Commercially available PVA (BP-17 and BF-17, Chang Chun, Taiwan) was dissolved in H₂O at 90°C to form 3.5% and 7% (wt/wt) PVA polymer solutions. After cooling, the 7% PVA polymer solution was then blended with 40 mg/mL, 20 mg/mL, 10 mg/mL, 5 mg/mL, 2 mg/mL, and 1 mg/mL gelatin solution (dissolved in H₂O) in the 1:1 volume ratio to give rise to different gelatin concentrations of the PVA-G polymer solution [PVA-G (20 mg/mL), PVA-G (10 mg/mL), PVA-G (5 mg/mL), PVA-G (2.5 mg/mL), PVA-G (1 mg/mL), and PVA-G (0.5 mg/mL)]. The 3.5% PVA polymer solution was blended with collagen to give rise to different collagen concentrations of the PVA-C polymer solution [PVA-C (200 ug/mL), PVA-C (100 ug/mL), PVA-C (50 ug/mL), and PVA-C (25 ug/mL)]. The PVA-CG polymer solution was prepared by adding collagen to the different concentration of the PVA-G polymer solution to give rise to different gelatin and collagen concentration of the PVA-CG polymer solution. For example, PVA-C (100 ug/mL)-G (5 mg/mL) denotes that the PVA-CG polymer solution has the final concentration of 3.5% PVA, 100 ug/mL collagen and 5 mg/mL gelatin dissolved in water. For cell cultures, 0.1 mL PVA, PVA-G, PVA-C, and PVA-CG polymer solution were coated in each 24-culture-well (area: 1.9 cm²) and dried for 7 days at room temperature as thin films.

To test the biocompatibility (adhesion and viability) of prepared PVA, PVA-G, PVA-C, and PVA-CG polymer substrates, NIH-3T3 cell line was used. This cell line was cultured in the DMEM/High Glucose medium containing 10% fetal bovine serum (FBS), penicillin G (100 IU/mL), and streptomycin (100 μg/mL) at the cell density of 1.5 × 104/cm². Cultures were maintained at 37°C in a humidified atmosphere of 95% air/5% CO₂ and were subpassaged every 3–4 days.

Human embryonic stem cell line (TW1), which was gifted from Dr. Dah-Ching Ding of Tzu Chi University and can be obtained from Bioresource Collection and Research Center (BCRC), Taiwan (Cheng et al., 2008; Ding et al., 2012; Chang et al., 2017), was maintained on mitomycin-treated mouse embryonic fibroblast (MEF) feeder layers in the embryonic stem cell culture medium [DMEM/F12 supplemented with 20% (vol/vol) KnockOut serum replacement (Invitrogen), 1 mM nonessential amino acids (Invitrogen), Glutamax (Invitrogen), penicillin/streptomycin (Invitrogen), 0.55 mM 2-mercaptoethanol (Invitrogen), 4 ng/mL recombinant human FGF2 (R&D Systems)] (Thomson et al., 1998; Hoffman and Carpenter, 2005). Cultures were manually passaged at a 1:6–1:8 split ratio every 5–7 days.

MEF were harvested from E13-E14 embryos of female ICR mice and cultured in Dulbecco’s modified Eagle’s medium (DMEM; Invitrogen) supplemented with 5% fetal bovine serum (FBS; Gibco), 2 mmol/l l-glutamine (Invitrogen) and penicillin/streptomycin (Invitrogen). For use as feeder cells, MEF were inactivated with 10 μg/mL mitomycin C (Sigma) for 3 h to arrest mitosis, thoroughly washed in phosphate buffered saline (PBS), and replated at 3 × 104 cells/cm² on 0.1% gelatin-coated tissue culture plates and reached confluence in 3–4 days (Cheng et al., 2008).

