Microfluidics for the study of mechanotransduction

Abstract

Mechanical forces regulate a diverse set of biological processes at cellular, tissue, and organismal length scales. Investigating the cellular and molecular mechanisms that underlie the conversion of mechanical forces to biological responses is challenged by limitations of traditional animal models and in vitro cell culture, including poor control over applied force and highly artificial cell culture environments. Recent advances in fabrication methods and material processing have enabled the development of microfluidic platforms that provide precise control over the mechanical microenvironment of cultured cells. These devices and systems have proven to be powerful for uncovering and defining mechanisms of mechanotransduction. In this review, we first give an overview of the main mechanotransduction pathways that function at sites of cell adhesion, many of which have been investigated with microfluidics. We then discuss how distinct microfluidic fabrication methods can be harnessed to gain biological insight, with description of both monolithic and replica molding approaches. Finally, we present examples of how microfluidics can be used to apply both solid forces (substrate mechanics, strain, and compression) and fluid forces (luminal, interstitial) to cells. Throughout the review, we emphasize the advantages and disadvantages of different fabrication methods and applications of force in order to provide perspective to investigators looking to apply forces to cells in their own research.

1. INTRODUCTION

Mechanical forces modulate biological processes across length scales. Extensive work in animal models has revealed that tissue-scale forces are necessary for development, homeostasis, and multi-organ function. Studies at the cellular length scale have identified molecules and protein complexes, such as focal adhesions [1] and adherens junctions [2], that respond to mechanical force. However, bridging these findings toward a complete model of the multiscale biological response to force has been limited by a lack of understanding of: 1) how organism- and tissue-scale forces are experienced by mechanosensitive elements at the molecular scale, and 2) how molecular mechanotransduction elements and pathways integrate at the tissue scale to govern multicellular processes and organ function. A key limitation in addressing these questions has been the lack of experimental methods that capture the architecture and the mechanical milieu of the in vivo microenvironment, particularly at the multicellular scale.

Here, we review recent advances in microfluidic technology that have been employed to address this experimental gap, and we highlight platforms that have been developed to culture cells and tissues in physiologically relevant mechanical microenvironments. Microfluidic devices further enable precise control over the chemical microenvironment, including the generation of stable molecular gradients, to more accurately recapitulate the in vivo microenvironment and to allow the study of how biochemical factors impact mechanotransduction. Given the recent interest and work in developing microfluidic platforms for cell culture and biological assays, we focus this review on microfluidics that have enabled novel investigation into the response of cells and tissues to mechanical force. We can direct readers to recent reviews that focus on other aspects of microfluidic platform design for biological applications including, microfluidic platforms for the study of cancer metastasis [3,4], vascular biology [5,6], and recapitulating organ-level function on a chip [7,8].

In this review, we broadly define mechanotransduction as a biological response to mechanical force mediated by molecular events, such as protein aggregation or phosphorylation. The forces, or inputs in the transduction cascade, experienced by cells are varied in nature and can be externally imposed (e.g. when a chondrocyte in articular cartilage is exposed to compressive forces due to mechanical loading of a joint [9]) or generated by internal actomyosin-dependent contractility (e.g. when a cardiomyocyte contracts during systole [10]). The biological responses, or outputs in the transduction cascade, span length and time scales from rapid protein folding and binding events [11] to intermediate changes in gene expression and cellular differentiation [12] to longer-term morphogenesis [12] and pathogenesis [13]. Importantly, the biological responses to mechanical perturbation are cell- and tissue-type dependent. For example, fluid shear stress causes blood vascular endothelial cells to align in the direction of flow [14] but causes vascular smooth muscle cells to align perpendicular to the direction of flow [15]. Despite this context dependence at the cellular scale, many of the molecular force sensors and effector pathways are conserved across organisms and cell types. There are a growing number of candidate molecular pathways and cellular signaling elements that were thought to operate independent of mechanics but are increasingly found to respond to force [16,17]. Consequently, we focus this review on mechanotransduction at the sites of cell adhesion, as the related molecular pathways have been extensively studied in microfluidic platforms, and we begin the review with a high-level overview of the elements and signaling pathways involved in mechanotransduction at sites of cell adhesion.

2. MECHANISMS OF MECHANOTRANSDUCTION

A. Cell contractility and mechanotransduction

In this review, we define the mechanical microenvironment to be the host of mechanical signals imparted on cells, including mechanical forces and the mechanical properties of paracellular proteins. Importantly, all living cells are contractile [18], and cells modulate contractile forces in response to applied forces or mechanical properties of the microenvironment [19,20]. Thus, cells exist in tensional homeostasis with the mechanical microenvironment [21], both applying forces to and sensing forces from neighboring cells and extracellular matrix (ECM). Cellular adhesion structures play a particularly important role in mediating this homeostasis by mechanically linking the microenvironment to the cellular cytoskeleton (Fig. 1). The bidirectionality of signaling at cell adhesions (i.e. transmitting cytoskeletal-generated forces to the ECM and microenvironment forces to the cytoskeleton), and reciprocity of tensional homeostasis (i.e. cell contractile forces are, in part, determined by microenvironment mechanics, which in turn are a function of cellular contractile forces) necessitates the study of mechanotransduction in physiologically relevant microenvironments. This need, which is not met by standard cell culture on 2D plastic or glass substrates, has motivated the development of microfluidic platforms described in this review that improve upon standard in vitro cell culture to enable the study of mechanotransduction in environments with physiologically relevant mechanical properties and architecture.

Figure 1.

The mechanical cellular microenvironment. (a) Forces that modulate biological processes via mechanotransduction are indicated with bold text and key mechanical elements of the cell are labeled. Red arrows indicate solid forces while blue arrows indicate fluid forces and flows that activate mechanotransduction pathways. (b) The adherens junction complex modulates mechanotransduction at cell-cell junctions and shares constuients of the adaptor protein plaque with focal adhesions. (c) The focal adhesion complex transmits mechanical signals across the cell membrane and is a key site of mechanotransduction.

