3D skeletal muscle fascicle engineering is improved with TGF-β1 treatment of myogenic cells and their co-culture with myofibroblasts

Cell and tissue culture

As shown in the experimental design listed in Fig. 1A, cultures consisting solely of human neo-natal dermal fibroblasts (passage 4 hDFs, PCS-201-010; ATCC, Manassas, VA, USA) or mouse myoblasts (pre-passage 4 C2C12s, CRL-1772, ATCC) were plated on standard cell culture plates at 1 k/cm² in growth media (GM) consisting of DMEM (Gibco, Waltham, MA, USA), 10% fetal bovine serum (FBS; Hyclone, Pittsburgh, PA, USA), 1% penicillin/streptomycin (P/S, Gibco). To promote differentiation of fibroblasts to myofibroblasts, some fibroblasts were incubated with 1 ng/mL TGF-β1 (Peprotech, Rocky Hill, NJ, USA) for 6 days between P4 and P5. C2C12s and P4 fibroblasts and myofibroblasts were trypsinized, resuspended in GM, and seeded at a 1:1 ratio on standard cell culture plates at 20 k/cm² or self-assembled into 3D tissue constructs. To develop 3D self-assembled tissues, cell suspensions were added into custom 2% agarose molds that were prepared by casting molten agarose onto a patterned PDMS mask. After solidification of the gel, the molds were sectioned and added to cell culture plates. Each tissue mold is composed of a hollow well with a 2 mm diameter central agarose post (Fig. 1A). When a cell suspension is added to the well, the cells will prefer to adhere to each other instead of the agarose, and self-assemble into a ring-shaped tissue structure that contracts around the central post. A cell seeding density of 350,000 cells/millimeter of post diameter is used per tissue sample. This density was selected through a pilot experiment determining the concentration of cells required to self-assemble tissue around the circumference of the agarose post, which is a function of the post’s diameter (C Malcuit, pers. comm., 2013). Hydrogels were produced from a protocol adapted from Langelaan et al. (2011) where 1.5 × 10⁶ C2C12s alone or a 1:3 ratio of fibroblasts/myofibroblasts and C2C12s were resuspended in 42% GM, 54% rat collagen type 1 (3 mg/mL; Corning Life Sciences, Corning, NY, USA), and 2.7% NaOH (0.1 M). The hydrogel mixture was poured between house-shaped Velcro anchors super glued to the bottom of a six well plate and gelled for 45 min before adding GM. After 24 h, the media were switched in both 2D and 3D conditions to differentiation medium (DM) composed of DMEM, 2% horse serum (Sigma-Aldrich, St. Louis, MO, USA), and 1% P/S, with or without 1 ng/mL TGF-β1. Media was changed every 2–3 days for 7 days, after which cells and tissues underwent qPCR, ICC or IHC, or mechanical characterization. See Fig. 1B for listing of experimental groups.

Gene expression analysis

Total RNA was isolated from cultures using TriZol reagent (Life Technologies, Carlsbad, CA, USA) according to the manufacturers recommended instructions. cDNA was reverse transcribed from 1 µg total RNA using qScript SuperMix (Quanta Biosciences, Gaithersburg, MD, USA) according to manufacturers instructions. qPCR was carried out using a 5 ng equivalent of cDNA in a 1X reaction of PerfeCTa SYBR Green SuperMix (Quanta Biosciences, Beverly, MA, USA) and 250 nM each (forward and reverse) custom oligonucleotide primers (Integrated DNA Technologies, Coralville, IA, USA) generated by using PrimerBlast (National Center for Biotechnology Information, Bethesda, MD, USA). Reactions were carried out on an Eppendorf RealPlex2 qPCR system, and fold changes in gene expression were calculated using the ΔΔCT method using species-specific GAPDH primers. Primer sequences are listed in Table S1. Human collagen 1 and α-SMA expression levels in the 2D FibCon condition were used as the reference group for fibroblast gene expression, and mouse myogenin gene expression levels of 2D MyoCon were used as a reference for myoblasts. N = 5 for all groups.

