The microbiota‐gut‐brain axis and epilepsy from a multidisciplinary perspective: Clinical evidence and technological solutions for improvement of in vitro preclinical models

Abstract Epilepsy is a common neurological disease characterized by the enduring predisposition of the brain to generate seizures. Among the recognized causes, a role played by the gut microbiota in epilepsy has been hypothesized and supported by new investigative approaches. To dissect the microbiota‐gut‐brain (MGB) axis involvement in epilepsy, in vitro modeling approaches arouse interest among researchers in the field. This review summarizes, first of all, the evidence of a role of the MGB axis in epilepsy by providing an overview of the recent clinical and preclinical studies and showing how dietary modification, microbiome supplementations, and hence, microbiota alterations may have an impact on seizures. Subsequently, the currently available strategies to study epilepsy on animal and in vitro models are described, focusing attention on these latter and the technological challenges for integration with already existing MGB axis models. Finally, the implementation of existing epilepsy in vitro systems is discussed, offering a complete overview of the available technological tools which may improve reliability and clinical translation of the results towards the development of innovative therapeutic approaches, taking advantage of complementary technologies.


| INTRODUCTION
In the last two decades, the bidirectional connection between the gut microbiota and the brain has been extensively investigated; this link is referred to as the microbiota-gut-brain (MGB) axis and involves underlying biological pathways including neural, endocrine, metabolic, and immune system. 1 Gut bacteria from different areas of the gastrointestinal (GI) tract can contribute to the central nervous system (CNS) development (e.g., neurogenesis, microglia maturation, and myelination), functions (e.g., cognition, mood, and behavior), and can Federica Fusco and Simone Perottoni contributed equally to the work. also influence the pathogenesis and progression of different childhood and adult brain disorders (e.g., Parkinson's and Alzheimer's diseases, schizophrenia, epilepsy, autism spectrum disorder, and multiple sclerosis). [1][2][3][4][5][6][7][8] Epilepsy is a common neurological disease, affecting 50 million people worldwide, and it is characterized by the enduring predisposition of the brain to generate seizures. 9,10 Etiology may be genetic, structural, metabolic, infectious, or immune, but up to 50% of cases are still classified as idiopathic. 9,11 The hypothesis of an interplay between the gut microbiota and epilepsy dates back to the beginning of the 20th century, with the concept of "Bacillus Epilepticus." 1,12 The widespread of newer approaches (e.g., next-generation sequencing, gnotobiology, and metabolomic) is allowing researchers to become increasingly aware of the role played by the gut microbiota in epilepsy, especially for those severe cases commonly refractory to conventional therapies, regardless of etiology. 1,[13][14][15][16] The need to dissect the biological mechanisms underlying the role of the MGB axis in epilepsy and the issues related to synergic model-

| THE MGB AXIS AND EPILEPSY
Starting from the antenatal and neonatal life, a wide variety of stimuli could act to mainly influence the composition of the host gut microbiota, which represents the thousands of microbial species inhabiting the GI tract. [17][18][19][20][21] Increasing evidence is highlighting the systemic modulation properties of this complex GI ecosystem, potentially affecting not only immune, but also neurological processes, and leading the way to the concept of a MGB axis. 1,20 The underlying mechanism of interaction seems to dwell on a bi-directional interplay. 22 Alterations in the gut microbiota provide the basis for a systemic immune activation, which elicits a mirrored inflammation in the CNS, while on the other hand neurological dysfunctions may induce systemic inflammation, which in turn directly acts on the gut microbiota features. 23,24 This set of alterations can frequently be the basis for neuronal excitability dysfunction and epileptogenesis, leading to seizure susceptibility and supporting the clinical implication of the MGB axis in epilepsy. 25,26 In this scenario, a central role of the blood-brain barrier (BBB) permeability has also been extensively reported, 27,28 supporting the concept that altered BBB permeability allows gut microbiota-produced metabolites and neuropeptides (particularly short-chain fatty acids [SCFAs] and the inhibitory gammaaminobutyric acid [GABA]) to reach the CNS to exert an effect on brain functions, and so on seizures' threshold in an epileptic context ( Figure 1). Other important elements of the MGB axis include the hypothalamic-pituitary-adrenal axis and the endocannabinoid system whose respective involvement in stress-response and neuromodulation may act on CNS functioning. 29,30 Therefore, the MGB axis is a multi-pathways network, still lacking full-blown comprehension. Proper pre-and clinical studies, primarily aiming to impact these complex pathways, may lead to the development of innovative therapeutic approaches.

