Foldable 3D opto-electro array for optogenetic neuromodulation and physiology recording

Abstract

This paper presents a thin-film, three-dimensional (3D) opto-electro array featuring four addressable microscale light-emitting diodes (LEDs) for surface cortex illumination and nine penetrating electrodes for simultaneous recording of light-evoked neural activities. Inspired by the origami concept, we have developed a meticulously designed “bridge+trench” structure that facilitates the transformation of the array from 2D to 3D while preventing damage to the thin film metal. Prior to device transformation, the shape and dimensions of the 2D array can be customized, enhancing its versatility for various applications. In addition, the arched base offers strong mechanical support to facilitate the direct insertion of the probe into tissue without any mechanical reinforcement. The array was encapsulated using polyimide and epoxy to ensure mechanical flexibility and biocompatibility of the device. The efficacy of the device was evaluated through comprehensive in vitro and in vivo characterization.

Introduction

Recordings of electrophysiological activity from neurons represent well-established essential sources of information for quantitative investigations of neural processes and for the development of biomedical systems and brain-computer interfaces^1,2. To better capture electrophysiological signals and understand the movement patterns and principles of muscles and nerves, many implantable devices have been evolving over the years through advances in microelectromechanical systems (MEMS) technology, deepening knowledge of neural activity and the brain^3–5. However, these devices have encountered many challenges during their continuous development, including the challenge of conventional 2D fabrication methods and the practical requirements of 3D shaping^6–9, accuracy and fidelity of neural signal acquisition, and collateral damage of implantable devices^10–13, among others. Under the constraints of these challenges, device design has become a delicate balancing act. In terms of the mechanical property, flexible devices are difficult to implant and require some means of strength reinforcement¹⁴. On the other hand, stiff devices may increase foreign object response and thus affect behavior and nerve signal quality in various ways^15,16. Choosing a less aggressive strategy, such as surface mounted, may improve biocompatibility and reduce tissue responses to implants but could also decrease signal acquisition accuracy due to the distance from the desired neurons and/or nerves¹⁷.

Origami-inspired designs hold the potential to revolutionize current paradigms. However, several challenges remain unsolved. On the recording side, conventional multi-electrode array technology primarily exists in a rigid, 2D format, thus limiting its functional interface to a small area in 3D form, usually restricted to the area near the bottom contact surface^18,19. Similarly, electrode arrays fabricated using traditional 2D microfabrication techniques are also difficult to accommodate customization of individual probes. As an alternative approach, foldable origami designs offer unique advantages such as reduced manufacturing costs and combination of favorable properties inherent in 2D materials with the versatility of custom 3D structures⁷. However, this approach comes at the price of potential metal cracking at folding sites, necessitating careful consideration in device structural designs. On the stimulation side, direct electrical stimulation (DES) is well-established but presents various challenges such as safety^20,21 and temporal/spatial precision^22,23. Optogenetic neuromodulation, utilizing calcium-sensitive fluorescent stains and genetically encoded calcium indicators, overcomes some of these limitations in neural recording and stimulation¹⁹. Although optogenetic based stimulation is one of the most desirable methods at this stage, integration of light delivery elements with electrode probes poses a new challenge. Using an optical fiber for light delivery is a straightforward and effective method, with the potential for further miniaturization of the device by integrating microelectrodes with the optic fiber^24,25. However, tethering of the optical fiber can increase the probability of infection, inflammation, or traumatic impact-induced mechanical failure^26–28, emphasizing the importance of wireless capability of implantable devices^2,29. Wireless devices can also minimize constraints on free movement of experimental subjects, thus minimizing the impact on subject’s behavior and facilitating more complex, long-term neural activity studies. For these reasons, integrating microscale light-emitting diodes (µ-LEDs) into neural recording probes emerges as a promising solution. However, µ-LEDs are by no means the perfect answer. If the LED is integrated into the probe and implanted in the tissue, the increased size may lead to increased foreign body reaction^16,30. LED approaches are also plagued by issues such as overheating^31,32 and Light intensity attenuation by distance.

To address the above challenges, this paper describes a qualitatively distinct type of three-dimensional (3D) neural interface device with unique properties that utilizes folding to complete the two-dimensional (2D) to 3D conversion to match the tissue of interest. Exploiting advanced planar electronic and optoelectronic technologies in this device, a thin-film, 3D opto-electro array (Fig. 1a) with 4 individually addressable µ-LEDs for optical neurostimulation and 9 penetrating electrode probes for simultaneous recordings of light-evoked neural activities was developed to achieve high-performance, high-resolution functions for neural interface applications. Inspired by the origami concept, a well-designed “trench + bridge” structure on a flexible polymer substrate can complete the folding and reduce metal damage during bending. The arched base design of the probe enhances the mechanical strength of the probe, allowing it to be inserted directly into tissue without mechanical reinforcement, thereby reducing tissue trauma. The opto-electro array can rely on this special hinge structure to achieve the conversion from a 2D substrate into a 3D configuration. Compared with the traditional 3D array fabrication, this origami folding technique can reduce the probe fabricating complexity and cost, especially in the production of relatively long probes (>1 mm). Moreover, fabricating a 2D opto-electro array by photolithography enables independently control the shape, and the length of individual probes, thereby forming an array of probes with different lengths to enhance the depth resolution of optogenetic stimulation and electrophysiology recording.

