Electrospun fibers were fabricated using a NEU-BM (Tong Li Tech, China) electrospinning instrument with modifications from a previously published protocol (Dziemidowicz et al., 2023). In brief, PCL (∼80 kDa, Sigma-Aldrich, MA, United States) was dissolved at 12% w/v in hexafluoro-2-propanol (HFIP, Sigma-Aldrich, United Kingdom) and stirred until complete dissolution of the polymer overnight. The polymer solution was then loaded into 10 mL syringes and ejected through a 0.8 mm inner diameter blunt-end needle at 2 mL/h charged to 5–7 kV for 2.5 h and 2 h, to produce random and aligned fibers, respectively. The fibers were collected on baking paper on a −3 kV negatively charged mandrel (25 mm in diameter) spinning at 500 rpm and 3,750 rpm to reach random and aligned orientation, respectively at 14 cm distance from the spinneret to the collector. The spinneret was scanning parallel to the collector over a distance of 50 mm at 50 mm/s speed to uniformly distribute the fiber on the mandrel. All electrospinning experiments were conducted at room temperature (RT; 18 °C–23 °C) and at 35% ± 10% relative humidity.

The fiber morphology was characterized using a TM3030 scanning electron microscope (Hitachi, Japan) at 15 kV and 100x/500x magnification, after sputter coating samples with gold for 60 s at a 20 mA current, on aluminum holders. When cells were seeded on the scaffolds, samples were fixed with 4% paraformaldehyde (PFA) for 15 min at RT, washed 3x for 5 min in phosphate saline buffer (PBS), rinsed in distilled water for 15 min, underwent alcohol dehydration grade (30%, 50%, 70%, 2% × 95%, 3% × 100%) for 5 min each and were left to dry overnight. The samples were sputter-coated with gold on carbon adhesive tabs and scanned the following day. Samples were analyzed in FIJI.

Neuroblastoma SH-SY5Y (ATCC) cells were routinely cultured in basal medium (Advanced DMEM:F12 (Gibco, MA, United States; 12634-010)) with 20% fetal bovine serum (FBS, Heat Inactivated, F9665, Sigma-Aldrich, MA, United States), 1% L-Glutamine and 1% Penicillin-Streptomycin (100 U/mL penicillin and 100 μg/mL streptomycin; 15140-122, Gibco, MA, United States) at 37 °C in 5% CO₂. Cell seeding was optimized to obtain high monolayer confluency on glass slides by day 7 (D7). We selected day 1 (D1) and D7 as timepoints for the experimental days to investigate the difference in cell behavior to the surface material. D1 and D7 were chosen to capture early attachment and later network formation, respectively. The cells were counted using Countess 3 (Invitrogen by Thermo Fisher Scientific, United States) and 10⁵ cells were seeded onto glass cover slips (13 mm in diameter) or scaffolds fitted into CellCrown™ inserts (Scaffdex, Finland) for 24 well plate: The cells were seeded in small volume (ca 150 µL) reaching 1.7 × 10⁵/cm² and left in incubator for 2 h to let the cells attach to the fibers. Later, the medium was topped up to 600 µL per well, and the cells were cultured for up to 7 days. Medium was changed on day 3. On experimental days 1 (D1) and 7 (D7) the supernatant medium was collected and stored at −20 °C for further assessment of LDH and cytokine release. The cell-laden scaffolds were visualised using SEM and quantification of mRNA expression change, and immunofluorescence of selected proteins via confocal microscopy.

Lactate dehydrogenase activity in the cell culture media was measured according to the manufacturer’s guidelines (11644793001, Roche, Switzerland). The working solution was prepared freshly before the testing and added 1:1 to the collected culture media thawed on ice, incubated on low rotation for 30 min on a shaker, protected from light for 30 min at RT. Absorbance was measured at 490 nm on a BioTek ELx800 absorbance microplate reader (BioTek, VT, United States). The LDH assay values of the cell medium of cell cultures on the fibers were normalized to the results of the LDH values of cells grown on glass.

