Advanced Biomimetic Nanostructured Microelectrode Arrays for Enhanced Extracellular Recordings of Enteric Neurons

Microelectrode surfaces covered with nanostructures derived from components of extracellular matrix, such as collagen fibers, have shown immense beneficial effects in promoting neuronal growth and cellular signaling. Synthetic nanostructures mimicking the features of biological nanostructures with durable conductive materials could promote the cell adhesion on microelectrode surfaces by providing topographical cues and simultaneously improve the charge transfer properties by reducing its global impedance. Therefore, an advanced nanostructuring method mimicking the structural and organizational features of natural collagen fibers onto metallic microelectrode surfaces has been presented here, which is adapted from previous technological achievements of the group and is based on nanoimprint lithography and gold electroplating. Surface characterization methods reveal an increase in surface area between 20% and 68% on the microelectrodes fabricated with two different nanostructure heights. Impedance spectroscopy measurements reveal reduction in impedance magnitude (at 1 kHz) between 22% and 41%, depending upon the nanostructure height and density on the microelectrode, which should subsequently modulate its charge transfer properties for biosensing application. Cell adhesion analysis performed with seal impedance measurements reveals a tighter coupling of enteric neurons on the nanostructured microelectrodes. Finally, extracellular recordings from enteric neurons exhibit a significant improvement in spike detection properties of the nanostructured microelectrodes.


Introduction
Since the discovery of electrical activity in neurons, researchers have desired to understand, analyze, and improve the fundamentals of electrophysiological signal transmission. With the introduction of microsystems engineering and thin film technology, the first microelectrode arrays (MEAs) were fabricated in early 1970s. [1] The usage of MEAs transformed the understanding of extracellular signal conduction, as the dimensions of the microelectrodes reached the resolution of a single neuron. [2] A major advantage of using MEAs for electrophysiological experiments in comparison to traditional patch clamp methods is the prevention of cell membrane being punctured during measurement, hence keeping the cell cytoplasmic shape undisturbed. Moreover, a single MEA can be used to measure activity from complete neuronal networks at the same time giving simultaneous information of a larger tissue and its interactions.
The first MEAs were fabricated with smooth metallic surfaces, where cells were expected to homogeneously spread and transmit signals. [1] But during following investigations it was discovered that the ability of neurons to adhere and consequently transmit electrical information was enhanced with increasing roughness of the electrode surface. [3] In fact, appropriate nanostructured biological interfaces have shown to increase the neuronal activity on MEAs, which could be based on higher expression rates of ion-channels that are involved in the electrical activity of the neurons. [4][5][6][7] From a biological point of view, neurons could prefer a rougher surface, as it mimics the nanostructured environment contributed by components such as extracellular matrix (ECM) proteins, bundled myelinated fibers, etc. present on its natural adhesion surface. [8] Consequently, development of MEA surfaces with nanorough and nanoporous topography started getting popular, which eventually led to introduction of microelectrode nanostructuring. [9] Early nanostructuring of microelectrode surfaces was performed with extracellular matrix biomolecules (such Microelectrode surfaces covered with nanostructures derived from components of extracellular matrix, such as collagen fibers, have shown immense beneficial effects in promoting neuronal growth and cellular signaling. Synthetic nanostructures mimicking the features of biological nanostructures with durable conductive materials could promote the cell adhesion on microelectrode surfaces by providing topographical cues and simultaneously improve the charge transfer properties by reducing its global impedance. Therefore, an advanced nanostructuring method mimicking the structural and organizational features of natural collagen fibers onto metallic microelectrode surfaces has been presented here, which is adapted from previous technological achievements of the group and is based on nanoimprint lithography and gold electroplating. Surface characterization methods reveal an increase in surface area between 20% and 68% on the microelectrodes fabricated with two different nanostructure heights. Impedance spectroscopy measurements reveal reduction in impedance magnitude (at 1 kHz) between 22% and 41%, depending upon the nanostructure height and density on the microelectrode, which should subsequently modulate its charge transfer properties for biosensing application. Cell adhesion analysis performed with seal impedance measurements reveals a tighter coupling of enteric neurons on the nanostructured microelectrodes. Finally, extracellular recordings from enteric neurons exhibit a significant improvement in spike detection properties of the nanostructured microelectrodes.
