Engineered extracellular matrices: emerging strategies for decoupling structural and molecular signals that regulate epithelial branching morphogenesis

Abstract

The extracellular matrix (ECM) is a heterogeneous mixture of proteoglycans and fibrous proteins that form the non-cellular component of tissues and organs. During normal development, homeostasis, and disease progression, the ECM provides dynamic structural and molecular signals that influence the form and function of individual cells and multicellular tissues. Here, we review recent developments in the design and fabrication of engineered ECMs and the application of these systems to study the morphogenesis of epithelial tissues. We emphasize emerging techniques for reproducing the structural and molecular complexity of native ECM, and we highlight how these techniques may be used to decouple the different signals that drive epithelial morphogenesis. Engineered models of native ECM will enable further investigation of the dynamic mechanisms by which the microenvironment influences tissue morphogenesis.

Graphical Abstract

Introduction

The dynamic interplay between developing tissues and their surrounding microenvironment is integral to the specification of multicellular form and function. Cells simultaneously secrete new matrix proteins and remodel the existing extracellular matrix (ECM), giving rise to spatial and temporal heterogeneity across multiple length and time scales (Figure 1). Owing to this remarkable complexity, the structural and molecular mechanisms by which the ECM influences the behavior of single cells and multicellular tissues remain to be elucidated. Technical challenges associated with live imaging further limit our understanding of how specific ECM proteins influence the behavior of cells and tissues in vivo. Engineered ECMs [1]―which we define as natural or synthetic materials designed to mimic the structural and molecular properties of native ECMs―will enable further studies of the dynamic reciprocal interactions between tissues and their surrounding non-cellular microenvironment.

Figure 1. Dynamic structural and molecular features of native ECM that may influence branching epithelial tissues.

Schematics of basement membrane thinning around an extending epithelial bud, turnover of ECM proteins, ECM remodeling by a fibroblast, local and global matrix stiffness as well as viscoelasticity, vasculature, ECM topology, and the binding/unbinding of soluble factors within the fibrous architecture of the ECM.

Previous review articles have provided detailed summaries of natural and synthetic ECMs in a variety of biological contexts [2–4]. Here, we provide an overview of recent developments in the design, fabrication, and implementation of engineered ECMs for studies of the morphogenesis of epithelial tissues. We begin with specific examples of the dynamic role that the ECM plays in the development of branched epithelial tissues. Next, we focus on examples of matrices that have been used to mimic the structural and molecular signals provided by native ECM during branching morphogenesis. Finally, we discuss emerging materials, fabrication techniques, and imaging methods that may enable future studies of the dynamic interactions between epithelial tissues and their matrix microenvironments. Although engineered ECMs are currently unable to mimic all of the stimuli provided by the native microenvironment, these modular systems will allow investigations of the role of specific facets of native ECM during branching morphogenesis. These investigations will permit researchers to decouple the roles of structural and molecular signals that regulate tissue morphogenesis.

1. Dynamic signals derived from native ECM during epithelial morphogenesis

Tree-like epithelial structures are generated through a developmental process known as branching morphogenesis. During branching, epithelial tissues undergo recursive rounds of branch initiation, elongation, and bifurcation as they remodel the surrounding microenvironment. In addition to an array of cellular mechanisms, the ECM provides distinct spatial and temporal signals during each stage of morphogenesis that regulate the branching program. Tissue explants or organoids, which are cultured on a porous membrane, submerged in culture medium, or embedded in a three-dimensional (3D) hydrogel, are routinely used to study branching morphogenesis ex vivo [5]. These systems have uncovered the relative roles of specific signals during each stage of the branching process. Here, we describe examples of ECM signals that influence branch initiation, elongation, and bifurcation during normal development of branched epithelial tissues.

