Gut-on-a-chip models for dissecting the gut microbiology and physiology

Microfluidic technologies have been extensively investigated in recent years for developing organ-on-a-chip-devices as robust in vitro models aiming to recapitulate organ 3D topography and its physicochemical cues. Among these attempts, an important research front has focused on simulating the physiology of the gut, an organ with a distinct cellular composition featuring a plethora of microbial and human cells that mutually mediate critical body functions. This research has led to innovative approaches to model fluid flow, mechanical forces, and oxygen gradients, which are all important developmental cues of the gut physiological system. A myriad of studies has demonstrated that gut-on-a-chip models reinforce a prolonged coculture of microbiota and human cells with genotypic and phenotypic responses that closely mimic the in vivo data. Accordingly, the excellent organ mimicry offered by gut-on-a-chips has fueled numerous investigations on the clinical and industrial applications of these devices in recent years. In this review, we outline various gut-on-a-chip designs, particularly focusing on different configurations used to coculture the microbiome and various human intestinal cells. We then elaborate on different approaches that have been adopted to model key physiochemical stimuli and explore how these models have been beneficial to understanding gut pathophysiology and testing therapeutic interventions.


I. INTRODUCTION
The human gut is one of the critical organs with important biological functions orchestrated by human intestinal cells and a rich microbial consortium, known as the gut microbiome. 1 The most wellknown function of the gut is the digestion of food and absorption of nutrients. 2 Being equipped with special enzymatic capacity, intestinal enterocytes are able to digest a variety of compounds such as amino acids, peptides, polysaccharides, and xenobiotics, [3][4][5] and benefiting from a unique metabolic diversity, the microbiome maximizes the digestion by breaking down complex carbohydrates and vitamins. 6 In addition to food digestion, the gut and its microbiome have broad roles in health regulation and disease prevention. 7 A notable example of such roles is microbial and epithelial barrier functions, which are essential for maintaining homeostasis, establishing a bidirectional relationship with immune cells, and protecting the body against pathogenic attacks. 8,9 The disruption of the gut barrier has been linked to various diseases, including gastrointestinal (GI) diseases, such as inflammatory bowel disease (IBD) and colorectal cancer, 10,11 in addition to non-GI diseases as versatile as cancer, diabetes, obesity, asthma, cardiovascular diseases, and respiratory infections such as COVID-19. [12][13][14][15][16][17] The eminent role of the gut in health and diseases has aroused tremendous interest among the scientific community in deciphering the gut's physiological complexities over the past decades; however, one of the major stumbling blocks in these attempts has been the limitations of conventional experimental techniques. 18 Among various methods, direct experimentation on human subjects has been proven to be difficult due to the inaccessibility of the GI tract, and imaging techniques such as colonoscopy have been invasive and lacked cellular-level resolution. [19][20][21] Analyzing the human microbiome has been conducted by sequencing fecal samples, but these are not representative of the entire gut microbial population, as many bacteria live in hardly accessible zones in the form of surface-attached communities, termed biofilms, inside the intestine. [22][23][24][25] Moreover, fecal samples provide little information about the spatial variation of microbiota across the intestine. 26 Animal models benefit from less strict

C. Microfluidic designs
Considering the multiplicity of cellular environments and the availability of various cellular models, several microfluidic designs have been thus far devised to study the gut. Depending on the complexity of design philosophies, the models can be broadly divided into mono-environment and multi-environment designs (refer to Table I for a summary of key models).

Mono-environment models of the gut
The simplest form of gut microfluidic models has been devices that mimic a single type of microenvironment and typically culture either microbial or intestinal cells. These devices have easily mimicked in vivo fluidic transport and shear forces. A simple implementation of this model is a device that hosts Caco-2 cells under the lumen hydrodynamics in a microchannel [ Fig. 3(a)], which was used in several FIG. 1. Schematic of a double layer organ-on-a-chip device. Each layer is built separately simulating a cellular microenvironment. A porous membrane can be used between the layers to create tissue interfaces across different microenvironments. Various analytical techniques such as microscopy as well as omics and histological analyses can be integrated with the devices to study biological processes (refer to Sec. II D for an explanation of analytical techniques).

