Collagen hydrogels with controllable combined cues of elasticity and topography to regulate cellular processes

A schematic of the ET-RaM process is shown in figure 1. PDMS molds were prepared in advance as follows. The precursor of PDMS (SIM-260) was mixed with a curing agent (CAT-260) (both from Shin-Etsu Chemical) at a 10:1 ratio, and then degassed and spin-coated on a silicon master mold (DTM1-1; Kyodo International) at 1000 rpm. After curing at 150 °C for 30 min, the molds were peeled from the master mold. A 5 mg ml⁻¹ collagen I solution (porcine skin, Collagen BM, #639-30861; Nitta Gelatin) was concentrated to 50 mg ml⁻¹ by evaporating the solvent at room temperature (∼25 °C). The gelatin (porcine skin, Type A, G1890; Sigma-Aldrich) and collagen peptide (porcine skin, Type A; Nitta Gelatin) solutions were prepared by dissolving in deionized water (supplied by a Millipore Milli-Q system) at 10 wt% by heating at 50 °C for 30 min. After pouring the solution into cell culture dishes (Iwaki 1000-035, non-treated, 35 mm; AGC Techno Glass) (figure 1, Step 1), micro-patterned flexible PDMS molds loaded onto a plastic cover such as polyethylene-terephthalate film were placed in the solution (figure 1, Step 2). The samples were stored overnight at 4 °C for collagen I and peptide and at 20 °C for gelatin. During storage, gelatin and peptide underwent physical gelation. The samples in sealed bags were irradiated with ⁶⁰Co γ-rays (⁶⁰Co No.2 Irradiation Facility of Takasaki Advanced Radiation Research Institute, QST) in air at 15 °C–20 °C at a dose rate of 10 J g⁻¹ h⁻¹ (J g⁻¹ = kGy) for ⩾10 J g⁻¹ and at 8 J g⁻¹ h⁻¹ for 8 J g⁻¹ irradiation (figure 1, Step 3). The samples were detached from the PDMS molds and immersed in PBS (pH 7.4, 10010023; Thermo Fisher Scientific) (figure 1, Step 4) at 37 °C for ⩾3 days for collagen I, or at 50 °C for 2–3 h for gelatin and peptide to remove non-cross-linked components and to allow them to equilibrate. Collagen hydrogels prepared by ET-RaM were used for cell culture after immersion in culture medium at 37 °C for 1 h to replace the absorbed PBS (figure 1, Step 5).

Zoom In Zoom Out Reset image size

Photographs of hydrogels and bright-field images of hydrogel microtopography were obtained using a digital camera (DMC-TZ85; Panasonic) and an inverted microscope (IX83; Olympus) with a digital camera (AdvanVision), respectively.

To determine the final collagen concentration in the hydrogels, the swollen samples after Step 4 were filtered through a stainless steel 200-mesh filter (mesh size: 75 μm) and weighed (W_s). After vacuum-drying overnight at 30 °C and weighing the dried sample (W₀), the final collagen concentration, C (%), was calculated as (W₀/W_s) × 100.

The elasticity of collagen hydrogels was measured using an indentation tester (RE2-3305B; Yamaden) at 37 °C following Step 5. The samples were compressed to 30% deformation with a 2 N load cell using a plunger (ϕ3 mm or ϕ8 mm) at 50 μm s⁻¹. The compressive modulus was then determined from the slope of the stress–strain curves over the strain range from 0% (surface) to the % value equivalent to a depth of 100 μm in each sample.

The chemical structures of gelatin before and after the ET-RaM process were compared using Fourier transform infrared (FT-IR) spectroscopy. As the pristine sample, 10 wt% physical gelatin gel was vacuum-dried overnight at 30 °C. The hydrogels produced by the ET-RaM process were washed in Milli-Q water at 50 °C for 2–3 h and vacuum-dried overnight at 30 °C. FT-IR spectra were collected using an IRAffinity-1 S instrument (Shimadzu) with a DuraSampl IR-II single-reflection diamond attenuated total reflection attachment (Smiths Detection) at a resolution of 1 cm⁻¹ and averaged over 32 scans.

