Round and square-shaped glass cover slips ( ∅ 22mm, Carl Roth, P235.1; 24 × 24 mm 2 , VWR, ECN632-1571) were cleaned following a modified RCA method (Kern Puotinen et al., 1970). Briefly, glasses were successively sonicated in acetone, ethanol, methanol and distilled water for 3 min each, then covered with a hydrogen peroxide solution (H2O:H2O2:NH3 aq. as 5:1:1), sonicated for 3 min, followed by incubation at 60°C for 30 min, and finally rinsed 10 times with distilled water and dried completely at 70°C. Round glass substrates were subsequently prepared to bind the hydrogels by incubation in 5% (V/V) vinyltrimethoxysilane (Sigma, 235,768) in toluene on a shaker for 18 h in the dark at RT. Afterwards the round glasses were washed consecutively with acetone, ethanol and distilled water before being dried at 140°C for at least 1 h (Hörning et al., 2017b).

For the polyacrylamide hydrogels, a 2% bisacrylamide solution (bAAm, Carl Roth, 3,039.2) was added to a 40% acrylamide solution (AAm, Carl Roth, 7,748.1) in distilled water at a crosslinker ratio of 2%. The polymerization of this solution was initiated by 10% ammonium persulfate (APS, Sigma, A3678) and N,N,N,N-tetramethyl ethylenediamine (TEMED, Carl Roth, 2,367.3). For the plane hydrogels, 20 µL of this polymerizing solution was then sandwiched between a vinyl-silanized glass cover slip ( ∅ 22 mm) and an RCA cleaned cover slip ( 24 × 24 mm²) and left to polymerize for 30 min at RT. For the mechanically-roughened hydrogels the same protocol was used, except that the square cover slip was mechanically roughened before the application of 40 µL of the polymerizing solution. After polymerization, the hydrogels were soaked in water for at least 48 h to remove residual chemicals. The hydrogels were about 60 µm (plane) and 100 µm (mechanically roughened) thick, as confirmed by microscopy. The elastic modulus was set to about 12 kPa by the concentrations of acrylamide and bisacrylamide solutions (Brock et al., 2022).

The E-modulus of the hydrogels were measured by nanoindentation using an atomic force microscope (AFM, NanoWizard, JPK Instruments, Berlin, Germany). A silicon nitride cantilever with an attached spherical colloidal probe (CP-PNP-BSG; 0.08 N/m; R = 5 μm, Olympus Optical) was used. The spring constant of the cantilevers was obtained through thermal noise measurements. The indentation curves were measured with an approach speed of 1 μm/s. The data were analyzed using nonlinear least-squares fitting to the Hertz model (Sneddon, 1965; Domke and Radmacher, 1998) with a customized (MATLAB, 2023) routine. Statistical significance was ensured by the quantification of the Young’s modulus E at 50 independent indentation sites in two 100 × 100 µm² areas for each hydrogel (Brock et al., 2022; Erhardt et al., 2024).

The hydrogels were functionalized by 3,4-Dihydroxy-L-phenylalanine (L-DOPA, Sigma-Aldrich, D9628) Brock et al. (2022). Briefly, L-DOPA was dissolved in freshly prepared TRIS buffer (10 mM, pH 10, Roth 4,855.2) at a concentration of 2 mg/mL for 30 min in the dark on a tube roller and sterilized through a 0.2 µm filter (Filtropur S0.2, Sarstedt 83.1826.001). Followed by a washing with TRIS buffer, 250 µL of this solution was added onto each hydrogel and incubated for 30 min in the dark at RT. To remove unbound L-DOPA, the samples were washed twice with PBS and finally incubated with human plasma fibronectin (Sigma, F2006) at different concentrations for at least 2 h at 37°C.

