‘Chip’-ing away at morphogenesis – application of organ-on-chip technologies to study tissue morphogenesis

Tissue morphogenesis describes the collection of processes by which cells arrange into forms that confer tissue-specific functions. Whether during embryogenesis, where a fertilized egg reproducibly gives rise to cells to form the body plan (Kyprianou et al., 2020), or the daily replenishment of intestinal villi (Krndija et al., 2019), tissue morphogenesis is the product of dynamic changes in cell mechanics and behavior that are coordinated in time and space with gene expression and cell fate specification. Biological control systems must therefore exist to confer fidelity and robustness to such processes. However, the nature of the signals and programs that ultimately direct specific cell behaviors within populations, and how such information is carried across time and length scales remain outstanding questions in diverse tissue morphogenic contexts.

Real-time microscopy of the dynamics of developing organisms has illuminated complex three-dimensional (3D) tissue changes, such as bending, branching, hollowing and extension, which arise from conserved cell behaviors including shape change, migration, extrusion, proliferation and apoptosis (Nelson, 2009; Zallen and Goldstein, 2017). Single-cell transcriptomics and lineage tracing methods have recently provided time-resolved, quantitative descriptions of gene regulatory networks and cell states that map to complex morphogenic events at the tissue and organism level (Marsh and Blelloch, 2020; Tyser et al., 2021). Yet, these detailed analyses have also contributed to an appreciation that biochemical and genetic instructions do not sufficiently explain self-organized, emergent tissue morphogenic behaviors that are instead orchestrated by tissue-level forces, mechanics and boundaries (Collinet and Lecuit, 2021; Heer and Martin, 2017) (Fig. 1A). For instance, in the developing zebrafish heart, proliferation-induced crowding and tension heterogeneity amongst cardiomyocytes triggers cell delamination; an event that in turn influences gene expression and fate commitment necessary to construct 3D cardiac trabeculae (Priya et al., 2020). Thus, gene expression, mechanics, and chemical and microenvironmental cues converge to instruct tissue morphogenesis (Fig. 1A), and an emerging area of emphasis is defining the contexts and mechanisms of their intersection.

Still, it is often experimentally challenging to evaluate the sufficiency of a mechanism to elicit change in tissue form within animal models. Recent advancements in microphysiological system technologies have provided an in vitro alternative to modeling developing organs and associated pathologies through the construction of miniaturized human tissue platforms, broadly termed ‘organ-on-chips’ (Kamm et al., 2018; Park et al., 2019). Organ-on-chips are microfluidic culture systems containing miniaturized native or engineered tissues with tailored organotypic architectures. This technology is generally founded in microfabrication approaches, such as photolithography techniques used in manufacturing circuitry on silicon wafers (Fig. 1B), which have been repurposed, driven by the core hypothesis that appropriately recapitulating key architectural features of a native tissue is necessary to faithfully reproduce tissue function in vitro. In contrast to conventional two-dimensional (2D) culture, an advantage of organ-on-chip systems is the ability to modularly introduce physical and chemical morphogenic inputs within a 3D biomimetic tissue, permitting the study of complex multicellular behaviors with high resolution and biological control (Fig. 1B; Box 1) (for a comprehensive review on organ-on-chip design see Leung et al., 2022). Combining organ-on-chips with advances in cellular systems, including primary organoid and pluripotent stem cell (PSC) derivatives, allows researchers to capture native tissue–tissue interfaces and the cellular complexity of organ-specific morphogenesis (Lancaster et al., 2013; Takebe and Wells, 2019). Similar to the advent of 3D culture techniques, which has accelerated the understanding of mammalian tissues, organ-on-chips offer biomedical researchers integrative experimental systems to understand how genetic, mechanical, chemical and architectural inputs together instruct human tissue morphogenesis in vitro.

In this Review, we emphasize organ-on-chip culture platforms and associated microphysiological systems as powerful experimental arenas for biological discovery, and we highlight recent examples of their application in identifying instructional programs during tissue morphogenesis. Progressing through several examples of coordinated multicellular behavior, we detail recent mechanistic discoveries uniquely revealed by these platforms. Additionally, we provide a perspective on how such platforms are poised to answer outstanding questions in distinct morphogenic contexts, and how these discoveries will be enabled by continued integration with complementary technologies.

