Comparing the effects of uncoated nanostructured surfaces on primary neurons and astrocytes

The long-term application of central nervous system implants is currently limited by the negative response of the brain tissue, affecting both the performance of the device and the survival of nearby cells. Topographical modification of implant surfaces mimicking the structure and dimensions of the extracellular matrix may provide a solution to this negative tissue response and has been shown to affect the attachment and behavior of both neurons and astrocytes. In our study, commonly used neural implant materials, silicon, and platinum were tested with or without nanoscale surface modifications. No biological coatings were used in order to only examine the effect of the nanostructuring. We seeded primary mouse astrocytes and hippocampal neurons onto four different surfaces: flat polysilicon, nanostructured polysilicon, and platinum-coated versions of these surfaces. Fluorescent wide-field, confocal, and scanning electron microscopy were used to characterize the attachment, spreading and proliferation of these cell types. In case of astrocytes, we found that both cell number and average cell spreading was significantly larger on platinum, compared to silicon surfaces, while silicon surfaces impeded glial proliferation. Nanostructuring did not have a significant effect on either parameter in astrocytes but influenced the orientation of actin filaments and glial fibrillary acidic protein fibers. Neuronal soma attachment was impaired on metal surfaces while nanostructuring seemed to influence neuronal growth cone morphology, regardless of surface material. Taken together, the type of metals tested had a profound influence on cellular responses, which was only slightly modified by nanopatterning.


| INTRODUCTION
A great deal of research has been performed on central nervous system (CNS) implants to help patients suffering from diseases such as amyotrophic lateral sclerosis (ALS), spinal cord injury or paralysis.
Effective long-term usage of such devices is limited by the defensive reaction of the CNS resulting in neuronal loss and glial scar formation.
These events lead to the signal obstruction between neurons and electrodes during long-term implantation, degrade the performance of the neural electrodes causing instability, and eventually, the failure of the implanted device. The main aims of implant development are to improve neuronal survival and unimpeded regeneration and extension of neurites, while preventing microglial and astrocyte activation by keeping them from attaching to the implanted surface [see (Adewole, Serruya, Wolf, & Cullen, 2019;Fernandez & Botella, 2018;Jorfi, Skousen, Weder, & Capadona, 2015;Kim et al., 2018) for review].
One of the recent strategies is the topographical modification of neural implant surfaces, as imitating the structure of the extracellular matrix (ECM) can influence the attachment and behavior of neural cells (Jeon, Simon Jr, & Kim, 2014;Kim et al., 2018). The micro−/nanostructure of the implant surface can have a selective effect on astrocytes and neurons, demonstrated previously both in vitro and in vivo (Berces et al., 2016;Moxon et al., 2004;Moxon, Hallman, Aslani, Kalkhoran, & Lelkes, 2007;Piret, Perez, & Prinz, 2015). Proposed explanations by which nanostructuring results in better biocompatibility include the formation of mechanical cues similar to the ECM, and/or the adsorption of growth factors and other molecules facilitating the survival of neurons. However, the exact mechanisms involved are not yet clear (Marcus et al., 2017).
Many of the commercially available neural implants use silicon as a carrier material and platinum for the electrodes (Kotov et al., 2009). Platinum and silicon have been extensively characterized both in vivo and in vitro for their biocompatibility with neuronal cells and tissue (Biran, Martin, & Tresco, 2007;Ereifej et al., 2011;Griffith & Humphrey, 2006;Mols, Musa, Nuttin, Lagae, & Bonin, 2017;Pennisi et al., 2009;Polikov, Tresco, & Reichert, 2005). A wide variety of nanostructure types and sizes of these materials has been created and tested so far (Jeon et al., 2014;Kim et al., 2018;Kotov et al., 2009;Marcus et al., 2017), often in combination with the application of different ECM-like surface coatings (von der Mark, Park, Bauer, & Schmuki, 2010). On the other hand, the exact modifications of the nanopatterned surfaces generated by biomimetic coatings are hard to describe which further complicates the interpretation of the experimental findings (Kim et al., 2018).
Previously, our group has established the fabrication of so-called black polysilicon (Fekete, Horvath, Berces, & Pongracz, 2014), referred to as nanostructured silicon in this article. Such surfaces were created by large-area, maskless, and cryogenic plasma etching. This technology could be integrated easily into the manufacturing steps of silicon-based multichannel neural microelectrodes (Fekete, 2015). Earlier, we demonstrated that neuronal survival was increased in the vicinity of an uncoated black polysilicon implant surface 8 weeks after implantation, while the rate of glial activation was unaffected by nanostructuring (Berces et al., 2016).
In an attempt to more closely examine the initial cellular reactions behind these effects, we investigated the attachment and growth of primary mouse astroglial cells and hippocampal neurons on these metal surfaces, until a confluent cellular layer was formed. Tested materials included vapor deposited polycrystalline silicon (referred to as the "flat" surface) and its nanostructured counterpart created by photolithography, as well as the platinum sputter-coated version of both of these surfaces. As we wished to investigate the initial effects of cell-surface contact, cell behavior was analyzed during the first 3 days in culture. In order to more directly compare our in vitro and in vivo results, no additional surface coating to facilitate cell attachment was applied.
We found that nanostructuring in itself did not have a marked effect on the attachment, spreading and proliferation of astrocytes, while neuronal growth cones seemed to differentiate between flat and nanostructured surfaces. On the other hand, the attachment of neuronal soma was highly impaired on both metal surfaces.