BOX I. Intestine cellular physiology.
A brief review of the gut physiology is provided here with specific focus on the important role of cellular microenvironment in organ-on-achip systems. The gut possesses an immensely diverse cellular system spanning different anatomical sections, including the lumen, epithelium, and sub-epithelium (Fig. 2). The microbiome consists of hundreds of species of bacteria, viruses, yeasts, and fungi that live in the lumen under the condition of the convective flow of the digested food and the fluctuating mechanical motion, termed peristalsis, that propels the food forward. [72][73][74] Lumen is characterized by an anaerobic environment, resulting from the oxygen consumption of microbes during food digestion. 74,75 The environmental variables themselves vary widely along the lumen; for example, the colon features a slower flow and less acidic pH compared to the small intestine, causing a denser microbial community. 26,76 Encircling the lumen are epithelial cells, a monolayer of closely adjoined cells shaped naturally into a 3D topography, consisting of villi (peaks) and crypts (valleys). 77 The bottom of the crypt is made of stem cells performing tissue renewal during which most differentiated cells migrate toward villi. 78 The differentiated epithelial lineages include columnar enterocytes (absorptive cells), which are responsible for food absorption, goblet cells, which produce a protective mucin layer, Paneth cells, which excrete antimicrobial peptides, and endocrine cells, which release gastrointestinal hormones. 77 Besides these roles, the tight junctions provided by the epithelium provide structural integrity for the gut tissue. 67 Beneath the epithelium is a niche consisting of structural and cellular components that are essential for maintaining its homeostasis. 79 The underlying ECM provides physical support for the epithelium and contains a structured network of collagen fibers, integrins, fibronectin filaments, laminins, and glycosaminoglycan. 79 The cellular components in this area, termed lamina propria, are diverse comprising structural elements (fibroblasts, fibrocytes, and vascular endothelial cells) in addition to blood and immune cells. 80 Lamina propria is nutritious, oxygenated, and rich in blood and lymphatic capillaries, which carry the absorbed nutrients off the gut. 67,81 Overall, the structure of the gut cellular system varies widely in terms of composition, architecture, physicochemical cues, and function across the organ's anatomy. It is clear that without a holistic approach to model cellular microenvironments, it is impossible to capture the physiological and functional complexity of the gut. studies for gut physiological investigations. 82,83 The simple channel model could be further modified to simulate environmental cues such as gut peristalsis. For example, the "minigut" device developed by Cremer et al. to study bacterial growth dynamics [ Fig. 3(b)] modeled intestinal wall contractions through successive pressure-actuated membrane valves along the channel. 76 In microbial studies, micro topographies have been embedded in various studies to simulate the gut epithelial villi topography, capturing microscale cell-surface interactions. 84,85 Mono-environment devices offer simple practical means for modeling biological processes in a specific niche; however, they often disregard the "big picture" interplay between different microenvironments.

Multi-environment gut-on-a-chip devices
In situations demanding more rigorous analysis of gut physiology, multi-environment gut-on-a-chip devices have undeniably been preferred. They have been typically composed of multi-layer devices with stacked interconnected microchannels. Here, we discuss some of the major schemes developed thus far (Table I).
a. HMI (host-microbiota interaction) and HuMiX (human-microbial crosstalk):. One of the highlighted design approaches for the gut-on-achip has been a scheme in which separate channels are allocated to microbial and intestinal cell environments. A pioneering device in this category has been the HMI platform proposed by Marzorati et al. [Fig. 3(c)]. 86 HMI contains a microbial and a Caco-2 cell culture channel partitioned by a microporous membrane with a deposited mucus layer. The segregated configuration of HMI has shown great advantage in reducing the cytotoxicity of a complex microbiome population on intestinal cells. 86 The device supported a viable coculture for multiple days following more than a week of stand-alone tissue culture. 86 Another example in this category has been a three-channel design, termed HuMiX, developed by Shah et al. 87 This device expanded on the HMI concept by simulating the basal side of epithelial Caco-2 cells, in a separate microchannel, termed perfusion chamber, and adding a porous membrane between the perfusion chamber and epithelial chamber on which epithelial cells were cultured [ Fig. 3(d)]. 87 Various assays demonstrated the HuMiX's ability to hold viable microbiota and intestine tissue cells mimicking numerous in vivo cellular signatures. 87 b. Emulate. A different gut-on-a-chip design approach features two separate channels for modeling the vascular and luminal environments separated by a microporous membrane on which epithelial cells are cultured [Fig. 3(e)]. 88 In this device, the microbial cells are injected into the lumen to be in contact with the epithelium being reminiscent of their real-life intimate arrangement. 88 This device design, which was pioneered by Ingber's group 88 and is being manufactured commercially by Emulate, Inc., has been used in numerous investigations. [89][90][91] The device in its typical configuration harbors chambers on the lateral sides of the microchannels to induce fluctuating vacuum that emulates intestinal peristalsis. 88 This platform has been shown to stimulate key phenotypic changes in the epithelium, such as the spontaneous development of villi topographies and the secretion of mucin. 88,92 The device was shown to maintain an effective viable tissue barrier over a few weeks in cultures of either intestinal cancer cell lines or organoid-derived cells, 71,88,93 and for at least several days in cocultures with the microbiome. 55 In some variations, the device has embedded endothelial and immune cells in the bottom channel to recapitulate the subepithelial zone. 71,94 c. Other designs. Other creative designs have also been developed building on earlier models. Shim et al. developed a platform that FIG. 2. Schematics of the gut cellular composition and structure. The microbiota lives in the lumen, which features an anaerobic condition and a convective fluid flow. Epithelial cells surround the lumen's outer surface and form 3D villi and crypts (the inset shows the cellular composition of a crypt as described in Box I). Below the lumen is the mucosa, which embeds aerobic blood and lymph arteries in lamina propria; this area is populated by a diverse cellular makeup ranging from endothelial cells to immune cells.  The experimental time refers to an example time frame where the device has been reported to stay functional (they might support longer times). Measurements could be collected at earlier times. b These studies are not explicitly designed as gut-on-a-chips but focus on microfluidic modeling of bacterial attachment on 3D villi and crypt-like topographies.

