changes in intestinal bacteria are now known to influence the initiation, progression, and treatment of diseases beyond those of the gastrointestinal system (covered in this Special Issue), e.g. cancer¹, atopic dermatitis², rheumatoid arthritis³, Parkinson’s⁴, etc. Follow-up mechanistic studies have the potential to uncover novel methods to improve precision medicine and thus enhance patient care.

In addition to their expanding utility as therapeutics (see Chapters 2, 3, 4 in this Special Issue), intestinal bacteria are also an emerging modifier of medications, affecting their efficacy and toxicity and thereby altering drug response in individuals. This burgeoning field termed “pharmacomicrobiomics” complements pharmacogenomics to examine interindividual variability in drug response. Unlike genetics, one’s microbiota are amenable to facile manipulation and modification and thus pharmacomicrobiomics poses an avenue to improve precision medicine and personalize drug response (Figure 1). This chapter delves deeper into how the gut microbiome impacts the absorption, distribution, metabolism and excretion of GI-relevant medications.

FIGURE 1:

Variability in drug response (efficacy and toxicity) result from variation in nonmodifiable factors (such as age, sex, and genetics), and modifiable factors such as environment, physiology, and microbiota. Microbiota are amenable to modification and manipulation, and thus can be leveraged to improve drug responses.

DRUG METABOLISM

Medications are chemicals that are not endogenously produced and thus are termed xenobiotics. Xenobiotics, often hydrophobic, require specialized biochemical processes to detoxify them to facilitate their efficient excretion. These biochemical processes are what comprise drug metabolism. Metabolic biotransformations generate drug metabolites that can be active, inactive, or toxic.

Most medications are actively altered by specialized classes of host enzymes which modify the compounds into forms that are easily excreted. Drug metabolism occurs in three phases, illustrated in Figure 2. Phase I, or functionalization consists of primarily oxidation, reduction or hydrolysis reactions that generate a polar metabolite. Some of these are further metabolized by addition of endogenous hydrophilic compounds such as glucuronic acid, sulfate, glutathione, etc. in Phase II, or conjugation reactions, which generate readily excreted hydrophilic conjugates. Phase III involves transport of the drug metabolites via specialized transport proteins that transport them in an ATP-dependent manner into or out of cells for excretion; Phase III is thought to be the final step of drug metabolism. Drug metabolism primarily occurs in the liver, although certain tissuespecific isoforms of drug metabolism are also found in the intestines, liver, pancreas, etc. ⁵.

FIGURE 2:

An overview of the metabolism of orally ingested medications (1) which are absorbed in the intestine (2) where the prodrug or parent drug first encounters intestinal bacteria (3). In the liver, medications are modified by host Phase I -III metabolism enzymes (4), then distributed through the blood stream (5) to the affected tissues where they exert their therapeutic effects (6). Drugs are excreted through the kidney for urinary elimination (7), or through the bile duct (8) into the intestines where they can be reabsorbed and subject to enterohepatic recirculation (9), or eliminated through feces; thus, Phase I/II drug metabolites encounter intestinal microbiota with a vast repertoire of biotransformation activities that carry out Phase IV metabolism (10).

Host drug metabolizing enzymes are complemented by a plethora of microbial enzymes that can further influence xenobiotic metabolism ^{1, 6-9}, in a process termed “Phase IV” drug metabolism ¹⁰. Recent findings also implicate Phase IV metabolism in the recycling of endobiotic substances e.g. hormones and neurotransmitters ^11-13. Emerging evidence implicates a rapidly expanding role for gut bacterial enzymes and Phase IV drug metabolism underlying interindividual variation in drug responses (toxicity and efficacy), a burgeoning field termed “pharmacomicrobiomics” ^{9, 14}.

Because Phase IV metabolism further alters byproducts of Phase I-III metabolism, it is capable of substantially altering drug responses. To date, Phase IV metabolism includes chemical reactions such as reduction, demethylation, decarboxylation, deamination, and deacylation, deconjugation of host Phase II metabolites and ring opening reactions^{6, 15}. Additional biochemical reactions that transform parent drugs or host-derived metabolites may be uncovered as this field of study continues to expand. Phase IV metabolism influences drug responses to a magnitude that we are only just beginning to uncover. Host-instrinsic factors such as age, sex, race, ethnicity, genotype of drug-metabolizing enzymes are known to influence drug responses; these factors are not amenable to change. In contrast, bacteria and their function that underlies Phase IV drug metabolism can be modified or modulated, and thus have the potential to improve drug responses. Although antibiotics pose the most direct method, this “scorched earth” approach can decimate beneficial functions of the gut microbiota. Sophisticated approaches to selectively and non-lethally modulate intestinal bacteria composition and function include utilization of live biotherapeutics, rationally designed bacterial consortia with defined functions, or small molecule inhibitors of specific bacterial enzymes mechanistically implicated in Phase IV metabolism.