| Design and manufacture of the in vitro test chips
Test chips for in vitro cell culturing were fabricated by standard MEMS processes in a way to contain all four different surfaces. In our study, flat polysilicon formed by low-pressure chemical vapor deposition or nanostructured polysilicon produced by cryogenic dry etching were used as seed layers and platinum deposited by DC magnetron sputtering was applied as coating material. To make well-defined interfaces between all four materials, we employed a specific chip design shown in Figure 1b. The manufacturing process is described in detail in an earlier publication by our group (Berces et al., 2018).
The four different experimental surfaces-flat polysilicon, nanostructured polysilicon, flat platinum, nanostructured platinum-were characterized by scanning electron microscopy ( Figure 1a). In case of the nanostructured silicon surfaces, the height of the nanopillars was between 520 and 800 nm and pillar density was 18-70 pillars/μm 2 , with pillar diameters of 80-150 nm.

| Animal handling
Wild-type CD1 mice were obtained from Charles River Laboratories (Wilmington, Massachusetts) and housed at 22 ± 1 C with 12-hr light/dark cycles and ad libitum access to food and water. All experiments complied with local guidelines and regulations for the use of experimental animals (PEI/001/1108-4/2013 and PEI/001/1109-4/ 2013), in agreement with local and EU legislation.

| Primary cell cultures
Primary astrocytes were prepared postnatally from 1 to 4 days old mouse pups essentially according to a previously described method (Tarnok et al., 2010). Cultures were maintained in HDMEM (Sigma) with 10% FCS (Gibco), 2 mM glutamine (Sigma), 40 μg/mL gentamicin (Hungaropharma, Budapest, Hungary) and 2.5 μg/mL amphotericin B (Sigma). Cells were allowed to proliferate and passaged at least twice with 0.05% trypsin -0.02% EDTA (Sigma) before being seeded onto test chips. Test chips were dry-heat sterilized at 180 C for 4 hr then placed in 24-well culture plates without any further surface treatment.
2.4 | Immunocytochemistry, microscopy, and image processing Cells were fixed with 4% paraformaldehyde (TAAB), permeabilized with 0.1% Triton-x-100 in phosphate buffered saline (PBS) and blocked using Samples were investigated by a Zeiss Axio Observer Z1 or LSM800 inverted fluorescence microscope. Images were captured by an AxioCamMR3 camera or GasP detectors using ZEN software.
Whole-chip scans were acquired by a mosaic-type image stitching technique using individual images of 10× magnification (obtained with a Plan-Neofluar 10×/0.30 objective). Individual images were captured by a Plan-Apochromat 63x/1.4 oil immersion objective and deconvoluted by the nearest neighbors method before z-projection.
To manually analyze the density of DAPI-stained astrocyte nuclei, the Cell Counter plugin of FIJI (Schindelin et al., 2012)

| Scanning electron microscopy
Cells were fixed with 2.5% glutaraldehyde (Sigma) + 5% saccharose in 0.1 M cacodylate buffer for 1 hr at RT and dehydrated using increasing Chips were coated diagonally with platinum, resulting in flat platinum (Pt; light gray) and nanostructured platinum (nano-Pt; dark gray) concentrations of ethanol (50, 60, 75, 90, and 100%), and amyl-acetate (Sigma). Dried samples were sputter coated with gold for scanning electron microscopy. Samples were imaged using a LEO XB1540 (Zeiss) scanning electron microscope. Tilt angles are stated in figure legends.
Surface characterization was performed using the ImageJ software.