| Preclinical data
Only a few preclinical studies have investigated the impact of the MGB axis on epilepsy. The attention has been mainly focused on dietary modification, microbiome supplementations, or manipulation of gut microbiota from mouse models, to evaluate pathological outcomes such as a reduction in seizures' frequency. 16 On the other hand, the number of studies demonstrating a link between gut microbiota alterations and increased neuronal excitability and seizures is constantly increasing, shedding light on the potential mechanisms, pharmacological targets, and treatments. [31][32][33][34] It is worth noting that literature studies can be divided into those demonstrating a direct or an indirect impact of microbiota manipulation on epilepsy features. Among those included in the first group, some studies reported an increased susceptibility to seizures that could be transferred by fecal microbiota transplantation as well as this latter can confer seizure protection and even be used to identify molecules (microbiota metabolites) as potential treatments. 14,34,35 Another example, that takes advantage of dietary treatments, is the ketogenic diet (KD), high fat and low-carbohydrate diet, inducing ketonemia and mimicking the positive effects of fasting on CNS functioning. The KD has long been used as a non-pharmacological treatment, mainly in the pediatric population affected by drug-resistant epilepsies. 36 Currently, available animal models are necessary and irreplaceable to further understand the mechanisms underlying epileptogenesis in the different forms of epilepsy 63 and to develop therapies to prevent the epileptogenic process, treat comorbidities, and drug-resistance.
In an extensive review focused on the Zebrafish use to study the MGB axis, three key advantages have been highlighted: (1) the possibility to perform genetic manipulation with zebrafish model being an ideal system to apply recent genome-editing technology; (2) the suitability for live in vivo imaging of host-bacteria interactions to monitor the spatial and temporal activities of immune-signaling components by exploiting the optical transparency of the transgenic zebrafish embryo throughout its development; (3) the existence of well-established protocols for germ-free experiments. 64 Moreover, Zebrafish and human microbiomes are known to have similar abundances of functional pathways despite substantial disparities in taxonomic composition. 65 As regarding brain activity studies in vitro, innovative techniques are rapidly rising to substitute and integrate animal models, fostering innovative modeling of epilepsy and improving animal welfare. 66 Indeed, over the last decades, high-density microelectrode array The first single electrode and non-invasive long-term EEG recording have been established only in 2013. 75 Recently, multichannel and noninvasive EEG recordings have been developed using embryonic zebrafish.
Indeed, Cho and colleagues have tested, in a pentylenetetrazoleinduced model of zebrafish, a non-invasive long-term multichannel EEG recording without embedding zebrafish in agarose. This method allowed measurements from each hemisphere of telencephalon and midbrain non-invasively, overcoming technical limitations in telencephalon analysis. 76 Moreover, a system named "Zebrafish Analysis Platform" has been recently designed and tested for long-term non-invasive high throughput multichannel electrophysiological monitoring, examining freely swimming zebrafish larvae autonomously and simultaneously within the microfluidic chamber array. 77 Therefore, integrating the emerging technologies with different animal models of epilepsy could greatly facilitate electrophysiological monitoring, high-throughput drug screening, and the accessibility to brain research in vivo with also a potential benefit for animal welfare.
Of notice, most of these in vivo models are inclusive of an informative gut microbiota too, thus offering from this point-of-view a suitable tool to study also in advanced technological systems the MGB axis' implications in epilepsy.
Nevertheless, investigating a complex relationship as the MGB axis role in epilepsy is still particularly challenging in in vivo systems, where several body districts can contribute to pathological outcomes.
Indeed, in vitro models with the help of innovative technologies could facilitate the clarification of molecular and mechanistic aspects at the basis of such an intricate interplay.