The importance of bacterial enzymes in bioconversion of medications was first observed with the example of salicylazosulfapyridine¹⁶ wherein the authors reported no conversion in the feces of germ-free rats administered the drug. More recently, pivotal research has emerged demonstrating that 271 orally administered medications are modified by human gut bacteria ¹⁷. This has been further expanded using high-throughput personalized approaches in a “Microbiome-Derived Metabolism” Screen to uncover personalized differences in bacterial drug metabolism of small molecule drugs¹⁸. Several extensive reviews describe how microbiota transform medications for diverse indications ^{1, 6, 15, 19-22} . Literature describing the reciprocal, i.e. how non-antibiotic medications alter microbiota composition is also well described ^{1, 23-25}. Here, we synopsize the current knowledge of bacterial Phase IV drug metabolism between GI-relevant medications, and the implications on drug response, comprising both drug efficacy and drug toxicity.

5-AMINOSALICYLIC ACID

5-ASA (mesalamine, mesalazine) is an oral medication used for the treatment of mild to moderate ulcerative colitis. To limit its small intestinal absorption, it was developed as sulfasalazine, a prodrug wherein 5-ASA is linked via an azo bond to sulfapyridine. Because of adverse side effects elicited by the sulfapyridine moiety, it does not comprise the next generation of 5-ASA products e.g. basalazide and olsalazine, which respectively are 5-ASA azo-bonded to 4-aminobenzyle-beta-alanine, or two 5-ASA azo-bonded to each other.

5-ASA exerts therapeutic effects for IBD via local (rather than systemic) effects through multiple mechanisms ²⁶. Patients treated with 5-ASA achieve fecal (luminal) concentration of 5-ASA ranging from 1-100 mM, with a median of 30 mM observed in many patients. These 5-ASA concentrations are known to inhibit the recruitment and chemotaxis of leukocytes by scavenging reactive oxygen metabolites ²⁷, and inhibiting the generation of inflammatory mediators and key chemotactic lipids ^{28, 29}. In epithelial cells, 5-ASA reduces nitric oxide synthesis by downregulating the expression of inducible nitric oxide synthase ³⁰, and impedes proliferation via blocking mitosis and inducing caspase-3 dependent apoptosis ³¹. Furthermore, 5-ASA modulates signaling pathways that influence inflammation and immunity such as inhibiting NF-kB ^{32, 33} and activating the nuclear receptor peroxisome proliferator-activated receptor gamma isoform (PPARγ)³⁴. PPARγ is critical for amelioration of murine experimental colitis by reducing inflammation and restoring intestinal mitochondrial respiration³⁵. This reverts hypoxic conditions within the intestines and prevents respiration-dependent expansion of the facultative anaerobic Enterobacterales like E. coli, a known signature of dysbiosis in IBD.

Two discrete mechanisms drive the inefficacy of 5-ASA, each with distinct underlying classes of bacterial enzymes. The first mechanism involves limited conversion of prodrug into the therapeutically active 5-ASA moiety; biotransformation into an inactive metabolite is the second mechanism.

Azo-bonded therapeutics rely on the release of the immunomodulatory 5-ASA moiety by functional azoreductase enzymes such as those expressed by intact intestinal bacteria. A recent search of the 4644 genome sequences in the Unified Human Gastrointestinal Genomes collection led to the identification of 1958 azo-reducing bacterial species ³⁶. Systematic genome comparisons of confirmed and putative azo-reducing bacteria led to authors to hypothesize the presence of additional, uncharacterized azoreductase-encoding bacterial strains. Analyzing sequencing data from serial stool collections, the relative abundance of azoreducing species (both known and putative) was found to fluctuate over time, likely as a function of host diet or other environmental factors. The variable, fluctuating abundance of azoreductase-encoding intestinal bacteria could possibly contribute to the variable efficacy of azobonded therapeutics. Functional azoreductase activity was confirmed in multiple strains of Fusobacterium nucleatum, Bacteroides fragilis, and Clostridium clostridioforme. Confirmation of functional azoreductase activity in F. nucleatum could be leveraged to develop colon-targeted chemotherapeutics for colorectal cancer, with which F. nucleatum is frequently associated ³⁷.