1.1. Branch initiation

In the developing mouse submandibular salivary gland, new branche s are generated through the formation of clefts on the basal surface of an epithelial bud. The clefts initiate as indentations, which subsequently progress inward in order to generate new epithelial buds (Figure 2a). Local expression of fibronectin drives cleft formation by decreasing cell-cell adhesions and increasing cell-matrix adhesions [6]. Fibronectin may also promote the accumulation of type III collagen fibrils, which are believed to stabilize the cleft [6]. In addition to ECM accumulation, which can change the mechanical and chemical signals locally, the bulk mechanical properties of the ECM can also affect branch initiation. To test the effects of matrix stiffness on epithelial branches, lung buds from embryonic day 11.5 mice were cultured on polyester culture well inserts or on 4% (storage modulus (G′)~2 kPa) or 8% (G′~30 kPa) acrylated hyaluronic acid hydrogels with decreased stiffness [7]. In this system, branch initiation of the lung bud was inhibited on the softer hydrogel substratum, although it remains unclear how stiffness affects the branching program of the lung. In contrast, a study of mammary epithelial branching morphogenesis found that increasing matrix stiffness from 680 Pa to 860 Pa did not affect the branching pattern but promoted branch initiation by increasing the magnitude of mechanical stress at branching sites [8]. How stiffness in the mammary gland affects the later stages of branching morphogenesis, such as branch elongation and bifurcation, remains poorly understood.

Figure 2. ECM protein expression and localization during branching morphogenesis.

a) Fluorescence images of an embryonic day 12 mouse submandibular salivary gland labeled for fibronectin. Scale bar represent 100 μm. Republished with permission of The Company of Biologists Ltd, from [60]; permission conveyed through Copyright Clearance Center, Inc. b) Fluorescence image of type I collagen around a 3-week-old mouse mammary epithelial end bud. Scale bar represents 50 μm. Reprinted from [15] with permission from Elsevier, copyright (2013). c) Fluorescence image of tenascin-C around a Hamburger Hamilton stage 36 branching avian airway. Scale bar represents 75 μm. Image provided by J.W. Spurlin III. d) Fluorescence image of type IV collagen around a developing epithelial bud in an embryonic day 13 mouse submandibular salivary gland. Scale bar represents 20 μm or 10 μm for the inset. Reprinted from [18] with permission from Elsevier, copyright (2014).

Soluble factors such as fibroblast growth factors (FGFs), which can bind to heparan sulfate in the ECM [9], affect the development of several branched organs. Using mammary epithelial organoids embedded in Matrigel, Zhang et al. found that FGF10 promotes mammary epithelial branch initiation when it originates from a point source (FGF10-soaked beads) [10]. Transforming growth factor-β (TGFβ) is another diffusible protein that forms complexes with the ECM [11] and has been shown to drive the accumulation of type I collagen and chondroitin sulfate [12]. During mammary epithelial branching morphogenesis, TGFβ acts as an inhibitory morphogen that can prevent branch initiation [13, 14]. Engineered ECMs may facilitate investigations of the role of specific temporal or spatial patterns of mechanical and soluble cues during branch initiation.

1.2. Branch elongation

During mammary epithelial branching morphogenesis, the tips of the epithelium, which are bulbous structures known as terminal end buds (TEBs), extend into the surrounding stroma. Brownfield et al. observed aligned collagen fibers around TEBs and suggested that these fibers may dictate the direction of branch elongation (Figure 2b) [15]. However, the underlying mechanism remains unclear because it is difficult to decouple the role of collagen fiber alignment from other stromal signals. In addition, diffusible signals in the stroma, such as FGF2, have been shown to promote the elongation of mammary epithelial branches ex vivo [10]. In the developing embryonic chicken lung, the ECM glycoprotein tenascin-C is secreted by the extending airway epithelium and transported throughout the surrounding mesenchyme as a result of mesenchymal fluidity (Figure 2c) [16]. The distribution of tenascin-C may alter the compliance of the ECM around extending branches in a manner that promotes airway morphogenesis [16].

In addition to signals from the surrounding stroma, the basement membrane, which forms a thin meshwork of proteins adjacent to the epithelium, can also regulate branch elongation. During expansion and extension of the epithelium, the basement membrane is proteolytically degraded by matrix metalloproteinases (MMPs). In mammary epithelial cell clusters embedded in a type I collagen gel, MMP14 was found to play a role in the extension of the mammary epithelium by locally softening the ECM [17]. In the mouse submandibular gland, proteolytic degradation results in the generation of dynamic perforations in the basement membrane at the tip of epithelial buds (Figure 2d) [18]. This degradation is thought to facilitate epithelial expansion by increasing the distensibility of the basement membrane. Testing these hypotheses will require the development of engineered ECMs that allow for spatial control of structural and soluble signals.