B. Cell-matrix adhesions

Integrins, which are the ECM adhesion receptors for adherent cells, transmit cell-generated forces and signals to the microenvironment (i.e. “inside-out” signaling) while also transmitting external forces to the cellular cytoskeleton (i.e. “outside-in” signaling) [22] (Fig. 1). Integrins are key transmembrane components of focal adhesions, which are plaques of protein complexes that establish a direct mechanical linkage between the cytoskeleton and ECM [23]. The aggregation of these adaptor plaque proteins and the assembly of focal adhesions modulates signaling activity of effector proteins [24], including central signaling cascades that regulate diverse cell processes such as cell survival [25], proliferation [26], and differentiation [27]. Focal adhesion assembly is, in part, governed by the magnitude of force across the adhesion complex [28,29]. Because mechanical equilibrium requires that forces applied to adherent cells must be balanced by reaction forces at matrix adhesions, integrins have been implicated as central components in the mechanotransduction of myriad diverse mechanical inputs including the endothelial response to fluid shear stress [30,31], tumor cell response to interstitial flow [32], fibroblast substrate rigidity sensing [33], and cardiomyocyte response to cyclic strain [34], among many others. Microfluidic platforms have enabled the study of many of these mechanical stimuli with high precision, and thus integrin-mediated cell-matrix adhesion signaling has been studied extensively using microfluidics, as detailed below.

C. Cell-cell adhesions

The structure of cell-cell adhesions varies with cell- and tissue-type, and even in cells of a given tissue, multiple adhesion structures can exist between two cells. That cell-cell adhesions are load-bearing was demonstrated by experiments in Drosophila, in which laser ablation of cell-cell contacts caused tissue retraction [35] away from the ablation site. Subsequent work has demonstrated that these cell-cell junctions share some structural and functional similarity to focal adhesions, including the local assembly of plaque proteins that connect to the cytoskeleton in response to mechanical force [36]. Some plaque constituents of adherens junctions and focal adhesions are shared, and there is growing evidence for molecular crosstalk between these adhesion structures [37]. This crosstalk suggests that mechanotransduction at sites of cell-cell adhesion can be modulated by cell-matrix adhesion, and recent data demonstrates that ECM mechanical properties regulate cadherin induction [38]. This integrated adhesion signaling demonstrates the importance of considering the microenvironment when investigating mechanotransduction at cell-cell contacts. Advances in microfluidics and organ-on-chip technology have enabled the culture of collections of cells in physiologically relevant mechanical microenvironments and has provided novel insight into mechanotransduction at cell-cell contacts, as discussed in the following sections.

3. FABRICATING MICROFLUIDIC TOOLS TO STUDY MECHANOTRANDUCTION

The design of a microfluidic device must take into consideration the number and types of cells to be studied, constituents and mechanical properties of the microenvironment, culture duration, mechanical inputs and biological outputs to be investigated, and integration with sensors or imaging systems. Prior to fabricating a particular device, several questions must be considered: how much time and cost is acceptable for preparing each device; what types of materials are compatible with the cells; what is the desired microenvironment; will sensors need to be integrated; does any processing need to be completed after addition of cells; and are there limitations on device size or the number of cells available. Therefore, understanding the biological questions that can be evaluated using microfluidic devices and contextualizing results requires an understanding of how devices and platforms are made. To establish a framework for understanding subsequent sections in this review, we provide a brief overview of microfluidic fabrication methods, including information relevant for designing bioanalytical devices. Strategies for fabrication can be broadly divided into two categories: 1) replica molding – methods in which a master mold is fabricated and copies of the mold serve as the final device for use with cells, and 2) monolithic fabrication – processes that assemble a complete device for direct use with cells. Generally, monolithic fabrication enables rapid prototyping of new designs at the sacrifice of throughput, whereas replica molding enables higher throughput of device fabrication at the expense of greater overhead in time and resources to make the master mold. Below, we provide an overview of the most common techniques used to generate microfluidic devices, emphasizing that this is a rapidly evolving field, and we direct readers to more complete reviews of advanced techniques that have yet to be broadly adopted [39].

A. Photolithography

The most common method for generating microfluidic devices is to fabricate a master mold with photolithography [40] and to fabricate final devices via replica molding with soft lithography [41] (Fig. 2). This process is conventionally referred to as microfabrication, and in subsequent sections, we will use “microfabrication” to refer to the two-step process of photolithography and soft lithography. Devices made primarily of silicon or glass are fabricated using photolithographic techniques developed by the electronics industry. Photolithography generally involves first coating silicon or glass wafers with photoresists, light sensitive polymers that change solubility in response to high-intensity UV exposure [42] (Fig. 2A). Photoresist-coated silicon or glass substrates are then exposed to UV light that is passed through transparency masks to spatially pattern UV exposure and decrease photoresist solubility [42]. Soluble photoresist is then removed in a development process, resulting in wafers patterned with structures determined by the transparency mask geometry and the thickness of the photoresist layer. An etching step removes silicon or glass not protected by the resist to create channels of desired size prior to removal of the resist. More complex steps can include multilevel structures in the substrate and addition of metal electrodes, optical waveguides or polymer photonics. Photolithography has the highest resolution of the techniques used to generate microfluidics and is capable of producing structures with minimum resolvable feature sizes of 250 nm [39]. Even more fine tuning of structure size or surface chemistry can be generated using electron beam lithography, reactive ion etching or ion implantation. However, photolithography is resource intensive, requiring access to expensive equipment for processing silicon or glass substrates [43]. Furthermore, the process is generally limited to the fabrication of layered 2D structures [43], although recent developments have relaxed some of these constraints to enable fabrication of 3D structures [44] or structures with overhanging features [45]. Because of the harsh conditions used for device fabrication, any protein-based materials (such as ECM) and cells must be added after the device is complete; this may be challenging if the only entrance to the device is a narrow channel. Less widely appreciated is the fact that silicon and glass surfaces tend to dissolve during extended contact with salt water [46], especially if they are not properly passivated. However, the high spatial resolution of structures produced using photolithography, the ability to create many devices on a single wafer, and the potential for integrating sensors directly into the devices remain attractive advantages.