| Statistics
The numbers of independent samples tested and the numbers of data points per experiment are noted in the legends of Figures 2 and 6. Statistical analyses for the results shown in Figure 2 were performed with SPSS Statistics (IBM). Normal distribution of the samples was evaluated using the Shapiro-Wilk test. Data was analyzed using one-way ANOVA tests with post hoc Bonferroni corrections or the nonparametric Kruskal-Wallis test with pairwise comparisons. A p-value equal to or lower than .05 was considered as a statistically significant difference.  (Figure 2g) after seeding. Astrocytes were visualized by phalloidin staining of the actin system. In agreement with our quantitative measurements, astrocytes spread more on platinum and achieved significantly higher confluence by 48 hr compared to the silicon surfaces. It is also evident from the images that astrocytes were generally smaller on silicon (Figure 2f,g).
Using widefield fluorescent microscopy, a notable difference in phalloidin (Figure 2f,g) fluorescent signal intensity over flat or nanostructured surfaces was evident. These differences are mainly due to the increased light absorbance of the nonreflecting nanostructured surfaces (Fekete et al., 2014). On the other hand, astroglial cells showed stronger fluorescence over silicon surfaces comparing to the same type of platina surfaces, especially 48 hr after seeding. Occasionally, neurites spread out to the noncovered uncoated metal surfaces, as well ( Figure 6). Scanning electron microscopy revealed that growth cones attached to nanopatterned surfaces had narrow lamellopodia separated by several filopodia. In contrast, growth cones attached to flat surfaces did not possess distinct lamellopodia and often tapered to a point (Figure 6a).

