Microfluidic tools to model, monitor, and modulate the gut–brain axis

The gut–brain axis (GBA) embodies bidirectional communication between the gastrointestinal tract (GI) and the central nervous system (CNS). Gut microbiota that colonize the GI tract play a crucial role not only in modulating gastrointestinal function but also in regulating immunological responses and neural function through the secretion of small molecules. Consequently, the disruption of the homeostatic state leads to gut dysbiosis, contributing to disorders such as obesity and neuropsychiatric conditions. Conversely, signals from the brain can influence the gut microbiota, creating an intriguing two-way communication.

The humoral (e.g., soluble factors in circulatory and lymphatic systems) and vagus nerves both take part in the GBA. The former involves the transport of bacterial products, metabolites, and other intestinal molecules through intestinal barrier into the circulation, which can influence the CNS upon transport through the blood–brain barrier (BBB). For the latter, vagal afferent neurons (VANs) play an especially crucial role in receiving signals from the peripheral organs, such as the intestine, and transmitting them to the CNS. The VAN cell bodies reside in the nodose ganglion, adjacent to the carotid artery. Inside the nodose ganglion, the VANs have a pseudounipolar morphology, with one branch of neurites extending to the peripheral organs (e.g., GI tract) and the other onto the brainstem. The molecules produced by the gut microbiota can directly modulate the activity of VANs, and this, in turn, has been hypothesized to influence many high-level functions of the CNS (e.g., mood, cognition).¹ However, it is difficult to establish causation between GBA mechanisms and their manifestations because higher-level dysfunctions result from many simultaneous physiological processes. Thus, there is a need for physiologically relevant platforms that can discriminate specific processes (e.g., humoral or vagal routes) to study the GBA. Traditional cell culture methods, where cells are cultured in a Petri dish, often fail to model the anatomical, hence physiological, compartmentalization, such as in the case of GBA.

Microphysiological platforms, usually built by incorporating cells and tissue in microfluidic devices, have been instrumental in studying distinct physiological processes in an isolated, hence more controlled, manner. Advantages of these devices include the ability to control cell–cell interactions and exposure to spatiotemporally varying signals (e.g., soluble factors)² and incorporation of sensors for a multitude of readouts (e.g., electrophysiological activity, oxygen, glucose, inflammatory markers),^3,4 as well as the use of mechanical forces to mimic in vivo conditions (e.g., peristaltic motion of the gut to increase mucus secretion).⁵ In order to set the ground for this perspective article, we point to recent review articles highlighting the use of microfluidic chips to model the GBA.^6–8 We envision a microfluidic system that embodies such critical features to enable a broad range of studies related to the GBA (Fig. 1). The proposed platform particularly emphasizes the ascending (sensory) vagal neuron-based communication between the gut and the brain, integrating the existing efforts that have mainly advanced individual components of the model (e.g., gut epithelium, cortical cell culture). While such a device will likely take years to conceive in its entirety, we hope that it will inspire further research toward its development. Here, we non-exhaustively describe the state-of-the-art in these technologies and discuss the utility and potential testbeds to study the interplay between the microbiota, neuroinflammation, and neurodegeneration.

CNS is modulated by signals originating from peripheral organs via afferent neurons as well as intrinsic signals within the brain. For the former, the gut microbiota and associated signaling molecules influence both the development/regeneration (e.g., myelination, neurogenesis)⁹ and neuroimmune processes (e.g., microglial response).¹⁰ For example, in a germ-free (GF) mice with partially eradicated microbiota, microglia (innate immune cells of the CNS) exhibited an impaired immune response. This was attributed to the absence of short-chain fatty acids (SCFAs) generated by the bacterial fermentation of dietary fibers in an intact microbiota.¹⁰ 

