Figure 1a shows a schematic diagram illustrating our basic experimental procedure. CNT patterns were prepared according to previously-reported methods.^20–21 Briefly, a self-assembled monolayer (SAM) of methyl-terminated 1-octadecanethiol (ODT) was first patterned on thin Au films on cover glass substrates by micro-contact printing, while leaving some bare Au surface regions unaltered (see Supplementary method). When the patterned substrate was placed in CNT suspensions (0.05 mg/ml in 1,2-dichlorobenznene), CNTs were selectively adsorbed onto bare Au regions forming CNT monolayer patterns. The CNT patterns were then placed in laminin solution (10–20 μg/ml) for 10–30 min so that laminin molecules were selectively adsorbed onto the CNT patterns. Laminin is one of the ECM components that is helpful for hNSC adhesion and growth. After cell seeding, the hNSCs grew preferentially along these laminin-coated CNT patterns in the culture media with growth factors, such as basic fibroblast growth factor (bFGF) and epidermal growth factor (EGF). For hNSCs, the growth factors (bFGF and EGF) are known to enhance hNSC growth and proliferation, while blocking the differentiation process. Afterwards, the substrate was placed in culture media without bFGF and EGF for two weeks to study the differentiation of hNSCs on laminin-coated CNT patterns.

Figure 1b shows the scanning electron micrograph (SEM) image of the prepared CNT patterns. It shows the well-defined CNT regions (darker square regions) as well as ODT-coated area (lighter region). The high resolution atomic force microscopy (AFM) image also confirmed the highly-selective adsorption of CNTs on bare Au regions (Figure S1a in Supporting Information).

When placed in laminin solution, CNT patterns selectively adsorb laminin molecules from solution. This was verified by immunochemistry (Figure 1c). For this purpose after the laminin adsorption, the substrate was placed in the fluorescent-labeled anti-laminin solution so that the anti-laminin molecules would bind to the laminin molecules on the substrate. The fluorescence image shows much stronger fluorescence intensity in the CNT regions (brighter green regions on Figure 1c) than on ODT regions, confirming the high-density adsorption of laminin molecules on the CNT patterns. It is also consistent with previous reports regarding the preferential adsorption of protein molecules to CNT sidewalls,^22–23 and the resistance of alkyl chains of the ODT SAM to laminin adsorption.²⁴ The CNT patterns with laminin coating were also investigated via an AFM topography image (insert in Figure 1c and Figure S1b in Supporting Information). It exhibited the average roughness of 26 nm, which is in the optimal range of surface roughness (20~50 nm) promoting the adhesion and longevity of primary neurons.^25,26 This result indicates that the nanotopographic cues of CNT network structures as well as laminin molecules adsorbed on the CNT patterns can synergistically induce the selective growth of hNSCs.

The biocompatibility of the CNT network structure as a substrate for hNSC growth was investigated via cell viability assay using flow cytometry. For the assay, the adherent hNSCs were detached from the CNT patterns after three-day growth and three-day differentiation period, respectively. After the three-day growth period, 98% of the cells were found to be viable (Figure 1d). The assay result of a three-day differentiation also exhibited nearly 97% of cell viability (Figure S2 in Supporting Information). This suggests the good biocompatibility of CNT patterns for hNSC growth and differentiation. Furthermore, we utilized the Western Blot method to confirm the protein expression of hNSCs before and after the differentiation (Figure S3 in Supporting Information). The results show that the hNSCs grown with the growth factors (EGF and bFGF) were positive for neural stem cell markers (nestin and SOX2), which shows that they just proliferated and undifferentiated. Meanwhile, those grown without these growth factors were positive for glial fibrillary acidic protein (GFAP) and neuron-specific class III beta-tubulin (TUJ1), which indicates that they differentiated.