as collagen, laminin, etc.) extracted from natural sources. [10] Although these coatings significantly improved the interaction of cells with electrodes, they were susceptible to adverse conditions such as temperature and pH. [11] Moreover, as the biomolecules are inherently insulating due to their hydrocarbon composition, they were responsible for increasing the surface impedance of the microelectrodes. [12] This led to the emergence of synthetic durable nanostructures on microelectrode surfaces with conductive materials, which not just enhanced the cell-electrode interface, but also improved its signal transduction properties. [13][14][15] Another important aspect of developing nanostructured MEAs is the mismatch of physical properties between the biosensor and the neural tissue. MEAs developed for in vitro application are generally fabricated with metallic electrode surfaces deposited on brittle substrates such as glass or silicon. This could lead to formation of glial scar tissue and subsequently reduction in efficacy of electrophysiological application. [16] Studies have shown reduction of glial scar formation induced by the presence of nanostructures. [17] Moreover, flexible and biocompatible substrates (e.g., polyimide and polydimethylsiloxane) deposited with softer nanostructured electrodes (e.g., carbon-based materials [18] ) would further help alleviate this issue.
The physical effects induced by nanostructures on MEAs during charge transfer can be effectively characterized by electrochemical impedance spectroscopy measurements. In MEA technology, impedance spectrum is an aggregate of charge transfer and diffusion processes occurring at the boundary between the electrode surface and electrolyte solution. Impedance measurements are particularly important for MEAs in biosensing application, as reduced impedance results in smaller thermal noise and enhanced signal-to-noise ratio. [19] Impedance of microelectrodes could be reduced by simply enlarging its active surface area dimensions. But this leads to loss in spatial resolution of measurement across the cell culture surface. Introduction of nanostructures is an intuitive way of increasing the effective surface area without changing the dimensions of the microelectrodes and methods such as cyclic voltammetry and mathematical fitting with electrical equivalent circuits have shown that the reduction in impedance due to the presence of nanostructures could be related to enhanced double-layer capacitance and reduced charge transfer resistance. [20,21] In recent times, impedance measurements have also been utilized to characterize the adhesion of cells to nanostructured microelectrodes. Spira et al. cultured Aplysia neurons on mushroom shaped nanostructured gold microelectrodes and designed a schematic electrical circuit analogue representing the neuron-nanostructure interface. In combination with electrophysiological experiments, they concluded that the impedance contributed by the seal between the neuron and electrode surface plays a major role in determining the strength of the coupling. [22] Likewise, Decker et al. measured a wide range of seal impedances for human embryonic kidney cells cultured on MEAs with nanostructures in multiple shape forms and correlated a strong dependency of cell adhesion with the dimension and contour of the nanostructures. [23] With the latest state-of-the-art thin film technologies, there are several options to fabricate organized synthetic nanostructures on microelectrodes with vast choices of materials, dimensions, isotropy, shape, etc. [24] The main purpose of the synthetic nanostructures is to mimic the natural roughness of the cell microenvironment. Therefore, nanotopographical features of the natural adhesion surface are perfect examples to derive inspiration for fabricating biomimetic nanostructures. The basement membrane serves as an adhesion substrate for overlying cellular structures and consists of several ECM biomolecules such as collagen, hyaluronic acid, proteoglycans, laminin, and fibronectin distributed in random spatial organization. [25][26][27][28] The dimensions of these biomolecules range from ten to a several thousand nanometers. [25][26][27][28] Therefore, it is necessary to study the interaction of cells with randomly organized synthetic nanostructures on microelectrodes that resembles the natural adhesion surfaces more appropriately than organized symmetric nanostructures.
Of the several biomolecules present in the ECM network the most prominent presence is that of collagen. [29] Collagen extracted from natural sources (e.g., rat tail) and coated on MEA surfaces has shown to enhance the neuronal surface adhesion and consequently its growth and differentiation. [30] In a recent study, we demonstrated a reproducible method to fabricate biomimetic nanostructures on macroelectrode surfaces (20 mm 2 ) replicating the dimensions and distribution of collagen fibers from bovine Achilles tendon into gold nanostructures using nanoimprint lithography (NIL) and gold electroplating. [31] NIL is a high efficiency nanostructure fabrication process with spatial resolution proven to be as small as 10 nm. [32] The biomimetic collagen-like gold nanostructures not only improved the wetting and impedance properties of the substrates, but also proved beneficial for cellular growth. Therefore, the same biomimetic nanostructuring process was applied to functionalize the surface of microelectrodes and subsequently its influence on neuronal signal transmission was evaluated.