A structural metagenomics approach³⁸ was used to define the atlas of human gut bacterial azoreductases using the known crystal structures of azoreductases expressed by E. coli, Pseudomonas putida, P. aeruginosa, and Klebsiella pneuomoniae ³⁹. This analysis identified five functionally critical and conserved residues which were used as “bait” to identify additional azoreductases by querying the Integrated Gene Catalog (IGC) from the Metagenomics of the Human Intestinal Tract (MetaHit) project⁴⁰ and the Integrative Human Microbiome Project (HMP2)’s Inflammatory Bowel Disease Meta-Omic Database (IBDMDB)⁴¹. From these two databases, the authors identified 196 unique azoreductases and their taxonomic provenance, and delineated biochemical parameters including catalytic efficiency using functional assays with purified enzymes.

Examination of the IBDMDB revealed a meager abundance of azoreductase-encoding bacterial species, irrespective of disease state. Azoreductase-encoding bacteria comprised less than 2% of all bacterial species in >80% of individuals analyzed; only 3% and 7% of non-IBD and IBD samples respectively had greater than 10% relative abundance of azoreductase-encoding taxa. This led the authors to conclude that the poor response of sulfasalazine is likely due to the “desolate landscape” of intestinal bacterial azoreductases. Thus, efficacy is directly linked to the presence of gut microbiota expressing functional azoreductases at adequate levels that convert prodrugs to enough 5-ASA to have therapeutic benefit for ulcerative colitis.

Prior to initiating treatment, quantifying the relative abundance of azoreductase -encoding bacteria or using functional assays to confirm azoreductase activity could aid in predicting treatment response to azobonded prodrugs. Coadministration of azoreductase-expressing non-inflammatory bacteria, or delivering recombinant azoreductases in colon specific drug delivery systems are complementary approaches to improve the efficacy of 5-ASA.

Even more than disease type or severity, 5-ASA emerged as the driver of differential metabolomes within the IBDMDB, underscoring the tremendous impact of medications in shaping the composition and functional output of the intestinal microbiota⁴². Metabolomic features included N-acetyl-5-ASA, a known inactive metabolite of 5-ASA generated in a cofactor- and pH-dependent manner by host cytosolic enzymes expressed in the intestinal mucosa⁴³. Additional putative acetylated forms of 5-ASA were also detected and were subsequently confirmed to be N-propionyl 5-ASA and N-butyryl-5-ASA⁴². Seeking the identity of enzymes responsible for producing these clinically inactive metabolites, the authors found that standard homologybased approaches were inadequate to identify the responsible acetyltransferases. Using paired metagenomic, metatranscriptomic and metabolomic data, the authors identified 12 acetyltransferase gene clusters belonging to the thiolase and acyl-CoA-N-acyltransferase protein families. Taxonomic analysis revealed that the acyl- CoA-N-Acyltransferase genes were nearly all derived from phylum Bacteroidetes with the exception of F. prausnitzii. Thiolase enzymes were exclusively encoded by members of the phylum Firmicutes. Heterologous expression, biochemical characterization, and structural elucidation of purified proteins uncovered substrate specificity and enzyme efficiency. Metagenomic carriage of 5-ASA metabolizing enzymes strongly predicted 5-ASA treatment failure, as defined by escalation to steroid usage.

Given this level of characterization, it is only a matter of time before selective, non-lethal inhibitors of this class of enzyme are developed. With careful preclinical evaluation, it is likely that such selective inhibitors can be developed into drugs used as an adjuvant to 5-ASA. Such compounds could extend the usage of 5-ASA into previously non-responsive users and may also reduce the administered dosage of 5-ASA-containing medications.

METHOTREXATE

Methotrexate is an immunosuppressant that is an established therapy for patients who have steroiddependent Crohn’s disease; it can also be used in combination with anti-TNF biologics. Although there is robust clinical evidence for its efficacy, it is only rarely used for CD⁴⁴.