Neurodegenerative conditions also often have microbiota in their etiology. For example, Parkinson's disease (PD), characterized by neural death in the substantia nigra and the emergence of Lewy bodies (LBs) accompanied by motor symptoms, is an example of a neurodegenerative disease in which gut microbiota might be implicated. Specifically, a pathogen or α-synuclein originating in the gut may be transported to the brainstem via the vagus nerve and lead to the formation of Lewy bodies as they aggregate with other proteins, resulting in neuronal toxicity and inflammation.^11,12 To model this postulated mechanism of the PD pathogenesis, that is, the propagation of aberrant proteins from the GI tract to the brain stem and then to a specific brain region via cell-to-cell transmission, a microfluidic platform is well suited. To that end, a microfluidic platform should enable the compartmentalization of neural cells into two distinct laterally positioned chambers while allowing axonal extension from one chamber to the other through microchannels. For example, in order to validate the propagation of α-synuclein (an aberrant protein implicated in the etiology of Parkinson's disease) in CNS via neurons-to-neuron transmission, primary cortical mouse neurons were cultured in a microfluidic platform where cell bodies were compartmentalized in a “soma” chamber with axons projecting through microchannels to an adjacent chamber [Figs. 2(a) and 2(b)].^13,14 When the soma chamber was treated with fluorescently labeled α-synuclein, uptake and transport of α-synuclein fibrils from the first-order neuron to the second-order neuron was observed in real-time. While this study was performed with cortical neurons, it suggests that the neuronal transport of aberrant proteins from the gut to brain via vagal afferent neurons is plausible taken together with findings from another study that demonstrated the transport of a surrogate protein (fluorescently -labeled cholera toxin) via microfluidically compartmentalized vagal afferent neurons.¹⁵ 

In addition to the transport and deposition of aberrant proteins, neuroinflammation is a key component of neuropathogenesis, involving multiple cell types such as neurons, astrocytes, and microglia. In this context, a microfluidic device with two chambers was developed to investigate the etiology of Alzheimer's disease (AD). The device had a central chamber for neurons and astrocytes, and the concentric outer chamber housed adult microglia that can migrate to the inner chamber via interconnecting microchannels [Fig. 2(c)].¹⁶ The model mimics the AD environment by culturing neurons expressing high levels of aberrant amyloid-beta (Aβ) peptides and astrocytes to study Aβ accumulation, hyperphosphorylated-tau formation, and the resulting activation and recruitment of microglia from outside chamber to the inside chamber. Although current in vitro CNS models, enabled by microfluidic compartmentalization, have provided important insight into neuropathological mechanisms, studies involving the influence of microbiota-derived factors on neurodegenerative and neuroinflammatory conditions are in their infancy.

The intestinal environment contains a multitude of microbes in contact with each other and the gut epithelial lining. Intestinal cells form a barrier to bacterial invasion and to control the transport of microbiota-derived compounds into circulation, which subsequently encounter the BBB for further selectivity in accessing the CNS cells. Transport across the epithelial barrier occurs via trans-cellular pathways including membrane diffusion, receptor- or carrier-mediated mechanisms, and paracellular pathways, where tight-junctions between the epithelial cells are critical for the barrier integrity.¹⁷ Additionally, the epithelium is innervated via VANs with neurite terminals situated in lamina propria (tissue at the basolateral side of epithelium) providing a direct communication route to the brain.¹ Thus, gut epithelium models are centered around including intestinal cell types (e.g., enteroendocrine cells, enterocytes, goblet cells, Paneth cells, M cells),¹⁸ bacterial cells (aerobic and anaerobic varieties),¹⁹ and recently neural cells (e.g., enteric neurons)²⁰ with the goal of capturing the influence of intestinal molecules on the epithelium and neural response. These cells are often cultured in a stacked (layered-channel) configuration to compartmentalize the distinct cell types within microfluidic channels representing apical and basolateral spaces,⁵ which is applicable to modeling various anatomical units (e.g., lung alveolus,²¹ BBB²²). This general microfluidic architecture provides a versatile environment that promotes the physiological function of the epithelium (e.g., barrier function, mucus secretion)^23,24 as well as the sustentation of anaerobic environments for relevant bacterial species [Fig. 2(d)].¹⁹ 

