Optimization of Freeze-FRESH Methodology for 3D Printing of Microporous Collagen Constructs

Abstract

Freeform reversible embedding of suspended hydrogels (FRESH) is a layer-by-layer extrusion-based technique to enable three-dimensional (3D) printing of soft tissue constructs by using a thermo-reversible gelatin support bath. Suboptimal resolution of extrusion-based printing limits its use for the creation of microscopic features in the 3D construct. These microscopic features (e.g., pore size) are known to have a profound effect on cell migration, cell–cell interaction, proliferation, and differentiation. In a recent study, FRESH-based 3D printing was combined with freeze-casting in the Freeze-FRESH (FF) method, which yielded alginate constructs with hierarchical porosity. However, use of the FF approach allowed little control of micropore size in the printed alginate constructs. Herein, the FF methodology was optimized for 3D printing of collagen constructs with greater control of microporosity. Modifications to the FF method entailed melting of the FRESH bath before freezing to allow more efficient heat transport, achieve greater control on microporosity, and permit polymerization of collagen molecules to enable 3D printing of stable microporous collagen constructs. The effects of different freezing temperatures on microporosity and physical properties of the 3D-printed collagen constructs were assessed. In addition, finite element (FE) models were generated to predict the mechanical properties of the microporous constructs. Further, the impact of different micropore sizes on cellular response was evaluated. Results showed that the microporosity of 3D-printed collagen constructs can be tailored by customizing the FF approach. Compressive modulus of microporous constructs was significantly lower than the non-porous control, and the FE model verified these findings. Constructs with larger micropore size were more stable and showed significantly greater cell infiltration and metabolic activity. Together, these results suggest that the FF method can be customized to guide the design of 3D-printed microporous collagen constructs.

Introduction

Scaffold composition and architecture are key elements in the design of biomimetic tissue scaffolds. The use of natural biomaterials such as collagen is often preferred due to good biocompatibility, ease of processing, presence of cell binding sequences (i.e., RGD), and compositional biomimicry with native tissue.^1,2 Tissues are comprised of complex, highly organized hierarchical structures that range from nanoscale to macroscale that provide the characteristic tissue properties.³ Therefore, scaffold microarchitecture, particularly at the cellular scale, is highly important and has been previously shown to govern cellular behavior, infiltration, differentiation, and nutrient exchange.^4–7 For example, scaffolds with thicker fibers and larger pore sizes have been shown to increase secretion of proangiogenic factors and stimulate myofibroblastic differentiation of adipose stromal cells.⁸ Further, polymeric scaffolds with pore size >325 μm have been shown to enhance cell infiltration and augment osteoblast differentiation and mineralization in vitro.⁷ On the other hand, smaller micropore size (i.e., 30 μm) has been shown to promote microvascularization.⁹ Therefore, biofabrication strategies that allow for the design and development of biomimetic tissue scaffolds with controlled microarchitecture are of paramount importance to guide cellular response and achieve functional tissue regeneration.

Electrospinning and freeze-casting are commonly employed to generate porous collagen scaffolds^10–12; however, these methods are associated with significant limitations. The use of high voltage and corrosive solvents during the electrospinning process have raised concerns of possible collagen denaturation.¹³ While freeze-casting is frequently used to generate porous collagen scaffolds without the use of organic solvents,¹⁴ scaffold geometry is constrained by the mold shape used for freezing the polymer solution.¹⁵ In addition, depending on the degree of porosity, freeze-dried scaffolds are associated with low strength and toughness.¹⁶ Extrusion-based three-dimensional (3D) printing allows for better spatial control over material deposition and, hence, is often implemented to generate porous scaffolds with complex geometries.^17–20 Cell-laden bioinks that include hydrogel-based materials, microcarriers, tissue spheroids, tissue strands, cell pellets, and components of decellularized matrices are commonly used for extrusion-based 3D printing applications.²¹ Typically, these bioinks allow for direct encapsulation of cells within the ink, thereby facilitating precise deposition of cells and allowing for the construction of complex multicellular structures.²² However, shear stress experienced during the extrusion of cell-laden highly viscous bioinks is a major concern that can compromise cell viability.²³ In addition, restricted cell–cell interactions and slower migration are possible limitations.^22,24 To circumvent these limitations, an alternative approach can be to seed cells on the constructs post-3D printing. However, printed scaffolds are devoid of microporous features below 100 μm due to limited resolution of the extrusion-based 3D printing technique;^25,26 and therefore, cell seeding onto these constructs typically results in cell adhesion on flat solid strut surfaces with limited cell infiltration to the core of the construct. Development of biofabrication strategies for the introduction of microporosity on the printed struts can allow for more efficient cell seeding of the construct after the printing process via greater cell infiltration, migration, and population of the 3D constructs.

Freeform reversible embedding of suspended hydrogels (FRESH) technique employs a thermoreversible gelatin support bath to enable layer-by-layer printing of stable tissue constructs by using soft hydrogel-based biomaterial inks.²⁷ The ability to print complex structures such as heart components and patient-specific menisci using highly concentrated type I collagen inks (35 mg/mL) and the FRESH printing method has been previously demonstrated.^28,29 Lee et al. have recently reported that the incorporation of gelatin microparticles (50 μm) into collagen inks before printing and subsequent leaching of the microparticles post-printing can yield collagen constructs with uniform microporosity.²⁸ Although typical needle diameters for 3D printing range from 100 to 800 μm, possible agglomeration of gelatin microparticles during extrusion can result in clogging of the needle. Recently, FRESH 3D printing was combined with freeze-casting in the Freeze-FRESH (FF) method to produce alginate constructs with hierarchical porosity.³⁰ The FF methodology involved 3D printing of alginate constructs in a support bath followed by freezing and lyophilization. However, use of the FF approach allowed little control of micropore size in the printed alginate constructs.

Herein, the FF printing technique is optimized for 3D printing of collagen constructs with tailorable microporosity. The FF technique was modified by melting the gelatin support bath before freezing to allow for better heat transfer via convective medium during the freezing process. Inclusion of the FRESH melting step also allowed gelation of the collagen prints, which is essential for printing stable collagen constructs by using the FF methodology. The micropore size was modulated by freezing the printed constructs in the melted gelatin support bath at −20°C and −80°C followed by lyophilization. The effect of different freezing temperatures on porosity, swelling, degradation, and mechanical properties of the 3D-printed constructs was assessed. In addition, Saos-2 osteosarcoma cells were seeded onto 3D-printed microporous constructs and the effect of pore size on cell morphology, infiltration, metabolic activity, and alkaline phosphatase (ALP) activity was investigated.