Rat cardiomyoblasts (H9c2 (2–1), < 15 passages, Sigma-Aldrich, 88092904) were maintained in Dulbecco’s Modified Eagle’s Medium (DMEM) low glucose (Gibco, 31885023), supplemented with 10% fetal bovine serum (FBS, Gibco, 1027010) and 1% penicillin-streptomycin (Gibco, 15140122) at 37°C and 5% CO 2 in a humidified atmosphere. Cells were kept at 60%–70% confluency and passaged every 2–3 days to retain differentiation potential. For single cell experiments, cells were plated at cell densities between 5 and 13 cells/mm² on 12 kPa substrates coated with different fibronectin densities ( ρ FN = 0.4 − 4.0 µg/cm²) and incubated for 24 h before fixation. For differentiation, cells were plated at a cell density of 100 cells/mm² on fibronectin-coated glass substrates (0.4 μg/cm²) and hydrogels (2.6 μg/cm²) and left to reach confluence in DMEM containing 10% FBS for 3 days. Then the medium was either left at 10% FBS for 14 additional days or switched to DMEM containing 1% FBS with or without the addition of different concentrations (10–1000 nM) of all-trans-retinoic acid (RA, Sigma, R2625). Two conditions for the 1% FBS medium was considered. Either the medium was left for 14 additional days unchanged, or the medium was changed and freshly supplemented with RA every 2–3 days for 2 weeks in total. RA was prepared in DMSO as a 10 nM stock, diluted before each use in DMEM with 1% FBS and added in the dark to prevent degradation.

The primary heart tissues were prepared following a method described in (Hörning et al., 2017b; Hörning et al., 2012; Erhardt et al., 2024). Briefly, hearts of 1-3-day-old Wistar rats were isolated and consecutively cleaned, minced, and enzymatically digested in five cycles using collagenase type I (Gibco, 17100017). The isolated cells from the last four cycles were pre-plated for 1 h in plastic dishes to reduce the fraction of fibroblasts. Cells were plated at a density of 2.6 × 10 ⁵ cells/cm² in Dulbecco’s modified Eagle’s medium (DMEM, Gibco, 11885084) supplemented with 10% fetal bovine serum (Gibco, 10270106), 1% penicillin-streptomycin (Gibco, 15140122) and kanamycin sulfate (80 mg/L; Gibco, 11815024). After 24 h, the medium was exchanged to minimal essential medium (MEM, Gibco, 11095080) supplemented with 10% calf serum (Gibco, 16,170–087), 1% penicillin-streptomycin, kanamycin sulfate (80 mg/L) and cytosine arabinofuranoside (ARA-C, 10 μM; Sigma-Aldrich, C1768). After 5 days of incubation, the cardiac tissues were fixed and stained, following the same protocol used for differentiated H9c2 cells.

Cells were washed with PBS and fixed with 4% formaldehyde in PBS (Thermo Scientific Chemicals, J60401.AK) for 10–20 min at RT. For F-actin staining, cells were then washed 3 times with 0.1% Tween 20 (Carl Roth, 9,127.1) in PBS for 10 min each, labeled with rhodamine phalloidin (0.25 U/mL in methanol, Alexa fluor 546, Invitrogen, A22283) and DAPI (1 μg/mL in PBST, Sigma D9542) for 1 h at RT in the dark, subsequently washed 3 times with 0.1% PBST for 10 min, and covered with ProLong Gold antifade reagent (Invitrogen, P10144) until observation. For antibody staining, after fixation, the samples were blocked with 400 µL 0.1% BSA (Sigma Aldrich, A9418) in 0.1% Saponin (Sigma, S1252) for 30 min, then incubated with 200 μL mouse monoclonal anti-sarcomeric alpha actinin (1:200; Invitrogen, MA1-22863) or cardiac troponin T monoclonal antibody (1:200; Invitrogen, MA5-12960) in 0.1% BSA in 0.1% Saponin for 1 h at RT, followed by 200 µL secondary antibody, Alexa Fluor 488 goat anti-mouse IgG (1:200; Invitrogen, A11001), rhodamine phalloidin, and DAPI in 0.1% BSA in 0.1% Saponin for 1 h at RT. ProLong Gold antifade reagent (Invitrogen, P10144) was used to preserve the samples until observation.