Human embryonic stem (ES) cells were then cultured on the PVA, PVA-C, PVA-G, and PVA-CG polymer substrates in the ES medium to observe if these blended PVA substrates could serve as mouse embryonic fibroblast feeder layer to maintain the self-renewing of human ES cells. For differentiation, the embryonic stem cell culture medium was replaced with FBS-containing medium (DMEM/F12 containing 10% FBS, penicillin/streptomycin) to see whether human ES cells can attach, grow and be induced to differentiate on the blended PVA substrates prepared in this study.

The cell adhesion and viability are measured by the method of the MTT assay to assess the ability of attachment, proliferation and survival of NIH-3T3 and TW1 cells on the PVA, PVA-G, PVA-C, and PVA-CG polymer substrates. The seeding density of NIH-3T3 for MTT test is 1.5 × 104/cm². At indicated time points (4 h, 1 day, 2 days, and 4 days), the culture medium was removed and incubated with 0.1 mL of MTT solution (2 mg/mL) for 3 h at 37°C. After incubation, the MTT solution was aspirated and the formazan reaction products were dissolved in dimethyl sulfoxide and shaken for 30 min. The optical density of the formazan solution was read on an ELISA plate reader at 570 nm.

AFM experiments were performed with JPK AFM Nano-wizard (Bruker Nano, Berlin, Germany). The AFM probe used was Aluminium reflex coating of cantilevers with silicon rotated tips (ContAI, Budgetsensors, Bulgaria). The spring constant for the cantilevers was 0.2 N/m, half angle 20 deg. and a tip radius of 10 nm. The AFM images were captured in contact mode at room temperature in dry environment and the resolution is 512 × 512 pixels. The scanning rate was set to 1 Hz. Both the height image and deflection image were recorded. Average Roughness values (Ra) and Z-average were then measured using JPKSPM data processing software. The measurement size was 100 um × 100 um. Before the force curves acquisition, the precise spring constant of the cantilever was calibrated by the thermal noise method in the air. To measure Young’s modulus and adhesion force, the force curves were recorded within a scan area of 100 um × 100 um, within which a grid range was 8 × 8 points. In total, 64 force curves were recorded. The indentation rate was set to 2 um/s. The batch processing of the curves was using baseline subtraction, contact point correction, vertical tip position correction, and applying elasticity fit. The Young’s modulus values were obtained by fitting Hertz/Sneddon model for cone indenter with half cone angle 20 deg. (Akbarzadeh Solbu et al., 2022; Asgari et al., 2022).

FTIR-ATR (Nicolet iS5, Thermo Fisher Scientific, United States) was used to confirm the absorption bands of amide I (1,600–1,700 cm⁻¹) and amide II (1,500–1,600 cm⁻¹) groups, the most prominent absorption bands of proteins, of the prepared PVA substrates. Briefly, samples are dried in a desiccator at R.T. and were sandwiched between the reflection crystal (KRS-5) and the back plate. Spectra were taken with air as the background using OMNIC spectra software (OMNIC 9.12.1019, Thermo Fisher Scientific, United States) in the range of 400–4,000 cm⁻¹ (Schwaighofer and Lendl, 2023).

