Innovative engineering approaches to model host-microbiome interactions in vitro

Abstract

The human microbiome plays a critical role in health and disease. Disruptions in microbiota composition or function have been implicated not only as markers but also as drivers of diverse pathologies, creating opportunities for targeted microbiome interventions. Advancing these therapies requires experimental models that can unravel the complex, bidirectional interactions between human tissue and microbial communities. This scoping review examines emerging engineering approaches to design in vitro platforms that successfully integrate host and microbial components to model these interactions. Compared to traditional in vitro and in vivo approaches, these advanced microphysiological systems offer greater experimental control, human-specific biology, and reduced cost and ethical concerns. Here, we identify key challenges in the creation of these in vitro models and innovative solutions to address them by leveraging microfluidics, biomaterials, and organoid technologies, among others. These strategies have enabled the development of co-culture systems that replicate critical features of host-microbiome interfaces, including mucosal barriers, oxygen and pH gradients, mechanical stimuli, and host cell diversity. We also describe how these physiologically relevant models are uncovering new insights into epithelial-microbiota crosstalk, immune modulation by commensal microbes, and systemic effects of microbiota and their metabolites across multiple body sites. We conclude by discussing opportunities to expand these systems in scale, complexity, and clinical relevance. As these models continue to evolve, they hold the potential to transform our ability to mechanistically probe microbiome interactions, personalize therapeutic strategies, and accelerate the translation of microbiome science into clinical practice.

Graphical Abstract

1. Introduction

Over the last twenty years, knowledge about the human microbiome has increased at an exponential rate, revealing its critical role in host health. Dysbiosis or dysregulation of the microbiome has been linked to a wide assortment of local and systemic conditions, including inflammatory bowel disease, bacterial vaginosis, acne, cardiovascular disease, metabolic syndrome, and autoimmune disorders [1-7]. Emerging evidence suggests that changes in microbial composition and function are not only a consequence of disease but may also play a causal role [8-11]. This growing recognition has spurred interest in microbiome-based therapies aiming to restore or manipulate microbial communities to improve health outcomes. The successful treatment of recurring Clostridioides difficile infections with fecal microbiota transplantation exemplifies the potential of such interventions to treat disease by reshaping the microbiome [12-14]. However, the development of targeted microbiome interventions requires a thorough understanding of the complex mechanisms that regulate human-microbiome interactions.

Existing clinical, animal, and in vitro models have significantly advanced our understanding of the human microbiome; however, these approaches are sometimes insufficient to fully characterize host-microbe interactions [8]. While clinical studies provide valuable insights into microbiome composition and disease associations, it is often difficult to determine cause and effect due to the complexity of confounding variables and the rightful ethical constraints of human experimentation. To address this challenge, current microbiome research relies heavily on animal models, which are time-consuming and costly, and at times fail to capture human biology [15-17]. Traditional in vitro models usually study microorganisms in isolation or in systems (e.g. cell culture on tissue culture plastic) that do not incorporate the properties of host tissues. As a result, the simplified nature of these models fails to reproduce the complex interactions between microbial communities and host cells observed in vivo. The inability to move from correlation to causation is a major roadblock in the development of microbiome-based therapeutics and diagnostics. To address these limitations, researchers are increasingly turning to advanced models that more accurately replicate human physiology and allow precise manipulation.

Engineered microphysiological systems have emerged as promising tools to model disease and uncover the underlying biology driving tissue homeostasis. These systems leverage biomaterials, microfluidics, organoids, and 3D printing technology to generate advanced in vitro models that replicate native tissue structure and function [18-23]. Engineered tissue mimics support the growth of multiple cell types in both 2D and 3D, and the simultaneous introduction of key mechanical and biochemical cues. Compared to animal models, microphysiological systems are highly controllable, time- and cost-effective [24,25], and accelerate the process of testing hypotheses while bridging the gap between conventional models and human biology. Furthermore, the recent FDA Modernization Act 2.0 allows “for alternatives to animal testing for purposes of drug and biological product applications” [26,27], thereby emphasizing the growing importance of these models for both basic science and translational applications.

In this scoping review of the literature, we explore recent advances in the design of in vitro systems to model human-microbiome interactions. First, we describe the major body sites colonized by microbiota and provide a broad overview of the principal modes of interaction between microbial communities and human hosts. Next, we define key design challenges specific to human-microbiome interfaces, identify innovative engineering strategies to overcome them, and highlight how these models are being used to uncover fundamental principles of human-microbiome interactions and inform therapeutic development. Finally, we conclude with perspectives on important considerations and future directions necessary to ensure that the next generation of in vitro models truly captures the complexity of the human microbiome.

2. Major microbiome sites in the human body

Microbial communities colonize numerous anatomical sites throughout the human body, forming complex and dynamic ecosystems that play critical roles in host health and disease [1-4,28,29]. These microbial populations vary significantly across body regions due to unique adaptations driven local environmental and physiological conditions that result in specialized ecological niches. Thus, the design of physiologically relevant in vitro models must consider the complex spatial organization, biochemical gradients, and site-specific host-microbiome interactions observed in vivo. Below, we highlight the key structural features, microbiome composition, and niche-specific characteristics of the body sites harboring particularly abundant and diverse microbial communities.

2.1. Skin

The skin, comprised of the epidermis, dermis, and hypodermis, serves as the first line of defense against the external environment [30]. The outermost layer, the epidermis, is avascular and contains layers of differentiated keratinocytes, pigment-producing cells and antigen-presenting immune cells [31]. Beneath it, the highly vascularized dermis contains nerve endings and fibroblasts responsible for extracellular matrix (ECM) remodeling [32]. Deepest of all, the hypodermis includes clusters of adipocytes partitioned by fibrous septa that operate as an endocrine organ and energy reservoir [30].

The skin microbiome includes both resident and transient microorganisms that are generally nonpathogenic and stable over time despite external exposures [33,34]. Although most microbiota research focuses on bacteria, other organisms also inhabit the skin. For instance, Demodex, a parasitic arthropod, is also present on normal skin [35] and Malassezia, a polymorphic yeast, constitutes 80% of cutaneous fungi [36]. Viruses are the least understood members of the skin microbiome.

Four main bacterial phyla are found on human skin: Actinobacteria, Firmicutes, Proteobacteria, and Bacteroides, with Corynebacteria, Propionibacteria, and Staphylococci being the most common genera [37]. Due to the limited nutrients found on human skin [38], resident microbiota utilize resources found in skin gland secretions, like sweat, to survive the harsh and acidic conditions of their microenvironment [31]. For example, Propionibacterium acnes (P. acnes) thrives in the sebaceous gland by secreting lipases to break down triglycerides in sebum, releasing free fatty acids that enhance bacterial adherence [39].

Skin bacteria distribution is influenced by environmental conditions such as moisture, temperature, sebum production, humidity, and pH that create distinct ecological niches. These niches are categorized as sebaceous (oily), moist, and dry [31]. Sebaceous skin sites, like the face and forehead, are rich in sebaceous glands, favoring lipophilic organisms like Propionibacteria [40]. The moist gluteal region supports aerobic organisms like Proteobacteria and Staphylococcus aureus, while the forearm, a dry area, hosts a greater variety of microorganisms, including β-Proteobacteria, Flavobacteriales, and Corynebacterium [37,41]. Overall, the bacterial composition at sebaceous sites is less diverse, even, and rich compared to moist and dry sites [37].

2.2. Oral Cavity

The oral cavity harbors the second most diverse microbiota due to constant exposure to the external environment and exogenous stimuli [42]. The oral microbiome includes over 600 bacterial species from six major phyla: Firmicutes, Bacteroidetes, Proteobacteria, Actinomycetota (formerly Actinobacteria), Spirochaetes, and Fusobacteria [43]. At the genus level, the most abundant bacteria are generally conserved across healthy individuals [44,45]. Other microorganisms include archaea, protozoa like Entamoeba gingivalis and Trichotomas tenax, viruses, and several fungi [46]. However, an increase in the abundance of several of these microorganisms is commonly associated with disease [1,46,47].

Differences in oxygen levels, pH gradients, and nutrient availability, create distinct oral microenvironments colonized by specialized microbial subtypes. Anaerobic bacteria like Fusobacterium nucleatum prefer the oxygen-deprived environment of the gingival sulcus, the pocket between teeth and gums [48]. Even within the same genus, different species adapt to occupy distinct oral habitats. For example, within the Streptococcus genus, S. mutans thrives on hard tooth surfaces, S. sanguinis in the supragingival plaque, and S. salivarius on the tongue dorsum [1,49]. Mastication and salivary flow may also modulate the oral microbiome at these niches, though their role is not well understood.

The microbiome plays a central role in oral health, with microbial dysfunction driving prevalent diseases. Dental caries is associated with an increased abundance of biofilm-forming and acid-producing bacteria [50,51]. In the subgingival space, microbiota dysbiosis triggers inflammation that can develop into gingivitis and periodontitis [52]. Oral microorganisms have also been linked to systemic diseases such as colorectal cancer, Alzheimer disease, aspiration pneumonia, and rheumathoid arthritis [1,47]. The translocation of periodontal pathogens to extra-oral sites is hypothesized to be responsible for these disease-promoting effects [53-56]; however, further research is needed to clarify their specific contributions.