The biodegradation properties of gelatin hydrogels were analyzed using collagenase (034-22363; Fujifilm Wako Pure Chemical). The hydrogels produced by the ET-RaM process were used after Step 4. The pristine samples and hydrogels in 35 mm dishes (initial volume of 1.5 ml) were incubated in collagenase solution (0.04% in PBS) at 37 °C. At different time points, the samples were filtered through a stainless steel 200-mesh filter and washed with Milli-Q water. After vacuum-drying overnight at 30 °C, the samples were weighed (W_t), and the percentage of remaining weight was calculated as (W_t /W₀) × 100.

Amino acid analysis of the gelatin hydrogels was carried out by acidic hydrolysis using a PICO-TAG Work Station (Waters) followed by high-performance liquid chromatography (HPLC). After irradiation in Step 3, the samples were washed in Milli-Q water at 50 °C for 48 h and vacuum-dried at 30 °C for 24 h. The samples and pristine gelatin powder (1 mg each) in separate glass test tubes were transferred to a reaction vial containing 0.2 ml of 1% phenol (161-01025; Fujifilm Wako Pure Chemical) in 6 M HCl (080-01066; Fujifilm Wako Pure Chemical). The reaction vial was tightly closed and heated at 110 °C for 21 h under vacuum. The test tubes were then vacuum-dried to remove any remaining HCl, and the obtained hydrolysates were filtered through a 0.45 μm membrane (SFCA033045S; AS ONE) after adding 20 ml of Milli-Q water to each tube. Amino acid standard samples were prepared from an amino acid mixture standard solution (013-08391; Fujifilm Wako Pure Chemical) containing 2.5 × 10⁻⁶ M of 17 amino acids diluted 1:20 with Milli-Q water and filtered through a 45 μm membrane. Borate buffer (pH 9.50) was prepared by mixing 0.05 M boric acid (021-02195) and 0.05 M NaOH (198-13765) (both from Fujifilm Wako Pure Chemical) at a 2:1 volume ratio. Each 10 μl sample solution was mixed with 20 μl of borate buffer and 20 μl of 0.05 M 4-fluoro-7-nitrobenzofurazan (342-04751; Dojindo Laboratories)/acetonitrile (015-08633; Fujifilm Wako Pure Chemical). The solutions were heated at 60 °C for 1 min and then cooled at −18 °C for 2 min; 50 μl of 0.3 M HCl was then added to terminate the labeling reaction. A 10 μl volume of each labeled sample was analyzed by HPLC (Model 1100; Agilent Technologies) with an InertSustainSwift C18 column (GL Sciences) and a fluorescence detector (2475; Waters) at 40 °C. The HPLC column was eluted with 0.1% trifluoroacetic acid (208-02741; Fujifilm Wako Pure Chemical) and acetonitrile with the gradient mixture ratio as recommended by GL Sciences (data no. LB316-0848) at a flow rate of 1.0 ml min⁻¹.

Madin-Darby canine kidney (MDCK) epithelial cells (RCB0995), HeLa human cervical carcinoma epithelial cells (RCB0007), and 3T3-Swiss albino mouse embryonic fibroblasts (RCB1642) were obtained from the RIKEN BRC Cell Bank. C2C12 cells (EC91031101) were obtained from DS Pharma Biomedical. Cardiomyocytes were obtained from neonatal Wistar rats (Japan SLC) according to our published protocol [31]. Animal experiments were approved by the Institutional Animal Care and Use Committee of QST (approval no. 18-T001-1) and were performed in accordance with the Fundamental Guidelines for Proper Conduct of Animal Experiments and Related Activities in Academic Research Institutions under the jurisdiction of the Ministry of Education, Culture, Sports, Science, and Technology of Japan.