Fig. 1. 3D opto-electro array overall design and test schematic.

a Layered schematic illustration of the origami-inspire 3D opto-electro array. b Schematic diagram of the folded device being inserted into brain tissue. c 2D schematic diagram of the device (right) and actual image of device (left), rectangular microelectrode (gold, impedance of 300 kΩ at 1 kHz). d Device fabrication flowchart. e Schematic diagram of the device folding and actual image of the folded device. f Schematic diagram of array insertion and actual image of in vivo test. g Schematic diagram of in vivo experiment setup and images of actual experiments

Result

Benchtop test

After all the fabrication and installation steps of the array were completed, the opto-electro array underwent benchtop testing to characterize the performance. In this study, the LED operated at 3 V with an input power of 15 mW. As shown in Fig. 2a, the opto-electro array demonstrated its basic ability to control each µ-LED individually. In addition, the overall light intensity of the array can be adjusted by tuning the On/Off cycle and frequencies of µ-LED light pulses, while the illumination region can be adjusted by selectively activating one or more LEDs at the same time. The optical images in Fig. 2b depicted the distribution of LED light in both fresh brain tissue and 1% agarose (Carolina Biological Supply Company, Burlington, NC, USA), while the plot in Fig. 2b illustrates the depth attenuation of LED light. When four LEDs were simultaneously activated at 3 V, the total optical power output was 5.77 mW, resulting in an overall optical intensity of 18.125 mW/mm². In a homogeneous phantom, the attenuation of incident light exhibited approximately linear behavior, with attenuation positively correlated with distance. However, a simple phantom test cannot fully elucidate the transmission of LED light within the brain. The light intensity did not simply propagate vertically downward in brain tissue; rather, it followed the internal structure of the brain tissue. The colormap of Fig. 2b shows that the light intensity exhibited a trend corresponding to the distribution of the frontal lobe and the internal bold vessels/tubular tissues. Notably, the light intensity with these tubular tissues was stronger than in surrounding areas, implying that light travels along these tubular structures. This result suggested that the intricate structure of brain tissue leads to highly non-uniform light transmission. Considering the attenuation of light through brain tissues, the light intensity was estimated to be approximately 15.4 mW/mm² at a depth of 1 mm (target penetration depth for in vivo experiments), which still exceeds the minimum required standard (1 mW/mm²) for ChR2 activation^33,34.

Fig. 2. Bench Test Results.

a The opto-electro array demonstrated capability of individually switch each LED in the LED array. b Optical images of µ-LED light penetration test in mouse brain (input power = 15 mW) and comparison of the analyzed optical intensity distribution between the fresh tissue and the 1% agarose. c Temperature measured in PBS using the FLIR® Infrared Thermal Imaging Camera at distance = 10 cm, infrared image changes (1) at 0 min, (2) at 10 min, (3) at 20 min and (4) at 30 min (input power = 20 mW) and the corresponding plot of measured temperature changes over time in PBS solution. d The electrochemical impedance of the electrodes measured over a broad frequency range from 1 to 100 kHz (n = 35). Data were plotted in mean ± standard deviation. e Background noise recorded in PBS where no significant difference was observed between LED On and Off

Further analysis of the colormap in Fig. 2b revealed that the transmission of the light was relatively stable within a certain distance (0.2 mm to 1.2 mm, or 3.3 mm to 5 mm). This may indicate the presence of relatively homogeneous regions within brain tissue, where light transmission approaches ideal conditions. However, at greater distance, the light distribution curve within brain tissue followed a trend similar to the idealized curve but had attenuated to 20–40%. This result highlights the challenges of LED light stimulation in deeper regions of brain tissue. Additionally, at close distances, there was a drastic variation in light transmission within brain tissue, suggesting that as light crosses different regions within the brain, it undergoes significant attenuation. This may be attributed to variations in the refractive index of different brain regions, leading to internal reflections and, consequently, light attenuation.