2.3. Gastrointestinal Tract

The human gastrointestinal (GI) tract consists of interconnected hollow organs that transport food for digestion, absorption, and excretion. Throughout the length of the GI tract, four overlapping layers of tissue form the gut wall: mucosa, submucosa, muscularis propria, and serosa [57]. The innermost mucosa layer encloses the lumen. Depending on the location, the mucosa can have specialized topographical features [58]. For example, tall intestinal villi are found along the small intestine to increase surface area for nutrient absorption. Meanwhile, the colon’s crypts protect proliferating intestinal stem cells for ongoing tissue renewal [58]. Goblet cells, Paneth cells, enterocytes, and enteroendocrine cells make up the mucosal epithelium and are responsible for mucus production, immunity, nutrient absorption, and luminal chemosensing. Surrounding the mucosa, the submucosa, muscularis propria, and serosa provide structural stability and GI motility, with smooth muscle facilitating peristaltic contractions [59,60]. Mechanical forces, tissue morphology, and mucosal secretions shape the luminal environment and influence the abundance and composition of the microbiome [58].

The gut lumen harbors the highest density of microbes in the human body with high spatial heterogeneity along the GI tract [61,62]. Microbial concentrations gradually increase from the stomach to the colon [63]. Reduced concentrations of microorganisms in the upper GI tract are attributed to the harsh acidic environment, rapid transit of luminal contents, high antimicrobial peptide concentrations, and high oxygen levels [64]. Thus, acid tolerant facultative anaerobes are prominent in the small intestine, in particular species belonging to the Lactobacillaceae and Enterobacteriaceae families [65,66]. In contrast, the colon, with its nutrient- rich environment and long transit times, sustains the highest microbial load consisting of diverse carbohydrate-degrading and mucus-adherent microbes [67]. There, mucolytic species, such as Akkermansia muciniphila and Bacteroides spp, can also forage host-derived glycans and colonize epithelial surfaces [65,68].

The human gut microbiota mainly consists of bacteria from the Firmicutes and Bacteroidetes phyla, with smaller proportions of Proteobacteria, Actinobacteria, and other phyla present [3]. The composition of the human gut microbiome is susceptible to both short-lived and permanent fluctuations influenced by both diet and age. Diets that are high in animal proteins and fats favor bile-tolerant microbes, such as Bacteroides species. Meanwhile, plant-based diets rich in fiber and carbohydrates are correlated to higher abundances of Prevotella [69]. Age also impacts microbiota diversity and composition [70]. Adult intestinal microbiota is dominated by anaerobic Bacteroidetes and Firmicutes phyla [71]. Meanwhile, the infant microbial diversity is highly dependent on early life colonization, such as type of birth delivery and their development stage. For example, Bifidobacterium longum outcompetes other microbes while an infant is breastfed exclusively, but as solids are introduced, their microbiota mimics an adult’s microbial diversity [72]. Other factors such as antibiotic usage, geographic location, diet, and lifestyle also modulate gut microbiome composition and function [4,73].

2.4. Respiratory Tract

The respiratory tract is continuously exposed to diverse microbial communities introduced by inhaled air that colonize distinct niches throughout the airway [74]. These microorganisms support respiratory health by stimulating the immune system and providing protection against pathogen colonization and invasion. Divided into the upper and lower respiratory tracts, the microbiota found in these regions exhibit distinct microbial compositions and functions [75].

The upper respiratory tract, the gatekeeper to respiratory health, and is comprised of distinct anatomical structures with specialized epithelial cell types and diverse microbial niches [76]. The upper respiratory tract is richer in microbes than the lower respiratory tract, with microbial density decreasing along the airway [77]. Closest to the external environment, the anterior nares are lined with keratinized squamous epithelium containing sebaceous glands that are enriched in lipophilic skin colonizers like Propionibacterium spp. [78,79]. The nasopharynx, covered by stratified squamous epithelium and epithelial cells, harbors diverse microbiota, including Moraxella, Staphylococci, Corynebacteria, and niche-specific genera like Haemophilus and Dolosigranulum [80,81]. The oropharynx, lined with non-keratinized stratified squamous epithelium, supports even more diverse bacterial communities, including Streptococci, Neisseria spp., and Rothia spp., and anaerobes such as Veillonella spp. and Prevotella spp. [82,83]. The upper respiratory tract also houses common viruses like rhinovirus and adenovirus [84,85] and fungi including Aspergillus, Penicillium, Candida, and Alternaria. [86,87].

The lower respiratory tract, including the trachea, bronchi, bronchioles, and alveoli, was once considered sterile. However, next-generation sequencing has revealed diverse microbial species, though contamination in low-density specimens requires careful interpretation [88]. The lung microbiota primarily originates from the upper respiratory tract via mucosal dispersion and micro-aspiration [89], with direct inhalation of ambient air may also contributing [90]. As a result, several bacteria present in the upper respiratory tract are also found in the lung. These include bacteria of the Moraxella, Haemophilus, Staphylococcus, and Streptococcus genera. Other species like Tropheryma whipplei are unique to the lower respiratory tract [91,92]. The lower respiratory tract virome includes Anelloviridae and bacteriophages, while the mycobiome features genera like Eremothecium, Systenostrema, and Malassezia [84,93,94]. Despite regional variations in oxygen, pH, and temperature, the lung microbiota shows minimal spatial diversity, suggesting that its composition is influenced by microbial immigration and elimination rather than being a stable, resident community [95].

2.5. Vagina

The vagina is a muscular organ of the female reproductive system that serves as a passage from the uterus and cervix to the outside of the body [96]. The walls of the vagina are made up of three layers: an outer adventitia layer made of connective tissue, a middle muscularis layer composed of smooth muscle, and an outer mucosa layer made of epithelial cells [96]. The inner layer of the vaginal wall is made up of stratified vaginal epithelial cells coated in a layer of mucus called cervicovaginal fluid, which is composed of water, proteins, lipids, and glycoproteins known as mucins [97-101]. The vagina is naturally a hypoxic environment because of lower oxygen concentrations in its blood supply, which allows both aerobic and anaerobic bacteria to coexist [101].

The vaginal microbiome includes many species of bacteria, fungi, and viruses but is often dominated by Lactobacillus species in reproductive-age women [97,101]. Lactobacilli are facultative anaerobes that produce lactic acid [100,102]. The production of lactic acid enhances the integrity of the epithelial barrier, elicits an anti-inflammatory response, and lowers the vaginal pH (~4.0), protecting against harmful pathogens [103,104]. The vaginal microbiome can be described by Community State Types (CSTs) which are defined by the microbes that dominate the microbiome [97,101,105]. Four of the five CSTs are dominated by a single species of Lactobacilli: CST I – Lactobacillus crispatus, CST II – Lactobacillus gasseri, CST III – Lactobacillus iners, CST V – Lactobacillus jensenii [97,101,105]. CST IV is characterized by an abundance of obligate anaerobic bacteria commonly including phylotypes Gardnerella, Prevotella, and Atopobium [97,101,105].

Shifts in the composition of the vaginal microbiome can occur due to changes in sexual activity, hormones, and stress. Sexual activity can introduce pathogenic bacteria into the vaginal microbiome and increase the vaginal pH to create an environment favorable to harmful bacteria [106]. Hormone changes caused by pregnancy, menstruation, contraception, and menopause can also impact host physiology and the vaginal microbiome [107]. High levels of estrogen are associated with increases in glycogen production and thriving Lactobacilli populations [107]. Stress triggers the production of cortisol which suppresses estrogen and can lead to a decrease in the presence of Lactobacilli in the microbiome [107,108]. Microbiomes not dominated by Lactobacilli species are associated with an increased risk of adverse health outcomes including STI acquisition, pregnancy complications, and pelvic inflammatory disease [97,100,101].

3. Human-microbiome interactions

The microbiome influences human health by actively interacting with host tissues at multiple biological levels beyond passive colonization. Microbial communities modulate epithelial barrier integrity, influence local and systemic host immune responses, produce bioactive metabolites that can reach systemic circulation, and actively reshape the local tissue microenvironment (Figure 1). These interactions are bidirectional, with host cells both responding to and influencing microbiome composition and function. Developing physiologically relevant in vitro models capable of capturing this level of complexity is essential to uncover the mechanisms driving host-microbiome interactions, which will facilitate the design of microbiome-based diagnostics and therapeutic interventions.

Figure 1. Overview of key interaction modes between complex microbiota communities and the human host.

The bidirectional influence of host-microbiome interactions includes direct colonization of epithelial surfaces, modulation of the immune system, production of metabolites, and modifications to the local microenvironment. IgA: immunoglobulin A; LPS: lipopolysaccharide; TMAO: trimethylamine N-oxide; SCFAs: short chain fatty acids; ECM: extracellular matrix. Created in BioRender.

3.1. Microbial interactions at the epithelium

Throughout the human body, the epithelium serves as a protective barrier and primary interface between host and microbiome. Epithelial cells form a selectively permeable barrier that regulates nutrient absorption and prevents the movement of microorganisms and other harmful substances into the underlying tissues [109]. At the same time, microbial communities at this critical interface also shape tissue homeostasis by regulating epithelial barrier integrity, biofilm formation, and colonization by pathogenic microorganisms [110].

Commensal gut bacteria can reinforce the stability of the epithelial barrier. For example, commensal Bifidobacteria, Akkermansia, and Lactobacilli upregulate the expression of tight junction proteins and stimulate mucus secretion in mouse and human hosts [111-114]. Conversely, GI pathobionts (i.e., symbiotic microorganisms that can cause harm under specific circumstances [115]) can release toxins and mucinases that degrade tight junctions and mucin glycoproteins, leading to increased permeability, gastric ulcers, and systemic inflammation [116,117]. Candida albicans, a fungal pathobiont, can also disrupt epithelial integrity when it transitions into its invasive hyphal form capable of breaking down tight junction proteins [118].

