Engineering human Gut-on-a-Chip culturomics for predictive pharmacomicrobiomics of first-pass metabolism Authors: Itsuma Nagao, Hyun Jung Kim Published: 2026 ABSTRACT AbstractFor orally administered drugs, intestinal first-pass metabolism influences systemic exposure and accounts for inter-individual pharmacokinetic differences. However, the mechanistic roles of gut microbiome-mediated biotransformation and patient-specific intestinal variability remain underrepresented in current regulatory frameworks. This Perspective explores how human-relevant culturomics platforms, including Gut-on-a-Chip microphysiological systems, allow for quantitative analysis of host-microbiome-drug interactions. We also discuss the potential of predictive intestinal ecosystem models for personalized pharmacomicrobiomics and next-generation translational drug development.Subject terms: Computational biology and bioinformatics, Drug discovery, Gastroenterology, Microbiology, Systems biology Intestinal first-pass metabolism Per-oral medications, administered as pills, capsules, or liquids, represent the most widely used route of drug delivery in modern medicine. Requiring no injections or clinical procedures, they can be self-administered at home, are straightforward to manufacture at scale, and are readily distributed through global supply chains. These practical advantages have made oral administration the default strategy in pharmaceutical development, consistently preferred by both patients and industry alike1,2. Yet despite this ubiquity, accurately predicting how an orally administered drug behaves in the body, how it is absorbed, distributed, metabolized, and eliminated, remains one of the most persistent and consequential challenges in contemporary pharmacology. Pharmacokinetic/pharmacodynamic (PK/PD) modeling and simulation are now integral to drug development, integrating in vitro, in vivo, and computational data to estimate dose-exposure-response relationships, first-in-human dosing, and therapeutic windows. However, the predictive accuracy of PK/PD models depends on biologically relevant experimental inputs for intestinal absorption, metabolism, and transport. Improving these experimental inputs is essential for the rational design of safe and effective medicines. Orally administered drugs are metabolized and absorbed in the intestine, where they undergo additional metabolism in the liver prior to entering systemic circulation3. This drug metabolism process involves coordinated Phase I biotransformation, including cytochrome P450 (CYP)-mediated oxidation as well as non-CYP enzymes such as flavin-containing monooxygenases (FMOs), esterases, and dehydrogenases, followed by Phase II conjugation reactions catalyzed by uridine 5′-diphospho-glucuronosyltransferases (UGTs), sulfotransferases (SULTs), glutathione-S-transferases (GSTs), and N-acetyltransferases (NATs). These pathways operate in parallel and collectively determine intestinal metabolic capacity. Among these pathways, CYP-mediated oxidation represents the dominant enzymatic component of first-pass drug metabolism, with previous analyses reporting CYP involvement in ~70% of metabolized drugs, compared with ~30% for non-CYP enzymes, including UGTs at ~15% and other individual enzymes at less than 5%4,5. Phase III transporters, including members of ATP-binding cassette (ABC) transporter families such as P-glycoprotein (P-gp), multidrug resistance-associated protein 2 (MRP2), breast cancer resistance protein, and bile salt export pump, regulate drug efflux, bile acid handling, and intracellular substrate availability6. The extent of this intestinal and hepatic first-pass metabolism can substantially reduce the bioavailability of orally administered drugs, posing a critical challenge in drug development. PK and bioavailability issues represent a major cause of clinical trial attrition, accounting for 15–25% of failures in Phase I clinical trials7. Although hepatic metabolism is the primary focus in early drug development, clinical PK analyses indicate that intestinal first-pass effects, encompassing Phase I/II metabolism and efflux transport, can account for more than 20% of total first-pass loss in a substantial subset of drugs8. These findings underscore the critical need to better model intestinal drug metabolism in a patient-specific manner, rather than focusing solely on hepatic processes. Intestinal first-pass metabolism also exhibits substantial inter-individual variability. Clinical studies have shown that drugs such as cyclosporine exhibit an average per-oral bioavailability of ~21%, but with a remarkably wide range from around 10% to 89%. Demographic factors, including age, sex, and ethnicity, have been reported to contribute to this inter-individual variability9. Similar variability has been reported for other per-oral drugs, including tacrolimus (varied from 5 to 93%)10 and midazolam (varied from 25 to 85%)11, with age reported as a contributing factor. This broad variability reflects patient-specific determinants that regulate intestinal first-pass metabolism. Developmental and age-dependent regulation influences baseline expression of intestinal metabolic enzymes12,13. Sex-specific hormonal signaling further modulates metabolic capacity14–16. Genetic polymorphisms, particularly in enzymes such as CYP3A4, introduce inherited differences in metabolic efficiency17,18. In addition, disease-associated inflammation can substantially reduce the intestinal drug-metabolizing capacity19–24. Collectively, these host-derived variables shape the baseline metabolic capacity of the intestinal epithelium. However, these host-derived variables do not fully account for the magnitude of variability observed clinically. Pharmacomicrobiomics: microbiome as a determinant of intestinal first-pass metabolism Although host-derived determinants represent a major regulatory layer of intestinal drug metabolism, they do not fully account for the inter-individual variability observed in intestinal PK. One such determinant is the intestinal microbiome, which encodes a diverse repertoire of metabolic enzymes capable of transforming xenobiotics in the intestinal lumen before absorption. The emerging field of pharmacomicrobiomics investigates how variation in the gut microbiome influences drug response and disposition. Pharmacomicrobiomics studies have identified several microbiome-dependent mechanisms of drug metabolism25. Strain-specific microbial inactivation of drugs provides a notable example of microbiome-dependent PK variability (e.g., digoxin inactivated by Eggerthella lenta26, levodopa metabolized by Enterococcus faecalis and E. lenta27, and 5-aminosalicylic acid (5-ASA) inactivated by microbial acetyltransferases28). Similarly, bacterial β-glucuronidases can deconjugate host-metabolized drugs in the intestinal lumen, regenerating parent drugs or active metabolites with pharmacological or toxic activity. Because β-glucuronidase activity is distributed across multiple gut bacterial taxa, including Bacteroides, Escherichia, and Clostridium species, β-glucuronidase-containing bacteria represent a major microbial mechanism of reactivation of host-glucuronidated drugs29. This reactivation mechanism has been reported for irinotecan30, mycophenolate31, regorafenib32, and nonsteroidal anti-inflammatory drugs33, highlighting how intestinal bacteria can drive clinically significant toxicity via enzymatic transformation in the gut lumen. Beyond direct drug metabolism, comparisons between germ-free and conventional mice have shown that the presence of intestinal microbiota alters the intestinal metabolic enzyme expression profiles in mice, indicating that microbiota can regulate host drug-metabolizing capacity in addition to directly transforming drugs in the gut lumen34. Together, these observations indicate that intestinal first-pass metabolism is not determined by host epithelial activity alone. Rather, it reflects host-microbiome co-metabolism, in which host epithelial absorption and enzymatic metabolism, microbial drug transformation, and microbiota-dependent regulation of host metabolic capacity jointly determine the fraction of a per-oral drug that reaches the portal circulation. Experimental systems designed to predict intestinal first-pass metabolism must therefore integrate both host epithelial cells and the intestinal microbiome to capture inter-individual variability in drug exposure. Current models to demonstrate intestinal first-pass metabolism A range of nonclinical models has been used to investigate intestinal first-pass metabolism, including in vivo animals, 2D in vitro static cultures, ex vivo surrogates, 3D intestinal organoids, and human Organ-on-a-Chip microphysiological system (MPS) models (Fig. 1). However, nonclinical models are limited to predict human per-oral bioavailability because they do not adequately reproduce human intestinal PK physiology (e.g., absorption, metabolism, and transport) and microbial contributions to drug metabolism. Comparative assessment of experimental models for investigating intestinal first-pass metabolism and pharmacomicrobiomics. Commonly utilized models, using animal surrogates, static in vitro cell cultures, ex vivo tissues, 3D organoid cultures, and Gut-on-a-Chip technologies, are systematically evaluated across key dimensions relevant to pharmacokinetics and microbiome-associated interrogations. These criteria encompass cost efficiency, sample accessibility, scalability, and high-throughput potential, recapitulation of intestinal mechanical dynamics (e.g., luminal flow and peristaltic motion), capacity for systemic pharmacokinetic assessment, fidelity of human-relevant drug-metabolizing activity, representation of microbiome-mediated drug biotransformation, and compatibility with advanced culturomics approaches. Relative performance is denoted as follows: −, minimal or unsupported capability; +, limited capability; ++, moderate capability; and +++, strong capability. Animal models Animal models have long been used to investigate intestinal drug metabolism35,36. In vivo dosing studies enable simultaneous evaluation of systemic PK and intestinal drug disposition under physiologic conditions. However, substantial interspecies differences in intestinal drug-metabolizing enzyme expression limit the predictive accuracy of these models for human intestinal first-pass metabolism37. For instance, midazolam, an orally administered sedative drug, exhibits per-oral bioavailability of less than 10% in mice, rats, dogs, and monkeys, whereas its bioavailability exceeds 30% in humans36, highlighting pronounced species-dependent differences. Midazolam is primarily metabolized by CYP3A4 in humans, whereas intestinal midazolam metabolism is mediated by species-specific CYP3A enzymes, such as CYP3A11/13 in mice, CYP3A9/18/62 in rats, and CYP3A8 in monkeys37. Cross-species studies indicate that intestinal CYP3A-mediated midazolam hydroxylation is higher in mice and monkeys than in humans38,39, suggesting that species differences in intestinal CYP3A-mediated metabolic capacity contribute to animal-human differences in midazolam bioavailability. Another example is aripiprazole, which exhibits per-oral bioavailability of less than 50% in mice and less than 20% in rats, dogs, and non-human primates, whereas its bioavailability exceeds 80% in humans40,41. Aripiprazole is primarily metabolized by CYP3A4 and CYP2D6 in humans41. However, the animal enzymes responsible for aripiprazole metabolism remain insufficiently defined. Because drug-metabolizing enzyme repertoires, substrate specificity, and catalytic activity differ across species, animal pharmacokinetic data may not provide sufficient mechanistic resolution to explain the enzyme-level basis of animal-human bioavailability differences. Furthermore, taxonomic and functional analyses reveal substantial interspecies differences in gut microbiota between humans and animals42. For example, dogs exhibit >60% overlap with the human gut microbiome, whereas mice show <20% overlap42. These differences in microbial composition and metabolic capacity likely result in substantial variation in microbiome-mediated drug metabolism across species. Attempts to model human gut microbial physiology using human fecal microbiota-transplanted animal systems have enabled partial reconstruction of human-associated microbial communities in vivo43. However, these models remain fundamentally challenged by interspecies host-microbiome incompatibility. Microbial colonization dynamics, epithelial adherence, immune crosstalk, and metabolic interactions are highly host-specific processes44, thereby limiting the fidelity with which human microbial functionality, ecological behavior, and pharmacologically relevant host-microbiome crosstalk can be recapitulated in animal hosts45. In vitro and ex vivo static culture models Intestinal epithelial cell lines have been widely used to model drug metabolism. These systems typically reproduce epithelial polarity by culturing cells on Transwell models that establish a luminal-abluminal interface and