D. Analytical measurements
The modular structure of organ-on-a-chip devices, be it monoor multi-environment, enables each compartment to be integrated with various analytical techniques ( Fig. 1). Being fabricated using transparent materials such as glass and PDMS, high-resolution visualization techniques, like confocal laser scanning microscopy and phase contrast microscopy, allow for in situ monitoring of cells and tissues. 98 Fluorescent and immunofluorescent staining of biomolecules can yield critical information on cell viability and cellular functions, 98 and cellular signatures can be obtained by omics approaches such as genomics, proteomics, transcriptomics, and metabolomics from microchannels. 87,99 Such data can be utilized to highlight physiological changes in each compartment. Furthermore, the combination of fluorescence microscopy with DNA probes can be used to perform genetic and transcriptomic mapping of cells and tissues using the fluorescence in situ hybridization (FISH) assay. 100 Considering the critical barrier role of the gut, barrier integrity assays can be conveniently incorporated into microchips. [101][102][103] These techniques include qualitative methods such as fluorescence immunostaining of tight junctions or quantitative ones such as permeability assessments by injection of fluorescent-labeled molecules or transepithelial electrical resistance (TEER) measurements. 101-103 TEER, which measures ionic conductance across a tissue layer, has particularly been highlighted due to its non-invasiveness and high accuracy. 104 In gut on-a-chip platforms featuring apical and basal channels, TEER measurements were performed across epithelium cultured on a porous membrane to assess tissue integrity (TEER of $4000 X cm 2 reported by Kim et al. 88 and $1000 X cm 2 by Shah et al. 87 were substantially higher than values reported in planar static cultures; for a detailed discussion of TEER, refer to the review by Srinivasan et al. 101 ).

III. GUT-ON-A-CHIP MERITS FOR SIMULATING ENVIRONMENTAL CUES
Gut-on-a-chip designs have so far appeared in different platforms and have exhibited remarkable advantages over conventional counterparts for simulating fluid flow, mechanical forces, and oxygenation conditions ( Table I lists the description of these parameters for various devices). Accordingly, gut-on-a-chip research, above everything, has provided new understanding of how these factors influence gut physiology. Herein, we touch on insights that different gut-on-a-chip models have given by their unique approach.