After the human ES cells were seeded on the control TCPS wells or blended PVA polymer substrates, total cellular RNA was extracted at indicated time points using Trizol reagent (REzol, Protech, Taipei, Taiwan) according to the manufacturer’s instructions. After cells were lysed, they were transferred to centrifuge tubes. Chloroform (0.2 mL) was then added to each tube and mixed with the sample. The sample tubes were incubated at room temperature for 5 min, and centrifuged at 12,000 g for 15 min at 4°C. The aqueous phase was transferred to new tubes, each containing 500 μL isopropanol. Samples were centrifuged at 12,000 g for 10 min at 4°C. The supernatant was removed, and the RNA pellet was washed with 500 μL 75% ethanol. The RNA pellet was then dissolved in 20 μL water containing 0.01% DEPC and stored in a −80°C freezer for further use. cDNA was prepared from RNA by using a reverse-transcription kit (Bio-Rad, Hercules, CA, United States). All PCR samples were analyzed by electrophoresis on 2% agarose gel containing 0.5 μg/mL ethidium bromide (Sigma). The primers used in this study are: β-actin (5′-CAC​CTT​CTA​CAA​TGA​GCT​GCG-3′, 5′-TGC​TTG​CTG​ATC​CAC​ATC​TGC-3′), Adpsin (5′-GCG​CAC​CTG​GCG​GCG​TCC​TG-3′, 5′-GCA​CTG​CGC​GCA​GCA​CGT​CGT​A-3), Gata4 (5′-CTA​CAG​GGG​CAC​TTA​ACC​CA-3′, 5′-AGA​GCT​GAA​TCG​CTC​AGA​GC-3), Gata6 (5′-CCT​CAC​TCC​ACT​CGT​GTC​TGC-3′, 5′-GTC​CTG​GCT​TCT​GGA​AGT​GG-3), GLUT4 (5′-GCC​ATT​GTT​ATC​GGC​ATT​CT-3′, 5′-CTA​CCC​CTG​CTG​TCT​CGA​AG-3′), hALBP (5′-AGT​CAA​GAG​CAC​CAT​AAC​CTT​AGA-3′, 5′-CCT​TGG​CTT​ATG​CTC​TCT​CAT​AA-3′), Sox9 (5′-GCC​ACG​GAG​CAG​ACG​CAC-3′, 5′-GCG​CCT​GCT​GCT​TGG​ACA-3′), Sox10 (5′-GAC​CGC​CGC​CAT​CCA​GGC-3′, 5′-AGT​AGC​TGC​TCA​CAT​GGC​CT-3′).

At indicated time points, the cultured cells were rinsed with phosphate buffered saline (PBS) three times, fixed with 4% paraformaldehyde for 30 min at 4°C, and then stained with filtered Oil Red O solution (0.5% Oil Red O in isopropyl alcohol) for 2 h at room temperature. After staining, the cells were washed three times with PBS and observed under an optical microscope. To quantify the percentage of Oil-Red-O-positive adipocytes, human ES cells differentiated on the PVA-CG polymer substrates were costained with Oil Red O and DAPI.

At indicated time points, cultured cells were fixed in ice-cold 4% paraformaldehyde in PBS for 20 min and washed three times in PBS. After fixing, cells were incubated with primary antibodies diluted in PBS containing 0.5% triton-X-100% and 10% bovine serum albumin for 24 h at 4°C. The primary antibodies and their dilution used in this study were mouse anti-microtubule-associated protein 2 polyclonal antibody (anti-MAP2; 1:500; Millipore), rabbit anti-glial fibrillary acidic protein polyclonal antibody (anti-GFAP; 1:500; Chemicon), goat anti-uncoupling protein 1 polyclonal antibody (anti-UCP1; 1:1,000; Santa Cruz), rabbit anti-nestin polyclonal antibody (anti-nestin; 1:500; Chemicon), rabbit anti-brachyury T polyclonal antibody (1:800, Abcam), mouse anti-α-fetoprotein monoclonal antibody (1:500; Millipore), rabbit anti-forkhead box D3 polyclonal antibody (anti-FoxD3; 1:1,000, Aviva), mouse anti-stage-specific embryonic antigen-4 (anti-SSEA4; 1:500, Chemicon), rabbit anti-octamer 4 polyclonal antibody (anti-OCT4, 1:500, Chemicon), mouse anti-cytokeratin 15 monoclonal antibody (anti-K15, 1:500, Invitrogen). FITC- and rhodamine-conjugated secondary antibodies were used to visualize the signal by reacting with cells for 1 h at room temperature. Cell morphology was viewed with a photomicroscope (Zeiss LAMBDA 10–2, Germany). Cells were also counterstained with DAPI. Immunostained cells were visualized by indirect immunofluorescence under the fluorescent microscope (Axiovert 100TV, Germany).

