Engineering fiber anisotropy within natural collagen hydrogels

Abstract

It is well known that biophysical properties of the extracellular matrix (ECM), including stiffness, porosity, composition, and fiber alignment (anisotropy), play a crucial role in controlling cell behavior in vivo. Type I collagen (collagen I) is a ubiquitous structural component in the ECM and has become a popular hydrogel material that can be tuned to replicate the mechanical properties found in vivo. In this review article, we describe popular methods to create 2-D and 3-D collagen I hydrogels with anisotropic fiber architectures. We focus on methods that can be readily translated from engineering and materials science laboratories to the life-science community with the overall goal of helping to increase the physiological relevance of cell culture assays.

INTRODUCTION

Inside the body, cells interface directly with a bioactive three-dimensional (3-D) network of fibrous proteins, polysaccharides, and proteoglycans called the extracellular matrix (ECM) (1–3). Cells dynamically remodel the structural components of the ECM in response to biophysical and biochemical cues to help maintain tissue homeostasis (4–6). Through reciprocal (i.e., inside-out and outside-in) signaling, ECM properties also influence cellular activities, including proliferation (7, 8), alignment (9, 10), motility (11–13), and differentiation (14–16). Given the important bidirectional relationship between cells and the ECM, there is considerable interest in creating in vitro scaffolds with tunable properties to support quantitative studies that relate ECM characteristics with cell behaviors.

Type I collagen (collagen I) is a fibrous biopolymer that makes up the largest fraction of collagen in the ECM (17, 18). The organization and structure of collagen I fibers help define the density and stiffness of the ECM and contribute to the structural integrity of the tissue. Collagen I fibers can also remodel into aligned (anisotropic) ensembles under mechanical stimuli, with fibers orienting themselves along the direction of the highest applied strain (19, 20). Anisotropic fibers have significant biological relevance and influence the degree of cell-substrate contact, induce cell polarization (21), and guide cell motility (22–25). Fiber anisotropy is also a defining feature in highly specialized tissues, including muscle (26), cardiac tissue (27), the corneal stroma (28–30), dermis (31–33), and cartilage (34–37). Collagen I anisotropy in tumor samples is also a prognostic indicator of breast cancer patient survival (38). Given the ubiquitous nature of collagen I in the ECM and the physiological importance of fiber orientation, significant efforts have focused on creating collagen I hydrogels with defined fiber microarchitecture.

Acid solubilized collagen that is extracted from a variety of species, including rat, porcine, bovine, and human, can be commercially purchased. Collagen I molecules self-assemble under entropic control when collagen I solution is neutralized and placed under an elevated temperature (typically >25°C). Collagen self-assembly extends across several length scales from nanoscale fibril formation to microscale fibril packing, to macroscale fiber formation (39–41). Fiber properties including length and thickness are sensitive to solution pH, temperature, and ionic strength (42–46), whereas the bulk mechanical properties are a function of collagen I concentration and the degree of cross linking in the gel. Based on the extensive flexibility in the parameter space, natural collagen I is a popular choice to mimic ECM properties in cell-based studies.

In this review, we focus on fabrication techniques that use mechanical manipulation to engineer natural collagen I hydrogels with controlled fiber anisotropy. These methods are divided into two general categories: 1) top-down approaches that apply strain to preformed collagen I hydrogels; and 2) bottom-up approaches that apply strain to collagen I solutions. These methods are summarized in Fig. 1. We also describe the biological applications enabled by these technologies and highlight advancements needed to further improve the physiological relevance of engineered collagen I biomaterials.

Figure 1.

This table provides an overview of the collagen I hydrogel fabrication methods that will be discussed in this review and compares key characteristics between the techniques.

DIRECT APPLICATION OF MECHANICAL STRAIN

Based on in vivo observations that cell-induced traction forces can reorganize fibers into anisotropic domains, several methods to externally apply strain have been explored using top-down and bottom-up approaches (20, 47–49).

Cell-Induced Anisotropy

Cells in vivo have been observed to remodel the surrounding ECM to display anisotropic fiber arrangement. The remodeling of fibers by cells has been attributed to cell-induced strains and directional deposition of collagen precursors in vivo (23, 25, 50–53). The ability of fibroblasts to remodel collagen fibers into anisotropic arrangements has been exploited to create aligned fibers in vitro, using a top-down approach. For example, Ray et al. (54) demonstrated that a rectangular collagen gel that was restrained by anchoring it to polyethylene spacers was remodeled by murine mammary tumor fibroblasts over a period of 2 days to create anisotropic collagen I fibers through lateral compaction, as seen in shown in Fig. 2.