Microorganisms at the epithelial interface can also form biofilms, structured microbial communities embedded in a complex polymeric matrix that adhere to host surfaces and are often associated with disease [119]. While some biofilms, like those produced by Staphyloccus mutants and Gardnerella vaginalis, contribute to the pathogenesis of diseases such as dental caries and bacterial vaginosis [1,120], others can be beneficial. For example, in the nasal cavity, colonization by Staphylococcus epidermis prior to adulthood leads to increased antimicrobial peptide production. Because S. epidermis forms robust biofilms, it can resist killing; thereby excluding pathogens and contributing to the successful maturation of the nasal microbiome [121]. Nonetheless, the role of biofilms in healthy microbiomes remains understudied.

Beneficial microbes play a critical role at the epithelial interface resisting pathogen colonization by outcompeting them for adhesion sites and nutrients. In the gut, commensal bacteria inhibit Clostridium difficile infection by producing metabolites that suppress its growth [122]. Similarly, certain oral bacteria species produce proteins that antagonize periodontal pathogens and inhibit biofilm formation [123,124]. Thus, dysbiosis or disruptions in the microbiome – whether by antibiotics, changes in diet, or disease – can lead to increased epithelial permeability and reduced pathogen resistance with significant consequences for the human host.

3.2. Production of metabolites

Microbiome-derived metabolites, intermediates or end products of microbial metabolism, are generated when microbes feed on dietary components or process host-secreted substrates, producing a diverse array of low molecular weight molecules [125]. These metabolites are integral to various physiological processes, including energy metabolism, cellular signaling, immune modulation, and biomolecule synthesis [126,127]. Microbiota-derived metabolites can be classified into three broad categories: (1) metabolites produced by the microbiome from dietary components, (2) those generated by the host and modified by microorganisms, and (3) metabolites produced exclusively by gut microbes [128].

The metabolites generated by microbial processing of host diets include short-chain fatty acids (SCFAs), trimethylamine-N-oxide (TMAO), and tryptophan metabolites [129]. SCFAs—such as acetate, propionate, and butyrate—are produced through the fermentation of dietary fiber [130]. Most SCFAs, and specially butyrate, are used locally by colonocytes as an energy source and promote epithelial barrier integrity in the colon [131]. Any unused SCFAs then enter hepatic and, eventually, systemic circulation where they regulate cholesterol uptake, glucose homeostasis, insulin resistance, and immune function [132]. In fact, dysregulated SCFA production has been linked to colorectal cancer, cardiovascular diseases, and metabolic disorders [133]. TMAO is produced when gut bacteria convert dietary choline and L-carnitine into trimethylamine, which is then oxidized in the liver [134]. Elevated TMAO levels are associated with an increased risk of cardiovascular disease, metabolic syndrome, and neurodegenerative conditions [135]. Finally, indole is generated when gut microbiota metabolize tryptophan, with host enzymes further modifying these compounds to generate serotonin in the brain or kynurenine in the liver [136]. Indole and its derivatives play a pivotal role in immune homeostasis, gut barrier integrity, and central nervous system regulation [137].

Among those metabolites generated by the host and modified by microorganisms, bile acids are the most widely studied. Primary bile acids like cholic acid are synthesized in the liver from cholesterol and later transformed by gut microbes into secondary bile acids, such as deoxycholic acid and lithocholic acid [138]. These secondary bile acids act as signaling molecules that influence glucose and lipid homeostasis through FXR signaling, and immune responses through TGR5 signaling [139]. In addition, secondary bile acids exert antimicrobial effects that shape the composition and stability of gut microbiota [140]. Dysregulation of bile acid metabolism, such as altered bile acid pools or excessive accumulation of hydrophobic bile acids, has been linked to insulin resistance, non-alcoholic fatty liver disease, and colorectal cancere [141].

Finally, gut microbes can also produce metabolites de novo, without host intermediaries. These include microbially-generated gases such as methane, and hydrogen sulfide, which can influence intestinal motility and epithelial cell signaling [142-144]. Additionally, the gut microbiome serves as a major source of B vitamins, with a significant portion of gut bacteria possessing biosynthetic pathways for their production [145]. B vitamins produced by the gut microbiome, including B1, B2, B6, and B9, play essential roles in energy metabolism, red blood cell formation, and neurological development [146]. Because these small molecules can easily access the blood stream, microbiota-derived metabolites play a crucial signaling role in health and disease both locally and systemically.

3.3. Modulation of the immune system

Homeostasis between the microbiome and the immune system relies on balance between defense against harmful pathogens and tolerance to commensal microbes. Immune responses are shaped by complex interactions with the microbiome, involving both innate and adaptive immune cells. Commensal bacteria stimulate epithelial cells to produce antimicrobial peptides, which maintain barrier integrity and prevent invasion by pathogens [147]. Microbiota also display microbe-associated molecular patterns, such as lipopolysaccharides and peptidoglycans, that are recognized by toll-like receptors on innate immune cells to regulate inflammatory responses [148,149]. Additionally, the microbiome can modulate adaptive immune responses by influencing the development and function of B cells and T cells. Imbalances in the microbiome can increase pro-inflammatory signaling cascades that can contribute to autoimmune and inflammatory diseases [150-153]. Understanding how interactions between the immune system and microbiome maintain tolerance to commensal bacteria while protecting against harmful pathogens is essential for developing therapies against inflammatory diseases.

3.3.1. Mucosal immunity

Mucosa-associated lymphoid tissues (MALTs) is a dispersed system of lymphoid tissue located in mucosal surfaces throughout the body, including the respiratory tract, gastrointestinal tract, urogenital tract, and oral cavity [154]. MALT plays an important role in mucosal immunity by providing protection against pathogens at barrier surfaces while maintaining tolerance to commensal microbes. Among these, the gut-associated lymphoid tissue (GALT), is the largest and most extensively studied. GALTs continuously sample luminal content – including dietary antigens, commensal and potential pathogens – serving a the first line of immune defense in the gastrointestinal tract [155]. Key components of GALT include Peyer’s patches, isolated lymphoid follicles, mesenteric lymph nodes, and scattered immune cells throughout the lamina propia [156,157]. Peyer’s patches are small, organized masses of lymphatic tissue found primarily in the lining of the ileum. Specialized epithelial cells within Peyer’s patches, called microfold (M) cells, initiate mucosal immune responses by sampling and transporting antigens from the intestinal lumen to underlying antigen-presenting cells [158].

Cytokine signaling and immunoglobulin secretion are also central to immune regulation within GALT.s. A diverse array of immune cells work in concert with commensal microbes to modulate mucosal responses. For example, commensal microorganisms stimulate tissue-resident regulatory T cells to secrete anti-inflammatory cytokines, such as interleukin 10 and transforming growth factor-β [157,159]. In parallel, B cells in the lamina propria produce secretory immunoglobulin A, which binds commensal microbes to prevent epithelial invasion and limit inflammatory responses [160-162]. Together, the unique composition of MALTs allows these tissues to maintain mucosal homeostasis by orchestrating a delicate balance between microbial tolerance and immune defense.

3.3.2. Regulation of innate immunity by lypopolysaccharides

Lipopolysaccharides (LPS), also known as endotoxins, are primary components of the outer membrane of gram-negative bacteria. They consist of three domains: a lipid anchor, known as lipid A, a core oligosaccharide, and a variable O-antigen polysaccharide side chain that faces outward [163,164]. Because LPS are not produced by eukaryotes, the human innate immune system recognizes them as microbe-associated molecular patterns [165]. Thus, they play a crucial role in mediating immune responses, particularly in inflammation and sepsis. More specifically, lipid A binds to immune cell receptors like Toll-like receptor 4, initiating a signaling cascade through the NF-kB pathway that culminates in the production of pro-inflammatory cytokines [166,167].

In healthy individuals, the gut microbiome is the primary source of LPS. Among gut microbiota, proteobacteria, particularly Escherichia coli (E. coli), are significant contributors to LPS production [168]. Under normal conditions, LPS from the resident microbiome does not trigger inflammatory responses due to effective regulation of LPS activity [169]. Recent studies have shown that LPS from gut-resident microbes can antagonize the E. coli LPS-TLR4 signaling pathway, with Bacteroidetes species producing antagonistic LPS that drive immune silencing [168]. This balance between different LPS types is essential for maintaining intestinal immune homeostasis.

3.3.3. Adaptive Immune Modulation

The microbiome plays a critical role in shaping adaptive immune system by influencing B cell maturation, T cell differentiation, and antigen presentation by dendritic cells (DCs). In mucosal tissues, B cells interact with microbiome-specific antigens that stimulate their maturation and promote the production of secretory immunoglobulin A (sIgA) [149,160]. sIgA coats commensal bacteria, limiting their adhesion to the intestinal lining and preventing microbial overgrowth and invasion. T-cell dependent sIgA production creates a more targeted and diverse antibody repertoire, enhancing the host’s ability to maintain immune surveillance while tolerating non-pathogenic microbes [149,160].

In addition to shaping antibody responses, commensal microbes influence T cell behavior through both metabolite production and direct interactions with host T cells. Gut microbiota-derived SCFAs, such as butyrate and acetate, stimulate Treg differentiation by increasing IL-10 production, thereby promoting immune tolerance [170,171]. SCFAs have also been shown to boost the formation of memory CD8+ T cells, enhancing long-term immune protection [172]. Specific microbial taxa can also drive the differentiation of effector Th17 cells, which can elicit both inflammatory and non-inflammatory responses [149,173]. For example, Bacteroides fragilis promote non-inflammatory Th17 responses, whereas Eggerthella lenta promotes inflammatory Th17 response [174,175].