Digital photomicrographs were taken in random fields at indicated experimental points and conditions. Quantification of collagen particle size was evaluated by the diameter of the particles. Total 30–50 independent readings of the particle diameter, fibril length, fibril covering area and cell spreading area were measured using NIH Image software (ImageJ), and the means and standard deviation were also calculated.

Cortical neural stem cells were prepared from pregnant Wistar rat embryos on day 14–15 according to a protocol detailed previously (Tsai and McKay, 2000). Briefly, rat embryonic cerebral cortices were dissected, cut into small pieces and mechanically triturated in cold Hank’s balanced salt solution (HBSS) containing 0.4 g/L KCl, 0.09 g/L Na₂HPO₄, 7H₂O, 0.06 g/L KH₂PO₄, 0.35 g/L NaHCO₃, 0.14 g/L CaCl₂, 0.10 g/L MgCl₂.6H₂O, 0.10 g/L MgSO₄.7H₂O, 8.0 g/L NaCl, and 1.0 g/L D-glucose in deionized water. The dissociated cells were collected by centrifugation and were resuspended in a serum-free medium (SFM) containing DMEM-F12, 8 mM glucose, glutamine, 20 mM sodium bicarbonate, 15 mM HEPES and N2 supplement (25 mg/mL insulin, 100 mg/mL human apotransferri, 20 nM progesterone, 30 nM sodium selenite, pH 7.2). The number of live cells was counted by trypan blue exclusion assay in a hemocytometer. Cerebral cortical neural stem cells were purified and cultured in T25 culture flasks (Corning, New York, United States) at a density of 50,000 cells/cm² in the above culture medium in the presence of bFGF at a concentration of 20 ng/mL. Cultures were maintained at 37°C in a humidified atmosphere of 95% air/5% CO₂. After 1–3 days of culture in vitro, cells underwent division and formed clust ers of cells, termed neurospheres, which suspended in the medium. Subsequently, adherent cells were discarded and suspended neurospheres were collected by centrifugation, mechanically dissociated and subcultured as single cell in a new T25 culture flask at a density of 50,000 cells/cm² in the fresh culture medium containing the same concentration of bFGF. These cells grew into new spheres in the subsequent 2–3 days. The procedure of subculture was repeated 2–3 weeks to achieve the purified cortical neural stem cells and proliferating neurospheres. The plasticity of these purified cortical neural stem cells was identified by the method of immunocytochemistry with anti-nestin, anti-NSE, and anti-GFAP, which has been reported in our previous publication²².

The protein expression of MAP-2, GFAP, and β-actin was determined at day 7 of cell culture. Rat cerebral cortical neurospheres cultured on the substrates were resuspended in cell lysis buffer (Thermo scientific) and sonicated. After centrifugation, the protein content was determined in the supernatants by a BCA protein quantification kit (Pierce Biotechnology, Rockford, IL). Sixty mg proteins from neurospheres were added to Laemmli sample buffer and boiled for 10 min. Proteins were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and blotted onto polyvinylidene difluoride membranes. Western blotting was performed using anti-MAP-2 (Chemicon), anti-GFAP (Chemicon), and anti-β-actin (Chemicon). The membranes were incubated with the primary antibodies overnight at 4°C. After extensive washing, the membranes were further incubated with horseradish peroxidase-conjugated secondary antibodies for 1 h. The blottings were then developed using an enhanced chemiluminescence detection system (Millipore).

All measurements were presented as means ± S.D. (standard deviation). Statistical significance was calculated using one-way analysis of variance (ANOVA) or using independent-samples Student’s t-test. The p values of <0.05 were considered statistically significant.