Oral methotrexate is absorbed in the small intestine by the reduced folate carrier RFC1, and primarily transferred into RBC, WBC, synovial cells and liver cells. In RBCs, polyglutamate groups are added to methotrexate by the enzyme folylypolyglutamyl synthetase. Methotrexate polyglutamate is an analog of dihydrofolate and competitively inhibits dihydrofolate reductase (DHFR), an essential enzyme responsible for folate metabolism. Thus, as an antimetabolite methotrexate inhibits the synthesis of purines and pyrimidines, the building blocks of DNA and RNA. Methotrexate polyglutamate also potently inhibits aminoimidazole-4-carboxamide ribonucleotide (AICAR) transformylase (ATIC), causing intracellular AICAR buildup and triggering adenosine release. Free adenosine can ligate to inhibitory GPCRs on multiple inflammatory cells including neutrophils, macrophages and T cells, thereby suppressing inflammation and immunity pathways ⁴⁵.

It was first observed that methotrexate can also be metabolized by gut bacteria in 1969⁴⁶ . Subsequently, bacterial carboxypeptidase glutamate 2 (CPDG2) such as that expressed by Pseudomonas stutzeri ⁴⁷ was found to cleave terminal glutamates from methotrexate, generating the inactive metabolites 4-amino-4-deoxy-N-methylpteroic acid (DAMPA) and 7-hydroxy-DAMPA ⁴⁸. Formation of these metabolites was also observed in Lacticaseibacillus casei⁴⁹. More recently, DAMPA generation was shown to positively correlate with high levels of glutamate and abundance of Prevotellaceae and Anaeroplasmataceae, collectively suggesting these bacteria encode functional CPDG2 ⁵⁰. The E. coli enzyme p-aminobenzyl-glutamate hydrolase (PGH) can also catalyze this reaction ⁵¹.

Since polyglutamated methotrexate remains sequestered within eukaryotic cells, the functional significance of bacterial glutamate removal is unclear. However, because methotrexate can also target the bacterial analog of DHFR, this function is likely critical for bacterial survival, and accordingly, glutamate hydrolysis by CPDG2 and PGH can be considered a bacterial detoxification process. The presence or absence of glutamate hydrolysis potential may be at the heart of Nayak et al’s obseration that methotrexate treatment reduces the relative abundance of B. thetaiotaomicron and B. fragilis in mouse models ⁵². It is possible that these Bacteroides do not encode functional CPDG2 enzymes, rendering them susceptible to methotrexate’s growth inhibitory properties.

The fact remains that CPDG2 is a highly efficient enzyme to the extent that recombinant CPDG2 produced in E. coli is clinically used an FDA-approved rescue therapy called glucarpidase for high dose methotrexate-related toxicity ⁵³. Glucarpidase lowers methotrexate levels by 97% within 15 minutes of administration. This is one of the best examples of Phase IV metabolism being implemented for improved drug response, in this case, for reducing drug toxicity.

CALCINEURIN INHIBITORS

Cyclosporine A and tacrolimus inhibit the eukaryotic Ca2+ and calmodulin dependent serine/threonine protein phosphatase calcineurin, thereby preventing dephosphorylation of the nuclear factor of activated T cells (NF-AT); this blocks T cell activation and prevent the transcriptional activation of pro-inflammatory cytokines. Oral or intravenously administered cyclosporine A and tacrolimus are well known immunosuppressants that prevent rejection of organ transplants. Although not a part of routine clinical practice, a few studies indicate limited efficacy of tacrolimus for inducing remission in corticosteroid-refractory ulcerative colitis ⁵⁴.

Tacrolimus is macrolide used in combination with mycophenolate in kidney transplantation, and its serum level is closely monitored. As with many oral medications, its pharmacokinetics are influenced by gut microbiota ^{55, 56}. Depleting microbiota with antibiotics leads to Constitutive Androstane Receptor (CAR)-mediated downregulation of the ABCB1 efflux transporter, which reduces serum tacrolimus levels to a third less than conventional mice with intact microbiota ⁵⁶. Polar microbial metabolites are thought to regulate CAR expression, although the specific metabolite is not known. Thus, transcriptional regulation of host drug metabolism enzymes is an additional mechanism through which microbiota impact host Phase I-III metabolism.