DCs are also heavily influenced by microbial signals. Tissue-resident DCs can extend dendrites through the epithelial barrier to sample luminal microbial antigens, which they then present to naïve CD4+ T cells [148]. This interaction guides the differentiation of CD4+ T cells and drives the production of cytokines such as IL-17 and IL-22 production, creating a pro-inflammatory response [149]. By influencing this crosstalk between the innate and adaptive immune systems, microbial communities help shape immunological memory and the magnitude of future adaptive immune responses.

3.3.4. Inflammatory Signaling

Microbiome dysbiosis, an imbalance in microbial composition, can provoke chronic inflammation by altering the balance between pro- and anti-inflammatory signals [176]. The loss of commensal microorganisms can result in decreased production of beneficial SCFAs, impairing regulatory T cell function, and the release of cytokines such as interleukin 6 (IL-6) and tumor necrosis factor-alpha (TNF-α) [150,151,177-181]. Concurrently, the overgrowth of pathogenic bacteria leads to elevated levels of microbial products that activate host pattern recognition receptors, particularly Toll-like receptors (TLRs), culminating in the release of additional cytokines [152,153,177-181].. While LPS are well-characterized activators of TLR4 signaling, other microbial ligands also trigger TLR pathways [182]. For example, flagellin from Salmonella typhimurium is recognized by TLR 5, activating chemokine and cytokine production via the MyD88/NFκB signaling pathway [183].

As dysbiosis progresses, the integrity of the epithelial barrier is also compromised, enabling microbial products and antigens to translocate into systemic circulation and activate macrophages and DCs, further driving the secretion of pro-inflammatory cytokines [184,185]. Certain microorganisms can also trigger the release of IL-17 by Th17 cells, leading to neutrophil recruitment that amplifies tissue inflammation [186-188]. Chronic exposure to these signals promotes systemic inflammation and can ultimately contribute to the development of autoimmune and inflammatory diseases, including inflammatory bowel disease, Crohn’s disease, rheumatoid arthritis, and asthma [150-153,178-181].

Although often associated with inflammation, TLR signaling can also induce anti-inflammatory immune responses. Lipoproteins, peptidoglycan, and lipoteichoic acid of gut microbiota are recognized by TLR2 receptors and support gut homeostasis by inducing IL-10 expression [189-191]. In the gut, the activation of TLR2/IL-10 signaling by polysaccharide A from commensal Bacteroides fragilis has been shown to suppress immune activation and reduce colitis in mouse models [192,193]. These findings highlight the context-dependent nature of host-microbial interaction, where both pro- and anti-inflammatory pathways are shaped by the composition and activity of the microbiota.

3.4. Modifications to the local microenvironment

Microorganisms actively modify their surroundings to establish colonization and promote their own survival. In doing so, they shape the local tissue microenvironment within the host. A key component of this microenvironment is the extracellular matrix (ECM), a complex network of proteins, glycosaminoglycans, and other molecules that provide mechanical support and biochemical signals to regulate tissue homeostasis [194]. Human-associated microbiota can secrete enzymes that actively remodel host ECM, altering its composition and mechanical properties [195-197]. For example, oral bacteria secrete trypsin-like peptidases that break down tooth enamel and Streptococcus species in the skin release a metalloprotease that cleaves human IgA, thus modulating the local immune response [198,199]. We have recently demonstrated that a wide variety of commensal gut bacteria can degrade multiple proteins and glycosaminoglycans found intestinal ECM [196]. Because the ECM plays a critical role in cell signaling and tissue integrity [200], microbial remodeling of the ECM can influence host cell behavior with significant consequences for host health.

Beyond ECM remodeling, microbial enzymes and metabolism also modulate physicochemical gradients within the microenvironment, particularly pH and oxygen levels. Fermentation by gut anaerobes produces SCFAs like butyrate and acetate that acidify the colonic environment [65]. Conversely, urease-producing bacteria, such as Helicobacter pylori in the stomach or Bifidobacterium infantis in the gut, hydrolyze urea into ammonia, neutralizing acidity to facilitate colonization [201-203]. These pH modifications can aid microbial survival but may also disrupt tissue homeostasis. Oxygen availability is similarly altered by microbial metabolism. During colonization of the infant gut, facultative anaerobes consume oxygen at the mucosal surface, creating hypoxic conditions that favor obligate anaerobes like Prevotella and Clostridia [204,205]. Disruptions to this ecological balance through diet, the consumption of antibiotics, or disease can destabilize these local pH and oxygen gradients, leading to shifts in microbial behavior and tissue function.

4. Identifying key challenges and engineering solutions to model host-microbiome interactions in vitro

Despite growing insights into the mechanisms that mediate host-microbiome interactions, critical gaps remain. These gaps persist, in part, because traditional models cannot fully recapitulate these processes within complex human tissue environments. Animal models have been instrumental in advancing our understanding of host-microbe interactions [206-208], yet they also present inherent limitations. First, many human commensals, including strict anaerobes and niche-adapted strains, fail to colonize animal hosts [16,209]. Gnotobiotic animals offer more experimental control over microbial colonization, but their immature immune systems and simplified microbiotas constrain their physiological relevance [210-212]. Interspecies variation in tissue anatomy, such as differences in intestinal villi, mucus thickness, or airway branching, can also lead to divergent microbial behaviors and host responses [17,213,214]. Furthermore, animal models offer limited spatiotemporal control over biochemical and mechanical cues, making it difficult to isolate variables like ECM stiffness, oxygen gradients, or immune cell recruitment. Finally, ethical and logistical constraints make it challenging to conduct high-throughput or longitudinal studies. Together, these challenges highlight the need for next-generation in vitro platforms capable of replicating the cellular, structural, and biochemical complexity of human tissues while allowing precise and reproducible manipulation of host-microbiome variables.

To evaluate how current platforms are meeting this need, we conducted a scoping review of the literature focused on microphysiological systems specifically designed to model human-microbiome interactions in vitro. We performed a systematic search across three databases: Web of Science, PubMed, and Scopus using keywords related to the microbiome and advanced in vitro models (e.g. tissue engineering, biomaterials, organoids, microfluidics, and organs-on-a-chip; Table S1). Review articles, conference abstracts, perspectives and opinion piece were excluded. After removing duplicates, 1,686 publications were screened using Covidence software (Figure 2), with 66 studies meeting our inclusion criteria for this review. The inclusion criteria were: (1) primary research article, (2) focused on the microbiome (either using whole communities or individual commensals), (3) inclusion of a human tissue component, and (4) use of an engineered in vitro model. Because this review concentrates on the application of in vitro models to study human-microbiome interactions, articles that did not incorporate both a host and microbial component were excluded. Studies that focused on pathogens, animal models, drug delivery or treatment strategies, and simple 2D systems (e.g. transwells without additional engineering) were also excluded. Approaches that describe the use of organoids were only included if the organoids were combined with other engineering approaches. Thorough reviews of current organoid technology can be found elsewhere [215-217].

Figure 2. Workflow of the methodology for the identification, systemic screening, and final inclusion of publications based on key terms and listed exclusion criteria.

In the process of screening and reviewing these studies, we identified eight key engineering challenges facing the development of the next generation of in vitro models (Figure 3). These challenges fall into three major categories: reconstructing tissue structure and architecture, integrating biochemical and mechanical cues, and incorporating diverse cell types and multi-tissue interactions. Below, we summarize design strategies that leverage microfluidics, tissue engineering, biomaterials, organoids, and 3D printing technology, among others, to overcome these challenges.

Figure 3. Major engineering challenges identified in the design of host-microbiome in vitro models.

These challenges fall in three main categories: recreating complex architecture, integrating biochemical and mechanical cues, and incorporating diverse cell types and multi-tissue interactions. Created in BioRender.

5. Recreating complex tissue architecture

5.1. Engineering specialized structural features

A key challenge for the design of more advanced in vitro models is replicating the specialized tissue structures present at host-microbiome interfaces. Features such as intestinal villi and crypts, gingival pockets, and stratified epithelia at the skin and cervix play important roles in nutrient transport, oxygen diffusion, host cell maturation, and microbial colonization. Engineering efforts to recreate these structural elements in the laboratory have leveraged techniques ranging from custom biomaterials scaffolds to microfluidic devices and bioreactors.

Reproducing the structural complexity of the intestinal epithelium remains a core challenge in developing physiologically relevant models to study interactions with the gut microbiome. Efforts to replicate the structure of intestinal villi within gut-on-chip devices have relied on both self-assembly and external stimuli to drive epithelial morphogenesis [218-220]. Devices that integrate organoids capitalize on their intrinsic potential for multilineage differentiation and self-organization [217]. When seeded on Matrigel, an ECM-based substrate, neonatal intestinal enteroids can self-assemble into 3D villus-like structures with appropriate apical-basal polarity [220]. The introduction of peristaltic-like mechanical forces can further enhance the formation of these structural features. Combining shear stress with cyclic mechanical strain has been shown to promote the organization of well-differentiated villi with basal crypt-like zones and apical brush borders even when using an immortalized cell line [218,219]. In fact, removal of mechanical stimulation led to epithelial flattening, loss of mucus production and microbial overgrowth [218].