The culture media were assessed for the content of fibroblasts growth factor (FGF-)1, FGF-2, follistatin and leptin using the Growth Factor Magnetic Bead Panel 1 (HAGP1MAG-12K 16-plex) (Merck, Kenilworth, NJ, United States). Factors were quantified with the Luminex 200 system (Luminex, TX, United States) employing xMAP technology and running Luminex xPONENT 3.1 software. The concentrations of the factors in the samples were calculated based on a logistic 5P weighted standard curve.

mRNA was extracted from cell lysate on ice using Dynabeads^® mRNA DIRECT™ Kit (cat. #61011, Ambion, Thermo Fisher Scientific, MA, United States) and transcribed to cDNA using cDNA kit ReverseAid First Strand cDNA Synthesis Kit (Ref K1622, Thermo Scientific, Lithuania) according to manufacturer’s guidelines using random hexaprimers. Differential mRNA abundance was quantified using 1 ng cDNA/reaction with 2x SybrGreen (SsoAdvanced Universal SYBR^® Green Supermix, cat.# 1725271, Bio-Rad, CA, United States) and a set of primers (Table 1) by qPCR (CFX Connect^® Real-Time System, Bio-Rad, CA, United States) run according to the manufacturer’s instructions. Ribosomal protein L19 RPL19 was used as a housekeeping gene (Cai et al., 2007). We compared mRNA expression of several structural proteins and common neural and neuronal markers, namely actin (ACTB), β3-tubulin (TUBB3), nestin (NES) doublecortin (DCX), neural cell adhesion molecule (NCAM). We also assessed the mRNA expression of gap junctional proteins connexin 43 (Cx43, GJA1) and connexin 31 (Cx31, GJB3) and ECM proteins, namely filamentous collagen 1 and 3 (COL1A1, COL3A1) and glycoprotein fibronectin 1 (FN1), and adiponectin (ADPN) (Table 1). Biological triplicates in technical duplicates of each sample (glass, random and aligned fibers) from 3 separate experiments were each normalized to their respective surface results from D1. For comparison among the cells cultured on different surfaces, we normalized the data to ACTB (in Supplementary Data 1).

Cells were seeded on scaffolds and glass coverslips at a density of 100,000 (100k) and grown for 1 (D1) and 7 days (D7). At the experimental timepoints, the samples were fixed in 4% PFA for 15–45 min, followed by washing with PBS. Cells at day 7 were permeabilized with 0.1% Triton X-100 for 120 min, compared to 15 min for day 1 samples, to account for increased layer thickness and ensure adequate antibody penetration for confocal imaging. Samples were blocked with 1% bovine serum albumin (BSA) in PBS (BSA-PBS) for 1 h and incubated with the primary antibodies (diluted to working concentration in BSA-PBS as specified in (Table 2)) for 1 h at RT. Samples were washed 3x with PBS before incubating with fluorophore-conjugated secondary antibodies (goat anti-mouse Alexa 568 and Alexa 647, and donkey anti-rabbit Alexa568 (A11031, A21236, A10042, Invitrogen, Thermo Fisher Scientific, MA, United States) diluted 1:500 in BSA-PBS. After 1 h incubation, coverslips were 3x washed with PBS, incubated with FITC-conjugated phalloidin (Phalloidin-Alexa 488 1:200; A12379, Invitrogen, ThermoFisher, MA, United States) for 1h, 3x washed with PBS, incubated with 300 nM DAPI for 10 min, mounted with VectaShield (H-1000, Vector Laboratories, CA, United States) and sealed with clear nail polish. We analyzed filamentous actin (F-actin), acetylated tubulin, β3-tubulin, Cx31, DCX, 24-dehydrocholesterol reductase (DHCR24), FN1, nestin, and SRY-Box transcription factor (SOX2).

Immunofluorescent samples were viewed by confocal microscopy using a Leica SP8 upright microscope fitted with HyD and PMT detectors, using objective 40x HC PL Apo CS2 40x/1.3 (Leica Microsystems, Germany). Excitation laser lines at 405, 488, 552, and 650 nm were used for DAPI, Alexa488, Alexa568, and Alexa 647 respectively in Leica Application Suite X, (version 3.5.6.21594, Leica Microsystems, Germany).

Collected images were analyzed using FIJI (Schindelin et al., 2012), and its plug-ins Bio-Format (Linkert et al., 2010) and 3D ImageJ Suite (Schmid et al., 2010). The confocal images were prepared for publication in Z and 3D projections.