1.3. Bifurcation

A bifurcation event occurs when a single epithelial tube splits into two new branches, which may form in-plane or out-of-plane relative to the previous round of bifurcation. Less is known about the mechanisms by which the ECM influences bifurcation of epithelial tissues. During bifurcation events in the developing kidney, fibronectin accumulates between newly formed epithelial buds [6]. The accumulation of fibronectin can act as a physical wedge-shaped barrier that helps split the epithelium into two distinct daughter branches. It has been suggested that a similar mechanism might drive bifurcation of the mammary epithelium during branching morphogenesis [19]. In the developing mammalian lung, ECM and smooth muscle cells have been found to accumulate in the same cleft and neck regions of a bifurcating bud [20]. However, it remains difficult to decouple the relative roles of specific ECM proteins and smooth muscle cells during epithelial bifurcation using conventional ex vivo organ culture.

Whereas tissue explants and organoids have been used to identify some of the ECM signals that drive epithelial morphogenesis, many questions remain about the mechanisms by which these signals affect branching. How are morphogen gradients established in the ECM? How are ECM components dynamically assembled and degraded during distinct phases of branching morphogenesis? How can we decouple the relative roles of the structural and mechanical signals that simultaneously affect epithelial branching morphogenesis? Engineered matrices can serve as powerful tools to identify and decouple the mechanisms by which ECM proteins affect the development of complex epithelial tissue structures.

2. Engineering matrices to study branching morphogenesis

A common strategy for studying branching morphogenesis involves incorporating epithelial organoids or explants into matrices that replicate specific facets of native ECM. These culture models enable systematic investigations of the impact of matrix composition and topology on tissue form and function. These models are also amenable to time-lapse confocal imaging, which is important for uncovering the dynamic role of the microenvironment. In this section, we highlight recent advances in the design and implementation of engineered ECMs that mimic structural or molecular signals found in native ECM.

2.1. Fiber alignment

A common structural feature observed in native ECM is the alignment of fibrous proteins such as type I collagen. As compared to heterogeneous networks of collagen, aligned networks of collagen provide directional information that impacts cell migration [21]. Moreover, aligned networks of collagen provide different mechanical signals to cells owing to alterations in the micrometer-scale stiffness and pore size relative to heterogeneous networks [22].

To understand the role of fiber alignment in epithelial morphogenesis, engineered ECMs with aligned collagen fibers have been generated by casting unpolymerized collagen into a strained elastomeric mold that is relaxed after polymerization [15]. Using this approach, mammary epithelial organoids were incorporated into aligned networks of collagen fibers. Organoids branched in the direction of collagen fiber alignment in a manner that depended on Rac1 GTPase activity (Figure 3a) [15]. One limitation of strained molds is that it is difficult to decouple the effects of collagen fiber alignment, collagen densification, and compression on tissue behavior. To overcome this limitation, recent work with 3D microextrusion printing has demonstrated the ability to 3D print cell-laden collagen-Matrigel hydrogels with aligned collagen fibers (Figure 3b) [23]. Shear and extensional flows generated during extrusion of the collagen-Matrigel ink align self-assembling collagen fibers in the printing direction. Epithelial aggregates that are incorporated within these 3D-printed networks generate branches that extend in the direction of collagen fiber alignment. In addition to type I collagen, the type IV collagen molecules that are present in Matrigel can be aligned during self-assembly by using flow-induced shear stress generated by the flow of media through a microfluidic channel [24]. ECM proteins can also be aligned indirectly. For example, surface topology can orient fibroblasts, which assemble fibronectin fibrils oriented in the direction of cell alignment [25]. While it is unclear if this approach can generate a matrix that is compatible with studies of branching epithelial tissues, cell-assembled matrices may provide a more physiologically relevant microenvironment for branching tissues. In contrast to incorporating cells or tissues into networks of pre-aligned ECM proteins, microfabricated tissues with specified geometry can be incorporated into heterogeneous ECM networks and used to study tissue-ECM interactions. In a recent study, microfabricated epithelial tissues incorporated into a collagen-Matrigel hydrogel were used to study the dynamics of tissue-induced type I collagen fiber alignment [26]. This approach revealed that tissue-induced collagen alignment is primarily a physical process and that the rate of alignment depended on the contractility of the multicellular tissue. Moving forward, microfabricated tissues offer a promising avenue for testing hypotheses about dynamic interactions between fibrous ECMs and tissues with a variety of geometries and cellular compositions.