While there are various examples of epithelial barrier models enabled by microfluidic compartmentalization, only a few studies emulate the innervation of gut epithelium with enteric neurons.^20,24 Microenvironments incorporating intestinal epithelial cells, and primary neuronal and enteric cells exhibit better differentiation, maturation, and barrier function as a consequence of innervation of enteric neurons and the epithelial cell types. Such systems utilize clear cover glass slides to visualize innervation between the enteric neurons and epithelial barrier using conventional light microscopy [Fig. 2(e)].²⁰ 

The BBB is analogous to the intestinal barrier whereby a neurovascular unit, involving vascular endothelial cells, astrocytes, and pericytes, provides a dynamic barrier between the brain and circulatory system.^25,26 Similar to microfluidic models of the intestinal epithelial barrier, the BBB is modeled via two layered laminar flow chambers comprising a brain environment (e.g., primary or induced pluripotent stem cell (iPSC) derived neurons and glia) and vascular lumen (e.g., endothelial cells) separated by a perforated membrane [Fig. 2(f)].²² This setup allows for the assessment of barrier function through the molecular transport rate and trans-epithelial-electrical-resistance (TEER) analyses, and investigating the influence of transported molecules (e.g., pharmaceuticals) on the neural cell response. Another commonly used model is the lateral-channel architecture that is advantageous for compartmentalizing cell populations in different chambers yet allowing their signaling or axonal connectivity via interconnecting microchannels. A recent study has utilized this architecture to create vascular lumen for the blood flow, CNS environment (e.g., containing parenchymal cells—neuron, astrocyte, pericytes, and microglia), and cerebrospinal fluid (CSF) chamber to investigate therapeutics for ischemic stroke [Fig. 2(g)].²⁷ Similar to the layered-channel architecture, the assessment of barrier function, live imaging, and protein assays are possible with these devices.

While there have been reports of microfluidic models of the BBB with neuronal integration, such studies for the gut epithelium-peripheral neuron integration are limited. Furthermore, current gut-on-a-chip systems are showing promise in mimicking in vivo environment but are yet to delve into the next phase of therapeutic or detrimental compound screening.

The vagus nerve—a critical conduit between the GI tract and CNS—is composed of vagal efferent neurons (VENs) and VANs in a 1:4 ratio. VAN cell bodies reside in the nodose ganglia and transmit sensory information from the GI tract to the CNS (more specifically to the brainstem and subsequently to cortical regions), influencing appetite, metabolism, and plausibly higher-level functions like mood. Conversely, VEN cell bodies are located in the brainstem, transmit signals from the CNS to the gut, control digestion and heart rate, and as recently demonstrated, the neuro-immune pathway, notably via cholinergic anti-inflammatory means.²⁸ Studying the specific roles of VENs and VANs in vivo is challenging due to other confounding physiological processes, motivating the need for their in vitro model counterparts. However, unlike CNS neural cell cultures, there are very few examples of in vitro VENs and VANs models. This is, in part, due to the non-trivial nature of harvesting VANs (e.g., surgical extraction of nodose ganglion) and small cell numbers, resulting in low seeding density. VENs are even more challenging to obtain since the cell bodies in the brainstem are not easy to identify and isolate. Existing acute VAN cultures combined with patch-clamp techniques have been useful for studying electrophysiological response to effector molecules (e.g., satiety hormone cholecystokinin).²⁹ However, since in most acute VAN cultures, the cell bodies and neurite projections are cultured in the same chemical environment (e.g., Petri dish), they do not accurately recapitulate the anatomical separation of VAN terminals (i.e., neurite terminals in both the gastrointestinal tract and the brainstem).