Despite the many advantages of these gut-on-chip devices, they typically rely on channels lined with epithelial cells on one side and thus lack the cylindrical architecture of the native gut lumen. Biomaterials-based models can address this limitation. Chen et al. fabricated a silk fibroin scaffolds with interconnected pores and a hollow lumen (2mm in diameter) patterned with threaded nylon rods [221]. This complex geometry supported the growth of a microvillus brush border on the luminal side, mucus production, co-culture with stromal layers in the bulk of the scaffold, and the maintenance of a gut-like oxygen gradient in long-term culture.

Beyond the gastrointestinal tract, similar engineering challenges arise in the oral cavity, where host-microbe interactions are shaped by the unique structure of the gingiva, which requires models that capture both epithelial stratification and interactions with the underlying connective tissue. Emerging models focus on capturing that spatial complexity. Microfluidic systems that maintain an air-liquid interface using vertically stacked chambers have been used to grow stratified human gingival epithelium on a lamina propria-like matrix in co-culture with gingival fibroblasts [222]. Compared to static transwells, the gingiva-on-chip model exhibited improved epithelial morphogenesis and barrier function. Other researchers have developed models of the gingiva using biomaterial scaffolds. Shang et al. developed a 3D reconstructed human gingiva model by combining immortalized human gingiva keratinocytes and fibroblasts on a collagen hydrogel [223]. In this case, exposure to biofilm from commensal oral bacteria was necessary to achieve increased epithelial thickness and stratification. Greater architectural complexity can be introduced with patterning techniques. Adelfio et al. designed a three-dimensional model of the periodontal gingival pocket by replica molding silk fibroin scaffolds using 3D printed dental resin teeth [224].

Efforts to model the stratified epithelium of the cervix have also leveraged natural biomaterial approaches. Łaniewski and Herbst-Kralovetz employed a rotating wall vessel bioreactor to culture endocervical epithelial cells on collagen-coated microcarrier beads, leading to the formation of 3D, polarized, mucus-secreting aggregates [225]. Upon inoculation with health- or bacterial vaginosis-associated bacteria, the model exhibited distinct cytokine and metabolite profiles that mirrored clinical observations, highlighting its potential to study host-microbiome interactions in the female reproductive tract.

Continued innovation in fabrication techniques will be essential to replicate the structural complexity of host–microbe interfaces with greater fidelity. For example, bioprinting can enable spatial patterning of multiple cell types in complex geometries. Ultimately, leveraging these advances to create models that integrate both host and microbial components will allow researchers to model structural–microbiome interactions in diverse mucosal tissues with greater control and biological relevance.

5.2. Establishing a mucosal component

Mucosal surfaces, including those of the gastrointestinal, respiratory, and reproductive tracts, are characterized by the presence of a protective mucus layer that serves as both a physical barrier and a biochemical interface regulating host-microbe interactions. Mucus is a complex aqueous material containing mucin glycoproteins, monosaccharides, lipids, proteins, salts, and extracellular DNA [226]. The rheological properties, thickness, and glycan composition of mucus vary by tissue and are dynamically regulated by host and environmental factors [227]. Given this complexity, accurately replicating mucosal layers in vitro poses a significant challenge.

The most common strategy to model mucosal interfaces in vitro involves incorporating mucus-secreting epithelial cell lines, such as goblet cell-like HT29-MTX cells [228]. The introduction of external cues, such as mechanical forces or biochemical signals, can further modulate endogenous mucus production by these epithelial cells. In microfluidic gut-on-a-chip models, for example, fluid shear stress stimulates epithelial differentiation, leading to the formation of villus-like structures covered in mucus [229]. Similar effects of fluid dynamics on mucus production have been observed in models of other organ systems. In a lung-on-chip model, continuous fluid flow that mimicked physiological shear conditions stimulated mucus secretion by A549 lung epithelial cells [230]. Likewise, varying perfusion flow rates in a cervix-on-chip platform modulated mucus production, indicating that hydrodynamic cues can influence epithelial phenotype and secretory function [231].

Although inducing epithelial cell differentiation generates physiologically relevant mucus, this approach can be time-intensive, variable, and difficult to scale. To overcome these limitations, recent efforts have focused on engineering synthetic hydrogel alternatives that recapitulate key biochemical and mechanical features of native mucus. These scalable mimics can involve exogenous natural materials (e.g. porcine or bovine mucins), synthetic polymers, or blends of both to generate hydrogels matching the viscoelastic and adhesive properties of human mucus [232,233]. For example, De Ryck et. al. developed a co-culture model where complex oral microbial communities were grown as biofilms on a mucus-mimicking agar-mucin layer and co-cultured with keratinocytes [232]. Using this model, they were able to establish that the presence of a microbiota delayed wound healing in a scratch assay. In a blended approach, Miller and Medina generated a tunable fluorine-assisted mucus surrogate (FAMS) with adjustable chemical composition and mechanical properties that can be decorated with porcine or bovine mucins to sustain microbial colonization and growth [234]. Caco-2 epithelial cells can also be incorporated into the FAMS to generate organoid-like models.

Despite their scalability, tunability and reproducibility, synthetic mucus-like materials remain underutilized in host-microbiome interaction studies likely due to their inability to fully replicate the biochemical complexity of native mucus. In particular, they often lack the diverse glycan structures found on naturally occurring mucins that mediate critical host-microbe interactions [235]. Most synthetic systems are designed to mimic specific properties (e.g. viscocity, elasticity, or adhesiveness) rather than the full spectrum of host-derived components and dynamic responses. To bridge this gap, recent approaches have leveraged natural mucins cross-linked with polyethylene glycol thiols [236], guar gum [112,237], polyglycerol [238], and polyvinyl alcohol [239,240] to model viral influenza A lung infections [241], study airway clearance [242] and block SARS-CoV-2 infections [243], among other applications. The high degree of control and reproducibility offered by these hybrid alternatives hold promise for future integration into microphysiological models. Ultimately, the choice between synthetic, hybrid, or cell-derived mucus models should be guided by the specific research question and the desired balance between biological fidelity and experimental or manufacturing control.

5.3. Integrating the ECM

Despite the ECM’s central role in regulating cell behavior and tissue homeostasis [244], extracellular matrix components are often omitted from in vitro models of host-microbiome interactions. As a result, these models may fail to capture critical ECM-mediated mechanisms that shape microbial colonization, immune signaling, and host responses. Recent work from our lab has demonstrated that commensal bacteria in the human microbiome can directly interact with and degrade various host ECM components [196], emphasizing the need to include the ECM in relevant model systems. Nonetheless, replicating the ECM in vitro presents significant technical challenges. Its composition varies across tissue types and includes a complex network of proteins, glycoproteins, and proteoglycans that are continuously remodeled in response to mechanical and biochemical cues [194,200]. Its temporal dynamics and spatial organization are difficult to capture using static culture systems. To overcome these limitations, researchers have developed several experimental strategies to recapitulate key ECM functions in microphysiological systems.

One common strategy for integrating ECM cues into microphysiological systems is through surface coatings composed of ECM-derived materials extracted from animal tissues or cell cultures [245,246]. These biomimetic surfaces promote cell attachment and polarization. In the Leaky Gut Chip model, coating microfluidic channels with collagen I and Matrigel enables the formation of a polarized epithelium responsive to both inflammatory cytokines and live probiotic bacteria [245]. In other models, ECM-derived materials are used as hydrogel substrates rather than as thin coatings, providing culture microenvironments of tunable stiffness and composition that support the formation of stratified epithelia [247].

ECM-based hydrogels can also be used to encapsulate cells (e.g. fibroblasts, leukocytes, or tumor cells) that are usually surrounded by ECM in vivo. For example, in a gingiva-on-a-chip model, primary human gingival fibroblasts were seeded within a 3D fibrin-based mucosal matrix before seeding human oral keratinocytes on top [248]. This approach facilitated the formation of a stratified gingival epithelium with stromal support, allowing for the study of host-material interactions within an oral microenvironment. Functionalized natural biomaterials functionalized can also be used for this purpose. Shelkey et al. encapsulated tumor cells and immune cells within a hydrogel mixture containing both methacrylated collagen and thiolated hyaluronic acid to mimic native tumor ECM. They then used these immune-enhanced tumor organoids to study the modulation of cancer immunotherapy by gut microbiome metabolites [249].

By integrating the ECM cues provided by these natural biomaterials with advanced bioreactor technologies, researchers have engineered increasingly sophisticated in vitro models that support multicellular self-organization. This type of approach has been used to recreate human intestinal mucosa by embedding fibroblasts and endothelial cells within a collagen-I matrix enriched with gut basement membrane proteins [250]. This construct is cultured in a rotating wall vessel bioreactor containing epithelial cells in suspension. Together, the combination of ECM cues and microgravity generated by the reactor promoted the self-assembly of these cells into an “inside-out” organotypic model, in which the luminal surface faces outward. Because the “lumen” is more accessible in this orientation, this model allowed researchers to easily assess the epithelial response to free versus microbiota-bound secretory immunoglobulin A antibodies [250].

Whether through ECM coatings or encapsulation in ECM-based hydrogels, these approaches bridge the gap between 2D cell culture and native tissue conditions, enhancing mechanistic insights on the role of the ECM as a mediator of microbial influences on host physiology. Looking ahead, the emergence of novel ECM-mimicking biomaterial strategies offers exciting opportunities to further advance model design. Incorporating emerging materials, such as ECM-based bioinks [251], protein-engineered matrices [252], dynamically stiffening systems [253,254], and stimuli-responsive hydrogels [255], will be key to model microbiome-ECM interactions in more dynamic and tissue-specific contexts.