Figure 3. Examples of synthetic ECMs and their application to control tissue behavior.

a) Schematic showing the alignment of collagen fibers within a strained elastomeric mold (left) and a branching mammary organoid within the aligned network (right). Scale bars represent 50 μm. Reprinted from [15] with permission from Elsevier, copyright (2013). b) Schematic of 3D-printed collagen-Matrigel inks (left) with aligned collagen networks (top right) that direct mammary epithelial branching of embedded cell aggregates (bottom right). Scale bars represent 100 μm. Adapted from Ref. [23] with permission from The Royal Society of Chemistry. c) Frequency map of branching from microfabricated mammary epithelial tissues embedded in type I collagen networks that were crosslinked with 50 mM D-ribose or their non-crosslinked controls. Scale bars represent 50 μm. Reproduced from [8] with permission from Oxford University Press. d) Mesenchyme-free branching morphogenesis of mouse lung explants [30]. Scale bars represent 200 μm. Copyright (2015) National Academy of Sciences. e) Schematic of heparin hydrogel containing hepatocyte growth factor. Reprinted from [38] with permission from Elsevier, copyright (2010). f) Schematic showing the synthesis procedure for hyaluronan cryogels (left) and fluorescent pattern of immobilized EGF in an irradiated region of the cryogel (right). Scale bars represent 100 μm. Adapted with permission from [39]. Copyright (2016) American Chemical Society.

While these studies demonstrate that it is possible to generate aligned ECM proteins ex vivo, uncovering the mechanisms by which fiber alignment influences the branching program of epithelial tissues will require reproducing specific 3D patterns of ECM alignment that are observed in vivo. Aligned ECMs will also need to be fabricated such that they support the morphogenesis of the epithelial tissues that are embedded within.

2.2. Matrix stiffness

Local and global variations in the stiffness of the ECM affect patterns of mechanical stress experienced by epithelial tissues during development. Stiffness can influence cell proliferation and apoptosis [27], migration [28], and differentiation [29], but the role of ECM stiffness in epithelial morphogenesis is not completely understood. By incorporating epithelial tissues and organoids into engineered matrices with tunable mechanical properties, the role of stiffness on the pattern and extent of branching can be assessed. To determine how this mechanical property affects mammary epithelial branching, Gjorevski et al. modulated the bulk stiffness of a type I collagen matrix by crosslinking with D-ribose [8]. Increased branching was observed for 3D microfabricated mammary epithelial tissues embedded within the crosslinked type I collagen matrices (Young’s modulus (E)~860 Pa) as compared to softer uncrosslinked matrices (E~680 Pa) (Figure 3c). In a study of mesenchyme-free branching of the airway epithelium, the bulk stiffness of hydrogels was modulated (soft: E~15 Pa; stiff: E~25 Pa) by varying the ratios of Matrigel and methylcellulose, which revealed that stiffness does not affect the spacing of branches (Figure 3d) [30].

To the best of our knowledge, engineered matrices have not been used to study how spatial patterns of ECM stiffness affect epithelial branching morphogenesis. Nevertheless, the ability to spatially pattern stiffness will be important for decoupling the role of this parameter from other features of the ECM such as fiber alignment, ligand density, and protein composition. For example, during the development of the embryonic lung, contractile smooth muscle tissue mechanically induces bifurcation [20] and lateral branching [31] of the airway epithelium. Smooth muscle-like cells may also drive morphogenesis of the mammary epithelium, although the role of this tissue in branching remains unclear [32]. Smooth muscle cells are highly contractile and stiffer than epithelial tissue [33], which makes it difficult to decouple the impact of mesenchymal stiffness and contractility on epithelial branching. Engineered matrices with light-responsive mechanical properties [34–36] offer an exciting avenue to test hypotheses about the effects of spatial and temporal patterns of stiffness in the regulation of epithelial branching. While it is unclear if these materials are compatible with ex vivo epithelial morphogenesis, the incorporation of branched epithelial tissues or organoids into matrices with tunable stiffness and viscoelasticity offers a promising approach for future investigations.

