“Microtopographical Patterns Promote Different Responses in Fibroblasts and Schwann Cells: A Possible Feature for Neural Implants”

Abstract

The chronic reliability of bioelectronic neural interfaces has been challenged by foreign body reactions (FBRs) resulting in fibrotic encapsulation and poor integration with neural tissue. Engineered microtopographies (e.g., surface patterning) could alleviate these challenges by manipulating cellular responses to the implanted device. Parallel microchannels have been shown to modulate neuronal cell alignment and axonal growth, and Sharklet™ microtopographies of targeted feature sizes can modulate bio-adhesion of an array of bacteria, marine organisms, and epithelial cells due to their unique geometry. We hypothesized that a Sharklet™ micropattern could be identified that inhibited fibroblasts partially responsible for FBR, while promoting Schwann cell proliferation and alignment. In vitro cell assays were used to screen the effect of Sharklet™ and channel micropatterns of varying dimensions from 2 – 20 μm on fibroblast and Schwann cell metrics (e.g., morphology/alignment, nuclei count, metabolic activity), and a hierarchical ANOVA was used to compare treatments. In general, Schwann cells were found to be more metabolically active and aligned than fibroblasts when compared between the same pattern. 20 μm wide channels spaced 2 μm apart were found to promote Schwann cell attachment and alignment while simultaneously inhibiting fibroblasts and warrant further in vivo study on neural interface devices. No statistically significant trends or correlations between cellular responses and geometrical parameters were identified because mammalian cells can change their morphology dependent on their environment in a manner dissimilar to bacteria. Our results showed although surface patterning is a strong physical tool for modulating cell behaviour, responses to micropatterns are highly dependent on the cell type.

1. Introduction

There is a growing need for bioelectronic neural interface technology to enable effective and chronic communication between a patient’s nerve fibers (i.e., axons) and computer platforms to provide a variety of therapeutic benefits (e.g., restoration of vision, improved prosthetic control). An ideal neural interface can restore both sensory and motor function through electrical stimulation of the nerves (input) and recording bioelectrical signals (output) from axonal impulses. Recorded bioelectrical signals are passed to computerized signal processors that can translate and direct desired motor function (e.g., movement in a robotic prothesis) or the reverse process where nerves are electrically stimulated to provide feedback to the patient for modulating neural activity and/or restoring sensory function.¹ There are several examples of neural interfaces that are successfully commercialized including cochlear implants to restore function of the damaged inner ear,² deep brain stimulators to manage Parkinson’s disease,³ and retinal prostheses to restore vision.⁴

Next generation neural interfaces aim to improve therapeutic outcomes by increasing device channel counts to interface with larger numbers of axons, using minimally invasive and less damaging implantation protocols, and improving chronic reliability so devices can maintain functionality for the patient’s lifetime.^5,6 However, a major challenge for many biomedical devices is the foreign body reaction (FBR) that can result in loss of function or device failure, and for neural interfaces these problems can include deterioration of recordable neural signals, loss of stimulation effectiveness, and increased background noise over the duration of the implant.⁷ The FBR can include non-specific protein absorption (i.e., biofouling), fibrotic encapsulation, chronic inflammation, and neuronal cell death near the implants.⁸ The FBR is thought to arise from a variety of interrelated factors including mechanical mismatch of the implant-tissue interface, implant biomaterial bulk and/or surface incompatibility, damage during implantation, and micro-motion of the implanted device.⁹ The FBR challenge must be overcome to allow for next-generation neural interface technologies to achieve their promised therapeutic potential. Here we focus on the FBR challenge for peripheral nervous system implants.

Acute inflammatory responses and cell death in the surrounding area of a peripheral nerve implant, initiated by the surgical implantation procedure, can impede neural regeneration and intimate nerve-implant integration.^10,11 When injury occurs because of the surgical implantation, immune cells (e.g., macrophages, and monocytes) are recruited to the injury site and promote fibroblast migration and activation that typically results in the formation of a fibrotic capsule, Figure 1 A. The fibrotic capsule will reduce signal-to-noise ratio of recordable action potentials and interrupt the neural interface-nerve communication. The severity of the fibrotic capsule is highly dependent on the material and micro/nano-topographical properties of the implanted device.^7,12

Figure 1. Foreign Body Reaction (FBR) in Response to Peripheral Nerve Interface (PNI) Implantation.

A) Surgical implantation of the PNI activates the immune system. Monocytes differentiate to microphages and recruit fibroblast. Finally, fibroblasts migrate close to the implant and encapsulate the device in fibrotic tissue that reduces the axon-electrode communication and the signal-to-noise ratio. B) One strategy to mitigate FBR is modification of the microenvironment (e.g., device surface) to inhibit fibroblast attachment and activation. Concurrently, encouraging Schwann cell activation and alignment can enhance the regenerative microenvironment resulting in better axonal out-growth and integration with the PNI. The combination of these two approaches were tested in this study.