Adapting the nanostructuring method from macroelectrode surfaces to microelectrode surfaces (≈756 µm 2 ) is indeed a challenging task. Firstly, the density of collagen fibers coated on silicon master in NIL has to be significantly amplified in order to see definite modifications in the microelectrode properties. Secondly, the fabrication process of functional nanostructured MEAs must be optimized for biosensor application, in order to detect microvolt range extracellular activity from neuronal membrane. Therefore, the nanostructured MEAs were transferred from wafer-scale to chip-scale by flip-chip bonding and were then characterized with surface microscopy methods and impedance spectroscopy. Afterward, enteric neuronal culture experiments were performed to observe the improvements in cell-electrode adhesion and extracellular action potential recordings.

Nanostructure Topography
Collagen-like gold nanostructured microelectrodes were fabricated with the process line illustrated in Figure 1a-h. The final MEA chip was assembled on a printed circuit board (PCB) with flip-chip bonding (Figure 1i). The collagen-like gold nanostructures (CLGNS) were grown on the microelectrode surfaces by www.advmatinterfaces.de controlled gold electroplating along the thickness of the nanoimprint resist (Figure 1d). The final thickness of the nanoimprint resist layer post residual layer removal was set to be 50 nm and it is dependent on the smallest reproducible nanofeature on the collagen coated silicon master. [31] Technically, the resist layer thickness can be larger than 50 nm to increase the CLGNS height, but that will lead to loss of information to be replicated from the collagen master. At the same time, the resist layer can be smaller than 50 nm for maximum information replication from the collagen coated master. But the CLGNS height would then be in the range of the roughness of the nanoimprint resist. Therefore, as a good compromise between maximum CLGNS height and the extent of information replication from the collagen master, 50 nm was chosen to be optimum thickness of the resist. Detailed information on the selection of 50 nm as the smallest reproducible feature of collagen coated master has been presented in our previous work. [31] Moreover, studies suggest enhancement of communication in neuronal networks for surfaces with nanoroughness in the range of 30 nm, in comparison to unstructured surfaces. [33] For the nanoimprint resist thickness of 50 nm on a 4-in. substrate wafer, the electroplating time t has been calculated by estimating the total nanostructured surface area and applying it in Faraday's law of electrolysis, where h is the CLGNS height, J is the electroplating current density (1 mA cm −2 ), F is the Faraday's constant (96 485 C mol −1 ), ρ is material density of gold (19.32 gm cm −3 ), M is the molar mass of gold (198 g mol −1 ), and z is the valence number of the gold ions. [34] · · · · By adapting the electroplating time in Equation (1), MEAs were fabricated with CLGNS of two different heights on two different substrate wafers: 30 and 50 nm. For easier understanding, microelectrodes with CLGNS height 30 nm are denoted as H1 and with CLGNS height 50 nm are denoted as H2. Figure 2a shows the scanning electron microscopy (SEM) images of the collagen coated silicon master stamp used for NIL and Figure 2b,c shows H2 nanostructured microelectrode with and without neurons cultured on the surface, respectively. The lateral width of the individual CLGNS was measured to be between 100 nm and 3 µm, and the length between 50 nm and 20 µm. The aspect ratio of the electroplated CLGNS is measured to be larger in comparison to the collagen fibers coated on the silicon master. This is because the nanoimprint cavity on the resist is made by the base of the collagen fiber, which has a maximum width compared to the rest of the fiber. Secondly, during the reactive-ion etching of the residual layer removal (Figure 1c), side walls of the resist cavities are partially etched by the reflecting reactive-ions leading to increased width of the cavities. [35] Although this cannot be eliminated with our approach, the overall random network structure of the collagen fibers is replicated onto the microelectrodes, which was the main goal of this study.