Figure 2.

Fabrication methods for microfluidic devices. (a) Photolithography involves using a transparency mask to pattern UV light and resultant crosslinking of photoresist, which is a photosensitive polymer (sample image of a silicon master fabricated with photolithography modified from [45]). (b) Soft lithography involves casting an elastomer on a master mold to create a negative copies of the master for use as a final microfluidic device (example of PDMS device modified from [168]). (c) Embossing is an alternative replica molding approach in which thermoplastic materials are heated and pressed against a master mold to form negative copies (example of thermoplastic device modified from [169]). (d) Laser cutting, milling, and cutting are methods for removing material from bulk plastic to directly fabricate channels or to create a master mold (example of micromilled channel modified from [170]). (e) Lamination involves bonding individual layers with heat and pressure to create multilayer devices (example of multichannel device modified from [68]). (f) 3D printing involves fabricating devices or molds through additive processes. Fused deposition modeling (FDM) involves directly patterning material ejected from a nozzle, and stereolithography (SLA) uses a laser or focused light source and movable stage to crosslink resin and build structures with 3D resolution (sample multicomponent device modified from [171]).

B. Soft lithography

Soft lithography is the process by which microfluidic devices are molded from silicon masters fabricated using photolithography [41] (Fig. 2B). Polydimethylsiloxane (PDMS) is an elastomeric polymer widely used for soft lithography due to its biocompatibility and ability to replicate structures with high fidelity [41]. Methods for soft lithography with PDMS have been widely reviewed elsewhere, and we point readers to a focused review for details of the technique including benefits and drawbacks [47]. In the context of fabricating microfluidics for mechanotransduction, PDMS is cheap, relatively inert, biocompatible, gas permeable, autoclavable, and optically transparent. These features enable cell culture on or in close proximity to patterned PDMS structures [47]. However, PDMS has been shown to absorb small molecules and drugs [48], and biocompatibility with tissue-specific cells remains a concern [49]. Its hydrophobic nature requires surface treatment to facilitate cell attachment, and although its gas permeability allows for oxygen transport to cells, evaporation within the device can greatly impact fluid composition at microliter volumes [49]. Adhesion to other materials can also be problematic, especially if the adhesion has to be performed after cell addition or has to endure in the presence of high internal fluid pressures or mechanical motion. Soft lithography is very popular for creating prototype devices in limited numbers, but PDMS is generally considered a poor choice for manufacturing devices in large numbers due to low throughput and reproducibility. These drawbacks have motivated the development of alternative methods for replica molding, which are discussed below, but the low overhead requirements of soft lithography contribute to its remaining the most commonly used method of producing devices for research purposes.

C. Embossing

A principal motivating factor in the use of PDMS for replica molding is the ability to pour the pre-polymer on to a master mold. Stiffer materials that are more conventionally used for cell culture, including polystyrene (PS), are not amenable to pouring but can be patterned to create channels using embossing methods. These approaches involve heating a thermoplastic polymer above its glass transition temperature and applying pressure against a master mold to form a negative structure [50] (Fig. 2C). A major advantage of this technique is the ability to create devices composed of stiff polymers with better biocompatibility and ease of use for fluid transport compared to PDMS. Examples include cyclic olefin copolymer (COC) [51], poly(methyl methacrylate) (PMMA) [52], and PS [53]. This fabrication process is easily adapted to automated manufacturing systems to decrease device fabrication cost, fabrication time, and human error [51]. The surface properties of these polymers can easily be modified, and they are more resistant to chemical solvents as compared to PDMS [54]. While the resolution of this process depends on the material and embossing parameters, structures 3 μm apart can be generated with PMMA [55], although devices must be made one layer at a time. As far as rapid prototyping, access to a mill or photolithography equipment is necessary for making the mold, but the embossing device can be easily built in most labs. However, this process is not suitable for prototyping, as a master mold must be fabricated before copies can be made. Additionally, most thermoplastics have low oxygen permeability, which can complicate cell culture within the platform [56].

D. Laser cutting & milling

While photolithography enables fabrication of master molds with high spatial resolution, the process of iteratively coating silicon with photoresist and crosslinking through transparency masks imposes limits on the types of structures and the sizes of features that can be manufactured. Furthermore, replica molding requires master molds that allow for the molded devices to be released from the master, further limiting the features that can be fabricated. To overcome these limitations, monolithic methods have been employed, whereby a material is patterned for direct use as a microfluidic device without replica molding. Laser cutting and milling are the two most common methods of monolithic fabrication, and both involve removing material from a block of glass or polymer to create voids or channels [57–59] (Fig. 2D). For laser cutting, the resolution of the cuts depends on the laser spot size, and current methods can create features with a minimum feature size of 700 nm [60]. Laser cutting requires access to expensive equipment and results in elevated temperatures in the etched materials which can negatively affect the precision of the laser cutting process [61]. An alternative method for directly fabricating channels in bulk materials is micromilling, where a drill bit or endmill is used to mechanically remove material [59] (Fig. 2D). Micromilling is the most commonly accessible monolithic method in most research settings, but the minimum feature size of 10 μm [62] is considerably larger than that of laser cutting. Additionally, milled surfaces are rough, which may impact fluid flow and inhibit bonding[63]. Knife plotters are an alternative version of micromilling, where full depth channels are cut out from a material. Although the resolution of knife plotting is only 30 μm, it is a faster and cheaper alternative to milling and laser cutting [64] (Fig. 2D). However, knife plotters can only cut soft polymer films and are incompatible with stiffer, brittle materials [63]. All three methods can interface with computer aided design (CAD) and computer aided manufacturing (CAM) software and systems to improve throughput and attenuate human error. One major disadvantage of laser cutting and milling approaches is that the requirement of access for the laser, drill bit, or knife plotter precludes fabrication of internal channels and structures. Using laser cutting or milling to create individual devices directly can produce high-resolution features, but their primary motivation is for the creation of high value, reusable devices. For this reason, these approaches are often implemented to process hard materials and in applications that can tolerate extended use, heat sterilization, and/or exposure to solvents.

