Characterization of spiral ganglion neurons cultured on silicon micro-pillar substrates for new auditory neuro-electronic interfaces

Micro-pillars were introduced in silicon substrates as described previously by Micholt et al (Micholt et al 2013). Briefly, in a cleanroom (Class 1000) environment, standard 8 inch silicon wafers were covered by a low temperature oxide layer, followed by a sintering step at 455 °C. Areas with diameters down to 1 μm and a minimal spacing of 600 nm were defined using standard I-line lithography. 3 μm high pillars were created by a timed reactive ion etch step (RIE), followed by 'piranha' (1:3 H₂0:H₂S0₄) cleaning. The substrate consisted of individual areas micro-patterned with hexagonal pillars of different dimensions. The pillar widths ranged from 1–5.6 μm (1, 1.2, 1.4, 1.6, 1.8, 2, 2.4, 2.8, 4, 5.6 μm) in the vertical direction, while the spacing ranged from 0.6–15 μm (0.6, 0.8, 1, 1.2, 1.4, 1.6, 1.8, 2.0, 2.4, 3.2, 4, 5, 7, 10, 15 μm) as shown in figure 1(A). Separate substrates were then diced in samples of 8 × 8 mm² to be used for cell culturing.

Zoom In Zoom Out Reset image size

The institutional animal care and use committee at the University of Split approved all protocols and experimental plans used in this study. Spiral ganglia dissection and culturing protocols were adapted from (Szabo et al 2003, Bostrom et al 2007, Richardson et al 2007, Vieira et al 2007, Clarke et al 2011). Spiral ganglion neurons (SGN) were isolated from postnatal day 5 rats (P5) and decapitated under cold anaesthesia after placing them on ice. Phosphate buffer saline (PBS) supplemented with 0.3% bovine serum albumine (BSA, Sigma-Aldrich) and 0.6% glucose (Sigma-Aldrich) was used as the dissecting buffer (Mattotti et al 2012). The skull was opened mid-sagitally under an operating microscope and the brain was removed. The temporal bone was harvested and transferred to clean dissecting buffer. Then the otic capsule was dissected, and the cochlea were identified and isolated. The organ of Corti and modiolar cartilage were removed and the spiral ganglia collected in the dissecting buffer.

Silicon micro-pillar substrates (MPSs) were cleaned overnight with acetone (Merck), sprayed with 70% ethanol (Medimon) and left to dry under sterile conditions. Glass coverslips were sterilized by autoclaving at 120 °C for 20 min. All samples were then placed in 24 well plates (TPP) for testing. All substrates were coated with 0.01% poly-D-ornithine (WT 30 000–70 000, Sigma) at room temperature overnight, cleaned with sterile water and allowed to dry.

The dissociation of SGN was performed enzymatically in 0.25% trypsin-EDTA (Sigma-Aldrich) at 37 °C for 20 min. The trypsinization was stopped by adding an equal volume of DMEM:F12 (Gibco) supplemented with 10% of Fetal Bovine Serum (FBS, Biochem). DNase (Sigma) was also added at 38 U ml⁻¹. The tissue was then triturated; large pieces were allowed to settle down and the cell suspension was collected. Trypsinization was repeated one more time for non-triturated tissue pieces. The cell suspension was then centrifuged for 5 min at 1000 rpm and the pellet was resuspended in the culture medium, which consisted of Neurobasal-A (Gibco) containing 1% Pen-Strep (Lonza), 0.5 mM L-Glutammine (Gibco), B27 (Gibco) supplement and 30 ng ml⁻¹ GDNF (rHu GDNF, AppliChem GmbH). In order to avoid potential interactions among multiple neurotrophic factors, as well as to keep the experimental design simple, the cultures were supplemented by a single neurotrophic factor. We chose GDNF as the preferred neurotrophic factor as it was found to strongly promote neurite outgrowth in P5 rat pup spiral ganglion explants (Euteneuer et al 2013) and to be the most effective among other neurotrophic factors in adult SGN (Boström et al 2010). Cells were counted and seeded in 100 μl volume at a density of 20*10³ cells/sample (≈300 cells mm⁻²). Cells were allowed to settle for 10–15' in the incubator, then the rest of the medium was added carefully in the well. Half of the medium was changed every 2–3 days.