6. Integrating biochemical and mechanical cues that support the growth of both host cells and microbes

6.1. Creating oxygen gradients

A major obstacle in modeling host-microbiome interactions, particularly in densely colonized tissues such as the gut, is recreating the drastically different oxygen requirements of aerobic host cells and anaerobic microbes. These steep oxygen gradients are critical for shaping microbial composition and host responses; yet, they are difficult to reproduce in vitro. Current modeling approaches to address this challenge vary in complexity, ranging from relatively simple transwell-based systems to biomaterials-based and microfluidic platforms (Figure 4).

Figure 4. Schematic representations of strategies used to generate and maintain oxygen gradients for host-microbiome in vitro models.

The three highlighted approaches— transwell systems, biomaterial scaffolds, and microfluidic platforms — each exhibit distinct oxygen distributions and range in complexity. Created with BioRender.

Transwell-based systems offer a straightforward solution for generating oxygen gradients due to their compatibility with standard cell culture materials. These platforms typically separate microbial and host components using membrane inserts, with microbes placed apically and oxygenated media delivered basally. Anaerobic conditions are maintained by sealing the apical space with an impermeable material to prevent oxygen diffusion (Figure 4). The Intestinal Hemi-Anaerobic Coculture System applied this approach to support co-cultures of colonic organoid monolayers with multiple anaerobic commensals, including Bacteroides fragilis, Bifidobacterium adolescentis, Clostridium butyricum, and Akkermansia muciniphila [256]. Kim and Allbritton developed a similar transwell-based model while featuring an air-liquid interface, which generated more in vivo-like cilia formation and epithelial differentiation, including increased goblet cell numbers and mucus production [257]. While transwell systems enable initial gradient formation, the resulting oxygen profiles are usually limited to two dimensions and are less suited for dynamic or long-term culture.

Biomaterials-based systems address some of these limitations by introducing 3D scaffolds that support more complex and spatially tunable oxygen gradients. In these platforms, lumen-like structures are fabricated via replica molding or 3D printing using natural biomaterials like porous silk [258,259] or gelatin methacryloyl [260]. Oxygen depletion can be achieved either by placing the constructs directly in an anaerobic environment, enabling short-term co-culture with facultative anaerobes [260], or through autonomous consumption by host epithelial cells embedded in the scaffold [258,259]. The latter strategy generates oxygen levels as low as 1% within the lumen and supports the growth of various obligate gut anaerobes. However, these static scaffold-based systems are limited by nutrient depletion and waste build-up over time, restricting their use for long-term studies.

Microfluidic devices offer dynamic control over oxygen gradients while simultaneously addressing limitations related to nutrient exchange through continuous perfusion. These models often feature parallel channels separated by thin membranes, with oxygenated and deoxygenated media perfused through separate compartments [229,261,262]. For example, the Intestine Chip developed by Jalili-Firoozine et al. consists of an upper chamber mimicking the lumen and a lower epithelial channel [229,261]. Anaerobic conditions are sustained by enclosing the chip in a customized anaerobic chamber flushed with humidified nitrogen gas and 5% CO₂, while normoxic media flows through the epithelial side [261]. This configuration achieves physiologically low oxygen concentrations (<0.5%) in under one hour and enhances the growth of multiple obligate anerobic gut genera compared to the same chip under aerobic conditions and conventional liquid culture [229]. In a different approach, Brasino et al. attained luminal oxygen depletion by incorporating oxygen-impermeable materials between the microbial and epithelial compartments, creating a passive oxygen barrier [262]. To further support the design and optimization of these organ-on-chip systems, computational modeling tools, such as finite element analysis (often implemented using COMSOL Multiphysics software), have been used to predict oxygen gradients and refine chip parameters, including channel dimensions, flow rates, and diffusion barriers [263-266].

Regardless of the method used to generate these gradients, accurate monitoring of oxygen levels is critical to validate model performance and ensure reproducibility. Approaches to measure oxygen in engineered microphysiological systems range from endpoint assays to continuous real-time monitoring. Endpoint methods (e.g. RNA sequencing and immunocytochemistry) assess cellular responses to oxygen but require destruction of the sample [260]. Alternatively, continuous monitoring techniques offer real-time measurements of oxygen concentrations throughout the experimental period using either direct contact or non-contact probes. Direct contact methods, including microsensors or needle-type probes, are useful in systems with accessible compartments [264], while non-contact approaches like fixed fluorescent spot sensors allow external detection of oxygen without disturbing the culture [267]. The more sophisticated VisiSens system uses planar fluorescent sensor foil sheets combined with live imaging to quantify and visualize spatial and temporal oxygen gradients, and can also track additional parameters, including pH and carbon dioxide levels [229].

6.2. Generating pH gradients

pH plays a crucial role in shaping microbial community structure and function across multiple body sites. Along the digestive and female reproductive tracts, spatial pH gradients create distinct niches of microbiota and protect against pathogens by promoting or inhibiting the growth of specific microbes [100,102-104,268]. Simulating these gradients in vitro remains a challenge, but doing so is essential to capture the spatial and functional complexity of host-microbiome interactions.

In organ on-a-chip models of the female reproductive tract, acidic conditions have been established through controlled perfusion of pH-adjusted media buffered with Hank’s Balanced Salt Solution to match the pH of the vagina or cervix [269,270]. This acidic pH did not affect epithelial viability or integrity. Additionally, these systems were able to sustain these pHs during colonization with commensal bacteria like Lactobacillus crispatus [269,270], and exhibited dynamic pH increases in response to a dysbiotic bacterial consortium [269]. Interestingly, the cervix model also reproduced a longitudinal pH gradient from 5.4 to 6.2 between inflow and outflow measurements, mirroring physiological pH transitions between the ecto- and endocervix [270].

A recent study extended pH gradient modeling to the gut by using acid-producing bacteria embedded in multi-layered hydrogel, rather than relying on external pH control [271]. This multilayered system was fabricated via spin coating and photo-initiated polymerization of alternating mucin-PSS-PEGDA and PAH-PEGDA. Each of these hydrogel layers harbored distinct compositions of intestinal bacteria that transformed the concentration of nutrients and metabolites found in each layer, generating a pH gradient due to the presence of acid-producing species [271]. Because of this complexity, this approach successfully supports the culture of complex intestinal microbial communities. Using this model, the research team was able to establish that patient-derived gut microbiota can impact anti-cancer drug efficacy [271].

Future work integrating real-time pH monitoring and further spatial tuning will be essential to capture the dynamic interactions between microbiota and host tissues in site-specific microenvironments.

6.3. Applying mechanical stimuli

Mechanical forces are essential regulators of tissue structure and function at mucosal surfaces, where dynamic stimuli such as fluid stress and cyclic strain modulate epithelial architecture, barrier function, and microbial colonization [272]. In vivo, these forces arise from organ-specific processes such as luminal or salivary flow and peristalsis [273]. Replicating these dynamic mechanical cues in vitro is challenging due to the need for precise control over multiple physical parameters, including flow rate, pressure, frequency, and direction of strain. Additionally, these forces must be applied in ways that support the long-term culture of both host cells and microbiota. This section highlights engineering approaches used to apply mechanical stimuli in vitro, with a focus on how fluid shear stress and cyclic deformation have been implemented to enhance the biological relevance of host-microbiome models.

Across multiple organ systems, fluid shear stress plays a crucial role at the epithelial interface by modulating epithelial cell behavior, nutrient transport, oxygen delivery, and even pH balance in microbial biofilms [274,275]. To replicate luminal flow in vitro, many microphysiological systems rely on syringe pumps that maintain constant shear conditions across epithelial surfaces. In intestinal chips using this approach, continuous flow promotes epithelial cell differentiation and strengthens barrier function under physiologically-relevant shear rates [219,276,277]. Computational modeling tools can also be used to simulate fluid velocity and shear stress profiles, helping researchers fine-tune device parameters to better mimic healthy or diseased intestinal conditions [276,277]. To simplify operation and reduce system complexity, several models have turned to pump-free flow strategies. Pressure-driven platforms have been applied in vaginal [278] and cervical [231] models sustaining epithelial and microbiota co-cultures. Alternatively, there are also osmotic flow systems that use polyethylene glycol (PEG) as the driving agent, with tunable flow rates based on the concentration of the PEG solution [279].

Extending these flow-based approaches to the oral cavity, custom microfluidic perfusion systems have been engineered to replicate salivary flow conditions that not only promote gingival epithelial morphogenesis, but could also be adapted to mimic disease states and mouth rinsing [222]. Similarly, Adelfio et al. mimicked salivary flow using a custom peristaltic pump bioreactor and demonstrated that physiological flow rates promote the growth of commensal oral microbiota in a gingival tissue model in long-term cultures [280,281].

Beyond flow parameters alone, the cellular and biochemical properties of the epithelial surface also influence how shear stress shapes host-microbe interactions. Eshrati et al. used atomic force microscopy (AFM) to compare the elasticity of cell monolayers between goblet cell-like HT29 MTX cells and enterocyte-like Caco2 cells, finding that the softer HT29-MX monolayer promoted greater adhesion of Lactobacillus rhamnosus under flow conditions [282]. This difference was attributed to the thick mucin layer on HT29 MTX cells, which likely enhances bacterial adhesion by altering the biochemical and mechanical properties of the epithelial interface. This study underscores how the combination of mechanical forces, cellular mechanics, and biochemical composition influences microbial colonization.