Materials and Methods

Preparation of gelatin slurry support bath

The gelatin support bath was prepared by adopting a protocol from previously published literature.²⁷ Briefly, 10 g of gelatin type A (Thermo Fisher Scientific, Waltham, MA) was added to 250 mL of phosphate buffer saline (PBS) in a 500 mL Mason jar, heated to 45°C, and mixed until completely dissolved. The mix was then gelled at 4°C for 24 h. After this, chilled PBS was added and filled to the brim of the gelatin containing jar and left to freeze at −20°C for about 1 h. The frozen mixture was then blended in the same container at 20 s intervals for a total of 60 s by using a household blender. The blended mixture was then transferred to 50 mL tubes and centrifuged at 3000 g for 5 min. The supernatant was discarded, and the tubes were filled with chilled PBS and vortexed to resuspend the gelatin. The centrifugation step was repeated a few times to ensure complete removal of soluble gelatin. The individual tubes were then mixed to distribute the gelatin evenly and stored at 4°C until use.

For 3D printing, the FRESH media were prepared by transferring two tubes of gelatin mixture into two 100 mL capped syringes and centrifuging at 180 g for 5 min. The supernatant was discarded, and the mixture from one syringe was transferred to the other by extruding with a plunger to minimize the formation of bubbles. Air was removed from the syringe, and the syringe was centrifuged once again. The ready-to-use FRESH was then extruded into Petri dishes and stored at 4°C.

3D printing of microporous collagen constructs using FF methodology

The 3D-printed microporous collagen constructs (15 mm × 15 mm × 1 mm) were fabricated by adopting a previously published protocol on the FF methodology with modifications (Fig. 1).³⁰ The REGEMAT 3D (Granada, Spain) bioprinting system was used to print pure collagen constructs by using highly concentrated neutralized type I collagen ink (Lifeink^® 200; 35 mg/mL; Advanced Biomatrix, San Diego, CA) using the printing parameters outlined in Table 1. The collagen ink was transferred by using a coupler to a 3 mL syringe and centrifuged at 850 g for 1 min to remove air bubbles. The capped syringe was then loaded onto the printer, and FRESH containing Petri dishes were placed onto the printer stage. The syringe cap was replaced with a 25G needle for extrusion, and the printer was zeroed to begin printing in FRESH. Once printing was completed, constructs were incubated at 37°C for 45 min to allow gelation of collagen as well as melting of the FRESH media. The printed constructs were then frozen at −20°C (FF −20) or −80°C (FF −80) within the melted FRESH media overnight. The frozen constructs were then lyophilized (Labconco, Kansas City, MO) for 24 h. The printed constructs were then recovered from the FRESH media by adding PBS at room temperature and then incubating at 37°C for about 45 min until the gelatin melted completely. The melted gelatin was removed by alternate washing of the constructs with deionized (DI) water and PBS by gently replacing the liquid using a pipette to preserve the integrity of the printed construct. Using a wide spatula, constructs were then carefully transferred to a new dish filled with DI water. The constructs were then frozen again at either −20°C or −80°C and subsequently lyophilized to obtain dry porous collagen constructs. Non-porous collagen constructs (control) were printed in a similar manner and recovered post-FRESH melting without the freezing steps. For qualitative assessment of print and shape fidelity, high-resolution images of the printed constructs were taken by using a Google pixel 3 camera mounted at a fixed distance.

FIG. 1.

Schematic illustration of fabrication of 3D-printed microporous collagen constructs using FF methodology. 3D, three-dimensional; FRESH, freeform reversible embedding of suspended hydrogels; FF, Freeze-FRESH. Color images are available online.

Table 1.

Parameters for 3D Printing of Collagen Constructs

Parameter	Value	Description
Tip diameter	0.25 mm	A blunt syringe tip for the print head
Tip gauge	25 g	A blunt syringe tip for the print head
Print shape	Cube	Outer limit of construct is square
Infill pattern	Mesh	Inner pattern is cross-hatched
Infill angle	45°	The angle at which the bioink was extruded
Flow speed	4 mm/s	Extrusion speed for optimal density

Assessment of microporosity of 3D-printed collagen constructs using scanning electron microscopy

Scanning electron microscopy (SEM) was used to assess the effect of different freezing temperatures on pore morphology (i.e., pore size, pore circularity, and pore size distribution) of the 3D-printed collagen constructs (N = 4 constructs/group). The control constructs were subjected to a rigorous sample preparation procedure that entailed dehydration in a graded series of ethanol solutions followed by critical point drying to retain structural integrity. The FF −20 and FF −80 constructs were dry after the FF process and were used directly for SEM. Constructs were put on stubs, sputter coated with gold, and imaged at 30 × and 100 × magnification with a JEOL JSM-6380LV SEM (JEOL, Tokyo, Japan). High-magnification SEM images were analyzed by using ImageJ to measure pore size and pore circularity by calibrating the line measurement tool to the image scale (NIH, Bethesda, MD).³¹

For pore size measurements, visible through-pores in the same focal plane were measured using the line tool by applying a consistent pore-selection criterion throughout the analysis. A minimum of 25 pores were measured per construct and four different constructs were analyzed per condition for at least 150 pore-size measurements per group. The effective pore diameter (d) was obtained by using Equation (1), where (l) is the pore long axis length and (s) is the pore short axis length.³² All measurements were assembled to generate frequency histograms and assess pore-size distribution. Pore circularity was measured by using the freehand area selection tool in ImageJ and calculated by using Equation (2) from an extended version of the “measure” command that calculates object circularity. A value of 1 indicates a perfect circle, and a value of 0 indicates an elongated polygon.

Assessment of swelling capacity of 3D-printed microporous collagen constructs

Swelling studies were performed to determine the effect of different pore size on the degree of fluid absorption (N = 8/group). The 3D-printed microporous constructs obtained by the FF method and critical-point dried non-porous control constructs were cut in half by using a sharp blade and weighed to measure the dry weight (W_d). Since the results for the swelling study are based on the initial weight of the construct, each construct was cut in two halves and used as separate samples. Constructs were then incubated in 500 μL of PBS at room temperature for 24 h. After this, constructs were removed from PBS by using tweezers, blotted twice on a Kimwipe to remove the excess liquid, and placed onto an analytical balance to obtain the wet weight (W_w). The swelling degree was calculated as the percent change in weight to the initial dry weight as shown in Equation 3:

Assessment of stability of 3D-printed microporous collagen constructs

An in vitro collagenase degradation assay was performed to assess the effect of different pore size on the stability of the 3D-printed collagen constructs (N = 8 constructs/group). The 3D-printed constructs were hydrated with PBS for 30 min before testing, blotted on a Kimwipe, and weighed (W_o). Constructs were then incubated in 500 μL of the collagenase solution (5 U/mL in 0.1 M Tris-HCl buffer and 5 mM CaCl₂; pH 7.4) for 2 h at 37°C under a constant stirring rate. Constructs were then removed from solution, blotted on a Kimwipe, and weighed again (W_f). The percent of residual mass post-incubation was calculated by using Equation (4):

Mechanical characterization of 3D-printed microporous collagen constructs

Compression tests were performed to assess the effect of different pore size on the compressive modulus of 3D-printed collagen constructs (N = 4 constructs/group) by using an MT G2 MicroTester (CellScale Biomaterials Testing, Waterloo, Canada). Constructs were hydrated in PBS, transferred to an acrylic platform, and placed into the testing chamber. A 0.3 mm diameter tungsten beam tipped with a 2 × 2 mm stainless-steel platen was used to compress the constructs. For each test, four measurements were taken per construct by applying a 10 μm/s loading rate until a 20% displacement of sample thickness was reached. The displacements at both the platen surface and the base of the tungsten beam were recorded. The compression force P extracted from the software is calculated by using the Euler-Bernoulli beam theory shown in Equation (5). In the equation, δ is the relative displacement of the platen to the base of the tungsten beam, E is the Young's modulus of the beam, L is the length of the beam, and r is the radius of the beam.

Stress was computed by normalizing the load with the area of the compression platen, and strain was determined by the ratio of δ to the original sample thickness. Stress–strain curves were generated, and the modulus was calculated as the slope of the stress–strain curve.

Finite element modeling of 3D-printed microporous collagen constructs

The compressive moduli of 3D-printed collagen constructs were further investigated by using finite element (FE) models reconstructed from SEM images. ImageJ and an open source MATLAB code “Im2mesh” were used to convert SEM images into FE models for microporous collagen constructs.^33,34 The virtual compression of constructs was performed in the commercial code ABAQUS (Dassault Systemes Simulia Corp., Providence, RI). The Young's modulus of collagen was 1.71 kPa, obtained from the compression test of non-porous control samples. The modulus of liquid medium was 0.62 kPa. The Poisson's ratio for both constituents of the constructs was 0.3. A compression displacement of 10% was applied to the top surface of the constructs.

Cell culture

Saos-2 human osteosarcoma cells (HTB-85; ATCC) were cultured in 75 cm² tissue culture flasks and maintained in Roswell Park Memorial Institute growth medium supplemented with 10% fetal bovine serum (FBS), 1% l-glutamine, and 1% penicillin/streptomycin. Cells were expanded for 2–4 days at 37°C and used for all the experiments. Printed constructs were sterilized in 70% ethanol for 30 min, transferred to an ultralow attachment six-well plate (one construct/well), and washed with sterile PBS. Saos-2 cells were seeded on top of the constructs at a density of 5000 cells/cm² based on the area of the well (50,000 cells/well). The culture medium was replaced 6 h post-seeding to remove unattached cells, and the remaining adherent cells were cultured for 7–14 days. For assessment of cell metabolic activity and cell morphology, cells were maintained in Minimum Essential Medium Eagle-Alpha Modification (α-MEM) containing 10% FBS, 10 mM beta-glycerophosphate, and 1% penicillin/streptomycin. For the assessment of ALP activity, culture medium was composed of α-MEM supplemented with 10% FBS, 10 mM beta-glycerophosphate, 10⁻⁷ M dexamethasone, and 1% penicillin/streptomycin. Culture medium was replaced every 3 days.

Assessment of Saos-2 cell metabolic activity using Alamar Blue assay

Alamar Blue (AB) assay was conducted to evaluate the effect of different pore size on cell metabolic activity (N = 8 constructs/group). At periodic intervals (days 1, 4 and 7), culture medium was replaced with a 10% solution of AB (Thermo Scientific) in α-MEM and incubated at 37°C for 4 h. The same constructs were tested at each time point. After this, 100 μL of AB solution was transferred in triplicate from each well into a separate 96-well plate. Fluorescence was measured by using a SpectraMax M2e plate reader (Molecular Devices, San Jose, CA) with an excitation wavelength of 555 nm and emission wavelength of 595 nm.

Assessment of Saos-2 cell morphology and infiltration on 3D-printed microporous collagen constructs

Confocal microscopy (Nikon) was used to assess the effect of different pore size on Saos-2 cell morphology (N = 3 constructs/group/time point). At days 1 and 7, constructs were fixed with 3.7% formaldehyde solution (with 0.05% Triton X-100 in PBS), washed twice with PBS, and incubated in permeabilization buffer (0.1% Triton X-100 in PBS). Then, constructs were washed twice with PBS and incubated in blocking buffer (1% bovine serum albumin and 0.05% Triton X-100 in PBS) for 30 min. After this, constructs were washed with PBS and stained with a working solution of AlexaFluor 488 phalloidin (1:25 dilution in PBS) (Invitrogen, CA) for 30 min. The stain was then removed, and samples were washed twice with PBS, wrapped in aluminum foil, and stored at 4°C before imaging. Quantitative analysis of cell morphology was performed using ImageJ by first converting the images to binary format and applying a manual threshold to select cells. After this, cell clusters and debris particles were excluded from the analyses, and cell size and cell shape were measured by using the particle analyzer toolbox. Cell infiltration into the collagen constructs was determined as the distance between the highest and lowest point of visible cells in the z-axis, which is given by the thickness of the z-stack needed to capture the entire cell layers within the construct. Measurements were obtained from four images per construct for a total of at least 12 measurements per group.

Assessment of ALP activity on 3D-printed microporous collagen constructs

To normalize ALP activity of Saos-2 cells, quantification of the total amount of DNA was first performed by using Quant-iT Picogreen dsDNA Kit (ThermoFisher Scientific) at days 7 and 14 (N = 3/group/time point). For cell harvest, collagen constructs were washed once with PBS and incubated in 1 mL collagenase solution (1 mg/mL) mixed in a solution of 5 mM CaCl₂, 0.1 M HCl, and Trizma^® base buffer pH 7.4 (Sigma Aldrich) at 37°C with constant stirring for 45 min to completely degrade the construct. Samples were then collected in microcentrifuge tubes and centrifuged at 850 g for 5 min to obtain the cell pellet. Cells were lysed by adding 1 × assay buffer to the cell pellet in the microcentrifuge tubes. Then, 50 μL of cell lysates from each sample and an equal volume of dsDNA reagent were added into each well of a 96-well plate in triplicate. The plate was then incubated at room temperature for 5 min covered from light with an aluminum foil. After incubation, fluorescence was measured at an excitation wavelength of 480 nm and emission wavelength of 520 nm by using a SpectraMax M2e plate reader. The DNA concentrations in cell lysates were obtained from a standard curve produced by using known concentrations of DNA.