A. The effect of fluid flow
An indispensable advantage of using microchip systems has been recapitulating gut hydrodynamic conditions. Various monoenvironment devices demonstrated that hydrodynamics significantly impacts the biological behaviors of microbial and intestinal cells. Using the minigut concept, Cremer et al. showed that the luminal fluid flow is a key determiner of bacterial growth kinetics, affecting the rate of metabolic processes and the spatiotemporal distribution of bacteria. 76 Secchi et al., 84 Valiei et al., 105 and Jahed et al. 106 indicated that fluid shear on posts and villi-like topographies can significantly impact bacterial interactions with these surfaces [ Fig. 4(a)]. Notably, these researchers discovered that shear force variation strongly influences bacterial attachment, 84 which, in the latter two works, was shown to instigate the formation of unconventional bacterial structures such as streamers and networks. 105 85 as fluid-protected niches, particularly the cavities inside crypts, create appropriate spots for the buildup of signaling molecules, promoting the formation of bacterial biofilms-this underscores the in vivo observations of crypts' pathogenic invasions. 107 Fluid flow affects intestinal cells through various mechanisms. Interestingly, reports unveiled enterocytes exposed to flow reproduce villi-like and microvilli architectures observed in the natural tissue, whereas those features were lacking under a static culture [ Fig. 4(c)]. 82 Moreover, the dynamic flow initiated important physiological effects such as mucin secretion, cell polarization, and the formation of tight junctions [Figs. 4(c) and 4(d)]. 71,82,88,108 Multi-environment devices simulating both apical and basal environments around the epithelium (as in the Emulate device) could better emulate the physiological effect of fluid flow. 82,88,95,97 Compared to static conditions, such bilateral flows better replicated the barrier permeability and phenotypic characteristics in both Caco-2 and organoid-derived culture models. 71,88 Intriguingly, while Caco-2 cells had been traditionally known to grow to enterocyte-like phenotypes, under a parallel microfluidic flow, they spontaneously differentiated into multiple lineages (absorptive, enteroendocrine, Paneth, and goblet cells) and underwent significant morphogenesis to form tall villi-like formations. 71,88,90,92 The morphogenesis has been recently attributed to the hydrodynamic removal of signaling molecules (particularly Wnt) from the basolateral side of the tissue, which could hint at a new physiological mechanism in the intestine development [ Fig. 4(e)]. 71,108 Finally, fluid flow has been found to be a crucial factor in modulating effective host-microbial interplay. Static cocultures have been shown to cause inharmonious growth of cells and the accumulation of undesirable cellular metabolites, eventually causing the death of epithelial cells and microbial overgrowth. 88 In contrast, the luminal fluid has been essential to extend a viable culture to a few weeks, much longer than cell longevities in Transwells. 55,87,88 Overall, previous works have unanimously demonstrated that fluid flow is an indispensable element in gut physiology without which a viable model of the human intestine is impossible.

B. Addition of cyclic motion
The ability to create cyclic motion is another unique addition to the gut-on-a-chip systems. The mechanical peristaltic motion, whether originating from gas-actuated channels or fluctuating pumps, has been shown to stimulate important physiological effects. 88,97 The minigut model demonstrated that intestinal peristalsis is responsible for mixing the food and microbial population, which is necessary to achieve efficient metabolism. 76 The lack of peristalsis, conversely, resulted in stratified flow leading to instability and microbial washout. 76 Peristalsis's effects on intestinal cells have been identified in a few studies, uncovering effects such as the elevated expression of key enzymes, including cytochrome P450 3A4 (CYP3A4), a pivotal effector in drug metabolism, sucrase-isomaltose, a critical catalyst for sugar decomposition as well as villin, a functional protein found in brush border membranes. 71,92 At the transcriptome level, peristalsis enhanced gene expressions related to ion, lipoprotein, and water transportation, as evidenced in the analysis of organoid-derived epithelial cells on chip. 110 In addition to its key role in digestion, chip models showed that mechanical pulsations induce important tissue responses. 88 When in vitro tissue cultures were exposed to fluid flow and a cyclic mechanical strain, the epithelial permeability measurements revealed more physiologically relevant values than when they were only subjected to  97,108 and the mucin production where the latter, in the case of organoid-derived cultures, recapitulated in vivo values. 71,88,90 In addition, the peristalsis promoted glycocalyx secretion and microvilli formation in epithelial cells. 97 In terms of host tissue-microbiome interaction, although the mechanistic impact of peristalsis is yet unclear, there is evidence to consider mechanical cues a potent factor in this interplay. 97 Kim et al. observed that bacteria exposed to flow without peristalsis grow more than twofold faster than the bacteria exposed to simultaneous flow and peristaltic motion [ Fig. 5(c)]. 94 This perfectly aligned with the observations that bacteria under a flow-peristalsis effect displayed substantially stronger enzymatic activity than static cultures. 88 Altogether, previous investigations, in tandem, confirm that the implementation of mechanical cues guarantees a higher fidelity in the in vitro modeling of the intestine.