During the natural regeneration process, Schwann cells (SCs) become activated and degrade myelin debris remaining from the nerve damage followed by alignment along the proximal stump of the nerve trunk to guide axonal extesion.¹³ In this process, in addition to biochemical cues such as neurotrophic factors, the topographical features of the surrounding extracellular matrix (ECM) play an important role in axonal guidance, regeneration, and outgrowth. These topographical features appear in different size and shape. While the diameter of axons in human varies between 0.2 to 15 μm, the topographical features in peripheral nerve ECM has a broad range between nanoscale (e.g., collagen fibrils diameter ~ 20 nm) to microscale (e.g., hierarchal fascicle structures diameter ~ 100 to 1000 μm).

Several strategies have been studied to improve natural integration of the neural tissue and to mitigate FBRs including refining device design and engineering parameters, using natural biomaterials, and mimicking the chemical, physical, and architectural microenvironment of the neural tissue.^14,15 Topographical surface features can be used to control the local microenvironment and responses of microorganisms and have been successfully used on medical devices and biomaterials (see comprehensive reviews by Curtis and Wilkinson¹⁶ and Hoffman-Kim et al.¹⁷). Micro- and/or nano-topographical features of varying dimensions, studied in biomedical technologies, include but are not limited to channels (grooves), pillars, pits (wells), aligned fibers, and biomimetic (i.e., nature-inspired) patterns.^17-19

The use of micro-topographical patterns have been reported to promote the regeneration of peripheral nerves.^20-23 During axonal regeneration, different cell types in the nerve undergo dynamic changes that are controlled by external cues. Like neurons, Schwann cells are of paramount importance for healthy nerve regeneration, produce repair-supportive trophic factors and cytokines,²⁴ and provide external natural topographical cues to guide axonal outgrowth.²⁵ Researchers showed that when Schwann cells are promoted to align by means of external topographical features like microchannels, they undergo differentiation and provide an oriented platform to set the stage for axonal guidance.^17,26 Moreover, it has been shown that microchannels modulate gene expression, neurite branching, and outgrowth of the neural cells^17,27,28 and can induce fibroblast trans-differentiation to neurons by upregulating the expression of actin cytoskeleton and non-muscle myosin-II.²⁹ Based on strong evidence that shows micropatterns are able to provide the proper physiological cues to control cellular responses, we hypothesized that simultaneously promoting healthy neural regeneration while mitigating fibrosis should be possible by means of micro-topographical features. However, it is not clear if the changes in the cellular responses are dominated by the cell type or the geometry and shape of the micro-topography (i.e., if a single microtopography can promote different cellular responses, in different cell types.).

We designed an in vitro study to address the above question and identify one or more micro-topographical size(s) and/or geometries that can inhibit fibroblast attachment, as a critical step to control fibrosis, and concurrently promote Schwann cells attachment and alignment to mimic the regenerative microenvironment, Figure 1 B. Parallel microchannels and Sharklet™ engineered microtopography geometries were used in this study as likely candidates to effectively control the cellular microenvironments, as previously hypothesized. Microchannels are commonly used to study cellular alignment as discussed earlier. Sharklet™ is a microtopography system inspired by the placoid scales of sharks and initially gained prevalence for its ability to control the biofouling of a wide variety of marine organisms.^30,31 Sharklet™ has also been shown effective at modulating the bio-adhesion of gram-positive bacteria, lens epithelial cells and keratinocytes.^32-34 For the first time, we explored the ability of Sharklet™ micropatterns to control cellular responses of Schwann cells and fibroblasts, for potential application in neural interfaces.

Polyimide was chosen as the substrate for in vitro analysis because it is one of the primary polymeric materials used for peripheral nerve interfaces in both animal and human trials.^35-37 Polyimide has an elastic modulus of ~3 GPa that is two orders of magnitude lower than other common neural implants materials (e.g., ~130 GPa silicon). Higher modulus devices with higher bending stiffness have been shown to exhibit increased FBRs.^38-40 Polyimide is not as ubiquitous as polydimethyl siloxane elastomers for fabrication of engineered micro-topographies, however, fabrication techniques are available and were used previously.⁴¹ In addition, the fabrication methods used in this study are fully compatible with standard microfabrication protocols and can easily be applied to polyimide nerve interface device fabrication.⁴² Microtopography size and spacings for both channels and Sharklet™ were chosen to range from 2 – 20 μm, which is comparable to the size range of the Schwann cells and fibroblasts. Pattern depth was fixed at 3–3.5 μm because it is the maximum depth that is compatible with most 10 μm thick polyimide-based peripheral nerve devices while being deep enough to impact cellular morphology. 18 different combinations of micropatterns ranging in size from 2-20 μm with geometries in Sharklet™ (SK) and channels (CH) were evaluated. Schwann cells and fibroblasts were cultured on these surfaces, and morphological changes in cytoskeleton and nuclei after three days in culture and the temporal metabolic activity of the cells over one week were measured and analyzed. Clear differences were observed between fibroblasts and Schwann cells in response to the micro-topographical features, and a specific micropattern was identified that promotes Schwann cells and simultaneously inhibits fibroblasts.