Figure 2c depicted the distribution of the LED-induced heat in phosphate-buffered saline (PBS), along with the corresponding localized temperature change over a 30-min period of continuous LED operation. To ensure the thermal safety of the array, the primary design consideration was to use an encapsulation layer to isolate heat from the tissue, ensuring that the LED-induced heat didn’t lead to excessive local temperatures. In addition, wide metal traces of the array could serve as a means of thermal conduction, dissipating heat effectively to areas farther from the brain tissue. During testing, the array was submerged in 37 °C PBS, and the four LEDs remained activated for 30 min. The instantaneous temperature changes were captured using a FLIR® Infrared Thermal Imaging Camera (FLIR Systems, Inc., Wilsonville, OR, USA) at 10 cm. Figure 2c showed that the temperature increase remained within a range of 2 °C over 30 min, indicating that the heat generated by the LEDs would not cause significant harm to the tissue over a short period. In addition, the thermal conduction using the wide trace design was effective, as evidenced in Fig. 2c. While the results showed that thermal management design may not completely prevent a warming trend in the LED area, it can mitigate temperature rise to some extent. It is noteworthy that during testing, the LEDs operated in continuous mode, whereas in in vivo experiments, the LEDs were pulsed for stimulation. The pulse stimulation could further mitigate the risk of thermally induced tissue damage. Therefore, Fig. 2c demonstrated closer to the worst-case scenario, and the temperature increase of LEDs was less than 2 °C after continuous operation for over 30 min (worst-case scenario), within the safety limit for biomedical implants³⁵. This temperature change is expected to be even lower when LED is operated in pulse mode.

The electrochemical impedance of the array was measured, as shown in Fig. 2d, with the impedance of the array generally staying within the range of 300 kΩ. Fig. 2e indicated that the light-induced artifacts were relatively small, as evidenced by the limited impact of LED light on the background noise recorded via the integrated electrode probes. This may be attributed to the relatively far distance between the LEDs and the electrodes.

Insertion and bending tests

For the convenience of testing and structural improvements, a single probe was employed in bending and insertion tests. A single probe was fabricated and constructed using the same method as the probes of the array to ensure experimental accuracy. A specifically designed arched base was incorporated to provide essential support to the shank of the probe, thereby structurally reinforcing the mechanical strength of the probe. Moreover, this arched base avoids drastic changes in the probe’s width, thereby preventing tearing issues during bending. With the arched base, the probes were expected to achieve a balance, with their mechanical strength slightly higher than the tissue’s strength to facilitate smooth insertion into the tissue while retaining maximum flexibility to minimize damage to surrounding tissue. Avoiding the use of additional strength enhancement strategies can further reduce system and surgical complexity. As shown in Fig. 3b, even without any strength enhancement, the probes could smoothly penetrate to a depth of 0.6 mm in 0.6% agarose at a speed of 0.1 mm/s. The bottom image in Fig. 3c showed that after insertion and retraction in 0.6% agarose, the probe arms remained straight.

Fig. 3. Insertion mechanics test setup, simulation, and results.

a An image of insertion test setup. b Video clips showing the insertion and retraction of a single probe in 0.6% agarose without any mechanical reinforcement. c The image of single probe inserted into 0.6% agarose with the PEG enhancement (top). The image of single probe inserted into 0.6% agarose with the tungsten guide (middle). The image of a single probe inserted into 0.6% agarose without any strength enhancement (bottom). d The average force measurement under different insertion strategies (n = 3), and the computed tomography (CT) image of a single probe insertion into fresh mouse brain tissue without any strength enhancement. e Finite element analysis comparing stress distribution between the bridge design (left) and the original design (right), including 1-wide bridge, 2-narrow bridge, and 3-original designs, alongside impedance measurements before and after bending. The side view of the simulation illustrates the formation of a distinct curvature by the bridge-type structures during bending

For comparison, as shown in Fig. 3c (top and middle), the probe enhanced with poly (ethylene glycol) (PEG) and tungsten guide remained relatively straight, indicating that the presence of strength enhancement also facilitated smoother insertion. Figure 3d (left) compared the insertion forces measured from the probe without and with the strengthening strategy. It is evident that the probe enhanced with both the PEG and tungsten guide exhibited insertion forces closely approximating those of the rigid tungsten guide itself. In contrast, with flexible structural enhancements using PEG, the force results were more aligned with those of the probe itself. Of particular interest was the finding that, despite the increase in cross-sectional area, the insertion force from tungsten-guided devices was lower than those from the probe-only and probe-PEG groups. This may suggest that during the insertion process, friction and the deformation of the flexible probe could be significant sources of stress.

While agarose mimicked the mechanical properties of brain tissue, it cannot replicate the complex internal structures of brain tissue or the bending issues that may arise due to differences in density. Therefore, the probe was inserted into the fresh mouse brain to validate its strength. As shown in Fig. 3d (middle and right), the probe was successfully inserted into brain tissue without any strength enhancement, resulting in a relatively straight insertion position. This result validates the feasibility of the probe’s strength and demonstrates that at a depth of 1 mm within brain tissue, the density variations and tissue strength are insufficient to alter the position of the probe.