Figure 2.

MDA-MB-231 cells seeded on decellularized matrices that are formed by fibroblast compaction over a period of 1–2 days. Compacted gels display fiber anisotropy and are able to induce elongation in cells, scale = 50 µm. [Reproduced from Ray et al. (54) with permission from Elsevier.]

The resulting anisotropic matrix was then decellularized using a lysing solution of Triton X-100, EDTA, and tris-HCl. Migrating breast cancer cells (MDA-MB-231) were found to exhibit increased directionality, persistence, and speed along the direction of fiber alignment. In addition to remodeling prefabricated collagen gels (top-down approach), the ECM deposited by fibroblasts has also been used to create substrates for engineered corneal stroma (55). In a recent study, Couture et al. cultured corneal fibroblasts for 35 days in fibroblast growth media with 50 mM ascorbic acid to upregulate the production of collagen by the fibroblasts. The deposited ECM formed sheets of aligned collagen that could be lifted off from the surface of the well and stacked to form a thicker, engineered corneal stroma. The inclusion of molecular crowding agents such as carrageenan prevented collagen precursors from being washed away by the media and promoted collagen fibrillogenesis and increased matrix deposition by 16 times (56). Fibroblast deposited collagen has been lifted off the surface on which it is deposited by using temperature-sensitive coatings and cross linkers (57, 58). Although cell-based techniques have been popular for tissue engineering applications and have resulted in anisotropic collagen I, they cannot precisely control the degree of anisotropy and require multistage processing steps. To overcome these limitations, alternative top-down strategies have been developed to apply strain directly to collagen gels.

Mechanical Strain Applied to Preformed Collagen I Hydrogels

Fiber anisotropy has also been achieved in collagen I hydrogels through the application of external (noncell-induced) strain (59). Collagen I hydrogels (0.5–4 mg·mL⁻¹) were allowed to self-assemble in a glass-bottomed dish (20 mm in diameter and up to 2 mm in height). After collagen self-assembly, two needles were inserted 1 cm apart as shown in Fig. 3 and were moved apart at linear velocities ranging from 0.125 to 12.5 µm·s⁻¹, corresponding to strain rates of 2.5 × 10⁻⁵ to 2.5 × 10⁻³ s⁻¹. These strain rates were similar to values found with in vivo fibroblast-induced remodeling (60). Mechanical strain was measured as the ratio of the distance moved by the needle (ΔL) to the starting gel length (L). Cyclic strain ranging from 4% to 16% was applied, and the resulting collagen I anisotropy was visualized using confocal reflectance microscopy (CRM). At higher strains (12% and 16%), anisotropy was seen along with a corresponding increase in fiber density (number of fibers per unit area). The density increase was attributed to the compression of the material normal to the applied strain direction. Fiber anisotropy was not observed at low strain (5%). Interestingly, induced anisotropy in noncross-linked networks was irreversible when applied strain exceeded 10%, whereas anisotropy in glutaraldehyde cross-linked gels was reversible below 15% strain. The authors hypothesized that the sliding fibers in noncross-linked hydrogels caused a plastic deformation in the gel after the threshold strain, whereas the chemical cross linking provided a restoring force that allowed the network to return to its prestrained state. This technique was also adapted to create anisotropic collagen substrates for guiding neuronal growth and extension (61). Cyclic strain to align collagen I fibers was also demonstrated in a later study by Nam et al. (62), who tested the effect of strain frequency, the duration of strain, and the amount of strain, on collagen fiber alignment. The authors reported achieving peak fiber alignment when the gels were exposed to a strain of 50 cycles·min⁻¹, for 12 h and 15% strain.

Figure 3.

The top panels show collagen fibers (black) aligned after being strained. The magnitude of strain is shown in the top right of each panel. Fiber density is higher in strained gels due to lateral compaction of gels, and alignment increases with increasing strain. Graphs in the insets show the fiber angle distribution from 0° to −180°. Images captured using confocal reflectance microscopy, scale bar = 50 µm. [Reproduced from Vader et al. (59) under a Creative Commons Attribution 4.0 International License.]