The development of tissue rudiments into arborized tubular networks occurs through a series of morphogenic stages collectively referred to as branching morphogenesis (Wang et al., 2017). These rapid developmental processes shape the lungs, glandular organs and vasculature into specific tissue architectures that confer specialized functions like gas exchange, fluid transport and secretion. Branching morphogenesis involves coordination between cell and extracellular matrix (ECM) dynamics to initiate and elaborate branched tissues. Although common cell behaviors, which broadly include migration, proliferation, shape change, neighbor exchange and differentiation, actuate these stages of branching morphogenesis, endpoint organ-specific tissue architectures are diverse. Indeed, studies of murine embryonic explant cultures have revealed that distinct combinations of cell behaviors and mechanical and molecular mechanisms govern branching in different organs (Yu et al., 2019). These differences suggest that reciprocal interplay between genetic programming and organotypic microenvironmental cues orchestrates branching morphogenesis. Here, we highlight recent examples in which this interplay has been mechanistically parsed in microphysiological systems.

During embryonic lung development, an initial hollow bud of pseudostratified epithelia surrounded by pulmonary mesenchyme bifurcates and proceeds through stereotyped branching events to produce thousands of hollow terminal tubules that form the bronchial tree (Metzger et al., 2008; Schnatwinkel and Niswander, 2013). Mesenchyme-free mouse embryonic lung explants cultured in 3D Matrigel and treated with fibroblast growth factors autonomously buckle and branch due to epithelial proliferation-driven mechanical instabilities but fail to develop into a form resembling native tissue, demonstrating a requirement for additional boundary regulation (Varner et al., 2015). In the developing lung, epithelial growth is directed into branches, in part, by bidirectional paracrine signals between the epithelium and differentiating mesenchyme that forms contractile airway smooth muscle. Additionally, the lumen of the developing lung is pressurized relative to the pleural chest cavity, and loss of this pressure differential is clinically associated with pulmonary hypoplasia (Kim et al., 2015; Nelson et al., 2017). How these biochemical and mechanical inputs are integrated across the epithelium and developing mesenchyme is central to understanding lung branching morphogenesis.

To capture this morphogenic process, an innovative organ-on-chip system designed to mimic the fetal chest cavity has been developed to apply calibrated fluidic pressures to developing embryonic mouse lungs (Nelson et al., 2017). In this system, dissected embryonic lungs are intubated through the trachea and immobilized within a polydimethylsiloxane (PDMS) microfluidic device consisting of two isolated chambers: a pleural chamber housing the lung tissue, and a luminal chamber that controls lung lumen pressure via intubation (Fig. 2A). Transmural pressure, defined as the fluid pressure differential between the lumen and pleural chambers, was found to positively influence epithelial transcriptional programs and synchronize mesenchymal airway smooth muscle differentiation and contraction frequency to enhance the rate of lung branching morphogenesis (Nelson et al., 2017) (Fig. 2B). Indeed, consistent with a model in which smooth muscle mechanically constrains lung branching, a recent study has demonstrated the ability to sterically direct the branching of embryonic mouse mesenchyme-free lung epithelium rudiments by locally changing mechanical properties of the ECM using photoinduced hydrogel cross-linking (Urciuolo et al., 2023). A follow-up study has leveraged the same chest cavity organ-on-chip system to understand the molecular mechanism by which transmural pressure influences the coordinated activities of the epithelium and mesenchyme during lung development (Jaslove et al., 2022). By cross-referencing RNA sequencing of embryonic lungs from within devices at low or physiological transmural pressure to regions of accessible chromatin within developmentally matched mice, the authors could identify putative transcription factor motifs sensitive to lung transmural pressure. Using a combination of loss-of-function approaches, computational modeling and in situ validation in embryonic lung tissue, the authors elucidated a mechanism in which positive transmural pressure leads to increases in the retinoic acid synthesis pathway in the lung epithelium that are dependent on YAP (also known as YAP1). Retinoic acid signaling in the proximal mesenchyme induces airway smooth muscle differentiation, thereby connecting epithelial growth and smooth muscle mechanics in the developing mouse lung (Fig. 2B) (Jaslove et al., 2022).