SGN were cultured during 1 and 7 DIV and fixed with 4% paraformaldehyde for 30 min. For immunocytochemical analysis, samples were washed three times with PBS, permeabilized with 0.1% Triton-X (Calbiochem) for 5 min and blocked with PBS containing 1% goat serum (GS, Biochem) for 90 min. The cells were then incubated at room temperature for 90 min in PBS 1% GS with mouse monoclonal anti-βIII tubuline (1:200, Millipore) to stain neurons (named Tuj1+ cells afterwards), and rabbit polyclonal anti-S100 (1:200, Sigma-Aldrich) to stain glial cells (named S100+ cells). Cells were then washed three times with PBS, incubated for 90 min in PBS 1% GS at room temperature with Alexa 488 goat anti-mouse (1:500, Molecular Probes), Texas Red goat anti-rabbit (1:500, Santa Cruz) secondary antibodies and 5 μg ml⁻¹ DAPI (Sigma-Aldrich). The samples were finally washed three more times with PBS and prepared for imaging with Imumount (Thermo Scientific). Imaging was carried out under an Olympus DP71 fluorescent microscope equipped with an Olympus U-RFL-T burner.

We first determined SGN presence on MPS depending on pillar dimension. Surface plots of SGN distribution on MPSs as function of pillar width and spacing were generated with SigmaPlot software with the aim of considering the cumulative percentages. Cumulative percentages were obtained from the cumulative frequency distribution (sum of SGN in each pillar area among different samples), divided by the total number of neurons and multiplied by 100. Subsequently, we grouped the pillar features according to spacing (S) into 4 bins (0.6–1.0 μm; 1.2–2.4 μm; 2.4–5.0 μm and 7.0–15.0 μm). The percentage of SGN in each bin was calculated for each experiment and averaged. Then we assessed the SGN presence on MPSs and compared it to the control glass coverslips. To quantify the total neuronal cell number, Tuj1+ cells were counted under the microscope on each sample (MPSs or glass coverslips) and averaged. Values were then normalized by sample surface area, which was 64 mm² for MPS and 132 mm² for control glass.

To describe the effect of the dimension of MPS on SGN morphology, morphometric analysis was carried out with the use of ImageJ software (NIH). A minimum of eight pictures for each sample was considered for analysis. We focused on three parameters: longest neurite length, number of sprouting, and neurite alignment.

Among the growing processes of an SGN, we focused on analyzing the longest neurite of each cell as morphologically most resembling an axon. However, we maintained this nomenclature in the text to avoid confusion. The longest neurite length was described in function of time (1DIV, 7DIV) and pillar spacing on MPS. Glass coverslips were considered as the control. The measurements of the longest neurite length were obtained by overlaying a segmented line on images of Tuj1+ SGN and quantifyied with the measuring tool on the calibrated images. The values were then represented in a Box-and-Whisker plot, where bars represent the minimum and the maximum values of the data distribution, the boxes are the interquartile range and the central lines indicate the median values.

The number of sproutings and the relation to neuronal morphology was determined on MPSs and control samples as a function of time (1DIV, 7DIV). To determine whether neurons were neurite-free, mono-, bi- or multi-polar, the number of neurites was scored manually in pictures for each Tuj1+ cell, following a classification described previously (Vieira et al 2007, Whitlon et al 2006). The values from the pictures of a single sample were summed together and the percentage calculated with respect to the total number of neurons. Percentages from different samples from the same conditions were averaged and plotted.

SGN alignment on MPSs as a function of pillar spacing was determined using the ImageJ Fast Fourier Transform (FFT)-Oval Profile plugin as described previously. Briefly, gray-scale images of Tuj1+ neurons were aligned to the horizontal bottom line and subsequently processed with the FFT function of the ImageJ software. The intensity and distribution of the pixels of the resulting image is related to the directional content. The 'Oval profile' plug-in performed the sum of the pixel intensities along a straight line from the centre to the edge of the FFT image (at angle θ, 0° ≤ θ ≤ 180°), which quantifies the relative contribution of objects oriented in that direction. Plots of the sum of pixel intensities along θ represent the directionality of the original image. Constant pixel intensities, independent of direction, represent random oriented images, while peaks along one specific direction represent an image preferentially aligned in that direction (Alexander et al 2006, Mattotti et al 2012). The values from the analyses of the orientation of each neuron were averaged between neurons in the same bin and plotted in the figure.

