Integration of Extracellular Matrices into Organ‐on‐Chip Systems

Abstract

The extracellular matrix (ECM) is a complex, dynamic network present within all tissues and organs that not only acts as a mechanical support and anchorage point but can also direct fundamental cell behavior, function, and characteristics. Although the importance of the ECM is well established, the integration of well‐controlled ECMs into Organ‐on‐Chip (OoC) platforms remains challenging and the methods to modulate and assess ECM properties on OoCs remain underdeveloped. In this review, current state‐of‐the‐art design and assessment of in vitro ECM environments is discussed with a focus on their integration into OoCs. Among other things, synthetic and natural hydrogels, as well as polydimethylsiloxane (PDMS) used as substrates, coatings, or cell culture membranes are reviewed in terms of their ability to mimic the native ECM and their accessibility for characterization. The intricate interplay among materials, OoC architecture, and ECM characterization is critically discussed as it significantly complicates the design of ECM‐related studies, comparability between works, and reproducibility that can be achieved across research laboratories. Improving the biomimetic nature of OoCs by integrating properly considered ECMs would contribute to their further adoption as replacements for animal models, and precisely tailored ECM properties would promote the use of OoCs in mechanobiology.

The (bio)chemical, mechanical, and structural properties of the extracellular matrix (ECM) are well‐known to guide cellular behavior and function. Yet, the integration of carefully designed and well‐controlled ECMs into Organ‐on‐Chip (OoC) platforms remains challenging. In this review, current state‐of‐the‐art design and assessment of in vitro ECM environments in OoCs are critically discussed.

1. Introduction

Initially considered to be an inert, passive support for cells, today the extracellular matrix (ECM) is established as a complex, dynamic network within all tissues and organs that can direct fundamental cell behavior, function, and characteristics.^{[ 1} , ^{2 ]} Its importance is illustrated by the variety of syndromes that arise due to abnormalities in ECM homeostasis, such as osteogenesis imperfecta or Ehlers–Danlos syndrome.^{[ 3 ]} Comprising of more than 300 proteins and carbohydrates which are produced, maintained, and remodeled by the cells, the ECM acts as a mechanical support and an anchorage point, provides tissues with structural stability, and allows for the separation of different tissue types.^{[ 2} , ^{4 ]} Furthermore, it serves as a reservoir of a variety of growth factors and bioactive molecules and plays an essential role in the regulation of their dose and availability, both spatially and temporally.^{[ 2} , ⁵ , ^{6 ]} This two‐way communication between cells and their non‐cellular network is so significant in directing cell differentiation, organization, and communication that the evolution of metazoan (i.e., multicellular animals with differentiated tissues) cannot be separated from the evolution of their ECM.^{[ 7 ]}

To date, more than 15 different biophysical and biochemical aspects of the ECM have been identified that affect/guide cell behavior.^{[ 1 ]} ECM stiffness, for instance, has been repeatedly shown to affect the differentiation of stem cells, regulation of osteoblast functionalization, and angiogenesis.^{[ 1} , ^{8 ]} The spatial distribution and density of the integrin‐binding cell adhesion motifs of the ECM are well known to influence the adhesion, shape, and migration characteristics of osteoblasts, human umbilical vein endothelial cells, and many more cell types.^{[ 9 ]} Variations in the composition of proteins that make up the ECM, (e.g., laminin, fibronectin, and collagen I or IV) cause alterations in the proliferation, morphology, and intercellular communication characteristics of various cell types.^{[ 10} , ^{11 ]} Moreover, differences in cell behavior with respect to ECM thickness,^{[ 12 ]} porosity, and pore size,^{[ 13 ]} viscoelasticity,^{[ 14 ]} viscous dissipation,^{[ 15 ]} hydrophobicity,^{[ 16 ]} and surface micro‐ and nano‐topography^{[ 17} , ¹⁸ , ^{19 ]} have been reported. As these examples demonstrate, in order to capture, understand, and direct cellular characteristics in a precise manner in vitro, it is of crucial importance to be able to quantify and modify the biophysical and biochemical properties of the surrounding ECM environment.

The majority of our knowledge regarding cell–ECM interactions stems from traditional in vitro cell culture studies, commonly performed on hydrogel and protein‐coated glass or polystyrene culture plates, as well as permeable supports such as cell culture inserts made of materials such as polyethylene terephthalate (PET). These methods are primarily static in nature and do not inherently allow for the incorporation and control of additional biophysical and biochemical cues, such as the application of fluid flow‐induced shear stresses or maintenance of steady nutrient/waste levels in the culture media. Organ‐on‐chips (OoCs) are cell culture microsystems engineered to address these limitations of traditional cell culture environments using established microfabrication techniques, cell culture methods, and microfluidics. By enabling the incorporation of various biophysical and biochemical signals, as well as the creation of well‐controlled cell and co‐culture conditions, OoCs allow researchers to mimic cell‐, tissue‐, and organ‐level features and functions of human (patho)physiology in cell‐relevant length scales.^{[ 20} , ^{21 ]}

While OoC platforms present advanced physiological cell culture possibilities, perfectly mimicking the complex, dynamic in vivo cell microenvironment on a single platform remains challenging. The OoC community is working on numerous aspects of the technology to create more and more advanced systems, which make use of alternative materials, microfabrication, and microfluidic technologies, as well as sensing and actuation elements.^{[ 22} , ²³ , ^{24 ]} Despite the increase in complexity and high maturity level of OoCs, the integration of well‐controlled and defined ECMs into OoC platforms remains challenging and the methods to construct, modulate, and assess ECM properties on OoCs remain underdeveloped. However, without an ECM whose properties can be reliably and reproducibly controlled and modulated, cell responses and related molecular processes studied on OoC platforms may lack important extracellular cues, which can result in misleading observations and conflicting results among similar studies.

In this review, we delve into current OoC technologies and focus on ECM integration into such platforms. We start with a brief presentation of commonly‐used, ECM‐mimicking biomaterials, followed by a summary of how the ECM environment is designed and characterized in traditional 2D and 3D cell cultures. Last, we focus on commonly‐used OoC architectures, evaluate the degree of ECM integration and control each architecture allows, and discuss prominent related research. The overarching goal of this article is to highlight the importance of a well‐controlled ECM microenvironment in OoCs, summarize state‐of‐the‐art devices and methods used for ECM integration, and initiate a discussion around how the OoC community could incorporate ECMs into OoC platforms in a more comprehensive manner.

2. Mimicking the ECM In Vitro

It has been long established that cells show aberrant behavior when residing on surfaces that do not represent their own physiological environment.^{[ 25 ]} In the past, cells were commonly cultured i) on flat, glass, or plastic 2D substrates forming cellular monolayers or ii) using the hanging drop method that enables scaffold‐free 3D cellular aggregate formation.^{[ 26} , ^{27 ]} As the importance of the ECM in regulating cell attachment, morphology, and behavior became more clear, culturing methods evolved to include further characteristics that would more closely emulate the in vivo cellular microenvironment. In order to construct and study in vivo cell–cell and cell–ECM interactions, coating the standard culture surfaces with cell‐adhesive motifs/proteins and using polymers of tunable mechanical, structural, and biochemical properties have become standard practice (Figure 1 ).^{[ 27} , ²⁸ , ^{29 ]} Today, 2D cell cultures established for ECM‐related investigations predominantly employ synthetic gels, most commonly polyacrylamide (PA) (Section 2.1.2) or polydimethylsiloxane (PDMS) (Section 2.2) as cell culture substrates. 3D cell cultures mainly make use of natural hydrogels, primarily collagen (Section 2.1.1) or polyethylene glycol (PEG)‐based synthetic or hybrid hydrogels (Sections 2.1.2 and 2.1.3). Specific chemistries and fabrication protocols for ECM‐mimicking materials have been extensively reported and reviewed.^{[ 25} , ³⁰ , ³¹ , ^{32 ]} To date, there is no single established biomaterial that can perfectly capture the full biophysical and biochemical complexity of the native ECM, while allowing for reproducible property manipulations. To facilitate further discussion related to ECM integration into OoC platforms, we briefly present some ECM‐mimicking materials, particularly, selected hydrogels and PDMS, as they are commonly used in in vitro investigations and OoC systems.

Figure 1.

2D cell culture systems incorporating ECM components. a) Cells cultured on rigid substrates (e.g., glass and polystyrene) whose mechanical and topographical properties cannot be easily altered. Cell attachment to such rigid substrates in the lack of adhesion‐enabling motifs (such as ECM proteins) takes place through electrostatic and hydrophilic interactions and does not involve integrin‐based attachment pathways.^{[ 33 ]} This mode of cell attachment might require additional modifications of the substrate surface, such as plasma treatment.^{[ 34 ]} b) Rigid substrates coated with ECM‐mimicking materials (such as hydrogels) to imitate a softer, in vivo‐like ECM. Here, a natural hydrogel containing cell‐adhesive ECM fibers (such as collagen) with added growth factors is depicted. The presence of ECM fibers enables cell attachment to the surface through biochemical interactions between integrins and ECM proteins. c) An alternative to rigid substrates is elastomeric substrates, most commonly PDMS, whose mechanical and topographical properties can be tuned. To enable long‐term cell attachment, ECM proteins can be adsorbed on the surface of such elastomeric substrates.