Fluorescence was observed by an AxioObserver SD confocal microscope (Carl Zeiss Microscopy GmbH Jena, Germany) equipped with a Yokogawa CSU-X1 spinning disk unit at × 40 (Plan-Apochromat 1.4 Oil DIC UV, Zeiss) and using an Axiocam 503 Mono CCD camera (Zeiss) at a resolution of Δ x = Δ y = 0.227 µm. The images were acquired and analyzed by the ZEN blue v2.3 software (Zeiss). The cell structures were visualized using the 405 nm (DAPI), 488 nm (Alexa fluor 488) and 561 nm (Phalloidin) diode lasers. The images were obtained as a 5 × 5 tile composition with 11–13 focal heights ( Δ z = 1 µm). The tiles were stitched and then orthogonally projected by the “fuse tiles”, “correct shading” and “orthogonal projection” features of the ZEN 2.3 software.

For the observation of calcium transients, the differentiated H9c2 cells were incubated with 200 µL of the fluorescence dye Fluo8 (8.3 µM in PBS, AAT Bioquest) at room temperature (RT) in the dark for 30–60 min. Then, the observation was conducted at RT in Tyrode’s solution (136.9 mM NaCl, 1 mM MgCl2, 2.7 mM KCl, 1.8 mM CaCl2, 0.4 mM NaH2PO4 and 5.5 mM glucose (Sigma-Aldrich, T2145)) with additional 2.7 mM KCl (final concentration of 5.4 mM) and 5 mM HEPES (Roth, 9,105.2). The pH level was adjusted to 7.4 using NaOH. The calcium transients were observed by a customized microscope setup (ThorLabs) equipped with a Kinetix sCMOS high-speed camera (Photometrics, 140 FPS and Δ x = Δ y = 48 µm/px after 4 × 4 binning) and a zoom objective (PlanApoZ 0.5 × /0.125 FWD 114 mm, Carl Zeiss). The samples were electrically stimulated using platinum electrodes ( ∅ 0.5 mm, 99.997%, Thermo Fischer Scientific/Alfa Aesar) and a modified version of the MyoPulser stimulator introduced by Ott and Jung (2023). The device was constructed using a motor controller ( 5 − 35 V, I max = 2 A) and a microcontroller (ESP32-S2-WROVER, Espressif) programmed with an Arduino IDE (2.3.2). A power supply unit with a voltage of 12 V and current up to 2 A was used. A customized Graphical User Interface (GUI) was implemented using the software Processing (The Processing Foundation). The device was enclosed in a customized 3D-printed chassis (PRUSA MK3s, Prusa Research), which was designed using the CAD software SolidWorks (Dassault Systemes) (Erhardt et al., 2024). The tissues were paced with monophasic pulses of 10 ms or 50 ms and amplitudes of ± 10 V. The electrodes were approximately 1 cm apart, resulting in an electrical field of 10 V/cm.

The quantification of actin filaments in muscle cells using AQuA has been described before using customized routines in MATLAB (2023) (Zemel et al., 2010; Inoue et al., 2015; Hörning et al., 2017a; Erben et al., 2020; Brock et al., 2022). Briefly, the Laplacian filter 0 − 1 0 − 1 4 − 1 0 − 1 0 (1) and n = 15 differently rotated anisotropic Gaussians with σ x = 2 px and σ y = 6 px were convoluted to elongated Laplace of Gaussian (eLoG) kernels from Equation 1. The kernels were applied to the original 5 × 5 tiled images and the maximum response of each pixel was calculated, as I max n , x , y = m a x e L o G n × I x , y . (2) Thereafter, from Equation 2 I max was processed by the binarized original images using the Otsu’s thresholding method Otsu (1979). Connected fibers of the same rotational direction with less than 10 px for single cells and 20 px for tissues were removed. The obtained actin fibers were colorized with a color scheme that corresponds to the local actin orientation angles, θ n .