| In vitro MGB axis and epilepsy modeling: Challenges for a technological integration
Several works studying epileptic activity modulation by external agents (e.g., molecules, toxins, antiepileptic drugs) couple in vivo and in vitro methods. [78][79][80][81][82] In particular, exploiting these latter for a deeper understanding of what is observed in vivo. The preparation of brain tissues, in vitro, and their manipulation for multiple biochemical and electrochemical analysis is more accessible to study in detail the neural networks and provide the possibility to perform mechanistic studies on the brain's molecular and cellular mechanisms which is not allowed by classic in vivo models due to their intrinsic complexity. 83 Especially for new antiseizure medications, in vitro models are considered important to link the pharmacokinetics (PK) properties, well predictable using in vivo methods, with the detection of the compounds anticonvulsant properties. 84 Moreover, to perform neural recordings to get detailed information on brain function and synaptic plasticity, recent in vitro systems offer higher spatial resolution and higher signal-to-noise ratio if compared to in vivo recordings. 85,86 Organotypic hippocampal slice cultures and induced pluripotent stem cell (iPSC)-based brain organoids are just a few examples of in vitro models that are spreading in epilepsy-related studies. 87 Research on the relationship between the brain-gut axis and epilepsy is still at the preliminary stage. Recent studies on murine models have shown a close relation between gut microbiota and the occurrence of multiple types of epilepsy. 88 However, most of the underlying mechanisms are still unknown. Progress is being made in developing strategies to investigate the bidirectional communication between the gut and the brain in general with animal models, like germ-free mice, being essential for deepening our understanding of how microbiota alterations shape brain pathophysiology. However, these models often fail to recapitulate human scenarios due to differences in microbiota profile, molecular mechanism, immune system, and brain function like extensively reviewed by Moysidou and Owens. 89 Therefore, the need for human reliable models led to the employment of conventional in vitro tools and the advent of more complex technological strategies for their improvement.
Recreating physiologically relevant cell culture environments and incorporating the main characteristic features of the pathophysiological in vivo conditions will help to elucidate mechanistic details of the MGB axis. To this respect, in vitro modeling with dynamic cell culture conditions has been considered the most promising approach to recapitulate the MGB complexity, also for integration with currently available animal models which are still essential in neuro-diseases studies, especially concerning epilepsy. 90 Recently, multi-organ platforms for MGB axis modeling were proposed 91,106 while is still missing their integration with brain-like compartments for epileptic seizure mimicking. This latter in particular holds many challenges and different technological strategies were explored. In the following section of this review, the technological approaches adopted by researchers for epilepsy studying and in vitro modeling are listed. A special focus will rely on the available systems and possible innovative technological strategies for an advanced brain in vitro model to be integrated within multi-organ platforms, towards the application for epilepsy research and drug discovery also in the field of the MGB axis and epilepsy connection ( Figure 2).

| Epilepsy organotypic cultures
The organotypic slice cultures of rodent brains have been used to study different aspects of neuroscience for several years. 107 [108][109][110] In the epilepsy field, organotypic cultures of rodent brain (i.e., acute hippocampal slice) have a main role to evaluate pilocarpine treatment, changes in ion concentrations, or potassium channel blockers. 111,112 These cultures approximately replicate the threedimensional architecture and local structure of brain cells, such as neurons, astrocytes, and microglia, as well as the neuronal connectivity and the glial-neuronal interactions in a brain area relevant for the disease.
F I G U R E 2 Body-on-a-chip platforms are the ultimate advanced tools with the potential to give alternative systems to replace animal models in drug development. They offer pathophysiological recapitulation of the entire human body in a single device for drug pharmacokinetic (PK) and pharmacodynamics (PD) analyses in advanced interorgan systems. Microbiota, gut, and liver organ-on-a-chip systems were recently developed with the demonstration of reliable inflammation scenario.  Figure 3. 45 Cerebral organoids constitute the most used 3D "top-down" in vitro system for brain disease modeling. 137  The absence of a vascular system in the organoid leads to the progressive necrosis of internal regions, mostly because of poor oxygen penetration, besides the fact that late development is highly dependent on blood vessels proximity, an obstacle that is stimulating a lot of interest to reach more reliable brain representation. 150,151 Another important organoid's limitation that deserves a mention is the inability to model later embryonic and fetal development, thus not making it possible to obtain full mature synapses and study circuits and connectivity, an important feature of all neurodevelopmental disorders including epilepsy-associated ones. Together with this issue, the lack of body axes makes it difficult for the organoid to display correct migration and organization in specialized brain regions. 