E. Lamination

To fabricate devices with more complicated internal structure, individually fabricated layers can be laminated into one final device (Fig. 2E). Standard lamination methods used to integrate paper or plastic layers into a single device involve using a laser cutter, mill, or knife plotter to remove material in individual layers, stacking the layers, and using heat and pressure or adhesive to meld the layers into a final device [39]. For thermal bonding layers polycarbonate (PC), PMMA, and COC are often used due to their melting properties [51–53]. Thinner layers can be made from plastic sheets with adhesive backings that can be more simply cut and fused. Because lamination is not a standardized process, labs either manually align features and fluidic ports across different layers or create custom alignment tools [63,65]. Additionally, most laminated devices utilize cutting machinery that creates full thickness channels and can only create extensions of 2D shapes. Despite these limitations, lamination remains a widely-used method of device fabrication for multi-layer systems due to its cost effectiveness and quick turnaround from design to device [66]. Lamination is easily performed on soft, flexible materials, allowing for the manipulation of device formation. Recent publications added various modifications to the technique, including folding the fluidic devices to form a microfluidic cube [40], and using paraffin wax and paper to fabricate devices [67,68]. Although lamination requires microfluidic devices to be individually assembled, its low cost and material flexibility allow for creative designs with integrated features for diverse functions [69].

F. 3D Printing

Recent advances in the techniques and materials used for 3D printing have the potential to revolutionize both master fabrication for replica molding and monolithic fabrication. As 3D printing is an additive process, channels are either created from replica molding and bonding in a similar fashion to microfabrication [70] or through the printing of sacrificial support structures that are removed in postprocessing [71]. The most commonly used 3D printing methods for microfluidic fabrication are stereolithography (SLA) and fused deposition modeling (FDM), also known as fused filament fabrication [72]. SLA uses UV light to crosslink resin on a registered stage [71] (Fig. 2F). Rastering the laser provides resolution in two dimensions, while moving the stage provides resolution in the third dimension [73]. While this method is capable of high resolution (50 μm in x, y, and z), uncrosslinked resin cannot be washed out of channels smaller than 400 μm [71]. Another drawback of SLA is that the resin used must be photocrosslinkable, optically transparent, and biocompatible, with biocompatibility remaining a significant challenge with currently available materials [74]. FDM involves melting and extruding a polymer through a nozzle, and by moving the nozzle with time, a device or mold can be fabricated layer-by-layer [72] (Fig. 2F). Using FDM, different materials can be printed within a device, allowing for flexibility of material properties within the device. Through these multi-material approaches, devices with valves, pumps, and sensors have been fabricated[75]-[76]. The resolution of this technique is limited to the size of the nozzle and by the motors and control system that govern nozzle translation [72]. The current minimal resolvable feature size with FDM is 10 μm [77]. Adhesion of printed layers can be challenging when fabricating fluidically sealed chambers [72]. The roughness inherent in the formation of the structures can be transferred if they are used as molds. While directly fabricating devices is appealing for prototyping, most 3D printers currently available are not geared for high throughput manufacturing – the Carbon 3D system being a notable exception [78]. Another development that will be important in the future is the evolution of bioprinters – 3D printers that print proteins and other ECM-type materials as scaffolds for cells in regenerative medicine applications [79,80]. Benchtop 3D printers are designed to be user friendly, and as 3D printers become more inexpensive and the variety of materials available increases, printed devices may become ubiquitous. Collectively, these fabrication methods can be used to create microfluidic tools to study biological phenomena in an isolated system.

G. Hybrid Methods

The integration of multiple fabrication methods has allowed the creation of hybrid systems with complex architectures that broadly extend the capabilities of microfluidic platforms. For example, multimaterial 3D printing has enabled printing structural and biological components from a single printhead. Lin et al. used a multimaterial 3D printing strategy to fabricate a 3D model of the renal proximal tubule, including vascular and epithelial compartments printed in complex, physiologically relevant architectures [81]. By integrating multimaterial 3D printing with milling and soft lithography, Skylar-Scott and colleagues created perfusable, vascularized channels directly in compacted cell aggregates. The cell aggregates behaved as a self-healing material allowing the printing nozzle to move throughout the medium without significant disruption to the cells and enabling direct printing in living tissue [79]. Another emerging strategy for fabricated perfusable channels with complex 3D architecture is to leverage the optical transparency of devices fabricated with photolithography to photopattern hydrogels embedded within a device. For example, Song et al. 3D printed photocrosslinkable hydrogels within a molded microfluidic housing to create spiral channels of varying widths to mimic stenosis [82]. This general strategy of combining 3D printing and soft lithography can be used to create systems with intricate architectures and the benefits of PDMS-based devices including biocompatibility and optical transparency [83]. As the resolution of 3D printers continues to improve and increasingly sophisticated inks become more widely available, the integration of 3D printed components to create hybrid systems will allow increasingly complex geometries to model physiological systems on chip.

4. Mechanotransduction of solid and fluid forces

While the advances in fabrication methods continue to enable the creation of more physiologically relevant microenvironments for multiple tissues, most mechanotransduction studies to date have focused on the cellular response to a single mechanical input or subset of inputs. To reflect current approaches, we divide this section into approaches that investigate the cellular response to solid vs. fluid forces. In vivo, solid and fluid forces are often coupled, but an advantage of microfluidic approaches is the ability to decouple these stimuli to determine key molecular mediators of the cellular response. We consider solid forces as those relating to the insoluble phase of the microenvironment, including ECM mechanics and architecture, substrate strain, and compressive forces. We divide the fluid forces discussion into approaches that interrogate the response of endothelial or epithelial cells to luminal flows and approaches that interrogate interstitial flows in various tissues. We also discuss how advances in fabrication methods allow for the integrating solid and fluid forces to more accurately recapitulate the tissue microenvironment (Fig. 3).