Mechanical Strain Applied to Collagen I Solutions

Fiber anisotropy has also been achieved by applying direct mechanical strain to collagen I in the solution phase through a carrier substrate. Manipulation of the collagen solution can be seen as a “bottom-up” approach to generating anisotropy. Kim et al. (63) developed an anisotropic collagen I network by depositing a neutralized collagen I solution at a concentration of 2.5 mg·mL⁻¹ onto a prestrained sheet of polydopamine-coated polydimethylsiloxane (PDMS).

The elastomeric PDMS sheet and collagen I solution were maintained in the strained position for 30 min (i.e., holding time) at room temperature and self-assembly was completed in the relaxed (zero strain) position at 37°C. The resulting fibers were oriented perpendicular to the direction of strain. The functionalized PDMS sheet could be either stretched or compressed before collagen solution deposition to induce anisotropy along either axis. The degree of anisotropy plateaued above 30% strain, with an orientation index (OI) of 0.5 (0 = isotropic and 1 = anisotropic). The holding time also had an effect on anisotropy, with maximal values observed between 5 min and 15 min. As shown in Fig. 4, the OI could be tuned as a function of applied strain. As the aligned collagen hydrogel was directly created on a PDMS sheet, the authors integrated microfluidic guidance channels loaded with cells directly into the PDMS-supported collagen I hydrogel to create a 3-D in vitro model of a mouse hippocampal neural network. They observed that embryonic neurons (E18.5 cells) displayed alignment on the collagen networks, and the alignment of the cells was found to match the OI of the underlying fibers.

Figure 4.

Panels showing collagen fibers aligned by direct application of strain during the self-assembly phase. The magnitude of applied strain is defined as the change in length (ΔL) over initial gel length (L) and is indicated above individual panels. The false coloring of the fibers indicates the angle of the fibers and corresponds to the colors seen in the angle distribution plot to the right of each panel. Images were captured using fluorescence microscopy. Scale = 50 µm. [Reproduced from Kim et al. (63) under a Creative Commons Attribution 4.0 International License.]

A similar technique was used to create a collagen I hydrogel with varying degrees of anisotropy to establish 3-D vascular networks (10). The endothelial cells responded to anisotropic fibers by forming longer microvessels and more lumens compared to non-strained isotropic controls. The authors also found that the deposition of basement membrane proteins, including collagen IV, was upregulated with an increase in fiber anisotropy. Other studies have demonstrated the use of anisotropy to guide smooth muscle cell and fibroblast alignment (64) and to explore the effect of contact guidance on vasculogenesis, where the alignment of fibers promoted the formation of endothelial networks and enhanced endothelial cell division (10). The authors hypothesized that pre-alignment of collagen I hydrogels mimicked the clustering of collagen fibers by endothelial cells as observed in vivo. Anisotropic collagen I hydrogels have also been used to enhance the persistence of invasive tumor cells by providing topographic guidance cues that mimic in vivo tumor-associated collagen signatures (11, 65).

The direct application of mechanical strain in both top-down and bottom-up approaches is a simple and relatively quick method to introduce anisotropy into collagen I hydrogels. The degree of fiber alignment can be directly controlled as a function of the strain applied to the gel or precursor solution. Collagen I fabricated by the application of direct strain creates uniform alignment throughout the entire hydrogel and is well-suited for applications that study cell motility (65), alignment (61), and differentiation (66). The resulting anisotropic hydrogels are also fabricated in an open format and can be directly used in cell culture applications. One disadvantage is that preformed collagen I hydrogels and collagen I solutions come into direct contact with external components such as actuators or needles that can introduce contamination if particular care is not taken.

Mechanical Strain Induced by Magnetic Fields

Anisotropic collagen I fibers can be achieved through exposure to magnetic fields (67–69). This method requires exceptionally high magnetic fields (as high as 12 T) and has not found widespread use due to safety concerns around sensitive laboratory equipment. A more accessible magnetic technique was demonstrated by Guo and Kaufmann (70) who incorporated surface-functionalized paramagnetic beads (1.5–2.5 µm in diameter) into a neutralized collagen I solution, which was then pipetted into a cavity. The solution was allowed to self-assemble next to a small magnet with a 2-G magnetic field. The motion of the paramagnetic beads along the magnetic field gradient applied local strain to the self-assembling solution and created fiber anisotropy in collagen gels with concentrations ranging from 1 to 3 mg·mL⁻¹. Interestingly, streptavidin-functionalized beads were found to induce the highest anisotropy in collagen gels, as seen in Fig. 5; the authors hypothesized that the interaction between the moving beads and the self-assembling gel was enhanced by the surface coating. In another study, anisotropy was created in hydrogels up to 3 mg·mL^–1 using 100-nm diameter beads and a larger magnetic field of 255 G (71). Anisotropic collagen I fabricated using the motion of magnetic beads has been extensively used to study cancer cell migration in a tumor-mimetic microenvironment (22, 72).