This series of studies represents a uniquely sophisticated application of a microphysiological platform to reveal mechanisms by which fluid pressure directs embryonic lung development, while maintaining native organotypic tissue architectures, cellular fidelity and physiological tissue–tissue interfaces. Additional proof-of-concept organ-on-chip studies to model lung morphogenesis have utilized more reductionist device designs. The first published organ-on-chip model was built to simulate the alveolar–capillary interface of the human lung using 2D epithelial and endothelial monolayers separated by a compliant, porous 2D membrane (Huh et al., 2010). Despite this relatively simple design, the ability of this device to introduce cyclic strain to mimic physiological breathing function, compartmentalized fluid flow and a proximal endothelium has elicited distinctive airway epithelial functions and morphogenesis in several contexts, including ciliary micropathologies induced by smoking (Benam et al., 2016) and altered non-small-cell lung cancer receptor tyrosine kinase activity, as well as tumor drug resistance (Hassell et al., 2018). Whereas these previous organ-on-chip examples utilized immortalized human cell lines, advances in induced PSC (iPSC) technology and primary tissue-derived organoids offer new access to the biology of human lung cells such as airway basal stem cells (Hawkins et al., 2021) and alveolar cells (Jacob et al., 2017). Similarly, building upon the foundational lung-on-chip system, recent organ-on-chip and bioprinting approaches have integrated 3D organotypic lung tissue architectures with cyclic mechanical strain mimicking breathing to generate culture environments that approximate the small airway and the lung alveolar–capillary unit (Grigoryan et al., 2019; Huang et al., 2021). As advances in these technologies continue to provide access to biological processes that closely resemble those in the native lung, biologists will be poised to answer outstanding questions, such as what additional factors are necessary to spatially pattern differentiation of airway smooth muscle during development, and how epithelial–endothelial juxtracrine interactions shape tissue form and cell fate within lung alveoli.

The mammary gland initially develops into a rudimentary branched network of tubules from an epidermal placode; however, the most significant branching morphogenesis and ductal expansion occurs postnatally. Mammary ducts are composed of bilayered tubes, consisting of basally positioned myoepithelial cells and inner luminal cells (Ochoa-Espinosa and Affolter, 2012). Postnatal branching morphogenesis initiates through bifurcation of existing duct termini or side branching from existing ducts. In contrast to the stereotyped architecture of the lung, branching events and endpoint tissue architectures are stochastic (Huebner and Ewald, 2014). During mammary branching morphogenesis, the epithelium transitions to a stratified organization, during which luminal cells lose apical–basal polarization, increase migration and intercalation, and expand via proliferation, filling the lumen and propelling the growing tissue into the surrounding stroma (Neumann et al., 2018; Nguyen-Ngoc et al., 2012). Despite these dynamic changes to tissue form, adhesion to the ECM maintains the basal positioning of smooth muscle-like myoepithelial cells in 3D human mammary organoids (Cerchiari et al., 2015) and organoids isolated from resected murine mammary glands (Sirka et al., 2018). Basal myoepithelial positioning is an essential mechanical boundary that maintains tissue integrity and the outward expansion of proliferating branch tips. Myoepithelial coverage is influenced by the physical and chemical composition of the surrounding ECM (Nguyen-Ngoc et al., 2012), suggesting that cell–ECM interactions play a significant regulatory role in mammary branching.

Despite mammary gland development being one of the most studied and arguably best characterized forms of branching morphogenesis, organ-on-chip approaches to model mammary branching morphogenesis have yet to be fully realized, and applications of 3D microphysiological systems have largely focused on examining the morphogenesis of murine breast explants or the interactions of human mammary cell line acini with 3D ECM of defined topography and mechanics. Within a microfabricated culture platform in which 3D collagen fiber alignment is defined during polymerization by asymmetrically straining confining PDMS boundaries, axially aligned collagen fibers direct murine mammary explant branching in a Rac1-dependent manner (Brownfield et al., 2013). Similarly, a recent model using spheroids derived from a human mammary cell line has demonstrated that collagen fiber polarization promotes epithelial branching by stimulating locoregional cell proliferation through β1 integrin (also known as ITGB1) and ERK signaling (Katsuno-Kambe et al., 2021). Indeed, tensile forces remodel and align collagen fibers to support invasive branching in 3D engineered mammary tissues and in vivo (Gjorevski et al., 2015; Levental et al., 2009). However, examination of the murine fat pad has revealed that mammary gland branching bias occurs in the absence of aligned collagen fibers. Using a 3D bioprinted model, it has been demonstrated that, although aligned collagen fibers can direct branch orientation, the local accumulation of ECM at branch bifurcations ultimately controls the rate and overall axial bias of the developing mammary gland (Nerger et al., 2021). While the densely fibrous human mammary microenvironment does differ from the adipose-rich stroma of the mouse (Wiseman and Werb, 2002), these studies highlight the importance of benchmarking mechanistic predictions from microphysiological systems against in vivo developmental processes.