allow directional transport measurements between apical and basolateral compartments. Intestinal cell lines offer several practical advantages, including robust barrier formation, experimental reproducibility, low cost, and compatibility with high-throughput screening formats46. Several intestinal epithelial cell lines have been employed for this purpose, such as Caco-2 and HT-29 cells. Caco-2 has become the most widely used system for evaluating intestinal permeability in drug development. However, parental Caco-2 express markedly lower levels of CYP3A4 than human intestinal tissues (>800-fold)47, and key xenobiotic-sensing nuclear receptors, including pregnane X receptor (PXR) and constitutive androstane receptor, are undetectable in parental Caco-246, limiting the ability to recapitulate in vivo intestinal first-pass metabolism. Several engineered Caco-2 models have attempted to compensate for the intrinsically low CYP3A4 activity of the parental line. Early CYP3A4-cDNA-transfected Caco-2 cells showed more than a 100-fold increase in CYP3A4 activity compared with the parental lines48. More recent piggyBac transposon-based models enabled experimental regulation of CYP3A4 expression and demonstrated CYP3A4 expression levels comparable to those observed in human small intestine49,50. However, they remain incomplete because they do not restore the full repertoire or physiological balance of intestinal drug-metabolizing enzymes, and artificially introduced CYP3A4 is not governed by the endogenous nuclear receptor-mediated induction machinery of human small intestinal epithelium, making it challenging to evaluate physiological CYP3A4 induction responses to drugs. To improve human relevance, ex vivo approaches using freshly isolated human intestinal tissue mounted in Ussing chambers have been explored51–53. In principle, these preparations preserve the function of major drug-metabolizing enzymes and key uptake and efflux transporters more faithfully than tumor-derived cell lines51–53. However, their translational utility remains constrained by practical and biological limitations. Ussing chamber assays typically requires freshly resected surgical specimens rather than small endoscopic biopsies, because each chamber requires an intact tissue area of ~1.0 cm2 for mounting and measurement51–53. As a result, experiments involving technical replicates or multiple conditions require relatively large fresh intestinal specimens. These requirements make stable, continuous tissue supply difficult. In addition, functional readouts can decline rapidly after tissue procurement. Even when freshly isolated intestinal tissue is maintained in ice-cold oxygenated buffer until the assay, some transporter activities and barrier function can decrease by approximately 50% within 4 h after procurement51. A further limitation of both conventional in vitro and ex vivo models is that they cannot maintain a physiological host-microbiome co-culture. Because these systems operate under static conditions without continuous medium perfusion, biochemical conditions in the culture compartment can rapidly become unstable when conducting host-microbiome co-culture. Nutrient depletion, pH decrease, accumulation of microbial organic waste products, secretion of toxic metabolites, and mismatched growth rates between epithelial cells and bacteria can compromise epithelial viability and barrier integrity54. In addition, establishing the steep oxygen gradient characteristic of the intestinal mucosa is challenging in conventional static systems, which are typically maintained under open-air culture conditions. Engineered Transwell-based models have addressed this issue by sealing the apical compartment with an oxygen-impermeable plug to create an anaerobic luminal environment while preserving basolateral oxygenation, thereby enabling short-term co-culture with anaerobic bacteria55. However, because these platforms remain static, they address oxygen control but do not fully resolve microbial overgrowth and biochemical instability during prolonged co-culture. Intestinal organoid models Patient-derived intestinal organoids developed from biopsy-derived intestinal stem cells provide a human epithelial system with greater physiological relevance than cell lines. These cultures are performed within extracellular matrix hydrogels and maintained in a Wnt-rich medium that supports stem cell maintenance and epithelial proliferation56. Because these organoids originate from patient-derived stem cells, they retain donor-specific transcriptional profiles, enabling recapitulation of patient-specific epithelial features57. In a 3D culture format using human biopsy-derived ileal organoids and colonoids, differentiated organoids showed PXR-mediated induction of CYP3A4 expression, a response that is largely absent in parental Caco-2 cells58. However, 3D organoids present structural and practical limitations. The apical surface faces an enclosed lumen, making direct apical drug exposure and real-time monitoring of absorption technically challenging. Microinjection of drug compounds into the organoid lumen may provide one approach to access the apical compartment, but this method requires technical expertise and is labor-intensive. To address luminal accessibility, organoid-derived monolayer models have been developed by seeding dissociated organoids on Transwell inserts. This format enables controlled apical drug exposure and basolateral sampling, allowing separate assessment of apical uptake transport, intracellular metabolism by drug-metabolizing enzymes, and efflux transporter activity59,60. By quantifying the amount of parent drug and metabolites that appear in the basolateral compartment, these models can also estimate intestinal drug bioavailability. For CYP3A4 substrates, organoid-derived monolayers have been reported to generate predicted bioavailability values that are comparable to human in vivo estimates59,60. Another approach involves induced pluripotent stem cell (iPSC)-derived intestinal organoids. iPSCs, which are established from patient-derived somatic cells, can be differentiated into intestinal epithelial lineages that generate multiple cell types, including absorptive enterocytes, goblet cells, Paneth cells, and enteroendocrine cells61. This strategy expands access to donor-specific intestinal models by enabling intestinal organoid generation from individuals for whom intestinal biopsy or surgical tissue is not clinically available, including healthy donors and patients without intestinal disease. Just