2. Materials and Methods

2.1. Micropattern Fabrication

Photomask Fabrication:

Pattern photomasks were designed using Layout Editor software (Juspertor GmbH, Unterhaching, GER) and exposed onto chrome-quartz photomasks (Nano-Film Microcircuit Technology, Westlake Village, CA, USA) using a DWL 66FS laser writer (Heidelberg Instruments GmbH, Heidelberg, GER). Exposed masks were developed in AZ400K (MicroChemicals GmbH, Ulm, GER) and etched in chrome etchant 1020 (Transene Company, Inc., Danvers, MA, USA). Two different masks were fabricated. The microtopography mask contained 19 individual 15-mm diameter patterned regions with the full matrix of engineered microtopographies ranging in size from 2 to 20 μm. The punchout mask contained corresponding non-patterned disks to protect the patterned regions produced from the first mask during polyimide etching.

Patterned Polyimide Fabrication and Characterization:

Engineered microtopographies were molded in polyimide. All fabrication work was performed in a class 100-1000 cleanroom at the University of Florida (Gainesville, FL, USA), Figure 2. 100 mm diameter, P type Si wafers (University Wafer, Inc, Boston, MA, USA) were patterned by deep-reactive ion-etching (DRIE) to make the molding surface. First, Si wafers were cleaned in a piranha solution composed of 3:1 (v/v) H₂SO₄ (<99%): H₂O₂ (33%) to remove residual organic contaminants. Next, wafers were coated with hexamethyldisilazane from vapor and cured at 112 °C to drive the silanization reaction to completion. Wafers were coated with a 0.8 μm layer of AZ1512 positive photoresist (MicroChemicals GmbH, Ulm, GER) as an etch-stop, soft-baked at 112 °C for 2 minutes, and exposed through the microtopography photomask at 88 mJ on a MA-6 mask aligner (SUSS MicroTec GmbH, Garching, GER). Exposed wafers were developed in AZ300 MIF (MicroChemicals GmbH, Ulm, GER) and washed three times in deionized water (DIW, produced in house, 18.2 MΩ). Patterned wafers were etched to a depth of 3.0 – 3.5 μm using the Bosch process in a Plasma Pro 100 DRIE (Oxford Instruments, Abingdon, UK). Wafers were cleaned of residual photoresist using PRS3000 (Avantor Performance Materials, LLC., Radnor, PA, USA) and organic residue using piranha solution, as before. Etch depth was confirmed by a Contour GT-I optical profilometer (Bruker Corporation, Billerica, MA, USA). U-Varnish S (UBE Industries, Ltd., Tokyo, JP) 3,4,3′,4′-biphenyltetracarboxylic dianhydride-p-phenylene diamine (BPDA-PPD) type polyimide precursor in N-methyl-2-pyrrolidone was spin-coated onto patterned Si wafers and cured at 450 °C in an N₂ atmosphere hotplate (Wenesco, Inc., Addison, IL, USA) to a final thickness of 10 μm confirmed using a F40 spectral reflectometer (Filmetrics, San Diego, CA, USA). Polyimide cured wafers were coated with a 25 μm layer of AZ9260 positive photoresist (MicroChemicals GmbH, Ulm, GER) as an etch-stop, soft-baked for 3 min at 112 °C, and aligned with/ exposed through the punch-out photomask at 2150 mJ on a MA-6 mask aligner. Exposed wafers were developed in 4:1 DIW: AZ400K and rinsed 3 times with DIW. Exposed PI was etched away in a Unaxis 790 (PLASMAtech, Inc., Erlanger, KY, USA) reactive-ion etch (RIE) O₂ plasma. Residual photoresist was stripped as before in PRS3000. Individual 15 mm diameter, 10 μm thick patterned polyimide disks were manually liberated from the DRIE Si wafer using forceps. Non-patterned, smooth polyimide controls were fabricated using non-patterned Si wafer substrates and the same procedures described above. Microtopography geometry and feature sizes were verified using a Nova NanoSEM 430 (FEI, Hillsboro, OR, USA) at 10 kV.

Figure 2. Micropatterns Fabrication.

A) Tilted scanning electron microscopy image of a deep-reactive ion-etching (DRIE) patterned Si wafer surface with the inverse Sharklet™ (SK2x2) pattern. B) Optical image of a patterned Si wafer containing 19 individual 15 mm patterned polyimide disks that have been revealed by reactive-ion etch (RIE) removal of the surrounding polyimide film. C) Top-down SEM image showing a molded polyimide sample with the Sharklet™ engineered microtopography (SK2x2). Scale bars = 15 μm. Cartoon schematics above represent the side-view of each step described.