Organ-on-chip systems that model a lumenized mammary ductal epithelium offer an alternative to conventional 3D acinar spheroid models and have demonstrated early promise in their application to study human mammary branching morphogenesis (Jimenez-Torres et al., 2016; Kutys et al., 2020). In a recent organ-on-chip system, recapitulating the architecture of a 3D mammary duct terminus proximal to a microfluidic channel positioned to deliver specific interstitial morphogen gradients was found to permit the induction of distinct forms of proliferation-driven or invasive branching morphogenesis from the ductal tissue (Kutys et al., 2020). Interestingly, the morphogen TGF-β1 (TGFB1) induces epithelial-to-mesenchymal 3D invasive branching with 3D organotypic tissues but induces apoptosis in the same mammary epithelial cells in 2D on substrates of similar compliance (Leight et al., 2012), highlighting the importance of appropriately capturing tissue architecture and spatial delivery of growth factors to elicit specific morphogenic responses. Whereas this example relied on an immortalized human mammary epithelial cell line, advances in organoid cultures of human normal breast tissue have recently facilitated preservation of notoriously difficult mammary epithelial lineages in vitro (Rosenbluth et al., 2020). Thus, cell and microfluidic platforms are now in place to begin to dissect the interplay between growth factor signaling, the ECM and mammary tissue mechanics, and to evaluate whether observations from murine organoids are representative of the instructive morphogenic cues and driving cellular processes in the postnatal human mammary gland.

Endothelial cells (ECs) assemble to form intricate vascular networks, which remodel throughout adult life. Development of the hierarchical network of vessels necessary to perfuse an entire organism occurs during embryonic development first through vasculogenesis, the process by which angioblast precursor cells differentiate into ECs, lumenize and self-organize into a primitive vascular network (Udan et al., 2013). Subsequent expansion of the vasculature from pre-existing vessels occurs through a specialized form of branching morphogenesis called sprouting angiogenesis. To initiate branching of new vessel segments, EC monolayers lining blood vessels must first break monolayer symmetry, degrade the surrounding basement membrane and progressively extend toward extracellular guidance cues. Leading ECs undergo further specification as tip cells, which direct stalk cells via leader–follower collective cell migration into the surrounding ECM (Senger and Davis, 2011). These branches then lumenize and anastomose with neighboring vessels to form vascular networks. In addition to the formation and remodeling of vessel networks, vascular morphogenesis also encompasses the dynamic regulation of vessel function, including vessel barrier function, which is the fluidic compartmentalization of the vessel lumen from the surrounding tissue. Vascular barrier function is an essential homeostatic behavior that is necessary for regulated permeability, clearance of fluids and solutes, and immune surveillance, and is dynamically controlled by cell–cell adhesion and other specialized behaviors of cells lining the tissue lumen (Yuan and Rigor, 2010).