Previous studies have demonstrated that, owing to the intrinsic properties of the PVA substrates, several anchoring-dependent cells cannot attach and survive upon the PVA substrates in vitro (Nuttelman et al., 2001; Schmedlen et al., 2002; Baker et al., 2012). Hence, in order to modify these effect of PVA substrates, we blended PVA substrates with extracellular matrixes collagen or/and gelatin to enhance the cell anchoring and survival on the PVA substrates. After blending, coating and drying, it is observed that these non-blended and blended PVA substrates show different surface topography (Figure 1). Without blending collagen and gelatin, PVA substrates reveal the dense and smooth surface structure similar with previous published results (Ino et al., 2013) (Figure 1A). Interestingly, it is found that after blending, collagen or/and gelatin can self-assemble to form nucleated particles or branched fibrils in the PVA substrates with the scale that can be observed under the optical microscope (Figure 1A). While the surface of PVA substrates blended with collagen (PVA-C) display nucleated particles, the surface of PVA substrates blended with gelatin (PVA-G) show the branched and elongated fibril structure (Figure 1A). Additionally, the surface of PVA substrates blended with collagen and gelatin (PVA-CG) show branched and intermingled fibrils (Figure 1A). These nucleated particles and branched fibrils observed on the PVA-C, PVA-G, and PVA CG substrates cannot be found on the collagen-coated or gelatin-coated wells (data not shown). The micrographs examined under the atomic force microscope also revealed that these nucleated particles and branched fibrils are embedded within the PVA substrates (Figure 1B). In PVA-C substrates, the number of nucleated collagen particles increases with the amount of collagen blended and are significantly different between groups (p < 0.05); however, the size of these particles does not change significantly among these groups (Figure 1C). In PVA-G substrates, when blended with the lower gelatin concentration (0.5 mg/mL), nucleated gelatin particles are formed in the PVA substrates (Figure 1D). However, as the blended gelatin concentration increased, these nucleated gelatin particles start to assemble or connect together to form branched and elongated fibrils; quantitatively, the total length and covering area per mm² of elongated gelatin fibrils in the PVA substrates increase with the amount of gelatin blended and are significantly different between groups (p < 0.05) (Figure 1D). Time-series observation shows that the number of branched and elongated fibrils in the PVA-CG substrates increases with time during the substrate preparation (Figure 1E), and the amount of branched fibrils in the PVA-CG substrates increases with the amount of collagen and gelatin blended (Figure 1F). According to the AFM analysis, it is found that the z-average and average roughness increased along with the formation of nucleated particles and branched fibrils as comparing to the non-modified PVA substrates. The z-average of PVA, PVA-C, PVA-G, and PVA-CG substrates is 37.27, 47.68, 90.63, and 39.87 nm, respectively (Figure 1G). The average roughness of PVA, PVA-C, PVA-G, and PVA-CG substrates is 11.37, 13.3, 20.1, and 19.0 nm, respectively (Figure 1G). It is also found that surface adhesion force showed no difference between PVA, PVA-C, PVA-G, and PVA-CG groups (Figure 1H). However, the Young’s modulus increased along with the particle and fibril formation in the PVA-C, PVA-G, and PVA-CG groups (Figure 1H). The spectra of FTIR analysis showed that increased band absorbance of the amide I band (1,600–1,700 cm⁻¹-originating from the C=O stretching and N-H in-phase bending vibration of the amide group) and the amide II band (1,500–1,600 cm⁻¹-arising from N-H bending and C-N stretching vibrations) in PVA-G and PVA-CG groups but not in the PVA-C group (Figure 1I; upper). Furthermore, It is found that the absorbance of amide I and amide II bands increased with the increasing collagen and gelatin concentration blended in the PVA substrates (Figure 1I; bottom). These results suggested that the presence and intercalation of collagen or/and gelatin in the PVA substrates after blending, and the assembly of collagen or/and gelatin into particles or fibrils may influence substrates’ surface topographical and mechanical properties.