The recorded movies were pre-processed by background subtraction, averaged in time (10 frames), and filtered in space with Gaussian blur (10 px) using a customized macro in ImageJ (1.54f) (Loppini et al., 2022; Erhardt et al., 2024). Further analysis was performed using customized routines in MATLAB (2023). The calcium transients were analyzed at a normalized calcium intensity of 50% of each individual wave to obtain the individual calcium transient duration (CTD) and calcium transient intervals (CTI). From the CTDs and CTIs, the mean values and standard deviation was calculated for each calcium transient.

The normalized calcium transients were calculated by the improved signal oversampling analysis (Erhardt et al., 2024). The periodic signals were pixelwise-stacked by equidistant time intervals, and the CTD was computed from the stacked period-2 calcium transient, as ⟨ CT ⟩ = ∑ i N / 2 CT i 2 i ⋅ T + τ , (3) where CT i are the calcium transients of the individual waves, T is the pacing period, and τ is the offset induced by frame rate inaccuracy of the camera ( τ ≤ 140  Hz ≃ 7 ms).

FFI was applied to the raw unprocessed fluorescence recordings, as introduced before (Hörning et al., 2017a; Loppini et al., 2022; Erhardt et al., 2024). Briefly, the images were decomposed pixelwise and transformed to the mathematically complex Fourier space, F x,y , as a function of the frequency f , i.e., I x,y ( t ) → F x,y ( f ) , with the intensity I x,y ( t ) at the spatial position ( x , y ) . From that, the amplitudes were calculated and spatially recomposed to a Fourier frequency-series. The amplitude map was selected at the respective frequency f T = T − 1 with the pacing period T , as | F ( f T ) | .

In order to identify the optimal differentiation condition, six differentiation medium compositions were tested and statistically analyzed (Figures 1A–G). For all conditions, cells were plated at the same density on glass substrates and left to reach confluence in DMEM containing 10% FBS for 3 days to ensure comparability. After this point of time, which was defined as t = 0 , the conditions were treated differently as described in Material and Methods. Briefly, the medium was changed to 1% FBS medium with or without regular addition of RA ( c RA = 10 − 1000 nM). Additionally, two conditions left with either 10% FBS and 1% FBS without further medium exchange were tested. Neither of those two has been investigated before, to the best of our knowledge.

For the comparison of the conditions, cells were fluorescence stained and observed on large 5 × 5 tiled images by confocal microscopy. From those images the differentiation rate was statistically quantified by an automated analysis routine (see Material and Methods). Figure 1A displays fluorescence-stained control cells at t = 0 . As expected, the differentiation marker sarcomeric alpha actinin (αA) was not detected, and thus, the cells are still undifferentiated at this point. This can be confirmed by the area ratio ⟨ A r ⟩ between the area of differentiated cells (αA signal) and the area of all cells (F-actin signal), shown in Figure 1B. For the undifferentiated cells, ⟨ A r ⟩ is close to 0%. Generally, ⟨ A r ⟩ represents the differentiation efficiency in terms of sarcomeric alpha actinin production. In this study, the highest ⟨ A r ⟩ of about 30% was found at the 10% FBS condition after 2 weeks, although previous studies indicate that serum reduction is necessary for differentiation of H9c2 (Ménard et al., 1999; Pereira et al., 2011; Suhaeri et al., 2015; Campero-Basaldua et al., 2023). However, the finding implies that FBS contains some components that support cell differentiation, and therefore, the differentiation rate is higher if the FBS concentration remains at 10%. In comparison, the 1% FBS condition without medium exchange exhibits only an ⟨ A r ⟩ of about 10%. Cells under this condition still form sarcomeric alpha actinin but less efficiently. In comparison, the 1% FBS condition with medium exchange leads to a significantly lower differentiation rate, despite the only difference being the frequency of medium exchange. A regular medium exchange seems to disturb cell differentiation as it possibly hinders cell-cell communication, which is necessary for effective differentiation. Consequently, all three conditions with RA display a very low ⟨ A r ⟩ as well, but also a relation between the respective RA concentration and ⟨ A r ⟩ . The higher the RA concentration, the lower ⟨ A r ⟩ and hence the differentiation rate. In contrast to previous findings, the addition of RA seems to disturb or even inhibit the differentiation of H9c2, at least in terms of αA formation.

