Highly Customizable 3D Microelectrode Arrays for In Vitro and In Vivo Neuronal Tissue Recordings

Abstract

Planar microelectrode arrays (MEAs) for - in vitro or in vivo - neuronal signal recordings lack the spatial resolution and sufficient signal-to-noise ratio (SNR) required for a detailed understanding of neural network function and synaptic plasticity. To overcome these limitations, a highly customizable three-dimensional (3D) printing process is used in combination with thin film technology and a self-aligned template-assisted electrochemical deposition process to fabricate 3D-printed-based MEAs on stiff or flexible substrates. Devices with design flexibility and physical robustness are shown for recording neural activity in different in vitro and in vivo applications, achieving high-aspect ratio 3D microelectrodes of up to 33:1. Here, MEAs successfully record neural activity in 3D neuronal cultures, retinal explants, and the cortex of living mice, thereby demonstrating the versatility of the 3D MEA while maintaining high-quality neural recordings. Customizable 3D MEAs provide unique opportunities to study neural activity under regular or various pathological conditions, both in vitro and in vivo, and contribute to the development of drug screening and neuromodulation systems that can accurately monitor the activity of large neural networks over time.

Keywords: 3D flexible implants; 3D microelectrode arrays; 3D printing; neural interfaces; two‐photon polymerization.

Figure 1

Application modalities of 3D MEAs. A) The in vitro design typically integrates both the culture dish and the 3D MEA directly on the chip, representing a stand‐alone device that can be used for spheroids, organoids, and acute neural slices. B) The in vivo design decouples the MEA from the neuronal tissue, where the MEA is fabricated separately on a (flexible) substrate and flip‐chip bonded to a printed circuit board. Attached to a micromanipulator, the probe can be lowered to penetrate the neuronal tissue in in vitro and in vivo experiments.

Figure 2

Fabrication of 3D MEAs. A) Fabrication of 3D electrodes on a planar MEA substrate (A_i) by printing 3D polymer templates (A_ii) and the template‐assisted electrodeposition of Au and PEDOT:PSS. By carefully controlling the current, a cap can be formed at the top of the pillar (A_iii). B) Template‐assisted electrochemical deposition of Au inside straight pillars of 35 and 65 µm height. During the chronoamperometry process a constant potential of −1.3 V was used. The current‐time curve exhibits the four stages of an electrochemical deposition process to fill pillars with different heights with Au, which was stopped when the measured current began to increase exponentially, thereby indicating that the Au filling reached the top end of the pillar. C) During the second deposition step, the current was fixed to −100 nA for 20 s (C_i) to create a smooth and small Au cap (C_ii). The following PEDOT:PSS deposition was carried out via cyclic voltammetry in 2–10 cycles (in (C_iii) number of cycles was 10) depending on the desired size of the PEDOT:PSS cap (C_iv). D) Focused ion beam (FIB)‐cuts of one individual pillar at the base, revealing a wall‐thickness of ≈4 µm (D_i) and top end (D_ii) of the pillar. E) Fabrication results showing a stiff 3D MEA device (E_i) and a zoom in picture of an array with 3D printed pillars of different heights (40–100 µm) (E_ii)). F) Fabrication results showing a flexible 3D MEA (F_i) and a zoom in picture of the 3D printed pillars with a height of 500 µm (F_ii).

Figure 3

Characterization of 3D printed pillar electrodes. A) Bode plot of electrochemical impedance spectroscopy measurements of an example probe with 12 electrodes after Au (gold) and PEDOT:PSS (blue) electrodeposition. Mean values are shown in solid lines and the mean +/‐ standard error mean in shaded area. B) Analytical calculation and COMSOL simulation of the critical buckling load following Euler's equation for a fixed‐pinned boundary condition for cone‐shaped, straight, and multisite pillars of different heights. In light red, an insertion force threshold between 0.5–2 mN is shown. C) COMSOL simulation of Von Misses’ stress showing the highest possible pillars according to the critical buckling load (490 µm for cone‐shaped pillars, 190 µm for straight pillars, and 420 µm for multisite pillars with 3 electrodes). D) Literature comparison of bending stiffness with other common materials and designs for implantable devices (see Table S1, Supporting Information for the references and calculations).

Figure 4

In vitro cell culture with 3D electrodes on planar stiff MEA. A) Scanning electron microscope (SEM) image of a 3D cell‐culture at 14 days in vitro (DIV) of primary cortical embryonic rat neurons within a 3D printed scaffold containing 3D printed pillars that sits on top of the 2D MEA to guide the growth of the neuronal culture. Zoomed in images show that cells do not avoid the 3D prints (ii, red) and cell growth is even enhanced on the scaffold in comparison to the baseplate (iii, blue). B) Electrophysiological recordings from different electrodes (B_i) showing bursting activity (B_ii) (zoom‐in (B_iii)) and different spike shapes (B_iv) of a 3D cell‐culture of primary cortical embryonic rat neurons at 14 DIV.

Figure 5

In vitro recordings of explanted rodent retinas. Retinal recordings using in vitro A) and in vivo B) approaches, exhibiting raw electrical signals (A‐B_i), spiking activity (bandpass filtered signal) (A‐B_ii), and local field potentials (lowpass filtered signal) (A‐B_iii) captured upon optical stimulation. Snapshots of individual optical responses in A and B reveal that the firing rate (blue trace) of the spiking signal (black trace) increases upon optical stimulation (A‐B_iv). In (A‐B_v) distinct neural waveforms are shown. Additionally, in both cases, averaged ERG‐like waveforms (blue) upon four to five stimuli representing the summed activity of the retina were recorded. C) Insertion of multisite pillars containing three electrodes with a height difference of 20 µm each. Recordings exhibiting the intraretinal placement of a multisite pillar following a stepwise insertion (Z₁–Z₄). Additionally, recording extracts at Z₄ (C_ii, red window) and spike waveforms captured by individual electrodes (C_iii) within a single multisite pillar are shown.

Figure 6

In vivo mouse recording. A) The surgical approach to insert the implant into mouse cortex. After a craniotomy and removing part of the dura (A_i), the flexible implant was placed on the cortical surface (A_ii). A wooden rod (A_iii) was used to push the implant inside the cortex, which subsequently remained in the tissue after retracting the wooden rod (A_iv). B) Recordings from an example electrode at the time of insertion, showing the raw (B_i), as well as filtered signals in the high‐ (300 to 3000 Hz, ii) and low‐frequency (up to 300 Hz, iii) range. The red line in (B_i), (B_ii), and (B_iii) denotes immediate spiking activity after insertion whereas the low‐frequency activity was largely unperturbed. C) Example of simultaneous recordings from three electrodes (C_i), (C_ii), and (C_iii) with sorted waveforms from multi‐unit and potential single‐unit recordings. D) Example recording from a pillar electrode with periodic whisker stimulation every 5 s (D_i) and foot‐pinch stimulation every 3 s (D_ii).