Folding is one of the fundamental challenges for all origami-inspired arrays. We employed two structural designs to optimize the integrity and performance of the array after folding. The folding of the probes was performed using a needle station, carefully folding them along the trench structures. The primary purpose of the trench design was to ensure that the folding occurred within the designated area rather than in an uncontrolled, relatively large area. However, subsequent impedance measurements indicated an increase in impedance for the array. This rise in impedance could affect the sensitivity of the probes, resulting in a decrease in the signal-to-noise ratio (SNR) of neural recordings. From the previous study³⁶, one major reason for this increase in impedance is the presence of cracks in the metal, especially thin film deposited metals, during the folding process. The increase in impedance is directly proportional to the number of cracks and the length of individual cracks³⁶. Based on these findings, the array was deliberately optimized for bending site structures. A key to this optimization lies in reducing the degree of bending to prevent the formation of metal cracks or decreasing the ratio of crack length to the width of the metal traces. To minimize the degree of bending in the metal portion, a bridging structure is employed at the bending locations of the array. This isolates the metal traces from the rest of the structure. When bending occurs, the bridging structure is not influenced by the main structure, thereby mitigating the stress caused by bending to some extent. On the other hand, to address metal damage, a relatively passive approach is used by increasing the metal area at the bending locations to compensate for metal damage. This reduces the ratio of a single crack length to the width of the metal, thereby lowering the impedance increase due to bending. The combination of these two approaches, along with the optimization of the trench structure, constitutes the “Bridge + Trench” structure employed in the array at the bending location.

As shown in Fig. 3e, finite element analysis in COMSOL Multiphysics® (Solid Mechanics Module) was conducted on different structures: the original design (right) and the two bridge structures (left). In Fig. 3e, the middle section shows impedance changes after bending for three designs: the original (Design 1), the bridge design (Design 2), and the wide bridge design (Design 3). In the original structure, stress concentrated mainly at the bending point, while in the bridge structure, stress was dispersed over a significantly larger region, and the stress in the metal trace area was greatly reduced. This demonstrates the effectiveness of the bridge structure in alleviating stress concentration issues. Lower stress levels are advantageous for maintaining the integrity of the metal traces during the bending. Additionally, Fig. 3e (middle) shows the impedance change chart, further demonstrating the effectiveness of the bridge + trench structure. Compared to the original design, which experienced a significant increase in impedance after bending, both bridge designs ensured stable impedance during bending. In other words, they protected the thin metal traces from damage during the bending. These results confirm that the bridge + trench structure can effectively protect the metal wires from damage during the folding. In addition, the COMSOL simulation in Fig. 3e (left) intuitively demonstrated the role of the “bridge + trench” structure during bending. Different curvature occurs in the metal trace portion and other parts of the probe during bending, resulting in a more gradual curve that reduces stress and protects the internal metal layer at the bending site.

When the probe arm is inserted into tissue, the tip experienced force transmitted along the shank, causing bending and deformation without support, affecting neural signal recording. The arched base supports the probe by extending the support distance while minimizing surface area, reducing overall size, and distributing force over a larger area to decrease bending stress. Figure 4 showed the COMSOL simulation results for the impact of different base designs on insertion, demonstrating that the arched bottom was the optimal solution for the base design which can achieve the most efficient force distribution in the smallest area. In contrast, the triangular base failed to effectively distribute the force, resulting in the stress concentrated at the transition area between base and shank. These simulation results further validate the effectiveness of the arched base in facilitating probe insertion.

Fig. 4. The COMSOL simulation results illustrate the stress distribution of probes with various base designs during forward insertion.

The right plots provide the detailed values of stress at varying distances from the probe tip to the base

In vivo validation of the optical stimulation and neurorecording capabilities

The schematic diagram in Fig. 5a provided an overview of the entire system for in vivo validation, including backend recording and LED control. On the right side were the actual images of the optical stimulation through the LED array. The entire system was centered around a custom designed printed circuit board (PCB). After implantation in the rat’s brain, the opto-electro array was connected to the data acquisition board via the PCB. The function generator was connected to the LED array through the same PCB to control the LEDs operation and provide transistor-transistor logic (TTL) signals to the data acquisition board to synchronize various input data. Crosstalk between the two systems was minimized through encapsulation and relatively large spacing between the signal line and the LED control line.

Fig. 5. Overall test framework and test results of in vivo tests.

a In vivo experiment setup. b Immunohistology analysis showing c-Fos (3,4) and mCherry (5,6) expressions in the stimulated and control visual cortex. (1,2) The combined images. Scale bar = 100 µm. c Neural activity during two cycles of LED stimulation (200 ms ON and 800 ms OFF), with each black bar representing a neural spike captured. Optical input power at 20 mW/mm². d Sorted neural spikes recording in 1 min during and after the LED stimulation. e Sorted neural spikes in 1 min before the LED stimulation. f Sorted neural spikes recording comparison between pure gold electrode and PEDOT enhanced

To validate the efficacy of LED for optical stimulation, immunohistology analysis using c-Fos biomarker was performed on both the stimulated and control primary visual cortex (V1). The results illustrated in Fig. 5b indicated a higher population of neurons (bright green in the image) expressing c-Fos biomarker in comparison with the control cortex. This up regulation in the stimulated cortex demonstrated that optical stimulation using the LED array could effectively excite transfected neurons in vivo. It is worth noting that the control cortex also exhibited some c-Fos expression, which could be attributed to spontaneous neural responses as well as neural responses caused by the mechanical pressure on the cortex during the surgical procedures. These immunohistology results proved that after completing the benchtop light intensity test and controlling the heat problem, the actual in vivo results aligned with the experimental expectations mentioned above. In addition, the placement of the LED array on the brain surface was able to deliver sufficient light for cortical stimulation, while minimizing the potential risk of tissue damages due to joule heating.