In order to investigate if these modified PVA-C, PVA-G, and PVA-CG substrates can improve anchoring and survival of cultured cells in vitro, we cultured NIH-3T3 fibroblasts on these modified substrates and examine the attachment and viability of cells by means of the MTT colorimetric assay. Quantitatively, as indicated by the results of MTT assay after 4 h of culture, the cell attachment were improved on the PVA-C, PVA-G, and PVA-CG substrates as comparing with those on the TCPS, non-blended PVA, collagen-coated and gelatin-coated wells (Figures 2A–C). Furthermore, the results of MTT assay after 1 day, 2 days, and 4 days of the culture show that the viability of NIH-3T3 cells on the PVA-C, PVA-G, and PVA-CG substrates increased with time, which suggested that these modified PVA substrates are biocompatible for NIH-3T3 cells in vitro (Figures 2D–F). Although the viability of NIH-3T3 cells on the PVA-C, PVA-G, and PVA-CG substrates are improved and significantly higher than that on the non-blended PVA substrates, they are not higher than those on the control TCPS, gelatin-coated or collagen-coated wells during the 4-day culture in vitro, assuming that NIH-3T3 cells are more adhesive but less proliferated on the modified PVA-C, PVA-G, and PVA-CG substrates.

As predicted, NIH-3T3 cells cannot attach and survive on the non-blended PVA substrates after 4 days of culture in vitro (data not shown). However, it is found that these cells can now adhere onto the PVA-C substrates and their morphological appearances are dependent on the concentration of collagen blended. Generally, NIH-3T3 cells cultured on the PVA-C polymer substrates blended with higher concentration of collagen (200, 100, and 50 μg/mL) reveal round and oval morphology without any cytoplasmic extension (Figure 3A). Instead, NIH-3T3 cells cultured on the PVA-C substrates blended with lower concentration of collagen (25 μg/mL) exhibit spindle and chained morphology (Figure 3A). On the contrary, the morphological appearances of attached NIH-3T3 cells on the blended PVA-G substrates show flatten spreading morphology exclusively regardless of the concentration of gelatin blended (Figure 3B). Likewise, NIH-3T3 cells also can attach on the PVA-CG substrates. Interestingly, three different types of cell morphology can be observed on the PVA-CG substrates (Figure 3C). In general, NIH-3T3 cells exhibit flatten spreading morphology on the PVA-CG substrates when the relative amount of gelatin is dominant in the PVA-CG substrates. On the contrary, when the relative amount of collagen is dominant in the PVA-CG substrates, NIH-3T3 cells exhibit round and oval morphology. Additionally, NIH-3T3 cells reveal another intermediate spindle and chained morphology on the PVA-CG substrates when the relative amount of collagen and gelatin are adjusted and balanced in the indicated groups. It is also observed that the spreading area per cell of NIH-3T3 cells on the modified PVA-C, PVA-G, and PVA-CG substrates depends on the concentration of collagen or/and gelatin blended (Figure 3D). In the PVA-C and PVA-G substrates, the spreading area per cell increases with the decreasing collagen concentration or the decreasing gelatin concentration, respectively. In the PVA-CG substrates, the spreading area increases with the decreasing collagen concentration and increasing gelatin concentration. Taken together, these data suggested that PVA substrates blended with collagen or/and gelatin can be utilized as culture substrates to improve cell adhesion and viability as well as to control cell shape and spreading in vitro.