Figure 5.

Confocal reflectance microscopy (CRM) images showing collagen I fibers with paramagnetic beads aligned under an applied magnetic field. A: collagen I without beads under an applied magnetic field. B: collagen I with unmodified beads. C: collagen I with streptavidin functionalized beads. D: collagen I with amine-functionalized beads. Collagen concentration was 2.5 mg·mL⁻¹ in all cases. Scale = 15 µm. [Reproduced from Guo and Kaufman (70) with permission from Elsevier.]

The self-aligning property of magnetic nanoparticles under a static magnetic field has been exploited to fabricate ECM fiber mimics within a Matrigel environment (73, 74). ECM proteins including tenascin, fibronectin, and laminin were conjugated to magnetic nanoparticles and mixed with a solution of Matrigel. Upon the application of a magnetic field, the beads assembled into high aspect ratio fiber-like ensembles with high anisotropy. The authors demonstrated that cells responded to contact guidance cues provided by the “fibers” similar to the behavior expected with natural fibers. They also used this approach to elucidate the roles of integrins, fascin, and myosin in sensing anisotropy and directing cell protrusions. Nanoparticle-based ensembles were also achieved in a solution of hyaluronic acid and plain culture media, although a continuous magnetic field was necessary to maintain the fiber alignment. Although the authors did not demonstrate control over native collagen fibers in their study, this technique is a practical and powerful approach to create ECM-coated fiber-like structures within a hydrogel of tunable stiffness. The use of this technique can be potentially expanded for the incorporation of aligned ECM-mimetic fibers within a native collagen hydrogel, or to create collagen-functionalized fibers within a different hydrogel environment.

The magnetic methods described in this section are robust and can be easily implemented with minimal equipment requirements (namely, paramagnetic beads and a small rare-earth magnet) and are thus very popular in practice.

FLOW-BASED METHODS

Microfluidic systems (i.e., channel networks with critical dimensions in the submillimeter range) are rapidly gaining popularity for in vitro cell culture studies (75–79). The key advantages of microfluidic systems include precise control over fluid flows, well-defined transport kinetics (due to the relative importance of diffusive transport), and reproducible experimental conditions (75, 80–82). As a consequence of miniaturization, flows in the channels tend to be laminar and flow profiles can be accurately predicted using analytical or computational methods. Laminar flows are characterized by the Reynolds number less than 2,100, with $Re=\frac{\rho uL}{\mu },$ ρ = density of fluid, u = velocity, L = characteristic length, and µ = fluid viscosity. The dimensions in microfluidic systems ensure the Re ∼ 1 (i.e., Stokes flows) and the stresses and strains on the fluid elements can be determined experimentally.

Solution viscosity is an important parameter in fluid flows and in collagen I solutions the viscosity changes rapidly during gel self-assembly. The viscosity of collagen I solutions before self-assembly has also been shown to be highly sensitive to changes in the solution concentration and temperature (83). The unfolding of long polymer chains also causes rate-dependent changes to the solution viscosity (84). To overcome nonlinear viscosity effects in collagen I solutions, flows can be characterized by viscosity-independent parameters, namely shear rate and the extensional strain rate. The shear rate (or the velocity gradient perpendicular to the flow direction) and extensional strain rate (the velocity gradient parallel to the flow direction) are important design parameters to engineer anisotropy into collagen I gels.

Microfluidic Methods

The use of microfluidic flows to create anisotropic collagen gels was first demonstrated by Lee et al. (85) using lithographically defined straight PDMS channels 40–100 µm in height and 10–400 µm in width. The authors pulled a 1.5 mg·mL⁻¹ neutralized solution of collagen I into the channels by applying negative pressure to the channel outlet, with a resulting fluid flow velocity of 5–10 mm·s⁻¹. Fiber anisotropy decreased as channel size increased and their results indicated a quantitative relationship between shear rate and collagen I anisotropy. The resulting gels were cross linked using UV, and the PDMS guiding channel was removed to enable direct access to the formed gel for cell culture applications.