2.1. Hydrogels

Hydrogels are hydrophilic, self‐supporting, water‐swollen 3D polymer networks that can be structured or patterned with the help of various microfabrication methods, such as micromolding, 3D‐printing, or electrospinning.^{[ 32} , ³⁵ , ³⁶ , ³⁷ , ³⁸ , ^{39 ]} Due to the similar structural properties to natural tissues and the possibility of functional modification, hydrogels are widely used in biomedical research.^{[ 25} , ³⁵ , ^{40 ]} They can be made up of synthetic, natural, or hybrid polymer chains, which are cross‐linked either by physical (non‐covalent) or chemical (covalent) bonds.^{[ 25} , ⁴¹ , ^{42 ]} This chemical makeup, together with the degree of monomer cross‐linking, determines some of the most important bulk characteristics of a hydrogel that are significant for biological applications, such as stiffness, viscosity, elasticity, roughness, wettability, hydrophobicity, swelling ratio, permeability, and tortuosity (a term that connects average pore size, pore size distribution, and pore interconnections).^{[ 35} , ⁴³ , ⁴⁴ , ^{45 ]} Typically, stiffer networks, which are characterized by a higher degree of cross‐linking exhibit smaller pore sizes, lower swelling/shrinking, and lower permeability.^{[ 1} , ²⁵ , ⁴⁰ , ^{43 ]} The inability to easily decouple such structural and physical properties poses a challenge for mechanobiological studies that make use of hydrogels, as it complicates the investigation of distinct ECM characteristics on cell properties and behavior (Figure 2a). For example, the behavioral changes exhibited by cells attached to stiffer or softer matrices might simultaneously be influenced by the different tortuosity of such matrices, as porosity and pore size impact hydrogel swelling, which in turn impacts cell migration, perfusion of nutrients and oxygen, and the removal of waste materials. Therefore, many groups are working on different methods to enable some degree of independence between structural and physicochemical properties for various hydrogels.^{[ 46} , ⁴⁷ , ^{48 ]}

Figure 2.

Interrelations between a) ECM material properties and b) cells and ECM, and their effect on cellular responses. This figure illustrates the strong interplay between properties (double arrows) that are often hard to decouple and control during ECM integration and related experiments.

An additional level of complexity related to the use of hydrogels as in vitro ECM stems from the interplay between hydrogel bulk properties and the characteristics of the cell–ECM interface (Figure 2b). To be able to sense the mechanical properties of the ECM and respond to it by remodeling the cytoskeleton and transducing forces onto the ECM, many adherent cells (e.g., epithelial, or endothelial cells) need to establish bonds to the matrix's cell‐adhesive motifs (e.g., the tripeptide Arginine–Glycine–Aspartate). These motifs are contained in the ECM adhesive proteins (e.g., collagen I or IV, laminin, fibronectin, etc.) and are most often sensed by cell–surface receptors (e.g., the integrins heterodimers).^{[ 49 ]} While for some hydrogels, it might be possible to decouple bulk hydrogel parameters, such as stiffness and tortuosity, from significant characteristics of the cell–ECM interface, such as topographical features, roughness, and hydrophobicity,^{[ 16} , ¹⁷ , ¹⁸ , ^{50 – c ]} for other hydrogels, the same bulk parameters might be entirely interconnected to the distribution, density, and bioactivity of cell‐adhesive ligands that take place at the cell–ECM interface.

2.1.1. Natural Hydrogels

Natural hydrogels such as collagen, alginate, fibrin, and Matrigel are composed of natural polymer chains, such as proteins or polysaccharides, which originate from plants, microorganisms, and animals.^{[ 45 ]} They are biocompatible, of low cytotoxicity, and advantageous to use, as they promote cell adhesion of many cell types (including sensitive stem cells) and do not require additional modifications to enable cell attachment.^{[ 40} , ^{51 ]} Moreover, cells can degrade natural hydrogels through a variety of enzymes, such as adamalysins, meprins, and matrix metalloproteases, which allows them to restructure their surroundings, as they would do in vivo.^{[ 25} , ^{52 ]} The ability of these hydrogels to be remodeled (unlike inert materials such as PA) is of great interest because in many pathological states (e.g., infection and fibrosis), human cells can get reprogrammed, altering the biophysical properties of their ECM in ways that are still under‐investigated but are likely important to determine and quantitate in order to understand disease. On the other hand, natural hydrogels present some significant drawbacks: i) their physical, structural, and biochemical properties are highly interlinked as matrix properties are commonly tuned by changing the protein concentration of the matrix,^{[ 36} , ⁴⁰ , ⁵³ , ⁵⁴ , ⁵⁵ , ^{56 ]} ii) adequate mechanical stiffness is hard to engineer,^{[ 25} , ⁵¹ , ^{56 ]} iii) they rapidly degrade,^{[ 40} , ⁵¹ , ^{56 ]} iv) they introduce a certain risk of contamination (immunogenic, pathogenic, etc.),^{[ 25} , ⁴⁰ , ⁵¹ , ^{57 ]} v) they suffer from batch‐to‐batch variability,^{[ 25} , ³⁶ , ⁴⁰ , ^{51 ]}, and vi) polymers of human/animal origin might be expensive and/or not easily available.^{[ 51 ]} Furthermore, ECM remodeling might complicate mechanobiological investigations, as gels do not maintain their original properties throughout the experiments.

A variety of natural hydrogels is available for applications in cell cultures, tissue engineering, and regenerative medicine, whose individual properties are extensively reviewed in the literature.^{[ 58 ]} In microfluidic cell cultures, the use of gels that are produced via the gelation of native ECM proteins is dominant (especially collagen gels based on the rat‐tail derived collagen type I). The composition of standard collagen gels is well‐defined, with established preparation protocols widely available in the literature and from commercial suppliers. Despite their simplicity and popularity, the preparation of collagen gels requires a certain level of expertise as the formation and architecture of the gels strongly depend on multiple experimental parameters, including ionic concentration, pH, and temperature.^{[ 59 ]} Furthermore, to stimulate cellular functions such as proliferation and differentiation, growth factors might need to be added to the prepared collagen gels (Figure 3a). This requires a deeper understanding of growth factors necessary for the specific application, as well as knowledge of growth factor adsorption and release characteristics of the gels themselves.^{[ 60 ]} One natural hydrogel that inherently captures the biochemical diversity of in vivo ECMs is Matrigel. Extracted from Englebreth–Holm–Swarm tumors in mice, Matrigel is considered to be a reconstituted basement membrane preparation that is rich in laminin, collagen type IV, and entactin.^{[ 61 ]} In addition, it contains a variety of growth factors that can be naturally found in Englebreth–Holm–Swarm tumors.^{[ 55 ]} While Matrigel is a cell‐derived hydrogel capable of representing the biochemical diversity of in vivo ECMs, it also has multiple disadvantages including an under‐defined and complex composition that limits property manipulation that can be desirable for mechanobiological research as well as challenges associated with in‐batch and batch‐to‐batch composition variability that has led to experimental uncertainties and lack of reproducibility.^{[ 53} , ⁵⁴ , ⁵⁵ , ^{62 ]} Furthermore, although biochemically complex, Matrigel cannot represent a wide range of tissues with various mechanical and biochemical properties. Decellularized ECMs (dECMs) are a notable class of natural hydrogels, which have gained significant attention in recent years and aim to overcome this limitation. dECMs are derived by removing the cellular and genetic material from human or animal organs/tissues while leaving the tissue‐specific ECM structure and architecture intact for further recellularization.^{[ 63} , ^{64 ]} Alternatively, they can be obtained via the decellularization of in vitro cell cultures, which secrete their cell‐specific ECM during the culture period.^{[ 64 ]} The decellularization process ideally allows for the preservation of tissue‐specific mechanical and biochemical properties, promising to capture the full complexity of the native ECM. Moreover, human‐derived dECMs, in contrast to ECM‐mimicking hydrogels derived from other species, might reduce unwanted cell responses during, for example, the testing of therapeutics, and lower the risk of xenogeneic disease transfer, as extensively reviewed by Kim et al.^{[ 65 ]} Despite this enormous potential, the technology surrounding the preparation of dECM scaffolds is yet to mature. The decellularization process, which can be carried out via multiple physical, chemical, and biological methods is greatly dependent on the tissue to be decellularized, and the protocols used can have a significant impact on the obtained scaffold properties. This necessitates the extensive evaluation/characterization of decellularized scaffolds, for example, the quantitative analysis of remaining cell components in the scaffold, evaluation of the 3D architecture, structure and topography of the matrix, as well as biochemical and mechanical analysis.^{[ 63} , ^{65 ]} In addition, extensive variability between the ECM properties of different donors, as well as the structural and molecular differences between organ/tissue‐ or cell‐derived dECMs are considered one of the main drawbacks of these hydrogels.^{[ 66 ]}

Figure 3.

Depiction of use of hydrogels and PDMS as ECM materials in 2D cell cultures. a) Traditional rigid culture substrates such as glass or polystyrene can be coated with natural hydrogels, commonly with collagen or Matrigel. The composition of collagen gels is better defined in comparison to Matrigel and can be modified by adding selected growth factors. b) Inert synthetic hydrogels, most commonly PA, are deposited on cell culture substrates and coated with cell‐adhesive proteins upon surface activation. c) PDMS can be deposited on a substrate or used as a standalone substrate. In both cases, PDMS is commonly coated with cell‐adhesive proteins to allow for long‐term cell attachment.