In addition to shear stress, the gut epithelium is subject to multiaxial strain during peristalsis. To replicate these forces in vitro, gut-on-chip platforms can rely on vacuum-actuated membranes to cyclically stimulate the epithelial layer, which promotes the formation of well-differentiated intestinal villi [219,246]. Other microphysiological devices employ convoluted microchannels to generate multiaxial deformations resulting in peristaltic-like stretching and improved mixing in the lumen [277]. The presence of both shear stress and cyclic strain at physiologically-relevant rates influences epithelial cell differentiation, mucus production, barrier function, and bacterial overgrowth [218,219,277]. These strategies have also been adapted for age-specific modeling of the gut. For example, the Neonatal-Intestine-on-a-Chip used repeated membrane contractions to simulate peristalsis in intestinal enteroids derived from premature infants, promoting epithelial maturation [283]. This dynamic system was also used to model necrotizing enterocolitis, capturing hallmark features of necrotizing enterocolitis, including inflammatory cytokine upregulation and epithelial barrier disruption driven by microbial dysbiosis.

Overall, these in vitro models provide a powerful tool for studying the complex interplay between mechanical forces, cellular behavior, and microbiome interactions across diverse tissues in a physiologically relevant context. While most existing platforms rely on microfluidics to replicate shear stress and cyclic strain, future systems could benefit from expanding the types of forces modeled (e.g. mastication, breathing, or uterine contractions) and the scales at which they are applied. Compared to microfluidic devices, larger bioreactor-based systems may allow for more complex tissue geometries, support higher flow rates and volumes, and enable multiaxial mechanical simulation across centimeter-scale constructs [284-286]. The integration of microbial communities into these systems would broaden opportunities to explore how mechanical forces shape host-microbiome dynamics in clinically relevant settings.

7. Incorporating Diverse Host Cell Types and Multi-Organ Interactions

7.1. Integrating physiologically relevant and diverse host cell types

Host-microbiome interactions are shaped by multiple host cell types with distinct functions that detect, interpret, and respond to microbial signals [287]. Equally important is the use of host cells that accurately reflect in vivo physiology, such as primary or tissue-specific cells, rather than immortalized cell lines or cells derived from non-human species. While simplified models using a single, often immortalized, host cell type have been instrumental in uncovering key mechanisms [288], they offer only a partial view of the complex cellular interplay involved in host-microbiome interactions. As interest grows in understanding how the microbiome influences host health and disease, there is increasing recognition of the need to design models that incorporate both physiologically relevant cell types and cellular diversity.

One of the fundamental decisions in model design is the selection of cells that retain key physiological characteristics. Primary cells, directly isolated from human tissues, more accurately preserve native morphology, differentiation status, and responsiveness to microbial stimuli compared to immortalized cell lines [289]. For example, a 3D endometrium model developed using primary human endometrial epithelial cells seeded on collagen-coated beads self-assembled into villus-like mucin-producing structures and mucus production, enabling the study of bacterial adhesion patterns to the endometrial lining [290,291]. Species origin is another critical factor for host cell selection, considering that microbial colonization, metabolites, and host responses are often species-specific [8,292]. In a comparative study, epithelial enteroids derived from both human and murine tissues revealed species- and organ- specific differences in circadian gene regulation in response to microbial SCFAs [293].

Induced pluripotent stem cells (iPSCs) offer a scalable alternative for generating human cell types that retain tissue-specific phenotypes. iPSCs can be differentiated into diverse lineages, providing access to cell populations that are otherwise difficult to isolate or maintain [294]. For instance, gut-brain axis chips have successfully incorporated iPSC-derived human neurons [295,296]. Similarly, in the context of host-microbiome models, iPSC-derived intestinal epithelial cells exhibit key features of native tissue, including barrier function, polarization, and responsiveness to microbial stimuli [297]. However, iPSC-derived cells often require prolonged maturation or complex media formulations to achieve full functional equivalence with primary cells, and their epigenetic memory may influence microbial responses [298,299]. Despite these limitations, iPSCs provide a valuable platform for disease modeling and patient-specific studies.

In addition to selecting the appropriate cell type and species, advanced in vitro models increasingly integrate multiple host cell types to mimic the multicellular interactions present in native tissues. Co-culture strategies typically fall into two broad categories: three-dimensional hydrogels or microfluidic devices. In the former, various cell types including intestinal epithelial cells, endothelial cells, and fibroblasts are seeded inside ECM-mimicking scaffolds, facilitating tissue-like self-assembly and maturation [248,250]. In the latter approach, epithelial cells are typically co-cultured with stromal or endothelial cells across porous membranes that create distinct compartments while enabling intercellular communication [218,231,300,301].

Derived from iPSCs or primary tissue, organoid- based systems provide an additional layer of physiological relevance by preserving native tissue architecture and cellular heterogeneity [217]. However, traditional organoid cultures present challenges for microbiome studies due to limited access to the lumen. To address these limitations, intestinal organoids generated from tissue explants can be manipulated and incorporated into microfluidic Intestine Chips, where the lumen becomes available for colonization with complex gut microbiota [220,277,302]. In an alternative solution, Williamson et al. used 3D printing, computer vision, and semi automation to develop a high throughput organoid microinjection platform capable of introducing and sampling both individual bacteria and complex microbiome communities [297].

While the strategies described so far increase cellular heterogeneity and physiological relevance in controlled environments, they still approximate rather than fully replicate in vivo tissue environments. Tissue explants offer an alternative by preserving the native architecture, extracellular matrix, and resident cell populations of the original tissue [303,304]. These explants can also be integrated with other emerging technologies. In a study by Donkers and colleagues, human proximal colon explants were placed in a microfluidic device to investigate the beneficial effects of inulin-derived microbial metabolites on epithelial barrier integrity [305]. Nonetheless, it is important to note that tissue explants offer limited scalability and tunability, which can constrain experimental manipulation compared to other engineered platforms.

7.2. Recapitulating multi-organ interactions

The microbiome influences host physiology beyond local tissue environments through metabolites and immune signals [131,306,307]. Among human-associated microbial communities, the gut microbiome is thought to exerts particularly broad systemic effects due to its large biomass, high metabolic activity, and continuous interactions with the immune, hepatic, and neural systems [308-310]. Because single-tissue models cannot capture this complexity, multi-organ experimental platforms have become increasingly important to investigate the systemic consequences of microbial homeostasis and dysfunction.

Given the central role of immune signaling in coordinating host-microbiome interactions, in vitro gut models are increasingly incorporating immune cell types [20,309]. These immune-competent systems have demonstrated how distinct immune populations modulate host responses to microbial stimuli, promoting either tolerance or inflammation [218,311,312]. For example, co-culture of colonic epithelial cells with antigen-presenting cells revealed that exposure to commensal Faecalibacterium prausnitzii increased pro-inflammatory gene transcription, which was attenuated by the addition of CD4+ naïve T cells [311]. In a similar model, introducing peripheral blood mononuclear cells into a human gut-on-a-chip exacerbated the response to LPS, leading to epithelial barrier disruption and increased adhesion of immune cells to the endothelium [218]. Additional work has shown that gut-on-chip platforms integrating tissue-resident macrophages and dendritic cells support immunotolerance to LPS and stable colonization by probiotic Lactobacillus rhamnosus, which prevents overgrowth of opportunistic pathogens such as Candida albicans [312].

Beyond immune cell integration, microfluidic platforms are well suited for modeling multi-organ interactions because they allow the culture of cell populations in distinct controlled environments. To study gut-liver-immune interactions, Jeon et al. developed a microfluidic system containing two distinct compartments: one to co-culture gut and hepatic epithelial cells, and a second for macrophages [313]. The two compartments were connected through a controllable microchannel, allowing regulated communication between the immune and epithelial components. LPS treatment of the macrophages triggered a pro-inflammatory response that propagated to the liver, which could be mitigated by treatment with the flavonoid luteolin. Notably, the presence of gut epithelial cells influenced how luteolin was absorbed and metabolized, highlighting the functional relevance of cross-tissue communication [313]. Related work has used gut-on-a-chip systems with epithelial and endothelial compartments to model the uptake and vascular transport of microbial metabolites produced by engineered synbiotics [314].

Microfluidic devices have also been used to model the gut-brain axis. In the NeuroHuMiX model, epithelial cells and iPSC-derived enteric neurons were cultured on opposite sides of a collagen-coated microporous membrane [315]. The addition of an additional mucin-coated nanoporous membrane enabled co-culturing with commensal probiotic bacteria like Limosilactobacillus reuteri. In a similar approach, modular dual perfusion devices of the gut and brain were connected to simulate the gut brain-axis [316]. Extracellular vesicles derived from commensal Bacteroides thetaiotaomicron could translocate from the apical intestinal compartment to the basal neural compartment through an endothelial layer, demonstrating potential applications to model the blood-brain barrier. More complex platforms have extended this concept to include modular gut, liver, and brain compartments that also contain tissue-resident immune cells (e.g. Kupfer cells or microglia) and circulating naïve CD4+ T cells [296]. Using this system, the research team showed that treatment with microbial SCFAs in the gut had anti-inflammatory effects in all three compartments. However, the same SCFA treatment upregulated Parkinson’s disease-associated gene expression and neuronal damage pathways in neurons carrying a Parkinson’s disease-linked mutation [296].

These examples illustrate how integrated microphysiological systems can be engineered to dissect the systemic impact of the microbiome, with growing implications for disease modeling and personalized medicine. Future innovations, including the incorporation of vascularized tissues [317], innervated organ models [318,319], and more complex immune and stromal populations [320-322], will further enhance the physiological relevance of these platforms and enable new insights into microbiome-driven mechanisms of health and disease.