Figure 5c–e demonstrated the signal acquisition capability of the array and its optogenetic stimulation capability. In the idle state (Fig. 5e), background noise was around 10 µV, and most spontaneous action potentials could reach −60 µV. Under LED stimulation, cells depolarized, and the extracellular action potential dropped to around −100 µV. Compared to the idle state, the action potentials during the stimulated state (Fig. 5d) exhibited larger magnitudes and faster firing rates. Figure 5f compared the recorded signal quality before and after depositing poly(3,4-ethylenedioxythiophene) polystyrene sulfonate (PEDOT: PSS) on gold electrodes. Although pure gold electrodes could record neural signals, they had relatively poorer signal quality due to the higher impedance compared to PEDOT: PSS coated electrodes. This could be attributed to the relatively higher electrochemical impedance of the gold electrodes, resulting in the attenuation of signals that were slightly distant or relatively weak action potentials, thus getting filtered out amidst background noise.

Figure 6 illustrated the multi-channel capability of the probes. Upon insertion into the brain, signals from different depths of the same area were recorded due to the varying lengths of the probes. Channel 1 (Ch-1), which was inserted deeper, recorded relatively stronger signals. Principal Component Analysis (PCA) indicated that the probe had detected signals from two clusters, suggesting the possibility of two neurons firing signals near the probe. On the other hand, Channel 2 (Ch-2), inserted at a shallower depth, recorded relatively weaker action potentials but captured a relatively larger number of neural signals due to its distance from the neurons.

Fig. 6. Multi-channel electrophysiological signals recorded with origami opto-electro array.

a Average action potential shape from the recording and representative time-dependent waveforms of isolated neurons from Ch-1. b Average action potential shape from the recording and representative time-dependent waveforms of isolated neurons from Ch-2

Overall, in the idle state, the opto-electro array demonstrated its ability to capture spontaneous response of neurons, confirming its capability for actional potential and multi-channel recording. Under LED stimulation, the array captured significantly intensified changes in action potentials, verifying its ability to optically stimulate neural activity via the µ-LEDs and electrically record light-evoked neural activity via the microelectrodes.

Discussion

In this study, we have successfully designed and tested a foldable opto-electro neural recording array that exhibited good performance and functionality. The results illustrated that this array efficiently addressed individual LEDs, as evidenced by the achieved design purposes shown in Fig. 2a. This feature provided the potential for precise experimental designs involving regional illumination. An intriguing discovery, as illustrated in Fig. 2b, was the distinctive light distribution between a homogeneous phantom and brain tissue. While light attenuated linearly in the phantom, its behavior in brain tissue was profoundly influenced by the intricate internal structure. Such divergence suggested that simple phantom tests were not robust enough to model light transmission intricacies within brain tissue. Moreover, the presence of relatively homogeneous regions within the brain tissue, inferred from stable light transmissions within specific distances, underscores the challenge of LED light penetration in deeper brain regions. At close distances, significant variation in light transmission was observed within brain tissue, indicating substantial attenuation when light crossed different regions within the brain. This could be attributed to variations in the refractive index of different brain regions, leading to internal reflections and subsequent light attenuation. These results emphasized the complexity of light transmission within brain tissue, underscoring the critical need for comprehensive understanding and careful consideration when designing optogenetic stimulation experiments.

Concerning the thermal protection of the array, utilizing an encapsulation layer proved to be a prudent approach, effectively isolating the LED-induced heat from sensitive brain tissue. Concurrently, the array’s wires act as a secondary line of defense, dissipating the heat and safeguarding the adjacent region. Experiments demonstrated that the heat generated by the LEDs caused only a minor temperature increase within brain tissue over a short period, ensuring the safety of the tissue during stimulation.

For insertion tests, achieving a fine balance between probe strength and flexibility was imperative. Figure 3 revealed that while probes could penetrate without enhancements, employing a PEG or tungsten guide leads to more precise insertions. Interestingly, increasing the cross-sectional area did not correspondingly increase the insertion force, suggesting friction and probe deformation were potential stress inducers.

Incorporating insights from origami-inspired designs, addressing the folding challenge was tackled through two main strategies: ensuring controlled folding through trench structures and mitigating potential metal damage. The innovative “bridge + Trench” structure, as elucidated in Fig. 3, not only reduced stress concentrations but also provided a more consistent impedance during bending, thereby effectively preserving the integrity of the thin film metal components during folding.