2.3. Soluble factors

The sequestration of soluble signals such as growth factors, chemokines, and cytokines is a critical feature of the ECM that allows for the establishment of morphogen gradients. Soluble factors can bind to negatively charged carbohydrate chains or specific epitopes within ECM proteins. Upon degradation of the ECM, these factors are released, which can initiate cell behaviors associated with different stages of branching morphogenesis. For example, during branching morphogenesis of the mouse submandibular gland, FGF10 binds to perlecan with high affinity and is released after cleavage of perlecan by heparanase [37]. To replicate this feature of the native ECM, growth factors have been incorporated into engineered matrices. Hepatocyte growth factor, which is known to regulate branching morphogenesis, can be sequestered and released from heparin-based hydrogels (Figure 3e) [38]. In addition to passive binding, spatial and temporal control of morphogen binding and release can be achieved by modifying proteins with a linker group that binds to a matrix in a photoresponsive manner. For example, epidermal growth factor (EGF) modified with iodoacetamide can be photopatterned in 3D within porous hyaluronan cryogels [39] (Figure 3f). One limitation of incorporating proteins into hydrogels using linker groups is that the stability and activity of a protein can be affected by the addition of the linkers [40]. To address this limitation, a recent study demonstrated that linker groups can be added in a site-specific manner using a sortase-tag-enhanced protein-ligation technique [41]. This approach allows growth factors such as EGF and FGF to be incorporated within poly(ethylene glycol) (PEG)-based hydrogels [41]. Recently, Cas12a was used to control the release of small molecules anchored to PEG hydrogels via a single-stranded DNA linker [42]. This approach allows molecules such as fluorophores to be released simultaneously at different rates, which may enable the reproduction of multicomponent concentration gradients observed in vivo.

We are unaware of any studies that have incorporated branching epithelial tissues or organoids into a matrix that is designed to control the spatial or temporal distribution of soluble factors. The ability to precisely control the distribution of soluble signals offers an opportunity to directly test hypotheses about how these signals affect the pattern and rate of epithelial morphogenesis. For example, during airway branching morphogenesis, highly O-sulfated heparin sulfate is found near the basement membrane while heparin sulfate that is low in O-sulfated glycosaminoglycans is found in the mesenchyme near regions of FGF10 expression [43]. The degree of O-sulfation in heparin sulfate is believed to affect the diffusivity of FGF-family proteins, whereby low O-sulfated heparin sulfate enables diffusion and refines FGF gradients. Engineered matrices that permit controlled release of FGF10 may allow for the investigation of the effects of different FGF gradients on tissue morphogenesis. Moreover, engineered matrices could be used to assess the use of patterned soluble factors for tissue engineering applications.

Concluding remarks and outlook

Despite the immense complexity of native ECM, a variety of structural and molecular features have been successfully reproduced ex vivo and used to study cell- and tissue-level behavior. The isolation of matrix proteins for the fabrication of single or multicomponent hydrogels has emerged as a common strategy for mimicking specific facets of native ECM. While this approach has provided insight into the role of individual signals during the formation of branched epithelial tissues, many of the dynamic mechanisms by which the ECM affects epithelial morphogenesis―especially those that involve a combination of physical and chemical modalities―remain poorly understood. Further progress on our understanding of the role of the ECM during branching morphogenesis will require addressing the following challenges:

1. Simplicity versus complexity: what are the minimum features required to mimic native ECM?

Whereas it is widely accepted that the 3D complexity of the microenvironment is important, the features necessary to completely mimic epithelial morphogenesis ex vivo remain unclear. Naïvely, the higher the complexity of the engineered ECM, the more accurately an in vivo process can be reproduced. However, as the complexity of the microenvironment increases, it becomes more challenging to interpret the behavior of the cells and tissues that are embedded within. As engineered ECMs incorporate additional features, it will be important to design modular systems [44, 45] that allow the effects of each to be parsed. To that end, one recent study used interpenetrating networks consisting of mixtures of Matrigel and alginate to control matrix stiffness independently of architecture and ligand density [46]. This study revealed that the composition of the ECM can enhance or nullify the impact of stiffness on the phenotype of human mammary epithelial cells. Moreover, Trappmann et al. developed a dextran-based hydrogel that allows matrix degradability to be tuned independently of stiffness [47]. This study revealed that matrix degradability regulates the mode of endothelial cell invasion. Looking forward, a key challenge will involve the design of engineered ECMs that present both structural features (such as stiffness, viscoelasticity, anisotropy, and pore size) as well as biochemical stimuli (such as the distribution of growth factors) to their constituent cells and tissues. These modular systems may provide deeper insight into how distinct signals collectively regulate the development of branched epithelial tissues.