SEM micrographs (n = 6 per fiber type and timepoint) were used to evaluate cluster size, morphology, migration, and coverage of the fibrous mat by the cells by measuring cell-covered areas on the fibers.

To evaluate cell cluster thickness and its incorporation to the fiber mat on D7, Z dimensions of 10 randomly selected 3D cell clusters that fitted scanning area per experimental group were marked and analyzed in micrographs from confocal microscopy via the Leica Application Suite X.

Mitotic index for D1 was calculated as a ratio of mitosis, manually counted, to total number of cells from 5 representative 3D micrographs from confocal microscopy in 3D ImageJ Suite, FIJI. Mitotic index for D7 was presented as total mitosis per scanned area due to technical limitations of the automated cell count. For simplification, D1 mitotic index was recalculated per scanned area.

Statistically significant differences in the quantifiable parameters were estimated using analysis of the variance (ANOVA, GraphPad Prism 9, CA, United States) with a confidence interval of 95% for group comparison in case of normal Gaussian distribution with Tukey post-hoc test, and as min–max and with z25 and z75% and nonparametric Kruskal-Wallis post-hoc test otherwise. In the case of comparison only of 2 groups, we used either a parametric t-test or nonparametric Mann-Whitney t-test, comparing ranks, depending on data distribution. Values of p ≤ 0.05, p ≤ 0.01, and p ≤ 0.001 were considered statistically significant among the groups that were compared. The data are presented as the mean ± standard deviation (SD) and the averaged values were determined from 3 independently prepared biological samples.

Aligned and random orientation morphologies (Figures 2A,B) were characterized by SEM from (Figures 3A,B, Fiber Characterization panel).

Confocal microscopy following FITC-phalloidin and DAPI staining enabled analysis of cytoskeletal architecture and nuclear distribution within clusters across the two fiber types. No significant difference in overall cluster thickness was observed between fiber types (38.4 ± 7.7 µm on random vs. 43.2 ± 5.5 µm on aligned, n = 10), while glass supported significantly lower vertical stratification (29.6 ± 9.9 µm; p = 0.0486 and p = 0.002 vs. random and aligned, respectively). Clusters on aligned fibers penetrated deeper into the mat (31.6 ± 3 µm vs. 25.5 ± 8 µm random; p = 0.04), though the protruding part above the fiber mat was similar (12.9 ± 6 µm vs. 11.5 ± 5.8 µm). The relative incorporation depth (as a percentage of total cluster thickness) was higher on aligned fibers (74.1% ± 10%) compared to random (66.16% ± 15%), but this difference was not statistically significant (Figure 3H).

Figures 4A–I shows cell morphology on D1, while Figures 4J–X depicts D7 outcomes. On D1, cells formed uniform monolayers on glass (Figures 4G–I) but clustered on fibers (Figures 4A–F). Over time, these aggregates expanded into larger structures (Figures 4J–O). On D7, glass-supported cultures retained a largely monolayer architecture with some vertical stacking (Figures 4P–R). In contrast, fiber-based cultures formed compact pseudospheroids (Figures 4J–L) and thick multilayer formations (Figures 4M–O). On random fibers, pseudospheroids were more defined and circular (Figures 4S–U), with a preserved dome-like structure at the base. On aligned fibers, clusters appeared more dispersed, often integrating deeper into the scaffold and decreasing in volume (Figures 4V–X). Pseudospheroids commonly reached diameters of ∼250 µm but remained under 50 µm thick, consistent with spheroidal proportions along the Z-axis. On glass, SH-SY5Y cells exhibited typical hexagonal morphology with broad, flat protrusions on D1 (Figure 4H), whereas cells on fibers displayed elongated shapes conforming to the underlying topography (Figures 4A,D). On random fibers, protrusions were primarily connected to nearby clusters along the shortest path (Figure 4B), similar to those observed on glass (Figures 4G,H). However, on aligned fibers, inter-cluster communication was limited; protrusions predominantly extended along the fiber axis, contacting clusters on the same fiber rather than following minimal-distance paths (Figures 4D,E,M). Prominent lamellipodia, reminiscent of neurites, aligned tightly to fiber directionality by D7 (Figure 4M).