2.1.2. Synthetic Hydrogels

Synthetic hydrogels, such as PA or PEG, are composed of non‐natural polymer chains. They offer better batch‐to‐batch consistency and properties that are easier to engineer but lack the cell‐adhesive sequences, a higher‐order ECM structure, and the endogenous growth factors that are necessary for cell attachment and growth.^{[ 40} , ⁶⁷ , ^{68 ]} In order to alleviate this problem, synthetic hydrogels are either blended with natural hydrogels (to create hybrid gels, see Section 2.1.3),^{[ 46 ]} or their surface is coated with cell‐adhesive amino acid sequences or proteins after chemical treatment (Figure 3b).^{[ 10} , ²⁸ , ^{69 ]}

2.1.3. Hybrid Hydrogels

Despite the lack of an exact definition, hybrid hydrogels can be a blend of different natural and synthetic hydrogels (e.g., interpenetrating polymer networks [IPNs]), or hydrogels incorporating microstructures, such as nanoparticles or graphene oxide.^{[ 40} , ⁴² , ^{70 ]} Hybrid hydrogels are engineered for tailored biological, structural, and mechanical properties.^{[ 71 ]} Jia et al., in their review of hybrid hydrogels, considered the native ECM itself as a “hybrid hydrogel, containing multiple structural and functional components interdigitated at all length scales.”^{[ 72 ]}

2.2. PDMS

PDMS has been one of the most commonly used materials in OoCs due to its biocompatibility, optical transparency, gas permeability, elasticity, and chemical inertness. Highly versatile, well‐established PDMS fabrication methods are used to create complete microphysiological systems at a low cost. When fabricating PDMS‐based OoC devices, multiple layers of PDMS are aligned and bonded or glued to other PDMS layers, glass surfaces (e.g., microscope slides), or polymers such as PET. Aside from being the device material, PDMS (with appropriate cell‐adhesive coating) can also serve as the substrate cells reside on; and therefore, its mechanical, structural, and biochemical properties are important (Figure 3c). For mechanobiological investigations, PDMS is often used as a cell culture substrate due to its tunable stiffness and structural features (e.g., surface topography), properties that are shown to affect in vivo cell behavior.^{[ 73 ]} The stiffness of PDMS can be modified mainly by varying the elastomer base‐to‐crosslinker ratio, which can be adjusted to match most types of soft tissue.^{[ 74} , ^{75 ]} For Sylgard 184 Silicone Elastomer, the most commonly used type of PDMS for biological applications, the manufacturer‐recommended elastomer base‐to‐crosslinker mixing ratio is 10:1, which yields a stiffness of ≈1–3 MPa depending on the curing conditions. This stiffness is at least three orders of magnitude higher than typical ECMs of most adherent human endothelial or epithelial cell types.^{[ 76 ]} In addition to the chemical composition and the mechanical properties of the ECM, topographical cues also influence cell behavior, motility, and stem cell differentiation.^{[ 1} , ^{73 ]} PDMS surfaces can be microstructured via standard soft lithography methods to include various geometries including ridges, pillars, and grooves to act as topographical cues for the cells cultured on them.^{[ 77} , ^{78 ]} Well‐defined pores and controlled porosities can also be generated in PDMS membranes of defined thicknesses, which can then be used as cell‐supporting membranes in OoC systems.^{[ 79} , ^{80 ]} In contrast to hydrogels, the structural properties of PDMS cell culture membranes, such as their porosity, can be decoupled from the mechanical properties of the material relatively easily. In addition, PDMS does not swell in aqueous solutions; therefore, no changes associated with swelling/shrinking need to be considered.^{[ 81 ]}

The PDMS surface is intrinsically highly hydrophobic and does not allow long‐term cell attachment and growth without modification. Simple surface modifications to improve hydrophilicity include exposure to plasma, which is known to affect surface stiffness, or the adsorption of ECM proteins.^{[ 82} , ^{83 ]} In fact, PDMS oxidization has been known to increase its stiffness, making it unsuitable for studies where strict control of stiffness is necessary. To maintain long‐term hydrophilicity, silanization of the oxidized PDMS surface by exposure to alkoxy‐ or chloro‐silanes, such as (3‐Aminopropyl)triethoxysilane (APTES), is commonly used.^{[ 84} , ^{85 ]} APTES treatment forms amine functional groups on the PDMS surface, which then acts as a coupling agent for further coatings such as glutaraldehyde, ECM proteins, alginate, or gelatin.^{[ 86} , ^{87 –c, 88 ]} As APTES is neither user‐ nor environmentally‐friendly, polydopamine is reported as an alternative.^{[ 82} , ^{89 ]} The PDMS surface is also modified using polymer brushes through “grafting from” and “grafting to” methods, as extensively reviewed by Hemmila et al. and Gokaltun et al.^{[ 85} , ^{90 ]} Recently, Gokaltun et al. suggested another straightforward method to improve PDMS surface hydrophilicity where a block copolymer comprised of PEG‐PDMS segments was mixed into the PDMS bulk and the PDMS mixture was cured as usual.^{[ 91 ]} Upon contact with water, these added segments rearranged to cover the surface with PEG groups. Using this method, the hydrophilicity of the surface can be retained for at least 20 months while the optical and mechanical properties of PDMS are preserved. Treatments to render the PDMS surface hydrophilic might also alleviate the widely‐known non‐specific small molecule ad‐ and absorption problem of PDMS; although, the relationship is not straightforward.^{[ 86} , ^{91 ]} Moreover, in contrast to hydrogels, which might act as a soluble factor reservoir due to their swelling behavior, PDMS might be unable to reproduce the biochemical properties of the native ECM.^{[ 92 ]} Therefore, results stemming from studies employing PDMS or hydrogels as the ECM materials should be compared with caution as these materials differ in nature and interact differently with cells.^{[ 48} , ^{74 ]}

3. Design and Assessment of In Vitro ECM Environments

Before we can discuss ECM integration and characterization in OoC systems, we need to introduce the most important methods used to design and assess the properties of ECM‐mimicking environments in traditional 2D planar cell culture dishes and the significance of cell culture dimensionality. Dimensionality, that is, whether cells grow as 2D cellular monolayers or 3D aggregates, is in many cases dictated by surface topography, presentation of the cell‐adhesive ligands, and the effective pore size of the extracellular matrix.^{[ 1} , ⁹³ , ^{94 ]} For the sake of simplicity, we will refer to 2D as flat and 2.5D as topographically structured surfaces on which the cells reside as monolayers and have access to cell‐adhesive motifs only through the underlying horizontal (x–y) plane, as depicted in Figure 4a. In 2D/2.5D settings, cells i) have homogenous access to nutrients and dissolved gases present in the medium,^{[ 94} , ^{95 ]} ii) can spread and migrate in the horizontal plane in an unconstrained manner unless the substrate surface is specifically micropatterned to prevent such behavior, and iii) are forced to display apicobasal polarity as a result of the spatial differences between the apical and basolateral sides of the cell membrane.^{[ 94} , ^{96 ]} While some of these features can be relevant for cells with inherent apical–basal polarity (for example, endothelial and epithelial cells), other cell types such as hepatocytes or mesenchymal cells might show unnatural behavior in 2D cultures.^{[ 95} , ⁹⁶ , ^{97 ]}

Figure 4.

Dimensionality of in vitro culture systems. a) 2D and 2.5D surfaces will be referred to as flat and topographically structured surfaces on which cells reside as monolayers. For 3D cultures, cells need to be effectively surrounded by cell‐adhesive ligands in 3D. b) 2D and 2.5D surfaces are easily accessible for microscopy. For the characterization of 3D cultures, confocal imaging is commonly needed. c) Cell and substrate properties are easier to assess in 2D/2.5D substrates, for example, using AFM (atomic force microscopy) (red tip). Surface features (e.g., roughness and micro‐topography) and cell properties (e.g., stiffness) can be defined by scanning the area of interest. In 3D systems, AFM cannot be used to assess internal structures and is only limited to providing surface information.

To create a 3D setting, cells can be cultured in pre‐fabricated 3D scaffolds or be embedded/encapsulated in pre‐polymer hydrogel solutions so that they are effectively surrounded by cell‐adhesive ligands in both the vertical and horizontal direction (Figure 4a). When scaffolds are used, the mesh size and the cell‐degradability of the matrix direct the dimensionality perceived by the cells and the degree of cellularization.^{[ 1} , ⁹⁵ , ⁹⁸ , ^{99 ]} When the encapsulation method is used, both the components of the pre‐polymer hydrogel solution as well as the polymerization method must be biocompatible to maintain cell viability and the substrate should be susceptible to cell‐mediated remodeling to allow for cell motility.^{[ 25} , ^{94 ]} In 3D settings, i) cells experience gradients of soluble factors, nutrients, drugs, and waste products due to the non‐homogenous structure of the matrix and the distribution of cell aggregates within it,^{[ 94} , ^{95 ]} (ii) the presentation of cell‐adhesive ligands in all dimensions prevents cell spreading and migration in an unconstrained manner, which imposes changes in cell morphology and surface‐to‐volume ratio,^{[ 94 ]} and iii) cells often do not have a defined polarity, which enables for a more accurate representation of, for example, liver or cancer models.^{[ 95} , ^{100 ]} Key differences between different cell culture dimensionalities are summarized in Table 1 below.

Table 1.