Pure cultures ^{17, 57} of representative gut microbiota including Clostridium spp. and F. prausnitzii and human stool samples ⁵⁷ consume up to 85% of tacrolimus ¹⁷, or generate microbial metabolites of tacrolimus⁵⁷. In contrast, the poorly soluble cyclic peptide cyclosporine A is highly stable when anaerobically incubated with human fecal lysates ⁵⁸. Thus, the functional output of the microbiota may be an important consideration when including using calcineurin inhibitors. Tacrolimus may have poor efficacy due to bacterial metabolism, which cyclosporine is spared from.

IRINOTECAN

Irinotecan is an anticancer drug used either alone or in combination with fluorouracil for the treatment of colorectal cancer. Irinotecan is a topoisomerase I inhibitor, and as such induces DNA damage and subsequent cell death in rapidly proliferating cells such as tumor cells and intestinal stem cells⁵⁹. Irinotecan usage frequently results in side effects of diarrhea which can often be dose limiting. Irinotecan is administered intravenously, where it is converted into the active metabolite SN38 by plasma carboxylesterases. SN38 is hepatically detoxified by Phase II conjugation to form the inactive SN38-glucuronide which is biliary excreted into the intestines, and is encountered by a class of bacterial enzymes called β-glucuronidase⁶⁰. Expressed by members of every intestinal bacterial phylum⁶¹, β-glucuronidase are carbon scavenging enzymes that hydrolyze diverse glucuronide conjugates resulting from Phase II metabolism of endo- and xenobiotics ⁶²; glucuronic acid enters the bacterial Entner-Doudoroff pathway and is subsequently shunted into the TCA cycle to generate ATP ⁶³.

Blocking the reactivation of SN38-glucuronide to SN38 alleviates diarrhea and weight loss in mouse models ⁶⁴, and does not impede the anticancer efficacy of irinotecan ⁶⁵. Small molecule inhibitors of bacterial β-glucuronidase can selectively target the bacterial ortholog, whilst sparing eukaryotic β-glucuronidase^{66, 67}. Importantly, these inhibitors do not have reported bactericidal activities ^{64, 68}, thus they maintain the compositional integrity of the microbiome ⁶⁵. Irinotecan’s anticancer efficacy in murine tumor models remained unimpeded with the usage of selective bacterial β-glucuronidase inhibitors, and reduced chemotherapy-related diarrhea ⁶⁵. Thus, specifically targeting bacterial enzymes relevant to drug metabolism may increase the usage of medications characterized by dose-limiting toxicity.

MYCOPHENOLATE MOFETIL

Mycophenolate mofetil (MMF) is primarily used for immunosuppression in solid organ transplantation to avoid rejection. It is usually combined with calcineurin inhibitors, and this combination is highly successful in avoiding rejection⁶⁹ . However, in about half of all patients receiving it, MMF is known for its gastrointestinal side effects such as diarrhea or constipation, nausea, abdominal pain etc., sometimes necessitating treatment cessation, which can eventually lead to transplant failure ⁷⁰.

MMF is a prodrug that is hydrolyzed to active mycophenolic acid (MPA) which is subsequently inactivated to MPA-glucuronide (MPAG) and acyl-MPAG by hepatic Phase II metabolism wherein uridine-diphosphate-glucuronosyltransferase (UGT) enzymes conjugate MMF to glucuronic acid. MPAG and acyl-MPAG are primarily renally excreted, but 10% is hepatically excreted into the gastrointestinal tract. Intestinal microbiota were found to drive MMF gastrointestinal toxicity⁷¹, which was alleviated with the antibiotic vancomycin ⁷². Phase IV metabolism by bacterial β-glucuronidase reactivation of MPAG was uncovered as the culprit for these side effects. The specific structural features of β-glucuronidases that reactivate MPAG have been elucidated using activity-based proteomics and metagenomics. Β-glucuronidases that bind the cofactor flavin mononucleotide preferentially reactivate MPAG and are thought to drive mycophenolate-induced gastrointestinal toxicity⁷³.

MMF was recently reported to induce the generation of mitochondrial reactive oxygen species, mitochondrial depolarization, reduced expression of tight junction protein, and apoptosis in intestinal epithelial cell lines; barrier defects were also observed in a mouse model ⁷⁴. B-glucuronidase reactivation of MPAG may increase local intraluminal MMF concentrations, causing topical injury akin that ultimately results in the development of diarrhea. Selective inhibitors that target flavin mononucleotide-binding β-glucuronidase may help alleviate mycophenolate-induced gastrointestinal toxicity.