For in vivo validation, results from Fig. 5 offered compelling evidence for the efficacy of the optical stimulation. A pronounced upregulation of c-Fos biomarker in the stimulated cortex compared to the control affirmed the array’s functional capacity. Also, it’s worth noting that while surgical procedures might induce some mechanical stress, the array’s underlying design ensures minimal tissue damage from heat. Furthermore, signal acquisition results in Fig. 5c–e highlighted the array’s adeptness in capturing neural activity both under the idle and LED-stimulated conditions.

In future work, further experimentation on probe insertion will still be a focus point. The reason why this article has always revolved around direct insertion into tissue without the aid of any kind is twofold. Firstly, doing away with guides will significantly reduce potential tissue damage. Secondly, if the strength of the probe itself is sufficient, it will facilitate the subsequent simultaneous insertion of the entre array. The implantation of the entire array into brain tissue will also be one of the focal points in the future. Inserting multiple probes at once and controlling their depth is challenging for flexible arrays. The opto-electro array can be folded and can smoothly accommodate the simultaneous implantation of 1–3 probes without relying on any reinforcements. In theory, simultaneously inserting 9 probes is feasible, but it’s still quite challenging in practice. This is limited by the insertion methods, the structure of the brain surface, as well as non-uniform force applied on the array. In the future, we will conduct further research on these issues. Regarding the structural aspect, the arched base design of the probe that was originally intended to prevent issues such as tearing caused by excessive width changes also provides additional support to aid in probe insertion. The mechanical study and finite element analysis of this part will be launched in the future.

In summary, our foldable opto-electro neural recording array showcases promising capabilities in optical stimulation and neurorecording. However, the intricate nature of brain tissue poses challenges in achieving uniform light distribution, emphasizing the need for sophisticated experimental designs and accurate modeling. Additionally, further studies could explore advanced materials and structures to enhance the array’s performance, promoting its application in neuroscience research and advancing our understanding of neural activities.

Conclusion

This paper presents an origami-inspired 3D opto-electro array that successfully integrates innovative mechanical design with functional optical and electronic elements for optogenetic neuromodulation and neurophysiology recording. The utilization of folding to transition a 2D array into a 3D form is skillfully managed through the implementation of the “bridge + trench” structure. This structural design not only facilitates the folding process but also effectively mitigates stress-induced damage to the metal layers within the probes, as demonstrated by COMSOL simulations and comparative analyses with the traditional designs (Fig. 3). The observed reduction in maximum stress and the maintenance of impedance post-folding serve as clear indicators of the efficacy of this new design. The base design with an arch shaped structure provides strong mechanical support to the probes while maximizing space utilization. Combined with the inherent material strength of polyimide, this allows the probes to penetrate tissues within a 1.5 mm distance without any mechanical reinforcement, thereby reducing tissue trauma. Moreover, the ability of the penetrating electrodes to capture spontaneous single-unit responses, along with the enhanced neural activity observed under LED stimulation (as depicted in Fig. 5), highlights the probes’ effectiveness in optogenetic neurostimulation and neurophysiology recording. The augmented expression of the c-Fos biomarker in stimulated neurons, compared to the control cortex (Fig. 5), further demonstrates the potency of LED light stimulation in inducing neural activity.

In summary, the incorporation of innovative structural design principles into origami probes, coupled with their successful application, represents a significant advancement in the realm of optoelectronic stimulation and recording arrays. The “bridge + trench” structure adeptly tackles the challenges posed by folding-induced stress and impedance fluctuations, thereby ensuring the structural robustness and functional reliability of the probes.

Method

Fabrication of the Origami opto-electro array

Figure 1d depicts the core fabrication process flow of the opto-electro array. Polyimide was chosen as the structural and encapsulation material because of its excellent biocompatibility, chemical stability, and mechanical properties^37–39. Specifically, a 5-µm thick layer of copper was electroplated on a 4-inch silicon wafer as a sacrificial layer. Adhesion promoter (VM652, HD Microsystems L.L.C., Parlin, NJ, USA) was spun on the wafer in advance to improve the adhesion of polyimide (PI 2611, HD Microsystems L.L.C., Parlin, NJ, USA). After that, polyimide was spun on the adhesion promoter at a spinning speed of 4000 rpm, resulting in the thickness of 6 µm. Then the whole wafer was soft baked on hotplate at 90–150 °C to thicken polyimide, followed by hard baking at 350 °C for approximately 30 mins to fully cure the polyimide. The cured polyimide can completely dissociate the carrier solvent, fully imidine the film, and complete polymer orientation, thereby optimizing electrical and mechanical properties. After polyimide was cured, a 200 nm copper layer was thermally evaporated on the polyimide layer as a hard mask. Photoresist (PR, S1813, Shipley, Marlborough, MA) was spun onto the copper to define the probe shape in the copper layer through ultraviolet lithography and copper wet etching. Followed by a reactive ion etching (RIE) at power of 200 W and gas pressure of 0.15 Torr for 20 min, the basic probe shape was defined with the copper mask. Gold was selected as the conducting material for electrodes, traces and contacts, and patterned using wet etching with a lithographically defined photoresist mask. Another polyimide layer was spun in the same way to encapsulate the probe and define the detection windows and mounting pads. As proof of concept, the electrode was designed with dimensions of 35 µm by 50 µm, which has been proven sufficient for detecting neural signals^40–43. Our current design featured only one electrode per probe to minimize the footprint size of each probe in order to improve the biocompatibility and reduce potential brain trauma⁴⁴.