as with biopsy-derived organoids, iPSC-derived intestinal models have been used in both 3D and Transwell monolayer formats to study intestinal drug metabolism. Although these studies did not directly compare organoids with human intestinal tissue, both formats showed substantially higher CYP3A4 expression than parental Caco-2 cells, and retained PXR-mediated drug induction responses62–64. In a Transwell monolayer, iPSC-derived intestinal epithelium has been used to estimate intestinal bioavailability for clinically used drugs, producing values that were consistent with reported human in vivo data63. However, compared to biopsy-derived organoids, iPSC-derived intestinal organoids require multi-step differentiation, increasing cost, time, and protocol-dependent variability. This process can generate heterogeneous cultures with variable epithelial maturation and lineage composition. Consequently, iPSC-derived intestinal organoids often retain a fetal-like or immature phenotype65–68, which limits their functional in vivo relevance for intestinal first-pass metabolism. Despite the advantages of these organoid-based systems, a critical limitation remains because both biopsy- and iPSC-derived models are typically maintained under static culture conditions. As such, these organoid models lack the mechanical cues present in vivo, such as luminal flow and peristaltic motion, which may further constrain epithelial differentiation and metabolic function. Organ-on-a-Chip models In recent years, Organ-on-a-Chip technologies have emerged as dynamic microfluidic platforms that enable controlled culture of human tissues under physiologically relevant conditions. Gut-on-a-Chip platforms culture intestinal epithelial cells, including conventional cell lines and intestinal organoid-derived epithelial cells, as a monolayer on an elastic porous membrane that separates apical and basolateral microchannels. Continuous perfusion of culture medium generates physiologically relevant luminal shear stress, while cyclic mechanical deformation mimics peristalsis69. Under these dynamic cues, initially seeded epithelial monolayers can undergo 3D morphogenesis and form villi-like intestinal microarchitecture70,71, with enhanced epithelial differentiation that more closely resembles the in vivo intestine compared with 3D organoids72. This hydrodynamic microfluidic platform enables compartment-specific access (e.g., luminal vs. vascular), facilitating targeted molecular and cellular analyses, including directional sampling of secretomes and controlled introduction of immune cells into the abluminal space73,74. Independent control of each microchannel also allows establishment of physiologically relevant oxygen gradients by supplying oxygen-depleted and oxygenated media to the luminal and vascular compartments, respectively75. In addition, vacuum-driven cyclic stretch permits precise modulation of peristalsis-like motions, enabling the recapitulation of diverse physiological and pathological mechanical cues69. This mechanically dynamic, physiologically relevant Gut-on-a-Chip platform supports robust host-microbiome co-culture by maintaining controlled microbial populations within a microchannel using continuous plug-flow bioreactor-type conditions. Steady-state operation enables regulation of microbial density through removal of unbound microbes, continuous nutrient supply, maintenance of stable pH, and clearance of microbial metabolic wastes, while preserving accurate oxygen gradients. Combined fluid flow and mechanical motion further facilitate active removal of excess microbes and host cell debris, stabilizing long-term co-culture of live microbial cells with host epithelial cells76. Early Gut-on-a-Chip models employed Caco-2 cells under fluid flow and cyclic stretching, demonstrating enhanced 3D intestinal morphogenesis and CYP3A4-mediated drug metabolism compared with static monolayers cultured in Transwells70. In a distinct microfluidic device design that applied fluid flow alone, exposure of Caco-2 cells to flow-induced shear stress was also reported to increase CYP3A4 activity by approximately twofold compared to static controls77. These physiological features were replicable in organoid-employed Gut-on-a-Chip, including both biopsy- and iPSC-derived organoids72,78,79. In these systems, gene expression of CYP3A4 reached levels comparable to human intestinal tissue72,79, and functional CYP3A4 activity was demonstrated by testosterone metabolism72. In addition, PXR-mediated drug induction in CYP3A4 activity was reported at levels comparable to those observed in the human intestine72. Despite these advances, Gut-on-a-Chip models remain insufficient for studying intestinal first-pass metabolism because most current approaches focus on drug metabolism in host cells and do not incorporate gut microbiota as contributors to drug transformation and epithelial metabolic regulation (Table 1). Table 1Comparative evaluation of Organ-on-a-Chip models for modeling intestinal first-pass drug metabolism.ModelCell sourceFluid shearMechanical motionGut microbiomeKey findingsRef.ApicalBasalGut-on-a-ChipCaco-2○○○×Dynamic Gut-on-a-Chip cultures induce 3D morphogenesis and CYP3A4 activity, whereas static Transwell systems lack CYP3A4 function.70Gut–Liver-on-a-ChipCaco-2HepG2○○××Apical and basolateral fluid shear increases CYP3A4 activity compared with static controls111Gut–Liver-on-a-ChipCaco-2HepG2○○××This study confirms the metabolism of paracetamol, but lacks validation against in vivo PK data112Gut–Liver-on-a-ChipCaco-2 + HT-29Primary hepatocytes×○××This study confirms the metabolism of mycophenolate mofetil but lacks validation against in vivo PK data114Gut–Liver-on-a-ChipPrimary jejunum epitheliumPrimary hepatocytes×○××This study confirms the metabolism of midazolam, but at lower levels compared to in vivo110Gut–Liver-on-a-ChipCaco-2, engineeredPrimary hepatocytes×○××Engineered Caco-2 cells expressing CYP3A4 and UGT1A1 enhance midazolam metabolism, albeit at levels lower than those observed in vivo113Gut-on-a-ChipiPSC-derived intestinal organoids○○××iPSC-derived intestine-on-chip shows enhanced maturation and increased expression of drug metabolism-related genes compared with organoids and Transwells.79Gut-on-a-ChipHuman duodenum organoid○○○×Duodenal organoid epithelium shows CYP3A4 expression with in vivo relevance during dynamic cultures72 Comparative evaluation of Organ-on-a-Chip models for modeling intestinal first-pass drug metabolism. Caco-2 HepG2 Caco-2 HepG2 Caco-2 + HT-29 Primary hepatocytes