Finally, we assessed the potential interaction of Tuj1+ with glial S100+ cells as manifested by being directly overlapped, both on MPSs and the control samples, as a function of time (1DIV, 7DIV). We defined two categories: SGN having the longest neurite in touch with S100+ glial cells and SGN not having an axon in touch with S100+ glial cells. The proximity of Tuj1+ cells with S100+ cells suggests an interaction between the two cell types. The S100+/Tuj1+ ratio and the number of axons of Tuj1+ cells in touch with S100+ glial cells were counted manually in fluorescent, double stained images.

For SEM imaging, samples were fixed in 4% paraformaldehyde for 30 min, washed three times, stained with osmium tetroxide for at least 2 h. Subsequently they were immersed twice over the course of 10 min in increasing concentrations of ethanol (steps of 10–30–50–70–90–100% EtOH) and brought to critical point dehydration. Then they were observed using a Philips XL30 scanning electron microscope.

For statistical analysis, data were transferred to spreadsheets and analyzed with the Statgraphics software (StatPoint Technologies Inc. Warrenton, US). Each experiment was repeated at least three times (see figure captions for details). Values in the plots were reported as averages with standard deviation bars, with an exception for the neurite length analysis (figure 4). To test the variance between multiple samples, one-way ANOVA was performed, followed by a post-hoc LSD test. A p value less than 0.05 was considered significant.

We assessed growth, behaviour and differentiation of spiral ganglion neurons (SGN) cultured on silicon micro-patterned substrates (MPSs). These silicon surfaces consisted of 150 different areas of 3 μm high micro-pillars with different widths and spacings. The shape of the pillars and their organization was designed in a hexagonal manner. The silicon substrate layout is shown in figure 1(A), whereas figure 1(B) shows scanning electron microscopy images of three areas with characteristic pillar designs.

We first analyzed the influence of pillar dimension on SGN presence. We immunostained spiral ganglion neurons with Tuj1+ after seven days in culture, and described their position on the substrate by fluorescent microscopy. In order to do that, we calculated the cumulative percentages among 12 different samples from three independent experiments (N = 3), for a total of n = 906 neurons (an average of 72.5 ± 20 neurons/MPS) and plotted the results in figure 1(C). The surface plot of SGN on MPS (figure 1(C)) revealed that SGN did not display a uniform distribution to the surface, but rather prefers certain specific pillar geometries. Subsequently, we grouped the pillar features according to spacing (S) into bins. We chose to analyze pillar spacing among pillar width since previously it has been found that pillar spacing was more influential in changing neuronal behaviour than pillar width (Micholt et al 2013). The percentage of SGN in the bins was calculated for each experiment and averaged (N = 3). The results revealed that the cell presence was enhanced on pillars with spacing between 1.2 μm and 2.4 μm. The percentage of SGN grown on pillar spacings between 1.2 μm and 2.4 μm was doubled (42 ± 3%) compared to SGN found on all the other pillar dimensions (22 ± 3%, on S = 0.6–1.0 μm; 17 ± 3% on S = 3.5–5 μm and 19 ± 3% on S = 7–15 μm) as shown in figure 1(D).

Next, we assessed the overall growth characteristics of SGN on silicon micro-pillars compared to the conventional culturing substrates, i.e. borosilicate glass coverslips. We considered Tuj1+ SGN grown on MPS and we compared cell numbers with the control glass coverslips. The pictures in figure 2(A) show an overview of the SGN cell culture. From a visual examination, it can be observed that neurons and non-neural cells are present both in control and MPS surfaces, and that neurons grow neurites in both conditions. The graph in figure 2(B) shows that there were no quantitatively significant differences between the total numbers of Tuj1+ grown on MPS (1.2 ± 0.3 cells mm⁻² equivalent to 72.5 ± 20 cells/sample, where sample area is 64 mm²; N = 3) and control glass surfaces (1.4 ± 0.4 cells mm⁻², equivalent to 181.2 ± 72 cells/sample, where sample area is 132 mm²; N = 3). This indicates that structured silicon does not change the overall growth of SGN in culture.