For now, we have demonstrated that PVA substrates blended with collagen or/and gelatin have better cell attachment and viability than non-blended PVA substrates. In our previous study (Young and Hung, 2005; Hung and Young, 2006), it is found that rat embryonic cerebral cortical neural stem cells cannot attach, survive and differentiation on the non-blended PVA substrates. Hence, we next want to examine if these modified PVA substrates can exert influence on the in vitro behavior of cortical neural stem cells. Similar with our previous published studies, the cortical neural stem cells, cultured as neurospheres, can suspend and grow in the absence of fetal bovine serum (FBS) or be induced to attach and differentiate in the presence of FBS on the control TCPS wells (Figure 4A). However, regardless of the absence or presence of FBS, these neurospheres cannot attach and survive on the PVA substrates; instead, they suspended above the PVA substrates and are finally subjected to debris and death (Figure 4A). In this study, it is also observed that these neurospheres cannot attach and survive on the PVA-G and PVA-CG substrates regardless of the absence or presence of FBS, or the concentration of collagen or gelatin blended in the PVA substrates (data not shown). Nevertheless, the attachment of these neurospheres can be achieved on the PVA-C substrates in the presence of FBS (Figure 4B). Without the addition of FBS in the culture medium, these neurospheres also suspended on the PVA-C substrate regardless the concentrations of collagen blended in the PVA substrates (Figure 4B). Interestingly, the behavior of these FBS-required attached neurospheres on the PVA-C substrates is collagen-concentration dependent. That is, these neurospheres reveal unspreading morphology when they are cultured on the PVA-C substrates blended with higher concentration of collagen [PVA-C (100 μg/mL)] (Figure 4B). However, when the amount of collagen blended in the PVA-C substrates decreases [PVA-C (50 μg/mL) and PVA-C (25 μg/mL)], it can be observed that these neurospheres start to spread and a great number of cells migrate out from the neurospheres and differentiate into process-bearing cells on the modified PVA-C substrates (Figure 4B).

In order to visualize the neuronal or glial differentiation of rat cerebral cortical neural stem cells on the PVA-C substrates, attached cells are immunocytochemically stained with anti-MAP-2 and anti-GFAP, which are specific for neurons and glial cells, respectively. Similar with previous studies (Young and Hung, 2005; Hung and Young, 2006), when FBS was present in the culture medium, most of the differentiated neural stem cells on the control TCPS substrates are immuno-positive for GFAP and immuno-negative for MAP-2 (Figure 4C), indicating that under this condition, neural stem cells are preferably induced to differentiate into glial cells. However, under the same culture condition, neuronal differentiation of neural stem cells can be achieved and increased on the PVA-C substrates at all indicated collagen concentrations (Figure 4C). Quantitatively, the ratio of differentiated MAP-2-immuno-positive neurons on the PVA-C substrates [PVA-C (25 μg/mL), PVA-C (50 μg/mL), and PVA-C (100 μg/mL)] are all significantly higher than that on the control TCPS wells (Figure 4D). Moreover, it is found that the ratio and number of differentiated MAP-2-immuno-positive neurons increase with the decreasing concentration of collagen blended in the PVA-C substrates (Figure 4D). Western blot analysis revealed that the MAP-2 protein expression on PVA-C (25 μg/mL) and PVA-C (50 μg/mL) substrates are higher than the control TCPS group, but is dramatically decreased on the PVA-C (100 μg/mL) substrate (Figure 4E). These results indicate that neuronal differentiation of rat cerebral cortical neural stem cells on the PVA-C substrates can be controlled in a collagen-concentration dependent manner.

Because of the ability of self-renewing and high plasticity, the pluripotent stem cells are the most promising cells for future cell-based therapies in the field of regenerative medicine and tissue engineering. Here, we also examine the effect of collagen- or/and gelatin-blended PVA substrates on the in vitro behavior of pluripotent human ES cells. First, we investigate if these modified PVA substrates can serve as mouse embryonic fibroblast feeder layers to maintain human ES cells in the state of self-renewing. Traditionally, human ES cells can attach and propagate in number without undergoing differentiation on the mouse embryonic fibroblast feeder layers in the ES culture medium; DMEM/F12 containing 20% KnockOut serum replacement (Figure 5A). However, under the same culture condition, it is observed that human ES cells cannot attach but suspend on the PVA, PVA-C, PVA-G, and PVA-CG substrates (Figure 5A), suggested that these non-blended and blended PVA substrates fail to function as mouse embryonic fibroblast feeder layers to anchor and support self-renewing of human ES cells.