Nevertheless, an opposite trend for the RA conditions is observed for the total number of nuclei N n (Figure 1C). The higher c RA , the higher N n . While RA fails to stimulate sarcomeric alpha actinin production, it seems to promote proliferation instead. A similar impression can be gained when considering ⟨ A c ⟩ , calculated as the total cell area divided by the total amount of nuclei (Figure 1D). While ⟨ A c ⟩ remains comparable to undifferentiated cells throughout all differentiation conditions, the two higher RA conditions exhibit a broader variation in ⟨ A c ⟩ . Cells under these two conditions might be in various stages of the cell cycle, resulting in both small and normal-sized cells, which again suggests an increased proliferation activity. On the other hand, N n is lowered for all reduced FBS conditions, indicating cell detachment, which could be explained by the lower amount of adhesion factors provided. In contrast, for the 10% FBS condition, N n remains comparable to the nuclei number of undifferentiated cells at t = 0 , implying dominant differentiation and cell fusion, as indicated by ⟨ A r ⟩ (Figure 1B).

To illustrate the above-mentioned results further, Figures 1E–G feature three typical examples of fluorescence-stained H9c2 cells after 2 weeks under the 10% FBS, 1% FBS, and 1% FBS with 10 nM RA conditions. The nuclei and F-actin channels demonstrate the difference in N n and the total cell number respectively, whereas the alpha actinin channel exemplifies ⟨ A r ⟩ . When comparing the cell morphology of the 10% FBS and 1% FBS condition without medium exchange, the cells exhibit an elongated shape and multiple nuclei in both. However, the 10% FBS condition leads to thick, voluminous cells, whereas the 1% FBS condition results in rather thin, spindle-like cells. This difference can also be detected in phase-contrast images (not shown). Since both conditions lack the addition of RA, the differentiated cells should be skeletal muscle cells according to the literature (Ménard et al., 1999; Branco et al., 2015). To confirm this assumption, these two conditions were investigated in more detail.

As indicated above, the influence of both differentiation conditions, i.e. 10% and 1% FBS, without RA and medium exchange was examined to determine whether the cells are cardiac or skeletal myocytes. In addition to αA, the cells were fluorescence stained with a cardiac Troponin T (cTT) antibody to distinguish heart and skeletal muscle cells. The cTT is highly specific for cardiac cells and only reacts with differentiated myocytes but not with undifferentiated myoblasts. Therefore, it also serves as an additional differentiation marker. Figure 2A compares ⟨ A r ⟩ in relation to both αA and cTT for control cells at t = 0 and differentiated cells after 2 weeks under the 10% and 1% FBS conditions without medium exchange. For both differentiation conditions, the cells are positive for cTT, and therefore, they are indeed cardiac muscle cells, contrarily to the initial assumption. Similarly, Figure 2B illustrates the nuclei ratio N r , which is defined as the nuclei number of the differentiated cells divided by N n and presents another independent way to quantify the differentiation efficiency. Both ⟨ A r ⟩ and N r show comparable differentiation rates for αA and cTT within the same differentiation condition. Thus, all of the differentiated cells are cardiomyocytes; there is no indication of skeletal muscle cells, which we also exemplarily confirmed for the c RA = 10 nM condition (see Supplementary Figure S1). This is in contrast to previous findings which claim that addition of RA is necessary to form heart cells (Ménard et al., 1999). Comparing the differentiation rate, N r is even higher than ⟨ A r ⟩ . Around 50% of the nuclei under the 10% FBS condition and 25% under the 1% FBS condition belong to differentiated cells, respectively. During differentiation, cells fuse and consequently possess multiple nuclei, however, the area does not correspond linearly to the number of nuclei per cell. For the control cells ( t = 0 ) , a slight difference between αA and cTT can be noticed for both ⟨ A r ⟩ and N r . Very few of the undifferentiated cells already expressed some level of cTT. On the one hand, cTT might be produced earlier than αA as soon as the cells reach confluence. On the other hand, a minority of H9c2 cells could occasionally produce cTT during some phases of the cell cycle and lose it again. To examine this hypothesis, undifferentiated myoblasts were observed 1 hour after cell seeding (see Supplementary Figure S2). At this point, no cardiac Troponin T was found. Hence, it is more likely that cTT is formed as soon as the cells reach confluence.