| DISCUSSION
Silicon and platinum have been characterized extensively in terms of biocompatibility for implantable device applications (Biran et al., 2007;Ereifej et al., 2011;Mols et al., 2017;Pennisi et al., 2009;Polikov et al., 2005), but fewer publications tested uncoated silicon or platinum surfaces. (Biran, Martin, & Tresco, 2005;Pennisi et al., 2009) As bare surfaces are less biocompatible, biomimetic coatings are often used to improve their performance (Adewole et al., 2019;Fernandez & Botella, 2018;Jorfi et al., 2015;Polikov et al., 2005). Part of previous research on the effect of nanostructuring on neural cells in vitro involved surfaces additionally treated with molecules such as poly-Dlysine/poly-L-lysine and laminin to aid cell adhesion and survival on otherwise biologically inert materials (Bugnicourt, Brocard, Nicolas, & Villard, 2014;Huang et al., 2018). Use of biomimetic coatings is also a promising strategy in itself for attenuating the negative tissue response to CNS implants (Aregueta-Robles, Woolley, Poole-Warren, Lovell, & Green, 2014;Jorfi et al., 2015), however, little is known about their persistence, longevity or adverse effects in an in vivo setting (Adewole et al., 2019;Chen, Canales, & Anikeeva, 2017;Cody, Eles, Lagenaur, Kozai, & Cui, 2018;He, McConnell, & Bellamkonda, 2006;Rao & Winter, 2009). So far, only the lack of coating degradation in response to the insertion process was shown (He et al., 2006). Therefore, it is important to examine whether the modification of implant surface topography in itself is capable of significantly affecting neural cell behavior.
Consequently, we extended our previous in vivo and in vitro studies by testing chronic responses to similar implant surfaces within the brain (Berces et al., 2016) or acute effects on immortalized neural stem cells and microglia (Berces et al., 2018), respectively. We aimed to compare how primary astrocytes and hippocampal neurons attach to flat or nanostructured silicon or platinum surfaces without additional coating. Our primary goal was to examine the effect of these materials on the spreading and proliferation of astrocytes and on neurite outgrowth within the initial days after seeding, until glial cells reach confluency.
In our study, neither bare silicon or platinum induced an acute cytotoxic effect on neurons or astrocytes during the experimental period, in agreement with previous studies (Ereifej et al., 2011;Kang et al., 2016). In accordance with previous studies, we found that direct attachment of primary hippocampal neurons to both uncoated metal surfaces was impaired (Khan, Auner, & Newaz, 2005;Piret, Perez, & Prinz, 2014). Earlier in vitro studies involving primary neurons that did not utilize additional surface treatment showed similar clusters of neurons sitting on top of glial cells in case of both nonstructured (Piret et al., 2014) and nanostructured silicon surfaces (Khan et al., 2005). In other cases, the morphology of the examined cells was clearly not neuronal (Ma, Liu, Xu, & Cui, 2005). There are also reports where neurons were apparently able to directly attach to uncoated silicon surfaces, but some of these surfaces also had a cytotoxic effect after 5 days of culture  or the presence of surface coating was not clearly stated (Kang et al., 2016).
In case neurites did attach to the tested surfaces, growth cone formation appeared to be promoted by nanostructuring. Similarly, an increase in neurite outgrowth on nanostructured relative to smoother surfaces has been demonstrated in several studies (Bugnicourt et al., F I G U R E 5 Scanning electron microscopy images of aggregated primary hippocampal neurons attached to primary astrocytes on flat polysilicon (Si) or platinum (Pt) surface 24, 48 or 72 hr postseeding. Bars denote 10 μm 2014; Moxon et al., 2004;Moxon et al., 2007). Kang and colleagues also showed enhanced and more directed neurite extension on silicon nanowires, however, growth cones on top of the nanowires were narrower compared to those on flat silicon (Kang et al., 2016). It is important to note that the apparent differences compared to our results might be explained by different surface composition and/or coating between the studies.
Our results regarding the behavior of astrocytes are in agreement with another study focusing on the effects of uncoated, nonstructured silicon and platinum surfaces on several aspects of glioblastoma behavior in vitro (Ereifej et al., 2011;Ereifej et al., 2013). In the report by Ereifej and colleagues, the observed effects of silicon and platinum were attributed to increased glial reactivity, which was not examined directly in our study. Due to differences in the reflectivity of the tested metal surfaces, we could not reliably compare the fluorescent intensity of anti-GFAP immunostaining as an indicator of astrocyte reactivity. However, our results showed increased proliferation as well as greater spreading of astrocytes grown on platinum, indicating that uncoated platinum provides a more suitable surface for primary astroglial cells than silicon. Nanostructuring itself did not affect cell spreading but cytoskeletal orientation was changed and resulted in thinner GFAP fibers, which were more radially oriented.
Pennisi and colleagues also found that the morphology and proliferation of a glial cell line was not markedly affected on either of the examined nanostructures (Pennisi et al., 2009). In other studies, changes in gene expression point toward a reactivity reducing effect of nanostructured surfaces (Ereifej et al., 2013;Ereifej et al., 2018).
In line with the above results, previous work by our group (Berces et al., 2016) and others (Chapman et al., 2017;Moxon et al., 2007;Piret et al., 2014) reported that nanostructuring affects primary astrocytes/glial cell lines and primary neurons/neuronal cell lines in a different manner. in vivo results also demonstrated no effect of microstructuring on astrocyte reactivity while the number of surviving neurons was positively affected by the topographical modification (Moxon et al., 2007). Neuronal attachment was unaffected by nanoporous gold surfaces, while astrocytic coverage was decreased compared to a nonstructured surface of the same material (Chapman et al., 2017). Additionally, Piret and colleagues were able to separate primary neurons from glia with vertically grown nanowires, although a marked separation of the two cell types was only achieved in case of alternating large contiguous flat and nanostructured surfacesnarrower arrays or single rows of nanowires did not induce such an effect (Piret et al., 2015). Interestingly, we observed that astrocytes frequently aligned along surface boundaries between flat and nanostructured regions, which might be explained by a sensitivity to steep changes in surface architecture. Horizontal line through data points shows median value. All data were obtained from four independently seeded test chips. Data points per experiment varied between 1 and 5 reported to best attach to and survive on surfaces with average surface roughness (Ra) values ranging from 20 to 100 nm. Both lower and higher values were found to negatively affect cell adhesion and viability, albeit to a different degree in different studies Fan, Cui, Hou, et al., 2002;Ma et al., 2005). Other studies concluded that pillar height of nanostructures influences the cellular adhesion and viability and can determine whether cells spread on top of the nanostructures or grow into the trenches between them (Choi et al., 2007;Piret, Perez, & Prinz, 2013). It can be speculated that the feature size of our nanostructured surfaces inhibited the adhesion of the neuronal somas. On the other hand, implantable device performance does not necessarily depend on the attachment of the neuronal somas, however, the unimpeded-and possibly guided-growth of neurites along the implant surface would be crucial to improve long-term functioning.
It must be noted that it is difficult to directly compare results from different groups related to this aspect due to large variations in fabrication methods, applied surface treatments, cleaning protocols and the resulting features themselves in terms of shape, size, and distribution. The exact feature dimensions different research groups choose to publish are also not uniform [see (Marcus et al., 2017) for review].
Therefore, further investigation is needed in order to clarify the effects of nanostructured materials on neural cell types.
In conclusion, we detected that nanostructuring of artificial silicon and platinum surfaces without any biomimetic coating do not affect the attachment and morphology of astrocytes. The type of surface material, on the other hand, had profound influence on cellular responses, further emphasizing that metal implants are less suitable for potential in vivo usage compared to other more promising materials Feiner & Dvir, 2018).