2. How can we dynamically and independently control ECM signals/features across multiple time and length scales?

Native ECM simultaneously provides a diverse array of local and global signals to epithelial tissues throughout morphogenesis. To understand and decouple the relative roles of interactions between the tissue and the ECM, we need to design matrices that are compatible with ex vivo culture models and that allow us to independently modulate specific signals, such as stiffness and growth factor concentration, in both time and space. However, many engineered ECMs are unable to modulate different signals with this level of accuracy and precision, and are not yet amenable to studies of epithelial morphogenesis in 3D. In future work, tunable photoresponsive materials [48–50] or materials with mechanical instabilities [51] may provide opportunities to manipulate the properties of engineered ECMs over relevant time and length scales. In addition, DNA-crosslinked hydrogels with reversible temporal control of matrix stiffness may allow researchers to understand the role of dynamic stiffness changes in epithelial morphogenesis [52]. However, it remains unclear whether these engineered matrices can support branching morphogenesis of epithelial tissues. Emerging biofabrication strategies such as 3D printing might also enable sufficient control of matrix properties in 3D [53]. Alternatively, 3D printing could be used to control the spatial deposition of materials that allow for temporal control of physical or chemical properties in order to generate the required spatiotemporal complexity. As techniques are able to more accurately and precisely control the properties of engineered matrices, it will also become important to obtain accurate measurements of native ECM and the branching epithelial tissues that interact with them. To that end, Serwane et al. recently developed a technique that uses ferrofluid microdroplets to measure the local mechanical properties of a tissue in vivo with a spatial resolution of ~50 μm and a temporal resolution on the order of seconds [54]. Non-invasive optical methods such Brillouin microscopy [55], which can approximate viscoelastic properties of cells and tissues using inelastic light scattering, might enable further improvements in spatial and temporal resolution of these measurements. In terms of ECM stiffness, a recent study demonstrated that the local stiffness of a cell-laden collagen network can be measured by using optical tweezers to manipulate 4.5-μm-diameter latex beads embedded in the network [56]. In future work, designing an engineered ECM that supports 3D epithelial morphogenesis and that can independently modulate structural and molecular signals in both time and space throughout multiple stages of morphogenesis will be a key challenge.

3. Long-term cell culture and imaging

Branching morphogenesis of epithelial tissues occurs over the timespan of days to weeks in vivo. However, many ex vivo studies of tissue development focus on the early stages of branching because current engineered ECMs can only facilitate short-term investigations of morphogenesis. To understand the dynamic role of the ECM during the later stages of development, longer culture times are needed. A key limitation with many engineered ECMs is that they do not incorporate vasculature needed to sustain tissues for long culture periods. 3D printing has emerged as a powerful strategy for fabricating perfusable vasculature within tissues that can be cultured on the timescale of weeks [57, 58]. However, it remains unclear whether 3D-printed vasculature is sufficient to support the long-term culture and morphogenesis of epithelial tissues. It is also challenging to image live tissues and the surrounding ECM proteins with high spatial or temporal resolution for extended periods of time due to photobleaching as well as phototoxicity that is caused by the generation of reactive oxygen species (ROS) [59]. Imaging techniques that limit light exposure outside of the focal plane, such as light sheet fluorescence microscopy and two-photon microscopy, can reduce photobleaching and phototoxicity [59]. Phototoxicity can also be reduced by adding antioxidants such as ascorbic acid or rutin to the cell culture medium in order to reduce ROS levels [59]. Looking forward, the ability to simultaneously track epithelial tissues as well as the synthesis, degradation, and transport of several ECM proteins in long-term cultures will enable deeper insight into the dynamic role of the ECM during morphogenesis.

Abbreviations

ECM

extracellular matrix

EGF

epidermal growth factor

FGF

fibroblast growth factor

MMP

matrix metalloproteinase

PEG

poly(ethylene glycol)

ROS

reactive oxygen species

TEB

terminal end bud

2D

two-dimensional

3D

three-dimensional

TGF

transforming growth factor