2.2. In Vitro Tests

Cell Culture:

Adult rat fibroblasts and adult rat Schwann cells were purchased from ScienCell Research Laboratories (Carlsbad, CA, USA). Cells were passaged to reach the desired number in a humidified incubator at 37 °C and 8% CO₂ (Thermo Fisher Scientific, Waltham, MA, USA). fibroblasts were cultured in Dulbecco’s Modified Eagle Medium (DMEM; Corning, Corning, NY, USA) supplemented with 10% fetal bovine serum (FBS; Atlanta Biologicals, Flowery Branch, GA, USA) and 1% penicillin/streptomycin (P/S; Gibco, Waltham, MA, USA). Schwann cells were cultured in DMEM supplemented with 10% FBS, 1% penicillin/streptomycin, 0.14% bovine pituitary extract (BPE; Gibco, Waltham, MA, USA), 0.1% forskolin (FSK; Sigma-Aldrich, St. Louis, MO, USA), and 0.1% fibroblast growth factor (FGF; Gibco, Waltham, MA, USA). Polyimide disks were adhered to 15 mm diameter, 0.25 mm thick circular borosilicate glass coverslips (Thermo Fisher Scientific, Waltham, MA, USA) using non-toxic RTV 734 silicone adhesive (Dow Corning, Midland, MI, USA) and placed in 24-well cell culture plates for in vitro testing. For sterilization, polyimide disks were exposed to 70% ethanol (1 ml in each well), and UV for 1 hour inside a laminar flow hood. Samples were rinsed twice with sterile phosphate buffered saline (PBS; Gibco, Waltham, MA, USA) and soaked in the culture media, described above for each cell type, and incubated for 2 hours prior to cell seeding. Cells (Schwann cells and fibroblasts) at passage 4 were seeded at a density of 13,158 cells/cm² (25,000 cell/well in 24-well plate) onto the polyimide patterned disks within 500 μl of relevant culture medium. Culture medium was changed the day after seeding and every three days thereafter.

Metabolic Activity Measurement:

An alamarBlue™ assay (Invitrogen, Waltham, MA, USA) was used to assess metabolic activity of the cells (n=3) at days 1, 3, and 7 of cell culture, according to the manufacturer’s instructions. Briefly, cell culture media were removed and replaced with relevant media for each cell type containing 10% alamarBlue™ reagent solution; samples were incubated for 2 hours in a humidified incubator at 37 °C and 8% CO₂. 100 μl of the incubated solution was removed from each sample and placed in a fresh 96-well plate to measure the fluorescence intensity at 530 nm (excitation), and 590 nm (emission) using a BioTek Synergy HT Microplate Reader (Winooski, VT, USA). The fluorescence intensity, which correlates with metabolic activity of the cells on the patterns, were normalized against the metabolic activity of those cells on smooth surface controls, for each cell type.

Visualization of Cell Cytoskeleton and Nuclei:

At day 3, culture media were removed and samples were rinsed two times with PBS and fixed in 4% paraformaldehyde (PFA). Cells were permeabilized with 0.1% Triton X100 (Sigma-Aldrich, St. Louis, MO, USA). To visualize the cell cytoskeleton, fixed cells were incubated with phalloidin (Invitrogen, Waltham, MA, USA), tagged with alexa fluor 660, diluted 1:1000, in PBS and 1% BSA, at room temperature for 20 minutes. Cell nuclei were stained with DAPI (Invitrogen, Waltham, MA, USA) at room temperature for 10 minutes. Samples were mounted onto microscopy glass slides and imaged with Zeiss Axio Imager.Z2.

Image Analysis:

Images were analyzed by CellProfiler™, an open-source image analysis software. To analyze the alignment, the Flip and Rotate module was used to line up the pattern horizontally with the x-axis of the image. This was done to calibrate the angle of patterns for accurate measurement of cells alignment with respect to patterns. To count the cell number, cell nuclei were identified using the Identify Primary Objects module. The features of the module were iteratively optimized and various quantitative features of each identified nuclei were measured using the Measure Object Size Shape module. The extracted data from the analysis were exported to an excel file to develop rose diagrams.