Model organisms, such as mice, chicken and zebrafish, have provided insight into the attractive and repulsive extracellular cues and intrinsic EC genetic programs underlying vascular morphogenesis, yet detailed mechanistic investigation remains difficult due to inherent in vivo experimental challenges (Hogan and Schulte-Merker, 2017; Meadows and Cleaver, 2015). 3D microfluidic organ-on-chip systems that approximate organotypic vessel architectures offer an important complement for parsing the intimate relationship between EC signaling and microenvironmental factors that pattern developing vessel networks. There are two primary organ-on-chip methods for generating models of vascular morphogenesis, which utilize either top-down or bottom-up assembly of vessel architectures (Fig. 3). Bottom-up assembly occurs through vasculogenesis, which relies on the ability of ECs to self-organize into a 3D vessel network resembling a capillary bed when exposed to appropriate growth factors or stromal cells. These microvascular network-on-chip platforms typically involve seeding EC-laden ECM hydrogels into channels within a microfluidic device that either directly interact with or are fluidically connected to stromal fibroblasts, which elicits EC self-assembly into perfusable microvessel networks (Fig. 3A). Use of these platforms has led to a detailed understanding of key stages of vasculogenic assembly: EC integrin engagement directs vacuolization or extracellular lumen expansion by a signaling interdependence between apical polarity complexes, cell division control protein 42 (Cdc42) and membrane type 1 matrix metalloproteinase (MT1-MMP, also known as MMP14) to form lumenized vessels (Davis et al., 2011). Recent organ-on-chip applications of self-organized microvascular networks have focused on microenvironmental factors that influence EC signaling underlying vasculogenic assembly. Embedding ECs within dynamic, stress-relaxing hydrogel networks leads to enhanced β1 integrin clustering, focal adhesion kinase (FAK, also known as PTK2) activation and matrix metalloproteinase (MMP) expression to promote robust vascular network assembly and deposition of basement membrane (Wei et al., 2020). Separately, delivering calibrated interstitial flow within a developing microvasculature network-on-chip has revealed that interstitial flow positively regulates EC MMP2 activity during vasculogenic assembly, resulting in improved vessel density, diameter and perfusion compared to that of networks cultured in static conditions (Fig. 3A) (Zhang et al., 2022).

Top-down vessel assembly approaches typically rely on soft lithography and sacrificial molding to generate pre-formed channels, in which endothelial monolayers are structured to form vessels. These human microvessel platforms most often involve casting 3D ECM hydrogels around a removable template, typically a stainless steel needle or PDMS rod suspended within a microfluidic device (Fig. 3B). Removal of the template after ECM hydrogel polymerization creates a 3D hollow channel, in which ECs are seeded, attach to the surrounding ECM and, within hours, assemble into endothelialized tubes. Connecting these engineered microvessels to a microfluidic handling system allows for the application of calibrated pressures and flows across the vessel lumen, thereby defining the hemodynamic shear stress imparted to the endothelium. Additionally, quantification of vascular barrier function can be measured in real time using conventional microscopy by measuring the diffusive flux of fluorescent molecules from the vessel lumen into the interstitial space (Polacheck et al., 2019). In one such application, bulk transcriptomic analyses of engineered human microvessels cultured under flow or static conditions has identified a previously undescribed role for the Notch1 receptor in promoting shear stress-mediated enhancement of vascular barrier function. Using CRISPR-Cas9 gene editing, the authors identified that barrier enhancement by Notch1 signaling does not involve canonical Notch transcriptional signaling, but instead occurs through a mechanism orchestrated by the Notch1 transmembrane domain that functions to directly stabilize VE-cadherin (CDH5)-based adherens junctions and associated cortical actin networks (Polacheck et al., 2017). Given the established prominence of Notch signaling in vascular biology, this surprising finding serves as a prime example of biological discovery that can be extracted by applying mechanistic molecular approaches to organ-on-chip systems that faithfully recapitulate tissue structure and function.

Human engineered microvessels also permit the in vitro study of angiogenesis – branching morphogenesis of new vessels from an existing vessel. Proof-of-concept applications of vessel-on-chip systems have demonstrated that extrinsic mechanical forces, such as fluid shear stresses (Galie et al., 2014; Haase et al., 2022; Song et al., 2012), local tensile stresses exerted by stromal cells (Sewell-Loftin et al., 2020) and ECM mechanics (Liu et al., 2021), can direct specific EC behavior to promote angiogenesis (Fig. 3B), although the majority of the associated EC signaling mechanisms remain to be determined. A recent multiplexed approach has been used to study how specific angiogenic cues influence EC behaviors during angiogenesis, utilizing an established two-channel angiogenesis-on-a-chip system consisting of a parent vessel and angiogenic factor gradient channel, which recapitulates sprouting angiogenesis from the stable, quiescent endothelium of an arteriole-scale parent vessel (Wang et al., 2020). Whereas vessels cultured without angiogenic factor stimulation invade minimally into the surrounding ECM, establishing a gradient of EC chemoattractant drives directional EC branching. Application of gradients of specific angiogenic factors that result in either excessive EC migration or proliferation produce sprouts with poor barrier function and limited perfusion. Thus, this study carefully illustrates that specific angiogenic factors are integrated into a balance of two fundamental EC behaviors, migration and proliferation, which dictates whether angiogenesis successfully produces perfusable vessels (Wang et al., 2020) (Fig. 3B). Still, much less is known about the signals and systems that ECs employ to sense and integrate angiogenic cues into these specific behaviors, and how these perceptions might change as a function of EC state (as monolayer, tip or stalk ECs). 3D angiogenesis-on-chip models have identified a critical role for contractility of non-muscle myosin IIA (MYH9) in coordinating interactions between stalk and tip cells during sprouting. Disrupting non-muscle myosin II activity using the inhibitor blebbistatin uniquely disrupts tip–stalk adherens junctions during angiogenic sprouting. Ablation of each non-muscle myosin II isoform via CRISPR-Cas9 has revealed that non-muscle myosin IIA controls tip–stalk junctional integrity, as well as the tensile traction forces that migrating tip cells exert on the surrounding ECM during sprouting (Fig. 3B) (Yoon et al., 2019). Taken together, these aggregate studies demonstrate that detailed investigations into EC control systems during vascular morphogenesis are possible through the combination of tractable organs-on-chips, molecular engineering and microscopy. In the near term, the continued application of vascular organ-on-chip systems might reveal the molecular cascades that trigger initial EC symmetry breaking in response to angiogenic stimuli and specify tip cells, as well as untangle the heterotypic cell–cell determinants of organ-specific microvascular bed architectures and functions.