Overview of the key differences among 2D, 2.5D, and 3D cell cultures

	Dimensionality of the cell culture
	2D	2.5D	3D
Cell growth	Cells grow in monolayers	Cells grow in monolayers	Cells grow as aggregates and/or spheroids
Presentation of cell‐adhesive ligands	Underlying horizontal (x–y) plane	Underlying horizontal (x–y) plane	In all three dimensions
Cell migration	In the horizontal plane, unconstrained	In the horizontal plane, constraints can be introduced by surface micro‐topography	Constraints introduced by the presence of ECM fibers in the vicinity
Cell polarity	Forced apical–basal polarity	Forced apical–basal polarity	No defined polarity
Access to nutrients, drugs, and dissolved gases	Homogeneous	Homogeneous	In gradients
Culture maintenance and access for characterization	Easy	Easy	Hard

Although 3D models are often considered as more in vivo‐like, considerable amount of knowledge on the interplay between cell behavior and ECM properties stems from traditional 2D cell cultures. A lot of studies over the past decades have shown that cellular behavior might also differ significantly when cells are placed on 2D versus 3D environments. For example, neutrophils not only use distinct receptors to adhere on 2D versus 3D matrices but also display different motility patterns.^{[ 101 ]} Despite such studies, 2D systems are widely accepted due to their practicality, the amount of control they allow over an experimental process, as well as the high degree of direct access for cell characterization. In 2D/2.5D settings, cells are cultured on top of already polymerized substrates, predominantly PA (Section 2.1.2) and PDMS (Section 2.2), which are modified to enable cell adhesion. PA and PDMS are well‐characterized and can be produced reproducibly using readily available protocols that also allow controlled tailoring of some of their properties.^{[ 102} , ^{103 ]} 2D matrices are easier to access, image (e.g., using microscopy), and evaluate during experiments, which allows for easier characterization of both materials and biological processes such as cellular pathways, cell migration, cell–cell, and cell–ECM interactions (Figure 4b). Moreover, irrespective of the used biomaterial, the validation of ECM properties (e.g., ligand density, topography) is significantly less challenging in 2D. This is also true for the characterization of local material properties (as opposed to bulk material properties) such as stiffness gradients, which can be assessed using methods such as AFM (Figure 4c).^{[ 25} , ^{104 ]}

In contrast, 3D systems are mainly restricted to natural hydrogels (primarily collagen), which offer limited control over their properties and suffer from batch‐to‐batch variations as discussed above. To establish better‐defined 3D matrices, the use of PEG‐based synthetic or hybrid hydrogels is favored over PDMS as cells cannot be encapsulated in the PDMS prepolymer solution.^{[ 99 ]} The lack of widely‐established, frequently‐used protocols to adequately and independently tune gel characteristics in 3D settings significantly complicates the interpretation and cross‐examination of results obtained from individual studies. For example, Mason et al. studied the behavior of bovine aortic endothelial cells (BAECs) embedded in 3D collagen gels, whose stiffness could be tuned independently of collagen density via non‐enzymatic glycation of collagen prior to polymerization.^{[ 105 ]} ECM stiffness could be tuned between 175 and 730 Pa with minimal effect on the inherent fiber architecture (i.e., collagen fibril arrangement and organization) of the matrix. In this study, BAECs exhibited significant increase in spreading and angiogenic outgrowth with increasing matrix stiffness at a constant collagen fiber concentration (Figure 5a). Although parameters such as the diameter of collagen fibrils or the porosity of the matrix were not evaluated in this work, the authors pointed out the important effect ligand availability and pore size can have on cell behavior. Similarly, Berger et al. designed a 3D gelatin methyl acrylate (GelMA)‐collagen I IPN in order to independently tune the stiffness and collagen ligand density for a greater stiffness range (2–12 kPa).^{[ 68 ]} Here, the total protein concentration in the matrix was kept constant by adding unmodified gelatin to the gels with lower collagen concentrations. In contrast to the findings of Mason et al., a gradual decrease in BAEC sprouting was reported as the matrix stiffness increased (at a constant collagen concentration) (Figure 5b). Although the authors attributed this variation in results to the stiffness range that was investigated, without comparable protocols, it is difficult to identify the exact mechanisms behind the conflicting cell behavior observed in these studies. Moreover, the analysis of 3D cell culture matrices is more complex and requires the examination of different length scales (i.e., macro, micro, and nano), as attentively reviewed by Lee et al.^{[ 98 ]} Established methods used to assess matrix properties in 2D settings, such as AFM, are limited to surfaces and cannot be easily transferred to 3D systems. To be characterized using microscopy, 3D matrices usually require the use of confocal imaging. While methods for characterizing cellular biomechanics, such as traction force microscopy (TFM), can be extended to measure out‐of‐plane traction forces in 3D matrices, intensive image processing and computation steps become necessary, difficulties in obtaining reference images arise, and the choice of the matrix material requires careful consideration because TFM relies on the ECM being purely elastic.^{[ 106 ]}

Figure 5.

Angiogenic outgrowth of BAECs in response to matrix stiffness. a) Stiffness of 3D collagen gels was tuned independently of the collagen density via non‐enzymatic glycation of collagen prior to polymerization. The gels depicted here contain varying concentrations of ribose (0, 50, and 100 mm) and a constant collagen density of 1.5 mg mL⁻¹. The number and length of the angiogenic sprouts increased with increasing matrix stiffness. Scale bar: 200 µm. Reproduced with permission.^{[ 105 ]} Copyright 2013, Elsevier. b) Total protein concentration of 3D GelMA‐collagen I IPN gels was kept constant by adding unmodified gelatin to the gels at constant collagen concentrations (i.e., 3 mg mL⁻¹ collagen + no gelatin, 1.5 mg mL⁻¹ collagen + 1.5 mg mL⁻¹ gelatin, no collagen + 3 mg mL⁻¹ gelatin). Increasing gel stiffness resulted in decreased invasion and sprouting for BAECs. Scale bar: 100 µm. Reproduced with permission.^{[ 68 ]} Copyright 2017, Elsevier.

To alleviate some of these challenges, contactless methods for the characterization of 3D matrix stiffness or ligand density have been reported,^{[ 107 ]} but these methods are not yet widely employed or are not suitable to use in the presence of cells. Establishing methods that can assess/validate matrix properties upon cell loading and matrix polymerization remains especially significant as the presence of cells in the gel during polymerization is likely to interfere with cross‐linking efficiency, leading to additional changes in matrix properties such as stiffness, permeability, and tortuosity (see Section 2.1).^{[ 108 ]}

4. OoCs and the ECM

In vivo cell behavior, function, and characteristics are guided by a multitude of factors including the physical, topographical, and (bio)chemical properties of the underlying/surrounding ECM,^{[ 1} , ¹⁰⁹ , ^{110 ]} extracellular physical cues such as stretch‐induced strain and flow‐induced shear stresses,^{[ 23} , ¹¹⁰ , ^{111 ]} as well as heterotypic or homotypic intercellular interactions.^{[ 112 ]} Cell culture platforms that provide researchers the ability to incorporate and control such cues (such as flow chambers and cell stretching platforms) are beginning to emerge. However, widely‐used cell culture platforms, such as traditional glass or polystyrene cell culture plates, lack such functionalities.^{[ 113 ]} To advance traditional cell culture setups for use in mechanobiological research, several groups have reported methods to incorporate additional stimuli. For example, Kohn et al. described a cone‐and‐plate viscometer setup in which endothelial cells could be exposed to shear stresses while residing on PA gels of various stiffness.^{[ 114 ]} Ortega–Prieto et al. used collagen‐coated polystyrene scaffolds to culture hepatic microtissues and recirculated the media over the scaffold via a pneumatically driven micro‐pump.^{[ 115 ]} In a similar approach, Kaushik et al. reported custom‐built bioreactors used to generate flow patterns at the bottom chambers of Transwells, while a mixture of endothelial cells and pericytes was seeded in a PEG‐based hydrogel placed in the upper Transwell chamber.^{[ 116 ]} While these methods partly allow the introduction of biomechanical cues in standard cell culture systems, they still lack a high level of control over experimental conditions, are not implemented in cell‐relevant scales, and may not allow for cell co‐culturing.

Microfabricated cell culture systems, most prominently microfluidics‐based OoCs, allow for precise control of experimental conditions and can be highly standardized, which promises a high degree of experimental reproducibility with minimal reagent consumption. Using microfabrication methods such as photolithography, soft‐lithography, and laser‐based microstructuring including 3D printing, OoCs can be fabricated with great flexibility in design, architecture, and functionality. Integration of microsensors for monitoring the cell microenvironment, state, and function (e.g., pH, cell metabolism) allow for further platform customization and the application of OoCs to a wide range of biomedical research areas.^{[ 117 ]} Commercial and lab‐made OoC platforms have been extensively used to identify mechanisms of action behind diseases, development of therapeutics, drug discovery and testing, and regenerative medicine.^{[ 20} , ^{118 ]} Yet, the application of OoCs to the area of mechanobiology has been very limited as is the level of control over the ECM environment that can be attained in OoCs. This is primarily due to the materials used and the OoC device architecture that dictate how the ECM is integrated into the system and how accessible it and the resident cells are for characterization.

In the following sections, we discuss three common OoC architectures, namely single‐channel OoCs, OoCs with multiple vertical channels, and OoCs with multiple planar channels (Figure 6 ), with respect to three important aspects of the ECM: its integration into the chip, the possibility for ECM modulation, and ECM accessibility for property characterization. Where relevant, we introduce and discuss emerging techniques used to overcome limitations related to ECM integration, property control, and characterization.

Figure 6.

Commonly used OoC architectures: single‐channel OoC, OoC with multiple vertical channels, and OoC with multiple planar channels. Each microfluidic channel is shown in a different color. a) Isometric chip view. b) Top view. c) Cross‐section view of the channel profile. In vertical OoCs, microfluidic channels are vertically separated by a porous membrane. In planar OoCs, channels are not physically separated but distinct flow patterns can be established using microstructured barriers at channel interfaces referred to as “phase guides.”