Primary jejunum epithelium Primary hepatocytes Caco-2, engineered Primary hepatocytes Collectively, these studies demonstrate that Gut-on-a-Chip platforms can reproduce key epithelial and mechanical features relevant to intestinal first-pass metabolism, CYP3A4 activity, PXR-mediated induction, and compartment-specific sampling. However, most current drug metabolism studies have evaluated these functions primarily within the host epithelium, without incorporation of gut microbiota as contributors to drug transformation and epithelial metabolic regulation. The next challenge is therefore to extend Gut-on-a-Chip technology from a host epithelial metabolism model to a microbiota-integrated host epithelium culture system that can evaluate both host- and microbiome-dependent variation in per-oral drug exposure. Engineering human intestinal culturomics for personalized prediction of first-pass metabolism Advanced culturomics in personalized human Gut-on-a-Chip models Advances in metagenomic sequencing have revealed extensive gut microbiome diversity, highlighting the significant discrepancy between detected microbial species and those that can be cultivable using conventional culture methods80,81. However, many human gut microbes remain difficult to culture in vitro due to symbiotic relationships and metabolic interdependence with the host microenvironment82,83. Recent advances in defined microbial community models have attempted to reconstruct simplified gut microbial ecosystems by combining culturable human gut bacteria. For example, the hCom model was constructed from 104 prevalent and culturable human gut bacterial species, providing a complex but defined community for controlled microbiome studies84. Similarly, simplified consortia such as Simplified Human Intestinal Microbiota (SIHUMI) combine a small number of fully sequenced, anaerobically culturable human gut-derived strains and enable qPCR-based strain-level tracking85. In another approach, fecal-derived polymicrobial communities have been cultured in continuous bioreactor systems. In mini-bioreactor arrays (MBRAs), approximately 15–25% of the taxa present in the original fecal inoculum were maintained as stable communities over 21 days86. Maintaining multi-species communities in in vitro co-cultures remains challenging, as competitive dynamics often result in “winner-takes-all” dominance. Hence, novel culturomics are crucial for expanding the repertoire of cultivable microbes by systematically incorporating host components (e.g., intestinal epithelial cells and mucus layers) to colonize bacterial species that have remained uncultivable with conventional culture87. A physiologically differentiated intestinal epithelium provides the host niche that microbes encounter in vivo, including a mucus interface for colonization, a tight junction barrier for compartmentalization, and epithelial defense programs such as antimicrobial peptide production. At the same time, mechanically dynamic flow and peristalsis-like motion are needed to control the physical and chemical environment by supplying nutrients, removing unbound bacteria and debris, and preventing uncontrolled microbial overgrowth. In addition to continuous nutrient supply, host-microbe nutrient exchange is also required to support epithelial viability and microbial metabolic activity. Host-derived soluble and insoluble products, including mucus glycans, shed epithelial debris, and secreted metabolites, can provide nutrient sources and ecological cues for microbial communities. In turn, microbial metabolites, such as short-chain fatty acids, can feed back to epithelial cells and influence barrier function, differentiation, and drug-metabolizing activity88. Accurate oxygen control is also essential for host-microbiome co-culture because host epithelial cells require oxygen from the vascular side, whereas many gut commensals require an oxygen-depleted luminal niche. This oxygen gradient also shapes microbial community composition and metabolic output89,90, making physiological oxygen control essential for maintaining microbiota structure and function during co-culture. Gut-on-a-Chip models provide a compartmentalized microfluidic architecture in which intestinal epithelium is cultured between luminal and vascular channels under continuous flow and peristalsis-like mechanical motion, thereby promoting formation of a differentiated, barrier-forming epithelium with a mucus layer (Fig. 2)70–72. In addition, selective perfusion of oxygen-depleted medium through the luminal channel and oxygenated medium through the vascular channel can establish a physiological oxygen gradient76, creating an anoxic microbial niche adjacent to an oxygen-supported epithelial interface. Thus, Gut-on-a-Chip models provide a practical path toward host-integrated culturomics by linking microbial culture to a living human intestinal niche. Conceptual framework for host-integrated culturomics using Gut-on-a-Chip platforms to model personalized intestinal drug metabolism. This schematic illustrates an integrated framework for engineering intestinal culturomics on chip to model first-pass metabolism as a host-microbiome co-metabolic process. Recapitulation of central physiological features, including oxygen gradients, mucus-supported microbial colonization, and intestinal mechanical dynamics (luminal flow and peristalsis), is essential for sustaining stable host-microbiome co-culture. Novel culturomics strategies enable advanced pharmacomicrobiomics by capturing microbial and host drug biotransformation. Longitudinal evaluation of patient-derived biospecimens, such as region-specific intestinal tissues, stool, blood, and luminal contents, across the small and large intestine supports individualized assessment of drug metabolism. This framework further extends to translational precision medicine applications, enabling systematic evaluation of clinically relevant variables, including food-drug-host interactions and inter-individual variability across sex, age, race/ethnicity, and disease status. Advanced Gut-on-a-Chip culturomics platforms uniquely support to recapitulate complex host-microbiome crosstalk by offering a controllable microphysiological environment that enables precise modulation of oxygen tension, disease-specific milieus, differentiated epithelial microarchitecture, mucus barrier interface, and steady-state balance of nutrients and metabolites. AP, apical; BL, basolateral. Personalized modeling of first-pass metabolism in gut epithelium In contemporary drug