Immunostaining

2D and 3D cultures were assessed qualitatively with immunostaining. 2D cultures were fixed with 4% formaldehyde for 20 min, and 3D cultures were fixed for 2 h. 3D self-assembled tissues were then submerged in OCT media, frozen for cryosectioning, sectioned at 30 μm thick, and applied to glass slides. Hydrogels remained unsectioned for whole-mount imaging. After fixing, 2D and 3D samples were permeabilized with 0.1% Triton X-100 in PBS for 20 min, and then incubated with a blocking solution consisting of 5% FBS in PBS for 20 min. Following blocking, samples containing myoblasts were stained with rhodamine phalloidin (R415, 1:40 dilution; Life Technologies, Carlsbad, CA, USA) and/or myosin heavy chain (MF-20, 1:50; Developmental Studies Hybridoma Bank, Iowa City, IA, USA), procollagen (M-38, 1:50; Developmental Studies Hybridoma Bank) or α-SMA (A 2547, 1:1,000; Sigma-Aldrich, St. Louis, MO, USA) primary antibodies. Primary antibody staining was followed by incubation with Alexa-Fluor 488 or 564 secondaries (1:250; Life Technologies, Carlsbad, CA, USA) and DAPI (1:50,000; Invitrogen, Carlsbad, CA, USA). Imaging of 2D samples was completed with an Olympus IX81 microscope (Olympus corporation of the Americas, Center Valley, PA, USA), 3D self-assembled tissues were imaged with an Olympus Fluoview FV1000 Confocal Microscope, and hydrogels were imaged with an Olympus Fluoview FV1000MPE multiphoton microscope. All images were processed with ImageJ (Version 1.48; National Institute of Health, Bethesda, MD, USA). The total area of procollagen staining was measured in 2D FibCon and FibTGFβ images and divided by the total number of nuclei to find the average area of procollagen staining per cell. Increased area procollagen signal was extrapolated as a change of in cell shape and increased cell spreading. Circularity of DAPI stained fibroblast nuclei was measured to compare changes in cyto- and nucleoskeletal shape. N = 13 images for each group. 2D MyoCon and CoCon images were also processed to measure changes in myotube alignment. In each image, the myotubes formed angles with a common axes (x or y). The standard deviations of the angles from each image were measured to compare the variability in each group, with smaller averaged standard deviations for one sample image indicating higher myotube alignment. N = 8–9 images for each sample.

Slack tests

The slack test method has been adapted by our lab to indirectly measure and compare tissue integrity, ECM maturity, and muscle fiber development between engineered samples. Two slack test methods were employed. In the whole tissue slack test, self-assembled tissue constructs were released from tension in culture and their shortening lengths were recorded over time. The initial lengths (μm, L_i) of rings were recorded across their longest axis while they remained on their posts. Rings were then removed from the agarose posts, suspended in PBS (37 °C) on a submerged horizontal microbeam from the Microsquisher system (CellScale, Ontario, Canada), and points across their longest axis for 27 min. The final lengths (L_f) were recorded and percent shortening (%) was calculated with the formula % = 100*(L_f/L_i). Tissues were then stored in their respective medias for 24 h at 37 °C and imaged with a stereoscope. N = 3–6 samples per group.

The cleaved tissue slack test was performed to release the tissues from internal tension, which may arise from tissue structuring patterns building passive forces. 3D MyoCon, MyoTGFB, CoCon, and CoTGFB tissue rings were cut with a scalpel to generate linear muscle constructs freed from their agarose posts and placed in a 37 °C bath of cell culture media. Tissue lengths were measured every 3 min for 15 min, then samples were allowed to slack over 24 h at 37 °C to give the final length (L_f). The initial length (μm, L_i) for each construct, the length of the uncut tissue sample still mounted on its post, was calculated as the circumference of the 2 mm agarose posts using the formula: L_i = 2 πr. The percent of original size (%) was determined with the formula: % = 100*(L_f/L_i). N = 4–5 samples per group.