Using a multilayer microfluidic device with guiding flow streams on the top and bottom, a collagen I solution (pH = 2, concentration = 5 mg·mL⁻¹) was flow focused between the steams to create a collagen sheet up to 1-cm wide and up to 5.2 µm in thickness (86). The collagen I sheet displayed a high degree of anisotropy and could be handled without the need for a carrier substrate. The ultimate tensile strength (UTS) was 2.7 MPa in the direction of alignment and 1.4 MPa in the orthogonal direction, with an elastic modulus up to 35.9 MPa. The collagen sheets were then strained (similar to the previous section) to achieve a UTS up to 5.3 MPa, with failure at 9% strain.

The effect of increasing shear rate on the deposition of fibers on a glass coverslip was studied by injecting collagen I solution into a gasket-based microchannel at shear rates ranging from 9 s^–1 to 500 s^–1 (87). The authors found that lower shear rates enabled the deposition of longer collagen I fibers while higher shear rates produced smaller curved (hooked) aggregates. Fiber anisotropy was analyzed using fast Fourier transform analysis as described by Sander and Barocas (88), with values ranging from 0 to 1 for completely isotropic and anisotropic fiber distributions, respectively. The authors hypothesized that the lower shear rate (9 s⁻¹) allowed more time for collagen molecules to diffuse to the surface of the glass coverslips, compared to higher shear rates (e.g., faster flows). A shear rate of 500 s⁻¹ resulted in fibers that were half the length of those formed at 9 s⁻¹. The anisotropy index was significantly lower for fibers deposited at a high shear rate of 500 s⁻¹, and the fibers did not display the characteristic periodic D-banding pattern typically found in collagen. Other studies have utilized shear flows to create aligned collagen fibers for studying corneal cell migration and spreading (89, 90).

As shown in Fig. 6, 2-D coatings were demonstrated by Lanfer et al. (91), who used shear flow in a microchannel to deposit collagen I fibers (0.2–0.8 mg·mL⁻¹) on a coverslip coated with poly(octadecene-alt-maleic-acid) (POMA). The desired solution was injected into a microchannel through a tube with an in-line heater to “precondition” the collagen before injection at shear rates of 8.3, 73.8, and 203 s⁻¹. The collagen I deposition was tested in solution form and using preformed fibers. Higher collagen I concentrations resulted in the adsorption of longer fibers on the coverslip with an increased density (i.e., number of fibers per unit area), which could be varied as a function of preconditioning time. Fully formed fibers were found to adhere better to the POMA coated coverslip under a slower flow rate but exhibited a loss in anisotropy. The authors reported the highest degree of anisotropy at a flow rate of 11 µL·min⁻¹, corresponding to a shear rate of 203 s⁻¹. This approach was also used to induce mesenchymal stem cell (MSC) differentiation into osteogenic, adipogenic, and myogenic lineages. The morphology of mouse muscle myoblast (C2C12 cells) myotubes on isotropic and anisotropic collagen I is shown in Fig. 7 (92).

Figure 6.

Images showing the alignment of collagen fibers induced by fluid flow and deposition of fibers on a Poly(octadecene-alt-maleic anhydride) coated coverslip. Neutralized collagen type I (concentration indicated above panels) was injected through a tube that was heated using an inline heater at a flow rate of 11 µL·s⁻¹, corresponding to a shear rate of 203 s⁻¹. The collagen was allowed to heat in the tube for 5 min before injection. Top row shows confocal reflectance microscopy (CRM) images and lower row shows atomic force microscopy images of the corresponding sample. [Reproduced from Lanfer et al. (91) with permission from Elsevier.]

Figure 7.

The morphology of C2C12 myotubes on isotropic and anisotropic collagen I fibers. Fluorescence imaging of MHC (red), nuclei (blue). A: cells cultured on random collagen for 7 days. B: cells cultured on aligned collagen for 7 days. Angle distributions for myotubes corresponding to random (C) and aligned (D). Scale = 120 µm, inset 30 µm. [Reproduced from Lanfer et al. (92) with permission from Elsevier.]

Extensional flows, characterized by their Weissenberg number (defined as $Wi=\dot{\mathit{ϵ}}\tau$, where $\dot{\mathit{ϵ}}$ is the extensional strain rate and τ is the relaxation time of the polymer chain), have also been used to create anisotropic collagen I hydrogels. Paten et al. (93) extended a needle from the surface of enriched collagen I droplet (concentration = 14 mg·mL⁻¹) and showed that extensional flows with Wi ≥ 1 could induce fiber self-assembly in the direction of applied strain.