151 Similar to iPSC-derived neural models, also organoids may be sim- trode level. 157 The shape and structure of the microelectrodes have been proposed in many forms according to the specific application or, for improvement of the recording performances. One example is represented by protruding 3D electrodes, for better slice tissue penetration near the living neurons, which demonstrated that the amplitude of evoked potential responses was significantly larger than those obtained with planar MEAs. 158 Moreover, the array topography can be easily customized as well. Epileptic seizures were found to spontaneously arise from a specific area on hippocampal slices and many models recently tried to predict the complex spatiotemporal network propagation. [159][160][161] To follow such fine and complex signal distribution, some groups developed custom-designed electrodes arrays with tissue-conformal configurations, creating epilepsy-specific highdensity MEAs that conform to the cytoarchitecture of the nervous tissue of interest. 162,163 Electrographic screening, using the described approaches allowed us to achieve interesting findings in the identification of antiseizure medications, giving more information about the effects of different drugs on epileptogenesis 152 as well as about epileptic seizures dynamics both in vivo and in vitro. 164,165 In the last decade, big efforts have been made in developing MEA devices for recording the spiking activity of live neurons cultured in vitro, with the final aim of more reproducible live neuronal networks. 157,166,167 Indeed the non-invasive nature of MEA recordings already showed significant improvements, over traditional long-term cultures of brain slices, for synaptic connectivity in cultured hippocampal networks. 168 Moreover, this approach is particularly useful when investigating the anti-seizure mechanisms of drugs that may not reach the inner region of the prepared slice efficiently; indeed it is possible to study their effect directly on cultured neurons by measuring seizures activity. 157 To enhance the development of specific patterns that lead to the formation of neural networks at the culture surface, innovative works proposed the use of micro-printing and soft lithography fabrication techniques with cell patterning approaches.
Most recently, specific geometric substrates were coupled to fully planar MEAs as a new technological solution for a better translation of MEA technology in advanced culture systems and also in vivo applications. 169,170 One of the major issues in MEA based epilepsy in vitro studies is the long-term maintenance of the brain sample survival within the recording. Maintaining the brain slice oxygenated and hydrated for long periods, and avoiding in this way hypoxia and necrosis, is challenging in traditional MEAs. In the next section, we will explore the impact of dynamic culture conditions on brain and epilepsy in vitro studies with a focus on the development of dynamic perfused chambers for long period neural activity recordings. Moreover, the dynamic solution with OOC devices may be also the key to build up a bi-modal platform, where one chamber hosts already available dynamic models of the gut microbiota or capable to culture a selected bacterial strain, 171,172  Dynamic culture systems not only aim to improve the diffusion of oxygen and nutrients to the cells but also to reach high-throughput outcomes after a long period (Table 1). At the same time, maintaining the sample vital for long periods represents a challenge as well as integrating the culturing environment with micro-fabricated structures like MEAs.
In the context of more complex fluidic systems like the already mentioned multiorgan platforms, in particular for modeling the microbiota-immune-CNS interaction, having a highly controlled dynamic in vitro system is fundamental to model the metabolites and neurotoxins transport between different compartments.
From an engineering point of view, modulating the velocity profiles and the shear stress acting on the barrier is as crucial as challenging to achieve. Moreover, as concerning epilepsy and body fluid flows, it was observed dynamic ictal perfusion changes during temporal lobe epilepsy. 180 Tuning the flowrates within the neurons culture microenvironment with physiologically relevant velocity levels would add a further level of complexity to epilepsy in vitro models. In such a scenario, brain-on-chip (BoC) devices represent an innovative approach for integrating biology and bioengineering. They consist of microfabricated platforms that reproduce the CNS physiological microenvironment and tissue mechanical properties and responses to stimuli, allowing the development of several devices for specific neurological disease modeling, but also the study of brain networks in their complexity.
The ideal BoC device could potentially include several innovative technologies, gaining all the advantages of dynamic perfusion, 3D culturing, and electrophysiological recordings.
Moreover, the "on-chip" approach offers flexibility, robustness, and the high-throughput monitoring and stimulation of neural cells. 181,182 Importantly, specific requirements in terms of cell types and biological outcomes to be assessed in the BoC, influence enormously both the design and fabrication process of the device.