The ALP activity was measured at days 7 and 14 (N = 3 constructs/group/time point) by using SensoLyte p-nitrophenylphosphate (pNPP) Alkaline Phosphatase Assay Kit (AnaSpec, Inc., Fremont, CA). A volume of 50 μL of the same cell lysates used for DNA quantification was added onto a separate 96-well plate along with an equal volume of pNPP solution in triplicate. Samples were incubated at room temperature for 60 min, and absorbance was measured at 405 nm by using a SpectraMax M2e plate reader. The ALP activity measured for each sample was normalized to the corresponding DNA content of Saos-2 cells to account for differences in cell number between constructs.

Statistical analyses

Results are expressed as mean ± standard deviation. Data for SEM analyses, swelling study, and degradation assay were analyzed by using one-way analysis of variance (ANOVA) followed by Tukey post hoc test for pairwise comparisons (JMP Statistical Discovery; SAS, Cary, NC). Statistical analysis for the mechanical characterization of collagen constructs was performed by using MaxStat (MaxStat Software, Germany) with one-way ANOVA and Tukey post hoc test. For the assessment of cell metabolic activity, cell infiltration, and ALP activity, statistical analysis was performed by using two-way ANOVA and Tukey post hoc test (JMP Statistical Discovery). Statistical significance was set at p < 0.05.

Results

Print and shape fidelity of 3D-printed collagen constructs using the FF method

The Computer-Aided Design (CAD) model employed in this study was a meshed pattern that consisted of four printed layers for the outer frame and two printed layers to make up the inner struts of the construct. The shape of the printed constructs for all conditions was consistent and indicated a good reproduction of the initial 3D CAD model (Fig. 2). Further, the print and shape fidelity of the control constructs was retained post-FRESH printing, recovery, and hydration. For the FF −20 and FF −80 microporous constructs, the outer frame of the printed construct remained intact; however, the inner struts were weaker and often broke during the multiple washing steps of the FF process. These results suggest that the print and shape fidelity of control constructs is better retained compared with microporous constructs.

FIG. 2.

3D model and images of non-porous control and microporous FF −20 and FF −80 printed constructs before removal from FRESH media and after rehydration. Color images are available online.

FF method allows modulation of micropore size in 3D-printed collagen constructs

The assessment of SEM images revealed that the use of different FF freezing temperatures impacted the microporosity and pore size of 3D-printed collagen constructs. The 3D-printed control constructs that were not subjected to freezing exhibited a solid non-porous surface morphology (Fig. 3A, D). The SEM images of FF −80 constructs showed higher density of micropores and smaller pore size compared with FF −20 constructs (Fig. 3B, C, E, and F). Quantitative analyses revealed that the average pore size of FF −20 constructs was significantly greater (two-fold; p < 0.05) than FF −80 constructs (Table 2). Pore circularity, although statistically significant, was deemed comparable between FF −20 and FF −80 constructs (Table 2). Quantification of pore size distribution revealed a non-normal dataset for both FF −20 and FF −80 constructs (Fig. 4). Although there is some overlap in the pore size distribution between the two groups, the FF −80 constructs had a more homogenous distribution, with most of the pores between 30 and 60 μm as opposed to FF −20 constructs, which showed a broader pore size distribution ranging from 28 to 346 μm. Together, these results indicate that it may be feasible to modulate the microporosity of 3D-printed collagen constructs by changing the freezing temperature during the FF fabrication process.

FIG. 3.

Assessment of microporosity in 3D-printed collagen constructs fabricated by using FF methodology. The SEM images at low and high magnification of non-porous control construct (A, D), FF −20 (B, E), and FF −80 (C, F). Scale bar: 500 μm for (A–C) and 100 μm for (D–F). SEM, scanning electron microscopy. Color images are available online.

FIG. 4.

Histogram for pore-size distribution of 3D-printed FF −20 and FF −80 constructs. Color images are available online.

Table 2.

Pore Size and Circularity of Freeze-FRESH 3D Printed Collagen Constructs

Group	Pore diameter (μm)	Circularity
FF −20	107 ± 56	0.776 ± 0.013
FF −80	57 ± 23^*	0.801 ± 0.008^*

Indicates p < 0.05 when comparing between FF −20 and FF −80 constructs.

FRESH, freeform reversible embedding of suspended hydrogels; FF, Freeze-FRESH.

Micropore size influences stability of collagen constructs

In vitro swelling and degradation assays were performed to assess the effect of different pore sizes on fluid absorption and stability of 3D-printed collagen constructs. Swelling degree is a measure of the percent increase in the weight of the construct due to fluid absorption. Swelling degree of non-porous control constructs was lower than the microporous constructs, although this finding was not statistically significant (p = 0.1; Fig. 5A). The FF −20 and FF −80 constructs showed a similar swelling degree. Expectedly, the microporous constructs degraded significantly faster than the non-porous control constructs (p < 0.05; Fig. 5B). The residual mass of FF −80 constructs was significantly lower than FF −20 constructs (p < 0.05), indicating that constructs with larger pore size were more stable. Together, these results suggest that the stability of 3D-printed microporous constructs is influenced by the average pore size.

FIG. 5.

(A) Swelling degree, and (B) in vitro collagenase degradation of 3D-printed microporous collagen constructs after 2 h (horizontal line connecting groups denote p < 0.05). Color images are available online.

Microporous constructs have lower compressive modulus

Stress–strain curves obtained from compression testing of 3D-printed collagen constructs are shown in Figure 6A. The compressive modulus of the non-porous control constructs was at least two-fold higher than the microporous constructs (p < 0.05; Fig. 6B). The compressive modulus of FF −20 constructs trended lower than FF −80 constructs, indicating that the larger pore size lowered the stiffness of the construct (Fig. 6B). However, these results were not statistically significant. These results indicate that the introduction of microporosity decreases the compressive modulus of 3D-printed collagen constructs.

FIG. 6.

Mechanical assessment of 3D-printed microporous collagen constructs. (A) Representative stress versus strain curves, and (B) Compressive modulus of 3D-printed collagen constructs (horizontal line connecting groups denote p < 0.05). Color images are available online.