The use of extensional flows was also used in a microfluidic format to create gradients in fiber anisotropy using a channel network with varying cross-sectional areas (94) (see Fig. 8). The device, consisting of an inlet region, a constriction, and an expansion channel, was used to generate a velocity gradient in the direction of flow that rapidly dropped off as the channel geometry increased according to the relation V = QA⁻¹, where V is the average velocity, Q is the flow rate, and A is the cross-sectional area. The resulting extensional strain rates from the end of the constriction into the expansion region decreased from 100 s^–1 to 30 s^–1 and produced a spatial gradient in collagen I anisotropy that was proportional to the local strain rate. Anisotropy was quantified as the coefficient of anisotropy (CoA), defined as the fraction of total fibers that were aligned ±15 in the direction of flow. With a changing extensional strain rate, the authors demonstrated gradients of CoA ranging from 0.74 to 0.35, over a 700-µm distance.

Figure 8.

A: schematic showing the top-down view of the microfluidic channel fabricated in Ref. 94. Channel was prefilled with collagen I from port 3. B: collagen I is flowed at a rate of Q = 500 µL/min from port 2, whereas ports 3 and 4 are closed as shown. Direction of fluid flow is indicated by arrows. C: yellow box in the schematic shows the imaged region of interest. D: confocal reflectance microscopy image showing a collagen gel with an alignment gradient is shown. The collagen fibers on the left side of the image are aligned, with a coefficient of alignment (CoA) of 0.74. The CoA value gradually decreases (due to a decrease in the strain rate experienced by the flow) to 0.37. CoA is defined at the fraction of fibers in a region that lie within ± 15° of the major axis of fiber alignment. Yellow dashed line delineates the region of anisotropy (CoA >0.5) from the region of anisotropy (CoA <0.5). Scale bar = 100 µm. [Reproduced from Ahmed et al. (94) with permission from John Wiley and Sons.]

Another technique to create anisotropic collagen I domains was based on injecting fully formed collagen I fibers into a microfluidic channel containing uniformly spaced pillar features. The collagen I fibers were preformed as bundles by rapidly bringing a cold neutralized collagen solution to physiological temperatures through the addition of water. The collagen I bundles were then injected into the microfluidic channel where the flow path was defined by the micropost distribution; the fibers formed local regions of anisotropy with defined curvature as seen in Fig. 9. MDA-MB-231 cancer cells exhibited significantly higher migration velocity and directionality on anisotropic collagen I compared to isotropic collagen I. To form 3-D tumor models, collagen I bundles were mixed with agarose to control the stiffness of the resulting hydrogel. Anisotropic collagen I fibers were able to induce invasive behavior in the MDA-MB-231 cells as characterized by longer protrusions and increased proliferation. As collagen I fiber properties are sensitive to the self-assembly conditions, in particular to pH and temperature (42, 45, 46, 96), the formation of fibers before injection can minimize experimental variability in situations where precise control over microstructure is required. This platform provides a distinct advantage by preforming collagen I fibers and using defined fluid flows between microposts to control collagen I anisotropy.

Figure 9.

A: schematic of device developed by Gong et al. (95) for fabrication of collagen bundles. B: collagen fibers wrapped around pillars of different shapes. Fiber angle distribution histograms can be seen below each panel. (Scale = 100 µm). C: DIC and fluorescently labeled images of MDA-MB-231 cells oriented along the aligned collagen bundles. (Scale = 200 µm). [Reproduced from Gong et al. (95) with permission from Elsevier.]

2-D and 3-D microfluidic techniques, although more complex than direct strain-based approaches, offer a great deal of versatility in engineering collagen fiber anisotropy (63, 86, 90, 97) by enabling excellent control over the degree of alignment and integration of a 3D gel or 2-D coating into a fluidic system. As microfluidic channels are a popular choice for in vitro tissue on chip systems, the direct integration of engineered collagen matrices into a channel adds a dimension of versatility to the already-existing toolkit of microfluidic cell culture assays (98–100).