Figure 3.

Advances in fabrication methods will allow the integration of many mechanical stimuli within a single device to more accurately recapitulate the native mechanical microenvironment.

A. Microfluidics for mechanotransduction of solid forces

A1. Substrate mechanics and architecture

Microfabrication and microcontact printing have long been used to make patterned substrates to study the effects of ECM stiffness and geometry on cell function and have been instrumental in defining the role of cell adhesion in mechanotransduction [84–88]. Key studies using these techniques demonstrated that cell area, integrin expression, and actin cytoskeletal structure of endothelial cells and fibroblasts were regulated directly by substrate stiffness [89]. Furthermore, it was observed that stiff substrates shifted the tensional homeostasis between cell-cell and cell-matrix adhesions in healthy epithelial cells towards increased focal adhesion formation and cell contractility, a transition that phenocopied malignant epithelial cells [21]. While these and many other studies demonstrating that cells are sensitive to substrate stiffness were instrumental in defining key mechanosensitive proteins and pathways [11,19], the physiological relevance of the experimental approaches was limited by the use of 2D culture substrates. The advent and development of the fabrication methods discussed above and microfluidics more broadly have enabled the extension of these studies into 3D environments that more accurately mimic the biomechanical and chemical cues that cells experience in vivo.

Cells embedded within tissue are confined in topographically complex 3D porous microenvironments, and cell migration requires navigating through geometric restrictions presented by the ECM. Because microfluidic approaches enable the fabrication of channels with dimensions similar to or smaller than single cells, microfluidics have been particularly effective in understanding how ECM architecture couples with stiffness to modulate cell motility. For example, Hung and colleagues used soft lithography to generate fluidic channels of variable widths ranging from 3 to 50 μm and applied migratory chemokine gradients to demonstrate that distinct molecular mechanisms govern 2D and 3D cell migration [90]. Stroka et al. defined a new mode of cell migration that depends on local osmotic flux, independent of actin dynamics, by culturing cells in microfabricated 3 μm-wide channels to simulate pores in confined environments [91]. In a hallmark study, Denais and colleagues microfabricated vertical pillars to constrict cell migration through pores ranging from 5–100 μm² in cross-sectional area within a microfluidic platform [92]. The authors demonstrated that migration through pores smaller than the nucleus induced rupture of the nuclear envelope and DNA damage, suggesting that cell migration through interstitial pore size can play a role in genomic instability and cancer progression (Fig. 4A) [92,93]. A subsequent study using microfabricated channels with cross-section smaller than the nucleus demonstrated that loss of nuclear membrane integrity is due in part to RhoA-mediated cell generated force and that nuclear membrane damage also contributes to altered cell migration [94]. Collectively, this work, enabled by photolithography and soft lithography, has provided novel insight into the role of the 3D ECM in cell migration and extended understanding of the role of the porous ECM in mechanotransduction.

Figure 4.

Microfluidic devices for applying solid and fluid forces to cells. (a) Patterning fluidic channels with dimensions similar to or smaller than a single cell enables the investigation of how ECM architecture influences cell migration. For example, patterning vertical pillars with gap sizes smaller than the nucleus enabled the observation that tumor cell migration through confined environments can lead to nuclear rupture and DNA damage (modified from [92]. (b) Introducing biomaterials with controlled mechanical properties into microfluidic platforms allows investigation into the interplay of matrix architecture and mechanics. Using hydrogels with controlled degradability in a microfluidic device with elucidated the role of matrix degradation in angiogenic sprouting (modified from [70]). (c) Platforms for applying strain to cell culture substrates have been developed using soft lithography to pattern compliant devices. In a PDMS-based microfluidic model of the lung, endothelial cells were found to align orthogonal to the direction of cyclic strain [100]. (d) By incorporating flexible substrates and applying pneumatic pressure in microfluidic platforms, the effects of compressive forces on cells can be investigated. For example, a multichannel microfabricated device was used to generate pressures up to 15 psi to investigate the effects of compressive forces on the cytoskeleton and nucleus (modified from [172]). (e) Fluid flow through microfabricated channels imparts fluid shear stress on cells cultured on the channel walls. Such platforms have been used to investigate the effects of fluid shear stress on endothelial cytoskeletal dynamics and barrier function (modified from [17,45]). (f) Using microfluidics to apply pressure gradients across hydrogels enables investigation of how fluid forces from interstitial flow impacts cells cultured within the hydrogel. By applying interstitial flow to endothelial cells in a microfluidic device, it was found that flow and VEGF signaling can drive angiogenic sprouting (modified from [162]).

Recent work has integrated biomaterials with tunable mechanical properties into microfluidic platforms to study the interplay of matrix architecture and mechanical properties in mechanotransduction. Seminal work by Burdick et al. used a series of serpentine microfabricated channels to produce hydrogels with gradients in crosslinking and peptide binding motif density within a microfluidic device to demonstrate that endothelial cells preferentially adhere to substrates with higher binding site concentrations [95]. More recent work by Trappmann et al. developed a microfluidic device with a non-swelling dextran methacrylate hydrogel containing matrix metalloproteinase-cleavable crosslinks to enable the modulation of matrix degradability and the generation of molecular gradients (Fig. 4B). The authors found that matrix degradation rate regulated multicellular invasion of vascular endothelial cells in response to a pro-angiogenic gradient [70]. In a similar study by Song and colleagues, photocrosslinkable sacrificial hydrogel microchannels were 3D printed in a number of geometries to further demonstrate that matrix degradability regulates angiogenesis [82]. Recent work has leveraged electrospinning to provide a high degree of control over the physical characteristics of the matrix. By changing the diameters of the polycaprolactone fibers independently of porosity and alignment, the authors found that increasing the diameter of the fibers reduced the permeability of dextran and reduced cancer cell protrusion size [96]. As increasing emphasis is given to studying mechanotransduction on mechanically dynamic substrates, it is likely that microfluidic devices will be developed from dynamic materials such as multi-step crosslinked PDMS [97] to provide relevant mechanical cues to cells in 3D space. Taken together, these studies, enabled by the integration of microfluidics with biomaterials, have provided insights into previously uncharacterized mechanisms of physiologic and pathophysiologic migration phenomena.