FE model validates experimental data for compressive modulus

The FE models reconstructed from the SEM images of 3D-printed microporous collagen constructs were employed to assess compression-induced stress and strain distributions in the constructs (Fig. 7A–K). After compression, uniform stress and strain was observed in the non-porous control construct, as expected. The compressive stress of the control construct at a strain of 0.1 was 0.176 kPa, which was validated with results from the experiments that showed a compressive stress of 0.171 kPa with a difference of 2.92%. On the other hand, the microporous constructs demonstrated heterogeneous stress and strain distributions. The peak compressive stresses of the FF −20 and FF −80 were 0.245 and 0.360 kPa, respectively. The peak compressive strain of the FF −20 and FF −80 were 0.011 and 0.030, respectively. It is worth noting that compression load led to the alignment of the collagen constructs that were associated with the directional trends in the stress and strain distributions. Specifically, the compressive stress distribution was directed along the loading direction, indicating the collagen elements carried the applied compression load (Fig. 7G, H). The compression strain distribution exhibited a directional trend perpendicular to the loading direction (Fig. I, J). These features might be used to guide the design of an anisotropic construct. In addition, the predicted compressive moduli from FE models agree with experimental results with differences of <5% (Fig. 7K).

FIG. 7.

(A–F) Finite element model construction of 3D-printed microporous collagen constructs. SEM images of FF −20 (A) and FF −80 (D), Segmentation of SEM images (B, E), and finite element models (C, F). (G–J) Compression-induced stress and strain distribution in FF −20 (G, I) and FF −80 (H, J). [compression stress (G, H) and compression strain (I, J)]. (K) Experimental and estimated compression modulus of 3D-printed collagen constructs. Color images are available online.

Assessment of Saos-2 cell morphology on 3D-printed microporous collagen constructs

Qualitative assessment of stained cell cytoskeleton showed that cells exhibited a combination of round and spread morphology on all constructs (Fig. 8). This finding was confirmed via quantitative analyses for cell size and cell shape (i.e., circularity), which showed comparable results between constructs at day 1 and 7 (Table 3). When comparing between time points, a significant decrease (p < 0.05) in cell size was observed for control and FF −20 constructs. Together, these results indicate that the presence of micropores had no effect on Saos-2 cell morphology.

FIG. 8.

Assessment of cell morphology via cytoskeleton staining by using Alexa Fluor 488 Phalloidin. (A–C) Day 1 for non-porous control construct (A), FF −20 (B), and FF −80 (C). (D–F) Day 7 for non-porous control construct (D), FF −20 (E), and FF −80 (F). Scale bar: 200 μm. Color images are available online.

Table 3.

Cell Size and Circularity of Saos-2 Cells Seeded on Three-Dimensional-Printed Collagen Constructs

Group	Cell size (μm²)		Circularity
Group	Day 1	Day 7	Day 1	Day 7
Control	885 ± 246	539 ± 31^#	0.73 ± 0.04	0.78 ± 0.02
FF −20	951 ± 229	518 ± 30^#	0.64 ± 0.11	0.77 ± 0.01
FF −80	813 ± 124	539 ± 39	0.71 ± 0.06	0.73 ± 0.02

Indicates p < 0.05 when comparing between time points for each construct group.

Micropores improve cell infiltration and metabolic activity of Saos-2 cells

Analysis of cell infiltration depth at day 1 revealed that cell infiltration was significantly higher in FF −20 constructs compared with FF −80 constructs and non-porous controls (p < 0.05; Fig. 9A, B). By day 7, cell infiltration was comparable for all groups. These results suggest that larger pore size allows better cell infiltration in 3D-printed microporous collagen constructs.

FIG. 9.

(A) Confocal images in XZ plane for visual assessment of effect of pore size on cell infiltration into microporous 3D-printed collagen constructs on day 1. (B) Quantification of cell infiltration measured as the distance traveled by the cells from the surface of the construct, and (C) Cell metabolic activity on 3D-printed collagen constructs (horizontal line denotes p < 0.05 when comparing between constructs at the same time point, “*” indicates p < 0.05 when comparing with day 1, and “#” indicates p < 0.05 when comparing with day 4 within the same group). Color images are available online.

The AB assay was performed to quantify the effect of pore size on Saos-2 cell metabolic activity. Results showed that cell metabolic activity increased with time on all constructs. Specifically, cell metabolic activity was increased significantly with time in both FF −20 constructs and FF −80 constructs at all time points (p < 0.05) and increased significantly between days 1 and 7 for the non-porous collagen constructs (p < 0.05; Fig. 9C). The FF −20 constructs exhibited significantly greater cell metabolic activity compared with the control at day 4 and 7 (p < 0.05), and FF −80 constructs showed significantly higher cell metabolic activity compared with the control at day 7 (p < 0.05). Cell metabolic activity trended higher on FF −20 constructs compared with FF −80 at all time points, although these results were not statistically significant. Together, results from the AB assay suggest that the presence of micropores enhances cell metabolic activity.

Assessment of Saos-2 cell ALP activity

ALP is an enzyme encoded by genes that are highly expressed in cells of mineralized tissue and plays a significant role in bone formation. Measuring ALP activity of Saos-2 cells can be indicative of cell differentiation into bone-forming osteoblasts.³⁵ ALP activity was comparable on all constructs at day 7 followed by a significant decrease from day 7 to 14 (p < 0.05; Fig. 10). At day 14, FF −80 constructs showed significantly lower ALP activity compared with controls (p < 0.05). Similarly, lower ALP activity was also observed in FF −20 constructs compared with controls, but this difference was not statistically significant. ALP activity was similar at both time points for the two FF constructs. Together, these results suggest that ALP activity in Saos-2 cells may decrease on the introduction of microporosity in 3D-printed collagen constructs and that the size of the micropores had no effect on ALP activity.

FIG. 10.

Assessment of ALP activity on 3D-printed microporous collagen constructs (horizontal line connecting groups denote p < 0.05, and “*” indicates p < 0.05 when comparing with day 7). ALP, alkaline phosphatase. Color images are available online.

Discussion

Application of the FF technique with alginate inks has been shown to yield 3D-printed constructs with hierarchical porosity.³⁰ Adoption of the same protocol to generate collagenous constructs is not feasible, because freezing the construct in the support bath immediately after printing does not allow the collagen molecules to polymerize. Therefore, the resulting construct is unstable and does not survive the subsequent steps in the process. The modified FF technique employed in the current study entailed incubation of the construct in the FRESH bath after the printing process at 37°C for 45 min to allow for the collagen molecules to polymerize, undergo fibrillogenesis, and form a stable 3D construct (Fig. 1). Incubation at 37°C also melts the FRESH bath into a liquid medium. This simple modification to the FF technique may have allowed better heat transfer to occur during the subsequent ice templating process, yielding 3D-printed constructs with more controlled microporosity. Microporous structures generated from freeze-casting are governed by the freezing rate, the final freezing temperature, the height of solidified fluid layer, and the thermal properties of the suspension in both liquid and solid state.^15,36 The freezing rate controls the number of ice crystal nucleation sites formed (i.e., number of pores), and the rate of heat diffusion away from the nucleation points determines the size of ice crystals (i.e., size of the pores).³⁷ Prior work with the FF technique and alginate inks did not achieve similar control over the porosity of the printed constructs as in the current study, possibly because the freezing step was performed with the support bath in a solidified state, resulting in comparable rates of heat transfer despite changing the freezing temperature. Attainment of greater control of microporosity in 3D-printed constructs is significant and can be leveraged for the fabrication of tissue-specific scaffolds for different biomedical applications such as bone and skin regeneration.