development, intestinal first-pass metabolism is predominantly evaluated using population PK approaches, in which PK parameters are derived from aggregated clinical datasets obtained across diverse populations91. Inter-individual variability in drug deposition is subsequently incorporated through a statistical modeling framework that integrates demographic, physiological, and clinical covariates, including race/ethnicity, age, sex, and biological laboratory parameters91. Although these models effectively characterize population-level variability in drug exposure, they do not directly represent the underlying biological determinants of intestinal drug metabolism, such as patient-specific differences in epithelial drug-metabolizing enzyme expression, transporter profiles, and gut microbiota composition. Consequently, population PK frameworks are not designed to reconstruct individualized intestinal metabolic phenotypes, highlighting the need for experimental platforms capable of modeling patient-specific intestinal pharmacology. Previous studies using human intestinal tissues have provided fragmented but important evidence that host biological variables shape intestinal drug-metabolizing capacity, including sex- and age-dependent differences in drug-metabolizing enzyme expression12–16, race/ethnicity-associated genetic variation in metabolic enzymes17, and inflammation-associated alteration of drug-metabolizing enzymes19,20,22,23. These findings collectively indicate that host variability should not be treated as background noise, but as a central determinant of intestinal first-pass metabolism. However, these observations have largely remained descriptive, based mainly on transcript or protein-level measurements in human tissues. Personalized modeling of intestinal first-pass metabolism now requires moving this variability into functional in vitro systems that can measure metabolic activity, epithelial drug transport, and changes in per-oral bioavailability under controlled experimental conditions. To functionally capture this host-specific variability, epithelial cells dissociated from patient-derived intestinal organoids should be integrated into the Gut-on-a-Chip platform. Such integration enables reconstruction of donor-specific intestinal epithelial biology in a controlled microphysiological environment, allowing direct experimental assessment of how host genetic variability influences intestinal first-pass drug metabolism (Fig. 2). Inter-individual variability in the human gut microbiome is also substantial. Large-scale studies such as the Human Microbiome Project have demonstrated marked differences in microbial community composition across healthy individuals92, which translate into variability in microbial drug metabolic capacity. Consistent with this concept, clinically relevant microbial reactions, including E. lenta-dependent digoxin inactivation and β-glucuronidase-dependent reactivation of irinotecan metabolites, vary among patients26,93. Because these differences arise from each patient’s microbial community structure and enzymatic repertoire, they cannot be fully captured without incorporating patient-derived gut microbiota into the experimental model. Therefore, personalized modeling of intestinal first-pass metabolism should not rely on patient-derived epithelium alone, but should also integrate patient-matched microbiota. To achieve this objective, Gut-on-a-Chip platforms should be designed to reconstruct donor-specific epithelial metabolism and patient-specific microbial biotransformation within the same compartmentalized microphysiological environment (Fig. 2). This design would enable personalized assessment of how host epithelial metabolism, microbial drug transformation, and their interaction shape individualized changes in per-oral bioavailability. The high cost of fully individualized organoid-chip testing represents a major barrier to its use as a universal screening strategy because the Gut-on-a-Chip model requires patient-specific organoid generation, chip culture, and functional validation. Therefore, fully personalized Gut-on-a-Chip testing should be prioritized for settings in which patient-specific intestinal first-pass metabolism is expected to highly influence safety, efficacy, or clinical development decisions. Examples include drugs for which small differences in systemic exposure can cause severe toxicity, such as anticancer agents, drugs with highly variable per-oral bioavailability, and therapies for rare diseases in which small patient populations limit large-scale clinical PK studies. In parallel, for broader drug development, a separate cost-conscious strategy is to use population-representative donor panels rather than fully individualized models for every patient. By selecting donors across clinically relevant variables such as age, sex, disease status, drug-metabolizing enzyme genotype, and microbiome-associated metabolic phenotype, these panels could capture meaningful variability in intestinal absorption and metabolism while reducing the cost and labor required for fully personalized testing. Advanced engineering for emulating a human physiological signature Beyond patient-derived epithelial cells and microbiota, personalized modeling of intestinal first-pass metabolism should also incorporate physiological contexts that shape drug metabolism in vivo. Dietary exposure represents a major determinant of intestinal drug absorption and metabolism. These effects can be categorized into three mechanisms. First, food-host interactions regulate intestinal enzyme and transporter activity (e.g., furanocoumarins in grapefruit juice irreversibly inhibit CYP3A494; isoflavones in soy-derived compounds induce CYP3A491). Second, food-drug interactions alter drug physicochemical properties and absorption (e.g., calcium forms insoluble complexes94; diet-dependent changes in luminal pH influence drug solubility95). Third, drug-drug interactions modulate intestinal enzyme activity (e.g., rifampicin induces CYP3A4, UGTs, SULTs, and GSTs via PXR activation96,97; ketoconazole inhibits CYP3A498). Although numerous studies have reported interactions between dietary factors and intestinal first-pass metabolism, accurate prediction of these effects remains challenging using conventional animal models. Laboratory animals differ from humans in habitual diet, gastrointestinal anatomy, gastrointestinal transit time, and luminal pH99. In addition, species-specific molecular regulation, such as differences in PXR ligand specificity, limits extrapolation