Tissue elastic moduli assessments

For calculations of the elastic stiffness of self-assembled engineered tissues, two methods of mechanical characterization were used: a compression-based system to generate slopes from stress–strain curves and atomic force microscopy (AFM). Tissue constructs were sectioned into 4–5 pieces with diameters between 250 and 500 μm and were compressed in a PBS fluid bath (pH 7.4, 37 °C) using a Microsquisher (Cellscale, Waterloo, Ontario, Canada). Tissues were compressed to 40% of their original diameter at a rate of 1% per second using microbeams with diameters between 0.2 and 0.3048 mm. Force–displacement curves were generated and the slope of the linear portion of the curve was extrapolated as the elastic modulus of the tissue. Slopes were averaged for each 3D condition with N = 8–12 section samples per group.

Tissue elastic moduli were measured by an Asylum MFP3D-Bio AFM (Asylum Research, Santa Barbara, CA, USA) through a nano-indention method using MFP-3D software (Version 13.04.77). Force-distance curves were determined using Eqs. (1) and (2):

(1)${\displaystyle F\left(\delta \right)=\frac{4\sqrt{R}}{3}\frac{E}{1-{\nu }^2}{\delta }^{3∕2}}$ (2)${\displaystyle F\left(\delta \right)=\frac{E}{1-{\nu }^2}\frac{2tan\alpha }{\pi }{\delta }^2.}$ Where F is the measured force, E is the local Young’s modulus, R is the cantilever’s tip radius (for a spherical tip), α is the cantilever’s tip angle (for a cone tip), ν is the Poisson’s ratio of the sample (assumed as 0.5 for an incompressible material), and δ is the sample indentation. Pyramidal tips with a nominal tip radius 20 nm, 200 µm in length, 20 µm in width, and a tip semi-angle of 15°on silicon nitride triangular V-shaped cantilevers with a nominal spring constant of 0.06 N/m (DNP-10; Bruker Inc., Camarillo, CA, USA) were employed. The recorded force-distance curves were analyzed in MATLAB and statistical analysis was done using SPSS (Ver. 17.0; IBM, Somers, NY, USA).

Statistics

Statistical analyses were performed with two-tailed unpaired t-tests and one-way ANOVAs using GraphPad Prism (Version 4; GraphPad Software) and statistical significance was defined as p < 0.05. Means are reported with standard error bars in bar graphs. Skew coefficients were calculated in Excel (Version 14.6.4; Microsoft).

Fibroblast to myofibroblast differentiation is observed in 2D cultures treated with TGF-β1, regardless of myoblast presence

2D fibroblast-only cultures displayed increased α-SMA staining in TGF-β1-treated samples (Figs. 2A and 2B) and increased α-SMA gene expression (p < 0.01, Fig. 3A), indicating their differentiation to myofibroblasts. There was no significant difference in collagen 1 transcription between any 2D conditions containing fibroblasts or myofibroblasts (Fig. 3B), but post-translational control of the protein differed with TGF-β1 treatment. Fibroblasts in the control group were spindle-shaped and displayed smaller areas of procollagen signal per cell nuclei (Figs. 2C and 2I), while TGF-β1 treatment significantly increased the area of procollagen staining per nuclei by 32.2 % (p < 0.0001, Figs. 2D and 2I) and enhanced the secretion of procollagen out of the cell (red arrow in 2D). Nuclei circularity was also increased in 2D FibTGFβ samples (p < 0.0001, Fig. 2J), indicating a widening of the nucleoskeleton accompanying increased cell spreading, compared to the more elongated nuclear shape in control conditions. Enhanced cell proliferation was also observed with TGF-β1 treatment (p < 0.0001, Fig. 2K), as recorded by average number of nuclei per field of view. α-SMA gene expression in 2D CoCon samples was similar to transcript levels of the fibroblast-only control group, but was significantly upregulated in 2D CoTGFβ conditions (p < 0.01, Fig. 3A).