Spin Coating

Spin coating, or the deposition of a thin layer of neutralized collagen I solution on a rotating substrate, has also been used to create fiber anisotropy. Aligned collagen fibers were produced by Saeidi by introducing 2.6 mg·mL^–1 collagen I at varying flow rates (0.3 to 1 mL·min⁻¹) through a heated tube onto a spinning substrate (750–3,000 rpm) (101). At 750 rpm, collagen I fibers were loosely arranged in arrays containing smaller fibers, and no discernible anisotropy at observed at 3,000 rpm. However, highly anisotropic fibers were observed at 1,000 rpm. This technique was also used to develop collagen I constructs with layers of fibers arranged orthogonal to the layers above and below, by the use of an offset disk setup. The collagen I film thickness was controlled by changing the rotational speed and the anisotropy varied as a function of the radial distance from point of injection. Spin coating is a favorable choice for creating coatings of fibers on large, flat substrates and where spatial changes in anisotropy are desired.

Extrusion

Extrusion is a rapidly growing technology for fabricating ECM materials with complex shapes (102, 103). Although bioprinting has been used to create a wide variety of shapes using materials such as gelatin (104), and polyethylene glycol (PEG) (105), collagen I offers desirable properties for tissue engineering such as low antigenicity and immunogenicity, and an in vivo like fibrous microstructure that presents native binding motifs to cells (17). Recent bioprinting approaches have demonstrated control over collagen I anisotropy by exploiting principles of mixed shear and extensional flows. In general, a collagen I printing platform features collagen I bioink (which can be a blend of collagen I, Matrigel, or other materials to impart structural integrity), an extrusion nozzle, and a translational stage. Fiber anisotropy, hydrogel thickness, and density can be tuned by changing the concentration and composition of the bioink, modulating the temperature, and tuning extrusion parameters such as the nozzle diameter, flow rate, and translational speed.

One of the early demonstrations showed that collagen I fibers could be aligned by extruding an acidic, concentrated collagen solution (30 mg·mL^− 1) from a 22-gauge needle at a flow rate of 3.2 mL·min⁻¹, while the needle was being translated across a glass substrate at a velocity of 340 mm·s^–1 (106). The substrate was immersed in a 10× solution of phosphate-buffered saline (PBS, pH = 7.4) to promote rapid neutralization and self-assembly of the collagen I strand. The attachment of the collagen to the glass coverslip likely aided the alignment process by imposing extensional strain as the needle rapidly translated and resulted in anisotropic collagen I. To produce collagen without fiber alignment, the needle was moved at a slower rate of 50 mm·s⁻¹ with a flow rate of 0.2 mm·s⁻¹. The authors also showed that fiber anisotropy could promote endothelial cell migration and induce cell elongation along the axis of fiber anisotropy. In a microfluidic extrusion setup, Haynl (107) and colleagues showed that collagen I with tensile strength up to 400 MPa could be fabricated by flow focusing an acidic collagen I solution coaxially with a hydrophilic poly(ethylene glycol) guiding stream to form a single collagen I strand. The extruded collagen I was released into a bath of neutralizing solution at neutral pH to complete self-assembly. In a later study, a collagen-based bioink containing a nonionic triblock polymer (3–6 mg·mL⁻¹) collagen and 40%–60% w/v Pluronic F-127, was extruded through a syringe needle heated to 37°C (108). The heated needle initiated collagen self-assembly before the collagen fibers were deposited and structural integrity was provided by the added triblock polymer.

To further control collagen I microstructure, Nerger et al. (109) defined interfaces between regions of anisotropy by varying nozzle size, print speed, flow rate, collagen ink composition, and substrate hydrophobicity. A Matrigel/collagen I bioink with 5.0 Matrigel and 0.8 mg·mL⁻¹ collagen was found to retain its shape during the printing process. Interestingly, anisotropy was only observed in Matrigel/collagen I bioinks and not in pure collagen, presumably due to Matrigel-induced molecular crowding effects that promoted fiber nucleation. By using a lower concentration of collagen, the authors overcame a significant hurdle of using expensive, highly concentration collagen I inks in their platform. As shown in Fig. 10, the extrusion process was well suited to fabricate defined interfaces between collagen gels with heterogeneous fiber properties.

Figure 10.

Images showing collagen gels with aligned fibers, produced by extruding collagen. A: schematic of extruded pattern. B: confocal reflectance images from left, middle, and center of pattern showing different regimes of collagen alignment. Collagen was mixed with Ficoll 400 or Matrigel to control fiber density by molecular crowding. C and D: demonstration of interface between drop-cast and 3-D printed collagen networks. [Reproduced from Nerger et al. (109) with permission of Royal Society of Chemistry (United Kingdom).]