of diet- and drug-mediated enzyme induction from animals to humans100. These limitations indicate that diet-associated first-pass metabolism should be studied in human-relevant systems where dietary components, epithelial responses, microbial activity, and drug metabolism can be experimentally controlled rather than inferred from animal feeding studies. Although Gut-on-a-Chip platforms have not yet fully reproduced the complexity of human dietary exposure, small intestinal organoids cultured in Gut-on-a-Chip platforms have been shown to express not only in vivo-comparable CYP3A4 expression, but also functional nuclear receptors involved in drug induction72. Therefore, Gut-on-a-Chip systems should be further developed as complementary platforms to evaluate diet-host, diet-drug, and drug-drug interactions in a human-relevant experimental context (Fig. 2). Intestinal first-pass metabolism is also strongly shaped by regional differences along the gastrointestinal tract. The stomach can influence per-oral drug absorption through gastric pH-dependent dissolution and acid-labile degradation101. In the duodenum, bile acids, phospholipids, and lipid digestion products can form mixed micelles and colloidal phases that increase the apparent solubility of poorly water-soluble drugs, while luminal pH further influences ionization and dissolution102. The jejunum is the primary site of intestinal drug metabolism, driven by high expression of enzymes and transporters, with human CYP3A4 gene expression generally peaking in the jejunum and decreasing in the duodenum, ileum, and colon103. Further downstream, intestinal drug transformation shifts along opposing host and microbial gradients. Host drug-metabolizing enzyme expressions decrease toward the ileum and colon, whereas microbial density increases from approximately 10³–10⁴ cells/g in the proximal small intestine to 10⁷–10⁸ cells/g in the distal ileum and 10¹¹–10¹² cells/g in the colon104,105. The ileum, therefore, represents an interface where declining epithelial drug metabolism overlaps with increasing microbial biomass, with a community enriched in facultative anaerobic taxa, including Enterobacteriaceae and Bacilli-class Firmicutes106. In the colon, dense obligate anaerobic microbiota, including Bacteroidaceae, Prevotellaceae, Lachnospiraceae, and Ruminococcaceae, favor microbial transformations such as reduction, hydrolysis, deconjugation, and β-glucuronidase-mediated reactivation106,107. To capture these region-specific physiological contexts, future Gut-on-a-Chip should incorporate region-specific patient-derived organoids and region-matched microbial communities. Small intestinal chips, particularly duodenum and jejunum, should recreate proximal intestinal environments, including relatively oxygenated conditions (32 mmHg)108 and digestive luminal components. Ileum and colon chips can support lower oxygen environments (11 mmHg)108 and dense microbial communities, enabling evaluation of host-microbiome co-metabolism. Beyond single-region models, linked multi-compartment Gut-on-a-Chip systems will be needed to model first-pass metabolism across the gastrointestinal tract. Such systems would allow drugs absorbed and metabolized in the proximal intestine to be distinguished from unabsorbed parent drugs or metabolites that transit to distal segments and undergo microbiome-mediated transformation. This sequential design would better recapitulate intestinal processing along the gut and improve prediction of region-dependent contributions to per-oral bioavailability (Fig. 2). Systems-level integration in mechanistic pharmacology A key advantage of Gut-on-a-Chip systems is their ability to experimentally separate the contribution of host epithelium, microbiota, biomechanical cues, and also other microenvironmental cells (e.g., Immune and stromal cells), to intestinal drug metabolism by comparing defined experimental conditions, such as epithelial-only culture, host-microbiome co-culture, flow or stretch stimulation, and immune or stromal cell addition under controlled settings. The separation of luminal and vascular channels further enables compartment-specific sampling of parent drugs, metabolites, epithelial cells, microbial communities, and secreted factors across multiple time points and drug concentrations. When combined with multi-omics, this framework links host regulation, microbial metabolism, and drug transformation at a systems level. In this framework, baseline host genomics can be used to test how genetic variants in drug-metabolizing enzymes and transporters influence per-oral bioavailability in patient-derived epithelial models. Transcriptomic profiling of collected epithelial cells can then define drug-induced host responses, including induction or suppression of drug-metabolizing enzymes and transporters. Proteomic and metabolomic analyses of epithelial cells and luminal or vascular effluents can quantify parent-drug depletion and metabolite formation. In parallel, microbial metagenomics or metatranscriptomics can link patient-specific microbial pathways to observed drug transformation and assess how drug exposure reshapes microbial community composition. To interpret these complex datasets, computational analysis, such as machine learning and artificial intelligence-based analysis, should therefore be used not simply to merge multi-omics datasets, but to rank the experimentally defined variables that explain patient-specific metabolic outcomes. By linking epithelial drug-metabolizing expression, transporter expression, microbial xenobiotic pathway, biomechanical cues, and measured metabolite profiles to parent-drug depletion, these models could identify whether a change in per-oral bioavailability is driven primarily by host epithelial metabolism, microbial biotransformation, or microenvironmental conditions. This approach would move Gut-on-a-Chip studies beyond cataloging molecular differences and toward mechanism-based prediction of patient-specific intestinal first-pass metabolism. Challenges and opportunities Although patient-derived Gut-on-a-Chip systems offer a promising strategy for personalized modeling of intestinal first-pass metabolism, several technical and translational considerations remain to be addressed. First, their scalability is limited by operational complexity. Each chip often requires dedicated tubing, continuous perfusion modules, and frequent medium exchange, making parallel experiments labor-intensive and challenging to maintain. This limited throughput restricts the number