Nuclear morphology also varied with substrate. On glass, nuclei were slightly elongated and oriented toward neighboring cells. On fibers, cells appeared rounder with nuclei conforming to scaffold topography; globular on random fibers (Figure 4C) and elongated on aligned ones (Figure 4F). Increased apoptotic nuclei were noted along the periphery of clusters formed on aligned fibers (Figure 4V). Interestingly, in D7 pseudospheroids, nuclear orientation remained random unless in direct fiber contact, indicating a more localized topographic influence.

To quantify the surface effect on cell proliferation, we assessed the mitotic index using confocal micrographs (n = 5, Figures 5A,B, Proliferation panel), lactate dehydrogenase (LDH) activity to evaluate cytotoxicity (Figure 5C, LDH panel), and cytokine release assessment (Figure 5D, FGF-2 panel). The mitotic index was significantly lower on random fibers compared to the glass surface. On average, the total number of cells on D1 was comparable between groups grown on fibers (149.4 ± 9.7 random vs. 145.4 ± 28.0 aligned, p > 0.99) with fewer cells on glass (127.2 ± 31.2. p = 0.31 vs. random, p = 0.77 vs. aligned).

Due to the dense organization of pseudospheroids on D7, automatic segmentation of total cell counts using FIJI’s 3D Image Suite was not feasible. Therefore, mitotic figures were quantified per scanned area and compared across conditions. The mitotic index on D1 (expressed as mitoses per 10⁶ μm²) was significantly lower on random fibers (0.42 ± 0.24) than on glass (1.00 ± 0.30; p = 0.03), with intermediate values on aligned fibers (0.67 ± 0.38; Figure 5A). No significant differences were found across substrates on D7 (random: 0.87 ± 0.44; aligned: 0.71 ± 0.32; glass: 0.68 ± 0.37; Figure 5B), although the trend was reversed.

The LDH assay, performed across three independent experiments with a total of 10 biological replicates, showed slightly lower values for cells on aligned fibers compared with random fibers at both D1 and D7; however, these differences were not statistically significant (Figure 5C, LDH panel), yet the values of aligned fibers were statistically significantly reduced when compared to glass on both, D1 and D7 (p = 0.04 and p = 0.04, respectively). The cells seeded on glass reached the highest cell confluency among the testing groups as visualized by confocal imaging, suggesting that LDH results reflected the metabolic activity of the alive cells rather than cell death caused by the cytotoxicity of the material.

LDH assay results showed slightly lower values on aligned fibers compared to random fibers on both D1 and D7, but these differences were not statistically significant (Figure 5C, LDH panel). In contrast, LDH release on aligned fibers was significantly reduced compared to glass at both timepoints (p = 0.04; Figure 5C), indicating decreased cell membrane damage on aligned scaffolds. Confocal imaging confirmed that cells on glass achieved the highest confluency. Thus, the elevated LDH levels on glass may reflect increased cell density, stress from crowding, or higher turnover, rather than direct material-induced cytotoxicity. Luminex cytokine analysis from 3 separate experiments containing 10 biological replicates total revealed stable release of FGF-2 on both fiber types, while a significant increase was observed on glass by Day 7 (p < 0.001; Figure 5D, FGF-2 panel), likely reflecting higher confluency due to limited surface area. SH-SY5Y cells under these conditions did not engage in acute wound-like signaling pathways, as indicated by the absence of EGF, angiopoietin-2, G-CSF, BMP-9, endoglin, endothelin-1, leptin, follistatin, VEGF-C, and VEGF-D.

We observed topography-dependent differences in mRNA expression directed by fiber organization, as shown in the Gene Expression panel (Figures 5E–G). To assess how 3D formation affects ECM remodeling and cell communication, we analyzed the expression of structural, neural, gap junctional, and ECM-related genes, along with stress marker adiponectin (ADPN) (Jung et al., 2006) in 3 sets of separate experiments in biological triplicates and technical duplicates.

On random fibers (Figure 5E), mRNA expression of actin (ACTB) and β3-tubulin (TUBB3) significantly increased from D1 to D7, up to 6-fold and over 20-fold, respectively (p < 0.001). In contrast, mRNA abundance of these markers remained relatively stable on aligned fibers, and glass (Figures 5F,G). Nestin (NES) expression showed little change in random fibers, slight downregulation on glass, and a notable decrease on aligned fibers by D7 (Supplementary Data 1C). NCAM levels were similarly stable across all conditions.