4.1. ECM Integration into Single‐Channel OoCs

Single‐channel OoCs consist of a single microfluidic compartment (Figure 7a). Most commonly, the microchannel is structured in PDMS and the device is sealed using a glass microscope slide. The use of glass makes the cell culture system compatible with established imaging techniques such as widefield epifluorescence microscopy and the use of inverted microscopes.^{[ 88} , ¹¹⁹ , ^{120 ]} This OoC device architecture is simple to produce; however, incorporating co‐cultures in a defined and controlled manner is challenging and application of other mechanical cues beyond flow‐induced shear stresses might not be possible due to the lack of a suspended membrane. In single‐channel OoCs, an ECM environment is commonly established by making direct use of the device materials, for example, by functionalizing the microfluidic channel as discussed below.

Figure 7.

ECM integration and characterization in single‐channel OoCs. a) Cross‐section of the microfluidic channel. b) Bottom OoC layer serves as stiff/soft/topographically‐structured substrate/ECM. c) Spatially‐defined ECM coatings and 3D ECM matrices can be integrated into single‐channel OoCs by microstamping and 3D printing prior to device sealing. Upon device sealing and sterilization, ECM can be introduced into the chip by flow. d) ECM environments can be accessed and characterized by methods such as microscopy or AFM (prior to device sealing).

4.1.1. Microfluidic Channel Acts as the ECM

The bottom surface of the microfluidic channel, most commonly made of glass or PDMS, can serve as the cell culture substrate. In this approach, cells assume the mechanical, topographical, and biochemical properties of the underlying substrate as their ECM. If the device fabrication method allows for a layered production approach, (i.e., the bottom layer of the microfluidic channel/cell culture substrate can be produced/prepared independently of the rest of the device) ECM‐related properties of the substrate can be flexibly modulated and characterized before the device is sealed. A layered production approach also allows for the use of a wider range of materials, which provides options concerning ECM properties such as stiffness; and thus, allows a more in vivo‐like representation of tissues/organs. For example, bone tissue is highly organized and its ECM is particularly stiff (10–25 GPa).^{[ 121 ]} The creation and integration of such stiff matrices into OoC systems are highly challenging. Recently, Tang et al. reported a hydroxyapatite (HA)‐PDMS‐based single‐channel OoC for bone tissue culture.^{[ 122 ]} In this system, HA, a bioactive ceramic mineral with properties that closely resemble the chemical and structural properties of human bones,^{[ 123 ]} was 3D‐printed as a paste, sintered, and employed as the bottom layer in OoC devices. Human fetal osteoblast cells cultured on these HA substrates demonstrated higher proliferation rates and higher osteogenic differentiation markers in comparison to PDMS, whose stiffness cannot imitate those of bone ECM (maximum reported PDMS stiffness is 10 MPa).^{[ 122} , ^{124 ]} PDMS stiffness can be adjusted to that of soft tissues (i.e., down to 100 kPa) by varying the elastomer base‐to‐crosslinker ratio and curing conditions (see Section 2.2).^{[ 104 ]} However, a high base‐to‐crosslinker ratio can result in ultrasoft PDMS matrices that are difficult to handle as standalone substrates; therefore, they are often integrated into single‐channel OoCs on stiff supports, such as glass slides. Using this technique, Das et al. produced 500 µm thick, ≈10 kPa stiff PDMS matrices (65:1 elastomer‐to‐crosslinker ratio) on glass microscope slides, which were then used as the bottom part in single‐channel OoCs.^{[ 125 ]} The authors also enhanced the functionality of the device by embedding fluorescent beads into the ultrasoft PDMS matrix, which enabled the use of TFM to map spatio‐temporal changes in cellular traction forces under both static incubation and dynamic flow conditions. However, tuning ECM stiffness by varying the elastomer base‐to‐crosslinker ratio might also have undesirable biochemical implications. For example, soft PDMS matrices can be more sticky and the effect of PDMS adhesiveness on protein adsorption and cell adhesion is still unclear.^{[ 12 ]} In addition, even at optimal, manufacturer‐recommended elastomer‐to‐crosslinker mixing ratios, curing PDMS is a time and temperature dependent process that does not reach 100% crosslinking.^{[ 85 ]} Using higher mixing ratios to match the stiffness of soft tissues can lead to an increase in uncrosslinked/uncured oligomers in the PDMS bulk, which then diffuse into the cell culture medium and interfere with cell behavior.^{[ 31} , ^{126 ]} To investigate this problem, Regehr et al. used Soxhlet extraction (a method that is used to isolate solutes from difficult‐to‐extract samples^{[ 127 ]}) to remove uncured PDMS oligomers from PDMS microchannels.^{[ 128 ]} After 24 h of extraction and although devices had lost 4% of their total mass, PDMS oligomers were still detectable in the microchannel using mass spectroscopy and could be identified in the plasma membranes of mouse mammary epithelial cells that were cultured in extracted PDMS microchannels. Subsequently, Glover et al. showed that such extraction methods have an effect on the mechanical properties and the geometry of PDMS parts,^{[ 129 ]} which further emphasizes the need for ECM characterization directly before and/or after integration into OoCs. As an alternative to such extraction‐based methods, which increase the amount of time necessary for device preparation and characterization, Palchesko et al. blended two types of commercial PDMS, Sylgard 184 and Sylgard 527, each prepared at their optimal elastomer‐to‐crosslinker ratio to control the stiffness while minimizing the amount of uncured monomers.^{[ 130 ]} With this method, they could achieve stiffnesses from 5 kPa to 1.72 MPa, with minimal effects on surface roughness and surface energy.

In addition to stiffness control, single‐channel OoCs in which the microfluidic channel serves as the ECM, can offer a significant level of control over the design of surface topography. In order to study the effects of topographical features on cellular behavior, well‐defined nano‐ and micro‐level structures, such as grooves, wells, pillars, or fibers have been patterned on PDMS surfaces using standard soft lithography methods to create 2.5D environments (Figure 7b).^{[ 131 ]} Alternatively, pre‐produced scaffolds with distinct porosity and topographical features can be integrated onto the microchannels to achieve 2.5D–3D environments (see Section 4.2.2).^{[ 132 ]} Furthermore, to enable cell adhesion and growth, the microchannel surface can be patterned with cell‐adhesive ligands in a controlled manner prior to OoC sealing, as discussed below.

4.1.2. Integration of ECM Proteins prior to OoC Sealing

Long‐term cell attachment in OoCs is of critical importance, especially because external mechanical stimuli such as flow‐induced stress and stress‐induced strain are often imposed onto the cells. Even though cells can attach to unmodified substrates such as glass or the sintered HA substrate discussed in Section 4.1.1 above through nonspecific electrostatic interactions, long‐term cell attachment can be significantly improved by introducing cell‐adhesive ECM components, such as collagen or fibronectin.^{[ 122} , ^{133 ]} In single‐channel OoCs, the integration of cell‐adhesive ECM motifs into the devices can either be realized before sealing the device, as will be discussed in this section, or after device closure, as will be introduced in Section 4.1.3.

Establishing an ECM environment in OoCs before sealing the systems allows for direct access to the ECM and offers similar design, structuring, and characterization opportunities as traditional cell cultures (Figure 7c,d). The main disadvantage of this method is that the presence of proteins and cell‐adhesive ligands in the channel greatly restricts the methods that can be used to seal OoCs as commonly used methods such as plasma treatment and high‐temperature bonding can alter the properties of the ECM.^{[ 131} , ¹³⁴ , ^{135 ]} To overcome this issue, Rhee et al. demonstrated the use of an indirect micropatterning method, which can be extended to ECM‐protein patterning in single‐channel OoCs.^{[ 135 ]} The authors first coated the surface of circular cover slips with poly‐L‐lysine (PLL), a polycation that acts as a glue facilitating 2D cell attachment or increasing cell adhesivity.^{[ 136 ]} They then masked defined sections of the coated surface using a PDMS piece and etched away the uncovered PLL layer using oxygen plasma, while simultaneously activating the glass substrate underneath. Subsequently, they bonded the plasma activated glass–PLL surface to an oxygen plasma‐treated PDMS part to seal the device. This method allowed for the functionalization of the OoC substrate prior to device sealing and the selective attachment of cells, as seen in Figure 8 .

Figure 8.

Indirect micropatterning method to create defined ECM patterns in single‐channel OoCs. Adapted with permission.^{[ 135 ]} Copyright 2005, Royal Society of Chemistry. a) Micropatterning process. b) Micrographs of patterned PLL lines, which act as a glue facilitating 2D cell attachment. Bright areas indicate the PLL‐patterned areas, black areas indicate etched sections. c) Primary rat cortical neurons cultures on PLL‐micropatterned surfaces.

In order to create 3D culture settings in OoCs, cell‐laden, hydrogel‐based bioinks can be printed into unsealed devices (Figure 7c). Lee and Cho bioprinted cell‐laden collagen and gelatin hydrogels of defined thicknesses into 3D‐printed poly(ε‐caprolactone) OoCs at exact positions to create spatially heterogeneous tissue models.^{[ 137 ]} To prevent cell death during device sealing, the authors 3D‐printed the complete device in a single step. Similarly, Abudupataer et al. bioprinted cell‐laden GelMA hydrogels into a polymethyl methacrylate‐based single‐channel OoC and sealed the devices using double‐sided tape, which had no influence on cell viability.^{[ 39 ]} Here, the viscosity and tortuosity of the non‐cellularized GelMA hydrogels were analyzed prior to device sealing by rheological measurements and scanning electron microscopy (SEM), respectively. While the measurement of gel properties without the presence of cells is a commonly‐used method to characterize 3D ECM settings, gel polymerization after cell loading is likely to occur differently than gel polymerization without cells.^{[ 108 ]} This, along with the challenges associated with the homogenous modulation of 3D hydrogel properties, significantly complicates the characterization of the 3D ECMs, as discussed in Section 3 above.