Scanning Electron Microscopy (SEM):

One set of samples at day 3 of culture were prepared for SEM imaging to determine how the morphology of the cells attached to microtopographies. A microwave-assisted procedure was used to prepare SEM samples for high quality/contrast imaging.^43,44 Briefly, samples were treated in Trump’s universal fixative solution that contains sodium cacodylate, formalin, and glutaraldehyde (VWR, Visalia, Ca, USA), and placed in a microwave at 250 W for 45 s, and then incubated under ambient conditions for 1 min. Next, samples were treated with 1x cacodylate solution in microwave at 250 W for 45 s, and then incubated overnight at 4 °C in vacuum. Samples were then treated with 2% osmium tetroxide in PBS for 20 minutes and then 45 s in a microwave at 250 W. Finally, samples were fully dehydrated by placing in ascending concentrations of ethanol (50%, 60%, 70%, 80%, 90%, and absolute ethanol), 10 minutes in each solution, and finished in a critical point dryer (CPD, Tousimis, USA) (Please see the full protocol of using CPD⁴⁵). Fixed and dehydrated samples were plasma sputter coated with 10 nm of Au/Pd and analyzed with a Nova NanoSEM 430 (FEI, Hillsboro, OR, USA) at 10 kV.

2.3. Statistical Analysis

Principal Components Analysis:

Principal components analysis⁴⁶ was used to generate a metabolic activity index by determining the principal component of variation across the multiple measurements of metabolic activity at day 1, 3, and 7, and the differences between day 1 and day 3, as well as day 3 and day 7. This principal-components analysis was conducted for each cell type. The percent of the total variance contributed by each principal component was calculated and the significance loadings at an α=0.05 were determined using a bootstrap procedure.⁴⁷ The first principal component from each principal component analysis was used as an index of metabolic activity.

Hierarchical ANOVA:

A hierarchical analysis of variance (also known as nested ANOVA) was used to compare the micropattern treatments across measurements of metabolic activity, nuclei count, and alignment while accounting for the measurement error resulting from multiple measurements. The model assumes the group mean for treatment i and measurement x is a latent state variable, μ_i,x, and observations (1…n) are repeated measures, y_i,x,n, of the group mean with some error, σ_obs,x (Equation 1).

$$y_{i,x,n}\sim \mathrm{N}({\mu }_{ix},{\sigma }_{obs,x})$$ (Eq. 1)

The group means for a given measurement were assumed to be random effects drawn from a normal hyper-distribution of possible group means (Equation 2).

$${\mu }_{i,x}\sim \mathrm{N}(0,{\sigma }_{hyper,x})$$ (Eq. 2)

The magnitude of the standard deviation of the hyper-distribution of group means, σ_hyper,x, describes the overall variation across treatment groups (higher values indicate less similar group means). Before conducting the hierarchical ANOVA, total nuclei count and alignment measurements were Z-standardized $\left(\frac{y-\overline{y}}{{\sigma }_y}\right)$ relative to cell type by micropattern type combination (e.g. Schwann cells on Sharklet™ were Z-standardized together). The metabolic activity index did not need to be Z-standardized, as it was the first principal component of the principal-components analysis. We assumed that σ_obs and σ_hyper both had half-normal priors of N (0, 0.5). The hierarchical ANOVA (Just Another Gibb Sampler) was implemented in JAGS in program R using the jagsUI package.⁴⁸ A separate hierarchical model was estimated for each cell type and pattern type. For each model, a prior probability distribution was initialized using a random draw from the prior distributions of each parameter and had four Markov Chain Monte Carlo simulations, each with 26,000 total posterior samples. Convergence was assessed when the potential scale reduction factor of each parameter was less than 1.05.⁴⁹ The median and 90% credible interval were calculated for each treatment group’s mean values for metabolic activity, nuclei count, and alignment for each cell type by pattern type. Group means for nuclei count and alignment were back-transformed from the Z-standardization for comparison. Posterior distributions of σ_obs and σ_hyper, were checked for shrinkage from their prior distributions. The hierarchical ANOVA Markov Chain Monte Carlo simulation results of the treatment group median for metabolic activity and total number of nuclei were summed to generate a relative cell type promotion index.

3. Results

3.1. Micropatterns

Optical and SEM analysis confirmed the microtopography geometry of the Sharklet™ and channels, with dimensions ranging from 2 – 20 μm on the polyimide test surfaces, Figure 3. Microtopographies were identified by values of “a” and “b”, where “a” indicates the feature width and “b” indicates the space between the features. The final height of all micropatterns was fixed at ~3 to 3.5 μm. Exact parity between pattern depths was not possible due to minor etch-rate variations during the DRIE Bosch process.

Figure 3. SEM Images of Micropatterns.

A) Both the Sharklet™ (SK) and parallel channel (CH) geometries have features ‘a’ μm wide spaced ‘b’ μm apart; however, the SK geometry is distinguished with diagonal cuts that produce 4 unique repeating features that are 8a – 2a μm long resulting in a diamond pattern. B) Side view of a representative SK pattern showing the depth of ~ 3 μm. C) SEM images of the 3x3 array of SK engineered microtopographies with feature widths ‘a’ ranging from 2 – 20 μm and feature spacings ‘b’ also ranging from 2 – 20 μm. D) The analogous 3x3 array of channels.