To further reduce the electrochemical impedance of gold electrodes for effective recording, PEDOT: PSS was electropolymerized onto the electrodes using cyclic voltammetry (CV) within a three-electrode cell setup. Microelectrode arrays acted as the working electrode (WE), a platinum wire as the counter electrode (CE), and a standard Ag/AgCl electrode as the reference electrode (RE)⁴⁵. The process involved a solution of 10 mM 3,4-Ethylenedioxythiophene (EDOT, Millipore Sigma, St. Louis, MO, USA) and 0.7 wt.% Poly (sodium 4-styrenesulfonate) (PSS, Millipore Sigma, St. Louis, MO, USA) in deionized water. This solution was deoxygenated by nitrogen purging followed by vacuuming before use. CV was conducted using a potentiostat (6149E, CH Instruments, Austin, TX, USA), with voltages ranging from −0.7 V to 1.1 V versus the Ag/AgCl RE and a scan rate of 10 mV/s. The PEDOT: PSS significantly reduced the 1 kHz impedance from 350 kΩ to 50 kΩ.

Integration of the µ-LED and array folding

After the device was released from the wafer, four µ-LEDs were mounted on their mounting pads. Applying small amount of tack flux (CHIPQUIK^® Tack Flux SMD291ST2CC6, Life solution Inc, Ancaster, ON) and low-temp solder paste (CHIPQUIK^® LOWTEMP LEAD-FREE SN42/BI58 Solder Paste SMDLTLFP, Life solution Inc, Ancaster, ON) on the corresponding electrode pads. After the solder was melted with a hot air gun, the µ-LED (Cree® TR2227™ LEDs, Cree, Inc. Durham, NC) were then gently placed and aligned on the pads. Having mounted the µ-LEDs, optical measurements of the light output were performed with a digital power meter ((PM100D, power meter and S120VC, photodiode sensor, Thorlabs, NJ, USA). The dimension of Cree® TR2227™ LEDs are 220 µm (width) × 270 µm (length) × 50 µm (thickness). The electrode probes were folded along the trench structure by a needle station. The specific steps for manually folding the probe are as follows. Lift the probe from the bottom, and while preparing to fold, observe whether the folding area is the designated region. Fold the probe centripetally, ensuring the folding angle slightly exceeds 90 degrees. During the folding process, only touch the base area and avoid touching the metal bridge, allowing the bridge section to fold naturally. The folding motion from the base should elevate the entire probe vertically. The trench design played an important role in inducing folding and limiting the folding area. After the probe is fully upright, apply a drop of epoxy at the tip of the needle and carry it to the LED area, ensuring complete coverage of the LED region with epoxy. Allow gravity and surface tension to spread the epoxy across the entire base of the probe, with a thickness of approximately 150 µm, thereby completing the LED encapsulation and securing the probe in place. Finally, the epoxy was cured at room temperature overnight (>8 h).

Finite element simulation in COMSOL Multiphysics®

For the pressing simulation, the COMSOL Multiphysics® solid mechanics module (COMSOL, Inc., Burlington, MA, USA) was applied, as shown in Fig. 7. The tips of the probes were fixed, and then 2 N force was applied from the bottom tip surface in the y-axis direction to exert press. For the bending simulation, we fixed the bottom tip surfaces of the probes and applied a −0.7 mm displacement along the axis direction displacement and a +1.1 mm vertical displacement from the edge tips. After setting up these boundary conditions, we used a stationary solver to obtain the results.

Fig. 7.

COMSOL setup diagram

Animal preparation and experiment setup

In vivo animal experiments were conducted to verify the surgical and functional applicability of the origami opto-electro array. The probe’s capability for neural recording was evaluated in vivo at various depths in the same brain region. All procedures were approved by the Institutional Animal Care and Use Committee (IACUC, 202200102 G2) at Michigan State University. Adult rodent subjects (Long Evans, Male, 500–700 g) received virus injection (AAV-hSyn-hChR2 (H134R)-mCherry; UNC Vector Core) in V1 to express neurons with light sensitive channelrhodopsin-2 (ChR2) prior to device implantation. The injection surgery was performed under an inhalation anesthesia (isoflurane and oxygen mixture, 1–4% vaporizer) while the subject was on a stereotaxic apparatus (World Precision Instruments, Sarasota, FL, USA). A 3–4 cm incision was created in the skin and three equidistant holes were drilled on each cortex using an electrical micro-drill (The Ideal Micro-Drill™, Roboz surgical Instrument Co., Inc., Gaithersburg, MD, USA). The AAV virus (10¹² to 10¹³ genome/mL, 1 µL per drilled location) was injected into the brain through the holes. After the injection was completed, the cortex was covered with gelfoam and then the skin was sutured closed. The injected rats were housed separately and given pain medications to reduce discomfort and antibiotics to prevent infections. After three weeks of post injection surgery, the subjects performed the experiment on the stereotaxic apparatus with the same anesthesia procedure mentioned above. The origami opto-electro array was surgically implanted into V1. A grounding wire was inserted under the skin to reduce artifacts. The array was connected via specially designed PCB to the Omnetics connector (Omnetics Connector Co., Minneapolis, MN, USA), then connected to the data acquisition board. The excitation of the µ-LED was controlled by a function generator and synchronized to the data acquisition board (Intan RHD USB Amplifier Evaluation System, Intan Technologies, Los Angeles, CA, USA). We demonstrated individual LED control in benchtop tests for precise optogenetic stimulation. At 3 V, the optical power of two LEDs is 4.9 mW, lower than the 5.77 mW of four LEDs. Therefore, in the current prototype stage, we use four LEDs in in vivo experiments to minimize variables and ensure effective optogenetic stimulation.