4.1.3. ECM Coating of Microfluidic Channels through Fluid Flow

Integration of ECM prior to OoC sealing significantly restricts the processes that can be employed to seal the devices. The commonly used alternative method is to introduce cell‐adhesive ECM proteins (e.g., collagen, fibronectin, and laminin) into the OoCs through flow following device sealing and sterilization. While this channel coating process is in principle simple, the success of the coating depends on many parameters such as the wettability of the channel surfaces and whether a homogenous, evenly distributed protein layer can be achieved on microchannel surfaces that are stable under flow conditions.^{[ 88} , ^{89 ]} Furthermore, ECM protein introduction into OoCs after device closure prevents direct access to the ECM; therefore, the modulation and characterization of ECM properties becomes difficult (Figure 7c,d).

In the presence of cell‐adhesive biomolecule coatings, the mechanical, topographical, and biochemical features of the ECM no longer solely depend on the microchannel material but also on the properties of the coating as a coating can change the bioactivity, roughness, and stiffness of a surface.^{[ 89} , ^{138 ]} The extent to which a coating affects the ECM environment depends on its thickness (which can be in the sub‐micron to well over millimeter range) as well as the concentration of the protein solution.^{[ 139} , ^{140 ]} Buxboim et al. investigated how “deeply” cells feel the stiffness of their microenvironment by culturing human mesenchymal stem cells on collagen‐coated PA gels of varying thickness, which were covalently attached to stiff glass substrates.^{[ 12 ]} They found that mesenchymal stem cells can sense a rigid surface located less than 5 µm beneath them and moderately respond to a rigid surface located 10–20 µm beneath them. Similarly, Kuo et al. showed that fibroblasts were able to sense stiffer glass or polystyrene substrates under compliant PA gels, when the gel thickness approached the lateral cell dimension.^{[ 141 ]} These results suggest that depending on protein coating thickness in OoC microchannels, cells might be primarily sensing and reacting to the stiffness of the coating they are cultured on or a complex stiffness that is a result of the coating and the underlying substrate itself. This emphasizes the need for standardized coating protocols and ECM characterization in OoC devices prior to cell seeding, which is practically challenging when the ECM microenvironment is established after devices are sealed.

4.2. ECM Integration into Vertical OoCs with Multiple Compartments

Vertical OoCs consist of vertically stacked microchannels that are separated from each other via a suspended porous membrane. This OoC architecture is commonly used to imitate the characteristics of in vivo barriers, which frequently include monolayers of epithelial and endothelial cells as it allows for the creation of physically partitioned co‐cultures, air–tissue interfaces, and the induction of apicobasal polarity. To create tissue interfaces and better mimic the organ‐level functions, more than one cell types can be cultured in the microchannels. Vertical OoC architectures allow conduction of high‐resolution imaging along the membrane axis, the application of external mechanical signals (e.g., shear stresses due to apically exposed fluid flow), and the introduction of oxygen and nutrient gradients across the microchannels. While it is possible to produce vertical OoCs with multiple microfluidic compartments separated by multiple membranes,^{[ 142} , ¹⁴³ , ^{144 ]} the most commonly used vertical OoCs make use of two microfluidic compartments separated by a porous membrane.

As illustrated in Figure 9 , cells can be introduced into vertical OoCs to create tissue interfaces in multiple ways: i) they can be cultured on the (coated) microchannel surfaces, assuming the properties of the microfluidic channel as their native ECM (similar to Section 4.1),^{[ 145} , ¹⁴⁶ , ^{147 ]} ii) cell‐laden hydrogels can be introduced into the channels to create 3D culture settings (discussed in Section 4.3),^{[ 142} , ¹⁴³ , ¹⁴⁴ , ^{148 ]} and iii) cells can be cultured as monolayers on the (coated) porous membrane, which then acts as the ECM. In the sections below, we will focus on different membrane types that can be integrated into vertical OoCs, whose mechanical, topographical, and biochemical properties determine the in vitro ECM environment.

Figure 9.

Tissue interfaces can be integrated into vertical OoCs in multiple ways. a) Channel cross‐section of vertical OoC system. In addition to cells cultured on the membrane, a second tissue type can be b) cultured on the (coated) microchannel surface, c) introduced to the channel in cell‐laden hydrogels, or d) cultured as a monolayer on the opposing side of the porous membrane. Depending on how selective (localized) ECM introduction into the OoC is, cells would also attach to other microchannel surfaces. A simplified case is demonstrated in this figure.

4.2.1. Synthetic Polymer Membranes

The majority of vertical OoC devices make use of synthetic porous polymer membranes, which are biocompatible but not enzymatically degradable by cells. Most popular membrane materials include commercially available PET or polycarbonate (PC) membranes, as well as porous, thin PDMS membranes, which are typically produced using standard photolithography and soft lithography methods (see Section 2.2). Such synthetic polymer membranes are most often independently purchased or fabricated, which necessitates their subsequent alignment and integration into the device. The success of the chip fabrication process (e.g., whether a tightly sealed device can be achieved) highly depends on the material of the membrane and the material of the chip body. For example, PDMS membranes can be successfully integrated into PDMS‐based devices using oxygen plasma‐initiated PDMS‐to‐PDMS bonding. The integration of PET and PC membranes is; however, more complex. PET and PC membranes have been bonded to PDMS‐based vertical OoCs by processes such as corona treatment (a surface modification technique that uses a low temperature corona discharge plasma to change surface properties) combined with high‐temperature bonding,^{[ 149 ]} oxygen plasma treatment,^{[ 146} , ^{150 ]} silanization combined with oxygen plasma treatment,^{[ 139} , ^{151 ]} silanization combined with high‐temperature bonding,^{[ 152 ]} and using PDMS mortar as a glue.^{[ 147} , ^{153 ]} When it comes to PDMS‐free OoCs, double‐sided adhesive layers,^{[ 154 ]} clamping,^{[ 155 ]} and other methods^{[ 22 ]} are used to integrate PET and PC membranes.

As synthetic polymer membranes are commonly purchased as independent layers, the possibility to modulate or specify their properties is very limited. Mechanically, suspended membranes need to have enough structural integrity to allow their production and handling, as well as to prevent membrane buckling. Topographically, micro‐level surface features are usually limited to pores. Properties such as pore size, porosity, and membrane thickness influence not only the mechanical properties of the membrane acting as the ECM but also the degree of direct contact and communication between cells cultured on each side, the exchange of molecules (such as hormones, drugs, and nutrients) between the two channels, and the migration of cells (especially relevant to circulating tumor cells, pathogens, and immune cells).^{[ 80} , ¹⁵⁶ , ^{157 ]} In addition, pore size and placement have been shown to affect cellular adhesion, morphology, and spreading, as well as the production of native ECM proteins by cells.^{[ 158 ]} While larger pore sizes are associated with increased direct cellular contact and enhanced transport/migration across the membrane, pore size needs to be carefully considered to avoid unintended cell migration across microfluidic chambers that could disrupt the integrity of the established cell monolayer. In the case of commercial PET and PC membranes, membrane thickness, pore sizes, and porosity are defined by the manufacturer and are highly standardized. Pores are created by track‐etching, a method that makes use of irradiation and chemical etching to define pore size, shape, and density.^{[ 159} , ^{160 ]} Commonly available pore diameters include 0.4, 1, 3, 5, and 8 µm. Porosity remains inherently low as higher porosity increases the incidence of merged tracks.^{[ 156} , ^{160 ]} Moreover, due to the inherent properties of the track‐etching process, pore distribution cannot be well‐controlled, leading to random, localized variations in pore density, as seen in Figure 10 .^{[ 156} , ^{161 ]} In addition, the relatively high thickness of commercial polymer membranes can limit cell–cell interactions through the membrane.

Figure 10.

Commercial PET cell culture membranes at 60× magnification. Reproduced with permission.^{[ 161 ]} Copyright 2013, Elsevier. Imperfections include regional variations in pore density and fused pores with larger‐than‐desired pore diameter. a) 1‐µm pores, 1.6% porosity. b) 3‐µm pores, 14% porosity. c) 8‐µm pores, 10% porosity. All scale bars are 10 µm.