Additionally, differentiation with 10% FBS not only increases the differentiation rate as quantified by ⟨ A r ⟩ and N r (Figures 2A, B) but also leads to larger cells. Figure 2C displays a connection between the number of nuclei ( N diff ) and the area ( A diff ) of differentiated cells within one recorded image. The slopes indicate the average differentiated cell area one nucleus occupies ( A c,diff ) as about 1500 μm² and 1100 μm² for differentiation media with 10% and 1% FBS. Consequently, cells differentiated with 10% FBS are almost 50% larger than cells differentiated with 1% FBS. This confirms the impression already exemplified in Figures 1E, F for αA. For cTT, the same difference in area size and morphology between the two conditions is depicted in Figures 2D, E. Again, the 10% FBS cells (Figure 2D) appear more voluminous and larger than the 1% FBS cells (Figure 2E). Hence, if the FBS concentration is reduced during differentiation, the cells visibly exhibit signs of starvation. This starvation might result in stress and thus negatively influence the differentiation efficiency. For comparison, Figure 2F shows cTT-stained primary cardiomyocytes which display comparable sarcomeric structures.

Another relation to primary cells appears when examining the differentiated 10% FBS cells more closely. Some cells developed sarcomere-characteristic striation (Figure 2G). While the striation is hard to detect in the orthogonally projected images, it can be seen clearer when going through the individual focal heights (see enlargements in Figure 2G). This finding strongly contradicts preceding studies which could not identify striation and even implied that differentiated H9c2 cells lack striation altogether (Suhaeri et al., 2015). In contrast to those studies, we used a differentiation condition that has not been considered before but led to a significantly improved differentiation. Comparing the 10% FBS cells with primary cardiomyocytes as regards αA again reveals strong likeness between the two (Figure 2H). This even implies that differentiated H9c2 cells might be functioning cardiomyocytes.

Due to the similarity of differentiated H9c2 cells to primary cardiomyocytes, the 10% and 1% FBS cells were examined for their calcium response, since it is part of the regulation of various functions such as contractility, hypertrophy, and gene expression (Aronsen et al., 2016). The differentiated cells were stained with the fluorescent dye Fluo-8 and stimulated by an electric field of about 10 V/cm to investigate the free intracellular Ca²⁺ concentration (Figure 3A). The monophasic stimulation applied by an stimulation duration τ s of 10 ms or 50 ms and a pacing period T ranging between 1000 ms and 240 ms. The recorded calcium signal was then used to extract the restitution properties of the cells at a single location (see Equation 3). Figure 3B shows an example of the relation between calcium transient duration (CTD) and the calcium transient interval (CTI) for cells at 10% FBS after 10 days of differentiation. At 240 ms, the pacing was stopped since the signal-to-noise ration reached a critical level. This typical restitution property can also be illustrated using a signal oversampling analysis of the calcium transient, where individual waves are stacked and averaged (Hörning et al., 2017b; Erhardt et al., 2024). Figure 3C displays three examples at different pacing periods. While the red line indicates the respective means, the black lines show the individual calcium transient curves. Even though the samples feature typical restitution properties for individual cells, i.e., a correlation between CTI and CTD for smaller T , and a CTD plateau for larger T, no wave conduction between the cells could be observed. Further, we did not detect any contractile motions in the tissue. However, considering the number of cells and the lower degree of sarcomere development in H9c2 cells compared to primary heart cells (Figures 2G, H), we are unable to conclusively state whether H9c2 cells are able to beat at this point.