Myogenesis is suppressed in 2D cultures with TGF-β1 supplementation and in the presence of fibroblasts and myofibroblasts, yet fibroblasts increase myotube alignment

Robust myotube formation was present in 2D MyoCon samples (Fig. 2E), decreased with exposure to TGF-β1 (Fig. 2F) or in co-culture with fibroblasts (Fig. 2G), and completely inhibited in myofibroblast co-cultures supplemented with TGF-β1 (Fig. 2H). MYOG gene expression in the 2D MyoTGFβ and 2D CoCon conditions were similar to each other and both displayed significant MYOG downregulation in comparison to 2D MyoCon. 2D co-cultures exposed to TGF-β1 showed the lowest MYOG expression of any group (p < 0.01, Fig. 3C). Despite diminishing myogenesis, fibroblasts organized myotube formation by increasing their alignment in 2D CoCon samples (p < 0.001, Fig. 2L).

In comparison to fibroblasts grown on plastic, self-assembled tissues containing only fibroblasts have a suppressed ability to assume a myofibroblast phenotype

In comparison to 2D fibroblast-only controls, expression of procollagen was dramatically downregulated in fibroblast-only tissues, and this was not improved by TGF-β1 exposure (p < 0.01, Fig. 3B). α-SMA staining was absent in 3D FibCon samples (Fig. S1A), but was somewhat increased with TGF-β1 treatment (Fig. S1C). 3D FibCon and FibTGFβ α-SMA gene expression was transcriptionally downregulated to an even greater degree than procollagen, yet these values were not significantly different from each other (p < 0.01, Fig. 3A). Fibroblast/myofibroblast-only tissues were quite fragile; they would easily rupture during handling (arrow, Fig. 4D) and would sometimes collapse, losing their annular shape (Figs. 4C and 4F). Despite their fragility, their surfaces were smooth and homogenous (Fig. 4S). However, the average thickness of constructs containing myofibroblasts treated with TGF-β1 was 33.2% larger than respective controls (p < 0.01, Fig. 5F), similar to the myofibroblast hypertrophy seen in 2D FibTGFβ.

TGF-β1 supplementation improves myoblast differentiation and alignment in myoblast-only self-assembled tissues

3D myoblast-only groups differed in their degree of myogenesis and their myotube formation patterns. 3D MyoCon contained myotubes that had some degree of organization but generally weren’t aligned (Figs. 6A and 6B), whereas myoblast samples treated with TGF-β1 had myotubes that were aligned along their circumferential axis (white filled arrow, Figs. 6E and 6F). Gene expression profiles indicated that 3D MyoCon samples had significantly lower myogenin expression than both the 2D MyoCon control and 3D MyoTGFβ condition (p < 0.01, Fig. 3C), but the 2D control and 3D TGFβ treated condition were not significantly different from each other. α-SMA was present in both untreated (Figs. 6C and 6D) and TGF-β1 treated (Figs. 6G and 6H) myoblast constructs. Observable in the α-SMA stained sections are fissures and breaks in the tissue (white asterisks, Figs. 6C and 6G). Additionally, the surfaces of myoblast-only tissues presented nodular syncytium of fused myoblasts lacking the anchorage necessary to elongate and form myotubes (arrow, Figs. 6I–6K), and this was macroscopically observable (unfilled arrows, Fig. 4T). Heterogeneous tissue patterning is also seen here, with many ripples along the tissue surface, large budges (filled arrow, Fig. 4T), and regions of varying density.