The microelectrodes fabricated in this study have a diameter of 30 µm and the lateral dimension of the individual nanostructures is in the range between 100 nm and 3 µm. Therefore, individual microelectrodes in a MEA are expected to have different nanostructure coverage densities and subsequently different increase in surface areas. Atomic force microscopy (AFM) was used to measure the surface profile of individual microelectrodes and then the density of CLGNS of 30 H1 and www.advmatinterfaces.de 30 H2 microelectrodes with 30 µm diameter were calculated with computational methods (Figures S2 and S3, Supporting Information). The overall increase in surface area varied from 20% to 63% for H1 nanostructured microelectrodes and 21-68% for H2 nanostructured microelectrodes. Due to the presence of such a large variance in the nanostructure density, the microelectrodes were further classified to low density (denoted as D1) and high density (denoted as D2). D1 microelectrodes have a CLGNS density in the range between 20% and 40%, whereas D2 microelectrodes have a CLGNS density in the range between 40% and 68%. Therefore, in this study along with planar unstructured microelectrodes, the CLGNS microelectrodes were grouped into H1D1, H1D2, H2D1, and H2D2. Examples of AFM surface profile of each type of microelectrode can be seen in Figure 3a-d and the distributions of the different types of CLGNS microelectrodes have been tabulated in Table 1.
It is important to realize that the CLGNS density on the microelectrodes is only dependent on the dimensions of the least reproducible collagen fiber on the master stamp, which is also the final nanoimprint resist thickness (Figure 1c). [31] It is independent of the height of CLGNS grown by electroplating. The distribution of CLGNS microelectrodes in Table 1 is an example for the structuring process done with a final nanoimprint resist thickness of 50 nm. The nanostructuring process was tested for master stamp with even denser collagen fiber network and thus increasing the CLGNS density on the microelectrode to be more than 68%. But for such high CLGNS density, a clear distinction between the nanostructures and the smooth electrode surface could not be observed and the structural information of the natural collagen was effectively lost.

Electrochemical Impedance Spectroscopy
Electrochemical impedance spectroscopy (EIS) is a quantitative and qualitative measurement method used to estimate the electrochemical signal transmission properties of a microelectrode, especially in a fluidic environment. In the previous section, a brief description of the topographical properties of the CLGNS on the microelectrodes has been presented and the nanostructured microelectrodes were grouped based on the structure height and density. In this section, the EIS measurements for each group of nanostructured microelectrodes in comparison to unstructured microelectrodes has been discussed.
In the impedance magnitude spectrum (Figure 4a), it can be observed that the nanostructured microelectrodes, in general, have lower impedance compared to unstructured microelectrodes within the frequency range of 1 Hz to 10 kHz. In the frequency range greater than 10 kHz, the impedance reduction induced by CLGNS becomes less significant. However, this frequency range is of little relevance in neuronal activity measurements. Upon further investigation of the different groups  www.advmatinterfaces.de of nanostructured microelectrodes, it can be observed that the height and density of the CLGNS have a considerable effect on the global impedance properties.
The impedance magnitude at 1 kHz of the nanostructured microelectrodes along with unstructured microelectrodes can be seen in the inset of Figure 4a and Table 1. The H2D2 microelectrodes show a maximum reduction of 41.3% in impedance magnitude, whereas H2D1, H1D2, and H1D1 show a reduction of 30.1%, 28.3%, and 22.7%, respectively. It can be inferred here that the height of the CLGNS plays a critical role in the impedance reduction, as H1 nanostructured microelectrodes show minimal impact in comparison to H2 nanostructured microelectrodes. The effect of the electroplated nanostructure height on the impedance spectroscopy signifies its importance of enhancement in charge transfer properties of the microelectrode, especially for electrophysiological application. The density also plays an important role in impedance properties, as densely nanostructured microelectrodes (D2) show a much significant impedance reduction for both H1 and H2 MEAs, in comparison to sparsely nanostructured microelectrodes (D1). Therefore, for optimum charge transmission properties it is reasonable to utilize CLGNS microelectrodes with maximum impedance reduction, i.e., with higher density and larger height (H2D2).
From the impedance phase plot (Figure 4b), the most significant effect of the CLGNS can be observed at the phase shift at a frequency of 100 kHz, which suggests that the increase in the double-layer capacitance could be the predominant contributor to the reduction of global impedance of the nanostructured microelectrodes. [36][37][38] Similar to impedance magnitude measurements, the maximum capacitive enhancement was observed for H2D2 microelectrodes, followed by H2D1, H1D2 and H1D1 CLGNS microelectrodes.