For the 1% FBS condition, no wave conduction was detected either, and the calcium transients exhibited weaker signals, i.e., lower signal-to-noise ratio. Figure 3D illustrates such an example along with a normalized amplitude of a 6 × 6 mm² field of view analyzed by Fourier transformation imaging (FFI) (Hörning et al., 2017b). Calcium transients were observed only at very few locations in the tissue (yellowish color). In contrast, for the same age of the tissue (16 days) but the 10% FBS condition, active areas with stronger signals were identified (Figure 3E). The strongest signal was detected after 9–10 days of differentiation (Figures 3C, F), implying a maximum maturity before 2 weeks. Thus, the starvation over the long time period causes a reduction of active, functional cells. Nevertheless, the differentiated H9c2 cells under the 10% and 1% FBS conditions show calcium transients similar to primary heart cells and are therefore functional cardiac cells, at least in terms of their calcium response.This finding does not contradict the increase upregulation of Ca²⁺ channels in the presence of 1% FBS Ménard et al. (1999), because the lower signal-to-noise ratio of the Calcium transients is mainly caused by the difference in the differentiation ratio of cells. Whether the Calcium transient amplitude on a single cell level changes under different culture conditions still remains to be investigated in more detail with a different observational approach.

As glass does not represent the optimal extracellular matrix (ECM) condition for muscle cells, artificial ECMs, such as hydrogels, were used in the past. They better mimic natural ECM conditions (Hörning et al., 2012; Engler et al., 2006; Engler et al., 2004; Brock et al., 2022) and influence differentiation (Engler et al., 2004; Smith et al., 2018; Denes et al., 2019). In order to identify the optimal ECM condition for H9c2 cells, the response of single cells to polyacrylamide hydrogels with an E-modulus of E = 12.4 kPa were tested initially. This was chosen since it matches the natural rigidity of muscle cells (Engler et al., 2004). In a previous study, we established ECM conditions for the C2C12 myoblast cell line and observed the largest cell area at ρ FN = 2.6 µg/cm² on 12 kPa hydrogels (Brock et al., 2022). For the experiments, the cells were seeded in low densities on either plane or mechanically-roughened hydrogels, which were functionalized with different fibronectin densities ( ρ FN = 0.4 − 4.0 µg/cm²). After 24h, the cells were fluorescence stained and statistically analyzed. Figure 4A compares fluorescence-stained cells on plane hydrogels that were coated with 0.4 and 2.6 μg/cm² fibronectin, respectively. The cells on the hydrogels with the lower fibronectin density appear smaller in size and more circular in shape. In contrast, the cells on the higher fibronectin density display a more complex spreading conformation by showing longer actin filaments, which indicates stronger adhesion to the coated substrate.

To compare the morphological cell response, the mean projected cell area ⟨ A ⟩ and the mean F-actin amount per cell ⟨ M ⟩ for ρ FN = 0.4 − 4.0 µg/cm² was statistically quantified (Figures 4B, C). Both ⟨ A ⟩ and ⟨ M ⟩ increase until about ρ FN = 2.6 µg/cm² and show a saturation thereafter. ⟨ A ⟩ can be parameterized by a Lorentz-like function, as L ρ FN = ⟨ A ⟩ = x 1 ρ FN x 2 ρ FN − x 3 2 + 1 + x 4 , (4) where x i are fitting parameters. The black solid line represents the respective fit of Equation 4 with an optimum at about ρ FN = 2.6 µg/cm² (Figure 4B). This analysis has been applied before for C2C12 cells on collagen (Zemel et al., 2010) and fibronectin coated hydrogels (Brock et al., 2022). In the latter study, the maximum was also determined at ρ FN = 2.6 µg/cm². For comparison, ⟨ A ⟩ and ⟨ M ⟩ are also depicted for single H9c2 cells on glass, displaying similar values as cells on hydrogels with higher fibronectin density.