Because FBS is an important factor for cells to anchor onto the substrates, we replace the ES culture medium with the FBS-containing medium (DMEM/F12 containing 10% FBS) to see whether human ES cells can attach, grow, and be induced to differentiation on the substrates prepared in this study. It is observed that, after 6 days of culture in vitro, the clumps of human ES cells now can attach and grow on the PVA-C and PVA-CG substrates as well as on the control TCPS substrates (Figure 5B). However, even in the presence of FBS in the culture medium, human ES cells cannot attach but suspend on the PVA and PVA-G substrates. According to the results of MTT assay, the viability of human ES cells on the control TCPS, PVA-C, and PVA-CG substrates increased with time and are higher than those on the PVA and PVA-G substrates at indicated time points (Figure 5C)

Morphologically, it is found that human ES cells cultured on the PVA-CG substrates are induced to differentiate into homogeneous populations with multiple Oil-Red-O-positive lipid droplets in the cytoplasm similar to brown adipocytes in appearance (Figure 5D). Quantitatively, the relative amount of differentiated lipid-droplet-containing cells on the PVA-C and PVA-CG substrates are significantly higher than that on the control TCPS wells (Figure 5E). The results of immunocytochemical stainings also demonstrated that most of the differentiated lipid-droplet-containing cells on the PVA-CG substrates are immunoreactive with anti-uncoupling protein 1 (anti-UCP1; a marker specific for brown adipocytes) (Klaus et al., 1991; Ricquier, 2017) and the amount of UCP1-positive cells on the PVA-CG substrates is more abundant than on the control TCPS wells (Figure 5F). The results of RT-PCR also found that as comparing with the control TCPS wells, the PVA substrates increased the expression level of adipocyte-related adipsin, hALBP and GLUT-4 mRNAs of differentiated human ES cells (Figure 5G), which suggested that PVA substrates might favor adipogenic differentiation of human ES cells.

In order to elucidate the lineage specification of human ES cells on the PVA-CG substrates, differentiated cells are immunocytochemically stained with anti-nestin (a ectodermal marker), anti-Brachyury (a mesodermal marker), and anti-α-fetoprotein (a endodermal marker). The results indicated that these differentiated human ES cells on the PVA-CG substrates were immune-positive with anti-nestin, but immune-negative with anti-Brachyury and anti-α-fetoprotein (Figure 5H), which suggested that PVA-CG substrates preferably induce human ES cells differentiation toward ectodermal lineage. The results of RT-PCR analysis also demonstrate that as comparing with control TCPS wells, PVA substrates can downregulate the gene expression of GATA4 and GATA6, an endodermal and mesodermal marker, respectively (Figure 5I). These results assumed that PVA-CG substrates might support ectodermal differentiation but suppress mesodermal and endodermal differentiation of human ES cells. Although adipocytes are thought to be derived from mesoderm generally, Billon et al. (2007) have demonstrated that brown adipocytes could also be derived from neuroectodermal neural crest during embryonic development. Here, it is found that mRNA expression of SOX9 and SOX10, two markers highly expressed in the neuroectodermal neural crest (Cheung and Briscoe, 2003; Haldin and LaBonne, 2010), were enhanced and increased on the PVA substrates with time during 2 days of human ES cell culture (Figure 5G). It is also observed that these differentiated human ES cells on the PVA-CG substrates were immune-positive with FoxD3, an important transcription factor expressed in the ectoderm-derived neural crest (Barembaum and Bronner-Fraser, 2005; Teng et al., 2008), but were immuno-negative with other ectodermal neural (MAP-2), glial (GFAP) and epithelial (K15 and K7) markers (Figure 5H). Taken together, these data suggested that PVA-C and PVA-CG substrates are biocompatible and can serve as suitable substrates to adhere, grow and enrich UCP-1-positive brown adipocyte differentiation of human ES cells through ectodermal lineage.