8. Key experimental readouts for evaluating host-microbiome interactions in microphysiological systems

A nuanced understanding of host-microbiome interactions in vitro requires not only sophisticated model design but also the intentional selection of biological and functional outputs. These outputs must capture the complexity of both host responses and microbial activity, enabling systems-level insights into host-microbe crosstalk and emergent behaviors. Because these systems vary widely in their design, cellular components, and microbial complexity, the choice of outcomes must be guided by the biological question and technical capabilities of each system. Experimental outcomes can be broadly classified into three overarching categories: host-specific, microbe-specific, and interaction-level measures. Commonly used assays within each category are summarized in Table 1. Key considerations when selecting readouts include their relevance to the modeled tissue and specific microbial community, spatial and temporal resolution, sensitivity, budget, compatibility with physical features of the model, and alignment with the ultimate experimental goals (e.g. mechanistic studies vs. therapeutic screening). Integrating multiple types of measurements can yield deeper mechanistic insights and strengthen the predictive power of in vitro findings. Ultimately, the standardizing and validation of critical assays, particularly against in vivo or clinical benchmarks, will be essential to ensure translational relevance and enable comparisons across studies and platforms.

Table 1. Experimental approaches to evaluate host-, microbiome-, and interaction-level outcomes in in vitro models.

Category	Measurement Type	Example assays/methods
Host	Epithelial barrier integrity	Transepithelial electrical resistance (TEER), FITC-dextran permeability assays, immunocytochemistry for cadherins and tight junction proteins (e.g. ZO-1, occludin, claudin-1)
	Viability, proliferation and metabolic activity	Live/dead staining, caspase activity assays, ATP-based assays, ki-67 staining, EdU-based assays
	Mucus production	Alcian blue or PAS staining; ELISAs, immunocytochemistry, or RT-qPCR for relevant mucins (e.g. MUC2, MUC5AC)
	Gene expression	RT-qPCR, RNA-seq (bulk or single cell), NanoString,
	Protein expression	Western blots, ELISAs, proteomics
	Morphology and spatial organization	Histology (e.g. H&E, Masson’s trichrome), immunofluorescent imaging, confocal microscopy, live imaging, quantitative image analysis of key structural features, spatial RNA sequencing
	Differentiation	Immunostaining, qPCR, or flow cytometry for lineage-specific markers,
	Immune activation	ELISA or Luminex for cytokine profiling, flow cytometry for immune cell phenotyping, macrophage polarization, T cell activation
	ECM remodeling	Zymography for MMP activity, biochemical assays (e.g. dimethylmethylene blue, hydroxyproline), gene/protein expression of MMPs and ECM components, PicoSirius red staining and quantification for collagen
Microbe	Viability and microbial load	CFU counts, qPCR for a specific 16S or 18S rRNA gene, viability stains (e.g. Live/Dead BacLight), fluorescent reporters, OD600 growth curves
	Taxonomic composition	16S rRNA (bacteria and archaea) or ITS (fungi) gene sequencing
	Community structure and function	Shotgun metagenomics, metatranscriptomics, functional annotation pipelines (e.g. HUMAnN, MetaPhlAn), metaproteomics
	Metabolic activity	Targeted metabolomics (e.g. SCFAs, bile acids), untargeted LC-MS or GC-MS, isotopic labeling
	Enzymatic activity	Functional assays for specific enzymes; fluorescent, colorimetric or luminescent reporters that detect substrate conversion or the accumulation of metabolic products
	Spatial organization	Fluorescence in situ hybridization (FISH), imaging of fluorescent strains
	Antimicrobial resistance	Minimum inhibitory concentration (MIC) assays, antibiotic challenge followed by viability staining or CFU counts, genomic detection of resistance genes via qPCR or metagenomics.
	Biofilm formation	Crystal violet staining for biomass quantification, confocal microscopy for biofilm architecture, EPS quantification assays, viability assays within biofilms (e.g., Live/Dead BacLight).
Interactions	Metabolic exchange	Stable isotope tracing (e.g., ^13C-labeled SCFAs or amino acids), m etabolic flux analysis, targeted metabolomics (e.g., host-converted microbial products or vice versa), measurement of co-metabolites only produced in co-culture (e.g., conjugated bile acids)
	Bidirectional signaling	Reporter assays for host signaling (e.g., NF-κB, TLR), detection of microbe-sensing pathways (e.g., PRRs), quantification of host-derived antimicrobial peptides or microbial evasion mechanisms (e.g., quorum sensing disruption)
	Emergent systemic effects	Quantification of multi-organ responses that arise only from host–microbe interactions, such as neurotransmitter and hormone production

9. Future directions

Advanced in vitro models that integrate host and microbiome components offer powerful platforms to study human-microbiome interactions with unprecedented precision. As outlined throughout this review, these systems have already enabled new insights into microbial regulation of epithelial differentiation, barrier function, immune signaling, and systemic host responses. Their value will only continue to grow in light of the FDA Modernization Acts and recent NIH guidance supporting the use of animal-free technologies in biomedical research and therapeutic development [27,323-325]. Similar legislative efforts are gaining traction internationally, highlighting a global momentum toward more predictive, ethical, and human-relevant model systems.

However, despite these advances, current models still exhibit significant limitations in scope and applicability. On the host side, most existing systems focus on the gut, with relatively few efforts to model other critical human-microbiome interfaces such as the cervix, skin, or lungs (Figure 5A). Broadening the repertoire of tissue types and incorporating additional cell populations – such as fibroblasts, endothelial cells, and immune cells – will be essential to increase the physiological relevance of these models, particularly in non-gastrointestinal tissues.

Figure 5. Scope of the host-microbiome in vitro models highlighted in this review.

Distribution of the (A) organs and (B) microbial components modeled in the microphysiological systems identified for this scoping literature review. (A) Most of the microphysiological systems focused on modeling the gut (78.8%). (B) Researchers focus primarily on studying interactions between human tissues and bacteria, either by exposing the models to single species of bacteria (34.9%) or small mock communities of bacteria (23.8%). A small mock community was degined as a community consisting of fewer than 10 bacterial species or strains.

On the microbial side, most in vitro systems have focused almost exclusively on bacterial species, sometimes focusing on single strains (Figure 5B). Yet, the human microbiome also includes fungi, archaea, viruses, and microbial eukaryotes that influence host physiology and disease [31,47,326,327]. Additionally, microbiomes are shaped by intrinsic and extrinsic factors, including diet, lifestyle, ancestry, and geography [44,328-333]. For example, individuals in the Global South often harbor more diverse gut microbiomes that include protozoa and helminths, which can modulate immune development and inflammation [334-336]. However, most microbiome studies and associated in vitro models rely on samples from the Global North [337-340], potentially limiting their generalizability. Similarly, host ancestry and sex shape cellular responses to disease-relevant stimuli [341-344], suggesting that these variables should be considered in the design and interpretation of host-microbiome models. Thus, increasing the representation of under-sampled populations and including diverse microbiota from a wide range of clinical and geographic contexts will be essential to build more equitable and predictive models [73,345,346]. Greater access to cell and microbiome biobanks will be critical to achieve this goal [347,348].

These limitations also point to new opportunities. Microphysiological systems are particularly well-suited for personalized medicine by enabling co-culture of patient-derived cells and microbiota to test individualized responses to therapy. As evidence grows for donor-specific host-microbiome interactions, the ability to capture biological heterogeneity will be crucial to predict patient outcomes, stratify treatment strategies, and develop targeted personalized interventions. Moreover, future systems will need to better support long-term co-culture and the study of dynamic temporal processes that are particularly relevant to understand the impact of the microbiome in chronic conditions.

However, to support widespread application, future host-microbiome in vitro systems must also become more scalable, standardized, and reproducible. Currently, most designs are custom-built and difficult to replicate across labs, which impedes broader adoption. There is, therefore, a pressing need for standardized protocols, scalable manufacturing pipelines, and interdisciplinary training programs, along with shared benchmarking frameworks to validate model performance against in vivo or clinical data. Close collaboration between scientific, engineering, and clinical communities will be essential to ensure translational utility, and robust model development and interpretation.

Reducing cost and complexity is equally critical to expand the reach of these technologies beyond specialized laboratories. This includes making platforms viable for under-resourced settings and industrial applications that demand high-throughput, reproducible workflows. Emerging efforts [234,256] to develop modular, open-source systems and simplified fabrication strategies offer promising solutions to democratize access and ensure that the next generation of models is not limited by geography or lack of infrastructure.

As the field of microbiome-based therapeutics matures, engineered tissue models will become increasingly important for translational research. Probiotics, synbiotics, and microbial consortia can be screened in these platforms to evaluate efficacy, safety, and host-specific responses [245,313,314,349]. Furthermore, advances in synthetic biology now enable the creation of engineering microbes and dynamic biosensors that respond to or influence host physiology [350-352]. When integrated into microphysiological systems, these tools offer new opportunities for research and development through real-time monitoring, iterative design, and functional validation.

In summary, engineered in vitro models have already reshaped our ability to study human-microbiome interactions. To realize their full potential, future efforts must continue to broaden their biological relevance, improve their scalability and reproducibility, and integrate diverse patient and microbial inputs. These advances will not only deepen our understanding of human and microbial biology but also accelerate the development of next-generation diagnostics and therapeutics grounded in the complex reality of human-microbe ecosystems.