of conditions and patient samples that can be tested, limiting statistical power. Addressing these constraints will require an advanced high-throughput MPS, automated fluid handling, standardized manufacturing, and a user-friendly platform to enable parallel experimentation without compromising patient-specific fidelity. The cost of generating individualized organoid-chip models also represents a major barrier to broad implementation. Strategies to reduce this burden, including automated fluid handling, standardized platform manufacturing, and biobanking of quality-controlled patient-derived organoids, will be necessary to make personalized Gut-on-a-Chip testing more practical and scalable. Gut-on-a-Chip platforms have enabled co-culture of intestinal epithelium with anaerobic gut bacteria75,76,109, supporting their use as host-integrated microbiome models. However, further validation is needed to determine how well complex patient-derived fecal microbiota preserve their taxonomic diversity and metabolic function during on-chip culture. Future studies should therefore assess community stability and functional microbial fidelity to strengthen Gut-on-a-Chip platforms as host-integrated culturomics systems for microbiome-dependent drug metabolism. Until such systems are fully validated, microbiome-conditioned drug media may provide a complementary screening strategy to identify host-independent drug metabolites produced by culturable microbial communities. However, this approach remains limited by the same culture bottleneck that motivates host-integrated culturomics, because uncultivable or host-dependent microbes, mucus-associated colonization, and epithelial-microbial feedback are not preserved. Thus, conditioned-media approaches can support metabolite screening but cannot replace direct Gut-on-a-Chip co-culture for modeling microbiome-dependent first-pass metabolism. Complete understanding of first-pass metabolism requires reconstruction of the sequential process by which drugs absorbed in the intestine enter the portal circulation and undergo transformation in the liver. Intestinal models in isolation cannot capture this downstream hepatic extraction, which determines systemic drug exposure. Several proof-of-concept studies of inter-organ interaction have coupled intestinal cell lines such as Caco-2 with hepatic cell lines, including HepG2, within microfluidic gut-liver models (Table 1)110–114. However, these systems rely on immortalized cells, limiting their ability to model personalized metabolic capacity. Integration of patient-derived Gut-on-a-Chip and Liver-on-a-Chip modules through controlled microfluidic coupling would enable reconstruction of sequential intestinal and hepatic metabolism, providing a more comprehensive and personalized platform for evaluating whole-body first-pass drug processing. A further limitation is that current Gut-on-a-Chip platforms are primarily optimized for PK interrogation rather than comprehensive PD assessment unless the pharmacological target resides within intestinal tissue. Thus, they should be viewed as complementary to in vivo animal models, which remain essential for systemic responses and target engagement. Integrating Gut-on-a-Chip-based PK prediction with animal studies may enhance translational success by enabling more informed selection of drug candidates prior to clinical trials. Moreover, emerging efforts to develop organoid-derived Gut-on-a-Chip platforms for experimental animals suggest that in vitro PK modeling may also refine and reduce animal experimentation by improving dose selection and study design115–117. Looking ahead, integration with interconnected organ-chip systems may extend in vitro modeling toward multi-organ PK and PD processes, further complementing animal models and improving human-relevant prediction of drug disposition. Conclusion and future directions Intestinal first-pass metabolism is a dynamic and individualized process shaped by host genetics, microbial ecology, luminal chemistry, and biomechanical forces, yet no existing model captures this integrated biology in a patient-specific manner. By combining patient-derived intestinal organoids with controlled microenvironmental engineering and microbiome co-culture, Gut-on-a-Chip platforms provide a mechanistically precise framework for modeling human intestinal PK in vitro (Fig. 2). Within this framework, advances in pharmacomicrobiomics can be experimentally interrogated through controlled host-microbiome systems, enabling direct evaluation of how microbial communities influence drug metabolism and PK variability. At the same time, emerging culturomics approaches expand the repertoire of cultivable gut microbes, enabling reconstruction of increasingly complex and physiologically relevant microbial communities within microengineered intestinal environments. Although current systems remain constrained by scalability, material limitations, and incomplete multi-organ integration, these challenges represent engineering frontiers rather than fundamental barriers. In synergy with animal models and clinical studies for systemic validation, the convergence of Gut-on-a-Chip technology, pharmacomicrobiomics, and advanced culturomics may ultimately transform intestinal first-pass metabolism from an empirical source of variability into a measurable and individualized parameter in drug development. Acknowledgements This work was supported in part by the Kenneth Rainin Foundation Innovator Award, the Crohn’s and Colitis Foundation of America Senior Research Award, NIH NCI IMAT program (1R33CA286797), Japan Society for the Promotion of Science Research Fellowship for Young Scientists (24KJ0651), and Japan Society for the Promotion of Science Overseas Research Fellowships. During the preparation of this work, the authors utilized generative artificial intelligence (AI) tools (e.g., Gemini) and illustration platforms (e.g., BioRender.com) in part for the creation of schematic figures. Author contributions I.N. and H.J.K. developed the concept and wrote and revised the manuscript. All authors reviewed and approved the final manuscript. Data availability No new datasets were generated or analyzed during the current study. No custom code was generated or used for data analysis. Competing interests The authors declare no competing interests. Footnotes Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Data Availability Statement No new datasets were generated or analyzed during the current study. No custom code was generated or used for data analysis.