The folding of the intestine is a conserved evolutionary process in vertebrates that results in crypt–villus architectures that facilitate intestinal self-renewal and nutrient uptake. During embryogenesis, the primitive midgut, which gives rise to the small intestine, exists as two general layers: a luminal endoderm layer and an outer mesenchymal layer. Although mechanisms of villus formation are contested due to species-specific differences, it is widely proposed that epithelial–mesenchymal cross talk between the two intestinal layers provides both signaling and mechanical cues that drive the formation of characteristic villi (Freddo et al., 2016; Shyer et al., 2013; Walton et al., 2012, 2016). These finger-like protrusions extend into the lumen and maximize the surface area required for absorption (Chin et al., 2017). Adjacent to these villus structures, non-muscle myosin II-dependent apical constriction of the epithelium results in initial invaginations into the mesenchyme, called crypts, which form to contain a stem cell niche that supports rapid tissue self-renewal (Sumigray et al., 2018). The most basal region of the intestinal crypt is populated by secretory Paneth cells and Lgr5^­-positive crypt base columnar stem cells. Daughter cells are generated and continue to rapidly divide as they translocate up the neck of the crypt through the transit-amplifying domain, undergoing lineage specification as they exit the crypt and ascend the sides of the villi (Clevers, 2013). This gives rise to patterned crypt–villus structures consisting of multiple differentiated epithelia that help mediate downstream intestinal function.

Advances in organoid technology have resulted in the ability to grow ‘mini-guts’ from a single isolated Lgr5-positive crypt base columnar stem cell (Sato et al., 2009). These intestinal organoids retain their self-renewing capacity, recapitulate the stem cell differentiation hierarchy and form characteristic crypt–villus architectures that resemble in vivo physiology (Sato et al., 2009). More recently, the application of organoid technology has uncovered key regulators of mechanical compartmentalization and cell fate patterning along the crypt–villus axis (Gjorevski et al., 2022; Pérez-González et al., 2021). Moreover, combining reductionist approaches with quantitative imaging has revealed insight into unique mechanical contributions to crypt morphogenesis (Pérez-González et al., 2021; Tallapragada et al., 2021; Yang et al., 2021). Isolation, dissociation and culture of in vivo murine intestinal crypts on soft 2D polyacrylamide compliant hydrogels results in monolayers that retain compartmentalized crypt–villus-like organization and permits measurement of regional cell traction forces via traction microscopy (Fig. 4A). The crypt-like compartment compresses the underlying substrate and exerts normal tractions downward and radial tractions inward. These tractions are driven by stem cell-specific apical constriction, resulting in an accumulation of apical actomyosin and changes in stem cell shape that lead to the invagination of the crypt-like compartment. Outside the stem cell crypt compartment, the transit-amplifying and villus-like domains exert upward normal tractions and outward radial tractions on the substrate, respectively, pulling cells out of the crypt, indicating that distinct mechanical behaviors coincide with these functional compartments (Fig. 4A). Moreover, proliferative zones within the stem cell crypt transition to collective migration in the transit-amplifying and villus domains, illustrating the capability of tailored 2D microphysiological systems to recapitulate regional behaviors observed in vivo (Pérez-González et al., 2021).