When the hNSCs were seeded on the laminin-coated CNT patterns in the culture media with the growth factors, they selectively adhered onto the CNT pattern regions and grew along the patterns (Figure 2a, 2d and 2g; left column in Figure 2). In this stage, the growth factors blocked the differentiation of hNSCs. When the substrate was placed in the culture media without the growth factors, the hNSCs started to differentiate (Figure 2b, 2e and 2h; middle column in Figure 2). The differentiation was confirmed by immunocytochemistry (Figure 2c, 2e and 2i; right column in Figure 2). Here, we used three different markers to look at cytoskeletal distributions on the CNT patterns after the differentiation: GFAP as an astroglial cell marker (Figure 2c), and neurofilament light (NF-L, Figure 2f) and TUJ1 (right in Figure 2i) as neuronal cell markers. We also performed an experiment to investigate the hNSC growth on CNT networks compared with that on conventional substrates such as coverglass (Figure S5 in Supporting Information). After the hNSC seeding on the laminin-coated CNT patterns that were prepared on coverglass (Figure S5A), we observed that the hNSCs grew selectively in the CNT regions (Figure S5B, S5C, and S5D). This result clearly indicates that the CNT network can provide a better extracellular environment for hNSC growth than conventional cell-culture substrates such as coverglass.

Depending on the geometries of CNT patterns, the hNSCs exhibited significantly different outgrowing behaviors during growth and differentiation (Figure 2a-2c). When the size of the CNT square patterns was large enough (300 μm × 300 μm, 200-μm spacing) to hold multiple cells, the hNSCs in the CNT patterns could maintain their cell-cell interactions and proliferated very well (Figure 2a). Eventually, they outgrew over the 200-μm-wide ODT SAM regions toward the adjacent CNT square patterns and formed the neural networks, where the cells grown on the distanced CNT square patterns were connected (Figure 2b). The fluorescence image clearly shows that the outgrowing astrocytes (green regions marked as GFAP) were connecting the hNSC population on the distanced patterns after differentiation (Figure 2c)

We then reduced the size of CNT square patterns (50 μm × 50 μm, 50-μm spacing) such that each square could hold only a single hNSC (Figure 2d-2f). In this case, the hNSC outgrowth was extremely restricted during the growth and differentiation process (Figure 2d). Even after we removed the growth factors to induce differentiation, the hNSCs did not exhibit any indication of major outgrowth over the ODT regions (Figure 2e). The fluorescence image of neuronal cytoskeletons (NF-L, red) does not show any outgrowing hNSCs from the patterns (right in Figure 2f). This result clearly shows that the cell-cell interaction can be controlled by the geometries of CNT patterns, which can be critical for hNSC growth and differentiation.

We also tested hNSC behaviors on CNT square patterns smaller (5 μm × 5 μm, 5-μm spacing) than individual hNSCs (Figure 2g–2i). Here, the hNSCs first adhered and outgrew over several CNT square patterns (Figure 2g). Note that each cell was bound strongly on the small CNT pattern regions and outgrew and extended over the ODT regions. In this case, the spacing of the CNT square patterns should significantly affect the cytoskeletal tensions of the individual hNSCs, which probably should affect the differentiation of stem cells.²⁷ After the differentiation, we could observe that the neuronal outgrowths extended and bound on the nearby CNT patterns (Figure 2h). The fluorescence image clearly shows the neuronal cytoskeletal marker (TUJ1, green) indicating the connected neural networks bound on the small CNT patterns. Overall, the results in Figure 2 clearly show that the size and spacing of CNT patterns can play a critical role in controlling the hNSC outgrowths during the growth and differentiation process, which can possibly affect cell-cell interactions or cytoskeletal tensions.

Furthermore, line-shape CNT patterns can be utilized to control the neuronal orientation with high precision (Figure 3). In the line-shape CNT patterns with a line width (30-μm width, 60-μm spacing) which can hold two or three cells, the hNSCs adhered (Figure 3a) along the line pattern. We observed that they differentiated to form neural networks along inside the line patterns (Figure 3b). Here, the differentiation was also confirmed by immunocytochemistry with the TUJ1 marker. When the CNT line width was narrowed down to about 5 μm, which can hold only a single hNSC, the hNSCs grew and differentiated into bipolar shapes along the individual CNT line patterns (Figure 3f). Significantly, this result indicates that we can control the orientation of hNSCs with a single-cell-level precision.