MDCK and HeLa cells were cultured in minimum essential Eagle's medium (MEM) (M5650; Sigma-Aldrich) supplemented with 10% fetal bovine serum (FBS) (SH30910.03; GE Healthcare Japan and 10437-028; Thermo Fisher Scientific, respectively), 100 U ml⁻¹ penicillin, 100 μg ml⁻¹ streptomycin (15140-122; Thermo Fisher Scientific), and 0.002 M L-glutamine (10378016; Thermo Fisher Scientific and G7513; Sigma-Aldrich, respectively). 3T3-Swiss cells, C2C12 cells, and cardiomyocytes were cultured in Dulbecco's modified Eagle's medium (DMEM, 08488-55; Nacalai Tesque) supplemented with 10% FBS, 100 U ml⁻¹ penicillin, 100 μg ml⁻¹ streptomycin, and 0.002 M L-glutamine (Sigma-Aldrich). A 2 ml volume of each cell suspension (1 × 10⁴ MDCK, HeLa, or 3T3-Swiss cells ml⁻¹), 5 × 10⁴ C2C12 cells ml⁻¹, or 1.5 × 10⁵ cardiomyocytes ml⁻¹ prepared using the Countess II FL automated cell counter (AMQAF1000; Thermo Fisher Scientific) was introduced onto collagen hydrogels in 35 mm cell culture dishes without any reagent coating, or polystyrene (PS) cell culture dishes (Iwaki 3000-035, treated, 35 mm; AGC Techno Glass) without protein coating. To induce the differentiation of C2C12 cells, the culture medium was replaced with differentiation medium composed of DMEM supplemented with 2% horse serum (26050-070; Thermo Fisher Scientific), 100 U ml⁻¹ penicillin, 100 μg ml⁻¹ streptomycin, 0.002 M L-glutamine (Sigma-Aldrich), and 1 μg ml⁻¹ insulin (I0516; Sigma-Aldrich) 2 d after seeding. The differentiation medium was replenished every 1–2 days. The cells were cultured at 37 °C in 5% CO₂.

Nuclei, the actin cytoskeleton, and sarcomeres were fluorescently labeled to analyze cell responses to collagen hydrogels. On day 3 (MDCK, HeLa, and 3T3-Swiss), 6 (C2C12), or 4 (cardiomyocytes) of cell culture, samples were washed with PBS, and cells were fixed with 4% paraformaldehyde (163-20145; Fujifilm Wako Pure Chemical) for 10–15 min. After washing with PBS, the cells were incubated in PBS containing 0.1% Triton X-100 (35501-02; Nacalai Tesque) for 5 min and washed with PBS. MDCK, HeLa, and 3T3-Swiss cells were incubated in PBS containing 0.2 μg ml⁻¹ tetramethylrhodamine B isothiocyanate (TRITC)-conjugated phalloidin (P1951; Sigma-Aldrich), 2 μg ml⁻¹ 4',6-diamidino-2-phenylindole (DAPI) (D523; Dojindo Laboratories), and 1% bovine serum albumin (BSA) (019-27051; Fujifilm Wako Pure Chemical) for 20 min. C2C12 cells were incubated in PBS containing 0.1 μg ml⁻¹ TRITC-conjugated phalloidin, 1 μg ml⁻¹ DAPI, and 1% BSA (P-6154; Biowest) for 20 min. To visualize sarcomeres, cardiomyocytes were incubated in blocking solution composed of PBS with 1% BSA (Fujifilm Wako Pure Chemical) for 60 min, followed by blocking solution containing mouse anti-α-actinin antibody (1:300, A7811; Sigma-Aldrich) for 60 min. After washing with blocking solution, cardiomyocytes were incubated in blocking solution containing Alexa Fluor® Plus 488-conjugated goat anti-mouse IgG (1:300, A32723; Thermo Fisher Scientific) for 60 min. To visualize the surface of collagen hydrogels, samples were washed with PBS and incubated for 10 min in PBS containing 1 mg ml⁻¹ of 0.2 μm green fluorescent microspheres (F8811) or 0.2 mg ml⁻¹ of 0.2 μm red fluorescent microspheres (F8810) (both from Thermo Fisher Scientific). The samples were then washed with PBS and examined with an upright microscope (BX61WI) with a disk scanning unit (BX-DSU), objective lens (LUMPLFLN 60XW) (all from Olympus), and a complementary metal oxide semiconductor camera (ORCA-Flash4.0 V3; Hamamatsu Photonics K.K.). Staining and observation were performed at room temperature (∼25 °C). Three-dimensional (3D) image stacks were deconvoluted using cellSens software (Olympus). Image processing and 3D rendering were performed using ImageJ software (National Institutes of Health).