3.2. Interaction of the Cells with Micropatterns

Cell Metabolic Activity and Adhesion:

Cell metabolic activity on each pattern was measured at days 1, 3, and 7 using an alamarBlue™ assay, and the results were normalized against the smooth surface control. Cellular density at day 3 was calculated by nuclei count using CellProfiler™ on DAPI stained samples. Unprocessed data for the metabolic activity is presented in the supplementary section, Figure Supp. 1. Schwann cells on CH and fibroblasts on SK had significant differences in metabolic activity at days 1 and 3 as well as days 3 and 7. The metabolic activity of the Schwann cells on Sharklet™ at day 1 was not significantly different from the smooth control; however, the metabolic activity was significantly higher on the micropatterns at day 7. From the hierarchical ANOVAs, pair-wise samples of the treatment group mean for metabolic activity and nuclei count were summed to generate a median generic activity index for each micropattern and used to assess the overall cell-micropattern interaction, Figure 4. Positive generic activity indices indicate higher metabolic activity and higher relative nuclei counts while negative generic activity indices indicate lower metabolic activity and lower relative nuclei counts. From the generic activity index, CH 20x10 noticeably inhibits both Schwann cells and fibroblasts, as do, to a lesser degree, SK 2x10 and SK 2x20. Conversely, CH 2x2, CH 20x20, CH 10x10, SK 20x10, and SK 10x2 promotes cellular activity and attachment of both cell types. CH 20x2, SK 20x2, and CH 10x2 increasingly inhibit fibroblasts with only a minor impact on Schwann cells (promotion or inhibition) generic activity. Metabolic activity is relatively unchanged on SK 10x10, CH 10x20, and CH 2x10.

Figure 4. Generic Promotion Index.

A) The median nuclei count as a function of the median metabolic activity index for each micropattern for Schwann cells, B) and Fibroblasts. Arrows indicate the median alignment of each cell type on each pattern. Results show an overall trend of the size (indicated by the point shape) and spacing (indicated by the color) of the micropatterns (solid points indicate channel (CH) patterns and open points indicate Sharklet™ (SK) patterns). C) The median metabolic activity and nuclei count effect sizes predicted from the hierarchical ANOVA were summed to create a generic promotion index representing the promotion and inhibition of the two cell types. Micropatterns are denoted by text with the first number indicating the size and the second number indicating the spacing. (Region 1) shows the micropatterns that inhibit both Schwann cells and fibroblasts. (Region 2) shows the micropatterns that inhibit the Schwann cells but promote fibroblasts. (Region3) shows the micropatterns that promote both Schwann cells and fibroblasts. (Region 4) shows the micropatterns that promote the Schwann cells but inhibit fibroblasts.

Cell Alignment:

Cellular nuclei alignment is shown in Figure 5 A and B. With the imaging techniques employed, it was impossible to precisely determine the shape and area of the cell body. Therefore, we used the alignment of nuclei as an analogue for overall cellular alignment by assuming that the orientation of the cell nucleus reflects the orientation of the cell body on micropatterned structures larger than 400 nm.^50,51 Figure 5 C highlights selective results from the whole array of microtopographies and demonstrates the impact of surface texture on cellular alignment. The complete cellular alignment data and staining images of the cells are presented in the supplementary section, Figure Supp. 2. 70-80% of the Schwann cells aligned within ± 10° to the direction of the long-axis for the following microchannels: CH 2x2, CH 10x10, CH 2x20, CH 20x20, CH 20x10, and CH 10x20. On the Sharklet™ microtopographies, only 50-60% of Schwann cells were aligned within ± 10° on SK 10x2, and SK 10x10 indicating that the channels were superior at Schwann cell alignment. In general, fibroblasts aligned to a lower extent on all microtopographies, compared to the Schwann cells. Only on CH 2x10 did 83% of fibroblasts align with the pattern, whereas the remaining channel microtopographies only aligned 25-66% of the fibroblasts. The alignment of fibroblasts on the Sharklet™ features was even lower at 14-39%. In general, microtopographies have varying effects on cellular alignment for different cell types, Figure Supp. 3. In particular, Schwann cells align to a greater extent than fibroblasts, and smaller channels provide enhanced alignment compared to Sharklet™ micropatterns of comparable dimensions.

Figure 5. Cell Alignment.

Cell alignment was measured based on the angle of the cell nuclei relative to the pattern feature’s long-axis. The schematic A) shows different examples of attached cells and their measured angles. B) Rose diagrams generated from the measured angle of the cells ranging from ± 90° (perpendicular) to 0° (parallel). C) Shows images of cells on selected micropatterns that impact the morphology of the fibroblasts and Schwann cells the most (CH 2x2 and CH 10x10) and the least (SK 2x20 and SK 20x2) compared to the smooth surface (control), (For the full panel results please see Supplementary Figures 2 and 3). Actin filaments (Red) stained by phalloidin-Alexa Fluor 660, and nuclei (Blue) are stained by DAPI. Magnification: 20x. Rose diagrams, for each cell/pattern is plotted under the correlated image.