When circle-shape patterns are connected with narrow-line-shape ones, we observed quite an interesting hNSC behavior during the growth and differentiation (Figure 3g–3i). After seeding in the culture media with growth factors, the cell bodies of hNSCs tended to adhere and proliferated on the circle-shape pattern regions (Figure 3g). After withdrawal of the growth factors in the culture media, they started to differentiate and the outgrowing neurites were observed mostly along the narrow line-shape CNT pattern regions (middle in Figure 3h). The differentiation was also confirmed via immunostaining (right in Figure 3i). Since the hNSCs first adhered and grew on the circle-shape patterns, their nuclei (blue regions) were mostly located on the circle regions, while the long neurites (red regions) extended along the line-shape regions (Figure 3i). This result indicates that the CNT patterns can be utilized to control both of the locations of cell nuclei and the direction of neurite growth, thus allowing us to control the structural polarization of the neuronal differentiation of hNSCs. Furthermore, the synapse formation of the neurons was checked by a neuronal presynaptic vesicle marker, synaptophysin (Figure S6 in Supporting Information). The results clearly show that the neurons differentiated from the hNSCs grown on the CNT patterns can also form the synapses, which are important for a neuron to pass a chemical/electrical signal to another neuron.

For future therapeutic applications such as regenerative medicine, it would be crucial to apply our strategy to a flexible and biocompatible substrate such as polyimide (PI) (Figure 4) which has been widely utilized for implantable neural devices such as three-dimensional artificial nerve conduits²⁸ and stimulating electrodes.^29–30 We prepared CNT patterns on thin-Au-film-coated PI substrates and performed the experiments of hNSC growth and differentiation on them (Figure 4a). We could achieve high-quality CNT patterns on the Au-coated PI substrates as shown in the SEM images (Figure S7a in Supporting Information). The immunofluorescence image indicates the highly-selective adsorption of laminin onto the CNT patterns on the PI substrate (green regions in Figure 4b). After seeding on it, the hNSCs adhered selectively onto the CNT pattern regions on PI substrates and proliferated (Figure 4c). Eventually, we achieved the orientation-controlled growth and differentiation of hNSCs along the CNT patterns on the flexible PI substrate (Figure 4D, Figure S7b in Supporting Information).

Finally, the structural-polarization-controlled differentiation of individual hNSCs can be achieved by CNT patterns comprised of one square- and one line-shape ones (Figure 5a). Here, the width of the line-shape region is much smaller than the size of an individual hNSC. After cell seeding we were able to observe the selective hNSC adhesion inside the square regions (Figure 5b). Then, the hNSCs on the square region outgrew along the narrow line-shape regions during the growth and differentiation stages (Figure 5c). The neuronal differentiation was confirmed by growth associated protein 43 (GAP 43, green in Figure 5d) which is known to be expressed in the growth cone regions of neural cells. We observed that GAP 43 was also highly expressed on the line-shape CNT regions, indicating the neurites outgrew along the line-shape regions (Figure 5d, Figure S8 in Supporting Information).

We carried out immunocytochemistry to check the neural lineages of the differentiated cells on these CNT patterns (Figure 5e). Here, GFAP and TUJ1 indicate astroglial and neural cells, respectively. To confirm their lineages, the relative fluorescence intensities of GFAP and TUJ1 from the cell nuclei on the square pattern regions were quantified using the similar method reported previously (Figure S9 in Supporting Information).³¹ The result shows that 20% of them were TUJ1-positive, whereas another 20% were GFAP-positive. It should be noted that the hNSCs were differentiated with controlled-structural-polarity on the CNT patterns, while maintaining their capabilities to differentiate into the main phenotypes in the nervous system, such as neuronal or astroglial cells.

In summary, we demonstrated a structural-polarization-controlled neuronal differentiation of hNSCs using the patterns of CNT monolayer coating. Due to the synergistic effect of CNT network structures for selective laminin adsorption and optimal nanotopography, we could effectively promote the selective growth of hNSCs on the CNT patterns. The result of the cell viability assay (>97%) suggested the good biocompatibility of CNT patterns for hNSC growth and differentiation. We also confirmed that CNTs could induce the adhesion and growth of hNSCs even better than conventional cell-culture substrates such as bare glass. Importantly, the structural-polarization-controlled neuronal differentiation was demonstrated at the level of an individual axon or neurite. Furthermore, we also applied it to flexible and biocompatible PI substrates, which should significantly expand the possible therapeutic applications of our method. Since CNT monolayer coatings can be applied to versatile substrates including flexible ones and provide a better cell-growth environment than conventional cell-culture substrates such as glass, our strategy should provide many new opportunities in various areas such as neural engineering, stem cell therapy, and regenerative medicine.