When cells are cultured directly on synthetic polymer membranes acting as their ECM, membrane properties and their interplay need to be carefully considered. For example, although commercial PET and PC membranes are commonly used as suspended cell culture membranes in vertical OoCs, their stiffness (typically between 2 and 3 GPa) is far from the physiological stiffness of most soft tissues.^{[ 104} , ¹⁵⁶ , ^{162 ]} Another parameter that needs to be taken into account depending on the application is the interplay between porosity and stiffness.^{[ 156} , ^{163 ]} Furthermore, an increasing amount of vertical OoCs is used to mimic in vivo breathing or peristaltic motions. In these devices, the effect of membrane elasticity, stiffness, and thickness on the amount of produced mechanical strain also needs to be considered.^{[ 164 ]}

In contrast to commercial PET and PC membranes, porous PDMS membranes are commonly fabricated in research laboratories. While pore size and distribution on PDMS membranes can be well‐defined, reproducible production of structurally stable membranes with features less than 5 µm remains challenging.^{[ 79 ]} Pore diameters produced via replica molding are commonly 8–10 µm with 25 µm center‐to‐center pore spacing.^{[ 80} , ¹⁶⁵ , ^{166 ]} By employing a dry etching method to structure the pores and a water‐soluble sacrificial layer to help transport the membrane, Quirós‐Solano et al. recently reported the production of PDMS membranes with pores as small as 2 µm in diameter, 3 µm center‐to‐center pore spacing, and a porosity range of 2–65%.^{[ 79 ]} Furthermore, this method allowed the production of ≈4 µm thick membranes, significantly thinner than commercial PET and PC membranes (typically 10–15 µm) or PDMS membranes produced via replica molding (typically 10–30 µm).^{[ 80} , ¹⁶⁵ , ^{167 ]} Such ultrathin membranes integrated into vertical OoCs might better imitate thin barriers, such as the interstitium between the alveolar epithelium and capillary endothelium (<1 µm),^{[ 80 ]} the blood–brain barrier (<500 nm),^{[ 168 ]} or the glomerular basement membrane (<500 nm) that separates the capillary endothelium and the urinary space.^{[ 169 ]} To produce porous membranes with sub‐micron thickness, the use of alternative materials, such as poly(L‐lactic acid),^{[ 170 ]} poly(lactide‐co‐caprolactone),^{[ 171 ]} and silicon nitride has been reported.

As discussed in Sections 4.1.2. and 4.1.3. above, long‐term cell attachment is of critical importance for the formation of tissues and organs in OoCs. Commercial PET and PC membranes can be treated with oxygen‐containing plasma to make their surface hydrophilic, which in turn facilitates protein adsorption and cell adhesion.^{[ 172 ]} For example, Abdalkader et al. reported a vertical OoC with an integrated PET membrane to mimic the human corneal epithelium where cell attachment to the PET membrane was realized solely through corona treatment of the membrane without any specific ECM‐protein coating.^{[ 149 ]} Yet, the presence of cell‐adhesive ECM ligands on synthetic polymeric membranes ensures better cell adhesion, especially in the case of PDMS that quickly becomes hydrophobic even after plasma treatment (see Section 2.2).^{[ 85 ]} Furthermore, in order to create multiple tissue interfaces in one vertical OoC, or similarly, in order to recapitulate the 3D structure of certain tissues through “sandwich” cultures (i.e., by “sandwiching” a cellular monolayer between two layers of ECM), a multi‐layer coating approach can be used. For example, to recreate the connective tissue between the alveolar epithelium and the microvascular endothelium in the lung‐on‐chip model introduced by Ingber and colleagues, fibroblasts embedded within thin collagen gels could be plated onto the porous membrane through fluid flow prior to the seeding of human alveolar epithelium.^{[ 80 ]} However, this method might suffer from the difficulties discussed in Section 4.1.3, namely, the uneven distribution of cell‐embedded gels across the membrane surface or limitations related to the modulation and the characterization of ECM properties following device sealing. Alternatively, vertical OoCs that allow for membrane handling before device sealing can be used, such as the one reported by Li et al.^{[ 173 ]} In this work, multiple cell types were sequentially layered to form a “sandwich” culture that mimics the 3D structure of the liver acinus, shown in Figure 11 . A “sandwich” culture involves the addition of cell‐adhesive ECM ligands over cellular monolayers (here, hepatocytes) to prevent forced apicobasal polarization and to create a more in vivo‐like environment for certain cell types.^{[ 95} , ^{103 ]}

Figure 11.

Example “sandwich” culture that mimics the 3D structure of the liver acinus. Reproduced with permission.^{[ 173 ]} Copyright 2018, Royal Society of Chemistry. LECM: liver extracellular matrix, LSEC: liver sinusoidal endothelial cells.

An in‐depth characterization of the membrane and its coating was conducted by Bennet et al., who reported a vertical OoC to mimic the corneal epithelium.^{[ 139 ]} Here, the authors aimed to reconstruct the epithelial basement membrane (felt‐like arrangement of fibers, ≈50 nm in size, and pores of ≈100 nm) and the underlying Bowman's layer (a supporting layer for the corneal epithelium, composed of ≈20 µm thick collagen fibrils with 10 µm pore size), whose physical properties influence the growth of the corneal epithelium. A PC membrane was used to represent the Bowman's layer while a UV‐crosslinkable membrane coating (fibronectin) was designed to recapitulate the epithelial basement membrane. The thicknesses of the uncoated and coated PC membranes were characterized by a non‐contact optical profilometer and found to match the in vivo thicknesses of the Bowman's layer and basement membrane of the corneal epithelium. To control basement membrane stiffness, the authors varied the UV exposure time used to crosslink the fibronectin coating (at a fixed fibronectin concentration). A bioAFM was used in contact mode to characterize the stiffness of both layers, which were found to closely match those of their in vivo counterparts. Moreover, topographical properties such as the fibronectin fiber groove size or the roughness of coated and uncoated membranes were also determined using bioAFM. Interestingly, in this study, Bennet et al. found the stiffness of a commercial untreated PC membrane to be ≈110 kPa in contrast to 2–3 GPa reported by Chung et al., which further emphasizes the need for standardized and consistent membrane/ECM characterization.^{[ 156 ]}

4.2.2. Scaffolds on/as Membranes

In order to better define surface micro‐topography or mimic the physiological 3D environment of the in vivo ECM in vertical OoCs, scaffolds can be used as independent cell culture structures or be integrated on cell culture membranes. A common application area is the recapitulation of the gut epithelium, which is characterized by an enormous surface area folded in microstructures known as intestinal villi and microvilli and crypts.^{[ 174 ]} To be able to mimic such 2.5D features in vertical OoCs, Shim et al. manufactured collagen‐based villi structures from water‐soluble, alginate‐based inverse molds by replica molding.^{[ 175 ]} The produced villi were bonded to commercial porous PET membranes using collagen as glue, and their height and shape were determined before and after cell seeding via confocal microscopy. Alternatively, the membrane itself can be a scaffold. For example, electrospun membranes can be produced with distinct thicknesses, fiber diameters, porosity, and mechanical strength that mimic the properties of in vivo ECM.^{[ 38 ]} Moreover, ECM‐adhesive proteins can be blended into the spun polymer solution, avoiding a subsequent coating step.^{[ 38 ]} Drawbacks of electrospun membranes include sealing challenges for leakage‐free OoC operation, membrane properties being affected during device sealing, and difficulty in controlling pore positioning (Figure 12 ).^{[ 38} , ⁷⁹ , ^{176 ]}

Figure 12.

SEM images of electrospun a–c) poly‐l‐lactide (PLA) and d,e) PLA‐gelatin methacryloylmembranes. Scale bar: 20 µm. Morphology of the membranes before device sealing (a,d). Morphology of the membranes after device sealing, which required a heat treatment for 15 min in a convection oven at 130 °C (b,e). Hydrophilicity of the PLA membrane surface increased significantly with a laminin coating (c). Adapted with permission.^{[ 38 ]} Copyright 2021, American Chemical Society.

4.2.3. Natural Polymer Membranes

Using synthetic polymer membranes to create communicating micro‐compartments in vertical OoCs is the established method adapted from traditional cell culture setups such as Transwells. However, the nature and properties of synthetic polymers often significantly differ from those of the native ECM.^{[ 177} , ^{178 ]} The in vivo ECM comprises a meshwork of random or oriented fibrils that can be remodeled by the cells through enzymatic degradation and ECM secretion.^{[ 179 ]} To better mimic these properties in vertical OoCs, the use of biopolymer membranes of natural origin has been suggested. In this section, we will focus on ECM protein‐based membranes, such as collagen‐ or Matrigel‐based membranes, which attempt to more closely imitate the composition of the in vivo ECM.^{[ 120} , ¹⁷⁷ , ¹⁷⁸ , ¹⁸⁰ , ¹⁸¹ , ¹⁸² , ¹⁸³ , ^{184 ]} These membranes are also commonly referred to as “ECM‐based”, “ECM‐derived,” or “biological” membranes. They are typically produced as standalone membranes via the vitrification (i.e., aseptic drying) of natural hydrogels (most commonly collagen I) and are rehydrated upon integration into devices. As ECM protein‐based membranes cannot be easily bonded to commonly used OoC materials, device assembly techniques include using vacuum to hold the membrane in place,^{[ 180 ]} sandwiching the membrane between two plasma treated pieces of PDMS,^{[ 181 ]} using a PDMS mortar,^{[ 177} , ^{178 ]} using double‐sided tape,^{[ 183 ]} or injection molding of monolithic devices.^{[ 178 ]}