There are some fundamental elements for BoCs design and manufacturing: microchannels, that can be used as neurite growth guide, as scaffolds, or basically to provide perfusion and biochemical gradients; microchambers, to allow spatial separation between different cell types or heterogeneous tissues formation; ECM components for three-dimensionality representation; electroactive components such as MEAs for stimulation and recording. All those elements are connected to specific functional features as dynamic mechanical stress, mass transport of solutes, analysis of neural electrical activity, and distribution of biochemical cues. [183][184][185][186][187][188] As for biological models to be included in the BoC, the device can novel treatment strategies. 190,191 Different BoC devices have been developed in the last decade, having one or more of the aforementioned features and various disease study applications (e.g., Alzheimer's disease, Parkinson's disease, traumatic brain injury). 175,[192][193][194] However, specific development of  OOC technology is an emerging field in bioengineering aiming to fill the gaps in drug screening by recreating human tissue models for more reliable predictions of drugs efficacy and safety in support of clinical trials. 96,195 The main characteristic feature of a specific organ is recreated within a sophisticated complex microenvironment using advanced technological approaches, some of which we recapitulated in this review. Up to date, several organs scenarios have been modeled in innovative OOCs: the microbiota-gut interface, the BBB, or the lung microenvironment are just a few examples. 196,197 In this review, we already went through OOCs applied to Brain physiology or disease modeling which are called BoCs and their main technological features.
As concerning epilepsy, in vitro modeling is recently approaching advanced technological routes (Table 2) T A B L E 1 List of currently used systems for flow perfused long-term cultures of brain organoids and brain slices applicable to epilepsy research represents the first case of microchannels incorporated into the well for brain slice maintenance. The micrometric dimension of the channels and the controlled flow velocity led to axonal growth and alignment along the channel direction. 199 This latter phenomenon was also reported by Shen and co-authors who proposed a neuro fluidic microdevice showing the influence of microfluidic constrain on the functional neural connectivity, which was measured by an integrated MEA inside the microchamber. 184 Liu and colleagues recently proposed μflow-MEA, the first epilepsy-on-a-chip system for ASMs discovery. 179  The microfluidic chip contained micro-wells in which the formation of homogeneous neurospheroids occurred. An osmotic micropump system connected to the outlet provided a continuous flow of medium that contained oxygen and nutrients. This microfluidic device retains two in vivo brain characteristics, the 3D cytoarchitecture and the physiological interstitial flow. Moreover, Wang and colleagues developed a simple and robust micro-device that allows generating hiPSCs-derived brain organoids in a controlled manner. 176 Through their platform, they examined the features of neural differentiation, brain regionalization, and the cortical organization in the brain organoids.
Soscia et al. 189 proposed a platform to reproduce complex neuronal cultures and record brain cells excitability The MEA integration inside these microsystems is perhaps one of the most challenging aspects. One interesting solution was proposed by Sharf et al. 206 They reported a system for monitoring the electrical activity generated by multi-cellular networks in a non-contact configuration. In this way, cells can be grown on conventional cell culture substrates and the recording electrodes array can probe different cultures in succession, without degrading its sensitive electronic surface.
Moreover, this configuration is particularly suitable for micro-channels thanks to the micrometric distance between the cells and the MEA.
Hence, considering the available epilepsy models and the technologies described above, it is evident that the integration of the biological aspects of brain cells culturing with the technologies that bioengineering can provide, including an OOC solution to model the microbiotagut compartment, constitutes the new frontier for reliable epilepsy disease modeling, contributing to the clinical translation also in a multi-organ approach.

| CONCLUSION: PERSPECTIVES AND CHALLENGES FOR INNOVATIVE EPILEPSY RESEARCH WHEN TAKING INTO ACCOUNT THE MGB AXIS PARADIGM
The microbiome plays a significant role in the health status of its host.
Since the gut microbiota was first proposed to influence human health over one century ago, our understanding of its role has immensely 3D culture inside a fluidic platform with other different organ models that may be 2D or 3D barrier models to recapitulate the systemic inflammatory process; (iii) introducing access points to the platform for sample manipulation; or (iv) the implementation of complex electrodes arrays inside epilepsy/brain compartment by fast production techniques.
In conclusion, besides many challenges to face, the promising outcomes from the newborn epilepsy-on-chip technology are supportive of the feasibility of new strategies for reliable epilepsy in vitro models to be integrated inside complex dynamic multi organs platforms, with a mid-term impact also on the molecular mechanism, drug and biomarker discovery and ultimately clinical translation.

DATA AVAILABILITY STATEMENT
Data sharing is not applicable to this article as no datasets were generated or analyzed during the current study.