Bioprinting approaches are well suited for creating large (centimeter scale) hydrogel constructs with complex shapes and defined fiber microstructure properties. Compared with direct strain-based and other flow-based techniques, bioprinting approaches require specialized equipment and suffer from resolution limits that could hamper widespread adoption. However, as the accessibility of bioprinters continues to increase, this technology could become more prevalent in biology-focused laboratories and help enable the exploration of increasingly complex biological questions.

SUMMARY

Fiber anisotropy in the ECM is a crucial biological cue that is important in a wide range of physiological processes. To help researchers identify techniques to study how cells respond to collagen fibers in vitro, we have described several popular engineering techniques that use the mechanical manipulation of natural collagen I hydrogels (top-down methods) or collagen I solutions (bottom-up methods) to create anisotropic fiber domains. These methods range in complexity from magnetic methods and the direct application of strain, to flow-based microfluidic methods and spin coating, and finally to extrusion and bioprinting techniques. Each approach has its advantages and limitations as shown in Fig. 1, and the goal of this review was to provide an overview of each technique and identify key experimental capabilities. These technologies were selected because they are flexible enough to incorporate additional collagen modifications either before or after the process, but their equipment requirements are not overly onerous and the methods remain accessible to researchers in the engineering, materials science, and life science domains.

The collagen triple helix is flanked by telopeptides at either end that may be cleaved during enzyme-assisted extraction of collagen (110). Significant differences exist between the microstructure, self-assembly rates, and stability of gels that are fabricated using acid extracted telocollagen as compared to enzyme extracted atelocollagen (110–113). In general, telocollagen self-assembles at a faster rate as compared to atelocollagen, which limits the working time with telocollagen solutions (111). Methods such as microfluidics, bioprinting, and strain require precise control over collagen self-assembly; therefore, bovine-derived type I atelocollagen has become a popular choice over the past 5 years. Due to the higher yield of bovine sources, bovine atelocollagen is also more economical for large-scale experiments. Collagen type I may also contain small amounts of impurities in the form of collagen IV or collagen II; however, commercial formulations are available with purity >99.9%. The effects of telopeptides and collagen source on the ability to define fiber alignment have not been well documented in literature and require more attention, but the general principle of applying strain to align fibers can be applied across collagens from any source.

FUTURE PERSPECTIVES

The engineering of collagen-based hydrogels has focused on replicating the complex biophysical environments found in vivo to increase the physiological relevance of cell culture models. Natural collagen I continues to be widely used because its properties can be precisely tailored to fit the needs of a particular application. However, the major challenge lies in the lack of independence between the structural properties. For example, stiffness can be changed by increasing the collagen concentration or by chemically cross-linking the material, but density (and porosity) is also impacted, and those changes affect the transport of biomolecules through the gel. Thus, it can be difficult to determine how one matrix parameter can influence cell behavior using only natural materials. Therefore, the integration of synthetic and natural materials is an active area of investigation (114–116).

To create more physiological cell culture platforms, there is much interest in controlling not only bulk properties and composition of a hydrogel, but also tailoring spatial variations in these properties. For example, cell motility is affected by gradients in matrix stiffness (durotaxis) (79, 117), tethered biomolecules (haptotaxis) (118, 119), and other local cues (ratchetotaxis) (120). Precisely presenting combinations of gradients within a 3-D hydrogel environment could help answer open questions related to the hierarchy between multiple guidance cues.

Finally, significant opportunities exist to develop hydrogel engineering approaches that can be effectively translated from a technology-focused environment to a life science laboratory. Strategies such as limiting the need for specialized equipment and collaboratively determining the required experimental capabilities can help make engineering solutions more accessible and help the research community collectively create more physiologically relevant model systems.

GRANTS

This work was supported in part by the National Institutes of Health National Institute of General Medical Sciences (NIGMS) Grant R35GM119623, National Heart, Lung, and Blood Institute (NHLBI) Grant R61HL154249, NIGMS Grant R43GM137651, and Rochester Institute of Technology Kate Gleason College of Engineering New Faculty Startup Funds.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.A., N.N.N.A., and M.-C.H. prepared figures; A.A., I.M.J., and M.-C.M. drafted manuscript; T.R.G. and V.V.A. edited and revised manuscript; A.A., I.M.J., M.M., N.N.N.A., M.-C.H., T.R.G., and V.V.A. approved final version of manuscript.