Cell Morphology:

SEM images of cells cultured on micropatterns after 3 days were captured and used to qualitative assessment of the cellular morphology on selected microtopographies Figure 6 A-D. In general, fibroblasts were flattened and their cytoplasm were expanded over and between the features of both the Sharklet™ and microchannels. Schwann cell extensions were stretched out and adhered on the top or edge of the Sharklet™ features and were highly elongated on microchannels. In addition, Schwann cells expressed more cytoplasmic projections (i.e., filopodia) compared to fibroblasts, Figure 6 E.

Figure 6. Cell Micromorphology.

SEM images of the cells after 3 days in culture on the micro-topographies. Cells were fixed and dehydrated before imaging to maintain their morphologies under vacuum. Fibroblasts on A) SK 2x20 and B) CH 20x10; Schwann cells on C) SK 2x20 and D) CH 20x10. On A) SK 2x20, the cell body of the fibroblasts are stretched between and over the features of the micropattern. In the case of C, and D) Schwann cells, the cell body remain smaller but pseudopodia-like cell extensions are stretched out and adhered on top and inside of features. On the parallel channels, B) CH 20x10, fibroblasts are flattened and their cytoplasm are expanded over the channel features. However, D) Schwann cells are highly elongated and stay in between the channels. Schwan cells express filopodia (slender cytoplasmic projections) all over the cell surface, but fibroblast express fewer filopodia only close to the edges. E) Schwann cells filopodia attached to the feature of a SK 2x20 pattern.

3.3. Model Fitting

All hierarchical ANOVAs converged with the potential scale reduction factor for all parameters less than 1.05. Posteriors of σ_hyper and σ_obs shrunk the greatest relative to the prior for alignment models with relative similar median values for Channel and Sharklet™ (Figure Supp. 4 and 5). Alignment models had the lowest σ_hyper indicating the least amount of variation in alignment values between micropatterns while metabolic activity models had the greatest σ_hyper indicating the greatest amount of variation in microtopographical patterns.

4. Discussion

Chronic reliability of peripheral nerve implants has been considerably challenged by foreign body reaction, which is mainly controlled by immune cells, fibroblasts, and Schwann cells. In this study, we have demonstrated that the in vitro behavior of two cell types, which contribute to regeneration around and encapsulation of peripheral nerve implants, are heavily influenced by engineered microtopographies. The specific microtopography geometry and feature spacing impact cellular metabolic activity, attachment, and morphology, and the relative magnitude of these changes is highly dependent on the cell type. Importantly, we found that dissimilar microtopographies evoke different responses in different cell types. CH 20x2 was shown to inhibit the metabolic activity of fibroblasts that are responsible, in part, for the FBR while promoting the metabolic activity and directional alignment of Schwann cells that are responsible, in part, for healthy neural regeneration.

In general, metabolic activity of Schwann cells was higher than fibroblasts when cultured on both channel and Sharklet™ micropatterns. The changes in metabolic activity of the Schwann cells correlate with variations in cell proliferation, differentiation, and morphology. Schwann cells have the capacity to function in more than one mode during regeneration (e.g., myelinating, repairing, and non-myelinating^52,53), and they have the ability to differentiate, migrate and change morphology depending on environmental cues.^54,55 This phenomenon, known as Schwann cell plasticity, allows these cells to support neural regeneration.⁵⁵ Within the natural regenerative environment of damaged adult nerve tissue after transient injury, reactivation of Schwann cells occurs through chemical signaling that initiates because of the disruption of basal lamina and later, Wallerian degeneration of the axons.^55,56 Studies have demonstrated that Schwann cells are able to detect the injury days before receiving signals for axonal death, in spite of the long distance from the cell’s soma that is located several decimeters away.^57,58 This suggests that Schwann cells are sensitive to the physical changes in their microenvironment, including mechanical and architectural cues.⁵⁴ The tube-like architecture of basal lamina, longitudinally organized collagen fibers in normal ECM, and the aligned structure of bands of Büngner during regeneration are examples of natural micro- and nano-topographical cues that regulate Schwann cells responses.^24,59 Previously, Mitchel and Hoffman-Kim studied the motility of Schwann cells on microchannels (30-60 μm wide) and showed increased cellular orientation and alignment along the channels compared to smooth controls.⁶⁰ Therefore, the higher metabolic activity of the Schwann cells, in comparison with fibroblasts reported here, could be the result of their intrinsic ability of reacting to the morphological changes in their microenvironment.