In self-assembled co-cultures, fibroblasts and myofibroblasts homogenize tissue surfaces, myofibroblasts improve myotube formation, and TGF-β1 enhances myogenesis

Addition of fibroblasts or myofibroblasts smoothened the surface of co-culture tissues (Fig. 4U), similarly to fibroblast- or myofibroblast-only samples, and these constructs lacked the studding of nodular syncytium seen in myoblast-only tissues. Co-culture conditions (Figs. 4M and 4P) appeared visually denser than their fibroblast- (Figs. 4A and 4D) and myoblast-only (Figs. 4G and 4J) counterparts. Similar to the hypertrophy observed in 2D and 3D myofibroblast conditions treated with TGF-β1, CoTGFβ tissues were 27.9% thicker than CoCon (p < 0.01, Fig. 5F). While human fibroblast/myofibroblast α-SMA gene expression was decreased in all 3D conditions compared to the 2D FibCon group, it was significantly higher in 3D CoCon and CoTGFβ compared to 3D fibroblast-only tissues (p < 0.01, Fig. 3A), and co-culture transcription levels were not influenced by TGF-β1 supplementation. A non-species specific α-SMA antibody was visible within myotubes, where CoCon samples showed α-SMA within punctate and short myotubes (Fig. 7A), while those in CoTGFβ tissues were elongated and more mature (Fig. 7C), indicating that TGF-β1 facilitated the development of bundles of myotubes aligned in parallel. Additionally, collagen 1 expression was increased in 3D co-cultures compared to 3D fibroblast/myofibroblast-only cultures (p < 0.01, Fig. 3B), but CoCon constructs were not significantly different from the 2D FibCon condition. Interestingly, collagen 1 gene expression was highest in 3D CoTGFβ samples, which was also the condition of highest MYOG expression (p < 0.01).

3D CoCon constructs containing fibroblasts frequently presented dense bulges with thickened regions of cells composed of differentiating myoblasts (Figs. 7E and 7F) between thinner regions absent of MyHC staining (Figs. 7H and 7I); this heterogeneous tissue architecture can also be seen macroscopically (arrow, Fig. 4M). CoCon constructs contained myotubes that were short but numerous. Some myotubes fused from one wall of the tissue to another, so that during sectioning, cross sections of myotubes are apparent (filled arrows, Figs. 7E and 7F). In contrast, CoTGFβ samples containing myofibroblasts had myotubes that were thicker and more multinucleated (Figs. 7G, 7J and 7K). These tissues also had many cross-sections of myotubes (filled arrow, Fig. 7J), but additionally contained much longer circumferentially aligned myotubes than in control co-culture samples (unfilled arrow, Fig. 7G). MYOG expression in 3D CoTGFβ constructs was significantly higher than all other groups (p < 0.01, Fig. 3C), while MYOG expression in 3D CoCon constructs was suppressed compared to 2D MyoCon (p < 0.01), and was not significantly different from 3D MyoCon. This data demonstrates that addition of fibroblasts with myoblasts in 3D co-culture does not improve myogenesis beyond what is observed in cultures consisting solely of myoblasts, and TGF-β1 supplementation is a myogenesis-promoting factor.

Zero-force velocity is greatest in co-culture with TGF-β1 supplementation in slack tests

Slack tests are a measurement of zero force velocity tissue shortening resulting from the recoil of stretched elastic elements and myofibrillar filament sliding in sarcomeres. Slack tests assess acto-myosin kinetics that are independent of Ca²⁺ activation, which provides biomechanical information about extracellular matrix networking, MHC isoform and muscle fiber type, and muscle organ characteristics (Claflin & Faulkner, 1985; Josephson & Edman, 1998; Reggiani, 2007). In whole tissue slack tests, the initial maximum lengths of self-assembled tissue constructs were recorded prior to removal from their posts (white arrow, Fig. 5A). Constructs were then released from tension (Figs. 4A, 4D, 4G, 4J, 4M and 4P), their lengths measured after 27 min (Figs. 4B, 4E, 4H, 4K, 4N and 4Q), and allowed to slack over 24 h (Figs. 4C, 4F, 4I, 4L, 4O and 4R). While all groups shortened significantly from their original size after 27 min (p < 0.001, Fig. 5B), tissues containing only fibroblasts or myofibroblasts shortened the least, with 3D FibCon remaining 87.3% and FibTGFβ 85.4% their original size. Fibroblast/myofibroblast tissues were the only cell-content matched tissue group that did not display significantly different final length values due to TGF-β 1 treatment. Myoblast tissues underwent significantly greater shortening than fibroblast/myofibroblast samples: MyoCon rings shrank to 69.4% and MyoTGFβ to 62.2% their initial diameter. Co-culture tissues underwent the greatest alteration, however. CoCon samples had the most significant shortening compared to all other groups (48.7%), and CoTGFβ was the second most shortened group (55.1%). Interestingly, after 24 h, CoTGFβ tissues (Fig. 4R) showed a greater degree of sample shortening than CoCon (Fig. 4O).