DCX mRNA expression increased on all substrates, with a modest 3-fold increase on random, 5-fold on glass, and a 30-fold increase on aligned fibers (p < 0.001). However, when normalized to ACTB, DCX expression decreased over time on random and glass, and remained stable on aligned fibers (Supplementary Data 1D).

Gap junction markers showed inverse patterns. Connexin 43 (GJA1) increased significantly, up to 60-fold on random fibers by D7 (p < 0.01) but remained unchanged on aligned fibers and glass. In contrast, connexin 31 (GJB3) increased markedly on glass (∼20-fold, p < 0.001) and aligned (∼15-fold, p < 0.05), with only a modest rise on random (∼2-fold) (also supported by Supplementary Data 1A).

Fibronectin (FN1) mRNA expression decreased on random fibers but stayed stable on aligned and glass surfaces (Supplementary Data 1C). Collagen mRNA expression varied: COL1A1 was undetectable on glass and minimally expressed on fibers, while COL3A1 increased on glass, decreased on random, and was stable on aligned (Supplementary Data 1A).

ADPN mRNA expression increased across all substrates, with the greatest fold change on glass (∼7-fold), followed by random (∼5-fold) and aligned (∼3-fold) fibers Supplementary Data 1B. GAPDH expression also rose from D1 to D7, with high variability, peaking at ∼350-fold on random, >100-fold on glass, and ∼20-fold on aligned fibers (not shown).

In summary, mRNA analysis revealed substrate-driven differences in gene expression, particularly among cytoskeletal (ACTB, TUBB3), gap junctional (GJA1, GJB3), and early differentiation and migration marker DCX. These patterns reflect how scaffold architecture can distinctly influence early neural gene programs and intercellular communication.

To assess cell and pseudospheroid morphology across the different substrates, we first performed 3D cytoskeletal and nuclear imaging using fluorescent phalloidin and DAPI staining, followed by confocal microscopy (D1 in Figure 6, D7 in Figure 7).

Aiming to examine protein localization within the pseudospheroids, we carried out immunofluorescent staining for two well-established markers of neuronal identity and cytoskeletal remodeling, namely acetylated and β3-tubulin (top panels, Figures 6, 7). Acetylated tubulin marks stabilized microtubules critical for cytoskeletal integrity and mitotic spindle function, especially in post-mitotic and elongating cells (Nekooki-Machida and Hagiwara, 2020) and β3-tubulin is a microtubule element associated with early neuronal differentiation (Spildrejorde et al., 2023). Acetylated tubulin was abundantly expressed on both fibrous substrates but showed lower expression on glass. On D7, cells on random fibers exhibited more uniform and widespread acetylated tubulin staining, whereas cells on aligned fibers displayed a distinct network of thin cytoplasmic extensions. The expression pattern of β3-tubulin on D7 was consistent with the qPCR findings. DCX expression increased from D1 (Figure 6) to D7 (Figure 7) on both glass and random fibers in agreement with qPCR data, while no such increase was observed on aligned fibers (see Supplementary Data 1A). Nestin expression remained stable from D1 to D7 on both fiber types but declined on glass by D7, also matching qPCR trends (middle panel on Figure 5, second panel on Figure 7). FN1 protein expression decreased on fibers between D1 (Figure 6) and D7 (Figure 7), while remaining stable on glass, mirroring mRNA results (lower panel on Figure 6, third panel on Figure 7).

On D7, we also detected DHCR24 and SOX2 expression (Figure 7). DHCR24 (24-dehydrocholesterol reductase) is involved in cholesterol biosynthesis and implicated in neuroprotection and cellular stress response. SOX2 is a transcription factor essential for maintaining neural progenitor identity and pluripotency, and its presence may indicate retention of undifferentiated or stem-like characteristics within the culture (Samara et al., 2022). DHCR24 was evenly expressed in cells on glass and random fibers but was restricted to the periphery of pseudospheroids on aligned fibers (Samara et al., 2014). SOX2 was observed in the nuclei of cells on aligned fibers and glass but was predominantly cytoplasmic in cells on random fibers, particularly at the edges of pseudospheroids.

Overall, the confocal microscopy data support the SEM observations and gene expression results, confirming distinct substrate-specific effects on cytoskeletal organization, stress responses, and early differentiation markers.