The FF method employed in the current study entailed two separate freezing steps. In the first step, the construct was frozen in the melted FRESH bath without physical manipulation to preserve the structure of the printed construct. The second freezing step was performed in water and was essential to remove the gelatin adhered on the surface of the construct after the first lyophilization process. The weaker inner struts of the microporous constructs often broke due to flow-induced shear during washing and rehydration of the constructs (Fig. 2, FF −80 rehydrated). One approach to strengthen the inner struts of the construct and prevent them from breaking may be to print additional layers of collagen (i.e., four layers), like what was done to print the outer frame of the constructs. Alternatively, use of a single-step freezing process followed by intermittent incubation (37°C) and washing of the construct in PBS to remove remnant gelatin instead of a two-step freezing process can help mitigate breaking of the inner struts of the printed constructs. Application of physical or chemical crosslinking post-FF method can also help better retain the print and shape fidelity of microporous collagen constructs.^38,39

In the conventional freeze-casting process, the size and orientation of the micropores can be controlled by modulating the freezing temperature and directionality of freezing.^11,40 Nucleation is the initiation step in the formation of ice crystals during the freezing process. The nucleation step is followed by a growth phase that is governed by the freezing rate and controls the size of the ice crystals. The freezing rate also controls the distribution of the collagen fibers within the porous constructs.¹⁴ During the freeze–drying process, a stable dendritic ice crystal-forming front spreads throughout the collagen, causing the fibers to concentrate in the non-crystalized channels between the ice crystals. On the other hand, collagen constructs formed without freeze–drying would have a more uniform fiber distribution with no localized concentration of collagen fibrils. Results from this study showed that freezing the constructs with the melted FRESH bath at −80°C yielded constructs with significantly smaller micropores (p < 0.05) compared with the constructs frozen at −20°C (Fig. 3; Table 2). Further, FF −80 constructs exhibited a homogenous porous structure, as evidenced by a more uniform pore size compared with FF −20 constructs (Fig. 4). The micropore size distribution reported in the current study is in agreement with prior work using collagen/polycaprolactone scaffolds.⁴¹ Lower freezing temperatures create a larger temperature difference between the freezing source and the solution freezing temperature, resulting in expedited nucleation of ice crystals followed by a limited crystal growth phase, thereby producing smaller, more uniform micropores.^32,37 On the other hand, freezing constructs at higher temperatures may result in a more prolonged crystal growth phase, which, in turn, causes more variability in micropore size and morphology, yielding scaffolds with a wider pore size distribution (Fig. 4). The extended growth phase can also trigger the merging of neighboring ice crystals and thereby result in the formation of large micropores from multiple nucleation sites. While micropore size analyses in the current study was performed using dry FF constructs, future studies may use cryo-SEM for a comparison of micropore size in the hydrated state.^42,43

Typically, scaffolds with a larger pore size are expected to allow for greater fluid permeability and nutrient flow.⁴⁴ Swelling results in the current work indicated that printed collagen FF constructs with different micropore size yielded similar swelling properties (Fig. 5A). These results are consistent with a previous study that showed that swelling properties of freeze-dried recombinant human collagen peptide-chitosan scaffolds are independent of the micropore size.³² In a separate study, the swelling and degradation properties of freeze-dried collagen/hyaluronan/chitosan scaffolds were reported to decrease with decreased scaffold pore size due to reduced penetration of the fluid or enzyme solution, resulting in slower scaffold degradation.⁴⁵ Contrary to this outcome, the current study indicated that constructs with a smaller pore size degraded faster, possibly due to thinner pore walls (Fig. 5B). Since all three scaffold types were printed with the same amount of material, variations in pore size will have an impact on pore wall thickness. Further, it has been previously documented that pore size and specific surface area have an inverse relationship,⁴⁴ which can explain the significantly faster degradation of collagen constructs with a smaller pore size and higher surface area. Together, these results suggest that by controlling the micropore size, it is feasible to maintain the fluid uptake properties while modulating the degradation properties of the 3D-printed microporous collagen constructs.

The assessment of mechanical properties showed that the compressive modulus of collagen constructs with a different micropore size was comparable (Fig. 6). This could be attributed to the load-sharing capacity of the liquid medium within the constructs, as the effective modulus of liquid medium is approximately one-third of the collagen modulus. Nonporous control constructs showed a significantly higher stiffness compared with FF −20 and FF −80 porous constructs. These results are in agreement with previous work on microporous alginate scaffolds using the FF technique.³⁰ Similar results have also been shown with collagen-based scaffolds wherein a change in average micropore size from 96 to 151 μm had no impact on the compressive modulus of the scaffolds.⁴⁶ In addition, FE models captured the local compressive mechanics of the constructs and showed heterogeneous stress and strain distributions in porous collagen constructs (Fig. 7). Compared with the control group, the porous constructs exhibited larger stresses and lower strains because of the presence of the liquid medium in the microporosities (Fig. 7G–J). The FF −20 constructs showed much lower strain compared with FF −80 constructs, which is indicative of increased resistance to the compression load due to the larger volume fraction of the liquid medium in the FF −20 constructs. Results also demonstrated the compression-induced alignment of the porous collagen constructs as well as the directional trends in their stress and strain distributions. The anisotropic stress–strain distributions showed different load-sharing patterns, wherein the collagen fibers along the loading direction experienced larger stresses and strains compared with the ones perpendicular to the loading direction. This observation could be leveraged to guide the design of novel constructs with regulated directional modulus. Moreover, virtual experiments using FE models might enable the optimization of porous constructs for the enhanced mechanical environment of cells.⁴⁷