Cell morphology was analyzed in terms of spread area, aspect ratio (length of major axis/length of minor axis), and orientation (angle from 0° to 90°) from bright-field images randomly acquired on an inverted microscope (IX70; Olympus) with a CellPad E digital camera system (Allied Scientific Pro) using ImageJ software (n = 3). For the proliferation assay, HeLa cells cultured for 2 h or 1–3 days were stained with 1 μg ml⁻¹ Hoechst 33342 solution (H342; Dojindo Laboratories). Nuclei were visualized using a light source (130 W mercury lamp, U-HGLGPS) and mirror units (U-MWU2) (both from Olympus), and the number of HeLa cells was counted from randomly acquired fluorescence images (n = 3). The viability of HeLa cells on day 3 was analyzed after staining with 1 μg ml⁻¹ calcein-AM (C396; Dojindo Laboratories) and 1 μg ml⁻¹ DAPI. Viable and dead cells were visualized using mirror units (U-MWU2 and NIBA; Olympus), and the percent viability was calculated (n = 3).

The orientation of the actin cytoskeleton was analyzed using a Fourier transformation-based method. An optical section of the actin cytoskeleton beneath the nucleus was first extracted from z-stack images and cropped to 128 × 128 pixels (13.9 μm²). The 2D Fourier spectrum for the cropped image was then calculated. After removing the direct current offset, a histogram in polar coordinates was generated from the Fourier spectrum data, which was approximated by an ellipse. The orthogonal direction of the major axis of this ellipse was defined as the orientation direction of the actin filaments. The angular difference between this and the major axis of the microgrooves on collagen hydrogels was defined as the orientation angle of the actin cytoskeleton relative to the microgrooves. The 2D Fourier spectrum was calculated with MATLAB (MathWorks) software, and other image analyses were performed using a custom-made macro in ImageJ software.

Data were analyzed using Welch's t-test and one-way analysis of variance followed by a post-hoc Tukey's honestly significant difference test using OriginPro2018J software (OriginLab). The analyzed datasets of cell responses are summarized in (supplementary table S1 (available online at stacks.iop.org/BMM/16/045037/mmedia)). All experiments were performed at least three times.

To produce collagen hydrogels without using chemical cross-linkers, we applied radiation-induced chemical cross-linking reactions. Although collagens are dominantly decomposed by ionizing radiation in a dried state [32–34], cross-linking can also be induced in the presence of water [32–35]. At low concentrations (<1%), the probability of decomposition is still higher than that of cross-linking [36], owing to the long intermolecular distances, and cross-linking only produces nano-/micro-particles in low yield. We found that at a certain concentration (a few to several 10 s%) in water, both in the solution state and physically formed hydrogel state, both ECM-derived native collagen and hydrolyzed collagens are predominantly cross-linked through a reaction with hydroxyl radicals generated in water. Water-containing 3D polymeric networks, that is, chemically cross-linked collagen hydrogels, are thus formed without requiring a cross-linking agent.

We developed an ET-RaM technique (figure 1, see section 2) after optimizing the production of collagen hydrogels by radiation-induced chemical cross-linking with high yield and high reproducibility. As an ionizing radiation source, we chose ⁶⁰Co γ-rays, which are commonly used to sterilize medical products. Owing to their high penetrability, γ-rays provide a uniform dose in samples, even in batch processing. Flexible PDMS molds placed before radiation cross-linking were easily demolded from the soft collagen hydrogels. ET-RaM successfully altered ECM-derived collagen I and hydrolyzed collagen (gelatin or collagen peptide) solutions into micropatterned hydrogels without using chemical cross-linkers. The irradiation dose required for gelation was roughly the same range as that usually adopted for radiation sterilization (4–25 J g⁻¹) [37], as described later.