With the collagen fiber network coated on the silicon master stamp used for the fabrication of CLGNS in this study, more than 70% of microelectrodes in the array are structured with D2 density. And since the most significant impact of the CLGNS on impedance properties was observed with H2D2 microelectrodes, only H2 nanostructured MEAs were exclusively used for cell adhesion evaluation and electrophysiological experiments. Figure 5a shows the electrical circuit analogue of the interface between neurons and nanostructured microelectrode. The charge transfer components of the cell membrane (extra-and intracellular ion channels) and the nanostructured microelectrode are shunted by a seal impedance which is inversely proportional to the distance between the two surfaces at the junction. Therefore, a larger seal impedance attributes to better surface coupling of cells to the microelectrodes. [39] Multiple studies have illustrated the importance of cell adhesion to the life events of cells such as growth, differentiation Table 1. Distribution of different nanostructured microelectrodes and their impedance magnitude at 1 kHz. The surface profiles of nanostructured microelectrodes measured for impedance spectroscopy can be seen in Figure S2 and Figure S3 (Supporting Information).

www.advmatinterfaces.de
and proliferation. [40][41][42] This is also true for electrogenic cells, which would eventually lead to enhanced extra-and intracellular signal transmission. [43] In this study, the seal impedance at 1 kHz was calculated with enteric neurons cultured on top of the MEAs on day 5 in vitro (DIV) to estimate the cell adhesion. 1 kHz frequency was specifically chosen for the seal impedance measurements as the electrogenic activity of the neurons (spikes, bursts, post-synaptic potentials) are recorded in the frequency range between 500 Hz and 3 kHz. [44,45] It is beneficial to use non-electrogenic cells (e.g., human embryonic kidney cells) for estimation of seal impedance on microelectrodes, as they would not interfere with the measurement potential across the cell-electrode junction. [23] But at the same time, it is also known that the adhesion mechanism of cells vary depending upon their type and morphology. [46] Since the microelectrodes are primarily used for sensing activity of electrogenic cells (e.g., neurons), it would be appropriate to use the same cells for characterization of the adhesion to CLGNS MEAs. Moreover, this estimation is only valid if the coverage of cells is uninterrupted on the MEA surface with maximum confluency, as presence of large number of alternate pathways between the potential source and the microelectrode would not result in accurate measurement of seal impedances. [23] Figure 5b shows the measured impedance magnitudes for 30 unstructured and 30 H2 nanostructured microelectrodes, before and after cell culturing. The impedance measurements were performed in Dulbecco's modified Eagle medium (DMEM) solution as electrolyte, as it was also used as culture medium for neurons. In Table 2, it can be observed that the impedance measured for H2 nanostructured microelectrodes without cells is lower than that of the unstructured microelectrodes, which fits to the EIS measurements presented in the previous section. Whereas, for the same microelectrodes, the impedance measurements with enteric neurons show a contrasting result. The H2 nanostructured microelectrodes show a mean increase of a factor 3.4 in impedance compared to unstructured microelectrodes, which implies that the seal gap between the cell membrane and the microelectrode surface is much smaller and the cell adhesion is much tighter. The variance of the seal impedance measured with enteric neurons is much larger than the variance without cells. This is due to the fact that on DIV 5 the neurons are still under the process of differentiation and the extracellular activity from the early neurons could act as internal current sources, which could interfere with the impedance measurements. Nevertheless, even with the minimal interference on impedance measurements, it can be suggested that CLGNS do indeed reduce the seal gap for cell coupling on the microelectrodes, which would subsequently improve the cell proliferation and development.

Enteric Neuron Electrophysiology
Reduction in global electrode magnitude due to the increased effective surface area and tighter cell coupling measured with seal impedance measurements could result in superior charge transmission properties of the CLGNS microelectrodes. Therefore, extracellular activity from enteric neurons was recorded from H2 nanostructured as well as unstructured microelectrodes and the differences in the action potential spike properties were analyzed. Figure 6a shows the confluent spread of enteric neurons on the MEA surface used for extracellular recordings on DIV 10. Overview of signal processing steps for extraction of action potential spike properties from extracellular recordings of enteric neuronal culture on MEAs can be seen in Figure S4 (Supporting Information). The extracellular activity recorded by the microelectrodes were analyzed in an active measurement time period of 30 s and the spikes were detected from baseline noise by thresholdbased method. Figure 6b,c shows the averaged cutouts of all spikes detected by the H2 nanostructured (n = 24) and unstructured (n = 12) microelectrodes. Figure 6d shows the comparison of the number of spikes detected in 30 s measurement period by the H2 nanostructured (n = 24) and unstructured (n = 12) microelectrodes. The nanostructures are expected to enhance the neuron adhesion to the electrode surface which then promotes the intra-and intercellular signaling. This subsequently would correspond to larger number of spikes detected by the  microelectrode. Nevertheless, no significant difference was observed between the two sets of microelectrodes. This could only the case because the effect of nanostructuring is confined to the microelectrodes and the rest of the cell culture is grown on unstructured passivation layer. Furthermore, the spike rate is proportional to not just the cell dynamics of the neuron in contact with the microelectrode, but the complete neuronal network in the vicinity of the microelectrode, which is comprised of the unstructured passivation layer.