C. Anaerobic condition
Although microfluidic devices that solely host intestinal cells are typically operated under aerobic culture media, microbial culture requires an anaerobic culture to grow in body-like conditions [ Fig. 6(a)]. A few studies have so far tested devices with oxygen gradient across microfluidic layers by regulating the media flow and the oxygen concentration. 55,86,87 HMI has been one of the successful platforms for creating differential oxygen level between epithelial cells and microbes, which illustrated that the spatial distribution of bacteria across the lumen and mucosa is heavily impacted by the oxygen availability [ Fig. 6(b)]. 86 Specifically, the strict anaerobes showed a particular tendency to accumulate inside the lumen and the upper layer of the mucosal biofilm, whereas bacteria with more tolerance to oxygen were concentrated at the microaerophilic niche near the base of the microbial entities. 86 Jalili-Firoozinezhad et al., who incorporated an anaerobic enclosure to control the oxygen in the Emulate chip [ Fig. 6(c)], 55 discovered that alteration of oxygen influences the richness and abundance of bacteria. Their observations clearly underpinned the fact that microbiome models could only resemble in vivo compositions when living in a controlled anaerobic niche [ Fig. 6(c)]. 55 Shah et al. further noted creating anaerobic conditions in the lumen substantially impacts the gut tissue physiology besides the microbiome. 87 Investigating Lactobacillus rhamnosus cultures in HuMiX, they detected substantial differences in gene expression patterns for Caco-2 related to metabolism, cellular homeostasis, and interaction with the immune system. 87 These studies suggest that the oxygen gradient across the gut tissue has profound impacts on cellular functions, substantiating that controlling oxic conditions can improve the mimicry of the microbiome and host interactions.

IV. APPLICATIONS OF GUT-ON-A-CHIP DEVICES
The ability to produce hydrodynamical, mechanical, and chemical properties of the gut has made the gut-on-a-chip an excellent platform for launching medical and biological investigations. Gut-ona-chip platforms have so far been used in important research domains, including the pathogenesis of diseases and therapeutic discoveries.

A. Mechanistic insight into diseases (pathology)
Most GI diseases are challenging to study due to their convoluted links to a variety of environmental and genetic factors. 111 Considering that animal models fall short to dissect etiological cause-effect relations due to inherent confounding effects present in complex organisms, 111 gut-on-a-chip platforms, which have engineered simpler structures, have been sought as potential candidates to model multifactorial disease scenarios. In gut-on-a-chip systems, multiple cellular or environmental variables of relevance can be evaluated, and physiological effects derived from each modulation can be characterized in terms of cellular morphology, multi-omics profile, tissue barrier function, cytokine releases, and other factors. [112][113][114] The insights gathered from these simulations can elucidate complex cell-cell, tissue-tissue, and tissue-environment interactions involved in GI diseases. [112][113][114] One well-demonstrated application of on-chip devices has been the simulation of GI bacterial infections as these devices can sustain microbial cells and human tissue cocultures. 89,91,115,116 In one of the earliest case studies, Kim et al. devised a microfluidic system to model the epithelial infection caused by Enterohemorrhagic E. coli (EHEC) [ Fig. 7(a)]. 115 The device was designed to have a modular configuration to simulate the scenarios involving the invasion of pathogenic bacteria into the gut tissue after contact with commensal bacteria. These simulations uniquely uncovered that the signaling molecules from commensal E. coli have a strong prohibiting effect on EHEC infection, pointing to a critical disease defense mechanism. 115 Later, using the Emulate platform to compare EHEC infection between mice and humans, Tovaglieri et al. illustrated that microbiome-derived metabolites can alter the phenotype of pathogenic bacteria (e.g., chemotaxis and motility), causing tissue damage [ Fig. 7(b)]. 89 In another study, Gazzaniga et al. used a similar platform to decipher the effect of the microbial composition against Salmonella typhimurium. 116 They discovered that specific strains in the microbial composite (in their case, Enterococcus faecium) can elevate the host tolerance against the infection. 116 The validity of gut-on-a-chips for infectious disease studies is also denoted by Grassart et al., who demonstrated that the ability of these devices to model tissue topography, fluid flow, and peristalsis is of high significance to model the Shigella bacteria's invasion mechanism [ Fig. 7(c)]. 91 Gut-on-a-chip designs have also exhibited noticeable applicability for modeling viral infections. 117,118 By incorporating a dualchamber model with vascular and luminal environments, Villenave et al. captured the full viral pathogenesis cycle of the coxsackievirus B1, including viral infection, replication, and propagation. 117 The researchers notably tracked the infection path upon viral injection through the lumen or blood channel, which unveiled that final tissue damage results from an apoptotic mechanism that is triggered by the virions traveling preferentially to the apical side regardless of the point of injection. Bein et al. and Gou et al. proved that the gut-chips present unique features to model coronaviruses infections, the causative agents of diseases such as the common cold (NL63) and COVID-19 (SARS-CoV-2) [ Fig. 7(d)]. 118,119 Importantly, both studies showed that angiotensin-converting enzyme 2 (ACE2) receptor, a key means of entry for coronaviruses, and protease TMPRSS2, a viral spike (S) protein priming element, are efficiently expressed in on-chip models. 118 Fig. 7(d)]. These simulations have unveiled powerful means to complement animal studies considering the disparity between animal and human physiology in response to viruses.
Furthermore, a number of attempts have been made to examine complex GI diseases such as IBD and colorectal cancer with gut-on-achip models. A mimicry of gut tissue inflammation as a precursor to IBD has been presented by Kim et al. in a gut chip involving immune cells and bacterial endotoxins. 94 The authors discerned a disease pathway based on the combined secretion of pro-inflammatory cytokines Interleukin 8 (IL-8), IL-6, IL-1b, and tumor necrosis factor alpha (TNF-a) by epithelial cells upon exposure to bacterial lipopolysaccharides (LPS) and PBMCs [ Fig. 8(a)]. 94 The critical role of the immune function in the intestinal barrier has been reinforced in the studies of Beaurivage 120 on the IBD and Apostolou 110 on the leaky gut [Figs. 8(b) and 8(c)]. By screening the effect of various cytokines, these studies identified that the signaling molecules including IL-1b, 120 TNFa, 120 interferon-gamma (IFN-c), 110,120 and IL-22 110 are major contributors to the morphological damage and the increased permeability of epithelium. Strelez et al. have further illustrated the utility of gut-on-achip systems for colorectal cancer explorations [ Fig. 8(d)]. 121 Once tumor cells were introduced into the lumen, they disrupted the metabolomic behavior of epithelium and endothelium, traversed the epithelium, and then partially dispersed in the blood circulating channel. 121 The model was consequently deemed suitable to assess tumor invasion and identify physicochemical markers in cancer research. 121 Overall, these findings highlight that the modular structure of gut-on-a-chips is extremely valuable in simulating gut pathophysiology by decomposing disease scenarios into less-complex case studies. These advantages as demonstrated in recent triumphs with gut-on-achip technologies herald a new era in modeling and comprehending complex diseases.