Scaffold microstructural features play an essential role in regulating cell behavior and fate.^8,48,49 Qualitative and quantitative analyses of confocal images of Saos-2 cells seeded on 3D-printed collagen constructs revealed that cell morphology was maintained with no differences on both non-porous control and porous constructs (Fig. 8). Decrease in cell size with time may be attributed to cell migration and remodeling of a high-density 3D collagen construct. Quantitative analyses of cell infiltration at day 1 revealed that cell infiltration was significantly higher (p < 0.05) on FF −20 constructs due to the presence of many large-sized pores, suggesting that cells migrate more efficiently into the construct, which may be a viable cell population strategy (Fig. 9A, B). By day 7, cell infiltration was found to be comparable between constructs, which may be attributed to initial cell-mediated remodeling of the uncrosslinked collagen construct. AB results showed that cell adhesion was comparable between the different conditions, as evidenced by similar relative fluorescence units at day 1 (Fig. 9C). At day 7, cell metabolic activity and proliferation was significantly greater (p < 0.05) on porous constructs compared with controls, possibly due to the availability of greater surface area in porous constructs to better support cell growth. These results are in agreement with prior work that show enhanced cell infiltration and proliferation on collagen scaffolds with a larger pore size.⁴⁰ Osteoblasts need to be in dense layers to differentiate and mineralize. Preliminary work to assess Saos-2 cell differentiation showed lower ALP activity on porous constructs compared with the non-porous control (Fig. 10). It is likely that the availability of a higher surface area in microporous collagen constructs resulted in a prolonged proliferative phase and delayed the onset of cell differentiation. Further, lower ALP activity on porous constructs may also be attributed to the lower stiffness of porous constructs compared with non-porous controls.⁵⁰

In conclusion, results from the current study demonstrate that the FF method can be customized to 3D-print collagen constructs with hierarchical microporosity by controlling the size of the macropores (i.e., pores between struts) in the 3D CAD model and modulating the micropores (i.e., pores on the struts) by employing different freezing temperatures. Although different micropore size had no effect on compressive properties, constructs with a smaller pore size degraded faster. Further, the introduction of microporosity via the modified FF approach enhanced cell infiltration and proliferation in 3D-printed collagen constructs. The FE model developed in this work allows to predict and validate the effect of microporosity on the mechanical properties of the collagen constructs. Future studies will entail performing longer-term cultures and employing mesenchymal stem cells to assess the effects of microporosity on cell functionality such as osteogenic differentiation and vascularization.

Protection for Human Subjects

Human subjects were not involved in this research.

Footnotes

Author Contributions

V.K., S.F., N.K., and T.S. conceived the idea and designed the study. T.S. and N.K. carried out majority of the experiments detailed in the article, including 3D printing of microporous constructs, SEM analyses, swelling ratio and stability, and in vitro cell experiments. P.D. performed the compression testing of the constructs. P.D. and L.G. generated and tested FE models, and drafted the methods, results, and sections for the computational work. T.S. and V.K. prepared all the figures and wrote the article with support from S.F. All authors read and approved the final article.

Acknowledgments

The authors would like to thank Ms. Gayle Duncombe for help with SEM and Mr. Trevor Schmitt for help with 3D printing and preparation of FRESH support media.

Disclaimer

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author Disclosure Statement

No competing financial interests exist.

Funding Information

This work was supported in part by a grant from the National Institute of Arthritis and Musculoskeletal and Skin Diseases of the National Institute of Health to V.K. (1R15AR071102).

References

Chevallay

, Herbage

. Collagen-based biomaterials as 3D scaffold for cell cultures: Applications for tissue engineering and gene therapy. Med Biol Eng Comput, 2000; 38:211–218.

Chang

, Lee

, Kuo

, et al. Three-dimensional collagen fiber remodeling by mesenchymal stem cells requires the integrin-matrix interaction. J Biomed Mater Res Part A, 2007; 80:466–474.

Christy

, Basha

, Kumari

, et al. Biopolymeric nanocomposite scaffolds for bone tissue engineering applications—a review. J Drug Deliv Sci Technol, 2020; 55:101452.

Liu

, Tamaddon

, Gu

, et al. Cell seeding process experiment and simulation on three-dimensional polyhedron and cross-link design scaffolds. Front Bioeng Biotechnol, 2020; 8:104.

Perez

, Mestres

. Role of pore size and morphology in musculo-skeletal tissue regeneration. Mater Sci Eng C, 2016; 61:922–939.

Dang

, Vaquette

, Shabab

, et al. Porous 3D printed scaffolds for guided bone regeneration in a rat calvarial defect model. Appl Mater Today, 2020; 20:100706.

Murphy

, Duffy

, Schindeler

, et al. Effect of collagen-glycosaminoglycan scaffold pore size on matrix mineralization and cellular behavior in different cell types. J Biomed Mater Res Part A, 2016; 104:291–304.

Seo

, Chen

, Ling

, et al. Collagen microarchitecture mechanically controls myofibroblast differentiation. Proc Natl Acad Sci U S A, 2020; 117:11387–11398.

Madden

, Mortisen

, Sussman

, et al. Proangiogenic scaffolds as functional templates for cardiac tissue engineering. Proc Natl Acad Sci U S A, 2010; 107:15211–15216.

10.

Babaie

, Bhaduri

. Fabrication aspects of porous biomaterials in orthopedic applications: A review. ACS Biomater Sci Eng, 2018; 4:1–39.

11.

Faraj

, van Kuppevelt

, Daamen

. Construction of collagen scaffolds that mimic the three-dimensional architecture of specific tissues. Tissue Eng, 2007; 13:2387–2394.

12.

Kim

, Choe

, Kim

. Injectable hierarchical micro/nanofibrous collagen-based scaffolds. Chem Eng J, 2019; 365:220–230.

13.

Zeugolis

, Khew

, Yew

ESY

, et al. Electro-spinning of pure collagen nano-fibres—just an expensive way to make gelatin?. Biomaterials, 2008; 29:2293–2305.

14.

Schoof

, Apel

, Heschel

, et al. Control of pore structure and size in freeze-dried collagen sponges. J Biomed Mater Res, 2001; 58:352–357.

15.

Rouhollahi

, Ilegbusi

, Florczyk

, et al. Effect of mold geometry on pore size in freeze-cast chitosan-alginate scaffolds for tissue engineering. Ann Biomed Eng, 2020; 48:1090–1102.

16.

Wegst

UGK

, Schecter

, Donius

, et al. Biomaterials by freeze casting. Philos Trans R Soc A Math Phys Eng Sci, 2010; 368:2099–2121.

17.

Guo

, Holzberg

, Lim

, et al. 3D printing PLGA: A quantitative examination of the effects of polymer composition and printing parameters on print resolution. Biofabrication, 2017; 12:024101.

18.