A2. Substrate strain

In mechanically active tissues, loading causes strain in the ECM that signals to cells through the same adhesion pathways involved in rigidity sensing [98]. Early studies interrogated the role of strain in various biological processes using bulk assays, including the commonly used FlexCell system [99], but these approaches require expensive, specialized equipment and do not allow simultaneous application of flow. Microfluidic technology and lithography approaches for patterning compliant materials have expanded the understanding of the role of strain in biological processes, particularly in the lung [100], gut [101,102], and cardiovascular system [103,104]. In a seminal study that first reported a now broadly used organ-on-chip microfluidic device including two cell culture compartments separated by a flexible membrane, a lung-on-a-chip platform was microfabricated to model the epithelial-endothelial interface at the alveolus and induce physiologically relevant strains (10% cyclic strain at 0.2 Hz) to both cell types (Fig. 4C). These devices elicited tissue level responses to bacteria, inflammatory cytokines, and nanoparticulates that more closely represent the physiologic microenvironment compared to conventional 2D co-cultures [100]. These devices have also been used to model a number of respiratory pathologies including cancer, asthma, chronic obstructive pulmonary disease, and pulmonary edema to great effect, providing insights into more effective drug delivery mechanisms [105,106].

In other studies, substrate strain was found to modulate physiologic drug-induced pulmonary edema [105], to mediate pathogenic bacterial growth in a gut inflammation model [102], and to improve recovery of injured skeletal muscle tissue [107]. Microfluidic strain models have also provided novel insight into how mechanotransduction in cells harboring pathogenic mutations contributes to disease progression. For example, a microfluidic chip induced uniaxial strains ranging from 9–16% on smooth muscle cells differentiated from patient-derived iPS cells and found that cells from patients with progeria upregulated inflammatory cytokines, integrin expression, and caveolin expression under high pathological strain [104]. Analysis of skeletal muscle repair with and without 10% cyclic strain revealed that physiological cyclic strain improved recovery from a combined oxidative stress and uniaxial strain injury by increasing myosin heavy chain expression and reducing reactive oxygen species production [107]. Collectively, through the use of techniques for patterning compliant, elastic materials, these studies demonstrate that in addition to solid forces and fluid stress, strain is a key activator of mechanotransduction pathways in microfluidic models of physiologic and pathologic settings.

A3. Compression

Despite the importance of loading in skeletal and connective tissue development and function [108–111], considerably fewer microfluidic platforms have been developed to study the role of compressive forces in mechanobiology. A central challenge is the development of platforms that recapitulate physiologically relevant force magnitudes and frequencies. At the length scale of tissues, conventional machine design can leverage higher strength polymers and metals to enable application of force via mechanical actuation [112–115], fluid pressures [116–119], and electromagnetic fields [118,120,121]. These established engineering approaches have enabled the generation of both normal and traumatic forces; however, these systems are restricted in throughput and require the use of large, costly equipment. On the scale of cells, indentation by atomic force microscopy [122,123] has been the most commonly applied method for probing the molecular response to compressive forces. While the resolution of this approach allows for investigation of subcellular structures, there are disadvantages in terms of throughput and the sample size that can be tested.

Microfluidic platforms have emerged as a viable approach for investigating the impact of compression at the cellular and subcellular scales. The most prevalent method of generating compressive forces in microfluidic devices utilizes pneumatic pressure to apply force in devices fabricated using multilayer soft lithography. Briefly, a pliable channel rests between a rigid substrate and elastomeric valve cell. Fluid pressure induces contact stress that deforms the dimensions of the channel cross section; compression on the channel develops as the expanding valve membrane indirectly presses the biological sample into the stiffer substrate [123–127] (Fig. 4D). Using such a platform, Hosmane et al. characterized how neurons respond to axonal injury and described two distinct responses based on the degree of compressive force. Damage caused by loading exceeding 95 kPa promoted axonal regrowth, while lower stress magnitudes stunted neuronal extensions and inhibited axonal regrowth [128]. One advantage of microfluidic devices is that they can integrate with imaging modalities for observation of highly dynamic upstream cell signaling events under compressive load. For example, He et al. leveraged the transparent properties of microfluidic materials and a fluorescence-based sensor of GTPase activity to elucidate the important role of Ca²⁺ currents in the detection of membrane tension changes with compression [129].

Recapitulating physiologically relevant compressive forces on tissues in microfluidic devices is challenging due in part because commonly used polymers such as PDMS fail at pressures exceeding a few MPa. High pressures (> 8 MPa) have been achieved in microfluidic devices, although these platforms focus on micro scale high pressure chromatography or reactor operations [130]. While the design of devices incorporating cells can benefit from these approaches, fabrication with higher strength materials requires concomitant changes in the fabrication processes that may jeopardize the biocompatibility of the final device. It remains to be seen whether high pressure microfluidics will become a staple of mechanobiology research, but the possibility of integration with standard imaging modalities and improved throughput would be beneficial for investigating the role of compressive forces in various biological contexts.

B. Microfluidics that induce mechanotransduction using fluid forces

Fluid transport regulates myriad developmental and homeostatic processes, and misregulation of fluid transport is not only a hallmark but often an underlying cause of the progression of many diseases. While interventional studies in animals have demonstrated that cells are capable of converting signals from fluid flow into biological responses, mechanistic investigation into the key molecular regulators of the response to flow are difficult in vivo because the effects of fluid forces cannot be decoupled from changes in nutrient exchange and other factors. Microfluidic platforms have been effective at addressing this experimental gap to define key physical parameters and molecular pathways that govern the cell response to fluid flow.