First, the chemical properties of the obtained collagen hydrogels were analyzed to determine whether the irradiation process influences the intrinsic functionality of collagen. Figure 2(A) shows the FT-IR spectra of gelatin before and after ET-RaM processing with irradiation doses of 60 and 160 J g⁻¹. The amide bands of polypeptides (A: 3297 cm⁻¹, B: 3075 cm⁻¹, I: 1630 cm⁻¹, II: 1537 cm⁻¹, and III: 1237 cm⁻¹) related to peptide bonds and helical structure remained essentially unaltered by ET-RaM. A slight shift to a lower wavenumber for amide A and amide II after irradiation at 160 J g⁻¹ suggests a small change in the secondary structure of gelatin. In vitro enzymatic biodegradation of the gelatin hydrogel was evaluated using collagenase (figure 2(B)). Biodegradability was maintained after the ET-RaM process, although it slowed down with increasing irradiation dose, that is, with increasing matrix density. These results were supported by the observation that the amino acid composition ratio of 15 amino acids, including R, G, D, and E, which comprise cell-binding motifs, were almost unchanged by ET-RaM at doses of 30 and 60 J g⁻¹ (supplementary table S2). These chemical analyses showed that ET-RaM ensured the maintenance of the intrinsic functionality of collagen.

Zoom In Zoom Out Reset image size

Collagen hydrogels with a regulated elastic modulus, E, were obtained by tuning the irradiation dose of ET-RaM from 8 to 120 J g⁻¹. E values determined by the indentation test (supplementary figure S1) ranged from 1 to 236 kPa (figure 3(A)) without the addition of any reagent. Collagen I and gelatin showed similar E values at the same irradiation dose, while collagen peptide, which has the lowest molecular weight among the three collagen types, required the highest irradiation dose to form hydrogels. Higher irradiation doses resulted in higher cross-linking densities, thus resulting in higher E values. E showed power law-dependence on the final concentration of collagen in the hydrogels, C, which ranged from 0.3% to 14%, except for the softest hydrogel obtained with collagen I (figure 3(B)).

Zoom In Zoom Out Reset image size

We obtained micropatterned collagen hydrogels ranging from 7 to 236 kPa. Microtopographies such as microgrooves with widths and spacings of 1, 5, and 10 μm and a height of 2 μm, were successfully formed on the hydrogel surfaces (figure 3(C)). The softest hydrogel in which micropatterns were observed had an E of 1.2 kPa, which corresponds to the stiffness of the human brain [11]; however, the patterns easily collapsed due to their honey-like fluidity (figure 3(D)). Interestingly, the volume of collagen I hydrogels reduced during soaking in PBS at 37 °C (figure 1, Step 4), and reached an equilibrium after 3 d (figure 3(E)). Since the compaction of collagen I was isotropic, microtopographies were formed with a fixed reduction rate; the width of the transferred microgrooves on hydrogels produced with 20 J g⁻¹ irradiation decreased from 5 and 10 μm to 3.6 and 7.2 μm, respectively. On the other hand, mold patterns were directly transferred to gelatin and collagen peptide hydrogels. As a result, our collagen hydrogels covered a broad range of elasticities and topographies of the native ECM of in vivo soft tissue (figure 3(F)), which were never achieved with conventional dishes, collagen hydrogels, or Matrigel [8, 38].

As described above, the developed collagen hydrogels can combine elastic, topographic, and compositional cues. We used collagen hydrogels to investigate cell responses to chemical and physical complex cues, especially soft topographic cues, such as those existing in vivo.