In this study, maximum and minimum peak-to-peak amplitudes are denoted as the largest and smallest spike amplitudes detected for microelectrode recoding channels respectively, in the 30 s measurement period. The mean peak-to-peak amplitude is denoted as the average of all spikes detected for a microelectrode recoding channel in the 30 s measurement period. Figure 6e shows the comparison of the averaged maximum, minimum and mean peak-to-peak amplitude (V p2p ) of the spikes detected by all H2 nanostructured (n = 24) and unstructured (n = 12) microelectrodes. Increase in the minimum V p2p (by 51%) for the H2 nanostructured microelectrodes is an indicator for the improved detection of diminished action potential spikes which are usually superposed by noise due to an insufficient seal between cells and electrode. Increase in the maximum V p2p (by 70%) is an indicator of the electronic amplification of the recorded signal which can be attributed to reduced electrode impedance. Increase in the mean V p2p (by 55%) is an indicator of the cumulative enhancement of the signal transmission properties of the microelectrodes induced by the CLGNS. Every microelectrode-cell membrane junction is characterized by a different seal gap (as illustrated in Section 2.3) Figure 6. a) Microscopy image of enteric neurons cultured on the MEA surface (inset scale bar 50 µm). Microelectrodes encompassed in blue box are H2 nanostructured and microelectrodes encompassed in red box are unstructured. Spike cutouts of neuronal extracellular activity recorded on DIV 6-7 from b) H2 nanostructured and c) unstructured microelectrodes. Each colored line represents the averaged cutout of all spikes detected by the corresponding microelectrode channel in a measurement period of 30 s and the black colored line represents the mean of all the averaged cutouts for the H2 nanostructured and unstructured microelectrodes. d) Comparison of number of spikes detected in 30 s measurement period by the H2 nanostructured (NS) and unstructured microelectrodes (US). e) Comparison of smallest (min), largest (max) and mean peak-to-peak amplitude of spikes detected for the H2 nanostructured and unstructured microelectrodes. f) Comparison of signal-to-noise ratio (ratio of mean peak-to-peak amplitude and root-mean-square baseline seal noise) measured for H2 nanostructured (NS) and unstructured microelectrodes (US). Analysis of statistical significance by one-way ANOVA (*p < 0.05, **p < 0.01, n (H2 nanostructured microelectrodes) = 24, n (unstructured microelectrodes) = 12).

www.advmatinterfaces.de
and its corresponding baseline seal noise. Therefore, the ratio of the mean peak-to-peak spike amplitude to its corresponding root-mean-square amplitude of baseline seal noise has also been analyzed for the H2 nanostructured (n = 24) and unstructured (n = 12) microelectrodes (Figure 6f). Improvement of the signal-to-noise ratio (by 35%) for the microelectrodes confirms the overall effect of the nanostructures in enhancing the transmission of extracellular signals from the neuronal membrane to the microelectrode. Summary of the spike properties measured for H2 nanostructured and unstructured microelectrodes has been enlisted in Table 3.

Conclusion
In this study, an advanced nanostructuring method for the functionalization of microelectrode arrays has been presented, which was adapted from the biomimetic replication of natural collagen fibers into gold nanostructures with nanoimprint lithography and gold electroplating previously developed by our group. The enhancement of signal transmission properties of the microelectrodes by the CLGNS can be envisioned in three perspectives. Firstly, the increase in effective surface area of the microelectrodes resulted in reduction of global impedance of the electrodes, which in turn resulted in reduction of baseline noise of the recordings. Secondly, presence of CLGNS resulted in reduction of the seal gap between the microelectrode surface and the neuronal membrane, which in turn reduces the probability of extracellular activity being dissipated by ohmic loss during transmission. Finally, CLGNS mimic the geometry of extracellular matrix biomolecules present in the natural adhesion surface of the neurons. Therefore, it could be assumed that neurons bind firmly onto CLGNS during their growth process which helps in formation of stronger cytoskeleton structure, as well as, robust focal adhesion complexes. Stable cell structure in turn regulates stronger neuronal signaling leading to amplified action potential signals. The biomimetic surface functionalization process of microelectrodes described in this study can be further improved by extending the nanostructures to the dielectric passivation layer surface along with the conductive electrode surface. Such a whole-surface biomimetic nanostructured microelectrode array would further enhance its electrophysiological abilities for extra-and intracellular recordings.