B. Discovery of therapeutics
Another major application of the gut-on-a-chip is assessing drug metabolism. In the case of orally administrated medications, the drug's bioavailability can substantially subside by its traveling through multiple organs before reaching the target tissue. 122 The intestine, in particular, has been identified as a critical determiner of drug efficiency in the first pass drug metabolism. 122 Drug digestion in the intestine is mediated by a host of drug-metabolizing enzymes and transporter proteins secreted by enterocytes, including CYP3A4 and P-glycoprotein, 123 and is additionally influenced by the metabolic activities of microbiota. 124 Currently, animal and in vitro testing are routinely conducted to predict intestinal drug digestion; however, the former exhibit substantial differences in physiology and microbial architecture, while the latter is deficient in physiological-level drug-digesting enzymes. 41,125 Recent evidence suggests that gut-on-a-chip devices could fill the existing voids in drug discovery by presenting multiple benefits. Data on various benchmark drugs such as verapamil and ifosfamide have shown that Caco-2 cells, once cultured in the intestine chips, expressed elevated drug-digesting enzymatic activity compared to static cultures. 83,92 When organoids were used as the cell source, the predictability of the drug responses even further improved. 125 Being tested against model drugs rifampicin and 1,25-dihydroxyvitamin D3, cells derived from human duodenum organoids exhibited an expression level of CYP3A4 and intestinal nuclear receptors that were akin to the physiological response. 125 In addition, gut-on-a-chip systems have shown to be effective for assessing other physiological effects, such as the drug absorption through the intestinal barrier. 126 This has been useful to test the efficacy of new drug formulas for oral administration as demonstrated for a modified version of the anticancer drug SN-38 (7-ethyl-10-hydroxy camptothecin). 126 Beyond broad-scoped drug testing, a handful of studies proved that gut-on-a-chip devices could be particularly apt for developing therapies for GI disorders, as they can simulate both the disease and the treatment strategy at once. In the gut inflammation chip, for example, the administration of antibiotics into the injured tissue induced a curing effect due to the suppression of Enteroinvasive E. Coli (EIEC) bacteria, which corroborated clinical studies. 94 In a different work, the assessment of the anti-inflammatory compound TPCA-1 elucidated therapeutic effects caused by the reduction of inflammatory cytokines. 127 Another related use case has been the simulation of NL63 coronavirus infection on a gut-on-a-chip to showcase drug responses of nafamostat and remdesivir in vascular injection (blood channel) and toremifene, nelfinavir, clofazimine, and fenofibrate in oral administration (luminal channel) scenarios. 119 The drug studies have been extended to probiotic formulations as well, which is a popular research topic among the medical commuinity. 128 An example work has been the injection of a probiotic mixture of Lactobacilli and bifidobacteria species in the gut inflammation chip, which alleviated pathogenic infections and improved the barrier function. 94 Nelson et al. characterized the effect of an engineered bacterium (synthetic biotics) for treating Phenylketonuria, a metabolic genetic disorder associated with the reduced digestion of phenylalanine in the body. 129 The guton-a-chip harvested the dose-dependent metabolism of phenylalanine and yielded results aligning with in vivo tests. 129 Overall, the preliminary drug discovery research based on gut-on-a-chips has been extremely promising due to the unprecedented advantages these devices offer. Accordingly, research on a broader range of drugs and therapeutics using these devices is anticipated in the near future.