Gelatin was used as the base polymer owing to the broad range of E and precise microtopography without the random fibrous structures of collagen I that could affect the cellular response. We analyzed the morphology and actin cytoskeleton organization of MDCK, HeLa, and 3T3-Swiss cells grown on uncoated 7, 47, 148, and 236 kPa collagen hydrogels produced by ET-RaM at irradiation doses of 10, 20, 40, and 60 J g⁻¹. These E values were selected to cover the range of soft tissues and to ensure precise micropatterning: 7, 47, 148, and 236 kPa roughly correspond to the stiffness of the human kidney, cardiac muscle, skeletal muscle, and skin, respectively [11]. The E values of collagen hydrogels originated from C of 0.3%, 1.8%, 6.2%, and 8.0%, respectively (figure 3(B)). Hydrogels with a flat surface or with microgrooves (width and spacing = 5 μm, height = 2 μm) were prepared for each E. The topographical cues of sub-micrometers to 10 μm were considered to act on the actin cytoskeleton [1]. Within this range, we selected 5 μm microgrooves, owing to its high reproducibility on all the above E values.

First, we investigated cell responses to elastic cues using flat collagen hydrogels. The morphology of MDCK, HeLa, and 3T3-Swiss cells changed drastically on day 1 after seeding depending on the E (supplementary figures S2–S4). Cells became rounded and formed aggregates over 3 days on the softest 7 kPa (C = 0.3%) hydrogels (figure 4(A) and supplementary figures S3, S4). We found that MDCK and HeLa cell aggregates partially infiltrated the 7 kPa hydrogels (figure 4(A) and supplementary figure S3). 3T3-Swiss cells extended pseudopodia and actively migrated into these hydrogels (figure 5 and supplementary figure S4), which were similar to their movement between micropillars [39, 40]. On the contrary, with increasing E (i.e. with increasing C), these cells chose to stay atop the collagen hydrogels and gradually flattened and spread over a larger area (figures 4(A), (B) and supplementary figures S2–S4). The morphology of HeLa cells became more spindle-shaped with increasing E. The morphology on stiff collagen hydrogels was different from that observed on a PS dish. The flattened cells and aggregates adhered to the hydrogels. Proliferation over 3 days was independent of E and comparable to that of cells grown on a PS dish (supplementary figure S5).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Next, we combined specific topographic cues with elastic cues from collagen hydrogels. Microgrooves (width and spacing = 5 μm, height = 2 μm) were formed on the collagen hydrogels with the same E (7, 47, 148, and 236 kPa) as the flat ones. Interestingly, these soft topographic cues induced all cell types to spread and become aligned parallel to the microgrooves, even on the softest 7 kPa (C = 0.3%) hydrogels, where cells formed aggregates on the flat surface (figure 4(A) and supplementary figures S3, S4), except for some 3T3-Swiss cells that migrated into the hydrogels. The morphological aspect ratio and orientation angle relative to the longitudinal direction of the microgrooves significantly differed from those of cells on flat hydrogels (figures 4(C), (D) and supplementary figures S2–S4). These results indicate that the cell morphology conformed to the microtopographic cues even when they are as soft as in vivo ECM. Stiffer hydrogels structurally constrained the alignment of the cells effectively, which was clearly shown by the higher aspect ratio and smaller variation in the orientation angle of MDCK and 3T3-Swiss cells (figures 4(C), (D) and supplementary figures S2, S4). In contrast, the response of HeLa cells to topography was independent of E due to the drastic morphological change observed on both flat and micropatterned collagen hydrogels (supplementary figure S3). The organization of the actin cytoskeleton was also affected by the soft topographic cues, with actin filaments in MDCK, HeLa, and 3T3-Swiss cells aligned parallel to the microgrooves (figures 4(E), (F) and supplementary figures S3, S4). The effect of the microgrooves on the actin filament orientation was independent of the E values, unlike that on the cell morphology.

Collagen hydrogels with tunable elasticity and microtopography can be utilized to induce and control diverse cell behaviors in vitro. Figure 6 shows rat cardiomyocytes and C2C12 mouse myoblasts seeded onto collagen hydrogels. Both cells adhered well to the hydrogels without any reagent coating and were aligned parallel to the microgrooves. Aligned sarcomeres (the actin-based internal structure) that drive muscle contraction were formed in cardiomyocytes (figure 6(A)). C2C12 myoblasts formed aligned myotubes after culture in differentiation medium for 4 days (figure 6(B)). These cells reacting to soft topographic cues recapitulated their morphology and cytoskeleton as observed in vivo.

Zoom In Zoom Out Reset image size