Microfluidic compartmentalization provides a means to mimic the anatomical layout of the VANs. Specifically, Girardi et al. have developed a microfluidic platform consisting of two concentric chambers connected by radial microchannels. This device physically separates the cell bodies of dissociated nodose ganglia (consisting of VANs and supporting cells) from the neurite terminals [Fig. 2(h)].¹⁵ Utilizing the anatomical compartmentalization, they have further demonstrated that fluorescently labeled cholera toxin (surrogate for aberrant proteins that can migrate to the CNS) was taken up by the VAN neurite terminals and traveled to the cell body. The compartmentalized VAN culture enables the study of the underlying mechanism of neurodegenerative diseases such as Parkinson's disease. Integration of surface multiple electrode arrays (MEAs) within the device allows for monitoring the electrophysiological activity of the VANs with a path toward obtaining the electrophysiological fingerprints of the impact of the gut-derived molecules on neuronal activity [Fig. 2(i)].³⁰ 

Once the appropriate microfluidic layout and cell types are selected, various modalities of sensors and actuators can be integrated into the microfluidic platform to monitor and modulate the cultures depending on the specific experimental objectives. A comprehensive review of such sensor and actuator technologies can be found elsewhere.^31,32 Most conventional bioanalytical techniques are routinely used in conjunction with microphysiological platforms, including high-content imaging, genomics, transcriptomics, and proteomics.³³ In addition, some sensing modalities can be directly integrated into microphysiological platforms, such as surface electrode arrays (electrophysiological recordings,³⁴ TEER measurements,³⁵ aptamer-based sensors against protein, nucleic acids, and small molecules³⁶). In terms of modulation strategies, the flow rate is often tightly controlled via miniature pumps and syringe pumps, or by adjusting hydrostatic pressure and channel fluidic resistances. Fluid flow is critical not only for delivering soluble factors to the cells and removing waste products but also for imposing controlled mechanical stresses to cells (e.g., peristaltic motion of intestines)⁵ to maintain their physiological functions. Laminar flow in microfluidic channels has also enabled innovative design to establish time-varying concentration gradients in microfluidic chambers to study cellular chemotaxis and stimulation.^2,37

Microfluidics are a versatile tool to model in vivo systems while allowing high-level control and engineering of cellular environments in vitro. A major advantage of in vitro models is reducing the complexity of the physiological system to focus on specific biological questions. In the case of the GBA, an in vitro model allows for decoupling the humoral and vagal routes of interaction between the GI tract and the CNS. However, as it is the case for all model systems, it is important to be cautious of any artifactual findings, reproducibility of modeled biological phenomena, and in vivo relevance of the results. To that end, there are continuous efforts to validate and improve these in vitro models. Table I provides the examples of such devices (particularly two distinct device architectures, that is, lateral- and layered channels, as illustrated in Fig. 2) entering the commercial space and enabling studies of the gut–brain axis.

Despite the significant advancements in the use of microfluidics for modeling the gut–brain axis, there is still exciting work to be done. In this section, we will comment on the major remaining challenges, which are possible paths of further research. The microphysiological models of CNS, peripheral nervous system (PNS), and GI tract all comprise many different cell types, some of which, like neurons, are post-mitotic, and, hence, the limited number of cells obtained from an animal reduces the scalability of the models. While the cortical tissue from rodents yields a large number of neurons, VANs obtained from dissociated nodose ganglia are significantly fewer. Low cell numbers not only limit the number of concurrent experiments but also pose a challenge for MEA-based extracellular recordings since the probability of a cell residing next to an electrode site decreases. It would be impactful if VANs can be obtained from iPSCs, as this would increase the number of cells without utilizing animals and allow for establishing human cell culture models for higher translational potential. iPSCs are expected to eventually replace animal studies, especially in light of a 2019 United States governmental report emphasizing the use of organ-on-a-chip platforms to bypass animal testing.⁴⁴ While there have been many protocols describing the differentiation of iPSCs to different CNS cell types (e.g., neurons, astrocytes, microglia),^45,46 such knowledge for obtaining VANs is in its infancy. As mentioned earlier, VENs are even harder to isolate since their cell bodies are located in the dorsal motor nucleus of the vagus at the brainstem. Therefore, iPSC-based approaches to obtain VENs would also be a significant advancement for the field and pave the way for a large number of studies on the inflammatory reflex,⁴⁷ where cholinergic neurons modulate inflammation at the gut. However, iPSCs-derived neurons also have some weaknesses. It is difficult to consistently generate large numbers of mature iPSC-derived neurons with proper functionality, which demonstrates distinct axons and dendritic extensions as well as firing action potentials. These features inform neural function (e.g., electrophysiological activity), which is central to information transfer and processing in the gut–brain axis. Nevertheless, further research in iPSC-derived VANs and VENs possesses a high potential impact for accelerating progress in in vitro models.