The mechanical, structural, and biochemical properties of ECM protein membranes based on polymers of natural origin are highly interlinked (see Section 2.1.1.). For example, key membrane characteristics such as stiffness, porosity, and permeability are commonly modulated via adjustments in protein composition and concentration, which at the same time affect the availability, spatial distribution, and density of cell binding sites. Moreover, the polymer composition can determine the structural integrity^{[ 178 ]} and thickness^{[ 182 ]} of the membrane, which again necessitates the individual characterization of membrane properties prior to application. To adjust the membrane stiffness independently of the polymer composition, Zamprogno et al. examined the effect of gelation temperature.^{[ 183 ]} In this work, non‐vitrified collagen‐elastin (CE)‐based hydrogel membranes were directly polymerized on a supporting gold mesh under different temperatures. CE membranes that were polymerized at 4 °C were stiffer in comparison to membranes polymerized at 37 °C (≈2 kPa vs ≈0.8 kPa). To adjust the membrane thickness, methods involving gel volume regulation,^{[ 177} , ^{182 ]} stacking of individual ECM‐based membranes,^{[ 177} , ^{178 ]} and enzymatic degradation of the membrane layer^{[ 180 ]} have been reported. One such example comes from Arik et al.^{[ 178 ]} who described the production of ≈2 µm thick collagen membranes using different concentrations of collagen, stacked and “bonded” to each other by a transglutaminase, an enzyme that can crosslink collagen.^{[ 185 ]} This approach allowed the fabrication of membranes that could mimic highly organized ECM layers in vivo. Instead, Puleo et al. created thinner membranes by enzymatically etching away 20–50 µm thick vitrified collagen membranes using a collagenase solution. Etching took place in OoCs from the basal side following device construction, membrane rehydration, and apical epithelial cell seeding. While enzymatic treatment of ECM protein‐based membranes can result in thinner membranes, it can also affect membrane permeability due to the degradation of collagen fibers that disturb the porous membrane structure.^{[ 178 ]}

Last, it is important to note that the properties of ECM protein‐based membranes should be evaluated in a hydrated state as swelling after exposure to aqueous environments affects the internal structure and properties of the membrane. Moreover, ECM protein‐based membranes are likely to show anisotropic material properties due to the heterogeneous distribution and alignment of protein fibrils. This necessitates a discussion regarding methods that could be suitable to characterize anisotropic membrane properties and to evaluate and interpret such data.

4.3. ECM Integration into Planar OoCs with Multiple Compartments

Planar OoCs have multiple microchannels in a single plane (side‐by‐side) that are not physically separated by solid sidewalls. The partitioning between fluid flow and different tissue constructs is commonly realized by the use of microfabricated barriers at channel intersections, referred to as “phase guides” (Figure 13 ). In planar OoCs, the ECM environment is commonly created in two ways: i) by making use of the underlying (coated) microchannel surfaces, as introduced in Section 4.1,^{[ 78} , ¹³⁵ , ^{186 ]} or, more commonly, ii) by employing hydrogels that are restrained in specific microchannels by the phase guides. Such hydrogels, sometimes referred to as “hydrogel membranes” or “hydrogel scaffolds”, are integrated into planar OoCs in a pre‐polymerized form through flow and are allowed to polymerize in the chip.

Figure 13.

Planar OoCs make use of barriers microfabricated at channel intersections, commonly named as “phase guides,” to spatially partition tissue interfaces. a) Channel cross‐section of planar OoCs. While a three‐channel architecture is commonly seen, it is possible to extend the number of channels to incorporate additional tissue interfaces into the chip. b) An exemplary three‐tissue interface in a planar OoC. The cell‐laden ECM gel, usually a natural hydrogel, is loaded into the central microchannel in a pre‐polymerized state and allowed to polymerize in the channel. Phase guides prevent the overflowing of the gel to the neighboring channels. After polymerization, cell seeding into the neighboring channels is realized through fluid flow. c) Natural hydrogels are also incorporated into planar OoCs due to their degradability and remodeling by cells. An exemplary application is the investigation of angiogenesis, which involves the migration and differentiation of endothelial cells (shown as “cell type I”) to form new blood vessels.

Hydrogels can be integrated into planar OoCs before cell seeding to act as vertical substrates that cells are seeded onto (e.g., to create a vascular layer) or in a cell‐laden form. In both cases, hydrogels in planar OoCs serve as a 3D ECM. In order to ensure cell viability and establish biological models that might require ECM remodeling (such as angiogenesis or tumor models), natural hydrogels are most commonly used. The use of natural hydrogels comes with limitations discussed above, mostly related to the interrelation between polymer composition and the mechanical and structural characteristics of the ECM, as well as the spatial distribution and density of cell binding sites.^{[ 25} , ⁵⁶ , ^{94 ]} Moreover, the challenges that are associated with the characterization of 3D matrices are discussed in Section 3. In planar systems, access to the gels upon integration into the system is practically not possible without the destruction of devices, which creates significant challenges in terms of ECM property modulation or characterization. Gel swelling also plays a prominent role as hydrogels polymerized in the systems will swell in a rather uncontrolled manner upon the introduction of cell culture medium.

5. Conclusion and Outlook

In vivo, the ECM is a dynamic, complex, and sophisticated network with tissue‐ and age‐specific biochemical composition, ligand presentation, mechanical, and structural characteristics.^{[ 6} , ^{187 ]} The quest to decipher the biochemical and mechanobiological functions of the native ECM in vitro and imitate them under varying (patho)physiological conditions is a direct function of the biomaterial used, its biochemical and mechanical characteristics, and the degree to which researchers can direct and assess its properties. While cell–ECM interactions are classically studied in traditional coated cell culture glass or plastic dishes, the importance of a suitably designed and characterized ECM is often overlooked in OoCs. This constitutes a great limitation of the OoC technology, which was established and popularized on the potential of creating biomimetic environments.

In this review, we set out to map the current state of ECM integration into OoC environments and identify the bottlenecks associated with it. In our view, controlled ECM integration into OoCs comes down to four interrelated aspects: 1) the materials used to mimic the native ECM; 2) the architecture of the OoC device; 3) the architecture of the ECM (e.g., 2D coating or 3D scaffold); and 4) the degree of ECM integration control and accessibility for characterization. Similar to traditional in vitro cell culture substrates, when it comes to choosing an ECM material for integration into OoCs, it is imperative to consider physicochemical properties as they can play a key role in cell fate and function. Beyond properties, ECM material choice in OoCs also requires careful consideration related to integration and accessibility for characterization. For example, if the ECM environment consists of a PET membrane coated with a hydrogel, the researcher needs to ensure that PET membranes can be tightly integrated into the OoC body and that the properties of the hydrogel will not be affected by the integration process. This depends on the OoC body materials, its architecture, and the fabrication/integration process, which also determine ECM accessibility for property tuning and characterization after device sealing.

This intricate interplay among materials, OoC device architecture, and ECM characterization possibilities significantly complicate the design of ECM‐related studies, the degree of comparability between works, and the level of reproducibility that can be achieved, especially across laboratories. To improve reproducibility and comparability, standardized methods/parameters need to be developed/established for ECM characterization before and after integration into OoC devices. Furthermore, while it is not realistic to use a single ECM‐mimicking material across all OoC platforms, it would be useful to establish a set of matrix properties to be reported in a standardized manner (e.g., stiffness, pore size, and coating thickness) along with the detailed material formulations/compositions and fabrication processes.

To reach the full potential of OoC‐based studies in terms of result relevance, transferability to the human situation, and the establishment of patient‐specific conditions on chip, development and integration of materials that can imitate the full biophysical and biochemical complexity of the native ECM is necessary, for example, dECMs. To develop an in‐depth understanding of the interplay between individual ECM properties and cellular function/mechanotransduction using OoCs, controlled and reproducible ECM property tuning needs to become an integral part of OoC platform design and OoC‐based study design. Here, the use of new materials and technologies, such as nano‐ and micro‐scale printing technologies (e.g., 2‐photon‐polymerization), as well as the use of dynamically‐tunable, well‐characterized synthetic hydrogels could enable the fabrication of better‐defined, reproducible, and accessible matrices. Last, careful consideration and standardization of experiments, as well as detailed methods reporting would significantly contribute to the cross‐examination of results between studies, enabling an improved insight into the collective data.

Conflict of Interest

The authors declare no conflict of interest.

Author Contributions

H.K. handled conceptualization, investigation, methodology, visualization, writing – original draft, and writing – review and editing. E.E.B. handled writing – review and editing. I.C. handled conceptualization, funding acquisition, methodology, project administration, supervision, and writing – review and editing.

Biographies

Hazal Kutluk completed her bachelor's degree in mechanical engineering at the Hacettepe University in Ankara, Turkey (2015), and her master's degree in microsystems engineering at the University of Freiburg in Germany (2018) with a focus on life sciences. She is currently a Ph.D. candidate at the Institute of Microtechnology and the Center of Pharmaceutical Engineering at Technical University of Braunschweig in Germany. Her research interests are focused on developing lab‐ and organ‐on‐chips for applications in life sciences and the investigation of cell and tissue mechanics.

Effie Bastounis is currently a group leader at the University of Tübingen. Prior to moving to Germany, she was faculty/staff at the University of Washington (2018–2021) and postdoctoral fellow at Stanford University (2014–2018). She graduated from the National Technical University of Athens with B.Sc./M.Eng. in electrical and computer engineering and from the University of California San Diego with a M.Sc./Ph.D. in bioengineering, where she made seminal contributions in the cell motility field. Her group work focuses on the development of rigorous and reproducible culture‐based methods to investigate host–pathogen interactions, drawing on biophysical, cell biological, and computational approaches.

Iordania Constantinou is a professor at the Institute of Microtechnology and the Center of Pharmaceutical Engineering at Technical University of Braunschweig in Germany. She graduated with a degree in mechanical engineering from the Cyprus University of Technology (2011), followed by a master's degree (2013) and Ph.D. (2016) in materials science and engineering from the University of Florida. She then worked as a postdoctoral researcher at the Center for Molecular Biology at Heidelberg University until 2018. Her current work focuses on the development of microsystems for applications in the life sciences.

Kutluk H., Bastounis E. E., Constantinou I., Integration of Extracellular Matrices into Organ‐on‐Chip Systems. Adv. Healthcare Mater. 2023, 12, 2203256. 10.1002/adhm.202203256