Intestinal cell lines and organoids can also be dissociated and seeded within microfluidic organ-on-chip systems with defined geometries to elicit the morphogenesis of crypt–villus tissue architectures, such as villus differentiation on a 2D porous membrane under flow and cyclic deformation (Kasendra et al., 2018; Kim and Ingber, 2013; Shin et al., 2019), or within a 3D lumen lined with periodic crypt architectures (Nikolaev et al., 2020). Additionally, providing a pre-patterned tissue geometry allows for tighter control over stochastic intestinal organoid formation processes, such as size, shape and location of the crypt. Elastomeric stamps have previously been utilized to generate microcavities of defined size and shape within ECM hydrogels (Nelson et al., 2006). Isolated Lgr5-positive intestinal stem cells self-organize into these geometrically defined microcavities to create highly reproducible lumenized intestinal organoid structures with characteristic crypt- and villus-like domains (Gjorevski et al., 2022). Remarkably, these tissue geometries determine regional differences in cell packing and cell morphology that coincide with spatial heterogeneities in YAP activity. Indeed, heterogeneities in YAP activation are known to play a role in intestinal symmetry-breaking events (Serra et al., 2019). At the curved ends of these microfabricated organoids, a YAP-mediated Notch–DLL1 lateral inhibition event occurs to establish the first Paneth cell, which provides critical support for the intestinal stem cell niche (Gjorevski et al., 2022) (Fig. 4B). This indication that tissue geometry alone can govern cell fate and the establishment of crypt–villus-like domains poses questions regarding how biochemical signaling mechanisms that are known to mediate morphogenesis in intestinal organoids, independently of mesenchyme, might work in tandem with structural cues from the microenvironment. For example, how crypt cells might be involved in bidirectional signaling with the mesenchyme to regulate remodeling of basement membrane to allow for invagination is an outstanding question. Furthermore, as these tissue geometries are predetermined, the temporal regulation of stem cell mechanotransduction by dynamic changes in crypt architecture might reveal associated regulatory mechanisms. Indeed, recent work has implemented photopatterned hydrogels to uncover a time-delayed shift in YAP nuclear translocation that occurs in response to imposed boundary curvature (Yavitt et al., 2023), which complements the observations described above linking YAP activity to establishment of the intestinal stem cell niche.

Integrating 3D patterning with microfluidics allows for constructing more complex mini-gut tissues containing multiple crypt- and villus-like compartments with a perfusable lumen (Nikolaev et al., 2020; Shin and Kim, 2022) (Fig. 4C). These systems provide a scaffold that permits a more faithful recapitulation of cell fate along the crypt–villus axis. Indeed, rare cell types that are often absent from conventional organoids, but found in the intestinal epithelium in vivo, have been identified along the villi of mini-guts (Nikolaev et al., 2020) (Fig. 4C). These mini-intestines and gut-on-a-chip platforms also show remarkable regenerative potential and are amenable to long-term disease modeling, which can provide insight into how complex inputs, such as parasitic infection or pathogenic bacteria, affect intestinal behaviors (Nikolaev et al., 2020). Further integration of immune cells, endothelial cells and other mesenchymal cells will begin to establish a comprehensive understanding of how chemical and mechanical signals are coordinated across tissue types to regulate crypt–villus patterning, architecture and function.

Model organisms have provided a quantitative and mechanistic foundation for understanding the multifactorial control of tissue morphogenesis. However, a comprehensive understanding of the interplay between gene regulatory networks, mechanics and microenvironmental factors across diverse morphogenic settings is just only beginning to emerge. These discoveries can be accelerated by new biomimetic platforms that faithfully recapitulate key architectural features of native tissue and provide modular control over cell signaling, behavior and microenvironmental parameters. Here, we highlight examples that emphasize organ-on-chip culture models as emerging in vitro experimental systems for biological discovery in the context of tissue morphogenesis. The continued development and integration of organ-on-chip models with rapidly advancing cellular, molecular and fabrication technologies will accelerate our conceptual understanding of how human tissues form, regenerate and deteriorate in the context of disease.