Zoom In Zoom Out Reset image size

In order to describe how pillar dimensions (width and spacing) influenced the growth directionality of neurites, SGN were stained with Tuj1+ and visualized (figure 3(A)). The orientation of the neurites on micro-pillar surfaces with different dimensions was analyzed using ImageJ software. FFT-Oval Profile analysis showed that the hexagonal pattern of the pillars had a strong influence on the directionality of neurite outgrowth for pillar spacings between 0.6 and 2.4 μm (figure 3(B)). On these specific micro-patterns, neurites preferentially oriented along three directional axes, spaced by 60° angle intervals (30°; 90°; 150°). The interaction of aligned neurites with pillars can be observed in SEM images (figure 3(C)). On the other hand, the topographic guidance was highly reduced for spacings of 3.5 to 5 μm, and completely lost for spacings above 7 μm, suggesting that there is a threshold value of inter-pillar spacing to elicit neurite contact guidance. We did not observe significant effects of the pillar width on neuronal outgrowth (data not shown), as was previously reported (Micholt et al 2013).

Zoom In Zoom Out Reset image size

Next, we investigated whether pillar geometry had an effect on SGN neurite length. For this analysis the longest growth process of every SGN was considered, and their length was measured by submitting confocal images into ImageJ. After 1 DIV, pillars with both inter-spacings of 0.6–1.0 μm and 1.2–2.4 μm promoted SGN growth more significantly, as neurite length was significantly higher compared to the control group (figure 4). Among all pillar geometries, spacings of 1.2–2.4 μm gave rise to the longest neuritis; significantly longer than the control group. As expected, at 7 DIV in cultures, all neurites were longer than on the first day, but the increase was larger for the micro-structured surfaces, suggesting faster growth on these surfaces.

Zoom In Zoom Out Reset image size

We assessed the morphology of the SGN cells on both MPSs and control surfaces. The morphological shape of the SGN was determined by the number of neurites sprouting from the SGN soma, yielding neurite-free, monopolar, bipolar and multipolar morphologies (figure 5(A)). No significant differences were present between control and MPS surfaces concerning the percentage of neurite-free neurons (20 ± 5% and 30 ± 10% respectively). After 1 DIV, the MPS induced 3-fold more SGN with a monopolar morphology (46 ± 17%) with respect to the control condition (14 ± 8%) and 8-fold less SGN with a multipolar morphology (2 ± 4%) with respect to the control condition (17 ± 4%) (figure 5(B)). After 7 DIV, the neurite-free neurons on MPS decreased to 14 ± 6%. Monopolar neurons were still more abundant on MPS than in the control condition (39 ± 11% and 19 ± 10% respectively) while the number of multipolar SGN was significantly increased on both the MPS and control surfaces. However, the number of SGN with multipolar morphology was still lower on the MPS (14 ± 6%), almost half of that on the control surfaces (29 ± 12%) (figure 5(C)). In comparison with the directed growth of SGN on MPS as described above, these results revealed that MPS encouraged neurons to have less of a multipolar morphology.

Zoom In Zoom Out Reset image size

Previous observations of SGN in in vitro cell cultures have shown that these neurons often grow their axons in contact with S100+ cells (Bostrom et al 2007, Rak et al 2011). We observed that S100+ cells were the most abundant non-neural cell type in culture (60 ± 18%) and their number was stable over time both on control surfaces and MPS. S100+ cells showed heterogeneous morphologies, suggesting the presence of different types of glial cells, most probably Schwann cells and satellite cells (Whitlon et al 2009, Liu et al 2012). The ratio of glial cells/neurons was 12.5 ± 6.0, and there were no significant differences between the control surfaces and MPSs (Total number of SGN analyzed was n = 1205 from 16 images of MPS and seven images of control conditions). Further, we analyzed the interaction of the longest neurites from Tuj1+ cells with S100+ cells after 1 DIV and after 7 DIV on both surfaces (figure 6(A)).

Zoom In Zoom Out Reset image size

As can be seen in figure 6(B), the longest neurites on an MPS were encouraged to grow more in contact with micro-pillars than to seek interaction with glial cells that were present in their surroundings. After 1 DIV, on the control surfaces 46 ± 15% of the longest neurites sprouting from SGN were in contact with S100+ cells, while on the MPS the percentage of longest neurites touching glial cells was almost 4 times lower (9 ± 2%). After 7 DIV, in both conditions the number of longest neurites touching/contacting S100+ cells increased significantly. However, on MPS the number of longest neurites in contact with glial cells was still more than halved (33 ± 15%) compared to control surfaces (78 ± 25%). It seems that MPS present SGN with a strongly organized environment, which makes them reach out less to other supporting structures in their environment.