Experimental Section
Fabrication of Nanoimprint Lithography Master Stamp: The master stamp for NIL was developed by adhesion and pattering of collagen fiber network on 4-in. silicon wafer. Bovine Achilles tendon collagen type I (C9879, Sigma-Aldrich Corporation) was dissolved in 0.01 m HCl and homogenized by blending (9000 RPM for 5 min) and subsequent filtration (20 µm nylon filter). The collagen solution was then spincoated (EBS 10, RRT Lanz AG) on an ultra-clean silicon wafer (SSP 100, Siegert Wafer GmbH). In order to obtain defined regions for nanostructuring, the collagen coated silicon master was then patterned by photolithography (MA/BA 6, SÜSS MicroTec SE) and the uncovered collagen fibers were removed by reactive-ion etching with O 2 gas flow of 30 sccm at 30 Pa and 100 W power for 60 s (SI 591 m, SENTECH Instruments GmbH). Optical microscopy image of the collagen coated silicon master stamp can be seen in Figure S1a (Supporting Information).
Nanostructuring of Gold Substrate Wafer: The nanostructured microelectrodes are fabricated with 200 nm evaporated gold as substrate electrode material and 20 nm evaporated titanium as adhesion layer (BAK-500, Balzers-Evatec AG) on 4-in. borosilicate glass wafer (Borofloat, Siegert Wafer GmbH). The evaporated gold substrate wafer was spin-coated with thermal nanoimprint (T-NIL) resist (mR-I 8030R, micro resist technology GmbH) which has a glass transition temperature (T g ) of 110 °C. The resist coated gold substrate wafer was then imprinted (Eitre 6, Obducat AB) with the aforementioned collagen coated silicon master at an imprint temperature of 185 °C and a pressure of 30 bar for 240 s (Figure 1a) and subsequently released at 85 °C (Figure 1b). An anti-sticking layer (BGL-GZ-83, PROFACTOR AG) was spin-coated prior to the nanoimprint processes to improve the collagen master release properties. Optical microscopy image of the nanoimprinted resist can be seen in Figure S1b (Supporting Information). The initial thickness of the T-NIL resist was measured to be 450 nm and the smallest reproducible feature of collagen coated master was measured to be 50 nm. Therefore, 400 nm of T-NIL resist residual layer was sequentially etched with reactive-ion etching with O 2 gas flow of 30 sccm at 13 Pa and 25 W power for 220 s (SI 591 m, SENTECH Instruments GmbH) until the cavities created by NIL process are open for gold electroplating (Figure 1c). The open cavities of the T-NIL resist were then filled with gold by galvanostatic electroplating process in a twoelectrode setup (potentiostat SP200, Biologic SAS) with a wafer-holder as working electrode and a platinum mesh as counter electrode in a sulfite base gold electrolyte solution (Goldbad SF, METAKEM GmbH) (Figure 1d). By estimating the total area to be electroplated, the 4-in. wafer was electroplated at 1 mA cm −2 for 72 s to fabricate nanostructures with height 30 nm (H2 MEAs) and 120 s to fabricate nanostructures with height 50 nm (H2 MEAs). Post electroplating, T-NIL resist was stripped with acetone solution leaving stand-alone gold nanostructures (Figure 1e). Optical microscopy image of the CLGNS on substrate wafer can be seen in Figure S1c (Figure 1f). Optical microscopy image of the patterned MEA can be seen in Figure  S1d (Supporting Information). For insulating the microelectrode regions, two different passivation materials were tested. 1 µm thickness of SU-8 epoxy-based polymer (Kayaku Advanced Materials, Inc.) deposited by spin-coating and a multilayered stack of alternating silicon oxide and silicon nitride with a cumulative height of 1 µm deposited by plasma-enhanced chemical vapor deposition (Plasmalab system 113,

www.advmatinterfaces.de