The organ-on-chip systems described here represent concerted efforts to use reductionist approaches to study human tissue morphogenesis in vitro, yet elucidating mechanistic links between gene expression, cell mechanics and environmental cues that together produce diverse and reproducible tissue forms remains a primary challenge. As an illustration, during embryonic renal morphogenesis, tubules undergo rounds of stereotyped branching to form the arborized structure of the kidney. Yet, whereas this process occurs through T-shaped bifurcations in vivo, ex vivo kidney organoids in 3D culture branch via trifurcations (Goodwin and Nelson, 2020). In these two settings, microenvironmental signals cooperate with genetic programming and local tissue mechanics to produce distinct tissue architectures. Although kidney organ-on-chip platforms are yet to transition from proof-of-concept to studies of morphogenesis, a recent study employing 3D physical and in silico models, along with in vitro murine explant experiments, has posited that kidney tubule tension is a necessary orchestrator of kidney tubule branching and arrangement (Prahl et al., 2023). While several of the examples discussed here involve single tissues, microfabrication and microfluidic approaches now offer methods to construct 3D heterotypic cell culture platforms with physiological tissue–tissue interfaces – boundaries that permit juxtracrine communication yet also support reciprocal morphogenic remodeling within supporting 3D microenvironments. Critical to the future promise of organ-on-chip platforms will therefore be the accommodation of somatic cells from primary human isolates or harnessing tissue-specific PSC derivatives for elucidating niche interactions essential for human developmental and regenerative morphogenic processes. Organ-on-chip platforms, combined with PSCs, are beginning to be employed to study human developmental morphogenic processes, including the formation of embryo-like structures and cell sorting processes during gastrulation (Girgin et al., 2021; Zheng et al., 2019). Another example involves transient reactivation of the embryonic-restricted ETS variant transcription factor 2 (ETV2), which reverts primary human ECs to a naïve transcriptional state, that permits autonomous formation of a robust and perfusable vascular plexus via vasculogenesis within a 3D microfluidic device (Palikuqi et al., 2020). Integrating these cell and engineering technologies offers the potential to understand the pathogenesis of congenital vascular malformations and dissect tissue-specific angiocrine communication between the parenchyma and vasculature that is essential for specialized tissue assembly and function.

Whereas micropatterning and microfluidic-based approaches offer control over tissue architecture, the ECM and soluble factors, tools being developed through molecular engineering and synthetic biology provide complementary methods to control gene expression, cell–cell interactions and cell signaling. These bottom-up approaches are not restricted to the on-demand control of specific genetic programs and signaling, but can also be used for the tailored reconstitution of cell–cell and cell–ECM interactions for targeted hypothesis testing (Johnson and Toettcher, 2018). In a recent example, using mosaic 3D spheroids expressing inducible CRISPR interference circuits to deplete E-cadherin (CDH1)-based cell–cell adhesion, 3D branching of a stratified epithelial tissue could be initiated by reconstituting strong cell–ECM adhesion and weak cell–cell adhesion within a peripheral sheet of epithelial cells (Wang et al., 2021). Furthermore, the adoption of optogenetics has allowed spatial and temporal control of specific signaling processes during complex cell and tissue morphodynamics. Recent examples include the application of an engineered light-inducible variant of the receptor tyrosine kinase Shroom3 to illustrate that spatiotemporal control of apical constriction can result in diverse forms of 3D tissue shape change (Martínez-Ara et al., 2022), and the phenotypic rescue of patterning mutations in the Drosophila embryo using optogenetic Ras stimulation (Johnson et al., 2020). The field of synthetic biology is rapidly advancing, and soon orthogonal genetic circuits that permit tailored responses to several distinct inputs with intracellular feedback will be readily available (Stevens et al., 2023; Toda et al., 2018). Similarly, the integration of engineered cellular behaviors with technologies that allow for real-time mapping of tissue stresses and cellular force generation, such as molecular tension sensors, tunable microdroplets or 3D traction force microscopy (Polacheck and Chen, 2016), can provide unique insight into how local and global mechanics drive reshaping and regulatory signaling.

Our ability to structure and control 3D human tissues that faithfully recapitulate native architectures and tissue morphogenic behavior is rapidly being made possible by advances in microphysiological systems engineering and stem cell technology. As showcased in recent studies investigating tissue morphogenesis, the use of organ-on-chip platforms establishes their feasibility, presenting a compelling argument for their continued adoption by cell and developmental biologists.

The authors thank Lakyn Mayo and Kyle Jacobs for manuscript critiques.