C. Clinical and industrial translation
The ability of the organ-on-a-chip to simulate diseases and drug metabolism has spurred enormous interest in commercializing these devices for clinical and industrial applications. 44 Currently, the drive to reduce reliance on animal models is strong in biomedical, food, cosmetics, and other related industries, 39,130,131 which has prompted regulatory bodies such as the US Food and Drug Association and the European Medicines Agency to seriously evaluate the technology 28,132 (for a list of organ-chip manufacturing companies, refer to Singh et al. 130 ). In the biomedical industry, in particular, the demand for a robust in vitro platform has been growing considering the present lengthy, complicated, and time-consuming drug discovery process. 133 The high attrition rate of drugs is mainly attributed to false predictions in preclinical tests, driving the cost of each new medicine up to one billion dollars. 134 TABLE II. Select studies on multi-organ-on-a-chip platforms involving the gut.

Authors
Type of the chip Example case study Choe et al. 142 Gut-liver First pass metabolism of apigenin Lee et al. 143 Gut-liver Gut absorption and liver metabolism of fatty acids De Gregorio et al. 144 Intestine-liver First pass metabolism of ethanol and the resulting hepatic damage Chen et al. 145 Intestine-liver Urea and albumin metabolism and CYP enzyme activity Chen et al. 146 Gut-liver Inflammatory inter-tissue crosstalk Maschmeyer et al. 147 Liver-intestine Repeated dose administration of troglitazone Prot et al. 148 Intestine-liver First pass metabolism of paracetamol Kim et al. 149 Gut-brain axis Exosomes transport across the gut barrier toward the blood-brain barrier Lee et al. 150 Gut-kidney Antibiotic treatment effect on hemolytic uremic syndrome in Shigaproducing E. coli infection Lee et al. 151 Gut-skin The effect of gut-absorbed fatty acids on the skin upon gut inflammation Kimura et al. 152 Lung-intestine-liver The pharmacokinetics studies of three anticancer drugs-epirubicine (EPI), irinotecan (CPT-11), cyclophosphamide (CPA) Ramme et al. 153 Intestine-liver-brain-kidney Generation of four organs from induced pluripotent stem cells Vernetti et al. 154 Intestine-liver-kidney-blood brain barrier-skeletal muscle Absorption, metabolism, and excretion of terfenadine, trimethylamine (TMA), and vitamin D3 Imura et al. 155 Intestine-liver-cancerous breast Absorption, metabolism, and bioactivity of cyclophosphamide, epirubicin, 17-b estradiol, and soy isoflavone Furthermore, mechanistic understanding of many diseases is still out of reach, due in part to the absence of rigorous models. 135 Considering the promising results obtained from single organ-onchip devices, a major initiative has focused on the development of composite devices to mimic broader physiological features, setting forth the concept of "human-on-a-chip" devices. 28,136 In these devices, separate compartments are devoted to different organs that are connected to each other by fluidic channels or fluid movers. 137 An example composite device is the gut-liver-on-a-chip, which, owing to the central role of the gut and liver in metabolism, has been proposed as a testing platform for drug toxicities and disease studies. 138,139 Other multi-organ devices involve the gut combination with the kidney, brain, and skin (Table II lists example studies using these devices). Both single-and multi-organon-a-chip devices are envisioned to become game changers in the future, yet much work remains to be done in standardization and validation of the platform against benchmark methods. 132,134,140,141

V. GUT-ON-A-CHIP LIMITATIONS AND FUTURE PERSPECTIVES
In Secs. II-IV, we described the main gut-on-a-chip technologies and elaborated on how they could be useful in pathology and drug discovery. However, just like any other biological model, gut-on-a-chips have their own shortcomings. In this section, we review some of the limitations of this platform aiming to illuminate potential future research directions.