The layout for VANs in culture is another critical consideration. Neurite projections of VANs innervate both the GI tract and extend to the brainstem; therefore, it is important to direct the maturation of these distinct anatomical phenotypes. In simple terms, the “gut” terminals should predominantly express the receptors, while the “brain” terminals should possess a glutamatergic phenotype to secrete glutamate in response to stimulation at the “gut” terminals. Microfluidic compartmentalization, possibly via a three-chamber device where the soma lies in the middle and the neurite projections extend into the adjacent chambers (as illustrated in Fig. 1), can enable exposing the individual neurite terminal groups to distinct microenvironments via customized media constituents and direct their appropriate biomolecular specialization. It is important to consider invasive nature of VAN isolation, that is, during the removal of the nodose ganglion, the cell body is detached from its projections to the gut and the brain. Put another way, the aftermath of the isolation procedure is similar to neuronal regeneration following axotomy. Girardi et al. showed that neurons exhibit remarkable potential to regenerate neurites and express at least Transient receptor potential vanilloid 1 receptors (synthesized at the soma and trafficked to the neurite terminals).¹⁵ There is a need for focused studies to determine types and spatial distribution of other receptors [e.g., aryl-hydrocarbon, serotonin, capsaicin, cholecystokinin (CCK)] at the terminals following neurite regeneration and eventually to guide the expression of specific receptors-of-interest.

Finally, for microfluidic barrier models, gut epithelium and BBB models should be coupled to accurately represent the GBA. This will be an advancement over the state-of-the-art, where “gut” and “brain” regions are modeled as isolated systems in their distinct culture dishes or microphysiological platforms. Integrating multiple microphysiological platforms gives rise to other challenges, such as exposing specific tissue to their optimal culture conditions (e.g., culture media composition, media renewal frequency) and controlling cellular invasion/migration across cultures while they are cultured in concert with other tissue systems, especially when inter-kingdom (bacteria and mammalian cell) co-cultures are used.^5,19 Microfluidics is ideally matched for this task, where the significant knowledge on flow control schemes and device architectures from the last few decades will be instrumental. Including immune cells in the gut compartment will further advance the biological relevance of the GI tract models, since the crosstalk between the epithelial cells, enteric neurons, vagal nerve terminals, and immune cells orchestrate inflammatory processes that underlie diseases such as Crohn's and is hypothesized to play a role in shaping higher-level function of the CNS such as mood. Conversely, signals from the brain and how they modulate inflammation in the gut have become a vibrant research area.

In conclusion, the microfluidic compartmentalization of the critical cell types and the ability to monitor and modulate the physiological processes with existing and emerging sensors and actuators is expected to create many opportunities to study the gut–brain axis, screen therapeutics (e.g., electrical stimulation, microbiota-derived compounds), and translate the findings to inform dietary recommendations and clinical approaches.

We gratefully acknowledge the support from the National Institutes of Health via NINDS/NIA R03-NS118156, NCCIH R21-AT010933, NIEHS P30-ES023513, and NIBIB R01-EB034279, from the National Science Foundation via DMR-2003849 and DGE-2152260, and from the University of California, Davis, via College of Engineering Next Level Research Seed Funding.

The authors have no conflicts to disclose.

H.K. and G.G. contributed equally to this paper.

Hyehyun Kim Conceptualization (lead); Writing – original draft (equal). Gregory Girardi: Conceptualization (lead); Writing – original draft (equal). Allison Pickle: Conceptualization (supporting); Visualization (equal); Writing – original draft (equal). Testaverde S. Kim: Conceptualization (supporting); Writing – original draft (equal). Erkin Seker: Conceptualization (lead); Funding acquisition (lead); Supervision (lead); Writing – original draft (equal).

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