In solid tissues, hydrostatic and osmotic pressure gradients drive flow of fluid through luminal structures such as epithelial ducts and endothelial vessels [131,132] (i.e. luminal flow) and through the pores of interstitial ECM [133] (i.e. interstitial flow). When recapitulating physiologic flows with microfluidics, the flow regime is most often characterized by a low Reynolds number, meaning that viscous forces dominate over inertial forces (for a detailed review of hydrodynamics in microfluidics, see [134]). In viscosity-dominated laminar flows, the shear stresses exerted by the fluid on boundary walls are significant and depend strongly on the radius of the channel or vessel through which the fluid is flowing [45]. Interstitial flow, however, involves flow through porous media, and flow is dependent on the viscosity of the fluid phase and the pore size and geometry of the solid phase [135,136]. Below, we review advances in microfluidic technology that have enabled recapitulation of both flow regimes in vitro. These approaches have been used to enable greater understanding of the molecular mechanisms that govern the mechanotransduction generated by fluid stresses.

B1. Luminal flow

Endothelial cells comprise the inner lining of blood vessels and are thus subjected to the mechanical stresses imparted by the flowing blood. The effects of fluid flow on vascular endothelial cells has been thoroughly investigated in vitro using shear rheometry [137,138] and parallel plate perfusion chambers [14]. While these systems have greatly furthered our understanding of the molecular mechanisms that regulate the endothelial cell response to flow [139], they often require large volumes of cells and reagents and are limited in the kinematics of flow that can be applied to cells. Microfluidic devices provide distinct advantages to these traditional assays including low reagent use [140] and parallelization for screening applications [141]. Many microfluidic devices have been developed to investigate the effects of shear stress on endothelial cells, and we direct readers to reviews that cover this technology in more detail [134], while here we focus on more recently-developed platforms that have advanced our understanding of how cell adhesion signaling contributes to the endothelial response to flow.

Mesoscale systems and commercially available platforms for applying flow to endothelial cells require culturing of cells on flat, stiff substrates, which influence the cell-cell and cell-matrix signaling pathways at sites of mechanotransduction. Recently, microfabricated and micromilled devices have been developed that enable the culture of endothelial cells on substrates with more physiologically relevant mechanical properties and architectures. Galie et al. used multi-layer photolithography to create parallel channels of increasing height to spatially vary the wall shear stress magnitude for a single flow rate. Using a two-part soft lithography process, these channels were added above polyacrylamide hydrogels with varying stiffness, and the authors found that endothelial cell alignment to flow depended on the underlying substrate stiffness, with cells on softer substrates aligning at lower magnitudes of shear stress [142]. By integrating a similar platform with live-cell imaging, the authors found that endothelial cells rapidly apply force to the substrate through cell-matrix adhesions in response to flow [143]. Through the use of photolithography and micromilling, these microfluidic platforms enabled novel, quantitative investigations into the role of cell-matrix adhesions in the cellular response to flow, and the results highlight the importance of matrix mechanics and architecture in mechanotransduction of fluid stresses.

A limitation of microfabrication in the study of vascular mechanotransduction is that devices are fabricated in planar layers, and while it is possible recapitulate physiologic substrate mechanics, these approaches do not allow for the culture of endothelial cells in native architectures. To address these limitations, Chrobak and colleagues used microfabrication and small steel needles to create 50 – 100 μm cylindrical channels in collagen hydrogels that could be seeded with cells to form perfusable vessels [144]. In this approach, the use of soft lithography allows precise and repeatable positioning of the needle prior to hydrogel polymerization and fluidic access to the hydrogel after the needles are removed [45]. Replica molding and ease of assembly allows high throughput culture of cells under flow in native architectures surrounded by matrix with physiologically relevant mechanics. A platform fabricated with this approach was recently used to identify that the Notch family receptors are key mediators of barrier function in response to fluid shear stress (Fig. 4E) [17]. Although these needle-based systems have provided insight into the role of matrix mechanics and architecture in the response to flow, they do not allow fabrication of branches or curved structures that are present in native vasculature, and while 3D printing approaches are capable of generating such structures (reviewed [145]), they currently lack the resolution to recapitulate the microvasculature physiologically relevant matrix.

Ductal epithelial cells are subjected to fluid shear stress from flowing glandular secretions and blood filtrate, and microfluidic platforms similar to those developed for the vasculature have been fabricated to investigate the molecular mechanisms of mechanotransduction in many type of epithelial cells (reviewed [146]). Integration of recent advances in the culture of tissue-specific epithelial cells and in the maintenance and differentiation of stem cells with microfluidic technology has engendered the development of microfluidic organs-on-chip. Interestingly, in many of these platforms, fluid shear stress has been found to aid in the development and maturation of organs and organ function. For example, using soft lithography to fabricate microfluidic channels that allowed simultaneous generation of fluid shear stress and cyclic substrate strain, Ingber and colleagues found that shear stress promotes apical-basal polarity and the formation of villi-like structures in microfluidic models of the gut [101,147]. Subtle flows were also found to promote maturation of liver aggregates to increase albumin production [148] and to promote vascularization of kidney organoids [149]. While many of the molecular mechanisms of mechanotransduction remain unclear, these studies, enabled by microfluidics, demonstrate the importance of perfusion and shear stress when modeling epithelial tissues ex vivo.