Oxford Instruments) were the aforementioned passivation materials (Figure 1g). No significant difference in the impedance properties of the microelectrodes was observed between the different passivation materials. The microelectrode regions and the external contact pads on the passivation layer were opened by reactive-ion etching with CF 4 -O 2 gas flow of 25 and 3.1 sccm respectively at 13.33 Pa and 100 W power for 12 min for 1 µm SU-8 polymer and 10 min for 1 µm silicon nitride and silicon oxide stack (SI 591 m, SENTECH Instruments GmbH) (Figure 1h). Optical microscopy image of the final MEA wafer can be seen in Figure S1e (Supporting Information). The 4-in. wafer was diced (DAD-2H/6T, Disco Corporation) in to individual MEA unit dies (11 mm × 11 mm), which were then used to build fully functional MEA chips by flip-chip assembly (MPL3100 Micro Placer, Essemtec AG) and silver adhesive paste (H20E-PFC1OZ, Epoxy Technology, Inc.) on PCB. In order to apply the cells and culture medium on the MEA chip, cylindrical glass rings were mounted on the PCB with silicone polymer (sylgard 184, Dow Silicones Corporation), resulting in a culture cavity volume of 1 ml (Figure 1i). Microscopy Methods: Structural dimensions of CLGNS on the microelectrodes were measured with AFM (Dimension IKON, Bruker Corporation) and topographical analysis of the measured profiles were performed with computer software (Gwyddion 2.60, Czech Metrology Institute). SEM (Supra 40 Gemini, Carl Zeiss AG) was used for imaging of MEA dies with and without neurons on the surface. Prior to imaging, the neurons were fixed in 5% glutaraldehyde solution (C5882, Sigma-Aldrich Corporation) on DIV 10 and then dried in a mixture of ethanol and hexamethyldisilazane (440191, Sigma-Aldrich Corporation) solutions of varied concentrations in sequential steps (see Supporting Information Data S5 for steps in cell fixation and drying for SEM imaging). The MEA dies (with and without cells) were sputtered (Sputter Coater 108auto, Cressington Scientific Instruments Ltd.) with 5 nm gold layer prior to imaging to reduce charging effects of the dielectric passivation insulator regions.
EIS Measurements: EIS measurements of nanostructured and unstructured microelectrodes were performed in a potentiostatic threeelectrode setup (potentiostat SP200, Biologic SAS) with a 2 mm gold disc as counter electrode and Ag/AgCl reference electrode in 0.01 m KCl electrolyte. An input voltage of 10 mV and a DC bias of 0.12 V was used for measuring impedance in frequency range from 1 Hz to 3 MHz. The EIS measurements were represented in Bode magnitude and phase diagram with the corresponding potentiostat software (EC-Lab v11.43, Biologic SAS).
Cell Adhesion Analysis: Adhesion of enteric neurons to the nanostructured and unstructured microelectrodes was analyzed by simultaneously measuring the impedance at 1 kHz of the 60 microelectrodes with MEA-IT60 device (Multi Channel Systems MCS GmbH) before culturing of cells and on DIV 5 of differentiation. To maintain continuity in the impedance measurements, DMEM medium (31331-028, gibco Inc.) was used as electrolyte for measurements without cells. The nanostructured MEAs were measured with Ag/AgCl wire (A-M Systems Inc.) as external reference.
Electrophysiological Experiments: The extracellular activity measurements of the enteric neuron cultured MEAs were performed on DIV 6-7 with MEA2100-Mini-60-System (Multi Channel Systems MCS GmbH) with Ag/AgCl wire (A-M Systems Inc.) as external reference. The signal processing was performed by the corresponding software (Multi Channel Experimenter V 2.19.0, Multi Channel Systems MCS GmbH) which includes a band-pass filter (frequency range 500 Hz-3 kHz) and threshold-based spike detector. Spike data processing from raw signal recordings done as explained in Figure S4 (Supplementary Data).
Statistical Analysis: Statistical significance of spike data analysis was done with Microsoft Excel software (Microsoft Corporation) using one-way ANOVA (*p < 0.05, **p < 0.01). Sample size (n) for H2 nanostructured microelectrodes is 24 and sample size (n) unstructured microelectrodes is 12. Error bars have been manually added to the graphs in Figure 6d-f.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.