Since plane substrates do not promote the formation of confluent tissue, hydrogels ( E = 12.4 kPa) that were mechanically roughened and thus feature micro scratches on their surface to support cell adhesion were tested, as a further step. Figure 4D illustrates examples of fluorescence-stained cells on roughened hydrogels at ρ FN = 0.4 and 2.6 μg/cm². Similar to cells on plane substrates, the cells on the lower fibronectin density seem slightly smaller than on the higher density. Nevertheless, the cells appear visibly larger and their actin filaments more oriented than on the plane hydrogels, especially for ρ FN = 0.4 µg/cm². The quantification of ⟨ A ⟩ and ⟨ M ⟩ for cells on roughened hydrogels (Figures 4E, F) displays a similar trend as for plane hydrogels (Figures 4B, C). ⟨ A ⟩ and ⟨ M ⟩ increase only up to ρ FN = 2.6 , which again indicates the optimum. Generally, roughened hydrogels lead to larger cells for all tested ρ FN , as indicated by the solid and dashed line in Figure 4E. This difference is especially noticeable for the cells cultured on low ρ FN . At ρ FN = 0.4 µg/cm², for instance, the cells are about twice as large on the roughened hydrogels. Thus, the micro pattern influences cell adhesion. In conclusion, a more natural ECM, provided through the hydrogel, seems to be preferable for cells. As the mechanical roughening of the 12.4 kPa hydrogels in combination with a coating of ρ FN = 2.6 µg/cm² leads to a maximum in both ⟨ A ⟩ and ⟨ M ⟩ for single H9c2 cells, this ECM condition was chosen for subsequent experiments on cell differentiation.

For the investigation of the differentiation dynamics of H9c2 cells on hydrogels, the optimum ECM condition, i.e., roughened hydrogel with E = 12.4 kPa and ρ FN = 2.6 µg/cm², in combination with the optimum medium condition (DMEM with 10% FBS) was used. Figures 5A–C show time-related comparisons between differentiation cells on glass and hydrogels at t = 0 and after 1 and 2 weeks of differentiation. The quantification of cells after 1 week was added since the cells expressed the highest calcium signals after 9–10 days of differentiation (Figure 3). The cells were stained with the differentiation marker αA. Again ⟨ A r ⟩ , N r , and N n were statistically quantified (see Figures 5A–C). In comparison to glass substrates (orange diamonds), significantly lower differentiation rates were observed after 1 and 2 weeks for cells cultured on the optimized hydrogels (blue circles) for ⟨ A r ⟩ and N r (Figures 5D, E). Despite the natural substrate condition, the differentiation efficiency did not improve. Moreover, no striation could be detected for cells on hydrogels. However, in a few cases a comparable differentiation ratio was observed, as on glass. As opposed to this, the cells on glass reach their maximum differentiation ratio of about 40% after 1 week. Nevertheless, no sarcomere-typical striation could be found for those cells either, suggesting that their differentiation is not entirely completed after all and that a longer differentiation period is advantageous.

Regarding the total nuclei number N n (Figure 5C), the number of cells observed on both glass and hydrogel at t = 0 is comparable, suggesting sufficient adhesion to the substrate. During differentiation, however, the confluent cells on the hydrogels cluster and detach, resulting in lower N n after both 1 and 2 weeks of differentiation. For primary cardiomyocytes, the roughened hydrogels provide sufficient adhesion to allow cardiac contraction (Erhardt et al., 2024), indicating that the proliferation and differentiation dynamics for H9c2 cells in confluent tissues are too strong, surpassing the adhesion forces and resulting in cell detachment. For comparison, N n decreases on glass as well after 2 weeks, which indicates that some cells detach during differentiation independently of the substrate rigidity.

To visually illustrate the difference between the differentiation ratios on glass and hydrogels, Figures 5D–F show fluorescence-stained cells after 1 week on glass (Figure 5D) and hydrogel (Figure 5E), as well as, after 2 weeks on hydrogel (Figure 5F). There are fewer differentiated cells on both hydrogels compared to glass and also some holes between the cells on the hydrogels which matches the lower N n and indicates detachment. Conclusively, the roughening of hydrogels enabled H9c2 cell differentiation, which was not possible on plane hydrogels, however, at a much lower differentiation efficiency than on glass substrates. Further improvements for cell attachment needs to be considered in future studies.