Immunohistology processing followed a standard c-Fos protocol, the anaesthetized experimental subject’s brain was stimulated for 45 min, followed by a 90 min survival period post stimulation. Subsequently, the subject was perfused with chilled saline and 4% paraformaldehyde, and the brain was post-fixed overnight at 4 °C in the same solution. Brain sections were then cut to 40-µm thick and chilled in 0.1 M PBS before being placed in culture dishes for immunohistology analysis. The sections were washed three times for 10 min each in PBS. Next the sections were soaked in PBS mixed with 1% NGS and 0.3% Triton ×-100 for 2 h at room temperature (23 °C). Rabbit mAb antibodies (one mL) were placed in 1.5 mL Eppendorf tubes, which were then clipped onto a rotating mixer and stored at 4 °C for 24 h incubation. After that, the sections were incubated in the dark with 20 antibodies (ThermoFisher A27034 goat anti-rabbit IgG superclonal Alexa Fluor 488 conjugate) for 2 h at room temperature. Finally, the sections were washed in 0.1 M PBS, mounted, cover slipped with anti-fade media, and stored in a cool dark place.

Insertion of the origami opto-electro array

Insertion experiments were conducted to assess the ability of the array for direct insertion into brain tissue and to simulate the stress encountered during insertion under different reinforcement methods. This study further validated the integrity of the folded probe during tissue penetration. The experimental groups were divided into three categories: the tungsten guided group, the PEG (Sigma-Aldrich, Inc., St. Louis, MO, USA) enhanced group, and the control group (no reinforcement). The probe was immersed in a solution containing PEG and then slowly withdrawn, allowing the PEG solution to form a uniform thin film on the surface of the probe. As the solvent evaporated, the remaining PEG film provided additional mechanical support, enhancing the stability and durability of the probe, and possibly also offering some improvement in biocompatibility. For in vitro experiments, 0.6% agarose was selected to simulate the mechanical properties of the brain tissue⁴⁶. Weighed out agarose powder for solution (1× phosphate buffered saline or 1x TBE Buffer (Tris-borate-EDTA)) to create a 0.6% solution, dissolved by heating (100 °C), then cooled to 50–60 °C before pouring into a casting tray to solidify. The insertion test platform was built using Thor Lab components (Thorlabs Inc. Newton, NJ, USA). In particular, a motorized stage (MT1-Z8 12 mm One-Axis Motorized Translation Stage, Thorlabs Inc. Newton, NJ, USA) was used as the advancement platform to drive the array insertion into the 0.6% agarose at a speed of 100 µm/s. A high-precision load cell (M3-012, Mark-10 Co., Copiague, NY, USA) with resolution of 500 nN was firmly attached to the 0.6% agar to measure the force applied during insertion.

Acquisition of electrophysiological data

The electrophysiological data were acquired and monitored using a commercial system (Intan RHD USB Amplifier Evaluation System, Intan technologies, Los Angeles, California) and its companion software (RHD2000 interface, Intan technologies, Los Angeles, California). The system recorded from all electrodes simultaneously with a sampling rate of 20 kHz and real-time display capabilities. After completing the recording of biological data, use MATLAB (The MathWorks, Inc., Natick, MA, USA) and the MATLAB-based offline spike sorting software ROSS⁴⁷ for data processing and visualization.

Author contributions

Conceptualization, Y.G. and W.L.; Methodology, Y.G.; Software, Y.G. and Z.J; Validation, Y.G.; Formal Analysis, Y.G.; Investigation, Y.G., X.L., Z.J., A.W.; Data Curation, Y.G.; Writing-Original Draft Preparation, Y.G.; Writing-Review & Editing, Y.G., Z.J., X.L., A.W., W.L.; Supervision, W.L., A.W.; Project Administration, W.L.; Funding Acquisition, W.L. All authors have read and agreed to the published version of the manuscript.

Conflict of interest

The authors declare no competing interests.