Fibroblasts were targeted as a primary contributor to the foreign body reaction. The metabolic activities of fibroblasts during the time points studied were less than Schwann cells on both patterned and smooth surfaces. At day 7 of culture, the increase in metabolic activity of fibroblasts on the majority of the micropatterns was less than one-fold relative to the smooth surface (control) compared to two to three-fold relative for Schwann cells. Fibroblasts are responsible for producing fibrotic tissue during FBR.⁶¹ A comparative in vitro study of fibroblast and macrophage responses to the micropillar micropattern suggested that attachment and proliferation of fibroblasts were substantially enhanced by increasing the pillars’ height, whereas macrophage adherence was significantly diminished by reduced pillars’ spacing.⁶² This implies that the response to the micro-topographical features are highly dependent on the cell type as was shown here for Schwann cells and fibroblasts. Therefore, more studies should be performed on macrophages and immune cells to more fully understand the influence of microtopographies on the FBR and to complete the in vitro model.

Bio-inspired Sharklet™ patterns of targeted dimensions result in antifouling surfaces against a variety of marine organisms and have been reported to reduce bacterial attachment by 87%-99% compared to smooth surfaces.⁶³ Unlike the work with prokaryotic cells,³¹ no clear trends could be found that explain how geometrical parameters such as feature width and spacing influence the metabolic activity and cell adhesion of the eukaryotic cells in this study. However, it was clearly observed that none of the 2 μm Sharklet™ features (e.g., SK 2x10, and SK 2x20) promoted generic activity (the index was generated by summing results from metabolic activity and nuclei counts; See Figure 4). For the first time, Sharklet™ micropatterns were studied with respect to the FBR for peripheral nerve interfaces. These tests failed to identify a Sharklet™ size range that completely inhibited eukaryotic cellular adhesion. This finding is in contrast to the effect of Sharklet™ on prokaryotic microorganisms and select marine organisms. Previously, Decker et al described a thermodynamic model that can predict the promotion or inhibition of organisms with respect to Sharklet™ micropattern geometry and feature size that was successfully tested for predicting bacterial adhesion, among others.³¹ However, the adhesion behaviors of the mammalian cells that were tested in this study did not fit this model for a variety of possible reasons including size and anatomical differences between eukaryotes and prokaryotes. In addition, bacteria tend to maintain their distinctive shape when adhering to a surface compared to eukaryotic cells because of their rigid peptidoglycan cell wall compared to the deformable mammalian cells’ phospholipid membrane.^64-66 In a recent study, researchers compared cell adhesion of human fibroblasts and three bacteria species involved in infection of oral implants as a function of surface roughness. Results showed the behavior of fibroblasts and bacteria is quite dissimilar, probably because mammalian cells and bacteria have different attachment strategies.⁶⁷ Based on the result of this study, we were unable to generate a predictive model that relate micropatterns to metabolic activity and alignment of the Schwann cells and fibroblasts conceivably because mammalian cell responses are more complicated than those of bacteria and highly sensitive to the microenvironmental factors. We assume to generate such predictive model, if it is at all possible, multiple geometers should be tested in presence and absence of chemical factors.

Parallel microchannels were found to be more effective at aligning cells compared to Sharklet™. However, the relative magnitude of this effect was different between the two cell types with Schwann cells being more highly aligned compared to fibroblasts. This can be explained through the differences in the number of surface proteins and localization of the receptors that trigger signaling cascades to activate responsive programs in a particular cell type. In this case, Schwann cells are shown to be more sensitive to micro-architecture of the surface than fibroblasts.¹⁷ SEM analysis supports this theory by showing that the filopodia on the Schwann cells cultured on micropatterns were more prominent than those observed in fibroblasts on the same features. Moreover, Schwann cells are naturally able to change their structure by elongating and flattening based on their physiological role, whereas fibroblasts generally maintain a spindle-shape under physiological conditions.

Conclusions

Parallel channels and Sharklet™ micropatterns of different size are able to induce different responses in Schwann cells and fibroblasts. Different cell types react dissimilarly to the same microtopographical features. For example, CH 20x2, was found to promote Schwann cell generic activity, while it inhibits fibroblasts. This engineered microtopography is a potential candidate for bioengineered surfaces for neural implants, in particular peripheral nerve interfaces, to mitigate FBRs. However, additional in vitro and in vivo studies are required to verify the function of these micropatterns, since the cellular responses could be influenced by mixed cell populations and the presence of the vascular and immune systems in vivo.

Surface patterning is a strong physical tool for modulating cellular responses that also has additional potential applications such as in cell sorting and in vitro model systems. However, when designing the microtopography to control the cell behavior, the application and cell type should be carefully considered first. Several studies have provided proof of concept when only studying how fibroblasts react to the micropatterns. However, this study has shown that the response to micropatterns is highly dependent on the cell type; to globalize the function of a specific micropattern on one cell type to other types is an oversimplification.