Since reference points for tracking changes in length were not conserved overnight in ring-shaped tissues, a cleaved tissue slack test was performed by cutting tissues containing C2C12s into linear constructs and plotting slacking lengths over time (Fig. 5C). Most of the length shortening occurred within the first 3 min and slowed considerably by 9 min, with co-culture samples displaying the highest velocities. After 24 h, 3D MyoCon samples shortened to 53.2% of their initial length, MyoTGFB to 49.8%, CoCon to 47.2%, and CoTGFB to 42.7% (Fig. 5E). Each experimental group was significantly different from the others (p < 0.01), except for MyoTGFB and CoCon. The slack tests demonstrated that while TGF-β1 exposure for fibroblast-only samples did not lead to significantly more zero-force shortening in comparison to untreated controls, TGF-β1 positively impacted shortening of myotubes within myoblast-containing tissues, and the interactions between TGF-β1-treated myofibroblasts and myoblasts promoted the greatest degree of shortening over time.

Tissue elastic moduli increase in co-cultures and with TGF-β1 exposure

Using the parallel plate compression system, the Young’s modulus for sectioned self-assembled tissues was determined to be 2.04 kPa for FibCon, 0.81 kPa for FibTGFβ, 3.32 kPa for MyoCon, 3.86 kPa for MyoTGFβ, 4.47 kPa for CoCon, and 6.55 kPa for CoTGFβ (Fig. 5H). FibCon and FibTGFβ tissues were significantly softer than CoCon and CoTGFβ (p < 0.001). While MyoCon and MyoTGFβ were also significantly softer than CoTGFβ (p < 0.001), TGF-β1 did not significantly increase stiffness of samples containing only myoblasts. Similarly, while TGF-β1 treatment increased the stiffness of CoTGFβ, it was not a significant increase with respect to CoCon. Consequently, AFM was used to increase the sensitivity of measurements of elasticity at the tissue surface. AFM generated Young’s moduli of 2.66 kPa for MyoCon, 3.64 kPa for CoCon, and 6.02 kPa for CoTGFβ, with all groups being significantly different from one another (p < 0.01, Fig. 5I).

Supplementation with TGF-β1 and addition of myofibroblasts enhances myogenesis in collagen 1-based hydrogels

MyoCon, MyoTGFβ, CoCon, and CoTGFβ hydrogels were imaged for myosin heavy chain. In the myoblast control condition lacking TGF-β1 supplementation, some MyHC was present in unfused mononuclear cells, but observable myotube formation was absent (Fig. 8A). In contrast, TGF-β1-treated myoblast hydrogels showed robust myotube formation, with somewhat inconsistent alignment orientations (Fig. 8B). Without supplementation with TGF-β1, co-cultures with fibroblasts and myoblasts showed small and immature myotube-shaped cells (Fig. 8C), in contrast to its TGF-β1-treated counterpart containing myofibroblasts (Fig. 8D). CoTGFβ hydrogels yielded the thickest and longest muscle fibers of any condition and possessed the highest degree of multinucleation. These muscle fibers also demonstrated the best alignment. CoTGFβ hydrogels were also more visibly contracted in their culture wells than other conditions (Fig. 8E), although MyoTGFβ hydrogels displayed a lesser degree of contraction. Contraction of MyoCon and CoCon hydrogels was not observed.