Hockaday

, Kang

, Colangelo

, et al. Rapid 3D printing of anatomically accurate and mechanically heterogeneous aortic valve hydrogel scaffolds. Biofabrication, 2012; 4:035005.

19.

Schipani

, Nolan

, Lally

, et al. Integrating finite element modelling and 3D printing to engineer biomimetic polymeric scaffolds for tissue engineering. Connect Tissue Res, 2020; 61:174–189.

20.

Laronda

, Rutz

, Xiao

, et al. A bioprosthetic ovary created using 3D printed microporous scaffolds restores ovarian function in sterilized mice. Nat Commun, 2017; 8:15261.

21.

Leberfinger

, Ravnic

, Dhawan

, et al. Concise review: Bioprinting of stem cells for transplantable tissue fabrication. Stem Cells Transl Med, 2017; 6:1940–1948.

22.

Ozbolat

. Scaffold-based or scaffold-free bioprinting: Competing or complementing approaches?. J Nanotechnol Eng Med, 2015; 6:024701.

23.

Blaeser

, Duarte Campos

, Puster

, et al. Controlling shear stress in 3D bioprinting is a key factor to balance printing resolution and stem cell integrity. Adv Healthc Mater, 2016; 5:326–333.

24.

Khoshnood

, Zamanian

. A comprehensive review on scaffold-free bioinks for bioprinting. Bioprinting, 2020; 19:e00088.

25.

Kim

, Lee

, Kim

. Strategy to achieve highly porous/biocompatible macroscale cell blocks, using a collagen/genipin-bioink and an optimal 3D printing process. ACS Appl Mater Interfaces, 2016; 8:32230–32240.

26.

Diloksumpan

, De Ruijter

, Castilho

, et al. Combining multi-scale 3D printing technologies to engineer reinforced hydrogel-ceramic interfaces. Biofabrication, 2020; 12:025014.

27.

Hinton

, Jallerat

, Palchesko

, et al. Three-dimensional printing of complex biological structures by freeform reversible embedding of suspended hydrogels. Sci Adv, 2015; 1:e1500758.

28.

Lee

, Hudson

, Shiwarski

, et al. 3D bioprinting of collagen to rebuild components of the human heart. Science, 2019; 365:482–487.

29.

Filardo

, Petretta

, Cavallo

, et al. Patient-specific meniscus prototype based on 3D bioprinting of human cell-laden scaffold. Bone Joint Res, 2019; 8:101–106.

30.

Wang

, Florczyk

. Freeze-FRESH: A 3D printing technique to produce biomaterial scaffolds with hierarchical porosity. Materials (Basel), 2020; 13:354.

31.

Rasband

. Circularity. 2000. Available at https://www.imagej.nih.gov/ij/plugins/circularity.html (last accessed February 8, 2021).

32.

Zhang

, Zhou

, Deng

, et al. Pore architecture and cell viability on freeze dried 3D recombinant human collagen-peptide (RHC)–chitosan scaffolds. Mater Sci Eng C, 2015; 49:174–182.

33.

Schneider

, Rasband

, Eliceiri

. NIH Image to ImageJ: 25 years of image analysis. Nat Methods, 2012; 9:671–675.

34.

. Im2mesh (2D image to triangular meshes). MATLAB Central File

Exchange

. 2020. Available at: https://www.mathworks.com/matlabcentral/fileexchange/71772-im2mesh-2d-image-to-triangular-meshes (last accessed September 25, 2020).

35.

Vimalraj

. Alkaline phosphatase: Structure, expression and its function in bone mineralization. Gene, 2020; 754:144855.

36.

Scotti

, Dunand

. Freeze casting—a review of processing, microstructure and properties via the open data repository, FreezeCasting.net. Prog Mater Sci 2018;94:243–305.

37.

O'Brien

, Harley

, Yannas

, et al. Influence of freezing rate on pore structure in freeze-dried collagen-GAG scaffolds. Biomaterials, 2004; 25:1077–1086.

38.

Lee

, Grodzinsky

, Spector

. The effects of cross-linking of collagen-glycosaminoglycan scaffolds on compressive stiffness, chondrocyte-mediated contraction, proliferation and biosynthesis. Biomaterials, 2001; 22:3145–3154.

39.

Buttafoco

, Engbers-Buijtenhuijs

, Root

, et al. First steps towards tissue engineering of small-diameter blood vessels: Preparation of flat scaffolds of collagen and elastin by means of freeze drying. J Biomed Mater Res Part B Appl Biomater, 2006; 77:357–368.

40.

Murphy

, Haugh

, O'Brien

. The effect of mean pore size on cell attachment, proliferation and migration in collagen–glycosaminoglycan scaffolds for bone tissue engineering. Biomaterials, 2010; 31:461–466.

41.

Wang

, Yang

, Koons

, et al. Tuning pore features of mineralized collagen/PCL scaffolds for cranial bone regeneration in a rat model. Mater Sci Eng C, 2020; 106:110186.

42.

Marmorat

, Arinstein

, Koifman

, et al. Cryo-imaging of hydrogels supermolecular structure. Sci Rep, 2016; 6:25495.

43.

Rahbani

, Behzad

, Khashab

, et al. Characterization of internal structure of hydrated agar and gelatin matrices by cryo-SEM. Electrophoresis, 2013; 34:405–408.

44.

O'Brien

, Harley

, Waller

, et al. The effect of pore size on permeability and cell attachment in collagen scaffolds for tissue engineering. Technol Health Care, 2006; 15:3–17.

45.

Lin

Y-C

, Tan

, Marra

, et al. Synthesis and characterization of collagen/hyaluronan/chitosan composite sponges for potential biomedical applications. Acta Biomater, 2009; 5:2591–2600.

46.

Harley

, Leung

, Silva

ECCM

, et al. Mechanical characterization of collagen-glycosaminoglycan scaffolds. Acta Biomater, 2007; 3:463–474.

47.

Lin

, Lampi

, Reinhart-King

, et al. Eigenstrain as a mechanical set-point of cells. Biomech Model Mechanobiol, 2018; 17:951–959.

48.

Murphy

, O'Brien

. Understanding the effect of mean pore size on cell activity in collagen-glycosaminoglycan scaffolds. Cell Adhes Migr, 2010; 4:377–381.

49.

Zhang

, Lu

, Kawazoe

, et al. Pore size effect of collagen scaffolds on cartilage regeneration. Acta Biomater, 2014; 10:2005–2013.

50.

Murphy

, Matsiko

, Haugh

, et al. Mesenchymal stem cell fate is regulated by the composition and mechanical properties of collagen–glycosaminoglycan scaffolds. J Mech Behav Biomed Mater, 2012; 11:53–62.