A. Miniaturization
Gut-on-a-chip devices are miniaturized versions of the human tissue, which cannot accurately recapitulate mechanisms that manifest at larger scales. For example, while food digestion in the intestine occurs progressively through the length of the intestine, gut-on-a-chip systems, due to small channel sizes, merely offer a short snapshot of gut metabolic and physiological processes and, thus, could overlook spatial variations of physiological parameters.

B. Time limitations
Recent research has witnessed an increase in the lifetime of tissue and microbial cultures in the gut-on-a-chip for up to a few weeks, yet these devices cannot mimic real-life physiological time scales. This may cause inaccuracies in modeling temporal changes in microbiome composition and diseases that develop over years. 156,157 Research into new techniques, e.g., cell culture renewal, is essential to solving this issue by improving tissue longevities in chip models.

C. Limitations of fabrication materials
The synthesis of microfluidic scaffolds still relies heavily on the usage of artificial polymers such as PDMS. These materials, despite having practical advantages such as gas permeability and ease of fabrication, can pose technical challenges due to their absorption capacity for numerous biomolecules. 158,159 This can disrupt the bioanalytical data derived from tissue cultures and interfere with drug studies. 158,159 Innovative remedies for the fabrication materials and methods are required to address this issue.

D. Diversity of cell types and complexity of the microbiome
Gut-on-a-chip devices have offered the possibility of culturing multiple cell types in one device; however, the human cellular system is still far more diverse and complex. The microbiome, for example, contains viruses, fungi, yeasts, and bacteria, and the gut mucosa is packed with fibroblasts, fibrocytes, endothelial, blood, and immune cells, which are all important in shaping the gut physiology (see Box I).
One hurdle for maintaining all the cells together, however, is their different requirement of growth media and microenvironmental conditions. Extensive modularization could potentially boost cellular diversification; however, further research is needed to tackle operational issues that may arise in the resulting devices.

VI. CONCLUSIONS
In this study, we reviewed how microfluidic devices provide an innovative approach to model gut physiology. We highlighted that gut-on-a-chip devices can tackle two critical limitations of the previous in vivo models simultaneously: (i) the inability to capture the function in three dimensions; and (ii) the modeling of physiochemical environmental cues which are key determinants of myriad physiological processes. So far, a variety of gut-on-a-chip schemes have been proposed. A common example of gut-on-a-chip devices is a multicompartment device where lumen and vascular systems are mimicked in juxtaposed microchannels, and a porous membrane in-between supports the growth of the gut epithelium. This platform allows for the simulation of fluid flow in the lumen and the mucosa as well as the differential hydrodynamic shear experienced by the epithelium. The peristalsis has been additionally incorporated by actuating gas pressure in adjacent ancillary microchannels or peristaltic liquid pumps. Moreover, the oxygen in each compartment has been successfully regulated through the adjustment of oxygen in the solution media and the incorporation of an anoxic enclosure. Numerous studies have demonstrated that the exposure of the cell to relevant hydrodynamics, mechanical forces, and chemical gradients results in the genetic and phenotypic expressions that mimic the in vivo condition, much more accurately than organoids and 2D models. Given the success of gut-on-a-chip platforms for reproducing a wide range of biological functions, these devices have been explored for a variety of clinical applications, especially in disease modeling and drug discovery. Numerous investigations have showcased that the gut-on-a-chip devices are highly meritorious for modeling a plethora of infectious and gastrointestinal diseases, in addition to testing various therapeutic interventions, including drugs and probiotics. Despite recent success, gut-on-a-chip research is still a nascent field, and extensive work remains to be conducted to overcome technical hurdles arising from fabrication limitations and operational issues resulting from miniaturization.

AUTHOR DECLARATIONS Conflict of Interest
The authors have no conflicts to disclose.

Ethics Approval
Ethics approval is not required.

DATA AVAILABILITY
Data sharing is not applicable to this article as no new data were created or analyzed in this study.