B2. Interstitial flow

Solid tissues are biphasic, consisting of a solid phase hydrated by interstitial fluid. Hydrostatic and osmotic pressure gradients drive the flow of fluid through the ECM, and this interstitial flow impacts the local transport environment and imparts forces on cells embedded in the ECM. Though interstitial flow is misregulated in several diseases including cancer and fibrosis [150,151], the mechanisms by which cells sense and respond to interstitial flow are difficult to study in vivo due to the low driving pressures and fluid velocities that are difficult to measure and manipulate. Furthermore, recapitulating interstitial flow in vitro requires the application of pressure gradients through 3D hydrogels, which is challenging using conventional tissue culture methods. Consequently, microfluidic platforms have greatly advanced the understanding of how cells sense and respond to interstitial pressure and flow in development and disease and have further defined the key physical parameters that govern cellular responses.

Interstitial fluid pressure is elevated in solid tumors [152], yet the effects of increased interstitial fluid pressure and resulting flow on tumor and stromal cells remains unclear. Informed by seminal work in transwell assays that demonstrated autologous chemical gradients can drive tumor cell migration in response to interstitial flow [153], a microfluidic platform was built to track live cell migration of tumor cells in response to interstitial flow [154]. This system was used to identify a novel mechanism by which fluid stresses imparted on tumor cells promote cell adhesion signaling via integrin activation to direct cell migration independent of chemical gradients [32]. Using a multi-component microfabricated system, Huang and colleagues demonstrated that interstitial flow promoted amoeboid migration, dependent in part on the constitution of the ECM [155]. A multi-part injection molded platform enabled a broader range of interstitial flow velocities to be applied to tumor cells and demonstrated that the tumor cell response to interstitial flow is heterogeneous, with subpopulations migrating upstream and downstream in response to flow [156]. In all of these models, tumor cells were seeded as single cell suspensions within the device prior to applying flow, which does not capture the physiologic tumor microenvironment. To address this limitation, Tien and colleagues adapted the needle-based devices used for investigating the vasculature to apply interstitial flow to tumor cell aggregates and found that cells migrated toward increasing fluid pressure [157].

Beginning with Folkman’s observation that angiogenic sprouting begins predominantly in the venules [158], there has been an interest in whether the transendothelial transport of interstitial fluid plays a role in vascular function or morphogenesis. Early observations into the lymphatic and blood vascular endothelial response to interstitial flow involved seeding cells in collagen gels and applying flow using a custom built meso-scale system. In these studies the authors found that in response to flow, lymphatic endothelial cells formed long cellular extensions, while blood vascular endothelial cells formed branched networks with lumens [159,160]. To more closely represent the in vivo microenvironment, a device was microfabricated by Hernández et al. to investigate the effects of transmural flow on monolayers of endothelial cells, and the authors found that transmural flow guides angiogenesis through increases in the length and degree of neoangiogenic sprouts, processes that depend on the activity of Src kinase [161]. Subsequent work in a multilayer microfabricated platform demonstrated that interstitial flow modulated angiogenesis in response to gradients in vascular endothelial growth factor (VEGF) [162], yet the mechanisms of flow-driven angiogenesis remained unclear. Using a similar platform, Vickerman and colleagues modulated the direction of flow across a monolayer of vascular endothelial cells, demonstrating that flow crossing from the basal to apical side of endothelial cells induces sprouting in a focal adhesion kinase (FAK)-dependent manner, suggesting that integrin-mediated mechanotransduction plays a role in flow-driven sprouting [163]. To further improve physiologic relevance and to remove stiff substrates that were necessary to spatially pattern hydrogels in previous approaches, Galie and colleagues adapted a microfabricated needle-based platform to demonstrate that shear stress between endothelial cells is a key physical parameter in the response to flow and that there exists a threshold above which shear stress induces angiogenic sprouting [164].

Vasculogenesis is the process by which single endothelial progenitor cells differentiate and coalesce to form perfused microvascular networks during development [165], wound healing, and cancer progression [166,167]. Recently, microfluidic platforms have been fabricated to recapitulate elements of vasculogenesis on-chip to enable the formation of perfusable microvascular networks ex vivo (due to extensive work in this area, we direct readers to a recent review [5]). While detailed analysis of mechanotransduction of fluid stresses is difficult due to the complicated and random geometries of the resulting networks, Kim and colleagues developed a multi-compartment microfabricated platform to enable the application of interstitial flow to microvascular networks and found that sprouting vessels require continuous interstitial flow to in the basal to apical direction to grow and maintain newly formed vessels (Fig. 4F) [168]. These platforms hold great promise in understanding the role of mechanical forces in vascular network growth and maintenance during homeostasis and disease.

5. Conclusions and future directions

The importance of mechanical forces on cell differentiation and function are increasingly appreciated. With that understanding comes the motivation to learn more about how to accommodate and manipulate the cell response to physical stresses. Ideally, such analyses will be performed nondestructively, in real time, and with multiple cell types organized in a 3D matrix that recapitulates the normal tissue environment—quite a tall order. Microfluidic devices currently used to measure mechanotransduction in cells are generally composed of PDMS and usually create 2D cell layers with stresses generated either by moving an underlying membrane or controlling the fluid pressures and shear forces. Though some real-time monitoring has been done using imaging of cell morphology, calcium dyes or fluorescent protein expression, most of the analyses involve terminal staining or analysis of membrane receptor or gene expression.

As investigators explore designs for tissue-on-chip systems, they are moving more and more to 3D configurations, either incorporating vasculature with predefined geometry or bioprinting ECM that fosters proliferation and differentiation of specific cell types (Fig. 4). The materials available for devices are expanding to include more biocompatible polymers compatible with low temperature fabrication. Sensor integration with microfluidic devices is moving beyond impedance, temperature, and pressure sensors to include polymer laser diodes and photodiodes for optical sensing; such sensors will facilitate real-time monitoring. CMOS-based imaging is also exploding so that it might be possible to put the “microscope” inside the microfluidic chip in the future. Programmable RFID devices and ultrasonic actuators could replace traditional mechanical actuators to create physical stresses and on-chip pumps could control fluid sheer forces. The opportunities for system integration to monitor mechanotransduction phenomena in cells are only a function of our imagination, combination of diverse expertise, and fabrication resources.